At the HIC/DOH conference in August 2000, a concept for a rudimentary
WHFS site-specific river forecast model was presented, based on
a suggestion from Lee Larson. MBRFC offered to assist with the development
of this concept and is working with Russ Erb to create the application.
Lee's primary objective was to create a simple model which could
generate forecasts for significant short-fused rainfall events.
It would replace headwater tables and possibly other local site-specific
tools now used by many WFOs. He also wanted to develop an algorithm
which would generate a complete forecast hydrograph, not just a
crest stage forecast . It also needed to use data, parametric information,
and functions which were either already available within the current
scope of AWIPS (particularly WHFS and the IHFS_DB arenas) or could
be easily created.
When Lee studied the parameters and data required to create and
use a headwater table, he produced a simple relationship that could
be used to calculate storm total hydrograph as a function of flood
discharge (from flood stage and rating curve), peak unit-hydrograph
discharge, the mean areal precipitation (MAP) over the headwater
area, and the current headwater guidance value (FFH). All of these
parameters and data are already in the WFO AWIPS hydro system or
can easily be installed.
For each site, a "flood" hydrograph is determined by multiplying
the ordinates of the unit-hydrograph by a factor of flood discharge
divided by the peak unit-hydrograph discharge. For an event, the
ordinates of this hydrograph are multiplied by a factor of the total
MAP divided by the appropriate FFH value. The resultant discharge
time-series can then be processed with existing AWIPS applications
to generate a forecast stage hydrograph and generate a product if
This concept has a few limitations and restrictions, most notably
the use of a total event MAP value. This approach assumes that there
is a linear relationship between MAP and runoff for a given FFH
value, which is not the case. The degree of non-linearity is affected
by the initial soil moisture conditions and how easily the site
is flooded, as reflected in the magnitude of threshold runoff. Although
these variables might be estimated for each site, it is difficult
to determine the degree of non-linearity for various conditions.
It also diminishes the flexibility of the model to account for significant
temporal distribution effects. As the duration increases, the potential
for distribution errors increases significantly. It does not account
for "multiple" events, cases where several hours of significant
rainfall occurs, followed by a short break and another round of
significant rainfall, all within the same FFH valid period. This
is a problem often encountered using headwater tables.
Additionally, the duration of the MAP and FFH values must always
be consistent. Since the MAP value used in Lee's equation is an
event "storm total", and FFH guidance is normally issued
as 1, 3, and 6 hour values, the MAP value used must likewise be
"fixed" at 1, 3, or 6 hours . Events with durations other
than these will not be processed as well as we would like.
A further issue is that the duration of the unit-hydrograph must
also be consistent with that of the FFH and MAP values used. This
means that multiple unit-hydrographs would have to be defined for
each site, increasing the development, installation, and support
Proposed Model Concept
In an attempt to offset the above-mentioned limitations and philosophical
considerations, the approach we are taking bases all calculations
on an hourly time-step. This requires only a single 1-hour unit-hydrograph
per site. This significantly reduces development time and minimizes
installation and support requirements. More importantly, it can
be applied to events with a variety of durations and precipitation
patterns without requiring any special decisions by the user.
The parametric information required for each site is...
* Basin boundary definition for the headwater site
* Flood stage
* Rating curve
* 1-hour unit-hydrograph
The data required for each event are...
* Current headwater guidance value
* Mean areal precipitation time-series for the headwater area
The model will consist of 3 modules or functions. The first will
retrieve parametric information and data for selected site and determine
the MAP time-series to be applied. The second will calculate the
discharge time-series for the event. The third will display the
stage hydrograph derived from the discharge time-series and provide
the user with tools to adjust the hydrograph and generate desired
Current constraints with WHFS and IHFS_DB limit this initial version
of the model to a single basin boundary per headwater site. For
now, locations which may actually have multiple zones contributing
to the site will have to use a composite total basin boundary definition.
It is recognized that this limitation adversely affects spatial
and temporal distribution of the runoff which degrades the performance
of the model to a degree. Later versions should take advantage of
more detailed basin boundary definitions.
Precipitation Processing Function
1. The user will select a headwater location from an option menu
of valid location Ids or by clicking on the site in a WHFS display.
2. The flood stage, rating curve, 1-hour unit-hydrograph and most
recent 1-hour headwater guidance (FFH) value will be retrieved from
the IHFS_DB for that location. If any of these data are missing
it will not let the user continue with this location.
3. The flood discharge will be calculated from the rating curve
using the flood stage.
4. The 1-hour threshold runoff will be calculated by dividing the
flood discharge by the peak unit-hydrograph discharge.
5. The user will press a "Calculate MAPs" pushbutton which
will then calculate a 1-hour MAP time-series on the fly (there is
no MAP preprocessor like NWSRFS) from the gridded Stage II data
in the IHFS_DB. The default duration of the time-series will begin
at the starting hour of the valid period of the FFH being applied
and end at the top of the current hour. The time-series will be
displayed for the user to view and/or edit. Individual hourly MAPs
may be adjusted, a multiplier may be applied to the entire time-series,
or user-specified hourly values may be added to extend the time-series
(such as adding a few periods of QPF).
6. The user will then be able to press a pushbutton to run the "Discharge
Time-series Generation Function". The threshold runoff value,
1-hour unit-hydrograph, 1-hour FFH value, and 1-hour MAP time-series
will be passed to the discharge time-series generation function
Discharge Time-series Generation Function
1. Based on the FFH and threshold runoff values, a "soil moisture
indicator" (SMI) will be calculated using rainfall-runoff equations
from the MBRFC API model.
2. Beginning with the first hour of the MAP data, hourly accumulated
MAP will be computed. Applying the SMI value using the equations
from step 1, hourly accumulated runoff will be determined, and incremental
hourly runoff values can be calculated.
3. The discharge time-series will be calculated using incremental
runoff and the 1-hour unit-hydrograph ordinates and returned to
the main program.
Hydrograph Analysis Function
1. The 1-hour discharge time-series will be displayed as a 1-hour
forecast stage hydrograph
2. The user will be able to adjust the forecast hydrograph. See
discussion later in this report about potential features in this
3. The forecast stage hydrograph will be saved in the IHFS_DB as
a forecast time-series along with the input parameters which produced
this output. Existing WHFS tools will then be used to create and
transmit any desired product.
Obviously there will be requirements for user options to adjust
or modify the discharge time-series computed by the model. These
adjustments may be in the form of direct changes to the time-series,
input data modifications, or returning to the precipitation processing
function for re-analysis of the MAP data.
Since this is an "event-oriented" model, there are certainly
issues surrounding the ability to account for initial flow conditions
at the start of the discharge time-series computation. How to manage
"baseflow" and hydrograph "continuity" from
one FFH valid time period to the next are foremost questions.
Potential adjustments are a user-defined baseflow definition and/or
point-and-click time-series changes similar to IFP hydrograph adjustment.
Options to specify the date/time of the FFH value to be used or
modifying the FFH value itself may be useful. Beta testing will
confirm a few initial ideas, but continued monitoring for enhanced
requirements after operational implementation will most certainly
be required. Current Project Status
Russ is creating the first and third modules and we are formulating
the middle section. Russ has made great progress with his 2 components
and we have completed the testing of the hydrograph generation equations
and validated that they yield very good results. Not only does this
approach provide a better "crest" forecast than headwater
table use, but the entire hydrograph can be used to determine other
attributes of the event that may not be available from current methods.
Information such as times to reach or fall below flood stage and
duration of flood, can be easily determined.
The intent is to complete the coupling of all 3 modules within the
next few weeks and install a test version on the WFO system here
in Pleasant Hill. We can then test both the physical functionality
as well as the hydrograph generation accuracy for future events
in their HSA. We can also run tests for any case studies in other
RFC's if we are provided the appropriate site parameters and event
data. More about that later.
Keep in mind that this model is not intended to produce forecast
hydrographs with the accuracy or completeness of a full-scale model
like NWSRFS. It is intended as a local PRELIMINARY, assessment for
significant events while keeping installation and support requirements
at an acceptable level. It definitely performs better than the headwater
table approach, and if it will provide at least similar results
as other site-specific models used around the country, it is a step
in the right direction. We realize that not all RFC's provide site-specific
guidance to their WFO's and that not all WFO's have a site-specific
model need. However, if you are interested in helping shape this
initial site-specific approach, please feel free to comment.