8/4/2019 7dj6P

1/33

PresenterPresenter

Eric WarmathEric Warmath,, NDOTNDOT

Using Remote Sensing and GIS toUsing Remote Sensing and GIS toImprove Runoff IndexImprove Runoff Index

Determination for UrbanDetermination for UrbanHydrologic ModelingHydrologic Modeling

8/4/2019 7dj6P

2/33

AcknowledgmentsAcknowledgmentsuu South Dakota State Univ. Brookings, SDSouth Dakota State Univ. Brookings, SD

Ms. Pravara ThanapuraMs. Pravara Thanapura, Principal Investigator, and Research, Principal Investigator, and ResearchAssociate, Engineering Resource Center (ERC), Ph.D.Associate, Engineering Resource Center (ERC), Ph.D. SStudent in thetudent in the

Geospatial Science and Engineering Program, andGeospatial Science and Engineering Program, and oowner of GeoTechwner of GeoTechConsulting, LLC.Consulting, LLC.

Dr. Dennis HelderDr. Dennis Helder,, Director of Engineering Research, DepartmentDirector of Engineering Research, DepartmentHead,Head, Department of Electrical Engineering and Computer ScienceDepartment of Electrical Engineering and Computer Science **

Mr. Kevin DalstedMr. Kevin Dalsted, Director, ERC, Director, ERC

Dr. Suzette BurckhardDr. Suzette Burckhard, Hydrologist, College of Engineering, Hydrologist, College of Engineering** Dr. Dwight GalsterDr. Dwight Galster, Statistician, Department of Mathematics, Statistician, Department of Mathematics**

Ms. Mary OMs. Mary ONeillNeill, Program Manager, Office of Remote Sensing, ERC, Program Manager, Office of Remote Sensing, ERC**

uu City of Sioux Falls, SDCity of Sioux Falls, SD Steve Van AartsenSteve Van AartsenSteve Van AartsenSteve Van AartsenSteve Van AartsenSteve Van AartsenSteve Van AartsenSteve Van Aartsen, GIS Supervisor,, GIS Supervisor, Thanks for all the dataThanks for all the data!! Jeff DunnJeff Dunn, City Drainage Engineer, for reviews and information, City Drainage Engineer, for reviews and information

Sam TrebilcockSam Trebilcock, Transportation Planner, for input and information, Transportation Planner, for input and information

* Denotes coDenotes co--authorauthor

8/4/2019 7dj6P

3/33

TopicsTopicsuu Presentation ObjectivePresentation Objective

uu

Research Contributions & DOT End User BenefitsResearch Contributions & DOT End User Benefitsuu Hydrologic Model BackgroundHydrologic Model Background

uu Runoff Methods & Runoff Index for Urban Drainage DesignRunoff Methods & Runoff Index for Urban Drainage Designand Analysisand Analysis

uu Composite Runoff Index Geographic Model for IndustryComposite Runoff Index Geographic Model for IndustryStandard Runoff Index CalculationsStandard Runoff Index Calculations

uu Mapping Impervious Area and Open SpaceMapping Impervious Area and Open Space

uu

GIS SpatialGIS Spatial ModelingModelinguu ResultsResults

uu ConclusionConclusionss

8/4/2019 7dj6P

4/33

Presentation ObjectivesPresentation Objectivesuu Show the potential value of the research, in urbanShow the potential value of the research, in urban

areas, for hydrologic engineers at DOTareas, for hydrologic engineers at DOTs nationally.s nationally.

uu Briefly describe the 2 most common runoff methodsBriefly describe the 2 most common runoff methods

and associated Runoff Indexand associated Runoff Index

uu

Describe the research:Describe the research: Integration of remote sensingIntegration of remote sensingand GIS for determining industry standard values ofand GIS for determining industry standard values of

NRCS CN & C using the composite runoff indexNRCS CN & C using the composite runoff index

geographic model developed by Thanapura in 2005geographic model developed by Thanapura in 2005--6.6.

8/4/2019 7dj6P

5/33

BenefitsBenefitsuu Lead to an improved scheme for determining theLead to an improved scheme for determining the

standard runoff index used in urban watershed runoffstandard runoff index used in urban watershed runoff

assessment methodsassessment methods -- the NRCS curve number andthe NRCS curve number andrational methods.rational methods.

uu Demonstrate a more precise, simpler, and efficientDemonstrate a more precise, simpler, and efficientapproach of calculating runoff index.approach of calculating runoff index.

uu Allow repeatability and consistency of the results byAllow repeatability and consistency of the results byremoving human error factors while increasing speedremoving human error factors while increasing speedand potentially reducing costs of analysis andand potentially reducing costs of analysis and

mapping for both methods.mapping for both methods.uu Lead to an improved scheme of urban imperviousLead to an improved scheme of urban impervious

surface detectionsurface detection a key indicator of the effects ofa key indicator of the effects ofnonnon--point source pollution runoff and of future waterpoint source pollution runoff and of future water

and ecosystem quality.and ecosystem quality.

8/4/2019 7dj6P

6/33

DOT End User BenefitsDOT End User Benefitsuu Beneficial to engineers involved drainage analyses andBeneficial to engineers involved drainage analyses and

designs in urban areas.designs in urban areas.

uu

Allow for identification of structures nearing designAllow for identification of structures nearing designcapacity and needing replacement due to the effects ofcapacity and needing replacement due to the effects ofincreased urbanization on a drainage basin.increased urbanization on a drainage basin.

u Identify sites for potential property damage or loss of life.

Table. Minor Structure Design Frequencies (Viessman and Lewis, 2003).

Return period, Tr Frequency = 1/Tr

0-400 ADT* 10 yr 0.10

400-1700 ADT 10-25 yr 0.10-0.04

1700-5000 ADT 25 yr 0.04

5000+ ADT 50 yr 0.02

Airfields 5 yr 0.20

Railroads 25-50 yr 0.04-0.02

Storm drainage 2-10 yr 0.50-0.10

Levees 2-50 yr 0.50-0.02

Drainage ditches 5-50 yr 0.20-0.02* ADT = average daily traffic

Type of minor structure

Highway crossroad drainage*

8/4/2019 7dj6P

7/33

Why NDOT & SDSU ?Why NDOT & SDSU ?uu One of the conference discussions at GISOne of the conference discussions at GIS--T 2004 inT 2004 in

Rapid City, SD, was getting more use out of imageryRapid City, SD, was getting more use out of imagery

and remote sensing at DOTand remote sensing at DOTs.s.uu I met the principal investigator, was impressed withI met the principal investigator, was impressed with

the research, and saw the potential benefits for NDOT.the research, and saw the potential benefits for NDOT.

uu Since Nevada is the fastest growing state and urbanSince Nevada is the fastest growing state and urbandrainage problems sometimes occur, the research isdrainage problems sometimes occur, the research isvery relevant to NDOT issues .very relevant to NDOT issues .

uu

We have tentative plans to work in the Las Vegas areaWe have tentative plans to work in the Las Vegas areathis year as part of the study.this year as part of the study.

8/4/2019 7dj6P

8/33

Hydrologic Models BackgroundHydrologic Models Backgrounduu There are 2 main hydrologic modeling methods used by theThere are 2 main hydrologic modeling methods used by the

majority of practicing engineers. These models weremajority of practicing engineers. These models were

developed for storm water calculation in engineering stormdeveloped for storm water calculation in engineering stormdrainage design and water resource planning and analysis.drainage design and water resource planning and analysis.

1.1. Natural Resource Conservation Service (NRCS), NRCSNatural Resource Conservation Service (NRCS), NRCS--CNCN

Method.Method.

2.2. The Rational Method.The Rational Method.

uu According to an EPA studyAccording to an EPA study ~86%~86%of private engineeringof private engineering

firms, water boards, and other government entities arefirms, water boards, and other government entities areusing one or both methods in their hydraulic engineering.using one or both methods in their hydraulic engineering.

uu NDOT and the City of Sioux Falls, SD use both.NDOT and the City of Sioux Falls, SD use both.

8/4/2019 7dj6P

9/33

Runoff Curve Number MethodRunoff Curve Number Method

uu The NRCS Runoff Curve Number (NRCSThe NRCS Runoff Curve Number (NRCS--CN)CN)

method:method:

Used to estimate runoff from storm rainfall.Used to estimate runoff from storm rainfall.

Well established in hydrologic engineering andWell established in hydrologic engineering and

environmental impact analysis.environmental impact analysis.

Widely used by practicing engineers andWidely used by practicing engineers and

hydrologists nationally and internationally.hydrologists nationally and internationally.

8/4/2019 7dj6P

10/33

Curve Number (CN)Curve Number (CN)

Industry Standard TableIndustry Standard Tableuu CN is an runoffCN is an runoff

index described inindex described in

detail in the TRdetail in the TR--5555(NRCS 1986).(NRCS 1986).

uu CN is a function of 3CN is a function of 3

factors:factors:

Hydrologic soilHydrologic soil

groupgroup

Cover complexCover complex

(Land Cover /(Land Cover /Land Use)Land Use)

Antecedent soilAntecedent soil

moisturemoisture

8/4/2019 7dj6P

11/33

Curve Number (CN)Curve Number (CN)uu Proposed runoff coefficients for the compositeProposed runoff coefficients for the composite

runoff calculationrunoff calculation for urban land usefor urban land use

recommended by McCuen in 2005recommended by McCuen in 2005

(Thanapura, 2006).(Thanapura, 2006).

Land Cover

Character of Surface1: A B C D

Impervious Areas 98 98 98 98

Open Spaces - Good Condition 39 61 74 80

Curve Number (CN) for

SCS Hydrologic Soil Group

8/4/2019 7dj6P

12/33

Rational MethodRational Method Developed by Mulvaney in 1851 and refinedDeveloped by Mulvaney in 1851 and refined

by Kuichling and others in the late 19by Kuichling and others in the late 19ththccentury.entury.

Is the preferred method for use in smallerIs the preferred method for use in smaller

drainage basinsdrainage basins (

8/4/2019 7dj6P

13/33

Runoff Coefficients (C)Runoff Coefficients (C)

Industry Standard TableIndustry Standard Table

uu C is a function of 3C is a function of 3factors:factors:

Land coverLand cover(Impervious and(Impervious and

Open space)Open space) Hydrologic SoilHydrologic Soil

GroupGroup

SlopeSlope

8/4/2019 7dj6P

14/33

Runoff Coefficient (C)Runoff Coefficient (C)

uu ProposedProposed rrunoffunoff ccoefficients for theoefficients for the ccompositeomposite rrunoffunoff

ccalculation defined from thealculation defined from the rrunoffunoff ccoefficient, Coefficient, C

rrecommended by the American Society of Civilecommended by the American Society of Civil

Engineers (ASCE), the Water Pollution ControlEngineers (ASCE), the Water Pollution Control

Federation (WPCF) in 1969, and McCuen in 2005Federation (WPCF) in 1969, and McCuen in 2005(Thanapura, 2006).(Thanapura, 2006).

8/4/2019 7dj6P

15/33

8/4/2019 7dj6P

16/33

Research ObjectiveResearch Objective

The objective of this study was to demonstrateThe objective of this study was to demonstrate

and evaluate Normalized Differenceand evaluate Normalized DifferenceVegetation Index (NDVI) data derived fromVegetation Index (NDVI) data derived from

QuickBird (QB) high resolution satelliteQuickBird (QB) high resolution satellite

imagery to mapimagery to map land coverland cover surfacesurface

characteristics such as impervious area andcharacteristics such as impervious area and

open space for runoff index numberopen space for runoff index numberdetermination in urban watersheds.determination in urban watersheds.

8/4/2019 7dj6P

17/33

Study MethodStudy MethodComposite Runoff Index Spatial Model

Digital Data & Pre-ProcessingData Merging and Integration

QuickBird NDVI Imagery & GIS LayersDecision and Classification

Image ClassificationUnsupervised ISODATA Algorithm & QuickBird NDVI

Classification Output

Accuracy AssessmentReference Data Ortho Photo

Reports and GIS Data

GIS Spatial ModelingThe Composite of Runoff Index CN & C calculations and comparisons

Reject / Accept Hypothesis

8/4/2019 7dj6P

18/33

Composite Runoff IndexComposite Runoff Index

Geographic ModelGeographic Model

In this study, the composite runoff index geographic model was applied todevelop a GIS spatial model for the composite runoff index calculation of boththe NRCS CN and C for the NRCS CN method and the rational method.

8/4/2019 7dj6P

19/33

Study AreaStudy Area

The QB multi spectral image (4-3-2)The QB NDVI Image

(Band 4 Band 3 / Band 4 + Band 3)

This is ~2900 acres in the southwest portion of the City of Sioux Falls, SDrepresenting almost all potential land use types.

8/4/2019 7dj6P

20/33

Digital DataDigital Datauu QuickBirdQuickBird

uu ~8 ft (~2.4 m) [April 26, 2004]~8 ft (~2.4 m) [April 26, 2004]

uu Blue, Green, Red, and NIR bandsBlue, Green, Red, and NIR bands

uu OrthophotosOrthophotos

uu 2 ft color (0.6 m)[April 23, 2004]2 ft color (0.6 m)[April 23, 2004]

uu 0.5 ft color (0.15 m)[May 20, 2002]0.5 ft color (0.15 m)[May 20, 2002]

uu QB NDVI (band4QB NDVI (band4--band3/band4+band3)band3/band4+band3)

(created using Erdas Imagine 8.7)(created using Erdas Imagine 8.7)

uu GIS LayersGIS Layers

Parcel, hydro, and street layers.Parcel, hydro, and street layers.

NRCS SSURGO (1:24K) soil dataNRCS SSURGO (1:24K) soil data

8/4/2019 7dj6P

21/33

PrePre--ProcessingProcessing

QB RegistrationQB Registration (2.39m/ 8ft)(2.39m/ 8ft)A Root Mean Square of 0.42m/1.38ft (0.69pixel)A Root Mean Square of 0.42m/1.38ft (0.69pixel)

The QuickBird multi image (4-3-2)on the left with the 2004 Ortho-

image (1-2-3) on the right.

The QuickBird NDVI image on the leftwith the 2004 Ortho-image (1-2-3)

on the right.

The registered 2004 images displayed at same scale and location.

8/4/2019 7dj6P

22/33

Classification ApproachClassification Approach

Why High Resolution NDVI?Why High Resolution NDVI?uu Reduce heterogeneous spectralReduce heterogeneous spectral--radiometricradiometric

characteristics within land use land covercharacteristics within land use land coversurfaces in the QB image.surfaces in the QB image.

uu

Normalize potential atmospheric effects withinNormalize potential atmospheric effects within

the image.the image.

uu Improve accuracy of mapping imperviousImprove accuracy of mapping impervious

surface and open space as used in thesurface and open space as used in theproposed runoff index calculation.proposed runoff index calculation.

8/4/2019 7dj6P

23/33

Classification ApproachClassification ApproachWhy Unsupervised Classification?Why Unsupervised Classification?

uu To maximize control over the menu ofTo maximize control over the menu of

informational classes.informational classes.uu Minimize human involvement and error whileMinimize human involvement and error while

expediting the process.expediting the process.

uu To maximize correlation between spectralTo maximize correlation between spectral

homogeneous classes and the informationalhomogeneous classes and the informational

categories (i.e., impervious area and opencategories (i.e., impervious area and openspace).space).

8/4/2019 7dj6P

24/33

Classification ApproachClassification Approach

QB multi spectral image (4-3-2) QB NDVI Image

Orthophoto (1-2-3)

8/4/2019 7dj6P

25/33

Sample DesignSample Design

Random Sampling MethodRandom Sampling Methoduu The conservative sample size equation is expressedThe conservative sample size equation is expressed

as follows (Congalton and Green, 1999):as follows (Congalton and Green, 1999):

n = B/ 4bn = B/ 4b22

Where: n =Where: n = the total samples of all classesthe total samples of all classes

B = the upper (B = the upper (/k) x 100th percentile of the Chi/k) x 100th percentile of the Chi--

squared distribution with 1 degree of freedom (squared distribution with 1 degree of freedom ())k = number of classesk = number of classes

b = Significance level =b = Significance level = +/+/-- 5%5% accuracyaccuracy

uu

To ensure unbiased sample selection.To ensure unbiased sample selection.uu To provide a statistically sound assessment ofTo provide a statistically sound assessment of

accuracy.accuracy.uu Over 500 points were used.Over 500 points were used.

8/4/2019 7dj6P

26/33

Classification SchemeClassification Scheme

Labels Rules

Land Use / Land CoverCharacter of Surface:

Impervious Areas If land area has < 25% covered with areas characterized by

vegetative open spaces then Impervious Area (1)

If land area > or = 75% characterized by impervious surfaces(e.g., asphalt, concrete, and buildings.) then Impervious Area (1)

If land area > or = 75% covered by bare land (e.g., bare rock,

gravel, silt, clay, dirt, and sand or any other earthen materials.)then Impervious Area (1)

Open Spaces Else if land area < 25% covered with areas characterized by impervious surfacesthen Open Space (2)

If land area > 75% covered with vegetation naturally existing or planted

(e.g., grass, plants, trees (leaf-on /leaf-off), forest, shrub, and scrub.)then Open Space (2)

Else Impervious Area (1)

8/4/2019 7dj6P

27/33

Accuracy AssessmentAccuracy Assessment

Map C ass Name

ID

11 Impervious area 1-40 1270.98 43.72 258 216 210 81.40 97.22 0.9429

Open space 41-100 1635.81 56.28 244 286 238 97.54 83.22 0.6734

12 Impervious area 1-45 1392.27 47.90 258 241 229 88.76 95.02 0.8976Open space 46-100 1514.53 52.10 244 261 232 95.08 88.89 0.7838

13 Impervious area 1-50 1526.54 52.52 258 261 247 95.74 94.64 0.8896

Open space 51-100 1380.25 47.48 244 241 230 94.26 95.44 0.9112

14 Impervious area 1-55 1677.57 57.71 258 285 250 96.90 87.72 0.7473

Open space 56-100 1229.22 42.29 244 217 209 85.66 96.31 0.9283

15 Impervious area 1-60 1842.09 63.37 258 306 251 97.29 82.03 0.6302

Open space 61-100 1064.70 36.63 244 196 189 77.46 96.43 0.9305

Overall Classification Accuracy = 87.65 % Overall Kappa Statistics = 0.7515

Overall Classification Accuracy = 95.02 % Overall Kappa Statistics = 0.9003

Overall Classification Accuracy = 91.43 % Overall Kappa Statistics = 0.828

Overall Classification Accuracy = 89.24 % Overall Kappa Statistics = 0.7857

Overall Classification Accuracy = 91.83 % Overall Kappa Statistics = 0.8368

Accuracy Assessment - Unsupervised Thematic Map#1(100 spectral clusters)Labeling

Criteria*

Areas

(Acres)Areas (%)

Reference

Totals

Classified

Totals

Number

Correct

Producers

Accuracy

Users

Accuracy

Kappa

Statistics

> Comparing the maps generated using five different labeling criteriashowed slight differences in overall accuracy results, with accuracy

improvement increasing toward the mid DNs of the unsupervisedQB NDVI. No major difference was found between 8 and 16 bit imagery.

> This pattern of change in classification accuracy showed thepotential correlations between increasing and decreasing of the DN

values and amounts of open space and impervious area.

8/4/2019 7dj6P

28/33

Accuracy AssessmentAccuracy Assessment

QB NDVI Thematic MapQB NDVI Thematic Map

8/4/2019 7dj6P

29/33

Geographic ModelingGeographic Modelinguu ArcView 3.3 geoprocessing was used to generateArcView 3.3 geoprocessing was used to generate

new polygons showing the relationship betweennew polygons showing the relationship between

impervious area, open space, hydrologic soil groups,impervious area, open space, hydrologic soil groups,and slopeand slope with the index valuewith the index valuess (NRCS CN or C).(NRCS CN or C).

uu The composite runoff index spatial model was usedThe composite runoff index spatial model was used

to develop spatial modeling for the runoff indexto develop spatial modeling for the runoff indexcalculation in the study area.calculation in the study area.

uu The results were compared toThe results were compared to industry standardindustry standard

values of NRCS CN and Cvalues of NRCS CN and C in order to validate thein order to validate theutility of the QB NDVI image.utility of the QB NDVI image.

8/4/2019 7dj6P

30/33

CN Results and ComparisonCN Results and Comparison

ReferenceThe City of Sioux Falls s Standard CN:The City of Sioux Falls, 2005. The Citys design standards, Chapter 11. Drainage improvements, Sioux Falls,South Dakota. Available athttp://www.siouxfalls.org/upload/documents/publicworks/designstandards/ch11.pdf .

Keyword: U30011. Accessed onFebruary 3, 2006.

Table. The Runoff Curve Number (CN) Results and Comparisons.

GIS GIS Vector Layer Descriptions Industry Standard Values of NRCS CN

ID Activity Code and Description CN (avg.) Impervious (%) (McCuen, 2005)

Hydrologic Soil Group B 5 10 100 Impervious (%)

1 11 Single family - Residential 1/8 acre (0.13 acres or 506 sq.m) 80 50 85 45 50 70 40

2 11 Single family - Residential 1/4 acre (0.25 acres 1012 sq.m) 76 41 75 45 50 70 40

3 11 Single family - Residential 1/3 acre (0.33 acres 1348 sq.m) 74 35 72 45 50 70 40

4 11 Single family - Residential 1/2 acre (0.5 acres 2023 sq.m) 74 34 70 45 50 70 40

5 11 Single family - Residential 1 acre (4047 sq.m) 68 29 68 40 45 65 30

6 11 Single family - Residential 2 acres (8094 sq.m) 76 31 65 40 45 65 30

7 31 Banks and Financial Institutions 92 82 92 88 90 93 95

8 33 Other offices 90 73 92 88 90 93 95

9 64 Warehousing, Distribution, and Wholesale 76 81 88 80 80 85 80

Total (avg.) 78 51 79 57.33 61.11 75.67 54.44

by StormFrequency, Years and Impervious Areas (%)

The Standard CN of Sioux Falls in 2005GIS Model CN Results

CN results compared to published values: NRCS/McCuen and City of Sioux Falls

8/4/2019 7dj6P

31/33

C Results and ComparisonC Results and Comparison

C results compared to the industry standard published values of C

8/4/2019 7dj6P

32/33

ConclusionsConclusionsuu The composite runoff index geographic model (Thanapura 2005/6)The composite runoff index geographic model (Thanapura 2005/6)

demonstrated that mapping techniques using high spatial resolutidemonstrated that mapping techniques using high spatial resolution satelliteon satelliteimagery, and GIS spatial modeling were successful in determiningimagery, and GIS spatial modeling were successful in determining a morea moreprecise, spatially representative runoff index (CN or C) in urbaprecise, spatially representative runoff index (CN or C) in urban watersheds.n watersheds.

uu Mapping impervious area and open space, using QuickBird NDVI satMapping impervious area and open space, using QuickBird NDVI satelliteelliteimagery generated with traditional unsupervised classification uimagery generated with traditional unsupervised classification using thesing theISODATA algorithm, is a more precise, simpler, consistent, and eISODATA algorithm, is a more precise, simpler, consistent, and efficient datafficient dataextraction approach. This is reflected in the fact that overallextraction approach. This is reflected in the fact that overall accuracy for theaccuracy for theQB NDVI thematic map produced wasQB NDVI thematic map produced was 9595%.%.

uu The finer resolution image and the mapping approach used in thisThe finer resolution image and the mapping approach used in this studystudyallowed for better discrimination in land cover/land use and morallowed for better discrimination in land cover/land use and more accuratee accuratespatially representative runoff index estimation compare to prevspatially representative runoff index estimation compare to previous studiesious studiesthat utilized medium resolution remotely sensed data ( Bondelidthat utilized medium resolution remotely sensed data ( Bondelid et al., 1981;et al., 1981;Singh, 1982; Slonecker et al., 2001).Singh, 1982; Slonecker et al., 2001).

uu Previous studies using medium resolution data demonstrated a sigPrevious studies using medium resolution data demonstrated a significantnificanttime savings in the ability to produce land cover but with accurtime savings in the ability to produce land cover but with accuracies only inacies only inthethe 7070--8080% range. That was not sufficient for urban areas.% range. That was not sufficient for urban areas.

8/4/2019 7dj6P

33/33