GIS 5100 - Lab 3 - Watershed Analysis

In Lab 3 we learned about modeling streams and watersheds using a provided Digital Elevation Model (DEM) from USGS.  We used a variety of tools in the Hydrology toolset within Spatial Analyst Tools.  The first step was to make the DEM by hydrologically correct by filling in all the sinks using the Fill tool.  That output was then used to create a flow direction surface raster (fdr) which used the D8 algorithm to create cells that had values of the possible 8 flow directions and their corresponding values (E-1, SE-2, S-4, SW-8, W-16, NW-32, N-64, NE-128).  The fdr was used as input for a flow accumulation raster which represents the number of upstream cells that flow into each specific cell based on the flow direction.  It is possible for the flow accumulation value to be zero if there are no upstream cells flowing into a particular cell.  A stream raster was creating using the Conditional Con tool and the flow accumulation raster and setting an arbitrary threshold to a value of >= 200.  This threshold determines the number of cells it takes to create a stream channel.  Using the Stream to Feature tool we were able to convert the streams raster to a vector feature which maintains the logical flow direction.  Individual stream links and a stream order raster were created with the Stream Link and Stream Order tools.

To create watershed delineation we used both the Watershed tool and the Basin tool.  The Watershed tool uses a pour point field which in our assignment was the Stream Link output raster.  A watershed delineation was created for each stream segment.  The Basin tool uses the edges of the DEM as watershed outlets instead of pour points.   There were quite a variety of tools in the Hydrology toolset.  

For the final deliverable, I compared my modeled stream network and watershed results from the DEM raster with the actual (or observed) stream network and watershed features that were provided.  I chose to model a watershed that was on the east side of the island as it had a large outlet into the ocean.  Instead of just using one outlet point based on the stream raster, I chose five pour points because they were in close proximity to each other.  There were 2,649 observed streams with total stream network length = 2,623km and 3,606 modeled streams with a total stream network length = 2526km.   There was a difference of 957 streams or 1.3 times more modeled streams than observed streams.  Most of the 957 streams however are along the coastlines or edges of the island.   The 957 modeled streams are lower order streams (1^st order) that were probably created because of the flow accumulation threshold value of 200 and the flatness of the terrain.  I also changed the threshold value to see how it affected the number of modeled streams.  Increasing the Flow Accumulation threshold from 200 to 300 reduced the number of modeled streams from 3606 to 2414 or just a difference of 235 streams between the observed count of 2649.  Increasing the threshold to 500 drastically decreased the level of detail and reduced the number of modeled streams to just 1442. 

There were 22 observed watersheds and 1789 modeled basins.  During the modeling, a larger number of smaller watersheds/basins were created around the coastline edge which accounted for most of the comparison difference.  The pour points feature with 5 outlet points created 5 distinct modeled watersheds that fell inside the observed watershed but did not have the same coverage area as the observed watershed which was much larger.  The observed watershed covered 99 square kilometers.   The respective modeled watersheds had a cumulative watershed area of 65 square kilometers (1.7+51.8+9.5+.8+.9). 

I was not aware of the ability of ArcGIS to create so much information from one DEM raster.

Comparison of Streams & Watersheds - Observed vs. Modeled