After the late October basin delineation meeting here at NSSL, we chose prototype areas in Virginia and south Texas. The NED data quality in Virginia is relatively low while the data quality in south Texas is high. Therefore, these two areas should represent the full spectrum of delineation results. In addition, both areas are coastal with relatively flat terrain, and thus should provide good insight into the extent to which burning in streams from the National Hydrography Dataset will be necessary.
In late November, EROS sent the NED and digital line graphics (DLG) hydrography data (including streams and lakes) for the prototype areas. (Figure 1 - Image of NED data for the prototype areas.) They also sent their AMLs (ARC/INFO scripts) for delineating sub-basins within the 8-digit HUC basins. These scripts are amazingly efficient, with the average HUC requiring only about 3 1/2 to 4 hours processing time. (Figure 2 and Figure 3 - Images of delineation status by HUC in Virginia and Texas prototype areas.) However, the assignment of Pfafstetter IDs is not included in these scripts. Paul Jendrowski has written Avenue scripts (for use in ArcView) to do Pfafstetter codification, but there is still more work to be done in this area.
One of the issues we have been working to resolve is whether or not we can use Pfafstetter codification from top to bottom. Our initial idea was to begin with the three-, four-, or five-digit Pfafstetter basins delineated from the 1-km resolution GTOPO30 elevation data, and add Pfafstetter digits as we further divided these basins. However, because of the coarse resolution of the GTOPO30 data, the basins derived from it may not be accurate enough to use as starting points. Seann Reed (OH) recently sent the 400-meter resolution flow direction grids he has developed for three River Forecast Center (RFC) regions. Using the Avenue Pfafstetter scripts, we delineated basins (approximately the size of the 8-digit HUC basins) with Pfafstetter codification. These flow direction grids seem to provide the added resolution necessary to derive reasonable starting point basins, and many of these basins actually correspond to the 8-digit HUC basins. Ideally, we would like to use the 400-meter flow direction grids to determine starting point basins for the entire country. However, this depends on the availability of the 400-meter grids. The other option suggested by Kris Verdin at EROS is to use the HUCs as the starting point basins and then use the Pfafstetter codification within the HUC framework. Thus, each basin would be identified by the 8-digit HUC it is in, plus its own Pfafstetter ID within the HUC basin. These sub-basin IDs would not be as efficient and meaningful as IDs based solely on Pfafstetter codification, but this may be our only alternative. We will be discussing these issues further with Kris Verdin, Seann Reed, and Paul Jendrowski.
Despite the Pfafstetter issues still to be resolved, the results we are seeing so far are very encouraging. The delineations we have done in Virginia and Texas look EXCELLENT!!! (Figure 4, Figure 5, and Figure 6 - Images of the location of HUC 12090205 in south Texas and the basins delineated in the northern portion and southern portion of this HUC.) At this point, it does not appear we will have to burn in streams to correct errors in the elevation data very often, if at all. There are generally a couple small basins in each HUC where the delineation could be improved slightly, but so far we have not seen anything to necessitate burning in streams. This will save quite a bit of time and manpower and allow us to concentrate on the full-scale delineation when the data arrives. The RAID disk arrived last week and our new workstations arrived a few weeks prior to that, so we are very close to being ready for full production mode. Ken, Gina, Seann, and I will be attending a NED-H collaborators meeting at the EROS Data Center on Jan. 25-27 to work out details relating to methodology and transfer of data sets. We will provide another web-page update after the meeting.