sacramento flood Risk: GIS Runoff Model

 

By: Chris Webb

What: Term Paper

For: CEE 6440 – GIS in Water Resources

Date:   December 7, 2007

 

 

 

 

 

1.0              Introduction

1.1       Project Abstract

            The goal of this project was to create floodplain maps for a forecasted storm event for Sacramento, CA.  This project had two major components.  The first step was to determine the runoff generated by a given storm.  Second, that runoff was routed through Folsom Dam and a corresponding floodplain map produced.  This paper focuses on the first step in the process.  A model was created in ArcMap to determine the runoff produced by forecasted precipitation and temperature.

1.2       Area History

            Folsom Dam is located about 25 miles upstream of Sacramento, CA.  This dam stores water from the American River and its watershed.  Folsom Dam was created because of the Flood Control Act of 1944.  The dam was built by the U.S. Army Corps of Engineers and finished in 1956.  The original design was supposed to be for the 500-year flood event.

            Mother Nature had other things in mind.  In 1951 shortly after construction on the dam had begun, a record flood occurred.  In 1956 near completion of the dam another record flood happened.  Engineers had said that it would take about one year to fill the new reservoir.  This record flood filled it in one week.  Eight years later, in 1964, yet another record flood event hit the American River.  Folsom Dam was reevaluated and determined to be designed for the 120-year storm as opposed to the 500-year storm as originally planned.  Record floods in 1986 and 1997 caused the actual design of the dam to be decreased to a 60-year storm protection.  In the past few years improvements have been made to the existing levees to increase the flood protection downstream of Folsom Dam in Sacramento to about the 100-year storm. 

            The Sacramento Area is a high risk area for flooding similar to New Orleans.  Much of the city is protected by levees.  Knowing what quantity of water will be coming from a storm event will allow Folsom Dam to be operated in way that will minimize flooding and property damage.

 

 

 

 

2.0              Objectives

            The objective of this project is to predict the runoff for a given watershed based on forecasted precipitation and temperature.  Specifically, this project will develop a model to predict runoff in the American River Basin.  This objective has three major components: collecting GIS data, building a base map in ArcMap, and developing a runoff model.

 

3.0              Project Description

3.1       Data Collection

            To achieve the objectives stated above several datasets needed to be collected.  To build a base map in ArcMap watershed boundary, flow line, and water body files were needed. The runoff model would need stream flow, precipitation, temperature, and elevation data.  Before data collection could begin the temporal availability of the various data sets had to be checked.  The temporal stream flow, temperature, and precipitation data were all available for the study area from January 1980 to February 1986.

            Datasets for the base map were obtained from the National Hydrography Dataset Plus (NHDPlus, 2007).  More spatial reference data such as roads, cities, and counties are easily obtainable from United States Geologic Survey (USGS) seamless dataset (2007a).  This data was not critical for this project.  Daily mean stream flow data was downloaded from the USGS National Water Information System (NWIS) (2007b).  Daily precipitation and average temperature were collected from DAYMET (2007).  Lastly, a 30 meter digital elevation model (DEM) was found from the National Elevation Dataset (USGS, 2007a). 

3.2       Base Map

            The watershed boundaries, flow lines, and water bodies were first added to ArcMap.  The American River North Fork and American River South watersheds were selected and exported as a separate layer in ArcMap.  The flow lines and water bodies had to be clipped to the size of the new watershed layer.  This was done by selection by intersection and exporting each layer as a new layer in ArcMap.  Next the DEM was imported and trimmed to the size of the watershed using the extract by mask command in ArcMap.  Figure 1 shows a screen shot of the base map.

Figure 1 – ArcMap Base Map

3.3       Runoff Model

(1)
 
            The next task was to develop a runoff model.  Equation 1 shows a basic runoff model that was assumed:

Runoff = Precipitation – ET

Infiltration was not accounted for in this model.  This assumes a worst case scenario that the ground is saturated and the slopes are steep.  A secondary model was evaluated that also ignored the effects of ET.  The results of the two models will be compared to determine which model best fits the study area.

            The input data of precipitation and ET for the model was found for a set of 13 artificial weather stations.  These artificial weather stations were randomly selected and the associated geographic coordinates found for each.  The weather stations were then plotted in ArcMap on top of the DEM and elevations for each station assigned.  An input spreadsheet was created with the location and elevation data for each weather station.  The input spreadsheet is shown in Figure 2.

Figure 2 – Input Spreadsheet

 

The precipitation and temperature data collected from DAYMET was found for each station location for the date of concern and entered into the green columns.  The blue column shows the ET, which is calculated based on the location of the weather station, precipitation, and temperature.  The ET calculations in this model are based on Shuttleworth’s equation (Stannard, 1993).

            Finding the runoff associated with storm represented in the input spreadsheet consists of 6 steps.  The first step is to add the completed input spreadsheet to the base map in ArcMap (see Figure 3). 

Figure 3 – Screen Capture of Adding Spreadsheets to ArcMap

 

Next the Add XY Data Tool is used to add the weather stations to the map display with the stations associated precipitation and temperature data (see Figure 4).

Figure 4 – Screen Capture of the Add XY Point Command in ArcMap

Step three is to interpolate to raster the precipitation and temperature data.  Fourth, the rasters have to be trimmed to the size of the watershed using the extract by mask command in ArcMap (see Figures 5 & 6).

Figure 5 – Precipitation Raster from ArcMap

 

Figure 6 – ET Raster from ArcMap

 

Fifth, the raster calculator is used to perform Equation 1 and then convert the raster values to units of flow in cubic feet per second by a multiplier (see Figure 7).

Figure 7 – Flow Raster from ArcMap

 

Lastly, zonal statistics are used to sum the runoff from each cell of the raster to produce a total runoff value (see Figure 8).

Figure 8 – Output Table from Zonal Statistics in ArcMap

 

            This process can be automated by using the model builder within ArcMap.  Figure 9 shows a screen capture of the model builder for the runoff model.  This model will be referred to as Runoff Model 1.  A similar model was built that assumed the ET to be zero.  This secondary model will be referred to as Runoff Model 2.  Figure 10 shows Runoff Model 2.

Figure 9 – Model Builder of Runoff Model 1

 

Figure 10 - Model Builder of Runoff Model 2

 

4.0              Results

            The model was calibrated using a historical rainfall event on October 28-29, 1981.  Stream flow data from NWIS showed that the flow in the American River changed from 674 cfs on October 27th to 2286 cfs on October 29th.  This produced a runoff of 1612 cfs.  Runoff Model 1 that accounted for ET predicted the runoff for the two day storm event to be 1392 cfs, an error of 13.2 %.  Runoff Model 2 that suggests all precipitation goes to runoff predicted a value of 1561 cfs, an error of only 3.2 %. 

 

5.0              Summary

            The results of Runoff Models 1 and 2 suggest that the ET and infiltration are both negligible for the study area.  The model has the ability to estimate runoff from storms longer than one day by simply running the model for each day of precipitation.  The objectives of this project were met.  This model would allow someone working at a reservoir to input a forecast of precipitation and temperature and receive an output of expected runoff that has reasonable accuracy.

            The model has a few short comings.  The biggest being that this model does not account for rain on snow and how that contributes to runoff.  Major flooding in watersheds with winter snow pack usually occurs during the spring runoff.  For this model to be an effective flood forecasting tool the ability to calculate snowmelt would have to be added into the model.  Another shortcoming of the model is the high risk of user error.  The model could be modified and automated through the use of programming to eliminate some user introduced error.

            In summary this project was successful in meeting the objectives set forth at the beginning, but as always there is room for improvement that could expand the application of the model.

References

DAYMET (accessed 2007, December 3). Daily Surface Weather Data. http://www.daymet.org.

 

NHDPlus (National Hydrography Dataset Plus) (Accessed 2007, December 1).  http://www.horizon-systems.com/nhdplus.

 

Stannard, D. I. (1993). Comparison of Penman-Monteith, Shuttleworth-Wallace, and modified Priestley-Taylor evapotranspiration models for wildland vegetation in semiarid rangeland. Water Resources Research, Volume 29, Issue 5, p. 1379-1392

USGS (United States Geological Survey) (Accessed 2007, December 1)a. The National Map Seamless Server. http://seamless.usgs.gov/website/seamless/viewer.htm

 

USGS (United States Geological Survey) (Accessed 2007, December 1)b. Surface Water Data. http://waterdata.usgs.gov/nwis/sw.