Assignment Two Part Two: Calibration for Flow Due April 3

Transcription

1 Assignment Two Part Two: Calibration for Flow Due April 3 The objective of part 2 of this project is to calibrate the tutorial SWAT model developed in part 1 for flow. Re-Running Part 1 The precipitation data we used in Part 1 was from two sites that were fairly distant from the watershed. There is better data from Rockwall County which is in the watershed. I will you 2 files: pcploc.dbf, and pcprock.txt. Put these in the folder AVSWATX\AvSwatDB\Example3. We will also reduce the number of HRU s (in order to make changing curve number easier) by using the dominant landuse and soil. Open the project you created in part 1 (sabine) and in the watershed view, click on Avswatx HRUs distribution and select Dominant Land Use and Soil instead of Multiple HRUs (page 33 of tutorial). Follow the tutorial as before but when defining the rainfall data (page 35 of the tutorial handout), instead of selecting pcpfork.dbf as the name of the rain gage location table (see top of page 36), select pcploc.dbf. The rest of the procedure is the same as the tutorial handout but use a daily time step, and start on 1/1/1978 (one year later) and end on 12/31/1979. USGS Observed Data Set The observ.txt file in the tutorial only covers the last 6 months of 1978 which was a dry year when there was very little flow. We will download the data from the USGS gage station for three years (1978 through 1980) and use 1979, which was a wetter year, for calibration (we don t use the first year as the model needs at least one year to warm up ). The USGS gage station number is given in the strflow.dbf file in the Example3 folder and it is , but this is the wrong number (it is below a reservoir to the east). The correct number is First, we go to the USGS website and download the daily stream flow file for our site. Go to check the box next to Site Number under Site Identfier, and hit Submit. Type in for site number, and check the box for 3 Streamflow, ft /s under Water Level/Flow Parameters. Near the bottom, click the radio button for date range and type in the dates and To check that you have the right data, choose Graphs of data under Output Options and Submit. You should see a graph of daily flow for the two years. If not, go back and check your settings. If you see the graph, go back and under Output Options, select Tab-separated data in the format YYYY-MM-DD, Save to file, and Submit. Create a folder under AVSWATX\sabine called Myfolder and save the file there under the name Download. Next we use Excel to get the USGS file into a text file that can be read by WDMUtil. Open Excel and then open Download (using the All files type). In the Text Import Wizard -Step 2 of 3, leave the tab delimiter checked, check the Other delimiter, and type in a dash to separate years, months and 1

2 days. Delete the first 28 rows of text, then delete all of the columns except the year, month, day and flow (only 4 columns). Save as a Text (Tab delimited) file under the name USGS.txt in Myfolder. Answer Yes to question about keeping format. Next we use WDMUtil to change the text file into a database file that can be read by GenScn. Open WDMUtil and go to File New and create a new file wdm in Myfolder. Go to File Import and select the USGS.txt file created in the steps above. The Script Selection window should open. Select Blank Script under Description and then click on Edit. Under the section on Header, remove the check mark next to Skip. Under the section on Column Format, select Tab Delimited. Then select the tab at the top for Data Mapping. Now we must indicate the column numbers for the entries. In the row for Value, under Input Column, type 4. For year type 1, for Month type 2, and for Day type 3. Since there is no entry for hour, hour will be a constant value of 24. In the row for hour in the top section, under Constant, type 24. In the row for Scenario, under Constant, type OBSERVED. In the row for Location, under Constant, type In the row for Constituent, under Constant, type FLOW-cfs. Click Read Data at the bottom and you should see that a database file is now listed in the Time Series section. Click on this time series and then click on the Write Time Series to WDM button in the lower right corner. In the Write to WDM window that comes up, put 1 in the box below Output DSN, and in Location box, and click Write. You should see a message that the data has been written to Now we need to change flow in cfs to cms in this file. Click on the Observed flow WDM file in the Time Series box and click on the Generate Timeseries button in the Tools section. Click on the Math tab at the top. Click the radio button for Div. Double-click the textbox to the right of the numerator label and select OBSERVED FLOW. Click in the textbox to the right of the denominator and enter a value of as the conversion factor. Change the text in the Constituent field from FLOW to FLOW-cms and the units field to m3/s. Save in the same wdm file you are working on (use the pull down menu to find wdm) and click OK. You should see a message that New data-set number 2 successfully stored on WDM file.... Now we will open GenScn and bring in the SWAT daily flow output and the observed database files. Open GenScn and click on File New Project. Use the pull-down menu under File Type and select SWATDBF. Click on Add from File. Navigate to C:\AVSWATX\sabine\scenarios\default\sim3\tablesout (it could be sim3 or larger in the string, whichever simulation has the daily time step using the new rainfall data) and select rch.dbf. Change the File Type to WDM, click on Add from File, and navigate to where you stored wdm and select it. In the GenScn window, in the Locations section, you should see the 16 reaches and It s a good time to save this project (File Save Project ), in AVSWATX/sabine/myfolder as Calibrated, for example. To select the time series for plotting, in the Locations section, click on reach number 11 and In the Scenario section, click on OBSERVED and tablesout:rch. In the Constituents section, click on FLOW_OUT and FLOW-cms. In the Time Series box, click on the Plus sign. Select both time series and click on Generate Graphs in the Analysis section, followed by Generate. 2

3 You should see the observed and SWAT predicted time series of flow for 1978, 1978, and Change the starting and ending dates so that you only have the year Use the Compare Two Time Series tool to get the value of the Model Fit Efficiency (Nash Sutcliffe) and RMS Error for the uncalibrated model (using just 1978 data). Print out the graph, show the NS and RMS Error and include this in the material you hand in. Base Flow Separation and the Groundwater Recession Parameter The recommended procedure for calibrating SWAT is given starting on page 687 of the SWAT 2000 Users Manual pdf file (look at the bottom middle screen to see the page number in Adobe Acrobat). The manual is in the BASINS\Docs\SWAT directory (swat2000.pdf). It suggests as a first step that you look at how SWAT partitions flow into baseflow and stormflow and compare that to the USGS data. To divide the USGS data into baseflow and stormflow, a Baseflow Filter Program (available at can be used. Download the instructions and program into a folder and unzip the program file. This will create a new folder baseflow which contains (among other files) the executable program (bflow.exe), an example input file ( prn), an example master input file (file.lst), and a journal article describing the process (baseflow_inst.doc). Use the same USGS file we downloaded before (Download) and bring this file into Excel, but this time keep the year, month, and day in one column (don t use - as a delimiter). Strip out all the text at the top and all of columns except YYYY/MM/DD in the first column and flow in cfs in the second column. Leave the first line blank (or put in date and flow as headers). To get rid of the slash marks in the dates, select the first date and use Format Cells Number Custom and type in yyyymmdd as the type. Then copy this format for all the date cells. Save the file as Formatted Text (space delimited) under the name prn in the baseflow folder. In the baseflow folder, open up file.lst with Notepad and you will see the following lines:!!input for baseflow program: 10!NDMIN: minimum number of days for alpha calculation 10!NDMAX: maximum number of days for alpha calculation 0!IPRINT: daily print option (0-no; 1-yes)!!Daily stream data files prn out The last line contains the name of the input file and optional output file. Change the last number in these file names to our USGS gage station ( ) and save the file in the same folder. Move prn into this folder. Change the 10 in the third line to 300 (keep the right-most zero in the same column as the original 0). Save this file. 3

4 With bflow.exe, file.lst, and prn in the same folder, double click on bflow.exe to get it to run. This should create a new file baseflow.dat with the SWAT baseflow recession rate parameter (ALPHA_BF). The file also shows three different estimates for the percentage of baseflow. According to the instructions, the percentage of baseflow in the SWAT predictions should fall somewhere within the rang of the three estimates. You can compare the percent baseflow of the observed data set to the partitioning between stormflow and baseflow in your SWAT predictions by looking in the output.std file in AVSWATX\sabine\scenarios\default\sim2\txtinout directory. This can be read with a text editor such as Notepad. Scroll to near the bottom and you will find a section called AVE ANNUAL BASINS VALUES. The total predicted average annual flow is TOTAL WATER YLD in mm. The baseflow part of total flow is approximately the value shown for GROUNDWATER (SHAL AQ) Q in mm. The storm flow is approximately the sum of SURFACE RUNOFF Q and LATERAL SOIL Q in mm. Dividing GROUNDWATER (SHAL AQ) Q by TOTAL WATER YLD gives you the predicted fraction that is baseflow. Viewing Parameter Values To see what the actual values of a parameter are in each HRU, (in SWAT view) go to Edit Input and select Subbasins data. You will see a window where you can select any of the subbasins from the first column. For a given subbasin, the land uses (that exceed the threshold percentage) appear in the second column and the soil mapping units appear in the third column. In the fourth column, if you scroll down and click on the management input file (.mgt), you will see a window that gives the curve number for moisture conditions II or average moisture conditins (CN2) in the upper right section. Other files can be selected from the fourth column to see additional parameters. Changing Parameter Values Once you start the calibration process, you will need to make changes to parameters such as curve number (CN2). Make sure you are in the SWAT view. You can can change parameters at the HRU level using Edit Input (described above). You can also change parameters at the basin level by going to Simulation and Run SWAT and then clicking on the Bsn button in the lower right. Start by changing the initial soil water content (FCCB) in the Basins inputs from 0 to 1 (indicting that initial soil water content is 100% of field capacity). This will speed warming up the model for flow. Also change the ALPHA_BF value in the.gw HRU file from the default to the value you found using the Baseflow Filter program. At this point, it is probably a good idea to run the sensitivity analysis to see which parameters are most important. You will need a new observed flow file (not the observ.txt file used in the tutorial which only covered the last 6 months of 1977 when there was very little flow). I will you a new observed file called myobs.txt. Put this file in the AVSWATX\AvSwatDB\Example3 folder and when you go through setting up the sensitivity analysis, select this file instead of observ.txt. We are 4

5 interested in the parameters that cause the greatest change in the objective function so this is the first line in the sensitivity report and in the sensresult.out file. Sometimes it s difficult to know whether a change to a parameter that you intended to make was actually made. You can check to see if the change was made by opening a blank project in ArcView, click on Tables in the left panel, and then use the Add button to open database tables in the sim directory of the scenario you have run. You should be able to see the parameter values for CN2 in each HRU by looking in the mgt1.dbf, for example. You can also look in the output.std file in the txtinout folder for each run. Viewing Output in Genscn Open the Genscn project you created to look at the observed flow (Calibrated.sta) and go to File and Edit Project. The model output is a SWATDBF file written to the tablesout folder for the simulation number you have run. For example, if your last run was sim3, change the file type to SWATDBF and click on Add from File. Then navigate in your project folder to scenarios\default\sim3\tablesout and select rch. Close the Edit Project window. In Genscn, select reach 11 for location, tablesout:rch for scenario, FLOW_OUT for constituent, and click on the plus sign in time series. You should see the scenario time series listed. You can compare this with the USGS observed flow or with another scenario file before you made a change. Flow Calibration Run SWAT for the period to Use the period to for the calibration (giving the model one year to warm up ). The key to making progress in calibration is changing one parameter at a time and being able to view the predicted flow before and after the change on the same graph. The best measure of a good fit is a combination of statistics, such as RMS error NS, and how it looks to you. Print out a graph of the final calibration time series for 1978 and give the NS and RMS error. Also include the graph and statistics for the uncalibrated run (see top of page 3 of these notes). Make a table that shows the parameters that you changed (default value and calibrated value) for you final calibration. 5