The Water Balance Post Processor

Download 
What is it? 
Objective 
Justification 
Methodology: AMS post-processing 
Methodology: GeoWRSI post-processing 
Methodology: Combining point data and raster data: co-interpolation 
Program operation 
Dependencies: system requirements 
Sample datasets 
Program Development 
Contact us 

The Water Balance Post Processor (WBPP) is a program that facilitates yield modeling and estimation from some water balance models.

The WBPP uses outputs from a water balance modeling program and reformats them into summaries that are compatible with historical yield records. It can basically be used to facilitate yield modeling and yield estimation from water balance models. The program has four primary uses:

1. Combined analysis of historical yield records and historical water balance analyses to produce data tables that are ready for regression modeling. This regression modeling would result in yield estimation models.  
2. Extracting summaries from up-to-date water balance analyses so that these summaries can be used as inputs into water balance models in order to derive up-to-date yield estimates and forecasts.
3. Combining outputs from the point-based FAO AgrometShell and the raster-based GeoWRSI into a set of improved water balance outputs
4.  Calculating summaries from water-balance analyses for specific regions of interest such as districts, provinces or countries.

Justification

Water Balance modeling programs such as the Agromeshell (AMS) and the GeoWRSI produce outputs that are useful in food security analysis, in particular for yield estimation. The outputs from these programs are geospatial datasets that are in point vector format (in the case of the AMS) and raster grid format (in the case of the GeoWRSI). These outputs are generally not compatible with historical yield data, which tends to be produced and archived by administrative units such as provinces and districts. Because of this difference, translating the outputs from these water balance models into a format that can be used directly for yield estimation is a laborious and time-consuming process. The WBPP considerably simplifies this process by enabling the user to select outputs from the water balance models, and automatically reformats these outputs into a format compatible with historical yield records. It also combines the water balance outputs with the historical yield records, thereby simplifying the regression process. Finally, it combines the point-based AMS with the raster-based GeoWRSI to produce enhanced water balance outputs.

The AMS produces a set of comma-delimited files, one for each year, which contain information on various water balance parameters for each station that was used in the water balance analysis. The following process is then undertaken in the WBPP with water balance parameters selected by the user and for the selected years:

1. reformat the AMS outputs for SEDI (Satellite Enhanced Data Interpolation) interpolation
2. interpolate the outputs using SEDI in Windisp to produce grids of interpolated water balance parameters. A blank background image is used, which basically makes the interpolation an Inverse Distance Weighting (IDW).
3. calculate the average value of the water balance parameters   for each polygon from a map (e.g. districts) which the user specifies
4. combine these average values with historical yield stats from a separate data file (but the yield stats have to be for the same districts that the AMS averages were calculated for),
5. reformat the combined file into a format which considerably simplifies regressions between yield and AMS outputs.

In the event that the user does not want to do a historical regression with yield, but only wants to aggregate the water balance data by administrative level, the WBPP also offers this option.

The GeoWRSI produces a set of raster grids, with each grid containing information on a specific water balance parameter. The following process is then undertaken in the WBPP with water balance parameters selected by the user:

1. calculate the average value of the water balance parameters   for each polygon from a map (e.g. districts) which the user specifies
2. combine these average values with historical yield stats from a separate data file (but the yield stats have to be for the same districts that the AMS averages were calculated for),
3. reformat the combined file into a format which considerably simplifies regressions between yield and AMS outputs.

As with the AMS, in the event that the user does not want to do a historical regression with yield, but only wants to aggregate the water balance data by administrative level, the WBPP also offers this option.

One strong advantage of the AMS is its ability to provide the most accurate water balance calculation at each station point for which data exist. In contrast, the advantage of the GeoWRSI lies in it’s ability to provide a good depiction of the variation in water balance parameters over geographic space. By combining these two outputs, the user can take full advantage of the strengths of these two approaches and integrate them to produce an enhanced output.

Co-interpolation of the AMS and the GeoWRSI outputs is an option in the WBPP. It is automatically done using the SEDI routine in Windisp. The corresponding AMS and GeoWRSI products are automatically matched using the filenaming convention specification from the two water balance programs. A batch file is created with all the commands required to carry out the commands, and Windisp is automatically called up to process the commands.

In order to operate the program, the user needs to launch Microsoft Excel and then click on the button “Activate Postprocessor”. This will launch the WBPP.

1. The user chooses whether to develop a model (requires historical yield data) or to run a model
2. The user chooses whether to process the outputs for
1. the AMS
2. the GeoWRSI
3. a combination of the AMS and the GeoWRSI
3. The user selects the directory with the input files and the directory where the outputs should be placed
4. The user selects the map file containing the administrative regions or other regions of interest that the outputs should be summarized by.
5. The user specifies a file with the “years” information (the years which the program should process outputs for). In the case where a season cuts across two years, the “year” specified should be the year in which the season begins
6. If the user has opted for AMS postprocessing, the user specifies the name of the station list from which the outputs were derived.
7. The user specifies the path of the Windisp program.
8. If the user has opted to develop a model, the user needs to specify the file containing yield information.
9. The user can select which water balance parameters should be postprocessed.
10. The user can define how the SEDI processing should be done.
11. Once all the inputs have been defined, the program automatically carries out the processing described in the methodology section and produces the relevant outputs as specified by the user.
12. The program automatically cleans up the intermediate files that were created during the program run, but the user can opt to save these intermediate files.

Each time the WBPP is launched, it automatically loads the last settings that the user specified the last time the program was run. In addition, the user can also save the settings (1 to 8 as described above) for future reference or future rerunning of the program, and can opt to load up a set of saved settings.

1. Operating system

The program can be operated from Windows XP and Windows 2000. It should also work on Windows ME. It has not been tested on Windows 98 or Windows NT

1. Windisp

The program requires Windisp 5.1 in order to work. The user needs to have installed Windisp 5.1 installed on their computer in order for the program to work. – it takes all the user specifications and generates a batch file which interpolates the AMS outputs and extracts stats by polygon. Once the program has generated the batchfile, the program then uses the shell command to call Windisp and run the batch file. After Windisp has run, the program then organizes the output in a second step

2. Microsoft Excel

The program requires Microsoft Excel to work. Microsoft Excel needs to be installed on the user’s computer. The program is based on VBA in Excel. It does not have any Excel dependencies (i.e., it makes no direct reference to Excel functions) except that it has to be launched from Excel.

Sample datasets to illustrate the program operation will soon be available.

This program was developed by the USGS/FEWSNET activity in collaboration with SADC RRSU

Conceptualization:

T. Tamuka Magadzire
Chris Funk

Program Writing:

T. Tamuka Magadzire

Program testing:

T. Tamuka Magadzire
M.S. Gamedze