|
giswr
> events
> feb 2000 giswr
conference > feedback from alan rea
Feedback from Alan Rea
Geodatabase Concepts for a National Spatial
Data Infrastructure (NSDI) Watershed Boundaries Dataset
(WBD)
Alan Rea, US Geological Survey
2/28/2000
Several federal agencies are collaborating to produce a
nationally consistent and seamless set of watershed boundaries
smaller than the current 2200+ cataloging units (8-digit
HUCs). This effort is tied to parallel efforts to develop
the National Hydrography Dataset (NHD) and the National
Elevation Dataset Hydrologic Derivatives (NED-H). The ultimate
goal is to revise these three national datasets to be vertically
integrated such that all three data sets are in agreement.
The current system of Hydrologic Units is a hierarchical
system with 4 levels. The WBD will extend the hierarchical
system two additional levels. Smaller drainage areas also
will be delineated using the NED-H and NHD such that a drainage
area or catchment will exist for each reach of the NHD.
This structure is summarized in the following table:
Hydrologic Unit Level Example
Region 1 11
Subregion 2 1113
Accounting Unit 3 111303
Cataloging Unit 4 11130302
Watershed 5 1113030207
Subwatershed 6 111303020703
Reach Catchment 7? ?use NHD Reach code?
Polygons at each level are nested within polygons at the
next higher level. Geometry should be shared between coincident
polygon boundaries. The dataset should be seamless nationwide,
so we will have shared geometry across the conterminous
US for about 150,000 6th-level subwatersheds, or eventually
2-3 million NHD reaches.
Ideally the Geodatabase implementation of this structure
should take advantage of the hierarchical nature of the
Hydrologic Units. In determining upstream units, we would
want to be able to start with the smallest size units we
have available (i.e. either 4,5,6th or 7? level), and select
units upstream from that until reaching the first unit at
the next higher level, then jump to that level, selecting
upstream units, and so on until we've gotten all upstream
units.
If we have to associate a line network with each polygon
at each level in the hierarchy, we could potentially do
that by constructing a "network" of links between
polygon labels at each level. We currently have enough information
to do that at the 4th level (8-digit units.) This would
almost certainly give better performance than if we used
the NHD or RF-1 to represent the network. This approach
would appear to fit in with the networking tools ESRI has
already developed, and could be considered a REALLY generalized
stream network.
In the Arc 7.X data model we have been modeling the hierarchical
structure outlined above using a polygon coverage with regions
defined for higher level Hydrologic Units. We also store
attribute information on the boundaries by building line
topology on top of polygon topology. We use line attributes
primarily for feature-based metadata for things like source
scale, etc. We haven't settled on a method for selecting
upstream polygons, though I'm aware of at least 4 different
basic methods to do that. The most likely of those to perform
well consist of preselected lookup tables (database hash
tables), or regions defined for each polygon. None of these
approaches look very good when we start to envision 150,000
to 3 million polygons in a single dataset. For something
like that, it would seem to be a real advantage to have
this structure built into the data model.
|
