Conservation Reserve Program (CRP) Model Layer Development

STATISTICS OF CRP AREAS FOR EIGHT WA COUNTIES

Statistics for CRP areas were obtained from NRCS for eight Columbia Basin region Washington counties:  Adams, Benton, Douglas, Franklin, Grant, Lincoln, Spokane, and Whitman.  Candidate pixels for consideration for identification as CRP lands were identified using a September '93 AVHRR bi-weekly composite image and the domain soils map.  Those pixels with a spectral signal consistent with a shrub-steppe vegetative cover and located in susceptible L series soils were designated as candidate CRP pixels .

SPECTRALLY DETERMINED CRP CANDIDATE PIXELS  

The candidate pixels defined by spectral and soil values were then reprocessed to be constrained by the NRCS statistics.  For each of the six counties with both candidate pixels and NRCS statistics, candidate pixels were randomly eliminated until the resulting CRP area matched the reported area.  The resulting pixels represent probable CRP lands, realistically distributed within the domain in accordance with available statistics and spectral and soil associations.

CRP MODEL LAYER AT 1 KM  

The process of randomly decimating the county specific spectral-soils pixels is discussed next.  The candidate pixels were reprocessed using a FORTRAN code run on a UNIX workstation.  This code read a file containing the candidate pixels, along with county identifiers.  County specific ratios of total area of candidate pixels (from the spectral-soil determination) to reported areas of CRP were used to establish criteria for each county for randomly eliminating candidate pixels.  The area of candidate CRP pixels always was greater than or equal to the NRCS reported area, in all of  the six counties with candidate CRP pixels.  The random pixel elimination process was initialized uniquely for each of 40 runs. For each of these runs the resulting error (as RMSE) was calculated across the counties for the area of the surviving pixels relative to the target area.  The best of the 40 runs was selected for re-export as a data layer in the GIS system and for use by the modelers.  The resulting data layer is shown above.