Watershed Analysis

            Watershed analysis is an essential device in GIS that performs DEM and raster data operations to delineate watersheds and define stream networks and basins. It is especially applicable in hydrology and water resources. The amount of watersheds produced depends on the spatial scale. More watersheds would be produced at a smaller scale. The thresholds chosen for the stream network and the quality of the DEM strongly influences the outcome of the watershed analysis. Also, algorithm for deriving flow direction plays a role in influencing the analysis. The objective of this lab assignment was to perform 2 components (basin and stream network), and compare this watershed analysis to a product of a different source. The area of focus was the Tibetan Plateau, which locates the world’s biggest group of high-altitude endorheic lakes.

The methods necessary to perform this lab correctly involves close attention to detail, since the steps are easy but a mistake would ruin the whole process. Spatial Analyst extension must be checked before proceeding with the tools. The Hydrology tool is located in Spatial Analyst under Arctoolbox. The first step is the make the fill, but this was already provided for student convenience, and was easily accessible from the class website. The second step is to make the flow direction raster with the fill DEM as the input. Then, generate the basin by using flow direction as the input. After this last step, the first component of the lab can be completed by converting the basin into a polygon using the raster to polygon tool. Stream network is a little more complicating than creating the basin. The output will show small individual stream sections. Flow accumulation raster tool is used under Hydrology, and the fill from previous steps is the input weight while the flow direction is the next input. Both is needed to create the flow accumulation raster. Float is kept as default. The flow accumulation raster is reclassified using the reclassify tool. It is reclassified into a stream raster file. The number of classes is changed to 3 and the method is changed to manual (in that order). The break values are changed also to 0, 500 (threshold), and 9999999999999999 (until the end). The first two new values are changed to “NoData”. The next step is to make the stream order. The stream order tool is under Hydrology. The reclassified raster is used for the input stream raster and the flow direction raster is used for the other input. Also, the method of stream ordering was changed to shreve. The stream files can then be vectorized into stream lines using the stream to feature tool. The symbology needs to be changed to show more clarity. Change the symbology to graduated colors and set the value to GRID_CODE and make it 5 classes. The colors for each class can be adjusted. It is important to consider moving the lake layer to the top of the contents. Now, other sources can be downloaded and compared with the finished stream network and basin. It could be a good idea to put the basin layer on top of the other source to see the difference more clearly. For this lab exercise, data for Asia was downloaded from Global Drainage Basin Database.

           

The watershed map created had an extensive amount of stream networks and water basins. Most of the streams have a stream order of 1 (green). This indicates that the water is flowing from a primary source. Therefore, more water can be collected. The streams are definitely contributors to the formation of lakes. Streams appear to be very detailed and lakes are clearly evident. The watershed map that was created has more detail in stream network when comparing maps. It even shows more accuracy. The data downloaded from the Global Drainage Basin Database has misleading data. There is empty data where it is supposed to be partly or wholly covered by lakes. This white area is not counted as basin. This result is probably due to the fact that the data was for the entire Asia. It does not show much detail for the area of study. The lakes are shown as points and there are just a few. On the other hand, the advantage of my map is the detail in stream networks and lakes. The basins for both maps do not match. These differences are likely due to conflictions on method type. My map was area-based while the downloaded data was pour-point. Perhaps, there was a z-limit difference during the process of the watershed analysis. Z factor was a huge issue when working with ArcGIS 10. Another important note to consider is that there shouldn’t be only one basin if there are parallel lakes. These discrepancies can be seen when closely comparing the basins for the two maps. For example, the big basin for my map at the bottom right corner contains several lakes. This is a reasonable error to catch when comparing it with other sources. In this case, the downloaded data has a more accurate map of basins. It also seems that the largest basins are produced in areas of the largest stream order. This is not the case for the downloaded data. Basins for this data are separated while my map has basins all in one big piece. Nevertheless, stream network and the number of lakes my map shows is significant when compared to the downloaded data. The downloaded data only shows 3 lakes, which is insufficient to provide a truly realistic watershed analysis.

           

The fill is an important step for the rest of the analysis to work properly. The step became an issue from the start and was resolved by the TA. Which z value was supposed to be used was one of the major questions regarding this issue. There was no problem in ArcGIS 9 but ArcGIS 10 seemed to not tolerate any z value errors. Further problems occurred with reclassification. No instructions were given about how we should input values for break values or new or old values. It took some trial and error to figure out which numbers fit best. And, of course there were unexplainable errors that occurred. Usually, these unknown errors are solved by redoing the procedure. Downloading the right DEM is important in working with watershed analysis. The resolution and quality is especially important. There were some complications with USGS data sources. It did not seem to work properly when performing fill or any other tools after that. It produced a blank output with stars (bright spots). So, again, this might have been another mystery of the fill problem.

Nevertheless, watershed remains an important topic in GIS. It requires the integration of knowledge and data. It is important in solving hydrological problems. The effects of watershed analysis can be very helpful. For instance, it can prevent vegetation and fertile soil from getting depleted by proper management of watersheds in a sustainable manner. Therefore, watershed analysis is crucial for environmental protection. And comparing personally created maps with other sources especially helps in explaining the differences of the best possible information available for watershed purposes. In this lab, both maps were needed for proper analysis. Downloaded data had more accurate basins, while the created map had more detailed drainage networks and lakes.