Some FAQ for the on-line LTHIA and watershed delineation.
May 27, 2005
Larry Theller
Basic Data
We have layers of streams and waterbodies from the US Census bureau, which is what we would call a 1: 50,000 scale layer. This is the best layer of streams and lakes we have, but it means it is not expected to be exactly lined up with things shown on new aerial photos, which are made at a scale of 1:12,000 ( roughly three times more accurate.)
|
|
|
We use an elevation model that has a point every 30 meters, and we stretch a sheet across that and follow the flow patterns that it describes. This was built from the contour lines on the old USGS topographic maps, which are great maps but not updated anymore. So they show elevation features as they were in 1964. They are a scale of 1:24,000.
These topographic maps show stream features, but are different size maps than those which were used by the Census. So our digital streams and lakes do not exactly match the streams and lakes on USGS topomaps because they have different sources and scales. This is expected and a common issue in GIS.
|
|
|
The catchment areas are calculated from the recent (2007) USGS 10 meter DEM, and flow direction is controlled by the addition of the High-resolution NHD data (vector layer of streams and lakes.) So water moves across a 10 meter resolution surface, which has the NHD High-resolution streams "burned" into it.
Our Landuse values come from a layer known as the National Landuse Cover Dataset, (NLCD) from 2001. It is arguably the best and most complete landuse we can get right now. So water features on this Landuse layer represent their size and location in 2000-2001 when the satellite images were taken which were used to create NLCD. The satellites of the day were nowhere as precise as today, and the size of water bodies in the NLCD is different from that in the Streams layer we generated from High-resolution NHD data, which itself is derived from old USGS maps corrected with aerial photography.
Landuse data derived from remotely sensed data will contain errors due to misclassification of the data as well.
|
|
This illustrates a typical difference between 2001 NLCD landuse data (the squares) and the NHD water body layer (Blue line.) In Particular an issue for small features. Scale is indicated by the 200 meter long line. |
|
|
|
The delineation
The delineation process follows the surface that was fit on points 30 meters apart. So a feature needs to be pretty big to be represented. A lake 300 feet in diameter only has a few points representing it. Small ditches or outlets are missed, and streams direct our delineation even if they are not in the correct spot according to aerial photos.
Questions and Responses:
1. I have tried using
delineation to create a watershed for a number of lakes and found a few
discrepancies with what we have used in the past--total watershed acreage is
fairly comparable, but lake acreage differs considerably.
What is the error
plus/minus on lake acreage and total watershed acreage.
Two different items. Water reported as landuse is from NLCD layer of 1992, whereas the watershed size is calculated on the fly from the USGS DEM. Accuracy of the DEM varies from very good over most of the state to quite poor in some areas. That said, it is a 30 meter posting and is not intended to delineate field-size areas accurately.
Most of the watershed
info that we have comes from the NRCS.
Some of our lake
acreage has been from USGS (e.g. Starve Hollow Lake in Jackson Co near
Vallonia--USGS 1959).
We used the new USGS 8 digit watersheds as boundaries for
LTHIA, and we will match the large watersheds pretty well. Our
Your NRCS lake data is much more accurate than NLCD, although out of date now, perhaps, as the land changes over time.
2. If this is easy to
explain, how is watershed acreage calculated on your website?
It is NOT. The watershed acreage is calculated from the watershed boundary that is estimated from the DEM. Standard techniques that are widely used within GIS tools are used in estimating the watershed boundary from the DEM for the outlet selected.
3. Sometimes the
entire lake is not captured when clicking on the outlet, according to the map
(e.g. Buffalo Trace Lake in Harrison Co near Palmyra; Lake Iola in Scott Co in
Scottsburg). Why is this?
The lakes are not included in the DEM as individual features, they are just low spots in approximately the right place. The program does not check for presence or absence of a lake. In this case the delineation followed the DEM, probably along an old stream in the Census layer.
4. Does a lake always
consist of one soil group?
No. No real connection, soils are mapped as “under” most small and medium lakes. Only very large bodies of water are in the soil map as water. So two or more soils may touch under a “lake.”
5. When two
"Water" listings exist under "Land use" and one is the
selected lake (via its outlet), is the other "Water" listing wetland
and/or other bodies of water in the watershed (e.g. Pisgah Lake in Johnson Co
near Edinburgh has a few ponds that drain into it)?
If you are asking about the water listings for a delineated watershed, water as other categories may appear multiple times in the table because they are located on different soil groups. These do not necessarily correspond to a specific water feature or lake.
6. Are rivers/streams
included in "Water"?
Yes if they are
big enough to be in the landuse layer.
7. How do you
recommend citing this website for watershed delineations?
The site should referenced as:
Choi, J.Y., B. Engel and L. Theller. 2005. Online Watershed Delineation. http://pasture.ecn.purdue.edu/~watergen/
This reference describes the technical aspects of the www site.
Choi,
J.Y. and B.A. Engel. 2003. Real time watershed delineation system using
web-GIS. Journal of Computing in Civil Engineering 17(3):189-196.