of 13
7/29/2019 Flood Warning Systems
1/13
NovaLynx Systems, Inc.
3235 Sunrise Blvd, Suite 3, Rancho Cordova, CA 95742
NovaLynx Systems, Inc.
NEXRAIN Corporation
9477 Greenback Lane, Suite 523A, Folsom, CA 95630
KEY ISSUES IN DEVELOPING AND MAINTAINING SUCCESSFULFLOOD WARNING SYSTEMS
THE IMPORTANCE OF HIGH QUALITY RAINFALL MONITORING INFLOOD WARNING SYSTEMS
CHOOSING A HYDROLOGIC MODEL FOR FLOOD FORECASTING
By
David C. Curtis, Ph.D.
NEXRAIN Corporation
9477 Greenback Lane
Suite 523A
Folsom, CA 95530
7/29/2019 Flood Warning Systems
2/13
NovaLynx Systems, Inc.
3235 Sunrise Blvd, Suite 3, Rancho Cordova, CA 95742
NovaLynx Systems, Inc.
NEXRAIN Corporation
9477 Greenback Lane, Suite 523A, Folsom, CA 95630
Preparing for the CostIntroduction
Flood warning systems help mitigate flood damages
through early detection of potential flooding and the
issuance of flood warnings. Upon receipt of flood warn-
ings by those individuals affected initiate response activ-
ities designed to protect life and property.
ALERT (Automated Local Evaluation in Real-Time)
flood warning systems and the related IFLOWS (Inte-
grated Flood Warning Observation and Warning Sys-
tems) are the most popular and widely used automatedflood warning systems in the U.S. today. These and
similar automated systems are now helping reduce flood
damages and save lives in more than 500 communities
throughout the U.S. Many more such systems are in
operation throughout the world.
Flood Preparedness Programs
Flood warning systems are only one component of a flood
preparedness program. A typical flood preparedness pro-
gram consists of the following components: planning,
flood threat detection and warning, flood response, flood
recovery, and review. After the flood review process, the
flood preparedness plan is revisited and updated.
Response planning and other preparedness plans are
completed well in advance of any flood activity. The
flood warning system detects a potential problem during
the early stages of the event. Once the warning is issued,
response activities begin as planned. When the water
begins to recede, tasks switch from response to recovery.
After the event, a review and evaluation team studies
community performance during the event and recom-
mends improvements to the preparedness plan.
As the time to execute the flood detection and warning
phase decreases, the time available for response increas-
es. This is a key objective of the flood warning system.
Therefore, goals of the flood threat detection and warning
component are to:
n minimize the time that it takes to detect a potential
problem,
n minimize the time to disseminate the warning, and to
n minimize the time to get populations at risk torespond.
Flood Forecast Value
The value of a flood warning system is derived from the
warning systems contribution to saving lives and prop-
erty. Implicit in any flood warning system is the flood
forecast. (i.e. the prediction of a flood threat based on
current weather conditions, rainfall amounts, and/or river
elevations etc.) Assessing the value of a flood forecast
involves two issues: accuracy and timing.
A flood forecast must be accurate to have value. A flood
forecast must also be timely to have value. Unfortunately,
it is generally not possible to simultaneously have the
most accurate forecast andthe most timely forecast. For
example, a very accurate forecast can be made by taking
a river level measurement (say 16.5 for discussion) and
forecast that the river will rise to a 16.51 in five minutes.
While this may be a very accurate forecast, it has little
value because it doesnt provide any time to respond and
mitigate damages.
On the other hand, if a flood peak of 16.51 is forecast to
occur two weeks from now, no one is likely to take it
seriously enough to engage flood response activities. In
this case, the forecast is timely but current technology
doesnt exist to provide a flood forecast on most streams
thats credible enough for people to take action that far in
advance..
To achieve added value in a flood warning system, some
investment in time, effort, and money is required. The
amount of investment should be in proper balance with
KEY ISSUES IN DEVELOPING AND MAINTAINING SUCCESSFULFLOOD WARNING SYSTEMS
Preparedness Planning
Flood Warning &Detection
ResponseRecovery
Review &Evaluation
Figure 1: The circular nature of a flood warning
By
David C. Curtis, Ph.D.
NEXRAIN Corporation
7/29/2019 Flood Warning Systems
3/13
7/29/2019 Flood Warning Systems
4/13
NovaLynx Systems, Inc.
3235 Sunrise Blvd, Suite 3, Rancho Cordova, CA 95742
NovaLynx Systems, Inc.
NEXRAIN Corporation
9477 Greenback Lane, Suite 523A, Folsom, CA 95630
some flood warning systems that trainees are frequently
overwhelmed with new information that they dont even
know what questions to ask. Follow-up training six to
nine months after the initial installation has proven to be
extremely effective and valuable.
Training on several different levels for different job
functions is required. Training for electronics technicians
on field sensors, telemetry, and computer equipment isrequired. Other users require training on central station
and computer software operation. The flood warning
system manager also requires more advanced training for
overall system configuration and management.
Training for new personnel and refresher training on an
annual basis keeps personnel up-to-date on the latest
equipment and software developments. Attendance at
group training sessions frequently brings together flood
warning system personnel from many communities that
eagerly share their experiences and ideas.
Regional and national users groups offer many opportu-nities for professional development for flood warning
system operators. Newsletters, meetings, Internet web
sites, and symposiums are excellent forums to learn and
exchange ideas. The costs of user group memberships and
conference attendance should be considered in flood
warning system training budgets and professional devel-
opment budgets.
Services and Support
A variety of services and support can be purchased from
manufacturers or third-party service providers. Thesecosts must also be recognized in the flood warning system
budget. Recurring costs for services essential to flood
warning system operation such as telephone lines and/or
data fees must also be included in the budget as necessary.
Maintenance
A car needs regular oil changes, tune-ups, and occasional
new tires to operate smoothly. In addition, car owners
have to face repairs for the dents and dings that inevitably
occur from normal use. Flood warning systems are no
different.
Flood warning systems need regular care. Batteries need
replacing, solar panels need cleaning, instruments need
calibration, and radios need tuning. In addition, accidents
occur. Hunters shoot gauges, vandals destroy antennas
and steal solar panels, and, occasionally, bulldozers run
over gauges. As unfortunate and unplanned as they are,
repairs and replacements are still required and must be
considered in operational budgets.
Computers and software evolve rapidly. Upgrades from
manufacturers are frequent and a fact of life. Users must
plan for the cost of updating these items several timesduring the lifetime of the project.
Personnel
Successful flood warning systems all have one common
element - a highly dedicated and skilled staff. Large flood
warning systems may require several full-time personnel
for operations and maintenance. Small systems may only
require one or two people assigned to the project part-
time. In any case, a commitment to adequate staffing
levels is vital to long term success.
Summary
Costs associated with flood warning systems arise from
several sources. Some sources such as capital equipment
costs are obvious. Others, such as training, system up-
dates, and staffing are equally important but often over-
looked. Consideration of these cost elements and other
that may be unique to an individual system must be
considered to provide more realistic total cost estimates
when considering flood warning system investments.
References
National Weather Service,Role of the National WeatherService, Flood Warning - Preparedness Programs, Flood Warning - Preparedness Programs, Flood Warning - Preparedness Programs, Flood Warning - Preparedness Programs, Flood Warning - Preparedness Programs, US
Army Corps of Engineers Short Course, Hydrologic
Engineering Center, Davis, CA March 21-25, 1994
Curtis, David C., Designing Rain Gauge Networks for
Automated Flood Warning Systems, Paper presented at
the Flood Plain Management Association Spring Confer-
ence, Solvang, CA, March 30-April 2, 1993
7/29/2019 Flood Warning Systems
5/13
NovaLynx Systems, Inc.
3235 Sunrise Blvd, Suite 3, Rancho Cordova, CA 95742
NovaLynx Systems, Inc.
NEXRAIN Corporation
9477 Greenback Lane, Suite 523A, Folsom, CA 95630
IntroductionConsistent rainfall data is, perhaps, the most significant
ingredient in developing accurate hydrologic analyses.
Without consistent rainfall data from storm to storm or
even within storms, accurate streamflow simulations and
forecasts are extremely difficult to achieve. Even though
most hydrologists readily acknowledge this fact, rainfall
records are rarely scrutinized to the degree necessary to
develop an engineered data set that best indexes the true
rainfall entering the watershed. This section will review
some of the factors that lead to inadvertent inconsisten-
cies in rainfall data, provide some insights into the sensi-tivity of streamflow simulations to rainfall errors, and
offer some suggestions to improve gauge and data man-
agement procedures in order to help improve data consis-
tency.
Historical Perspective
Some of the problems associated with rainfall measure-
ment have been known for hundreds of years. For exam-
ple, a demonstration performed in 1769 showed that a rain
gauge located on the top of a 30 foot tall house caught just
80% of amount measured in a ground-level rain gauge.Similarly, a gauge on top of a 150 foot abbey tower caught
just over 50% of the ground level catch. It took until the
late 19th century to fully understand that the reduced rain
gauge catch associated with height above ground was due
to the turbulent airflow around exposed gauges in strong
winds. (Frisinger, 1977)
Similar observations were reported a century later by
Symons (1881). Symons compared rain gauge catch at
various elevations with the catch at two inches above
ground level. Symons results as shown in Figure 2
represent conditions prevalent at the time of his experi-ments. He did not consider the gauge catch variation with
varying wind conditions as later researchers would.
Inadvertent Inconsistencies
Inconsistent rainfall records had been brought about by a
variety of actions, many of them well intentioned, and
most of them quite inadvertent. Rain gauges have been
moved short distances to accommodate the wishes of a
cooperative observer who wanted to raise vegetables
where the gauge stood. Or, the observer might want to put
in a walk way, or build a chicken coop, or provide space
for children to play; at other times, a rose garden might be
desired so that it wouldnt be necessary to look at that
eyesore called a rain gauge. It would be difficult to
determine how many records have lost consistency be-
cause someone wanted to beautify the area around that
dented, dirty old can with an attractive planting.
The National Weather Service occasionally contributed
to this confusion by moving gauges as vegetation altered
site characteristics - the correct procedure would have
been to keep the height of the surrounding vegetation
constant. Unfortunately, the cost and politics of such
actions were well beyond the capability of a government
agency. Another problem impairing data consistency
occurred when an observer terminated their observation
program. At such times, the gauging equipment would
frequently be moved to another property in the general
area. A location which was considered by the NWSsubstation specialist as being a compatible site. All too
often, the term compatible was used to describe any site
which had its mail delivered by the same post office.
Unfortunately, many of these records were published as
a continuing record with inadequate documentation of the
change in location.
Even the best intentions of those who were running the
rainfall data program led to a variety of inadvertent
THE IMPORTANCE OF HIGH QUALITY RAINFALL MONITORING INFLOOD WARNING SYSTEMS
Figure 2: Variation of gauge catch with height for a given set of windconditions as report by Symons (1881)
0
2
4
6
8
10
12
14
16
18
20
GaugeHeight(ft)
84%86%88%90%92%94%96%98%100%Percent of Catch at 2 " Elevation
Variation of Gauge Catch with Heightfrom Symons (1881)
By
David C. Curtis, Ph.D.
NEXRAIN Corporation
7/29/2019 Flood Warning Systems
6/13
NovaLynx Systems, Inc.
3235 Sunrise Blvd, Suite 3, Rancho Cordova, CA 95742
NovaLynx Systems, Inc.
NEXRAIN Corporation
9477 Greenback Lane, Suite 523A, Folsom, CA 95630
Figure 3: Expected gauge undercatch due to 15 M.P.H. wind
A
B
10 mph
20 mph
0
5
10
15
20
Height(ft)
0 2 4 6 8 10 12 14 16 18 20 22 24
Velocity (mph)
Expected Gauge Undercatch Due to WindLogarithmic Wind Profile
ALERT Rain Gauge15% undercatch
12% undercatch
Velocity Profile
Relative positions of two rain gauges in a wind profile
with winds of 15 mph at the height of the ALERT gauge.
15 mph
Ground ConditionsMowed Grass and
Wet Soil
changes in record through the necessary, but frequently
injurious, effort to maintain and improve the quality of
data collection. This occurred when equipment wore out
and was replaced with equipment which had an orifice at
a different elevation and/or was designed with a shape
that had different aerodynamic properties. Changes which
confused the hydrologic consistency of the data might be
as obscure as installing a platform so the observer didnt
have to stand in the mud, or as physically obvious as
installing a wind shield to improve the catch under
windy conditions. But all of these actions had one thing in
common, they altered the exposure of the gauge orifice to
the wind and in so doing modified the representativeness
of the resulting rain catch.
Many different factors that affect rain gauge records have
been identified and, to some degree, quantified. Some of
these factors are reviewed in the following sections.
Natural Variation in Rainfall
Rain gauge measurements taken by identical gauges
located a few feet apart have experienced differences as
much as 20%.
Size of the Tipping Bucket
At the end of the storm, the expected amount of unreport-
ed rainfall is one half the tipping bucket size. If all or part
of this rainfall evaporates before the next tip, this amount
of rainfall is unrecorded. In areas with 50 storms per year,
a 1 mm (0.04 in.) tipping bucket might fail to report 25
mm (0.98 in.) of rainfall over an annual cycle. Under the
same conditions, a 0.01 in. tipping bucket gauge might
fail to report 0.25 in. of rainfall during the same annual
cycle.
Tip Time
Tipping buckets miss a small amount of rainfall during
each tip of the bucket due to the bucket travel and tip time.
As rainfall intensities increase, the volumetric loss of
rainfall due to tipping tends to increase. At rainfall inten-
sities above six inches per hour, 1 mm tipping buckets will
under report rainfall in the range of 0-5% depending on
how the gauge was calibrated. Smaller tipping buckets
can have higher volumetric losses due to higher tip
frequencies.
Gauge Height
The height of the gauge above ground can have a dramatic
affect on gauge catch. As gauge height increases from
ground level, wind speed at the gauge orifice increases
due to decreasing frictional effects on the air stream
caused by the ground surface. Larson and Peck (1974)
show results that indicate wind induced undercatch is on
the order of 1% for each mile per hour of wind at the gauge
orifice. Assuming a logarithmic wind profile with height(Figure 3), a 15 mile per hour wind at the standard
ALERT gauge height of 10 feet could be expected to
induce a 15% loss compared to a 12% loss if the same
gauge were just 4 feet high.
Discrete Exposure
Gauges located in an area with variable protection rela-
tive to different wind directions will produce different
results. For example, consider Figure 4 with two gauges
located in the same general area. In the wind field shown,
gauge AAAAA
is protected by nearby vegetation and experienc-es a 10 m.p.h. hour wind. However, gauge BBBBB, located a few
feet away, may experience 20 m.p.h. winds due to more
direct exposure to the general wind field. Under these
conditions, Gauge AAAAA might experience a 10% reduction in
catch due to wind but gauge BBBBB could experience a 20%
Figure 4: Location of gauge relative to local wind field affects gaugeperformance.
7/29/2019 Flood Warning Systems
7/13
NovaLynx Systems, Inc.
3235 Sunrise Blvd, Suite 3, Rancho Cordova, CA 95742
NovaLynx Systems, Inc.
NEXRAIN Corporation
9477 Greenback Lane, Suite 523A, Folsom, CA 95630
reduction in rainfall catch due to higher wind speeds at its
location.
Time Variability
Site exposure conditions that change with time will also
affect rain gauge performance. As vegetation grows and/
or changes relative to the gauge site or as the number, size,
and shape of nearby buildings change, site aerodynamics
change over time and can produce significant changes to
the rain gauge performance. Figure 5 shows how growth
of vegetation can modify wind at the gauge and alter
precipitation catch.
Gauge Change
Each gauging system has its own unique rainfall measure-
ment characteristics. Both external and internal gauge
system attributes affect gauge performance. External
factors include site and locational characteristics. Inter-nal factors include the physical, mechanical, and electri-
cal characteristics of the gauge itself. As long as these
attributes stay the same, gauge records remain consistent.
Any errors or biases present in a consistent rainfall record
are overcome in hydrologic model calibration.
Table 1 presents some of the important systematic errors
identified in a World Meteorological Organization report
that are associated with rain gauges. Each gauge has a
specific set of systematic errors. If a gauge is changed
either for a new model or a new gauge type, these
systematic errors change, making the record inconsistent.
Gauge Calibration
Rain gauge calibration has an impact on record consisten-
cy, especially if the method of calibration changes period-
ically. For example, if a static calibration is used one time
and a dynamic calibration is used another, rainfall mea-
surement will be inconsistent.
Static calibration of an ALERT tipping bucket usually
sets the 1 mm tipping bucket to tip after the accumulation
of 72.94 grams of water ( i.e. the weight of 1 mm of water
across the area of the 12.0 in. ALERT gauge orifice).
Dynamic calibration, on the other hand, sets the bucket
to tip the correct number of times associated with a certain
rainfall rate, usually 6 in./hr. for ALERT gauges.
Unfortunately, it's impossible to have an ALERT tipping
bucket calibrated to tip at exactly 1 mm (72.94 grams) andandandandand
be calibrated for zero error at a rate of 6 in./hr. In other
words, you cant have your cake and eat it too!
Table 1: Main components of systematic error in precipitation measurement and their meteorological and instrumental factors listed in orderof importance. (Sevruk, 1982)
a: 20 mph
b: 15 mph
c: 10 mph
ab
c
Figure 5: Growth of vegetation over time significantly changes windvelocity at gauge site and alters catch characteristics.
rorrEfotnenopmoC edutingaM srotcaFlacigoloroeteM srotcaFlatnemurtsnI
noitamrofeddleifdniwoteudssoL
ecifiroeguagevoba%05-2
gnirudmireguagehttadeepsdniW
.llafniarfoerutcurtsehtdnallafniar
fohtpeddna,aeraecifiro,epahsehT
.rotcellocehtdnamireguagehthtob
sllawlanretninognittewmorfsessoL
reniatnocehtnidnarotcellocehtfo
.deitpmesitinehw
%01-2
fotnuomadna,epyt,ycneuqerF
ehtfoemitgniyrdeht,noitatipicerp
.gniytpmeeguagfoycneuqerf,eguag
fohtpeddna,aeraecifiro,epahsehT
nI.rotcellocdnaemireguagehthtob
egadna,lairetam,roloceht,noitidda
.reniatnocdnarotcellocfo
morfnoitaropaveoteudssoL
.reniatnoc%4-0
noitarutas,noitatipicerpfoepyT
folevelehttadeepsdniwdna,ticifed
lavretniemitehtgnirudmireguageht
dnanoitatipicerpfodneehtneewteb
.tnemerusaem
ehtfonoitalosidnaaeraecifiroehT
emosni,dnaroloceht,reniatnoc
ehtro,rotcellocehtfoegaeht,sesac
.lennufehtroepyt
tuohsalpsdnanihsalpS %2-1 .deepsdniwdnaytisnetnillafniaRreniatnocehtfohtpeddnaepahsehT
.noitallatsnieguagfodnikehtdna
7/29/2019 Flood Warning Systems
8/13
NovaLynx Systems, Inc.
3235 Sunrise Blvd, Suite 3, Rancho Cordova, CA 95742
NovaLynx Systems, Inc.
NEXRAIN Corporation
9477 Greenback Lane, Suite 523A, Folsom, CA 95630
The reason for this paradox is that in dynamic operation,
the tipping bucket takes a finite amount of time (e.g. on
the order of 0.5 sec) to tip. If the bucket is calibrated to tip
at exactly 1 mm, rain will still accumulate in the bucket
until the bucket moves past the midpoint of the tip and the
rain begins to accumulate in the second bucket. This
extra rain accumulating in the first bucket during the tip
is unmeasured. A bucket calibrated to tip at exactly 1 mm
will tip fewer times and under report rainfall at higher
rates.
Dynamic calibration takes the tip time into account im-
plicitly. In order to calibrate a tipping bucket to have zero
error at a rainfall rate of 6 in./hr., the bucket must be
calibrated to tip at a lower volume (e.g. 69.6 grams,
approximately). This lower volume plus the volume
associated with the tipping time will total 1 mm and the
tipping bucket will exhibit zero error at the dynamic
calibration rate of 6 in./hr.
Calibrating a tipping bucket to zero error at 6 in./hr. usinga smaller volume to initiate a bucket tip implies that the
tipping bucket might tip more frequently at lower rainfall
rates and, therefore, over report the rainfall. However, at
least in some ALERT tipping buckets, the tipping time is
slightly longer at lower rainfall rates which compensates
for the lower calibrated volume. A properly functioning
ALERT tipping bucket dynamically calibrated to zero
error at 6 in./hr. shouldnt over report at lower rainfall
rates by more than 1-2%.
Implications for Hydrologic Analysis
How systematically must the precipitation indexing ca-
pability be maintained in order to be useful in determining
runoff? Figure 6 shows the effect which an inconsistent
gauging network would produce in evaluating the storms
contribution to peak discharge. Assuming moderately dry
Figure 7: Errors in rainfall estimates produce relatively greater errorsin runoff estimates.
0
200
400
600
800
TotalRunoffError(%)
0 50 100 150 200
Total Rainfall Error (%)
Magnification of Precipitation ErrorSCS Runoff Model
8"
6"
4"
2"
NoError
Magnification of precipitationerror in runoff estimates
Figure 6: Errors in rainfall estimates can generate large errors inestimates of peak discharge
0
100
200
300
400
500
600
700
800
900
1000
PercentVariability
0 1 2 3 4 5 6 7 8 9 10 11 12Catchment Rainfall (inches)
Variablity of Peak DischargeDue to Errors in Rainfall Estimates
Small California Coastal Catchment
Moderately Dry
(Runoff using rainfall x 1.05) - (Runoff using rainfall x 0.95)Runoff using rainfall x 0.95
x 100
initial soil moisture conditions, a variety of storms of
various magnitudes were analyzed to determine the rela-
tive difference in the expected contribution to peak flow
if the precipitation was overvalued by 5% compared to
being undervalued by 5%. These relatively small changes
in precipitation indexing capability produce errors whichare inversely related to the quantity of runoff. The propor-
tional variation in peak flow can readily exceed three
hundred percent when runoff volumes are small. In a
major flood event, with high runoff volumes, the runoff
error converges very slowly toward the rainfall error. This
convergence is however, so slow that for the storm
rainfall values used to produce Figure 6, the proportional
runoff variation is just dropping below 20% for very large
storm rainfall volumes. Thus, for even major events, we
can conclude that inconsistency in evaluating precipita-
tion which is as little as 5%, can have substantial impacts
upon runoff determination.
Figure 7 again shows how rainfall errors are magnified.
When soils are nearly saturated, runoff nearly equals the
rainfall and the runoff error is just slightly larger than the
rainfall errors in large events. However, for smaller
events and dryer soils, the effect of errors is much more
pronounced. In this case, a 100% rainfall error is magni-
fied by a factor of 3.5 to 4 for a 2 inch event.
Summary
Unfortunately, there are so many factors which can influ-ence the accuracy of precipitation measurements that no
one has yet been able to devise a gauge that will consis-
tently measure true rain. The measurements which
have been judged most accurate are those which have
been observed with pit gauges. But pit gauges are ex-
tremely expensive to operate and have major constraints
which substantially limit their application to research
projects. The best that can be hoped for is that the gauging
7/29/2019 Flood Warning Systems
9/13
NovaLynx Systems, Inc.
3235 Sunrise Blvd, Suite 3, Rancho Cordova, CA 95742
NovaLynx Systems, Inc.
NEXRAIN Corporation
9477 Greenback Lane, Suite 523A, Folsom, CA 95630
equipment will operate close to the scale of reality and
with a degree of consistency which will provide a stable
index to the rainfall-runoff process. Inasmuch as it takes
many years of consistent data to define the rainfall-runoff
relationship, it is a mistake to continually modify an
operational gauge seeking a better approximation oftrue
rain.
There are two essential elements in the effective applica-tion of rainfall data to streamflow forecasting. The rain
gauge network must adequately index the precipitation
falling on a catchment and it must do so in a consistent
manner which does not alter the indexing capability with
time. The first of these elements is achieved by installing
an adequate network of appropriately sited and consis-
tently measuring precipitation gauges. Regardless of the
density of the gauge network, a second element is neces-
sary. A consistent representation of the runoff regime is
dependent upon effective systematic maintenance of
both the equipment and the gauging site. Only through
these steps can the investment in real-time data produce
the flood warning capability for which the flood warning
system was intended.
References
Alena, Thomas; et. al.,Measurement Accuracy Enhancement
of Tipping Bucket Rain Gauges at High Rainfall Rates, Fifth
International Conference on Interactive Information and
Processing Systems, AMS publication, 1989
Frisinger, H. Howard, The History of Meteorology: toThe History of Meteorology: toThe History of Meteorology: toThe History of Meteorology: toThe History of Meteorology: to
18001800180018001800, Science History Publications, AMS, 1977, pp 88-91
Larson, Lee and Eugene L. Peck,Accuracy of Precipitation
Measurements for Hydrologic Forecasting, Water Resourc-Water Resourc-Water Resourc-Water Resourc-Water Resourc-
es Researches Researches Researches Researches Research, Vol 10, No. 4, 1974
Sevruk, B., Methods of Correction for Systematic Error in
Point Precipitation Measurement for Operational Use,
Operational Hydrology Report 21Operational Hydrology Report 21Operational Hydrology Report 21Operational Hydrology Report 21Operational Hydrology Report 21, World Meteorological
Organization, Geneva, Switzerland, 1982
Symons, G. J. On the Rainfall Observations Made Upon
Yorkminster by Professor John Phillips, F.R.S, BritishBritishBritishBritishBritish
RainfallRainfallRainfallRainfallRainfall, pp 41-45, 1881
Weiss, L. L. and W. T. Wilson, Precipitation Gauge Shields,
International Union of Geodesy and GeophysicsInternational Union of Geodesy and GeophysicsInternational Union of Geodesy and GeophysicsInternational Union of Geodesy and GeophysicsInternational Union of Geodesy and Geophysics, GeneralAssembly of Toronto, 1957, Volume 1, pp 462-484
7/29/2019 Flood Warning Systems
10/13
NovaLynx Systems, Inc.
3235 Sunrise Blvd, Suite 3, Rancho Cordova, CA 95742
NovaLynx Systems, Inc.
NEXRAIN Corporation
9477 Greenback Lane, Suite 523A, Folsom, CA 95630
IntroductionThe most common application programs in automated
flood warning systems are the runoff and river forecast
programs. These programs use observed and, in some
cases, forecast rainfall amounts to compute the amount of
water that will enter the stream system.
Forecast Models
The purpose of a forecast model is to estimate future river
flows and elevations based on observed or forecast amounts
of rainfall. In flash flood situations, certain portions of the
forecast hydrograph are more important than others.Accurate forecasts of the rising limb, the time to hydro-
graph peak, and the magnitude of the peak are critical.
These are the elements of model output that have the most
impact on the flood warning. The model implemented in
a flood warning system must consistently perform well in
these three areas.
Before model selection, one very important element,
rainfall estimation, must be considered. The volume of
water under the rising limb of a flash flood hydrograph is
primarily surface runoff. Basins with short response
times are often characterized by low infiltration rates andsteep slopes which efficiently generate runoff. Because
these basins efficiently generate runoff, especially during
periods of high intensity rainfall, the volume of runoff is
very sensitive to the volume of rainfall. This implies that
the output of a flash flood forecast model will also be very
sensitive to the rainfall inputs.
Flash flood forecast sensitivity to rainfall inputs serves to
emphasize the importance of establishing a good mea-
surement system first. The phrase commonly heard in the
computer industry, Garbage in. Garbage out. is equally
applicable to flash flood forecasting. Good model perfor-mance, no matter what model is used, cannot be expected
without a good measurement system. The implication for
forecast system design is to invest in the measurement
and detection systems first, then consider hydrologic
models.
There are many different hydrologic forecast models in
use. The most commonly used models in local flood
warning systems fall into two categories: simple index-
type models and conceptual rainfall-runoff models. In-dex models keep a running index that reflects current
moisture conditions. The moisture index, a time of
year index, current rainfall, and rainfall duration is
generally all thats needed to estimate surface runoff
with these models. Conceptual models attempt to
provide a more physically-based approach to basin
modeling by more explicitly accounting for
evapotranspiration, interception storage, retention stor-
age, infiltration, surface runoff, percolation, interflow,
etc.
Table 2 shows the most widely available models forlocal flood warning systems.
CHOOSING A HYDROLOGIC MODEL FOR FLOOD FORECASTING
API ModelAPI ModelAPI ModelAPI ModelAPI Model The API (Antecedent Precipitation Index)
Model was developed by the National Weather Service
and has been used in various forms since the 1950s. The
antecedent precipitation index reflects the current soil
moisture based on recent rainfall. A high index means
high soil moisture content while a low index indicates dry
conditions. The API for a given period is used with a
rainfall-runoff relationship, the rainfall amount, and the
storm duration to estimate runoff. A unit hydrograph is
applied to distribute the runoff. At each computational
period, the index is updated based on the additional
rainfall and by a seasonally dependent factor. The season-
ally dependent factor empirically accounts for changes inthe rainfall-runoff relationship due to seasonal changes in
evapotranspiration, infiltration, etc.
Complex basins can be modeled by applying the API
technique to individual sub-basins that are hydrologically
homogeneous. Outflows from sub-basins can be routed
downstream and combined with other tributary flows and
inflows calculated by the API model for local areas.
Table 2: Flood Forecast Models
Index Models Conceptual ModelsAPI Sacramento Soil Moisture
ADVIS HEC1-F
Flood Advisory Tables SSARR
By
David C. Curtis, Ph.D.
NEXRAIN Corporation
7/29/2019 Flood Warning Systems
11/13
NovaLynx Systems, Inc.
3235 Sunrise Blvd, Suite 3, Rancho Cordova, CA 95742
NovaLynx Systems, Inc.
NEXRAIN Corporation
9477 Greenback Lane, Suite 523A, Folsom, CA 95630
Many versions of the API model exist. Most National
Weather Service River Forecast Centers that use API
have added modifications to customize the technique
for conditions in basins within their area of responsibility.
At least eight different implementations of API are used
by the National Weather Service.
The API model is simple and relatively easy to under-
stand. It is also relatively easy to adjust. Forecasters can
easily change model parameters or model runoff based on
his or her assessment of the current event to improve
model performance.
ADVISADVISADVISADVISADVIS The ADVIS program (Sweeny, 1988), devel-
oped by the National Weather Service for local flood
warning, includes an API model as its primary hydrolog-
ic forecast technique. (All National Weather Service
implementations of API are available in ADVIS) ADVIS
is a simplified implementation of hydrologic modeling
that produces output appropriate for the local user de-
pending upon what type of information is available. For
example, ADVIS output includes:
Categorical forecasts for ungaged
watersheds. Categorical forecasts are general forecasts of
minor, moderate, or severe flooding based on the
antecedent precipitation index and rainfall estimates.
Crest stage forecast. ADVIS will generate a
crest forecast if the unit hydrograph peak is available.
Forecast hydrograph. Where the complete unit
hydrograph is available, ADVIS generates a complete
forecast hydrograph.
The ADVIS program is intended to address relatively
simple hydrologic situations at the local level.
Flood Advisory TablesFlood Advisory TablesFlood Advisory TablesFlood Advisory TablesFlood Advisory Tables Flood advisory tables are used
to provide a quick estimate of peak stage forecasts usingindices produced by the API or other modelling tech-
niques. The tables are computed in advance for a variety
of antecedent conditions. The current index can be com-
puted on-site or provided by a local National Weather
Service office. Local users apply the current index with
the latest rainfall estimate to the table to determine the
estimated peak stage. An estimated time to peak is usually
available based on previous analysis of basin response
times.
Sacramento Soil Moisture Accounting ModelSacramento Soil Moisture Accounting ModelSacramento Soil Moisture Accounting ModelSacramento Soil Moisture Accounting ModelSacramento Soil Moisture Accounting Model
The Sacramento Soil Moisture Accounting Model is aconceptual model designed as a comprehensive represen-
tation of the hydrologic processes of the upper soil man-
tel. Evapotranspiration, direct runoff from impervious
areas, surface runoff, percolation, interflow, and two
types of base flow are explicitly represented. Runoff
calculated for each period is distributed using a unit
hydrograph.
Figure 8: Sacramento Soil Moisture Accounting Model
Evapotranspirat ionDem and Precipitation
E T
E T
E T
E T
E T
Pervio us Im p e rvio us D ire c t Runo ff
Sur face Runof f
Interflow
Tens ionWater
FreeWater
Excess
Percolat ion
1 -PFre e PFre e
TotalC h an n e l
Inflow
DistributionFunction
Stream Flow
SubsurfaceDischarge
SideTotal Base
Flow
SupplementalBase Flow
PrimaryBase Flow
TensionWater
FreeP
FreeS
Reserve
Sacramento Soil Moisture Accounting Model
7/29/2019 Flood Warning Systems
12/13
NovaLynx Systems, Inc.
3235 Sunrise Blvd, Suite 3, Rancho Cordova, CA 95742
NovaLynx Systems, Inc.
NEXRAIN Corporation
9477 Greenback Lane, Suite 523A, Folsom, CA 95630
Each hydrologic process is represented by a function orseries of functions with adjustable parameters. The model
is calibrated with historical rainfall and streamflow data
by adjusting parameters until the model output adequate-
ly represents basin response. The model is applied to
individual basins that are hydrologically homogeneous.
Complex basins are modeled by combining outflows
from individual basins using a variety of available routing
techniques.
HEC1-FHEC1-FHEC1-FHEC1-FHEC1-F The Hydrologic Engineering Center (HEC)
has developed a forecasting system for COE offices that
is also available for local flood warning systems. Theforecast technique uses an initial and uniform loss rate to
compute runoff which is applied to a unit hydrograph to
produce a basin forecast. Results from each basin can be
combined and routed to develop forecasts for complex
systems. HEC1-F uses observed streamflows to set prop-
er loss rate parameters.
HEC1-F can be calibrated relatively easily. Most of the
necessary parameters can be easily obtained from maps.
Infiltration parameters and certain characteristics of the
unit hydrograph can be estimated initially. During a flood
event, HEC1-F evaluates model performance against
observed stream flow and automatically adjusts the ap-
propriate parameters.
HEC1-F is the forecast version of HEC-1, a widely used
hydrologic design tool. Many different public and private
organizations throughout the United States have usedHEC1 to generate flood hydrographs for a variety of
purposes from bridge design to flood plain mapping. As
a result, many local engineers understand the model and
the transition to HEC1-F is relatively easy.
SSARRSSARRSSARRSSARRSSARR The Synthesized Streamflow and Reservoir Reg-
ulation (SSARR) model was developed jointly by the
National Weather Service and the COE. It is a tool used
by the respective agencies in the Pacific Northwest for
flood forecasting and reservoir regulation. The SSARR
model provides a continuous accounting of soil moisture
to determine how much of the incident rainfall andsnowmelt will become runoff. Three phases of runoff are
computed: direct runoff, interflow, and baseflow. Each
phase is routed through a series or cascade of linear
reservoirs to produce the total streamflow.
Hydrologic Model Selection
Choosing the appropriate hydrologic model is a task
open to much debate. A widely cited study by the World
Meteorological Organization indicated that the API tech-
nique, the Sacramento model, and the SSARR model all
gave about the same results in humid climates. However,explicit soil moisture accounting models like SSARR and
the Sacramento model were clearly superior to the API
model for arid and semi-arid climates. In humid environ-
ments, soil moisture conditions are less variable than in
arid or semi-arid climates. The added complexity of the
explicit soil moisture accounting models to handle wide
ranging conditions doesnt contribute significantly to
model performance when conditions are relatively stable.
However, when conditions are rapidly changing, some
researchers have found that explicit soil moisture ac-
counting models offer a significant performance advan-
tage.
When reviewing studies comparing the complex explicit
soil moisture accounting models with simpler index ap-
proaches, an important insight was noted. While the
simpler models performed well, statistically, compared
to the explicit soil moisture accounting models, signifi-
cant deviations occurred at key points. These deviations,
while significant, were rare and tended to have little effect
on the overall statistics. However, the deviations were
Figure 9: The SSARR Model
AreaElevation Prec ip ita tion Tem pe rature
Snow
Accum ulaton
Snowm eltRainfall
MoistureInput
Soi lMoisture
Runoff
BaseFlow
DirectRunoff
Su bSurface
Surface
SM I
BI I
S-SS
Routing
Rou ting Ro uting
Evaporat ion
Stream flow
SSARR Mo del
7/29/2019 Flood Warning Systems
13/13
NovaLynx Systems, Inc.
NovaLynx Systems, Inc.
NEXRAIN Corporation
References
Burnash, Robert J. C., Ferral, R. L., and McGuire, Rich-
ard L., A Generalized Streamflow Simulation System,
National Weather Service, California Department of Water
Resources, Sacramento, California, 1973
Dotson, Harry W., John C. Peters, Hydrologic Aspects
of Flood Warning - Preparedness Programs, Technical
Paper No. 131, U.S. Army Corps of Engineers Hydrolog-
ic Engineering Center, Davis, CA, August 1990
Federal Interagency Advisory Committee, Guidelines on
Community Local Flood Warning and Response Sys-
tems, National Technical Information Service, Spring-
field, VA, August 1985
Kitanidis, P. R., and R. L. Bras, Real-Time Forecasting
with a Conceptual Hydrologic Model, 2: Application and
Results, Water Resources Research, Volume 16, No. 6,1980, pp. 1034-1044
Linsley, Ray K., Max Kohler, and Joseph L. H. Paulhus,
Hydrology for Engineers, Third Edition, McGraw-Hill,
1982
Nemec, J., Design and Operation of Forecasting Opera-
tional Real-Time Hydrological Systems (FORTH), World
Meteorological Organization, Geneva, Switzerland, un-
published manuscript, Figure 12, 1984
Pabst, Art, State-of-the-Art Flood Forecasting Technol-
ogy, Proceedings of a Seminar on Local Flood Warning
- Response Systems, US Army Corps of Engineers, 10-12
December, 1986
Sittner, Walter T., Catchment Response in the Computer
Age, Computer Applications in Water Resources, Harry
C. Toro, ed., ASCE, 1985, pp. 749-757
Sittner, W.T., C.E. Schauss, and John Monroe, Contin-
uous Hydrograph Synthesis with an API-Type Hydro-
graph Model, Water Resources Research, Vol. 5, No. 5,
1969, pp. 1007-1022
Sorooshian, Soroosh,Identifiability in Conceptual Rain-
fall-Runoff Models, Computerized Decision Support
Systems for Water Managers, Proceedings of the 3rdWater Resources Operations Management Workshop,
ASCE, 1988, pp 172-183
Sweeny, Timothy L., Flash Flood Hydrologic Forecast
Model, ADVIS, Computerized Decision Support Sys-
tems for Water Managers, Proceedings of the 3rd Water
Resources Operations Management Workshop, ASCE,
1988, pp. 683-692
U.S. Army Corps of Engineers, General Guidelines for
Comprehensive Flood Warning/Preparedness Studies,
Hydrologic Engineering Center, Davis, CA, October 1988
U.S. Army Corps of Engineers Hydrologic Engineering
Center, Floodway Determination Using Computer Pro-
gram HEC-2, Training Document No. 5, January 1988
U.S. Army Corps of Engineers, Water Control Software:
Forecast and Operations, Hydrologic Engineering Cen-
ter, Davis, California, December 1989
U.S. Army Corps of Engineers, SSARR Users Manual,
North Pacific Division, Portland, Oregon, May 1991
World Meteorological Organization, Intercomparison
of Conceptual Models Used in Operational HydrologicForecasting, Operational Hydrology Report No. 7, Annex
III, p 47, Geneva, Switzerland, 1975
frequently observed when extreme hydrologic conditions
existed. The complex models could manage the extremes
where the simpler approaches were not capable. These
rare events are precisely the events that offer the greatest
potential for hazard mitigation.
The choice of models in specific situations remains
difficult. After all the analysis of which model performs
the best for a given basin, it ultimately depends upon thecapabilities and resources of local users. Complex models
requiring a high level of support might be appropriate in
cases where local skills and resources can handle it.
However, the same model may be entirely inappropriate
in situations with lower levels of local hydrologic skill
and resources.
To summarize model selection:
Choose a model that is within the capabilities of the
local user to understand, operate, and maintain,
Choose a model that is appropriate for the localhydrologic regime, and
Choose a model that will provide the best estimate of
the rising limb, the time to peak, and the flood peak.