Final Project
CEE6440 GIS in Water Resources, Fall 2010
Streamflow Forecasting for Environmental Purpose at Bear
River Watershed
By
Omar Alminagorta
I.
Introduction
Wetlands
are sensitive to changes in the hydrology. Changes in the precipitation (liquid
or solid) can impact in water supply of wetlands and consequently affect the
vegetation and habitat of species in these ecosystems. Wetland Managers has the
necessity to know better understanding about the forecasting flow to improve
wetland management plan, especially for the summer months, which are high
competence with other water users (agriculture, industrial, urban) (Olson 2008).
Climatic
and stream gage stations collect information at specific points in watersheds
of the US. This information at specific points can be converted to surface data
using the spatial interpolation of GIS. Surface data allow us to analyze
climatic variables and surface flow grid by grid.
Previous GIS model (Knight et al. 2001) replicate monthly runoff from climate condition. Knight (2001) shows that a simplified GIS approach can be useful to find the relationship between climatic variables and runoff.
The
present project has the objective of forecast the quantity of surface
flow based on spatial interpolation of precipitation and soil moisture. Forecasting
Analysis is accomplished at seven points at the Bear River Watershed. These
points were selected considering water supply to three main wetlands in the
Bear Watershed (Figure 1) : (1)
Cokeville Meadows National Wildlife Refuge, (2) Bear
Lake National Wildlife Refuge and (3) Bear River Migratory Bear Refuge.
Fig.1 Bear River Watershed
II.
Methodology
Methodology was developing in three main steps: data collection, spatial analysis and simple regression.
2.1 Data Collection
Table
1 shows the type of data collected and their respective source. Fifteen weather
stations (SNOTEL) and seven stream gages station distributed at the Bear River
Watershed were selected and evaluated. Details of each station are presented in
Appendix 1.
Table 1 Data Collection and Source
|
Data |
Source |
|
ü Bear River Watershed Boundary |
Bear
River Information System(1) |
|
ü USGS 12 Digit HUC Boundaries |
|
|
ü USGS 8 Digit HUC Boundaries |
|
|
ü 30 Meter National Elevation Dataset |
|
|
ü National Hydrography Dataset |
|
|
ü Precipitation Accumulation (in) |
Natural
Resource Conservation System(2) |
|
ü Soil Moisture Percent |
|
|
ü Stream gages flow (cfs) |
U.S.
Geological Survey(3) |
Figure
2 shows the location of 15 SNOTEL stations and 7 stream gages with their
respective drainage area. Precipitation (liquid and solid) and soil moisture percent
were collected from the SNOTEL station, while monthly flow was collected for
stream gages stations. A monthly analysis was performed. Average Monthly
of three years was considered (2007 to 2009) to accomplish the present study.
Fig 2 Location of Climatic and Streamflow gages stations as
well as their respective drainage area
Considering that each weather station
provide different period of time of data collection and different data
information, it was necessary identify incomplete, incorrect or inaccurate
data. For example in the stream gage station 10028500, average data from
previous months and near stations were considered to complete missing
information.
2.2 Spatial Analysis
Weather information is presented as punctual data for each weather stations distributed at Bear River Watershed. Thus, I used the spatial interpolation to obtain the surface data of precipitation and soil moisture for each month.
Once
the excel data is input to the GIS, there are some intermediate steps to get the
statistics of the spatial analysis. Table 2 shows series of tools to accomplish
the spatial analysis.
Table 2. Tools used in the Spatial Interpolation
|
N |
Tool |
Description |
|
1 |
Make Feature Layer |
Select the specific month to analyze |
|
2 |
Save to Layer File |
Create a layer from selected month |
|
3 |
IDW |
Implement to inverse distance weighted interpolation |
|
4 |
Int |
Change from float values to Integer values (to change from
Raster to Polygons) |
|
5 |
Extract by Mask |
Get the Mask for specific site analyzed |
|
6 |
Raster to Polygon |
Convert from raster to polygon |
|
7 |
Zonal Statistics as Table |
Get statistical values |
Considering
that all steps presented in Table 2 need to be implemented by each month (12)
and by each variable (precipitation and soil moisture), I used the Model
builder tool of GIS to automate these Arc tools. Figure 3 presents the
elaboration of the model.
Fig.
3 Application of model builder to spatial analysis
2.3 Simple Regression:
The
main idea was to find a simple relationship among spatial interpolation of
precipitation, soil moisture; also consider the site location and the variation
by month to predict the streamflow. Simple Linear Regression was accomplished
to correlate these
variables.
Equation 1 was used to
forecast the mean flow for each month. This equation shows that stream flow at
time “t” can be estimated using precipitation, soil moisture, and streamflow at
time “t-1”. Also location and specific month are considered in the model.
Qt= f( PPt-1, SMt-1, Qt-1, SL,M) (1)
PP= Spatial Interpolation of Precipitation
SM= Spatial Interpolation of Soil Moisture
Q = Stream Flow
SL = Site Location
M = Specific month
t = time
III.
Results
and Discussion
Inverse distance-weighted interpolation (IDW) was used in the present study. Thus, I assume that unknown climatic value of a station is influenced more by nearby know values than those farther away. The interpolation was applied for each month and each variable. Figure 4 shows some examples of the results of the spatial interpolation.
Fig 4. Spatial Interpolation of precipitation (A) and soil moisture (B) for January and June respectively
Obtained the spatial interpolation and using equation 1, three linear models were tested (Table 3). Model 1 shows the relationship among precipitation, soil moisture and streamflow; the R-squared obtained is equal to 0.64. Model 2 uses the same variables that Model1 plus the site location; the R-Squared is equal to 0.66. Finally model 3 incorporates the season variability and the R square is equal to 0.68.
Table 3 Relationship between different variables to predict streamflow
|
Type
Model |
R2 |
Equation |
|
Model1 |
0.64 |
Q (t)=
-25.5PP(t-1)+12.80 SM (t-1) + 0.72 Q (t-1) -133.36 |
|
Model2 |
0.66 |
Q (t)=
-14.58PP(t-1)+16.70 SM (t-1) + 0.59 Q (t-1) +40.58SL -348.47 |
|
Model3 |
0.68 |
Q (t)=
-25.264PP(t-1)+11.84 SM (t-1) + 0.6 Q (t-1) +40.42 SL-16.90 M -129.58 |
From these results we can imply that the variable precipitation doesn’t contribute too much to the model. It may be due to the type of interpolation that I use. Probably the IDW interpolation is not the most adequate to represent the precipitation data. Other potential weak of the model may be the few data that I consider for the model; take 15 weather stations for a watershed area of 7,500 square miles may not represent the precipitation conditions in the watersheds.
Flow from one month before and the specific site are the most relevant variables. This is because both variables are influenced by different characteristics such as drainage area, elevations, land use and water level of reservoirs.
Results for the model 3 for observed flow and predicted flow are illustrated in Figure 5
Fig 5 Flow measure vs Flow Predicted
Figure 6 shows the variation of observed and predicted flow during the different time and sites. Fig.5 and 6 show that the precipitation, soil moisture and streamflow do not have higher correlation among them. However, a simple linear analysis provide a roughly approximation of the average flow for the next month.
Fig. 6 Measured versus predicted streamflow
Forecasting analysis needs to be improved, Thus is necessary to consider in the model more years of data, incorporate more relevant variables such as level of reservoirs, elevation and also is necessary a better statistics analysis in the correlation (analysis with no linear regressions) .
IV.
Conclusion
ü Spatial interpolation and builder model analysis of GIS are useful tools to accomplish hydrological analysis. In the present study, spatial interpolation was used to know climatic information in unknown areas and model builder was used to systematize my work.
ü Simple regression was developed to predict the average monthly flow. Among the three models tested. Model 3 provides better results (R2=0.68). This is because model 3 incorporates the location of each site and also the variable seasonal.
ü Among the different variables used for the model, the streamflow (one month before) play the major role. A second important variable is the location site, which represents the characteristic of each drainage area (elevation, land use).
ü The advantage of used model is that relatively simple, straightforward to use, easy to calibrate, and provide reasonably accurate. The disadvantages are that model does not represent all known physical processes that affect streamflow such as level of reservoirs in the watershed, elevation and groundwater influence.
V.
Reference
(1)Bear River
Information System
Gregory Knight, Heejun Chang, et al. (2001), A Simplified Basin Model For Simulating Runoff: The Struma River GIS
(2) Natural Resource Conservation System
http://www.wcc.nrcs.usda.gov/snow/
Olson, B. E., 2008, Annual Habitat Management Plan, Bear River Migratory Bird Refuge. Brigham City, UT: U.S. Department of the Interior, Fish and Wildlife Service.
(3)U.S. Geological Survey
VI.
Appendix
1
|
Stations |
Site ID |
Code |
Site Name |
Long. |
Lati. |
Elev |
state |
|
SNOTEL Stations |
1 |
374 |
BUG LAKE |
-111.42 |
41.68 |
7950 |
UT |
|
2 |
455 |
DRY BREAD POND |
-111.54 |
41.41 |
8350 |
UT |
|
|
3 |
517 |
HAYDEN FORK |
-110.88 |
40.80 |
9212 |
UT |
|
|
4 |
579 |
LILY LAKE |
-110.80 |
40.86 |
9156 |
UT |
|
|
5 |
582 |
LITTLE BEAR |
-111.83 |
41.41 |
6544 |
UT |
|
|
6 |
634 |
MONTE CRISTO |
-111.50 |
41.47 |
8960 |
UT |
|
|
7 |
1013 |
TEMPLE FORK |
-111.55 |
41.79 |
7406 |
UT |
|
|
8 |
823 |
TONY GROVE LAKE |
-111.63 |
41.90 |
8386 |
UT |
|
|
9 |
730 |
SALT RIVER SUMMIT |
-110.91 |
42.51 |
7760 |
WY |
|
|
10 |
471 |
EMIGRANT SUMMIT |
-111.56 |
42.36 |
7390 |
ID |
|
|
11 |
484 |
FRANKLIN BASIN |
-111.60 |
42.05 |
8085 |
ID |
|
|
12 |
493 |
GIVEOUT |
-111.17 |
42.41 |
6930 |
ID |
|
|
13 |
761 |
SLUG CREEK DIVIDE |
-111.30 |
42.56 |
7225 |
ID |
|
|
14 |
677 |
OXFORD SPRING |
-112.13 |
42.26 |
6740 |
ID |
|
|
15 |
509 |
HAMS FORK |
-110.68 |
42.15 |
7840 |
WY |
|
|
Stream Gages Stations |
1 |
10016900 |
EVANSTON |
-110.96 |
41.27 |
6730 |
WY |
|
2 |
10020300 |
BELOW RESERVOIR, NEAR WOODRUFF |
-111.01 |
41.51 |
6400 |
WY |
|
|
3 |
10028500 |
BELOW PIXLEY DAM, NEAR COKEVILLE |
-110.99 |
41.94 |
6185 |
WY |
|
|
4 |
10038000 |
BLW SMITHS FORK, NR COKEVILLE |
-110.97 |
42.13 |
6140 |
WY |
|
|
5 |
10068500 |
PESCADERO |
-111.36 |
42.40 |
5950 |
ID |
|
|
6 |
10092700 |
IDAHO-UTAH STATE LINE |
-111.92 |
42.01 |
4420 |
ID |
|
|
7 |
10126000 |
NEAR CORINNE |
-112.10 |
41.58 |
4204 |
UT |