+ All Categories
Home > Documents > Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The...

Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The...

Date post: 09-Feb-2020
Category:
Upload: others
View: 10 times
Download: 2 times
Share this document with a friend
98
Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area Component 3 Flood Risk Analysis PHILIPPINE ATMOSPHERIC GEOPHYSICAL AND ASTRONOMICAL SERVICES ADMINISTRATION GEOSCIENCE AUSTRALIA Badilla, R. A. 1 , Barde, R. M. 2 , Davies, G. 3 , Duran, A. C 1 , Felizardo, J. C. 4 , Hernandez, E. C. 5 , Ordonez, M. G. 6 , Umali, R. S. 6 1. Philippine Atmospheric Geophysical and Astronomical Services Administration (PAGASA) 2. Metropolitan Manila Development Authority (MMDA) 3. Geoscience Australia (GA) 4. Department of Public Works and Highways (DPWH) 5. Laguna Lake Development Authority (LLDA) 6. Mines and Geosciences Bureau (MGB)
Transcript
Page 1: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area

Component 3 – Flood Risk Analysis

PHILIPPINE ATMOSPHERIC GEOPHYSICAL AND ASTRONOMICAL SERVICES

ADMINISTRATION

GEOSCIENCE AUSTRALIA

Badilla, R. A.1, Barde, R. M.

2, Davies, G.

3, Duran, A. C

1, Felizardo, J. C.

4, Hernandez, E. C.

5, Ordonez,

M. G.6, Umali, R. S.

6

1. Philippine Atmospheric Geophysical and Astronomical Services Administration (PAGASA) 2. Metropolitan Manila Development Authority (MMDA) 3. Geoscience Australia (GA) 4. Department of Public Works and Highways (DPWH) 5. Laguna Lake Development Authority (LLDA) 6. Mines and Geosciences Bureau (MGB)

Page 2: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

ii Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

Page 3: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the iii Greater Metro Manila Area – Flood Risk Analysis

Contents

Executive Summary .................................................................................................................................. v

1 Introduction ............................................................................................................................................ 1

1.1 Background of the Study.................................................................................................................. 1

1.2 Area of the Study ............................................................................................................................. 2

1.2.1 Topography and Location ........................................................................................................... 2

1.2.2 Climate ....................................................................................................................................... 4

1.2.3 Population................................................................................................................................... 5

2 Literature Review ................................................................................................................................... 8

2.1 Hydraulic Studies ............................................................................................................................. 8

2.2 Mines and Geosciences Bureau Flood Susceptibility Mapping .....................................................10

2.3 General Approach to Flood Risk Analysis .....................................................................................11

2.3.1 Flood Hazard Information .........................................................................................................11

2.3.2 Exposure Information ...............................................................................................................12

2.3.3 Vulnerability Information ...........................................................................................................13

2.4 Introduction to flood inundation models .........................................................................................13

2.4.1 Rainfall Runoff Models .............................................................................................................13

2.4.2 Hydraulic Models ......................................................................................................................14

3 Methods ...............................................................................................................................................18

3.1 Data ................................................................................................................................................18

3.1.1 Elevation Data ..........................................................................................................................18

3.1.2 Hydrological data ......................................................................................................................19

3.1.3 Exposure Data and Vulnerability Curves .................................................................................23

3.2 Software Selection .........................................................................................................................23

3.2.1 Data Processing and Analyses ................................................................................................23

3.2.2 Flood inundation modelling ......................................................................................................24

3.3 Flood Inundation Model Development and Calibration ..................................................................25

3.3.1 Rainfall Runoff Model ...............................................................................................................25

3.3.2 Hydraulic Model ........................................................................................................................27

3.4 Design Flood Estimation ................................................................................................................36

3.4.1 Synthetic Storm Time Pattern ..................................................................................................36

3.4.2 Spatial Variation in Extreme Rainfalls in the Pasig-Marikina Catchment. ................................39

3.4.3 Catchment-Averaged Extreme Rainfall Frequency Analysis ...................................................41

3.4.4 Frequency analysis of high water levels in Laguna Lake .........................................................42

3.4.5 Relation to extreme rainfalls .....................................................................................................43

3.4.6 Design Flood Boundary Conditions ..........................................................................................46

3.5 Damage Calculation .......................................................................................................................47

3.5.1 Computation of ‘damaged floor area equivalent’ in a single exposure polygon, for a single building-type and a single depth .............................................................................................49

3.5.2 Computation of the inundated floor area in each exposure polygon. ......................................50

4 Methods ...............................................................................................................................................52

4.1 Hydrology .......................................................................................................................................52

Page 4: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

iv Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

4.1.1 Regional Extreme Rainfall Frequency Analysis .......................................................................52

4.1.2 Catchment Averaged Extreme Rainfall Frequency Analysis ...................................................55

4.1.3 Design Storm Temporal Pattern ...............................................................................................55

4.1.4 Laguna Lake Water Level AEP Curve .....................................................................................56

4.2 Hydraulics ......................................................................................................................................57

4.2.1 Model Calibration .....................................................................................................................57

4.2.2 Design Flood Scenarios ...........................................................................................................66

4.3 Damage Estimation ........................................................................................................................67

4.4 Patterns of Flood Hazard and Risk ................................................................................................73

4.4.1 Limitations of the Analysis ........................................................................................................74

5 Conclusion ...........................................................................................................................................77

6 Recommendations ...............................................................................................................................78

References .............................................................................................................................................79

Appendix A - Parameters used in the Rainfall Runoff Model Sub catchments ......................................82

Appendix B - Estimated building costs per m² (‘000s of Peso), for different combinations of Building type, L4_USE and L5_USE. .....................................................................................................84

Appendix C - Floor heights for different building categories, based on field survey data collected by PHIVOLCS .........................................................................................................................................92

Page 5: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the v Greater Metro Manila Area – Flood Risk Analysis

Executive Summary

Metro Manila is highly vulnerable to flooding, as demonstrated by the events of Tropical Storm Ondoy

(2009), and more recently the enhanced monsoon rainfall (Habagat) events of 2012 and 2013. This

report details a flood risk analysis for the Pasig-Marikina Basin (which is the major river system in

Metro Manila), developed collaboratively by technical specialists from the Governments of the

Philippines and Australia, as part of the Greater Metro Manila Risk Analysis Project. The study

includes development of flood hazard maps for scenarios with annual exceedance probabilities of

between 20% and 0.5% (corresponding to average recurrence intervals of about 1/5-1/200 years). For

each of these scenarios, maps describing the building damages (damaged floor area and damage

cost) and number of people with inundated homes are also developed.

Page 6: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

vi Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

Page 7: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 1 Greater Metro Manila Area – Flood Risk Analysis

1 Introduction

1.1 Background of the Study

The Philippines is one of the most flood-prone countries in the world. For the last ten years, there have

been over 60 reported major floods in the Philippines. Nearly 14 million people have been affected

and the death toll has reached more than 700 people with damages estimating over $400 million (EM-

DAT International Disaster Database).

Metropolitan Manila, the economic centre of the Philippines, is considered the most susceptible city to

flooding. Owing to its geographical location, low elevations, high density of population and

infrastructure, Metropolitan Manila has greater exposure to flooding impacts than most other parts of

the Philippines. This was illustrated during the passage of Tropical Storm Ondoy (Ketsana) in Greater

Metro Manila Area on 26 September 2009 which brought 455 mm of rainfall for 24-hr to its

catchments. This event caused severe flooding, resulting in many casualties (464 dead, 529 injured

and 37 missing), with direct impacts on around 5 million people; and damages to infrastructure and

agriculture of 11 billion Pesos (NDRRMC, 2009).

Given the continued growth of population and infrastructure that is expected in Metropolitan Manila, it

seems likely that events with even more severe impacts could occur in future. To enhance the

capacity of the Government of the Philippines Technical Agencies to understand and manage future

natural hazard risks in this area, a new disaster risk reduction initiative was developed collaboratively

between the Governments of the Philippines and Australia in 2010. The $30 million dollar BRACE

Program (Building the Resilience and Awareness of Metro Manila Communities to Natural Disasters

and Climate Change Impacts) aimed to reduce the vulnerability and enhance the resilience of Metro

Manila and neighbouring areas to the impacts of natural disasters. The program included four

components related to disaster risk reduction. One of these components was dedicated to “Enhancing

Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind, and Earthquakes for Greater Metro

Manila Area”, which became known as the Greater Metro Manila Area Risk Analysis Project (GMMA-

RAP). The GMMA-RAP involved Government of Philippines Agencies from the ‘Collective

Strengthening of Community Awareness of Natural Disasters’ (CSCAND) collaborating with scientists

from Geoscience Australia (GA), as well as other agencies from the Government and University

sectors in the Philippines, in the development of natural hazard risk information for the Greater Metro

Manila Area (GMMA).

The anticipated outcomes of the project were:

Basic datasets fundamental to natural hazard risk analysis (including a high resolution digital

elevation model) are available in GMMA for the analysis of natural hazard risks and climate

change impacts.

Technical specialists from the Government of Philippines technical agencies are better able to

assess the risk and impacts from floods, tropical cyclone severe wind, and earthquakes in the

Pasig-Marikina River Basin, and have an improved understanding of these risks.

Local Government units in GMMA are better informed about the risks of floods, earthquakes and

tropical cyclone severe wind.

Page 8: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

2 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

The enhancement of relationships between the Government of the Philippines technical agencies,

AusAID and Geoscience Australia.

This report details the results of the flood risk analysis that was undertaken within the GMMA-RAP.

The flood risk analysis technical working group included representatives from the Philippines

Atmospheric, Geophysical and Astronomical Services Administration (PAGASA), the Mines and

Geosciences Bureau (MGB), the Metropolitan Manila Development Authority (MMDA), the Laguna

Lake Development Authority (LLDA), the Department of Public Works and Highways (DPWH), and

Geoscience Australia (GA). It focusses on the flood risks in the Pasig Marikina River Basin, which

covers much if not all of the Greater Metro Manila Area, including the relatively large Marikina, Pasig

and San Juan River systems.

1.2 Area of the Study

1.2.1 Topography and Location

The Philippines is a tropical archipelago, consisting of 7,107 islands situated southeast of mainland

Asia. It covers a land area of around 300,000 square kilometres, with a north-south extent of

approximately 1,850 km, and an east-west extent of approximately 1,000 km. The Philippines is

typically mountainous (Figure 1.1), a result of high seismic and volcanic activity associated with its

location on the ‘ring-of fire’. The mountains often contain deeply incised river valleys, and extensive

tropical forest cover. The larger islands can contain larger rivers, and therefore they also feature some

significant areas of alluvial plains (e.g. the Pampanga River Delta in Luzon). In contrast, the smaller

islands tend to exhibit high relief in their centre, with a narrow, discontinuous rim of lowlands along the

coast.

Figure 1.1. Topography and location of the Philippines.

Page 9: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 3 Greater Metro Manila Area – Flood Risk Analysis

Figure 1.2. Regional setting of Manila. The left pane shows Landsat imagery. The right pane shows SRTM elevation data.

Manila is the capital of the Philippines. It is situated on the west coast of Luzon (Figure 1.2), bounded

by Manila Bay to the west, and Laguna Lake to the east. Northeast of the city are the Sierra Madre

Mountains, while to the Northwest is the Pampanga river delta. To the south is the Taal Volcano.

Figure 1.3. Key rivers around Metro Manila. Left pane is LANDSAT image, right uses SRTM elevation data.

Much of Manila is built on floodplains and deltas associated with the Marikina and Pasig Rivers, and

on coastal plains around the edges of Manila Bay (Figure 1.3). The Marikina River flows south-

southwest through the Marikina valley, and drains the mountainous upper catchment. It is connected

to Laguna Lake via the man-made Mangahan Floodway, and the Napindan River. The Marikina River

may be divided into the ‘Upper Marikina’ upstream of the junction with the Mangahan Floodway, and

the ‘Lower Marikina’ downstream of this. At the junction of the Mangahan and Marikina Rivers is the

Page 10: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

4 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

Rosario weir, a flood-control structure with 8 sluice gates which is generally closed during the dry

season (forcing all water from the Upper Marikina through the Lower Marikina), but opened during

floods to allow water from the Marikina River to pass through the Mangahan Floodway to Laguna

Lake. The low-lying land in the vicinity of the Mangahan Floodway and Napindan River rivers

represents the old Marikina river-delta system, formed where sediments from the Marikina River were

deposited as floodwaters flowed to Laguna Lake.

The Pasig River is connected with both the Napindan and Lower Marikina Rivers, and provides a

hydraulic connection between Manila Bay and Laguna Lake (Figure 1.3). It flows from the

Napindan/Marikina/Pasig junction through a topographic constriction, then a low-gradient, highly

urbanised coastal plain. The San Juan River system is a highly urbanised tributary to the Pasig River,

which drains much of Quezon City.

1.2.2 Climate

The climate of the Philippines is tropical and maritime. It is characterized by relatively high

temperature, high humidity and abundant rainfall. Fluctuations in rainfall are mainly due to the

disturbances in the monsoon flow, the easterly wave, the Intertropical Convergence Zone (ITCZ),

tropical cyclones and local weather systems. The intensity of rainfall is influenced by latitude,

topography, exposure and the season. The spatial variation of rainfall differs from one region to

another, depending upon the direction of the moisture-bearing winds and the location of the mountains

(Estoque, 1956; Kintanar, 1984; Patvivatsiri, 1972; Asuncion and Jose, 1980). The mean annual

rainfall of the Philippines varies from 965 to 4,064 mm annually.

Figure 1.4. Climate map of the Philippines based on the Modified Coronas Classification.

Page 11: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 5 Greater Metro Manila Area – Flood Risk Analysis

Based on the distribution of rainfall shown in Error! Reference source not found., four climate types

are recognized.

Table 1.1. Types of Climate in the Philippines (Source: PAGASA, Climate Map of the Philippines).

Type Description

I Two pronounced seasons: dry, from November to April, and wet during the rest of the year. Maximum rain period is from June to September.

II No dry season with a very pronounced maximum rain period from December to February. Minimum rainfall occurs during the period from March to May.

III No very pronounced maximum rain period with dry season lasting only from one to three months, either from the period from December to February or from May to March.

IV Rainfall is more or less evenly distributed throughout the year.

Metro Manila in general experiences two pronounced seasons (Type I Climate Type), i.e., dry from

November to April, and wet during the rest of the year. The maximum rain period is from June to

September. In Science Garden (Diliman, Quezon City), the mean annual rainfall is 2,574.4 mm (1981-

2010). Historical records of climatological extremes of rainfall (1961-2010) in Science Garden show

that the greatest 24-hr rainfall occurred on 26 September 2009 during the passage of Tropical Storm

Ondoy in Metro Manila with 455.0 mm of rainfall. This record exceeded the normal monthly values for

September (1981-2010) in Science Garden which is 451.2 mm.

1.2.3 Population

Metro Manila comprises 16 cities and 1 municipality, with populations described in Table 1.2. The

population comprises 13% of the national population, which is 11,855,975 against 92,337,852, based

on the National Statistics Office, 2010 Population Census. It has a density of 18,925 per km², although

this varies spatially (Figure 1.5). The growth rate (based on the of 2010 census) is 2.02%. Taguig,

Parañaque, Las Piñas, and Pasay are considered to have relatively higher growth rates from 1990 to

2010. San Juan has negative growth rate due to migration or relocation of informal settlers to

municipalities in Rizal. Among the cities of Metro Manila, Manila is considered the most densely

populated, followed by Pateros, Mandaluyong, Caloocan, Navotas, Malabon, and San Juan. The

population sprawls from western central part towards the north, south and east. The most eastern

parts where population is relatively sparse are portions of Montalban, San Mateo, and Antipolo, mostly

in the upstream of Marikina River and its tributaries. These municipalities are outside Metro Manila.

Page 12: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

6 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

Table 1.2: Population over time in cities around Metro Manila (Source: 2010 Census, National Statistics Office).

Cities

Total Population Population

Density (per km

2)

Growth Rate

1990 2000 2010 Area km2

1990-2000

2000-2010

1990-2010

Las Piñas 297,102 472,780 552,573 41.54 13,302.19 4.75 1.57 3.15

Makati 453,170 471,379 529,039 27.736 19,074.09 0.39 1.16 0.78

Malabon 280,027 338,855 353,337 15.76 22,419.86 1.92 0.42 1.17

Mandaluyong 248,143 278,474 328,699 11.26 29,191.74 1.16 1.67 1.41

Manila 1,601,234 1,581,082 1,652,171 38.55 42,857.87 -0.13 0.44 0.16

Marikina 310,227 391,170 424,150 21.5 19,727.91 2.34 0.81 1.58

Muntinlupa 278,411 379,310 459,941 46.7 9,848.84 3.14 1.95 2.54

Navotas 187,479 230,403 249,131 10.77 23,131.94 2.08 0.78 1.43

Parañaque 308,236 449,811 588,126 47.69 12,332.27 3.85 2.72 3.28

Pasig 397,679 505,058 669,773 31 21,605.58 2.42 2.86 2.64

San Juan 126,854 117,680 121,430 5.94 20,442.76 -0.75 0.31 -0.22

Valenzuela 340,227 485,433 575,356 44.58 12,906.15 3.62 1.71 2.66

Caloocan 763,415 1,177,604 1,489,040 53.33 27,921.25 4.43 2.37 3.39

Pasay 368,366 354,908 392,869 19 20,677.32 -0.37 1.02 0.32

Pateros 51,409 57,407 64,147 2.1 30,546.19 1.11 1.12 1.11

Quezon 1,669,776 2,173,831 2,761,720 161.12 17,140.76 2.67 2.42 2.55

Taguig 266,637 467,375 644,473 47.88 13,460.17 5.77 3.26 4.51

Total 7,948,392 9,932,560 11,855,975 626.456 18,925.47 2.25 1.78 2.02

Page 13: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 7 Greater Metro Manila Area – Flood Risk Analysis

Figure 1.5. The spatial distribution of population density around Metro Manila (based on the Exposure Database, as detailed in the report on Exposure Information Development).

As early as 1942, the first recorded flood instance seriously affected the lives of the residents in

Manila. Major floods subsequently occurred in 1948, 1966, 1967, 1970, 1972, 1977, 1986, 1988,

1995, 1996, 1997 (Bankof, 2003). The degree of seriousness of drainage issues was illustrated in

August to September 1999 when Manila was stricken by knee-deep to neck-deep floods which

rendered most of the roads in Metropolis unpassable leaving thousands of people stranded in the

commercial and business centres of Makati and Manila (DICAMM, 2005). Typhoon Winnie came in

2004, leading to at that time the highest observed water elevations in Sto. Nino in the Marikina Valley

(EFCOS Staff, Personal Communication). TS Ondoy came in 2009, leading to the worst flooding in

recent memory. Typhoon Pedring came in 2011, causing moderate flooding in the Marikina Valley,

and a storm surge that lead to widespread coastal inundation. Since then, large floods also occurred

in 2012 and 2013 due to enhanced monsoonal rains (‘Habagat’ events).

Over the years, this problem has been enhanced in Metropolitan Manila and suburbs by rapid urban

expansion, inadequate river channel capacity and disappearance of waterways due to increase of

colonies of informal settlers. And in the core of Manila, inadequate drainage capacity was also one of

the key factors that contribute to the perennial flooding. Land Subsidence has also been identified as

important, given that much of Metro Manila is already near mean sea level (Rodolfo and Siringan,

2006).

Page 14: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

8 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

2 Literature Review

2.1 Hydraulic Studies

As early as 1952, local engineers from the then Ministry of Public Works, Transportation and

Communication formulated a Plan for the Drainage of Manila and Suburbs (MPWTC, 1952). Methods

used to compute the capacities of the drainage and pumping stations included the rational formula,

uniform flow formula and the mass runoff computation. This was still the era of slide rule. From this

report, the mean tide values of Manila were described. The report mentioned the Bureau of Public

Work’s Datum from a series of observation over a period of 19 years (1901-1919). This datum has

been a source of confusion amongst hydraulic modellers ever since, and the datum has not been

updated up to now.

In the 1970s and 1980s, Japanese consultants employed storage function models and other models

simulated through FORTRAN. As part of planning of flood control projects in Metro Manila in the early

1980’s, the National Hydraulics Research Center (NHRC) was tapped by the Japanese consultants to

probe the hydraulic conditions in the Pasig–Marikina River and Laguna Lake Complex through a

physical model. A downscaled physical model was laid out in the NHRC laboratory to examine the

effects of Mangahan Floodway and the Hydraulic Weir on flooding in the downstream Pasig River.

This simulation served as one of the bases for the planning and design of flood control structures at

different hydraulic conditions.

In 1983 and 1986, the Napindan Hydraulic Control Structures and Mangahan Floodway were

completed, respectively. To synchronize their operation along with other flood control structures, a

telemetered monitoring system, the Effective Flood Control Operation System (EFCOS), was

established in 1993. The system monitored the rainfall and water levels in the Pasig-Marikina River

and Laguna Lake Basins, while simulations gave forecasts for the operation purposes. The Japanese

storage function model was used in the hydrological modeling. River cross-sectional data of 1986

were inputs in the non-uniform and non-steady flow simulation using FORTRAN language. Rainfall

and water level data were inputted using the text editor (MIFES) and read by the executable

programs. Then graphical presentation of the results were shown in Lotus 123 and disseminated to

the officials of DPWH.

Parallel to this activity, the Pasig River Rehabilitation Secretariat, under the Department of

Environment and Natural Resources, acquired the EFCOS data for its clean-up project of Pasig River.

They simulated the flushing of high water level of Laguna Lake via Napindan Channel to Pasig River

using Mike 11. Seeing the need for more dense hydrological data, the EFCOS was enhanced with

more rain and water level gauges in 2001. Mike 11 and Arc View 3.2 simulated the rainfall and water

level forecasts in 2D animation. A related model was developed for flood forecasting, including data

assimilation capabilities based on Kalman Filtering (Madsen and Skotner, 2005). Unfortunately the

Mike 11 forecast model at EFCOS required an annual renewal of license, and due to budgetary

constraint and agency’s priorities it was not sustained.

In 2005, the Japan International Cooperation Agency (JICA) completed an update of drainage plans in

the Core Area of Metro Manila (DICAMM, 2005). The focus of simulation was on the drainage

discharges only, where the DHI Mouse was suitable for this purpose, and was used in combination

Page 15: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 9 Greater Metro Manila Area – Flood Risk Analysis

with ESRI GIS software. For the Japanese consultants, the Storage Function Model, Flowca, etc.,

were very widely used especially in the crafting of different Master Plans and Feasibility Studies in

Metro Manila. A related study around this time was by CTI Engineering (2005), which assessed

drainage in the Mangahan/Napindan/Taguig area, using rational-method rainfall-runoff type models

combined with models of the performance of pumping stations. The focus was the development of

operation rules for pumping stations.

Badilla (2008) developed a simulation of river flows in the Upper Marikina, based on linking the HBV

rainfall-runoff model with the DUFLOW 1D unsteady hydraulic model. Emphasis was placed on

predicting water levels and the associated uncertainty. The model was calibrated to match river gauge

levels at Sto Nino, and showed reasonable capacity to predict flood peaks measured during the 2003-

2004 wet seasons.

Tropical Storm Ondoy came in 2009, and caused extreme flooding in Metro Manila. The event

triggered the World Bank to finance the updating of Master Plan in Metro Manila and Surrounding

Areas (WBCTI, 2012). In laying out the flood extent of Ondoy and the countermeasures for this study,

Mike Flood and HEC-RAS were used for the Pasig Marikina River Basins. Many cross sectional data

came from the Pasig River Rehabilitation Commission, the DPWH, and others. However, there was a

problem related to the reference datum because of varied references used by different elevation

datasets, which created confusion. Discrepancies in some datasets lead the consultant to use the

2002 survey data from DPWH. Since Mike 21 requires extended cross sectional points from the bank,

the one meter interval contour was the most likely source for modelling. For the surrounding areas

without the cross-sectional data, the consultant relied mostly on the hydrologic analysis and past

studies.

Soon after Tropical Storm Ondoy, Abon et al. (2010) developed a hydraulic model of flow in the

Marikina River. Their semi-distributed model combined the SCS Curve-Number loss model with the

SCS Unit Hydrograph, using the HEC-HMS modelling environment. With this they could well simulate

the timing of Ondoy’s flood peaks in the upper portions of the Marikina River, as compared with the

observed timing of the flood peak based on interviews of affected residents. However, the simulation

of the time-to-peak in the lower-gradient parts of the Marikina was less accurate, which the authors

attributed to other sources of floodwaters and backwater effects that were not included in their model.

Muto et al. (2011) considered the impact of climate change on flooding in Metro Manila, using a

mixture of storage function models, 1D and 2D flood modelling. They included extensive analyses of

economic damages predicted under various scenarios related to climate change and the development

of structural measures to reduce flooding. This study concluded that flood damages in Metro Manila

would continue to be highly significant under any scenario, but could be greatly reduced by continuing

implementation of structural measures proposed in the Master Plan.

Santillan et al (2012) developed a near-real-time forecast model for the Upper Marikina River, based

on the integration of HEC-HMS and HEC-RAS. Remotely-sensed data was classified and used to

parameterize the roughness and runoff coefficients in the model. They also developed methods to

automatically run the model and broadcast the predicted flood extents on the web. The model showed

good performance in simulating a flood peak during June 2012.

As of 2013, there are also several projects disseminating model-based online flood hazard maps for

the Philippines (including Manila). Key sites include Project NOAH (www.noah.dost.ph), and

Nababaha (www.nababaha.com). The description at the Nababaha site suggests these maps are

Page 16: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

10 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

based on direct-rainfall 2D hydraulic modelling, combined with the use of remote sensing information

to estimate relief and terrain roughness.

2.2 Mines and Geosciences Bureau Flood Susceptibility Mapping

The Mines and Geosciences Bureau has been conducting geohazard mapping and assessment since

the 1960's. But it was in 1999 when the Cherry Hills landslide incident emphasized the need for

hazard assessment as a subcomponent of the Environmental Compliance Certificate (ECC), which is

a requirement for land development projects. The subcomponent is called the Engineering Geological

and Geohazard Assessment Report (EGGAR). Following the implementation of these requirements,

the Guinsaugon landslide incident in February 2006 leads the government to take further action to

address the geohazards problem in the Philippines.

Through a directive from the President of the Philippines, the Geohazard Assessment and mapping

Program was immediately implemented by MGB with the main objective of identifying areas

susceptible or vulnerable to various types of flood and landslide hazards and to increase public

awareness of such hazards in order to mitigate the negative impacts.

The major activities are the Geohazard Mapping at 1:50,000 and 1:10,000 scale, conduct of

province/municipal-wide Information and Education Campaign (IEC), provision of hazard maps and

threat advisories and identification and assessment of relocation sites, assessment of evacuation

centers and establishment of community-based early warning system.

Prior to the field survey, a desk study of the area which includes the interpretation of available aerial

photographs and land satellite imageries is done. The data are then validated in the field by a team of

geologists. An example map for Metro Manila is in Figure 2.1.

In the field, a field data sheet is completed to standardize the procedure of the assessment. A hand-

held Global Positioning System or GPS is used to get the location of the observation point. For flood

assessment, the susceptibility ratings are based on the topographic location, land cover, flood type,

depth and duration of inundation, flood frequency, direction of flood water and cause of flooding

among others. These are verified in the field through anecdotal accounts of residents living in the

area. Photos of the affected areas, the recorded flood height, as well as the source of flood water are

also taken as these are important in the preparation of geohazard assessment report.

Page 17: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 11 Greater Metro Manila Area – Flood Risk Analysis

Figure 2.1. Flood and landslide susceptibility map of Metro Manila from the Mines and Geosciences Bureau.

2.3 General Approach to Flood Risk Analysis

Flood risk analysis involves the combination of: 1) flood hazard information, which describes the

likelihood and intensity of a flood event; 2) exposure information, which describes the distribution of

people or elements ‘at-risk’ to a flood event; and 3) vulnerability information, which describes how the

exposed elements would be affected if subject to a given intensity of flooding. The typical form of each

of these inputs is described in the next three sections.

2.3.1 Flood Hazard Information

Flood hazard information consists of one or more flood scenarios, each with an annual exceedance

probability (AEP) which describes the likelihood a flood event larger than the scenario occurring in any

given year. For example, if a flood scenario is assigned an AEP of 1/20 (=0.05), then ideally 95% of all

years should experience floods smaller than the scenario, while 5% of all years will have floods larger

than the scenario.

Page 18: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

12 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

The flood scenarios typically provide an estimation of the inundated area (‘flood extent’), peak flood

depths and/or peak flood discharge, based on a flood inundation model. Depending on the model

complexity, other information may also be provided, such as the flood duration, the peak flow velocity

or momentum over the floodplains. All such information should be considered an estimate only, which

is susceptible to errors both because of model simplifications and uncertainties in the required input

data. Typical input datasets include topography, historical hydrological information, and estimates of

the ‘flow roughness’ of the terrain.

The AEP for the flood scenario is usually assigned based on a statistical analysis of hydrological

records at the site, such as the peak river discharge, the rainfall intensity and duration, and/or the

peak water levels somewhere within the flooded region. Exactly what sort of information is needed

depends on the geography of the study site. However, most sites will at least need information on

either river discharge or rainfall (the latter may be converted to river discharge using a rainfall-runoff

model). As an example of when other information might be required, consider a flood scenario for a

river which flows into a lake. If the lake level has a large ‘backwater’ effect on water levels in the river,

then it may be necessary to account for the likelihood of high lake levels, as well as the rainfall and/or

discharge inputs, to develop the flood scenario.

The flood scenario would then be created by inputting the hydrological information into a flood

inundation model. As a simple example, a 1/100 AEP flood scenario may be developed by estimating

the river discharge which is exceeded in only 1% of all years (based on statistical analysis of historical

discharge records), and then running a hydraulic model with this discharge to estimate the resulting

flood extents. Often hazard analyses include flood scenarios with several AEPs (e.g. 1/10, 1/25, 1/50,

1/100).

Even for a single AEP, generally more than one flood scenario is possible. For example, even if the

average rainfall intensity is fixed by the AEP, the spatial and temporal distribution of the rainfall time

series would have some effect on the flood behaviour, leading to an infinite number of flood events

with the same AEP. Other factors may also affect the results, such as assumptions regarding the

failure of levees, or blockages of drains by debris.

It is usually not possible to treat the full range of flood scenarios associated with a single AEP. Most

often a single scenario is taken to be broadly representative of the flooding for a given AEP (FEMA,

2003; Scawthorn et al., 2006). Alternatively, some studies treat this issue using a ‘Monte-Carlo’ type

approach where models for the variation in hydrological inputs are developed, and a large number of

scenarios are modelled to attempt to cover the range of flood behaviour with the given AEP (Aronica

et al., 2011; Christian et al., 2012).

2.3.2 Exposure Information

Exposure information describes the spatial distribution of ‘elements at risk’ from a flood, such as

people, buildings, and critical infrastructure. Examples might be spatial information on population

densities, the distributions of buildings of different types, and the locations of critical infrastructure

such as hospitals and schools. See the report on Exposure Information Development for more

information.

Page 19: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 13 Greater Metro Manila Area – Flood Risk Analysis

2.3.3 Vulnerability Information

Vulnerability information describes how the ‘elements at risk’ as susceptible to damage or loss when

affected by a flood with particular properties. For example, a common type of vulnerability information

is a ‘depth-damage curve’ or ‘stage-damage curve’, which describes the damage to a building as a

function of the depth of flooding above the building floor. See “Development of vulnerability curves of

key building types in the Greater Metro Manila Area, Philippines” (Pacheco et al., 2013) for more

information.

For a single hazard scenario (e.g. AEP=1/100), the flood, exposure and vulnerability inputs can be

combined to compute a number ‘damage’ metrics associated with the flood event. For example, the

cost of building damage could be computed by 1) Calculating the depth of inundation at each building

based on the flood scenario; 2) Estimating the associated damage to the building, using a depth-

damage curve; 3) Aggregating the results to compute the total damage cost, or map the spatial

distribution of damage. Similarly, one might compute the number of people whose homes were

inundated, or the damaged building floor area.

Flood risk analyses explore the results of one or more flood scenarios and damage metrics at the site

of interest, to develop an overall picture of the flood risk. The scenarios and damage metrics that are

analysed can vary widely, depending on the purpose and resources of the study. Often the results will

be aggregated over politically useful areas, such as local government administrative boundaries. For

example, Middelman (2002) estimated the building damage cost associated with a single hazard

scenario (a widespread 1%AEP flood event) in South East Queensland, Australia, and mapped the

results aggregated to census districts. This demonstrated the relatively high building damages

associated with such an event (~1% of the total cost of the building stock), and their highly non-

uniform spatial distribution. Where a range of hazard scenarios are considered, they may be

associated with a range of AEPs (Poretti and Amicis, 2011), levee breaches (Castellarin et al., 2011),

or climate change (Dumas et al., 2013). A suite of hazard and damage maps and aggregated damage

estimates may be produced. If a very large number of events can be simulated, it may be possible to

directly compute the probability of damage levels, or the average annual damages (Apel et al., 2004;

Bouwer et al., 2009). Because the latter approaches require the simulation of many scenarios, they

require large scale computing resources except when based on very simple and computationally

efficient flood models.

2.4 Introduction to flood inundation models

Flood inundation models are one of the fundamental tools of flood risk analysis. They are very diverse,

and so a brief introduction to them is included here. Flood inundation models most often consist of a

combination of two separate models: 1) a rainfall-runoff model and; 2) a hydraulic model. The rainfall

runoff model is used to simulate the transformation of rainfall through the catchment to the river

system, while the hydraulic model simulates flows within the river system and floodplains.

2.4.1 Rainfall Runoff Models

Given an input rainfall time series, a rainfall runoff model is used to predict the time series of discharge

(water outflow) from one or more catchments (Beven, 2001; Pechlivani-dis et al., 2011). This outflow is

then fed into a hydraulic model to perform flood inundation computations.

Page 20: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

14 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

Broadly, rainfall runoff models may be classified based on their spatial and temporal complexity.

Spatially, they are classified as one of:

1. Lumped models: These assign a homogenous description to the entire catchment. As a result,

they have the lowest spatial complexity and input data requirements of any rainfall runoff

model. Despite their simplicity, in some situations they can perform well, while enabling a

more rapid assessment of a site’s runoff characteristics than more sophisticated methods

(Beven, 2001).

2. Semi-distributed: Semi-distributed models describe the catchment with an interconnected set

of lumped models. They thus allow some treatment of the spatial variation of catchment

properties and rainfall inputs at a site, while retaining much of the simplicity of lumped models.

They are commonly used in applications (e.g. Moramarco et al., 2005; Abon et al., 2010).

3. Distributed: Distributed models represent the limiting case of a semi-distributed model, where

the catchment is approximated as a continuous region (although in practice, it will split into

small ‘cells’ or ‘pixels’). They have the strongest data input requirements, and also the most

flexibility in terms of treating variations in rainfall and catchment properties, which can be

advantageous in some situations (Pechlivanidis et al., 2011). However, uncertainties in the

definition of input parameters and process descriptions mean that distributed models do not

consistently outperform simpler lumped or semi-distributed rainfall runoff models, although this

is an ongoing area of research (Beven, 2001; Ghavidelfar et al., 2011; Pechlivanidis et al.,

2011).

From a temporal perspective, rainfall runoff models may be classified as either:

1. ‘Event Based’ models: Event based models are suitable for simulating a single flood event

(e.g. of 48 hours duration). Because of their limited temporal extent, they focus on modelling

the ‘direct runoff’ or ‘quick runoff’ which forms the bulk of the flood peak, and are not

concerned with detailed modelling of the ‘baseflow’ which is affected by water stored in the

catchment over longer periods. They are most often used for flood hazard assessment when

individual rainfall events are the major cause of flooding.

2. ‘Continuous’ models. Continuous models are designed to be applied to longer-term rainfall

series. While these also simulate the ‘quick flow’ described above, they also focus more on

accurate simulation of the baseflow component of runoff than do event based models.

Typically this is done by accounting for the time evolution of soil moisture conditions and

longer-term catchment storage. As such, continuous models are useful for simulating runoff

over longer time periods. This can be important for flood forecasting where the antecedent soil

moisture conditions can be estimated to improve flood prediction (Berthet et al., 2009). Such

information is not usually available in hazard applications, because they refer to hypothetical

future events. However, continuous models are sometimes used in flood hazard work to

create a synthetic time-series of flood events over long time periods (e.g. Winsemius et al.,

2012). Flood statistics are then estimated directly from the simulated time series.

2.4.2 Hydraulic Models

Hydraulic models are used to simulate the movement of water through channels and over floodplains.

As with rainfall runoff models, a wide variety of hydraulic models exist. The following classification is

based partly on Woodhead et al., 2007:

Page 21: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 15 Greater Metro Manila Area – Flood Risk Analysis

1. 0D models: These use GIS type methods to interpolate the water surface elevation between

known sites, using topographic data. They are typically used for a ‘quick’ estimate of likely

flood extents, or in situations where other flood modelling methods would be too

computationally demanding (e.g. Global Flood Modelling, Winsemius et al., 2012). Typical

computational time is seconds.

2. 1D channel network models: These simulate the water surface elevation and discharge along

‘channels’ (i.e. user-defined flow paths). Floodplain inundation may be simulated using

‘extended cross-sections’ which account for the flow and storage of water over floodplains, or

using a ‘quasi-2D’ network of channels to represent the floodplain (e.g. Dung et al., 2013). The

flow direction is restricted to the along-channel direction, and so must be pre-defined when the

model geometry is set up. They are widely used for flood modelling, but are most suited to

situations in which the flow directions are well known and flows occur dominantly in channels.

Typical computational time is minutes

3. 1D+ models: These extend 1D models with storage areas, which are conceptually ponds

represented with a volume-water level relation. Storage areas provide another mechanism to

represent floodplain storage and flow, which allow flow in multiple directions (as opposed to

just up-or-down channels). Flow between storage areas and other storage areas or channels

is modelled using simplified momentum relations, such as a weir relation or diffusive flow

relation. They are widely used in flood modelling applications over large areas, or where

computational time is limited (e.g. Castellarin et al., 2011). Typical computation time is minutes

to hours, depending on the geometric detail.

4. 2D- models: These simulate the motion of flood waters in 2 horizontal dimensions, using mass

conservation and simplified 2D flow momentum conservation equations. Two dimensional

models are more computationally demanding than 1D models, but have the potential for

greater accuracy in situations in which flow paths cannot be well schematised using the 1D

approach. 2D- models are widely used in research, and in applications (e.g. Neal et al., 2012).

Typical computational time is hours to days or longer, depending on the geometric detail.

5. 2D models: These solve the 2D shallow water equations or some variant, and simulate the

motion of flood waters in 2 horizontal dimensions (e.g. Horrit and Bates, 2002). The theory

underlying 2D models requires less approximation than does the theory of 1D or 2D- models.

This has the potential to support greater model accuracy, so long as the model is run with a

‘sufficiently fine’ mesh resolution to: 1) well describe the flow topography and; 2) avoid large

discretization errors in the numerical method. In practice, meeting these criteria can be

computationally demanding if the model needs to be run over large areas of complex terrain,

or resolve fine topographic features (e.g. small channels or streets). If overly coarse mesh

resolutions are used, then 2D models may be either more or less accurate than simpler but

more computationally efficient approaches (e.g. 1D or 1D+), and performance will depend

heavily on the nature of the site and details of the model setup. 2D models are widely used in

flood modelling applications. Typical computational time is hours to days or longer, depending

on the geometric detail. Parallel computation is often used to help manage the computational

demands.

6. 1D-2D models: These combine 1D and 2D flow models for efficient representation of flow both

in channels and over floodplains. They are widely used in flood modelling applications, and

are often easier to apply with acceptable accuracy over large areas than 2D models (Syme et

al., 2004). This is because many computations which are demanding with a fully 2D approach

(e.g. accurate modelling of channel flow) can be performed efficiently in 1D. Typical

Page 22: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

16 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

computational time is hours to days or longer, depending on the geometric detail. Parallel

computation is often used to help manage the computational demands.

7. 3D models: These simulate fully 3-dimensional flow using e.g. the Reynolds Averaged Navier

Stokes Equations. These models are rarely used in flood hazard applications at present

because they are so computationally intensive, although are often used in ocean and lake type

applications where the model resolution does not have to be as detailed as generally required

for flood inundation modelling. Typical computational time is hours to days or more, depending

on the geometric detail. Parallel computation is often used to help manage the computational

demands.

The suitability of each type of model for a given application depends on the site characteristics, the

needs of the user, and the available computational power. Usually both the model accuracy and the

computational demands increase as the model resolution is refined (i.e. mesh size or cross-sectional

spacing is reduced, more topographic detail is added). For an equivalent level of topographic detail,

3D models are typically the most computationally intensive, followed by 2D and 2D-, 1D+, 1D, and 0D

models. Similarly, 3D models are based on the least restrictive physical assumptions, followed by 2D

and 2D-, 1D+, 1D, and 0D.

In all cases, model results will depend on the decisions made when setting up the model. For all

model types (except 0D), results will typically improve with calibration against observed data. This is

usually achieved by ‘tuning’ the model friction coefficients to obtain acceptable agreement with

observations of flood depths or flood extents, while keeping them within physically acceptable ranges.

In 3D and 2D models, the model resolution (and hence accuracy) will often be limited by the available

computational power. Long model run times are an option, although this may make the iterative

process of model calibration difficult. For 1D and 1D+ models, computational time is less of a problem:

however, the modeller must ‘schematize’ the flow paths into channels (and storage areas in the 1D+

approach). To get a good solution in a complex situation, the schematization process may be very

time consuming with many iterations, and requires good judgement on the part of the modeller. Even if

well designed, the assumptions underlying this approach are not appropriate for all flow problems (e.g.

complex floodplain flow paths which change strongly during a flood event).

If estimates of velocities over the floodplains are required, then 3D or 2D/2D- models should be used.

In urban areas, accurate velocity simulation typically requires a very fine mesh resolution (e.g. 1-2m),

and hence a long computational time over large areas (Smith and Wasko, 2012). If coarser resolutions

are used, then velocity estimates are unlikely to be accurate.

On the other hand, if only water depth estimates are required, then coarser resolution 2D or 1D

modelling may be appropriate. Horritt and Bates (2002) compared a 2D, 2D- and 1D model for

simulating flow routing and inundation over a 60km reach in the River Severn, and found that all

models performed quite well, given the uncertainties in inflows, topography and validation data.

However, the predictive power of the 1D model was best, followed closely by the 2D model, while the

2D- model performed less well. Aureli et al. (2006) compared a fully 2D and 1D+ model of levee-break

flooding in the Po River, Italy. They found that the flooded areas, maximum depths and the discharge

flowing through the breach were quite comparable between the two approaches, noting that these

were the most important variables for hazard assessment. However, the detail of the dynamics

differed just after the initial breach, reflecting limitations underlying the 1D+ model.

Page 23: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 17 Greater Metro Manila Area – Flood Risk Analysis

Another issue relates to the importance of accurately simulating channel flow, and other small flow

paths. Some 2D models require a very high resolution (and hence long computational times) to do this

accurately, whereas it is usually straightforward in 1D or 1D/2D models. In some sites, this can have

an important impact on modelled flood extent, since small flow paths may be the main conveyer of

floodwaters. For example, Syme et al. (2004) compared the peak depth predictions of four hydraulic

models in Bristol, England, which variously employed 2D, 1D/2D, 1D+ and 2D- approaches. All

models showed differences in simulated flood depth time series at point locations, reflecting the

different assumptions on which they are based. However, simulated inundation extents were most

similar for the 1D/2D and 1D+ models, whereas the flood extents for the 2D and 2D- were quite

different from these and from each other, suggesting that the ability to resolve fine flow paths was

important at that site.

Page 24: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

18 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

3 Methods

3.1 Data

Among the data used in this study are:

1. Elevation data;

2. Hydrologic data (rainfall and water level)

3. Exposure data; and

4. Vulnerability data

3.1.1 Elevation Data

The key elevation data sources used in this study are:

1. The LIDAR Digital Elevation Model (DEM) for Metro Manila: This is a LIDAR-derived ‘bare-

earth’ elevation raster dataset, with a pixel size of 1m2, which was produced by FUGRO from

the raw LIDAR point data. The latter was flown in March 2011, and has a vertical accuracy of

+/- 15cm (1 standard deviation) in bare earth areas, and a point density of approximately 4

points per m2. Spatially the LIDAR DEM covers all areas of the Pasig Marikina Catchment

included in the hydraulic model in this study. Although considered accurate in terrestrial areas,

the LIDAR DEM is not accurate in ‘water areas’ such as rivers, lakes and ponds, which were

underwater at the time of data capture. It was used as the main source of elevation data in this

study - chiefly for defining the hydraulic model geometry outside of the water areas, and for

the computation of catchment boundaries to guide the rainfall runoff and hydraulic model

schematization. The technical report for Component 1 – High Resolution Elevation and

Imagery contains further details about the acquisition and processing of LiDAR elevation data.

2. Shuttle Radar Topography Mission (SRTM) DEM: This is a globally available mid-resolution

elevation dataset, with a pixel size of around 90 m, and a nominal vertical accuracy of 15 m

(although this varies depending on the nature of the land surface). It was used to compute

catchment boundaries in parts of the upper Pasig-Marikina Catchment which are not covered

by the LIDAR DEM.

3. River Cross-Sectional Surveys: These were variously provided by DPWH, the World Bank

/CTI flood modelling team, MMDA, and the ‘Study on Drainage Improvement the Core Area of

Metropolitan Manila, Republic of the Philippines’ (DICAMM, 2005). The cross-sectional

surveys included the Pasig, San Juan, Marikina, Mangahan and Napindan Rivers, and various

creeks in the western part of Manila. The DPWH data included major bridges as of 2002.

Cross-sections were typically spaced at 50-200 m intervals along the rivers, and were used to

estimate river bed elevations for the hydraulic model, in ‘water areas’ where the LIDAR DEM

was not suitable.

Page 25: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 19 Greater Metro Manila Area – Flood Risk Analysis

3.1.1.1 Vertical Datum Issues

Around Manila, most topographic and water level data is reported with respect to a local vertical datum

(termed ‘DPWH datum’). In DPWH datum, mean sea level is often reported as being at 10.47m (e.g.

DICAMM, 2005), although other reports suggest 10.6m (RFCOSMM, 2002). The variation in MSL with

respect to DPWH datum may reflect subsidence around Manila (e.g. DICAMM, 2005). In the LIDAR

DEM, vertical elevations are reported with respect to MSL.

While the study team is not in a position to resolve the ambiguities in vertical data around Manila, a

method was required to translate between the DPHW vertical datum and the LIDAR DEM MSL. To

support this, the team compared cross-sectional survey data and LIDAR DEM data in multiple sites

throughout the basin, and also compared the LIDAR DEM with observed high water levels in Bay

Breeze, Taguig in November 2011 (where the water elevation could be simultaneously measured with

respect to the location of the wet-dry boundary on the LIDAR DEM, and with local water level gauges

in DPWH datum). Based on these comparisons, the team found a constant offset of 10.5 m was

reasonable for translating between the LIDAR DEM elevations and other measurements in DPWH

datum. This is close to the reported values of MSL in DPWH datum, as expected. However, it is worth

noting that if differential subsidence is occurring over Manila, it may be more appropriate to apply a

spatially varying adjustment to integrate old survey data with the newer data. This would require a

good understanding of the spatial distribution of subsidence around Manila, and has not been pursued

in the present study.

3.1.2 Hydrological data

3.1.2.1 Rainfall

The hourly rainfall data was provided by the EFCOS project in MMDA (stations shown in 3.1). The

data covered the period 2002-2011, although coverage at individual stations was often less because

of instrument malfunction. In addition, daily rainfall data was provided by PAGASA (stations shown in

3.1), with record lengths varying from less than 10 up to 50 years.

Figure 3.1. Location of Stage and Rain Gauge measurements at key sites around Manila.

Page 26: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

20 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

3.1.2.2 Water Levels

Hourly water level data at gauges in key rivers around Manila was provided by the EFCOS project in

MMDA (stations shown in Figure 3.1). The data covered the period 2002-2010, although coverage at

individual stations was often less because of instrument malfunction. This was supplemented with

daily water elevation data for Laguna Lake covering the period 1919-2010 (provided by LLDA), and

tide gauge measurements from Manila South Harbour (Port Area) in 2008-2009 (provided by

NAMRIA), which were used to cross-check results from some of the EFCOS stations. In addition,

hourly water level observations in September-October 2009 were obtained from Labasan Pumping

Station, both inside and outside the flood defences that protect Taguig and Pateros from high water

levels in Laguna Lake. Hourly water level observations during part of Tropical Storm Ondoy were also

obtained from MMDA, based on manual gauge recordings.

In addition, depths on the floodplains of Manila during Tropical Storm Ondoy were estimated using

data provided by the Philippines Flood Hazard Maps project (www.nababaha.com). This is collated

from citizens reports of flood depths, reported categorically as ‘No flooding’, ‘Ankle deep’, ‘Knee

Deep’, ‘Waist Deep’, ‘Neck deep’, ‘Top of head deep’, ‘1-storey high’, ‘1.5-storeys high’, ‘2-storeys or

higher’. To compare these with flood model depths, they were assumed to correspond to depth ranges

of (0-0.1), (0.1-0.25), (0.25 – 0.7), (0.7-1.2), (1.2-1.6), (1.6-2.0), (2.0-3.0), (3.0-4.5), (4.5+) metres

respectively. The boundaries between these categories are necessarily subjective, and given the

nature of the underlying data, not all records are expected to be accurate. Despite these limitations,

the data gives a useful picture of inundation over the floodplains and is complementary to the gauged

water levels in the rivers.

3.1.2.3 Vertical Datum Issues

While the EFCOS water level data was nominally in DPWH datum, some of the gauging stations

appear to have a ‘drift’ in their vertical datum over time. EFCOS staff indicated that funding has not

always been available to maintain the stations, which has compromised the quality of the data in some

instances.

The present study is most affected by data quality during the flood model calibration event (Tropical

Storm Ondoy, September 2009). To check this, the EFCOS water level data at Fort Santiago was

compared with independent water level measurements at Manila South Harbour (Port Area) in 2008

and 2009. This data exhibits has a stable mean sea level in these years (difference of 2mm in 2008

and 2009). Hydraulically, low water surface gradients are expected in Manila bay and the mouth of the

Pasig River except during high river discharge events, and thus most of the time Manila South

Harbour is expected to have fairly similar water levels to the Fort Santiago gauge. The 2008 data was

used to provide a dry-season (low river discharge) comparison of the two gauges, as the Fort

Santiago gauge was not functioning during the dry season in 2009. MSL for the Manila South Harbour

data was computed from the data itself in 2008 and 2009, so the comparison does not depend on the

assumed datum at Manila South Harbour.

Figure 3.2 shows that if the Fort Santiago gauge is assumed to be in DPWH datum (MSL~=10.5) in

2008 and 2009, then the water levels at Fort Santiago are unrealistically elevated about 0.5m above

those in Manila South Harbour in both the wet and dry seasons. They also regularly exceed commonly

used design high water levels for Manila bay in both the wet and dry seasons (~0.9m above MSL,

DICAMM, 2005; WBCTI 2012). This is unrealistic given the close proximity of the Fort Santiago gauge

to Manila South Harbour (Figure 3.1). For example, during periods of low river flows there is a tidally

Page 27: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 21 Greater Metro Manila Area – Flood Risk Analysis

induced reverse flow from Manila Bay into the Pasig River, which must require the Manila Bay water

levels to be slightly above the Fort Santiago water levels at some time during the tidal cycle.

If instead the Fort Santiago data is assumed to have a datum of ‘DPWH less 0.5m’ (MSL ~=11.0m) in

2008 - 2009, there is good agreement with the Manila Bay gauge during periods of low river flows (the

dry season), and in wet season periods without strong floods (i.e. prior to Tropical Storm Ondoy, 26

September 2009) as shown in Figure 3.2. The observed truncation of low water levels at the Fort

Santiago Gauge in 2008 is assumed to reflect problems with gauge maintenance. Physically, the

‘DPWH less 0.5m’ datum is more reasonable than the DPWH datum. Thus, the ‘DPWH less 0.5m’

datum (MSL=11.0) is used in this study as a first approximation to convert the raw Fort Santiago

gauge data a MSL datum for the Tropical Storm Ondoy Calibration event. Future work to accurately

establish the vertical datum for stage gauges in Manila should be undertaken, and gauges should be

monitored for any apparent datum shifts which may indicate instrument malfunction. The adjustment

proposed in this study should be considered to be a ‘rough approximation’ only, and cannot replace

such detailed work.

Figure 3.2. Comparison of measured water elevations at Manila South Harbour and Fort Santiago, during a month the wet season (top), and dry season (bottom), 2008-2009.

Page 28: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

22 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

An equivalent comparison of raw water level data at the Fort Santiago gauge with Pandacan and San

Juan was undertaken. All these gauges show similar raw water levels during the dry season (Figure

3.3). During the wet season the low tide level is truncated upstream, while high water levels are more

similar at all stations except when there are significant river flow events. This behaviour is consistent

with our general expectations of tidal dampening in estuaries with and without river flow (Savenije,

2005), and suggests that there are no large datum differences between these stations (in 2008-2009).

Thus we conclude that Fort Santiago, Pandacan and San Juan are probably all in the ‘DPWH less

0.5m’ datum during the 2008-2009 period. When reported later on, water levels at the San Juan and

Pandacan stations are provided for the ‘DPWH less 0.5m datum’, and the ‘DPWH’ datum values also

reported in parentheses.

Figure 3.3: Comparison of water levels recorded at Fort Santiago, Pandacan and San Juan gauges in 2008 and 2009, during the dry season (top) and wet season (bottom).

We also assessed gauges in Laguna Lake and the Marikina River. The station at Napindan was not

functional during Tropical Storm Ondoy. Comparison of the EFCOS gauges at Angono and Rosario

weir suggest that the latter are in the same datum as each other. Further, the Angono record agrees

with independent Laguna Lake level measurements taken by LLDA, and manual measurements at

Labasan Pumping station, which are both reported as relative to DPWH datum. Thus, these stations

are presumed to be in DPWH datum. We do not have independent measurements at the gauges

upstream of Rosario, but they are believed to be in DPWH datum also (EFCOS, personal

communication).

Page 29: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 23 Greater Metro Manila Area – Flood Risk Analysis

It is worth summarising the effect of the datum issues on the present study:

1. For the flood model calibration against Tropical Storm Ondoy, the Pasig River mouth boundary

condition is taken from the Fort Santiago Data, with the adjusted vertical datum ‘DPWH less

0.5m’ which matches with the Manila South Harbour data in 2008/2009.

2. The Laguna Lake boundary condition is based on data in DPWH datum. The Manila Bay

boundary conditions (except for the Pasig River Mouth) are based on the data for the Manila

South Harbour Station, where MSL is computed from the data itself.

3. When comparing the model against data, observed water levels at Pandacan and San Juan

are reported in both the ‘DPWH less 0.5m’ datum as well as the ‘DPWH’ datum. The modelled

results turn out to fall between these values.

4. If the model of the Ondoy flood developed in this study is run with the ‘DPWH datum’ Fort

Santiago data as a boundary condition, then the flood extent in most areas is only slightly

affected, as the upstream river discharge and rainfall is of greater significance. However, early

on in the event, there appears to be too much flooding in the downstream reaches of the Pasig

River and surrounds.

If would be very useful for further work to tie water level gauges around Manila into a consistent

vertical datum, and to ensure the availability of funding so that this can be maintained over time.

3.1.3 Exposure Data and Vulnerability Curves

Damage calculations in this study make use of the Exposure Database (described in the report for

Exposure Information Development). This provides information on the distribution of building types

and population densities in Manila. We also use the Flood Vulnerability Curves, which model the

damage to a building as a function of the local flow depth. These are described in “Development of

vulnerability curves of key building types in the Greater Metro Manila Area, Philippines” (Pacheco et

al., 2013). In addition, estimates of the replacement cost of buildings were used, based on data

reported in Appendix A.

3.2 Software Selection

Decisions on software to use for this project were made collaboratively by the study team. An

emphasis was placed on the use of free and/or open source software, to ensure its availability to all

participants both during the project and into the future.

3.2.1 Data Processing and Analyses

The open source geographic information system “Quantum GIS” (QGIS) was used to visualise and

process spatial data in this project (www.qgis.org). The study team found this software to be intuitive

for new users with some existing GIS background. Bugs were sometimes a problem, particularly early

in the project, but many were fixed as updated versions of the software were released. For batch

processing of vector and raster data, the team made direct use of the GDAL and OGR command line

tools (www.gdal.org), which are distributed with QGIS and were found to work very reliably.

In addition, the team made heavy use of the open source programming language R (www.r-

project.org) for general computation, statistical analyses and data processing. The R language is

Page 30: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

24 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

supplemented with several thousand freely available packages, of which the study team made heavy

use ‘lmomRFA’ and ‘lmomco’ for hydrological frequency analyses, and ‘raster’, ‘sp’, ‘rgdal’ and ‘rgeos’

for working with spatial data. Most of the study team had limited experience in this sort of

programming prior to the project, but over time became more familiar with the language and its

capabilities, and with using and modifying scripts to run analyses on different data sets or with

alternative processing options.

3.2.2 Flood inundation modelling

3.2.2.1 Rainfall Runoff Model

In selecting a rainfall-runoff modelling approach for the project, the team noted that:

1. Floods in the Pasig Marikina Basin are typically generated by storms with a relatively short

duration (from < 1 to a few days of intense rain). The time-delay between rainfall and flooding

throughout the basin has been estimated as ranging from 0-10hrs (Abon et al. 2010).

2. The land-uses in the Pasig Marikina basin are quite diverse (ranging from urban areas to

natural forested areas and grasslands), and as a result are expected to have strong variation

in their hydrological response to rainfall. Considering also the focus of the study on flood

hazards, the study team reasoned that an event based, semi-distributed rainfall runoff model

would be appropriate. This approach has been taken in several previous studies in the basin

(e.g. Abon et al., 2010; Muto et al., 2011).

The team chose to use HEC-HMS as a rainfall-runoff modelling tool in the present study. This was

because it can be used to implement a wide range of event-based semi-distributed rainfall runoff

models; is freely available, and extremely widely used. As such, it was considered a useful tool for the

team to learn to use, both in the present study, and potentially for future work.

3.2.2.2 Hydraulic Model

As discussed above, the suitability of a hydraulic model for a particular application depends on the

purpose of the study, the computational resources available to conduct the study, and the nature of

the site. The team assessed these factors as follows:

1. The purpose of flood modelling in the present study is to predict the peak flood depths

associated with relatively large flood events in the Pasig Marikina Basin, to support the flood

risk analysis. Flood velocity information was not considered essential.

The computational resources were limited to laptop computers which were available to the

study team. In terms of software, the team had a strong preference to use free or open source

software, as it would sustainably allow the project participants to use the software into the

future. Most commercial flood software requires ongoing annual licence fees in addition to the

upfront cost of the software, and this has created software access problems in the past for

the project participants.

The study site, Metro Manila in the Pasig Marikina Basin, is moderately large (~ 300-400 square

kilometres). It consists of an interconnected network of rivers, creeks, and minor drains, which

are partly affected by hydraulic structures such as the Mangahan Floodway, pumping stations,

and river walls (levees). Additionally, it contains large areas of urbanised floodplain,

which are subject to inundation during floods.

Page 31: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 25 Greater Metro Manila Area – Flood Risk Analysis

Hence, the study team reasoned that a 1D/2D approach would be optimal for the study area, and that

if this was not possible, it would also be appropriate to adopt a 1D+ approach supplemented by

detailed 2D modelling at sites of particular interest. During the initial stages of the study, the team

could not identify any free or open source 1D/2D modelling packages (although we note that towards

the end of this study, a 1D/2D- model ‘Flo2D’ became free). However, several widely used and freely

available 1D+ models were identified, including SWMM and HEC-RAS. In addition, the open source

2D models ANUGA and Delft3D were known to individual members of the study team, and were

considered as options for detailed 2D work.

Ultimately the team chose to use the HEC-RAS 1D+ hydraulic model, because it is widely used for

flood inundation studies (including previous studies in Metro Manila); is not too computationally

demanding in an area the size of Metro Manila; can interact well with HEC-HMS for rainfall runoff

work; supports a wide variety of hydraulic structures; includes storage-areas for increased flexibility in

floodplain modelling; is freely available; and has a good graphical interface. While some preliminary

exploration of both ANUGA and Delft3D for detailed 2D modelling of small areas was undertaken,

there were technical challenges in applying both, and ultimately insufficient time to develop these

models in addition to the HEC-RAS model. Further modelling of this type could be considered in

future. However, evidence presented later in this report suggests that the 1D+ model provides

reasonable estimates of peak flood depths in the channels and floodplains, and is thus appropriate for

risk estimation in the present project.

3.3 Flood Inundation Model Development and Calibration

3.3.1 Rainfall Runoff Model

To support a semi-distributed modelling approach, catchment areas in the Pasig-Marikina basin were

divided into separate sub-catchments with different hydrologic properties. The sub-catchment

delineation was based on catchment boundaries computed from both SRTM data, and the LIDAR

DEM raster data (with the latter down-sampled to 10mx10m to make the computation tractable).

Catchment delineation calculations used the GRASS GIS module r.watersheds. The computed

catchment boundaries were combined with user judgement as to the appropriate locations for

boundary inflows into the hydraulic model, in order to produce the final sub-catchments.

For reasons explained later in the report, the final hydraulic model had to be split into two (a ‘Marikina’

model and a ‘Pasig-San Juan’ model). Hence, the rainfall-runoff model was also set up to support two

different hydraulic model configurations (Figure 3.4). Inflows from the sub-catchments were fed into

the hydraulic model in the form of either: a) discharge boundary conditions, b) uniform inflow sources

into channels, or c) uniform inflow sources into storage areas (Appendix A).

Page 32: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

26 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

Figure 3.4. The sub catchment structure in the rainfall-runoff models used to support the Marikina (left) and Pasig-San Juan (right) HEC-RAS models.

The hydrology of the sub-catchments was modelled using the SCS Curve Number loss model

combined with the SCS unit hydrograph. This is a widely used event-based method for modelling flood

scenarios (HEC-HMS TRM, 2000; Abon et al., 2010, WBCTI, 2012). One advantage of using this

approach is the existence of heuristic methods to estimate the model parameters. This is required in

the absence of calibration data.

The SCS Curve Number loss model is used to estimate how much incoming precipitation is

transformed into runoff within each sub catchment. The ‘excess rainfall’ (total precipitation less

storage) of a rainfall event is modelled as a function of the precipitation and potential storage capacity

of the catchment:

(1)

Here Pe is the time-integrated ‘excess rainfall’ since the start of the event (mm), P is the time-

integrated precipitation since the start of the event (mm), and S (mm) is the ‘potential maximum

retention’ of the watershed, which describes the capacity of the watershed to store water.

The parameter S is related to the watershed characteristic via the ‘Curve Number’ CN as:

(2)

CN theoretically varies from 0 – 100, with higher CN values occurring in catchments with limited

storage (e.g. heavily urbanised areas), and lower CN values occurring in more pervious areas.

Importantly, CN may be estimated by calibration, or alternatively as a function of land-use and soil

type if calibration data is unavailable (HEC-HMS TRM, 2000, Appendix A). In the present study data

on soil types and land use in the Pasig Marikina basin was used to assign a CN value to each sub-

catchment in the model (Appendix A).

Page 33: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 27 Greater Metro Manila Area – Flood Risk Analysis

To compute the outflow hydrograph for the catchment, the cumulative excess rainfall is routed through

the SCS unit hydrograph. Conceptually, this models the time-delay between the arrival of excess

rainfall in the catchment, and the catchment outflow. It depends on a single parameter t_lag as well as

the catchment area (HEC- HMS TRM, 2000). The t_lag parameter controls the lag-time or time to

peak of the watershed. Large values of t-lag are likely in larger catchments or those where the

transport of water through the catchment is delayed by e.g. low drainage slopes, hydraulically rough

surfaces, and tortuous channel paths.

In general, t_lag can be set by calibration against observation data, or using heuristic procedures

which relate it to the sub-catchment geometry and land-use. In the present study data is not available

to support calibration in most sub-catchments, and thus heuristic methods are used, as described in

the HEC-HMS manual. This is based on estimating the length and slope of channels and overland

flow paths in the catchment, and was done interactively in GIS using the LIDAR and SRTM DEMs,

and imagery.

As the rainfall runoff model parameters were set with a heuristic procedure, they were not directly

‘calibrated’ to match data. However, they were used to provide input to the hydraulic model, which was

calibrated to match flooding observed Tropical Storm Ondoy in September 2009. For these calibration

runs, the rainfall in each sub-catchment was taken from a nearby rain gauge, as described in

Appendix A. For design flood simulations, a design rainfall time series was applied, as described in the

section on design flood estimation.

3.3.2 Hydraulic Model

3.3.2.1 Model Theory

HEC-RAS can be used to simulate flow in a linked network of one-dimensional channels and storage

areas. Channel flows are modelled using a variant of the 1D St. Venant Equations (HEC-RAS TRM,

2010). The model can account for cross-channel variations in the flow velocity and bed roughness,

provided that the water surface elevation across the channel is constant. Several options are available

for modelling channel junctions, and the energy method is used herein. The energy method is also

applied in modelling flow momentum losses associated with bridges.

Within the 1D+ framework available in HEC-RAS, floodplain inundation may be simulated by either:

1. Extending river cross-sections to cover parts of the floodplain. A higher hydraulic roughness is

then typically assigned to the floodplain portion of the cross-section. This is a reasonable

approach when the water elevation over the floodplain is expected to be the same as the

water elevation in the channel.

2. Storage areas: These consist of a ‘pond’ type region with a water level-volume relation, in

which the water surface elevation is assumed to be constant. They can be connected to

channels or other storage areas as described below.

3. Using a network of channels over the floodplain to simulate flow paths.

Lateral weirs are used to connect storage areas to channels and/or other storage areas. These

simulate the exchange of flow based on the water levels in the two connected elements, according to

a broad-crested weir equation:

Page 34: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

28 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

(3)

where Q is the discharge (m³/s), C is a user-defined weir drag coefficient, L is the length of the weir,

and H is the upstream hydraulic head above the weir. For weirs with complex geometries, the weir is

broken into segments and Equation 3 is applied separately to each. When the weir is submerged,

HEC-RAS modifies Equation 3 to include backwater effects, which cause reductions in the flow over

the weir (HEC-RAS TRM, 2010). This ensures that no flow occurs between connected elements with

the same water elevation, which physically is reasonable. For a standard broad-crested weir,

recommended values of C are around 2.6- 3.1 (HEC-RAS TRM, 2010).

For linkages between storage areas on low-gradient floodplains, flows are generally more quiescent

(lower Froude number) than implied by the typical broad-crested weir model, and a lower weir drag

coefficient may be appropriate. In such cases H will be well approximated by the depth of flow over the

weir and L*H approximates the cross-sectional area of the weir-flow. In the absence of backwater

effects, the weir equation is then equivalent to setting the mean flow velocity to C*H(0.5)

, i.e. fixing a

constant flow Froude number. The form of this relation is identical to the Chezy friction model for

uniform flow, which also states that velocity scales with the square root of the flow depth (Chaudhry,

2008). The latter is a reasonable model for floodplain flows in the absence of backwater effects, and

can be used to provide a first-estimate of an appropriate weir drag coefficient, which may be refined by

calibration. With this method, the coefficient C is equal to the square root of the bed slope divided by

the Chezy bed roughness.

HEC-RAS solves the flow equations approximately, using an implicit finite difference numerical

method. In theory, the difference between the approximate and ‘true’ solution to the underlying flow

equations becomes vanishingly small as the model resolution (cross-sectional spacing) becomes finer,

and the model time-step is decreased. In practice, model accuracy is also affected by data errors, and

can be influenced by numerical instabilities (i.e. erroneous oscillations in the flow behaviour which are

an artefact of the numerical solution method). The latter can usually be controlled by 1) reducing the

model time-step or cross-sectional spacing, 2) adjusting some numerical method control parameters,

or 3) smoothing over rapid changes in the model geometry or boundary conditions, so long as the

changes still provide a good description of the flow situation.

The numerical method used by HEC-RAS is most widely applicable and stable for modelling sub-

critical flows (where the flow velocity is less than about 3.1 times the square root of the water depth).

HEC-RAS can also model super-critical flows (using the ‘mixed-flow’ option); however, it does so by

supressing the inertial terms in the flow equations (HEC-RAS TRM, 2010). This is reasonable for a

wide class of super-critical flows that vary sufficiently slowly in space and time, and is expected to be

reasonable in the present study when super-critical flows occur (which is relatively rare compared with

subcritical flows). Theoretically, it is less appropriate for extreme flow events such as dam-breaks,

where inertial effects are more significant. However, the latter situation does not occur in the present

study.

3.3.2.2 Construction of the model geometry

The main river systems are modelled as a linked network of one-dimensional channels and storage

areas (Figures 3.5 - 3.6). The channel cross-sections often extend well beyond the main river channel,

and are assigned a higher hydraulic roughness than the main channel regions. In addition, broad

channels are often placed in overland flow paths where there is no river, as these regions can have

defined flows during floods. Storage areas are also used to represent floodplain storage, and flows

Page 35: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 29 Greater Metro Manila Area – Flood Risk Analysis

between these are modelled using lateral weirs, with calibrated drag coefficients (see below). These

approaches to model schematization allow flood inundation to be simulated over both channels and

the floodplains (e.g. Dung et al., 2013).

Figure 3.5: Example of model schematization in the Upper Marikina River, showing channels (white cross sections) and storage areas (polygons). Note that some of the channels at the bottom centre of the image are not associated with real rivers, but represent possible overland flow paths.

The cross-sectional elevation was taken from the LIDAR data in all areas, except the water areas

where the LIDAR cannot penetrate. In these regions, cross-sectional surveys were used. Aside from

the MMDA surveys, the data are not recent, and errors may be introduced due to post-survey changes

in the cross-sectional geometry. However, the data was the best available to the project at the time of

model development.

Data entry proceeded by inputting an initial cross-sectional profile (making use of survey data in the

channel), and then replacing this with LIDAR outside of the water areas. It was then necessary to

manually check and correct the connection of the two datasets, as in some instances, unphysical

‘topographic steps’ would occur at the boundary of the two datasets. This could be due to changes in

the topography over time, or to errors in either dataset. In some places, survey data was not available,

and the cross-sectional data was estimated by interpolation from neighbouring cross-sections, using

HEC-RAS’s automated interpolation algorithm. Even if survey data is lacking, this technique can be

useful to improve the accuracy and stability of the model’s finite difference approximation of the

underlying flow equations. In overland flow channels which would otherwise go dry outside of the

main flood event, small ‘pilot’ channels were cut into the model geometry. This prevents complete

channel drying, which cannot be simulated with HEC-RAS (drying tends to cause strong instabilities

and a termination of the program).

Storage areas were initially defined with boundaries which corresponded to local high-points in the

topography. The latter were identified from catchment boundaries computed from the LIDAR data

using the GRASS GIS algorithm r.watershed. This ensured that topographic barriers between parts of

the floodplain were respected by the storage area routing. In some instances, storage areas were

Page 36: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

30 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

further subdivided to allow for a more gradual change in the water surface elevation over the

floodplains, based on the results of initial simulations. Connections were made between storage areas

and other storage areas or channels using lateral weirs. The elevations of storage areas and lateral

weirs were taken from the LIDAR DEM. A procedure was developed whereby the user could define

the location of the storage areas in GIS, save the result to a shapefile, and then use an R script to

compute the associated elevation-volume relation, and the elevations of the connections between the

storage area and other storage areas or the main channel. The data was then automatically inserted

into the HEC-RAS geometry file. In many cases the resulting lateral weir elevations were manually

corrected to reflect e.g. the elevation of flood parapet walls, which are too small to be resolved by the

LIDAR DEM data.

Bridges were added at known locations, with the geometry defined based on data from DPWH.

Pumping stations were added at known locations, based on site visits to Taguig, and information in

DICAMM (2005) and WBCTI (2012).

In low-lying areas to the north and south of the Pasig River, pilot channels were used to prevent

channels from drying, to prevent numerical ‘blow-up’ in the HEC-RAS model (as the HEC-RAS solver

cannot treat wetting and drying). Underground drains were not included, as there is evidence that

these have often blocked by garbage and silt (DICAMM, 2005), and their present status is unknown to

the study team.

3.3.2.3 Splitting the model into the Marikina and Pasig /San Juan regions.

HEC-RAS was found to crash when running models with a large number of storage areas (~ 1000),

due to memory overflow errors. In the present study, it was found that using large numbers of storage

areas gave a better representation of floodplain flows, as it allowed a more gradual change in the

water surface elevation. To circumvent the limitation on the number of storage areas, the model was

split into two separate regions: A ‘Marikina’ model which includes a simple representation of the Pasig

and San Juan river systems, and a detailed Pasig / San Juan model. The model was then ran by 1)

Running the Marikina model using the original boundary conditions, and accounting for inflows from

the San Juan River, then 2) Running the Pasig / San Juan model, using already computed flows from

the lower Marikina River as a boundary condition. Results from the Pasig / San Juan model were used

for all mapping in that region, while results from the Marikina model were used elsewhere. The

schematization of channels and storage areas in each model is shown in Figure 3.6.

Figure 3.6. Schematization of channel cross-sections (orange) and storage areas (blue) for the Marikina (left) and Pasig / San Juan (right) models.

Page 37: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 31 Greater Metro Manila Area – Flood Risk Analysis

3.3.2.4 Roughness and weir drag coefficients

The hydraulic roughness values were determined partially by calibration, while keeping them within

physically reasonable ranges. Unless otherwise stated, channels were given a roughness of 0.03, with

their floodplain regions (typically urbanised) given a value of 0.2. The exceptions are:

1. In the Upper Marikina River, between Wawa Dam and the junction of the tributary from Mount

Oro, the channel was given a value of 0.04, consistent with the enhanced bed-roughness here

(boulders on the channel bed).

2. Throughout the Upper Marikina River (upstream of the Rosario Weir, but most especially

upstream of Tumana), there are considerable areas of non-urbanised floodplain (e.g.

agricultural land). These were given a roughness of 0.05, except where upstream flow

blockages were observed which would prevent them from conveying much flow, in which case

a value of 0.2 was used.

3. The tributaries of the Upper Marikina River (except for the Nangka River) were given uniform

values of 0.05 for both the channels and the floodplains, reflecting the less-densely urbanised

nature of their floodplains, and the fact that their channels are not so cleanly defined and are

steep in parts. The Nangka river has more densely urbanised floodplains, so it was given a

channel value of 0.05 and a floodplain value of 0.2. The higher roughness in the tributaries

was also useful to enhance the model stability.

4. Tributaries to the San Juan River had their channel roughness set to 0.05, while retaining the

value of 0.2 for their urbanised floodplains.

For lateral weirs connecting channels with storage areas, the weir drag coefficient was set to 3.0. A

uniform value of 0.2 was used as a weir drag coefficient for flow between two storage areas on the

floodplains. This value is supported by analogy with the uniform flow Chezy equation for areas with a

bed slope ~ 0.0016, and Chezy coefficient of ~ 5, where the latter estimate is equivalent to a Manning

roughness of 0.2 for urban floodplains with a depth of 1m. These values were estimated to match a

key overland flow path in the HEC-RAS model around East Marikina and Cainta, and were found to

give reasonable agreement with observed floodplain depths during Tropical Storm Ondoy.

3.3.2.5 Boundary Conditions for the Tropical Storm Ondoy Calibration

For the Tropical Storm Ondoy simulation, inflows from the rainfall runoff model were imposed as

described in Appendix A. At the lower boundary of the Napindan River and the Mangahan Floodway, a

water level boundary condition was imposed based on water level observations from the Angono

gauge. At minor rivers flowing into Manila Bay, a downstream water level boundary condition was

imposed based on observations at the Manila Harbour South gauge. At the downstream end of the

Pasig River, a boundary condition was imposed using observed water elevations at Fort Santiago,

with the vertical datum estimated as described previously. In the Pasig / San Juan HEC-RAS model,

the water elevation and discharge at the downstream end of the lower Marikina was forced using the

stage and flow as computed from the Marikina model.

3.3.2.6 Model Setup and Calibration Approach

The model was set up iteratively, by gradually inputting more channels along with a crude storage

area representation. Often different parts of the model were worked on by different members of the

study team, and then integrated. As new channels and storage areas were added to the model, it was

re-run with idealised boundary conditions, to help detect any gross instabilities or errors introduced by

the new geometry (e.g. errors which could cause the model run to terminate). These could normally be

Page 38: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

32 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

fixed by making corrections or small adjustments to the new geometry, or the cross-sectional spacing.

This iterative strategy made it possible to gradually develop a complex model schematization, and

eliminate the main sources of model ‘blow-up’. Following the initial input of geometric and roughness

data, the model was run and iteratively adjusted to remove large instabilities or obvious errors in the

flow behaviour. These were typically related to input errors (or very rapid variations) in the models

geometry or roughness; insufficient cross-sectional spacing; or the use of overly large storage areas.

A short model time step (3s) was used to increase the model stability, while the non-linear solver was

allowed a maximum of 10 iterations to converge in each time-step. As super-critical flows occurred in

some reaches (typically in steep tributaries of the main river system), the model was run using the

‘mixed-flow’ option, with a Froude number threshold for the elimination of inertial terms set to 0.95.

The model was calibrated to predict peak depths during Tropical Storm Ondoy, both in the channel

and over the floodplains. Once the model was running with reasonable stability, the general strategy

for model improvement was to 1) note discrepancies between the model and observations; 2) consider

physically why these might be occurring, and look for any model input errors that might be

responsible. If this didn’t fix the problem, consider; 3) adjusting friction coefficients in broad areas of

the model to improve the agreement with data, while keeping them within physically reasonable

ranges. This led to the definition of Manning’s n and drag coefficients as described in the previous

section.

3.3.2.7 Taguig-Pateros-Lakeshore Region ‘bathtub’ models.

In the area south of the Napindan River and along the Laguna Lakeshore, flooding occurs largely due

to the ponding of standing water associated with flood events and high lake levels (CTI, 2005; Muto et

al., 2011). Regions in this area modelled for this study are shown in Figure 3.7.

Part of the area around Taguig and Pateros includes a large number of flood defence structures

(Figure 3.7). This includes pumping stations (total capacity of 27m³/s), floodgates on rivers which

connect to the Napindan River and Laguna Lake, a Parapet wall (elevation 3.6 m above MSL) along

the Napindan River, and a dyke along the Laguna Lakeshore (elevation 4.5 m above MSL) (CTI,

2005).

Page 39: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 33 Greater Metro Manila Area – Flood Risk Analysis

Figure 3.7. Regions around Laguna Lake modelled using the 'bathtub' approach. Red polygon: Areas inside flood defence structures. Yellow polygon: Areas flooded by high Laguna Lake levels.

If functioning perfectly, these defences will hydraulically isolate the area inside the flood defences from

the Lake and the Pasig-Marikina River system, at least while the water elevation in Napindan Channel

remains below 3.6 m above MSL. In that situation, flooding is caused by the rain-induced inflow being

larger than the outgoing flux due to pumping.

Page 40: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

34 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

At the other extreme, if the flood defences completely fail, or in areas that are not protected by them,

the water elevation will equilibrate with Laguna Lake. As the coverage of the flood defences is

incomplete, and their historical performance has been mixed, two sets of models are developed for

this study. In one model, the flood defences function perfectly, and hydraulic isolate the area inside the

red polygon in Figure 3.8 from water in Laguna Lake and Napindan channel. In the other, the flood

level is determined by the Laguna Lake water elevation.

3.3.2.8 Model 1: Flood defences function perfectly

The model here applies to the red-polygon region in Figure 3.8, and represents the situation in which

flooding is caused by rainfall onto the catchment, offset by the operation of pumping stations. Inflows

from Laguna Lake and the Napindan River are treated as negligible. Previous studies have

constructed similar models for this region (CTI 2005; Muto et al., 2011). The main advantage for the

present study is the availability of better quality elevation data (LIDAR DEM), which improves the

accuracy of the stage-volume relation (Figure 3.8), and hence the computed inundation.

Figure 3.8. Stage-Volume and Stage-Area relations for regions in Figure 3.7, computed from the LIDAR data.

In the model, the volume of water inside the flood defences is computed as:

where V is the volume of water inside the flood defences (m³), t is time (s), I is the rate of inflow

(caused by rain into the catchment, routed through a rainfall runoff model), and PS is the rate of water

Page 41: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 35 Greater Metro Manila Area – Flood Risk Analysis

removal by pumping stations. The stage was related to the volume via a stage-volume curve

computed from the LIDAR data (Figure 3.8).

The inflow hydrograph at the pumping stations was computed using the same methods as described

in the Rainfall Runoff Section. One exception is that in the t_lag computation, the channel velocity

scale was set to 0.9 m/s (following Muto et al, 2011) instead of begin estimated, because the slope-

based approach is not valid for channel slopes approaching zero as occur in low-land parts of Taguig.

For simulating the Tropical Storm Ondoy event, the input rainfall was taken as the observed Science

Garden hourly rainfall, scaled by a factor so that the total rainfall was equal to the catchment average

rainfall (i.e. 0.789 for Tropical Storm Ondoy). For the design flood scenarios, the same rainfall series

as employed in the HEC-RAS model was used. The pumping station operation was assumed to begin

when the stage exceeds 1.0 m above MSL, and end when the stage falls below 0.3m above MSL,

which is consistent with observations of pumping station behaviour in the lead up to Ondoy (Figure

3.9). For simplicity, it is assumed that the pumping stations begin pumping at full capacity (27m³/s).

The initial water elevation was set to 1m above MSL. Note that this model has no calibration

parameters to ‘tune’ its performance to the data.

A time series of stage observations both inside the flood defences and in Laguna Lake was obtained

from Labasan pumping station (Figure 3.9). Occasional spikes in the stage data are assumed to

reflect data-entry errors. Water levels inside the flood defences remained significantly below those in

Laguna Lake prior to and during Tropical Storm Ondoy, highlighting the impact of the flood defences.

Also of interest are the regular oscillations in the water level in the range 0-1m. These demonstrate

that even with the floodgates closed, water drains into the flood defences from some combination of

catchment baseflow and leakage from Laguna Lake / Napindan River.

Figure 3.9. Observed water levels in Laguna Lake and behind the flood defences during September -December 2009. Measurements provided by Labasan Pumping Station.

Page 42: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

36 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

3.3.2.9 Model 2: Flood Defences Fail

The previous model assumes that the flood defences (pumping stations, dykes and gates) work to

keep water levels in Taguig-Pateros hydraulically disconnected from water levels in Laguna Lake.

However, this is not necessarily realistic: for example, during the 2012 Habagat event, the pumping

stations ran out of fuel. Maps of the inundation extents provided by Taguig City Council suggest that

the peak water level inside the flood defences was similar to the peak Laguna Lake level, presumably

due to a combination of rainfall and leakage. More generally, for lake levels above 3.6m, water will

overtop the parapet walls along the Napindan River, eventually causing water elevations inside the

dyke to equilibrate with the Lake level. In general, the water levels inside the flood defences will not

rise above the lake level, because if this were to happen, the floodgates would be opened to allow the

water to flow out to Laguna Lake. Thus, the scenario where the flood defences fail and the water level

equilibrates with the lake level is considered a ‘worst-case’ scenario, which is nonetheless realistic. In

this case, annual exceedance probabilities can be estimated from the corresponding water level return

periods in Laguna Lake, described later in this report.

3.4 Design Flood Estimation

To provide suitable input to the flood inundation model in the present study, the design flood events

need to specify 1) Rainfall time series for each sub-catchment in the rainfall-runoff model, and 2)

Water level time series for Laguna Lake and Manila Bay. These should vary appropriately to reflect

the AEP of the design flood.

This section describes methods of statistical analysis to support design flood estimation. It includes

methods to define design-storm temporal patterns, investigate spatial variations in extreme rainfall,

estimate the magnitude of catchment-averaged extreme rainfalls and lake-levels, and an approach to

parameterising the relationship between the latter two variables.

As with any statistical methods, the approaches used here are most reliable for estimating events

which are within the range of the observed data (‘interpolation’), rather than for estimating events

outside of the range of observations (‘extrapolation’). This applies to both estimates of the extreme

values of quantities, and to their confidence intervals. For example, it is likely that the ‘true’ magnitude

of the 1/30 AEP 2-day catchment-averaged rainfall total is within the confidence limits that we

calculate. However, the estimate for the 1/200 AEP 2-day rainfall depth (and its confidence interval)

will inevitably be less reliable, because it requires extrapolating the statistical model well outside the

range of the observed data.

3.4.1 Synthetic Storm Time Pattern

In this section, a synthetic storm pattern is developed based on an existing Rainfall Intensity Duration

Frequency (RIDF) analysis at the Science Garden. The latter was developed by PAGASA, and is

available on their website. It describes the AEPs of rainfall depth, for storms with durations ranging

from 10 minutes to 24 hours.

The RIDF curves are shown in Figure 3.10, along with straight-line fits for each return period. These

are useful for design-storm construction, at least in the vicinity of the Science Garden, because for any

given storm duration and AEP, they allow the corresponding rainfall depth (or equivalently, average

Page 43: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 37 Greater Metro Manila Area – Flood Risk Analysis

rainfall intensity) to be estimated. They can also be used to develop design-storm temporal patterns,

as outlined below.

Figure 3.10. Rainfall Depth, Duration Frequency Curves for the Science Garden Rain Gauge, developed in a previous PAGASA study.

Design storms require a rainfall temporal pattern. There are many approaches to developing these

(e.g. Ellouze et al., 2009). A common approach is to artificially construct a pattern which, for a given

AEP (e.g. 1/50), contains a range of shorter duration sub-storms (e.g. duration 1, 2, 3, 6, 12 and 24

hours) which all independently have a rainfall depth equal to the design rainfall depth for their duration

(e.g. peak 1 hour intensity equal to the 1 hour RIDF curve at AEP=1/50 on Figure 3.10; peak 2 hour

intensity equal to the 2 hour RIDF curve at AEP=1/50 on Figure 3.10; and so on). This approach is

taken in the present study, using the analysis at the Science Garden.

A single dimensionless design storm temporal pattern is developed for all AEP’s less than or equal to

1/5. The use of a single dimensionless temporal pattern is justified by the observation that, for each

AEP below 1/5, the ratio of the rainfall depth to the corresponding 24 hour rainfall depth is nearly

constant (Figure 3.11). This implies e.g. that the 1 hour design rainfall depth is always approximately

30% of the 24 hour design rainfall depth, independent of the AEP. Similarly, the 2 hour design rainfall

depth is always approximately 45% of the 24 hour design rainfall depth, and so on.

Page 44: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

38 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

Figure 3.11. Ratio of rainfall depth to 24hr rainfall depth against duration, for each AEP <=1/5

Figure 3.12 shows the dimensionless design storm pattern constructed using the mean dimensionless

rainfall depth curve in Figure 3.11. It uses hourly time increments, and thus only applies for durations

of 1 hour and above. The dimensionless design storm pattern is used to construct design storms for

the catchment as a whole, by rescaling the total rainfall to agree with the 1-day rainfall depth with the

desired AEP. By construction, the design storm then contains sub-storms with the same AEP for

durations from 1, 2, 3,6,12 and 24 hours. Later, this design storm pattern is also extended to 2-day

rainfalls, by appending 24 hours of constant rain, with intensity chosen so that the 2-day rainfall depth

AEP (computed independently as described later) is also satisfied by the design storm. Methods to

estimate the 1-day and 2-day extreme rainfalls throughout the catchment are described subsequently.

Page 45: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 39 Greater Metro Manila Area – Flood Risk Analysis

Figure 3.12. Dimensionless design storm pattern developed from the Science Garden RIDF curves

3.4.2 Spatial Variation in Extreme Rainfalls in the Pasig-Marikina Catchment.

An analysis of spatial variations in extreme rainfall intensities over the Pasig-Marikina catchment was

undertaken. The aim was to determine whether variations were so significant that they needed to be

accounted for in the design storms, or alternatively, whether a simpler ‘spatially constant design storm

intensity’ could be used over the entire catchment.

The analysis was restricted to maximum 1-day and 2-day rainfalls at rain-gauges in the Pasig-Marikina

Catchment. The temporal restriction was necessary because data recorded at shorter intervals was

not as widely available in space and time. Previous studies suggest that the 2-day rainfall total is an

appropriate guide for expressing the magnitude of a rainfall event in the Pasig Marikina basin, as

flooding rains most often occur over less than 2 days duration, and the typical time-delay between

peak rainfall and peak flooding throughout the basin is much less than 2 days (WBCTI 2012). The

spatial restriction was chosen because of our focus on the Pasig Marikina Basin, and because the use

of a larger region can reduce the likelihood that the assumptions underlying the statistical analysis will

hold.

To investigate spatial variations in extreme rainfall, annual exceedance probabilities of the 1-day and

2-day peak rainfall depth were estimated at each station with sufficient data, using an index-flood

procedure (Hosking and Wallis 1997). Similar techniques are widely applied to estimate extreme

rainfalls and river discharges as input to flood studies (e.g. Fowler and Killsby 2003; Kjeldsen et al.,

2008; Perica et al., 2011).

In the index-flood procedure, the data are assumed to be drawn from a ‘homogeneous region’, in

which the frequency distributions at each station are identical apart from a scale factor, which varies

from site to site according to the model:

(4)

Page 46: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

40 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

Here is the intensity exceeded with frequency F at station ‘i’; q(F) is a probability distribution (the

‘regional growth curve’) which is identical within the homogeneous region; and is the scale factor

(termed the ‘index flood’) for Station i.

To see the implications of the index-flood assumption, consider a homogenous region containing

Station A and Station B. If the 10% AEP 1-day rainfall depth is 100mm at Station A and 150mm at

Station B, then the index-flood assumption implies that for any AEP, the 1-day rainfall depth at Station

B will be 1.5 (=150/100) times that at Station A. If this is not true, then Station A and Station B are not

from the same homogenous region, and alternative methods of analysis must be applied.

Hosking and Wallis (1997) describe the computational methods used in this analysis. This includes

methods for estimating the values of q(F) and , methods for checking the validity of the

‘homogeneous region’ assumption, and methods for computing uncertainties in the associated

extreme rainfall estimates. These computations are implemented in the open source software ‘R’

within the package ‘lmomRFA’, which was used for the present analysis.

The data was first converted into annual maximum rainfall series at each station. The index-flood

method used here requires that the data are statistically independent within each station, and this is

easiest to achieve by only selecting the maximum rainfall event for each calendar year (as the

calendar year contains one entire wet season, which does not usually overlap with the next year). For

each station, the maximum 1-day and 2-day rainfall was computed for every year in the record, so

long as that year contained > 90 days of wet season rainfall. Here the ‘wet season’ is defined as being

from May-October inclusive (184 days in total). Years for which the record covered less than 90 days

of wet-season rainfall (e.g. because of instrument malfunction) were excluded from the analysis,

because of the high chance that they did not record the peak rainfall event in that year. Finally, only

stations with > 10 years remaining were included in the analyses. The final datasets contained seven

stations with a total of 222 acceptable data-years (Table 1.1).

Table 1.1. Stations used in the Regional Frequency Analysis, and number of years of acceptably complete data.

STATION Port Area

NAIA Tipas Pasig Mt Oro Science Garden

Sitio Tabak

Number of accepted data years

47 36 20 34 16 49 20

Next, the L-moments for each station were computed, and the homogeneity of the region was checked

by computing the ‘discordancy statistic’ D_station for each station, and the ‘homogeneity statistic’ H for

the region (Hosking and Wallis, 1997). Hosking and Wallis (1997) suggest that stations may be

considered sufficiently consistent with the region as a whole if D_station<Dcrit, while the region may

be considered as acceptably homogeneous if H<1, possibly heterogeneous if 1<H<2, and definitely

heterogeneous if H> 2. More recently, Wallis et al. (2007) suggested that for rainfall data, H<2 should

be considered acceptably homogeneous, noting that the H<1 guideline does not account for non-

statistical sources of variability which often exist in rainfall data.

Even in homogeneous regions, it is possible for the discordancy and homogeneity statistics to suggest

non-homogeneity due to the presence of isolated extreme observations. This could be detected by re-

computing the homogeneity and discordancy statistics with the isolated extreme observation removed.

If the latter appear to be acceptably homogenous, it is recommended that the homogeneity

Page 47: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 41 Greater Metro Manila Area – Flood Risk Analysis

assumption is accepted, and that the analysis proceeds including the extreme observation, which will

contain useful information on the behaviour of extreme events in the region (Hosking and Wallis 1997,

p 70; Fowler and Killsby 2003).

The best-fitting probability distribution was chosen from a selection of 5 different distributions, by using

an L-moment based goodness-of-fit test, and viewing the data-set on an L-moment ratio diagram.

Return period curves were computed for every station, with confidence limits computed using a

parametric boot-strap from a synthetic region with the same number of stations as included in the

analysis (see function ‘regsimq’ in the R package ‘lmomRFA’ version 2.4). The quantile function at

each site in the synthetic region was of the same type as the fitted regional distribution, but with

parameter values taken from the parameter estimates for each individual station. The bootstrap

calculations account for a constant correlation between the extreme values in the same year at each

pair of stations, which is estimated as the average correlation among stations in the data. Results

were checked to assess the sensitivity to this value. The fitted return period curves and observations

were plotted graphically to provide a further visual check on the quality of the fit.

3.4.3 Catchment-Averaged Extreme Rainfall Frequency Analysis

The AEPs of extreme rainfall at individual stations will generally differ from the AEPs for the catchment

as a whole (Allen and DeGaetano, 2005). This is because rainfall is spatially variable. For example, in

a large catchment, a heavy rainstorm may have a small ‘patch’ of particularly intense rain which

exceeds the 1-day 1% AEP intensity, while in other areas, the storm has a 1-day 5% AEP intensity.

While individual rainfall gauges do measure the effects of an intense rainfall patch, the catchment

averaged rainfall will ‘average-out’ these extreme patches and thus not be as significantly affected,

especially for larger catchments. Hence, spatially-averaged extreme rainfall depths tend to be less

than at-a-station extreme rainfall depths.

For design flood scenarios, the ratio of the catchment averaged rainfall depth to the point rainfall depth

is termed the ‘Areal Reduction Factor’. It depends on the chosen catchment, the chosen storm

duration, and the storm AEP. This can be used to construct extreme rainfall intensities for the

catchment as a whole, by multiplying the at-a-point extreme rainfall intensities with the areal reduction

factor. Such catchment-averaged extreme rainfall intensities are generally considered to be more

appropriate than at-a-point extreme rainfall intensities for design flood estimation, especially for larger

catchments.

To estimate the return periods of 1-day and 2-day catchment averaged rainfall, a grid of 1kmx1km was

generated within the Pasig-Marikina Catchment, including Taguig and sites North / South of the Pasig

River, which are treated in this study (Figure 3.13). The 1-day and 2-day rainfall data was interpolated

from the rain-gauges to each cell using inverse-distance-weighted interpolation, so long as at least 3

stations had non-missing data for that day. Otherwise the data was treated as missing. The same rain-

gauges as included in the regional frequency analysis were used. Irrespective of their distance to a

cell, stations were given a zero weight on days they were missing data (so they do not affect the

calculation). For each day, the catchment-averaged rainfall was then computed as the average of the

interpolated values. Finally, the annual maximum 1-day and 2-day rainfall was computed for the entire

catchment.

Page 48: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

42 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

Figure 3.13. Pasig-Marikina Catchment including all areas treated in HEC-RAS model (polygon). Red + are rain gauges with > 10 years of extremes recorded. Blue dots are the 1kmx1km grid at which rainfall was interpolated to compute the catchment averaged rainfall.

The catchment-averaged data was then analysed using the same methods as applied for the regional

frequency analysis (as these also work for a single station). For consistency with that analysis, a

Generalised Normal distribution was fit to the 1-day data while a Pearson Type 3 distribution was fit to

the 2-day data. In both cases these distributions were ‘good-fits’ to the data based on the Z-statistic

(Hosking and Wallis, 1997).

The analysis was also repeated with the catchments broken into separate regions for the San Juan

Catchment, the Marikina Catchment, the Taguig-Pateros region catchment, and an area to the North

and South of the Pasig River. This turned out to have little effect on the results, and so for simplicity

the whole-of-catchment analysis was used for the design rainfall construction.

3.4.4 Frequency analysis of high water levels in Laguna Lake

The annual exceedance probabilities (AEP) for Laguna Lake water levels were computed using daily

lake level data, for which incomplete records are available since 1919. The dataset includes water

level observations at Angono, Kalayaan, Los Banos, and Looc. These were checked visually to ensure

consistency among stations, and exclude any obvious errors.

A single ‘annual maximum lake level’ data series was constructed from the maximum of all the station

observations in each year. Because of gaps in the recorded data, coverage was restricted to the

years 1919-1922, 1946-1956, 1958-1963, 1965-1967, 1972-1978, 1985-1986, 1988-2007 and 2009.

Page 49: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 43 Greater Metro Manila Area – Flood Risk Analysis

To estimate AEPs of peak lake level, an extreme value analysis was conducted. Probability

distributions were fit to the annual maximum lake level dataset using the method of L-moments

(Hosking and Wallis, 1997). The analysis was conducted using the software R, with distribution fitting

done by the package ‘lmomco’. Where required, the empirical AEP’s of data points were estimated

using Cunane’s plotting position formula. Ninety-percent confidence intervals for the fitted quantiles

were computed via a parametric bootstrap (Kysely, 2008), using one hundred thousand bootstrap

samples. For the final fits, the convergence of the confidence interval estimates was confirmed by re-

running the fit with different numbers of bootstrap samples (50000, 25000, 10000), and checking that

changes in the confidence intervals were negligible.

Probability distributions tested on the dataset include the two parameter ‘Gumbel’ distributions, and

the three parameter ‘Generalised Extreme Value’, ‘Log Pearson Type 3’, and ‘Pearson Type 3’

distributions. In each case, the goodness-of-fit of the statistical model to the data was checked by

examining the linearity of quantile-quantile and probability plots, and also by plotting the data L-

moments on a L-moment ratio diagram (Hosking and Wallis, 1997), which can be used to select

among candidate distributions. On this basis, the Generalised Extreme Value distribution was selected

for the final analysis, although all of the tested distributions were found to give quite similar predictions

for the quantiles of interest.

3.4.5 Relation to extreme rainfalls

For the design flood simulations, an artificial water level time series is required at Laguna Lake. This

has to be supplied as a boundary condition to our hydraulic model, because our rainfall runoff model

does not simulate 80% of Laguna Lake’s catchment (which is outside of the Pasig-Marikina Basin).

The design water level time series is constructed to reflect the relations between the lake water

elevation and the design rainfall. For example, visual investigation of Figure 3.14 suggests that often

high lake levels may be preceded by high rainfall in the Pasig-Marikina Catchment. However, this is

not always true, as most of the Laguna Lake catchment lies outside the Pasig-Marikina catchment.

Figure 3.14. Catchment-averaged 2-day rainfall (black) with Laguna lake water levels (red). Green lines denote days with the annual maximum rainfall.

Page 50: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

44 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

To simplify the problem of constructing the lake level time series for the 2-day design flood events, it is

assumed that the lake level increases with the same time pattern as observed during Tropical Storm

Ondoy at Angono station from midday 25/09/2009 to midday 27/09/2009. However, the latter series is

translated and scaled to set the initial and final lake levels for the simulation. The initial and final lake

levels are related to the observed annual maximum catchment-averaged 2-day extreme rainfall as

described below. This enables the computation of the lake levels time series once the catchment

averaged rainfall for the design event is known.

Figure 3.15 compares annual 2-day maximum catchment-averaged rainfall against the initial Laguna

Lake levels (1-day prior to that rainfall event). There is no strong relation between these variables,

which might be expected as the rainfall event does not directly affect lake levels the day before.

However, the associated change in lake level from 1-day before to 1-day after the storm is positively

related to the annual maximum catchment-averaged 2-day rainfall (Figure 3.16). Physically, this is to

be expected, both because high rainfall events in the Pasig-Marikina Catchment contribute inflow to

Laguna Lake, and also because these rainfall events are likely to be correlated with high rainfall

elsewhere in the Laguna Lake catchment.

Figure 3.15. Annual maximum 2-day catchment averaged rainfall vs. Laguna Lake level the day before the storm.

Page 51: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 45 Greater Metro Manila Area – Flood Risk Analysis

Figure 3.16: Annual maximum 2-day catchment averaged rainfall vs. the change in Laguna lake level from 1-day before the storm to 1-day after.

To construct the design flood lake level time series, the starting lake level is taken from the upper 90%

prediction interval in Figure 3.15, while the rise in the water level is taken as the upper 90% prediction

interval in Figure 3.16. This approach provides a conservative but realistic scenario, where both the

initial lake level and the rise in the lake level are higher than average, although not unrealistically so.

For example, during Tropical Storm Ondoy, the Lake level rose from ~ 2.3 to ~ 3.3 m. For the same

catchment-averaged 2-day rain (432 mm), the present method predicts a change in water level from

2.3 to 3.47, which is only slightly conservative as compared with observations during Tropical Storm

Ondoy (Figure 3.17).

Page 52: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

46 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

Figure 3.17. Comparison of the changes in Laguna Lake during Tropical Storm Ondoy, and the design storm boundary condition with the same 2-day catchment averaged rainfall.

It is worth noting in that for the largest few observed rainfall events, both the initial lake level and the

change in lake level fall above the regression line, although within the prediction intervals (Figures

3.15 and 3.16). This may be purely by chance, or it may indicate that the relation between rainfall and

the lake levels becomes stronger for more extreme rainfall events. The latter seems physically

reasonable, as extreme events may be spatially larger on average, and thus be associated with more

rain throughout the Laguna Lake catchment. If correct, the method used herein will be less

conservative for large events than for small ones. Regardless, within the range of the data the current

method is reasonable and conservative.

3.4.6 Design Flood Boundary Conditions

3.4.6.1 Manila Bay Water Levels

For all design flood scenarios, Manila Bay Water Levels are set to 0.9m MSL, which is approximately

equal to the highest astronomical tide in Manila Bay, and previously used design high water level for

Manila Bay (DICAMM, 2005; WBCTI, 2012). This water level is assumed to be independent of the

AEP. Future work may consider the relations between high water levels in Manila Bay and extreme

rainfalls in the Pasig Marikina Catchment, which could allow for e.g. accounting for Storm Surge. This

would require long time series data from Manila Bay, which is not available for the present study.

3.4.6.2 Laguna Lake Water Levels

The Laguna Lake water level time series is imposed for each design flood based on the rainfall AEP,

as described in the section on the relation between high lake-levels and extreme rainfall.

Page 53: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 47 Greater Metro Manila Area – Flood Risk Analysis

3.4.6.3 Design Flood Rainfall

For the design flood events, the catchment averaged rainfall was imposed on each sub catchment in

the rainfall-runoff model, with the rain time variation based on the design storm pattern described

previously. The rainfall intensity was scaled so that the 24 hour rainfall total was equal to the 1-day

rainfall total from the catchment-averaged frequency analysis, and a constant rate of rainfall was

appended for another 24hrs, with intensity chosen so that the 48 hour rainfall was equal to the 2-day

catchment-averaged rainfall for that return period.

The use of a single catchment-averaged rainfall throughout the Pasig Marikina Basin is justified by the

results of the Regional Frequency Analysis (see Results). It was found that the confidence intervals of

extreme rainfalls throughout the basin were generally overlapping, and that differences between

nearby stations were often as large as differences between far-apart stations. Thus it is reasonable to

neglect these differences as a first approximation, considering that they may be dominated by random

variation.

3.5 Damage Calculation

The damage calculations involve integrating outputs from the flood inundation model with the

exposure information and vulnerability models. The vulnerability models take the form of stage-

damage curves (Figure 3.18). These describe the damage fraction (i.e. ratio of damage to building

replacement value) as a function of the ‘peak flood depth minus the floor height’, for a range of

different building types.

Figure 3.18. Depth-damage curves developed for different building types by UPD-ICE for the present study. See “Development of vulnerability curves of key building types in the Greater Metro Manila Area, Philippines” (Pacheco et al., 2013) for further information.

Page 54: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

48 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

The full distribution of peak flood depths in each exposure polygon is used in the analysis. The peak

flood depth information is stored on a raster with a 10m pixel size, which typically leads to a range of

flood depths occurring in each exposure polygon (Figure 3.19). These peak depths are binned into

10cm classes (0m, 0.1m, 0.2m, …), with all non-inundated areas given zero depth. The exposure

polygons are then rasterised to 10m pixel size, with pixel values taking the ID value for the associated

polygon. Finally, a cell count of each depth-class in each exposure polygon is computed (via cross-

tabulation), resulting in a histogram of depth values for each exposure polygon.

Figure 3.19. Exposure polygons with the peak flood depths overlayed. Clearly most exposure polygons have a range of depths.

Damages are computed in several ways, all of which account for the multiple building-types that occur

in each exposure polygon. Initially in each exposure polygon, for each building-type / peak-depth

combination, the ‘damaged floor area equivalent’ in the floodable storeys is computed. For example, if

100ha of a given building type experiences a depth of 2m which causes 20% damage, then the

associated ‘damaged floor area equivalent’ would be 20 ha ( = 20% x 100 ha). In practice the depths

in each exposure polygon are variable, and so the total damaged floor area for each building-type is

computed as a sum of the damages in each depth-class. The ‘inundated floor area’ for each building

type is also computed. The details associated with computing these quantities are described in the

following sections.

Given these inputs, several useful measures of flood impact may be calculated for each exposure

polygon:

1) Damaged floor area equivalent: This is taken as the sum of the damaged floor area

equivalents for each building type.

2) Building damage cost: This is the sum of the damaged floor area equivalent per building

type multiplied by the associated building replacement cost per m².

3) Population with Inundated Homes: This is taken as the sum of the inundated floor area

multiplied by the population density per m² of floor area.

Page 55: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 49 Greater Metro Manila Area – Flood Risk Analysis

When mapping the above quantities, it is useful to transform them via division by the exposure

polygon ground area. Visually, this removes the confounding effect of exposure polygon size on the

results. For illustration, consider two exposure polygons with the same flood depths, building types,

and building densities. Suppose one is large (100 ha) and one is small (1ha). Then the large polygon

will have damage measures being 100 times the small polygon. However, if we divide by polygon

area, then they will both show up as having the same intensity of damage. Typically, the latter gives a

visually clearer indication of the spatial distribution of impact.

3.5.1 Computation of ‘damaged floor area equivalent’ in a single exposure polygon, for a single building-type and a single depth

The computations are described in a step-by-step fashion.

1. For each building-type in the exposure polygon, estimate the footprint-area (i.e. ground floor

area) covered by each building storey-category (1-storey, 2-storey, 3-7 storeys, 8-15 storeys,

16 – 25 storeys, 26 – 35 storeys, 36+ storeys):

Footprint-area vs. storey-category and floor-area vs. storey-category information is available in the

exposure database. The latter is also provided per building type. In each storey category, the

distribution of building-type by footprint-area is assumed to be the same as the distribution of

building-type by floor-area.

For each building-type / storey-category combination, compute:

Replacement cost per m^2, using building costs estimates reported in Appendix A.

‘Total floor area in storeys 1, 2 and 3’: Assume that only the lower 3-storeys of any building are

vulnerable to flooding, as it is very rare for flooding to exceed this level.

i) For 1-storey and 2-storey buildings, this is equal to the total floor area.

ii) For storey-categories ‘midrise’ or larger, this is calculated as 3x (footprint-area for

the associated category).

Nominal Floor height:

i) Localised building survey data suggest most buildings have a floor height

categorised as either 0 or 0 – 0.25 m (Appendix C). Thus we assume a uniform

floor height of 0.125m for every building type. This is subtracted from the flood

depth when computing the depth for the vulnerability curves.

Floor area associated with the given depth class:

i) Assume that all buildings are uniformly distributed over the polygon. Then the floor

area associated with this depth class = (Total floor area in storeys 1, 2 and 3) x

(Proportion of the exposure polygon in this depth class).

ii) Note that often, not all of this floor area will actually be inundated. When computing

damages, this is accounted for in the vulnerability curves. When computing

inundated floor area, this is accounted for using the inter-storey height.

Select a vulnerability curve for this ‘building-category’ / ‘storey-category’ combination, and compute

the associated damage fraction given the depth and floor height.

iii) Vulnerability Curves for different building-types / storey categories are noted in 3.2.

The methods used to develop these are described in “Development of vulnerability

curves of key building types in the Greater Metro Manila Area, Philippines”

Page 56: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

50 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

(Pacheco et al., 2013). The mid-rise curves are taken as applying to the floodable

fraction of buildings with 3 or more storeys.

Compute the damaged floor area equivalent for this depth as the sum of the (Flooded floor area for

this depth) x (damage fraction for this depth) for each storey category.

Table 3.2: Vulnerability Curves used for each building type / storey height combination. Missing values did not

feature in the exposure database. See Figure 3.18 for a plot of the vulnerability curves.

Building Type Vulnerability Curve

Type

Vulnerability Class

(1-Storey)

Vulnerability Class

(2-Storey)

Vulnerability Class

(3+-Storey)

W1 H W1-L-1 W1-L-2

W2 H W1-L-1 W1-L-2

W3 H W3-L W3-L

N H N-L-1 N-L-2

CHB C CHB-L-1 CHB-L-2

URA C MWS-L MWS-L

URM C MWS-L MWS-L

RM1 C MWS-L MWS-L

RM2 C MWS-L MWS-L

MWS C MWS-L

CWS C CWS-L

C1 C C1-L-1 C1-L-2 C1-M

C2 C C1-M

C4 C C1-M

PC1 C C1-L-1 C1-L-2

PC2 C C1-L-1 C1-L-2 C1-M

S1 C S1-L-1 S1-L-2 S1-M

S2 C S1-L-1 S1-L-2 S1-M

S3 C S1-L-1 S1-L-2

S4 C S1-M

3.5.2 Computation of the inundated floor area in each exposure polygon.

1. Assume that all buildings are uniformly distributed over the polygon, and thus the distribution

of depth experienced by the buildings is the same as that of the polygon as a whole.

2. Get the floor height and inter-storey height for this polygon from the exposure database. Floor

heights are assumed to be 0.125m, as discussed above. Inter-storey heights give the typical

height of each building storey.

Page 57: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 51 Greater Metro Manila Area – Flood Risk Analysis

3. Compute the floor area in the 1st storey (ground floor) as the sum of the building footprints.

The flooded fraction is equal to the fraction of depths in the polygon which are above the floor

height.

Compute the floor area in the 2nd

storey as the sum of the building footprints in buildings that are 2-

storeys and above

The flooded fraction is equal to the fraction of depths in the polygon which are above the floor

height plus the inter-storey height.

Compute the floor area in the 3rd storey as the sum of the building footprints in buildings that are 3-

storeys and above

The flooded fraction is equal to the fraction of depths in the polygon which are above the floor

height plus the twice the inter-storey height.

The inundated floor area is taken as the sum of all the (floor-areas x flooded-fractions) on storeys

1, 2 and 3.

Page 58: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

52 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

4 Methods

4.1 Hydrology

4.1.1 Regional Extreme Rainfall Frequency Analysis

The station annual maximum data were initially checked for homogeneity and discordancy using the H

and Dcrit statistics. This gave H=1.21 and 1.14 for the 1-day and 2-day rainfalls respectively,

suggesting the region is ‘possibly heterogeneous’ (or ‘acceptably homogenous’ according to the

weaker criterion of Wallis et al. 2007). All stations had D<Dcrit except for the NAIA station in the 2-

year rainfall analysis. Further investigation of the NAIA station was thus undertaken. This revealed the

presence of a single extreme observation in 1972 (472 and 291 mm of rain over consecutive,

total=763mm). Such high values are reasonable in Manila; for example, during Tropical Storm Ondoy,

the Science Garden station recorded 455mm of rain in 24 hours.

As noted in the Methods, it is possible for isolated extreme observations to give the appearance of

non-homogeneity in a region, and in this instance the analysis should proceed while including the

extreme observation. To check the significance of the NAIA outlier, the analysis was re-ran with 1972

data excluded from the NAIA station. This led to all stations having D_station<Dcrit, and H = 0.94/0.63

for the 1-day and 2-day rainfalls respectively, indicating that the region is acceptably homogeneous.

Thus, the analysis proceeded including the extreme observation at the NAIA station.

For the 1-day rainfall a Generalised Normal distribution was fit to the data, while a Pearson type-3

distribution was used for the 2-day rainfalls. These had the best fit to the data based on the Z-statistic,

although the results were not strongly sensitive to the use of other good-fitting distributions.

Confidence intervals were calculated using an inter-station correlation of 0.65, as estimated from the

data itself.

Table 4.1 and Figure 4.1 show an example of the AEP curve for a single station (Science Garden).

Figures 4.2 and 4.3 show the estimated 2-day 1/10 and 1/100 AEP peak rainfall events for stations

included in the analysis, along with their 90% confidence intervals.

For low AEP extreme rainfall events the 90% confidence intervals vary around +-30-40% of the point

estimate, due to the relatively short period of the data used in the analysis. This range is equivalent to

large changes in the estimated return periods. For example, at the Science Garden the upper

confidence limit of the 2-day 1/10 AEP rainfall is approximately equal to the lower confidence limit of

the 2-day 1/100 AEP rainfall (Table 4.1). There is also considerable overlap in extreme rainfall

estimates among stations for a given AEP, and it is difficult to distinguish any patterns (Figures 4.2

and 4.3). While extreme rainfalls are higher at stations north of the Pasig River, the differences are not

large compared with the variations between neighbouring stations, and could be an artefact of the

short nature of the records. Pairwise t-tests on the station data did not suggest any differences in the

mean annual maximum rainfall at each station, which is the ‘index flood’ value used to scale the AEP

curves. Similarly, a Kruskal-Wallis test did not suggest significant differences in the median annual

peak rainfalls between sites. In sum, differences in the observed extreme rainfall patterns do not seem

Page 59: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 53 Greater Metro Manila Area – Flood Risk Analysis

inconsistent with statistical randomness, and so for simplicity the spatial distribution of extreme rainfall

in the design storms may be approximated as constant.

Table 4.1. Extreme values of 2-day rainfall at the Science Garden (with 90% confidence intervals) based on the

Regional Frequency Analysis.

AEP 1/5 1/10 1/25 1/50 1/100 1/200

Rainfall Depth (mm) 319 387 475 534 594 654

Lower 90% CI 283 334 390 425 459 490

Upper 90% CI 360 454 592 688 791 898

Figure 4.1. AEP curve (and 90% confidence intervals) for annual maximum 2-day rainfall at the Science Garden, based on the Regional Frequency Analysis

Page 60: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

54 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

Figure 4.2. AEP 1/10 2-day rainfall depths (and 90% confidence limits) at stations used in the rainfall frequency analysis.

Figure 4.3. AEP 1/100 2-day rainfall depths (and 90% confidence limits) at stations used in the rainfall frequency analysis.

Page 61: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 55 Greater Metro Manila Area – Flood Risk Analysis

4.1.2 Catchment Averaged Extreme Rainfall Frequency Analysis

The AEP curve for 2-day catchment-averaged rainfall is shown in Table 4.2 and Figure 4.4. For a

given AEP, the catchment-averaged extreme rainfall values are generally smaller than the at-a-station

values, as expected on physical grounds. For example, they are about 82-93% of extreme rainfall

values estimated at the Science Garden. The magnitude of this ratio is consistent with results reported

elsewhere for catchments of similar size (Allen and DeGaetano, 2005).

Table 4.2. Extreme Values of Catchment Averaged Rainfall, and 90% confidence limits

AEP 1/5 1/10 1/25 1/50 1/100 1/200

2-day total

Catchment Averaged Rainfall (mm)

284 339 408 454 500 545

Lower 90% CI 258 302 352 383 411 444

Upper 90% CI 316 384 476 541 606 679

Figure 4.4. AEP curve and 90% confidence limits for 2-day catchment averaged rainfall totals

4.1.3 Design Storm Temporal Pattern

Figure 4.5 gives an example of the design storm temporal pattern used in this study, for the case of a

1/100 AEP event. It consists of the 24 hour temporal pattern described previously, scaled to match the

1-day 1/100 AEP rainfall total, followed by 24 hours of constant rain, with the rate determined so that

Page 62: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

56 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

the total storm depth matches the 2-day 1/100 AEP rainfall total. The most intense rainfall occurs

around 6-18 hours into the storm event. There is a slight increase in the rainfall rate in the final 24

hours (as compared with the rate at hour 24), which can be attributed to differences in the statistical

methods used in the previous RIDF analysis and in the present study. However, the peak flood depths

are largely determined by the most intense rainfall which occurs in the first 24 hours of the design

storm.

Figure 4.5. Design storm temporal pattern for an AEP 1/100 event

4.1.4 Laguna Lake Water Level AEP Curve

The Generalised Extreme Value distribution was selected for the final analysis, although all of the

tested distributions were found to give quite similar predictions for the quantiles of interest. The AEP

curve for extreme water levels in Laguna Lake is shown in Table 4.3 and Figure 4.6.

Page 63: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 57 Greater Metro Manila Area – Flood Risk Analysis

Table 4.3. Selected extreme values and 90% uncertainty limits for water levels in Laguna Lake (above Mean Sea

Level)

AEP 1/5 1/10 1/25 1/50 1/100 1/200

Lake Level 2.45 2.83 3.35 3.75 4.18 4.62

Lower 90% Confidence Limit

2.26 2.55 2.89 3.12 3.32 3.50

Upper 90% Confidence Limit

2.66 3.13 3.86 4.54 5.33 6.27

Figure 4.6. AEP curve for water levels in Laguna Lake

4.2 Hydraulics

4.2.1 Model Calibration

During calibration the HEC-RAS models were compared with peak depths during Tropical Storm

Ondoy, both in the channel where comparisons were made against EFCOS gauge data, and over the

floodplains.

The modelled and gauged water level peaks are reported in Table 4.4, with time series shown in

Figure 4.7. Results at Fort Santiago and Angono are not presented, as they are essentially forced to

agree with the data by the model boundary conditions. Several gauges broke during Tropical Storm

Ondoy, and in that instance the model is compared with the peak gauge value that was recorded prior

to breakage. In most instances, the model peak agrees reasonably well with the observations, with

Page 64: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

58 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

differences of 10-30cm. The modelled stages tend to rise and fall earlier than in the data. This

probably reflects the fact that the rainfall runoff model was not calibrated, does not include baseflow,

and is forced with rainfall data based on a small number of gauging stations. At Montalban the flood

peak was not recorded by the gauges, but independent evidence from a YouTube Video suggests the

modelled peak is reasonable (Table 4.4). The only large difference occurs at Nangka; however, this is

not of great concern because the gauge broke well before the flood peak, as explained in the notes of

Table 4.4.

Table 4.4. Comparison of measured and modelled peak water elevations (metres above mean sea level) in the

Pasig-Marikina Rivers during Tropical Storm Ondoy. * The gauge broke or otherwise failed to record the maxima.

In that case the model result at the time of the peak is reported, and the modelled peak is reported in the

parentheses. At bridges, the peak is reported as the mean of the upstream and downstream water level peaks.

SITE Data Model Notes

Pandacan 2.6 (3.1) 2.9 Data value in parentheses explained in ‘vertical datum issues’ section

San Juan 4.9 (5.4) 5.1 Data value in parentheses explained in ‘vertical datum issues’ section

Napindan 3.2* (4pm) 3.4 (4pm) (3.6 max)

Manual Observations (MMDA) which ceased at 4pm.

Rosario 7.4 7.7

Sto. Nino 11.7* (6pm) 11.6 (6pm) (11.7 max)

Nangka 12.0*

(12 noon)

12.8 (12 noon)

(15.4 max)

The modelled peak occurs at 3.30pm, 3.5 hours after the gauge broke (and > 3m above the last gauge value).

Hence the gauged value is not representative of the flood peak.

At the time of gauge breakage, the model is rising rapidly, and a small timing error in the input hydrology could be

responsible for the discrepancy. An equivalent gauge error could also be responsible.

Montalban 19.2* (4pm) 19.1 (4pm) (19.8 max)

Modelled peak occurs at 2pm. Gauge fails to record a value at 2pm or 3pm.

A peak of around ~ 20.0m can be independently estimated from a YouTube video of Ondoy Flooding @ Rodriguez Bridge between 2-4pm on 26 Sept 2009, by locating the

wet-dry boundary on the LIDAR data. See

https://www.youtube.com/watch?v=IT6kKqjDtok

A peak of around ~ 20.0m can be independently estimated from a YouTube video of Ondoy Flooding @ Rodriguez Bridge between 2-4pm on 26 Sept 2009, by locating the

wet-dry boundary on the LIDAR data. See

https://www.youtube.com/watch?v=IT6kKqjDtok

Page 65: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 59 Greater Metro Manila Area – Flood Risk Analysis

Figure 4.7. Comparison of modelled and measured water levels at key gauging stations during Tropical Storm Ondoy.

In the dyke regions of Taguig, the storage model was used to simulate flooding. Figure 4.8 compares

the simulated and observed peak depths at Labasan Pumping Station. The rate of rise and the peak

are reasonably well simulated; however, the model predicts a more rapid rate of drawdown than

evident in the data. This is likely to be caused by the model overestimating the real pumping station

capacity behind the dykes in Taguig, which is rated at 27m³/s but will be less because of pump failure

during Tropical Storm Ondoy. Another factor is the leakage of water through the flood defences at

Taguig, which was a problem during Tropical Storm Ondoy because of incomplete construction of

dykes. Regardless, the peak is well simulated, and this is of greatest importance for hazard

estimation.

Page 66: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

60 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

Figure 4.8. Comparison of modelled and observed water elevations behind the flood defences in the Taguig-Pateros area, during Tropical Storm Ondoy. Data was collected at Labasan Pumping Station.

Figures 4.9-4.13 compare the modelled peak depths with point observations of inundation depths on

the floodplains. The peak depth maps combine the two HEC-RAS models, the ‘flood defences

functioning’ model for the area behind the dykes in Taguig-Pateros, and the Laguna Lake water levels

to model lakeshore areas. Areas outside the modelled region are shown semi-transparently for

reference. There is broad agreement between with the simulated and observed patterns of inundation

during Tropical Storm Ondoy. Deep flooding is predicted and observed along the Marikina and San

Juan Rivers, to the east of the Mangahan Floodway, and along parts of the Laguna Lakeshore. Broad

areas of moderate to deep flooding also occur in western Manila to the North and South of the Pasig

River, and in the Marikina – Cainta Region and Taguig-Pateros Region. Similar inundation patterns

have been modelled in other recent studies of the area (WBCTI, 2012; www.nababaha.com).

Page 67: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 61 Greater Metro Manila Area – Flood Risk Analysis

Figure 4.9. Modelled and observed (points) depths during Tropical Storm Ondoy. Areas outside the model region are shaded semi-transparently. The subsequent figures provide a close-up comparison.

Page 68: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

62 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

Figure 4.10. Comparison of modelled and observed depths during Tropical Storm Ondoy around the San Juan and Pasig Rivers

Page 69: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 63 Greater Metro Manila Area – Flood Risk Analysis

Figure 4.11. Comparison of modelled and observed peak depths Tropical Storm Ondoy in the Lower Marikina / Mangahan / Napindan Area

Page 70: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

64 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

Figure 4.12.Comparison of modelled and observed depths during Tropical Storm Ondoy in the Upper

Marikina Area

Page 71: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 65 Greater Metro Manila Area – Flood Risk Analysis

Figure 4.13. Comparison of modelled and observed peak depths during Tropical Storm Ondoy along the Laguna Lakeshore area.

Page 72: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

66 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

4.2.2 Design Flood Scenarios

Inundation maps for design floods with AEPs of 1/5, 1/10, 1/25, 1/50, 1/100 and 1/200 are reported in

the final risk map products. As an example, Figure 4.14 shows the simulated 1/200 AEP peak flood

depths. The MGB flood susceptibility zones are also overlaid, for comparison and reference outside

the extent of the hydraulic model.

Page 73: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 67 Greater Metro Manila Area – Flood Risk Analysis

Figure 4.14. Modelled peak depths for an AEP 1/200 flood event.

4.3 Damage Estimation

For each AEP scenario, we computed the damaged floor area equivalent, building damage cost, and

number of people with inundated homes (Table 4.5 and Figure 4.14). For reference, computed values

Page 74: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

68 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

for Tropical Storm Ondoy are also shown. Example maps with this information are shown for the

design flood with an AEP of 1/200 (Figures 4.15-4.17). Similar maps for other AEPs are provided in

the set of final risk map products.

Table 4.5. Total damages estimated for each of the design flood scenarios, and Tropical Storm Ondoy.

AEP

Damage Metric

1/5 1/10 1/25 1/50 1/100 1/200 Tropical Storm Ondoy

Building damaged floor area equivalent (ha)

125 193 303 411 538 651 446

Building damage cost (million Pesos)

10682 16299 26431 36713 48596 59064 41097

Number of people with inundated homes (thousand

people)

705 967 1349 1665 1958 2164 1756

Page 75: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 69 Greater Metro Manila Area – Flood Risk Analysis

Figure 4.14. Damage estimates for each AEP flood scenario.

Page 76: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

70 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

Figure 4.15. Building damage intensity computed for the 1/200 AEP flood scenario.

Page 77: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 71 Greater Metro Manila Area – Flood Risk Analysis

Figure 4.16. Building damage cost estimated for the 1/200 AEP flood scenario.

Page 78: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

72 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

Figure 4.17. Number of people with inundated homes, estimated for the 1/200 AEP flood scenario.

Page 79: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 73 Greater Metro Manila Area – Flood Risk Analysis

4.4 Patterns of Flood Hazard and Risk

Large areas of Metro Manila are vulnerable to severe flood inundation (Figures 4.9-4.14), with depths

of one to several metres being widespread during large events. Fundamentally this is because much

of Manila is built on naturally flood prone lands, including floodplains along the Marikina, Pasig and

San Juan Rivers, tidal flats along Manila Bay, and various lakeshore and deltaic landforms around

Laguna Lake. The land surface in these areas was originally built by sediments deposited during

flooding, and would always have been flood prone. Significant efforts to reduce this flooding have

been made, such as the construction of the Mangahan Floodway, numerous pumping stations, flood

gates, drains and dykes. However, Manila remains very flood prone, and urban development has also

contributed to flooding by constricting or obstructing overland and river drainage pathways, reducing

soil infiltration capacity, and accelerating land subsidence in some areas.

In the hypothetical 1/200 AEP scenario, the patterns of flooding are qualitatively similar to those

observed during Tropical Storm Ondoy, but with greater flood depths and extents (Figure 4.14). The

deepest inundation (3+m) occurs along the Upper Marikina and San Juan Rivers. Widespread

inundation of ~ 0.5-2 m depth also occurs east of the Marikina River and Mangahan Floodway. This is

caused by a mixture of inflows from the local catchment, overflow from the Marikina River, and high

river levels in the Mangahan Floodway which inhibit drainage due to backwater effects. Flooding

occurs in the Lakeshore and Taguig-Pateros regions, due to high water levels in Laguna Lake. In the

areas west of the San Juan River, north and south of the Pasig River, flooding is widespread but

typically shallower than in other regions ( ~0.2 – 1.2m), and is driven by a mixture of local rainfall,

inflow from rivers and Manila Bay, and the flat topography which promotes relatively slow drainage.

Regarding flood damages, there is a clear difference in the spatial patterns when measured in terms

of the damaged floor area equivalent, the cost of building damages, and the population with inundated

homes (Figures 4.15-4.17). For large flood events (e.g. AEP 1/200), the damaged floor area

equivalent shows patches of particularly intense damage around the Marikina River near Tumana,

along the banks of the Mangahan Floodway and the San Juan River, and at various locations along

the Lakeshore and Taguig-Pateros regions. Highly damaged areas are characterised by the

simultaneous occurrence of deep flooding, dense settlement, and a large proportion of ‘Makeshift’ and

‘Wooden’ buildings. The latter are more intensely damaged by deep flooding than are other building

types (Figure 4.16). However, they are also less expensive to replace, and so the building damage

costs are comparatively more evenly spread out within zones that experience deep flooding. In terms

of the number of people with inundated homes, large parts of the city have around 10-50 thousand

people per square kilometre, with the most intense patches occurring at sites of with high population

density in predominantly low rise housing.

In less extreme events (e.g. 1/10 AEP), deep flooding is concentrated along the margins of the Upper

Marikina and San Juan Rivers, and floodplain flows are much less extensive. Moderate flooding

occurs along lakeshore areas and low-lying parts of Taguig, and along drainage paths east of the

Upper Marikina and Mangahan rivers. The damaged floor area is still intense around Tumana, and

remains significant in many areas bordering the Marikina and San Juan Rivers, and the Mangahan

Floodway. Many of these areas also have dense populations with inundated homes. The damaged

building costs are relatively more evenly distributed, due to the lower replacement cost of the most

vulnerable building types.

All damages increase strongly with increasing AEP (Table 4.5, Figure 4.14). Tropical Storm Ondoy

falls between the 1/50 and 1/100 AEP for every damage measure used herein. For the 1/200 AEP

Page 80: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

74 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

scenario, the building damages are around 40% greater than Tropical Storm Ondoy, and the

population with inundated homes is around 20% greater. While these damages are substantially larger

than those experienced during Ondoy, latter can serve as a reasonable ‘mental picture’ for the

patterns of inundation and damages expected from large flood events in the Pasig Marikina Basin.

4.4.1 Limitations of the Analysis

The flood hazard and risk information developed herein can assist in understanding the large scale

patterns in flood hazard and risk in the Pasig-Marikina Basin. However, this should be done with an

appreciation of the limitations in the underlying datasets, methods and models used.

Broadly, it is suggested that basin-scale work such as presented above can be usefully supplemented

with smaller scale flood studies to support local flood management decisions. Smaller scale studies

have a greater capacity to ground-truth input data, and to develop and test detailed hazard and

vulnerability models, whilst still drawing on inputs from larger scale work such as the present. Such

studies are the most appropriate to support robust local-scale flood management decisions, especially

if discrepancies are found between observational data and the results of larger scale studies, or

between multiple large scale studies.

Several other recent large-scale flood studies exist in the Pasig Marikina Basin, and they provide

inundation maps and/or damage estimates covering some of the same areas treated in this study.

Because they use different data and methodologies to the present study (and to each other), the

results of all these studies vary to some extent, although the general patterns of flooding in the basin

are broadly consistent between recent studies (WBCTI, 2012) and the present work. It is suggested

that decisions about large-scale flood management issues should draw on the results of multiple

studies where applicable (e.g. DICAMM, 2005; CTI, 2005; WBCTI, 2010; www.nababaha.com; Muto et

al., 2011). This will assist in making flood management decisions robust to the limitations of any single

study, and highlight areas which may need more investigation before consensus can be reached.

The key limitations to the present study are now outlined.

4.4.1.1 Hydrology and Hazard Scenarios

To define rainfall intensities in the hazard scenarios, the present study develops estimates of extreme

rainfall frequencies based on a statistical analysis of data from a limited number of rain gauges around

the Pasig Marikina basin. The statistical uncertainties in these estimates can be quite large (Table 4.2,

Figure 4.4), but are unavoidable without the existence of longer data records. They do not account for

possible future climate changes, which could alter the likelihoods of extreme rainfall events. For rare

events, the statistical uncertainties probably underestimate the true uncertainty, because they assume

the correctness of the fitted statistical model even when extrapolating to rare events. The underlying

rain gauge data has incomplete spatial coverage, which can lead to significant errors in the estimation

of basin-wide rainfall during any single event (Heistermann et al., 2013). This is not explicitly

accounted for in the computed confidence intervals. In the future, the combination of longer term rain

gauge records with calibrated radar data will probably help to better constrain extreme rainfalls in the

Pasig Marikina basin.

Similarly, the analysis of extreme lake levels in the present study is based on a limited dataset, and so

inevitably has significant statistical uncertainties (Table 4.3, Figure 4.6). It also does not account for

Page 81: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 75 Greater Metro Manila Area – Flood Risk Analysis

the uncertain impact of future climate changes or modifications to the Pasig River on annual maximum

lake levels.

The hazard scenarios developed in this study are based on a hypothetical combination of an input

rainfall time series, high Laguna Lake water levels, and a high water level in Manila Bay. A single

scenario is examined for each AEP, although in reality, an infinite number of possible scenarios could

be constructed, which would lead to differences in the pattern of flooding and damage. The spatially

constant design rainfall pattern is best suited to modelling areas flooded by a large fraction of the

catchment, and will tend to underestimate extreme localised rainfalls, and hence the AEP of flooding

in areas affected by small upstream catchments. In the Taguig-Pateros region, the degree of flooding

is significantly influenced by assumptions about the performance of flood defence structures, which in

practice have shown mixed performance. More detailed analysis based on Monte-Carlo type

modelling could be used to partially account for these factors, however, this process would still rely on

making assumptions about the probabilities of various hazard scenarios.

4.4.1.2 Flood modelling

The hazard scenarios are transformed into flood inundation maps using rainfall runoff and hydraulic

models. These approximate the actual flow processes with a linked set of simpler models (lumped sub

catchments with SCS runoff and routing models, networks of one-dimensional channels and storage

areas). The way these are defined depends partly on the subjective judgement of the modelling team,

and will have some effect on the results. In all cases, alternative models of these processes exist, and

it is unclear to what extent the results of the present study would be changed by using other models

(although this can partly be assessed by comparison with other studies). The models used in this

study are theoretically most appropriate for modelling channel and near-channel flows, and the results

might be further refined by the development and calibration of high-resolution 2D or linked 1D/2D

methods over the floodplains.

The models rely on input topographic data, and information on river structures. While the LIDAR DEM

provides relatively good elevation data for the present study, it cannot resolve submerged regions

(e.g. riverbeds), or fine topographic structures such as parapet walls. The information on the channel

geometry and flow structures available to the present study was incomplete, often around 10 years

old, and may not always provide a good description of the present-day geometry of the basin. Further,

in future the topography of Metro Manila will evolve, and this will affect flooding in ways that can’t be

anticipated in the present work.

The flood inundation model was calibrated to peak depths during the Tropical Storm Ondoy event, and

shows reasonable performance compared with data. However, it has not been tested in modelling

other events. Even in the calibration event, differences between the observations and the model do

occur (Figure 4.9-4.13), although the observations themselves may not always be accurate. In

addition, the model extent is limited to the Pasig Marikina basin, and excludes ‘upland’ regions such

as higher elevation parts of Quezon City, Taguig, and Cainta, which were included in the rainfall runoff

model in the present study. Flooding has been reported in these areas, but is not simulated in the

present model, which is most appropriate for simulating riverine flooding, driven by flows transported

from the main river systems.

4.4.1.3 Exposure and Vulnerability Inputs

The Exposure data is developed through a combination of subjective user judgement, and the

downscaling of other datasets to the exposure polygons. All these steps have limitations, which are

Page 82: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

76 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

outlined in the report for Exposure Information Development. For example, the distribution of building

types in each exposure polygon is estimated partly from 10 year old building information, which is not

thought to be highly suitable for present day Manila. In general, limited independent data exists with

which to check the properties in the exposure database, and it is expected that some inaccuracies

must remain.

The estimates of building replacement cost used in this study (Appendix A) are based on estimates

from two other sources, which were combined based on engineering judgement. In some instances

the cost estimates from each source would substantially disagree with each other.

The stage-damage vulnerability curves used in this study were developed with computational and

heuristic procedures, the limitations of which are described in “Development of vulnerability curves of

key building types in the Greater Metro Manila Area, Philippines” (Pacheco et al., 2013). In some

instances, large differences could occur between the computational and heuristic curves, and so

engineering judgement has been used to select the most appropriate curve. To our knowledge these

have not been tested against independent damage data.

4.4.1.4 Risk Analysis

As with the hazard scenarios developed for this study, the risk analyses are based on a suite of

scenarios. This approach doesn’t fully account for the diversity of storm spatial and temporal patterns

which contribute to the hazard. A more advanced approach could use Monte-Carlo analysis with a

suite of design storms. This would require work on defining the structure of the ‘random’ inputs to

define the hazard events, and further automation of the process of running models.

The damage metrics used in the present study are limited to estimates of the building damaged floor

area equivalent, damage cost, and number of people with inundated homes. Other damages not

treated in this study include indirect economic damages, the physical impacts of flood hazard on

people, or the increased spread of disease due to flooding. These may all be quite significant.

Page 83: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 77 Greater Metro Manila Area – Flood Risk Analysis

5 Conclusion

The results of this study include the statistical estimates of: a) Extreme rainfall frequencies at a

number of sites in the Pasig-Marikina River Basin based on a regional frequency analysis; b)

Catchment-averaged extreme rainfall frequencies; and c) Frequency of high lake levels in Laguna

Lake.

These statistical estimates were used to simulate design flood scenarios in the Pasig-Marikina River

Basin using the combined capabilities of HEC-HMS and HEC-RAS models. The model was shown to

perform reasonably well in simulating the flood depths associated with Tropical Storm Ondoy

(Ketsana). The calibrated model was then used to simulate the range of design flood events with

AEPs of 1/5 to 1/200 years. The damages associated with events were estimated using the recently

developed exposure database for Metropolitan Manila (refer to report for Component 2 - Exposure

Information Development) and the building vulnerability models (refer to report for Development of

vulnerability curves of key building types in the Greater Metro Manila Area). Damage estimates were

based on the ‘damaged floor area equivalent’, the ‘building damage cost’ and the ‘number of people

with inundated homes’.

The calibrated model was compared with the observed data both in the river and from the floodplain.

The result of the model calibration agrees well with the observed water level data from EFCOS. The

model results also typically agreed with the reported spot-depth data collected by nababaha.com

except for areas outside the model domain.

Aside from the observed data from EFCOS and nababaha.com, the result of this study is broadly

consistent with the results of other hydrological studies in the area such as the results from DICAMM

2005, CTI 2005, WBCTI 2010 and Muto et al. 2011. With these observations, it can be concluded that

this study was able to simulate the TS Ondoy event and the resulting hazard and risk maps with

different AEPs can provide useful inputs to support contingency planning of the local government units

and other applications relating to flood risk mitigation.

Page 84: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

78 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

6 Recommendations

Here we suggest several avenues for future work which could enhance our understanding of flood

hazards and impacts in Metro Manila:

1. Re-measurement of the vertical datum of river gauges in Manila and maintenance of this over

time.

2. Developing methods to estimate damages beyond building cost and number of people

inundated.

3. Investigation of Monte-Carlo methods for treating uncertainty in the flood risk analysis,

allowing for a more thorough exploration of the flood scenarios that are possible for a given

AEP.

4. Extending the risk analysis to other parts of Metro Manila which are covered by exposure data.

This would either require the development of more flood inundation models and extension of

the hydrological analyses, or alternatively, the use of other existing flood models.

5. Development of a capability for 2D and/or linked 1D/2D flood modelling in the CSCAND

agencies, to support more advanced hydraulic analyses.

6. Investigate methods for simulating flash flood hazards in the Pasig Marikina River Basin.

Page 85: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 79 Greater Metro Manila Area – Flood Risk Analysis

References

Abon, C.C., David, C.P.C. and Pellejera, N.E.B., 2010. Reconstructing the tropical storm Ketsana flood event in Marikina River, Philippines. Hydrology and Earth System Science Discussions 7:6081-6087.

Allen, R. J. and DeGaetano, A.T. 2005. Areal Reduction Factors for Two Eastern United States Regions with High Rain Gauge Density. Journal of Hydrologic Engineering 10(4): 327-335

Apel, H., Thieken, A.H., Merz, B., Bloschl, G. 2004. Flood risk assessment and associated uncertainty. Natural Hazards and Earth System Sciences 4:295-308

Aronica, G. T., Candela, A., Fabio, P. and Santoro, M., 2011. Estimation of flood inundation probabilities using global hazard indexes based on hydrodynamic variables. Physics and Chemistry of the Earth, doi:10.1016/j.pce.2011.04.00

Asuncion, J. F. and A. M. Jose, 1980: A study of the characteristic of the northeast and southwest monsoons in the Philippines. NRCP Assisted Project I-B-12.

Aureli, F., Mignosa, P. and Ziveri, C. 2006. Fully-2D and quasi-2D modelling of flooding scenarios due to embankment failure. Proceedings of River Flow 2006.

Badilla, R.A., 2008. Flood modelling in Pasig-Marikina River Basin. MSc Thesis, International Institute for Geo-Information Science and Earth Observation, Enschede, The Netherlands.

Bankoff, G. 2003. Vulnerability and Flooding in Metro Manila. IIAS Newsletter 31.

Berthet, L., Andreassian, V., Perrin, C. and Javelle, P. 2009. How crucial is it to account for the antecedent moisture conditions in flood forecasting? Comparison of event-based and continuous approaches on 178 catchments. Hydrology and Earth System Science 13:819-831.

Beven., K., 2001. Rainfall Runoff Modelling – The Primer. Wiley

Bouwer, L. M., Bubeck, P., Wagtendonk, A. J., Aerts, J. C. J. H., 2009. Inundation scenarios for flood damage evaluation in polder areas. Natural Hazards and Earth System Science. 9:1995-2007

Castellarin, A., Domeneghetti, A., Brath, A. 2011. Identifying robust large-scale flood risk mitigation strategies: A quasi-2D hydraulic model as a tool for the Po river. Physics and Chemistry of the Earth 36:299-308

Chaudhry, M. 2008. Open Channel Flow. Springer, New York.

Christian, J., Duenas-Osorio, L., Teague, A., Fang, Z. and Bedient, P., 2012. Uncertainty in floodplain delineation: expression of flood hazard and risk in a Gulf Coast watershed. Hydrological Processes. doi: 10.1002/hyp.9360

CTI, 2005. Study on Drainage System Operation for Metro Manila Flood Control Project: West of Mangahan Floodway.

DICAMM, 2005 The study on drainage improvement in the core area of Metropolitan Manila, Republic of the Philippines. Japan International Cooperation Agency, Metropolitan Manila Development Authority, Department of Public Works and Highways, The Republic of the Philippines.

DL&SI., 2010. Spon’s Asia-Pacific Construction Costs Handbook, Fourth Edition. Davis Langdon and Seah International.

Dumas, P., Hallegatte, S., Quintana-Segui, P., Martin, E., 2013. The influence of climate change on flood risks in France – first estimates and uncertainty analysis. Natural Hazards and Earth System Science 13:809-821.

Dung, N.V., Merz, B., Bardossy, A. and Apel, H. 2013. Flood hazard in the Mekong Delta – a probabilistic, bivariate, and non-stationary analysis with a short termed future perspective. Natural Hazards and Earth System Science Discussions 1:275-322.

Ellouze, M., Abida, A. and Safi, R., 2009. A triangular model for the generation of synthetic hyetographs. Hydrological Sciences Journal 54:287-299

Page 86: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

80 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

Estoque, M. A., 1956: Circulation and convergence over the tropical northwestern Pacific from May to December. Proceedings of the Eight Congress, 1953. Volume II Geology and Geophysics Meteorology. Quezon City, University of the Philippines, National Research Council of the Philippines.

Federal Emergency Management Agency (FEMA), 2003. Guidelines and specifications for flood hazard mapping partners. FEMA’s flood hazard mapping program, Washington DC.

Fowler, H.J. and Kilsby, C.G. 2003. A regional frequency analysis of United Kingdom extreme rainfall from 1961 to 2000. International Journal of Climatology 23:1313-1334.

Ghavidelfar, S.. Alvankar, S.R. and Razmkhah, A. 2011. Comparison of the Lumped and Quasi-distributed Clark Runoff Models in Simulating Flood Hydrographs on a Semi-arid Watershed. Water Resource Management 25: 1775-1790

HEC-HMS TRM, 2000. HEC-HMS Technical Reference Manual. US Army Corps of Engineers Hydraulic Engineering Center.

HEC-RAS TRM, 2010. HEC-RAS Technical Reference Manual. US Army Corps of Engineers Hydraulic Engineering Center.

Heistermann, M., Crisologo, I., Abon, C.C., Racoma, B.A., Jacobi, S., Servando, N.T., David, C.P.C. and Bronstert, A. 2013. Using the new Philippine radar network to reconstruct the Habagat of August 2012 Monsoon event around Metro Manila. Natural Hazards and Earth System Science 13:653-657

Horritt, M.S. and Bates, P.D. 2002. Evaluation of 1D and 2D numerical models for predicting river flood inundation. Journal of Hydrology 268 87-99.

Hosking, J.R.M. and Wallis, J.R. 1997. Regional Frequency Analysis: An approach based on L-moments. Cambridge University Press.

Kintanar, R.L., 1984: Climate of the Philippines, PAGASA Technical Report

Kysley, J. 2008. A cautionary note on the use of nonparametric bootstrap for estimating uncertainties in extreme value models. Journal of Applied Meteorology and Climatology 47:3236-3251

Madsen, H. and Skotner, C., (2005). Adaptive state updating in real-time river flow forecasting—a combined filtering and error forecasting procedure. Journal of Hydrology 308:302-312

Middelman, M., (2002). Flood Risk in South East Queensland, Australia. 27 Hydrology and Water Resources Symposium, Institute of Engineers Australia, 20-23 May 2002, Melbourne.

Moramarco, T., Melone, F., and Singh, V.P. 2005. Assessment of flooding in urbanized ungauged basins: a case study of the Upper Tiber area, Italy. Hydrological Processes 19:1909-1924

MPWTC, 1952. Plan for the Drainage of Manila and Suburbs. Report of the Ministry of Public Works, Transportation and Communication

Muto, M., Morishita, K. and Syson, L. 2011 Impacts of Climate Change upon Asian Coastal Areas: The case of Metro Manila. Japan International Cooperation Agency

NDRMMC., 2009. Final Report on Tropical Storm Ondoy (Ketsana) and Typhoon Pepeng (Parma). National Disaster Coordinating Council.

Neal, J., Keef, C., Bates, P., Beven. K. and Leedal, D. 2012. Probabilistic flood risk mapping including spatial dependence. Hydrological Processes doi: 10.1002/hyp.9572

NSO, 2010. 2010 Census of Population and Housing, National Capital Region. National Statistics Office.

Pacheco, B.M et al., 2013. Development of vulnerability curves of key building types in the Greater

Metro Manila Area, Philippines. Institute of Civil Engineering, University of the Philippines Diliman,

Quezon City.

Patvivatsiri, P., 1972: The rainfall distribution associated with the intensified SW monsoon of August 31st to September 4th, 1970. PAGASA Technical Series No. 16

Pechlivanidis, I.G., Jackson, B.M., McIntyre, N.R. and Wheater, H.S. 2011. Catchment Scale Hydrological Modelling: A review of model types, calibration approaches and uncertainty analysis methods in the context of recent developments in technology and applications. Global NEST Journal 13(3): 193-214

Page 87: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 81 Greater Metro Manila Area – Flood Risk Analysis

Perica, S. et al., 2011. NOAA Atlas 14 Precipitation-Frequency Atlas of the United States. Volume 6, Version 2.0, California.

Poretti, I. and Amicis, M. D., 2011. An approach for flood hazard modelling and mapping in the medium Valtellina. Natural Hazards and Earth System Science 11:1141-1151

RFCOSMM, 2002. Report on Technical Guidance for the Project for the Rehabilitation of the Flood Control, Operation and Warning System in Metro Manila.

Savenije, H. H. G. 2005. Salinity and Tides in Alluvial Estuaries. Delft, The Netherlands.

Scawthorn, C., Blais, N., Seligson, H., Tate, E., Mifflin, E., Thomas, W., Murphy, J. and Jones, C., 2006. HAZUS-MH Flood loss estimation methodology 1: Overview and Flood Hazard Characterization. Natural Hazards Review, 7(2):60-71

Smith, G. and Wasko, C. 2012. Two Dimensional Simulations In Urban Areas – Representation of Buildings in 2D Numerical Flood Models. Australian Rainfall and Runoff Project 15.

Syme., W.J., Pinnel. M.G. and Wicks., J.M. 2004. Modelling Flood inundation of Urban areas in the UK using 2D/1D hydraulic models. Proceedings of the 8

th National Conference on Hydraulics in Water

Engineering.

Kjeldsen, T.R., Jones, D.A. and Bayliss, A.C. 2008. Improving the FEH statistical procedures for flood frequency estimation. Joint Defra/Environment Agency Flood and Coastal Erosion Risk Management R&D Program.

Wallis, J.R., Schaefer, M.G., Barker, B.L. and Taylor, G.H. 2007. Regional Precipitation frequency analysis and spatial mapping for 24 hour and 2 hour durations for Washington State. Hydrology and Earth System Science 11:415-442

WBCTI, 2012. Master Plan for Flood Management in Metro Manila and Surrounding Areas. Final Draft Report. The World Bank; CTI Engineering International Co., Ltd; Woodfields Consultants, Inc.

Winsemius, H.C., Van Beek, L.P.H., Jongman, B., Ward, P.J. and Bouwman, A. 2012. A framework for global river flood risk assessments. Hydrology and Earth Systems Science Discussions 9:9611-9659

Woodhead, S., Asselman, N., Zech, Y., Soares-Frazao, S., Bates, P. and Kortenhaus, A. 2007. Evaluation of Inundation Models. Limits and Capabilities of Models. Floodsite Report T08-07-01

Page 88: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

82 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

Appendix A - Parameters used in the Rainfall Runoff Model Sub catchments

Sub catchment X* Y* Area

(km^2) CN

T_lag (min)

Boundary type*

Rain Gauge

(Calibration)*

wawa 528647.5654

1628279.332 287 74 168 Discharge Boso Boso

mt_oro_1 519975.7758

1632800.436 34.98 77 60 Discharge Mt Oro

Nangka 517089.2535

1618391.966 51.585 79 133 Discharge Aries

lower_marikina 507888.4813

1612425.73 8.175 79 28 Uniform lateral

Aries

north_west_of_nangka

511570.2903

1628053.333 24 81 73 Uniform lateral

Aries

east_of_montalban 519972.3158

1624260.754 25.5 74 43 Discharge Aries

North_of_nangka 517289.4862

1622747.779 24.8 77 44 Discharge Aries

Mangahan_napindan_area

510152.9448

1610971.532 16.683 92 10 Storage Aries

Pasig_01 504857.6887

1610519.985 15.6 93 52 Uniform Lateral

Science Garden

Pasig_02 499264.7122

1613715.972 73.148 93 10 Storage Science Garden

Sanjuan_1 503570.8552

1613682.93 3.45 92 33 Discharge Science Garden

Sanjuan_2 502109.1387

1614740.969 2.705 92 4 Uniform lateral

Science Garden

Sanjuan_3 506119.879 1613797.266 4.7 92 31 Discharge Science Garden

Sanjuan_4 503863.8376

1615787.891 8.7 92 33 Uniform lateral

Science Garden

Sanjuan_5 501144.3831

1617137.218 3.4 92 36 Uniform lateral

Science Garden

Sanjuan_t6 505525.39 1617675.493 15.2 92 33 Uniform lateral

Science Garden

Sanjuan_S7 503966.0222

1620631.795 10.22 92 53 Discharge Science Garden

Sanjuan_8 501397.7127

1620064.582 2.113 92 4 Uniform lateral

Science Garden

Sanjuan_9 501017.718 1622414.917 2.4 92 51 Discharge Science

Page 89: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 83 Greater Metro Manila Area – Flood Risk Analysis

3 Garden

Sanjuan_10 505239.8782

1623174.906 18.6 92 98 Discharge Science Garden

Sanjuan_11 502565.8436

1621401.599 2.866 92 5 Uniform lateral

Science Garden

Sanjuan_12 504184.3382

1624652.662 5.367 92 44 Discharge Science Garden

Sanjuan_13 501960.6673

1623160.832 3.26 92 18.6 Uniform lateral

Science Garden

Sanjuan_5b 502692.5084

1618319.422 1.1 92 17 Uniform lateral

Science Garden

Sanjuan_5c 500102.917 1619360.888 7.5 92 113 Discharge Science Garden

Mt_oro_2 516295.592 1633859.883 45.5 77 60 Uniform lateral

Science Garden

Dilian_crk 504822.9894

1608317.274 2.5 80 15 Discharge Science Garden

Maricaban 504653.0092

1606121.874 5 80 17 Discharge Science Garden

Calatagan 503463.1475

1608820.98 1.72 80 9 Discharge Science Garden

Casili_Crk 497819.8038

1621015.634 1.76 92 16 Discharge Science Garden

Sunog_Apog 498686.0266

1620432.303 1.456 92 16 Discharge Science Garden

Marik_tumana 510666.498 1624254.092 27.6 78 28 Uniform lateral

Science Garden

SANJUAN_ALL 505775 1620271 82.992 92 192 Uniform lateral

Science Garden

East_mangahan 517101.4099

1614057.54 103.7 90 48 Storage Aries

*’X’, ‘Y’ are the x and y coordinates of a point in the sub catchment polygon;

*‘Boundary type’ describes how the outflow hydrograph from the sub catchment was applied as a

boundary condition in HEC-RAS. ‘Discharge’ refers to a point inflow discharge boundary condition;

‘Uniform Lateral’ refers to uniform lateral inflow into a channel; ‘Storage’ refers to the use of uniform

lateral inflow in all storage areas within the catchment polygon, at a rate proportional to area of each

storage area.

* Rain gauge (calibration) refers to the rain gauge that was used to assign rainfall to the catchment

during the Tropical Storm Ondoy Calibration Run. Not all gauges in the monitoring network were used,

because of instrument malfunction.

Page 90: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

84 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

Appendix B - Estimated building costs per m² (‘000s of Peso), for different combinations of Building type, L4_USE and L5_USE.

The latter land-use classification variables are provided in the exposure database. Estimates are based on DL&SI. (2010), and Muto et al. (2011)

L4_USE L5_USE W1 W2 W3 N CHB URA URM RM1 RM2 MWS CWS C1 C2 C4 PC1 PC2 S1 S2 S3 S4

Formal Settlements

Residential 4.0 1.2 5.9 6.5 6.5 5.2 7.8 12.2 16.0

18.0 20.0 30.0 35.0 15.0 32.0

Formal Settlements

Mixed Residential and Small Commercial

5.6 5.6 2.0 1.2 6.5 7.2 7.2 8.0 8.0 6.0 9.0 13.2 18.0

20.0 20.0 20.0 30.0 35.0 15.0 32.0

Formal Settlements

Small Commercial

5.6 1.2 7.4 8.1 8.1 6.6 9.9 11.0 20.0

22.0 30.0 37.0 15.0 32.0

Informal Settlements

Mixed Informal Settlements

4.0 1.2 3.0 3.0 3.0 5.2 7.8 12.2 16.0

18.0 30.0 35.0 15.0 32.0

Education Schools 4.3 4.3 13.2 22.0 25.0

27.0 27.0 32.0

Education Universities 6.5 7.8 23.0 27.0 27.0

Education Vocational Colleges

6.5 22.0 27.0

Education Day Care Centers

13.2 22.0

Page 91: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 85 Greater Metro Manila Area – Flood Risk Analysis

L4_USE L5_USE W1 W2 W3 N CHB URA URM RM1 RM2 MWS CWS C1 C2 C4 PC1 PC2 S1 S2 S3 S4

Health and Welfare

Hospitals 7.5 25.0 36.0 38.0

Health and Welfare

Health Centers

30.0

Health and Welfare

Aged Care Centers

4.3 20.0

Health and Welfare

Rehabilitation Centers

4.3 20.0

Health and Welfare

Orphanages 4.3 7.3 12.2 18.0

20.0 35.0 35.0 35.0

Government Administration

15.0 25.0 30.0

30.0 30.0 35.0

Government Services 6.5 1.2 7.0 7.0 7.0 7.0 10.0 25.0 30.0

30.0 35.0 35.0 15.0 35.0

Government Accommodation

7.0 25.0

Government Operations 25.0 30.0

30.0 35.0 35.0 35.0

Emergency and Defense

Police 25.0 30.0

35.0

Emergency and Defense

Fire and Rescue

25.0 30.0

Page 92: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

86 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

L4_USE L5_USE W1 W2 W3 N CHB URA URM RM1 RM2 MWS CWS C1 C2 C4 PC1 PC2 S1 S2 S3 S4

Emergency and Defense

Ambulance 25.0

Emergency and Defense

Armed Forces

15.0 25.0 30.0

30.0 35.0 35.0 35.0

Cultural Places of Worship

6.3 12.0 15.0

Cultural Places of Assembly

12.0 15.0 35.0

Cultural Cemeteries 4.3 12.0 15.0 30.0

30.0 35.0 35.0

Leisure Exhibitions 40.0 40.0

40.0 50.0 50.0 50.0

Leisure Indoor Sports 25.0 42.0 45.0

Leisure Outdoor Sports and Playgrounds

25.0 42.0 45.0 45.0 15.0

Energy Production

Electricity 15.0 20.0 15.0

Energy Production

Gas 4.3 10.0 15.0 20.0

Page 93: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 87 Greater Metro Manila Area – Flood Risk Analysis

L4_USE L5_USE W1 W2 W3 N CHB URA URM RM1 RM2 MWS CWS C1 C2 C4 PC1 PC2 S1 S2 S3 S4

Energy Production

Liquid Fuels 4.3 10.0 15.0 18.0

20.0 20.0 22.0 32.0

Water Supply Potable Water Storage

15.0 18.0

20.0

Water Supply Treatment 15.0

Water Supply Transmission 15.0 18.0

20.0

Water Supply Urban Supply 20.0 25.0

Communications

Telecommunications

20.0

Communications

Broadcasting 20.0 23.0

Communications

Postal Services

20.0 23.0

Waste Management

Solid Waste 12.0 15.0 20.0

Waste Management

Liquid Waste 12.0 15.0 20.0

Waste Management

Hazardous Waste

12.0 20.0

Page 94: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

88 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

L4_USE L5_USE W1 W2 W3 N CHB URA URM RM1 RM2 MWS CWS C1 C2 C4 PC1 PC2 S1 S2 S3 S4

Transportation Road Transport

4.3 1.2 15.0 20.0 20.0

Transportation Rail Transport

15.0 20.0

Transportation Air Transport 4.3 4.3 25.0 35.0 20.0

Transportation Marine Transport

5.9 6.5 25.0 30.0

Transportation Cargo and Storage

4.3 4.3 12.0 15.0 20.0

Heavy Industry Manufacturing

4.3 4.3 13.0 17.0 18.0

20.0 20.0 22.0 25.0

Heavy Industry Processing 4.3 4.3 13.0 17.0 18.0

20.0 20.0 22.0 25.0

Heavy Industry Mining 17.0 18.0

20.0 20.0 22.0 25.0

Heavy Industry Construction 17.0 20.0

Major Commercial

Retail 5.4 20.0 30.0 35.0 30.0

Major Commercial

Wholesale 5.4 17.0 25.0 30.0 15.0

Page 95: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 89 Greater Metro Manila Area – Flood Risk Analysis

L4_USE L5_USE W1 W2 W3 N CHB URA URM RM1 RM2 MWS CWS C1 C2 C4 PC1 PC2 S1 S2 S3 S4

Major Commercial

Office 5.4 5.4 18.0 28.0 32.0

Major Commercial

Accommodation

18.0 28.0 32.0

Major Commercial

Mixed Major Commercial

5.4 5.4 18.0 28.0 32.0

Major Commercial

Markets 5.4 5.4 18.0 28.0 30.0

Major Commercial

Tourism Facilities

5.4 3.9 5.9 7.0 28.0 30.0 30.0

Food Security Government Grain Storage

20.0

Food Security Private Storage

10.0 14.0

Flood Control Flood Gates 10.0 14.0

Flood Control Flood Monitoring Stations

10.0 15.0

Flood Control Flood Pumping Stations

15.0

Page 96: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

90 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

L4_USE L5_USE W1 W2 W3 N CHB URA URM RM1 RM2 MWS CWS C1 C2 C4 PC1 PC2 S1 S2 S3 S4

Rural Residential

Yet to be defined

4.3 1.2 5.9 6.5 6.5 5.2 10.0 15.0 18.0

18.0 20.0 22.0 15.0 25.0

Agriculture Yet to be defined

4.3 1.2 5.9 6.5 6.5 5.2 10.0 15.0 18.0

18.0 20.0 22.0 15.0 25.0

Aquaculture Yet to be defined

4.3 1.2 5.9 6.5 6.5 5.2 10.0 15.0 18.0

18.0 20.0 22.0 15.0 25.0

Horticulture Yet to be defined

4.3 1.2 5.9 6.5 6.5 5.2 10.0 15.0 20.0 15.0

Forestry Yet to be defined

4.3 1.2 5.9 6.5 6.5 5.2 10.0 15.0 20.0 15.0

Livestock Yet to be defined

4.3 1.2 5.9 6.5 6.5 5.2 10.0 15.0 20.0 15.0

Livestock Feedlots 4.3 1.2 5.9 6.5 6.5 5.2 10.0 15.0 20.0 15.0

Market Gardening

Yet to be defined

4.3 1.2 5.9 6.5 6.5 5.2 10.0 15.0 20.0 22.0 15.0

Mixed Farming Yet to be defined

4.3 1.2 5.9 6.5 6.5 5.2 10.0 15.0 18.0

18.0 20.0 22.0 15.0 25.0

Poultry Yet to be defined

4.3 1.2 5.9 6.5 6.5 5.2 10.0 15.0 20.0 15.0

Vacant Areas Yet to be defined

4.3 6.5 10.0 15.0 20.0

Natural Areas National Parks

4.3 4.3 6.5 10.0 15.0 18.0

18.0 25.0 20.0

Page 97: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the 91 Greater Metro Manila Area – Flood Risk Analysis

L4_USE L5_USE W1 W2 W3 N CHB URA URM RM1 RM2 MWS CWS C1 C2 C4 PC1 PC2 S1 S2 S3 S4

Reserved Areas Yet to be defined

4.3 1.2 5.9 6.5 6.5 5.2 10.0 15.0 20.0 15.0 25.0

Reserved Areas Urban Parks 4.3 6.5 10.0 15.0 18.0

18.0 20.0 22.0 25.0

Reserved Areas Greenbelts 4.3 6.5 10.0 15.0 20.0

Reserved Areas Buffer Zones 4.3 6.5 10.0 15.0 20.0

Reclamations

Page 98: Enhancing Risk Analysis Capacities for Flood, Tropical ... · 1.1 Background of the Study The Philippines is one of the most flood-prone countries in the world. For the last ten years,

92 Enhancing Risk Analysis Capacities for Flood, Tropical Cyclone Severe Wind and Earthquake for the Greater Metro Manila Area – Flood Risk Analysis

Appendix C - Floor heights for different building categories, based on field survey data collected by PHIVOLCS


Recommended