Skip to main content
ARS Home » Pacific West Area » Corvallis, Oregon » Forage Seed and Cereal Research Unit » Research » Publications at this Location » Publication #206639

Title: Spatial Display of Agronomic Data While Preserving Grower Confidentiality

item Mueller Warrant, George
item Whittaker, Gerald

Submitted to: ESRI Conference Proceedings
Publication Type: Other
Publication Acceptance Date: 6/26/2006
Publication Date: 9/30/2006
Citation: Mueller Warrant, G.W., Whittaker, G.W., Young, W. 2006. Spatial Display of Agronomic Data While Preserving Grower Confidentiality. ESRI Conference Proceedings. Map ID 250.

Interpretive Summary: Data from a decade of field inspections conducted by the Oregon Seed Certification Service (OSCS) were shared with us for analysis of trends over space and time in the severity of weeds of Oregon’s grass seed industry. The data were shared with the understanding that no confidential business information would be released to the public. Sophisticated procedures to locate the probable boundaries of individual grass seed fields were developed and used on data from Linn County, Oregon. Knowledge of field locations allowed us to characterize crop rotation patterns used by the certified grass seed growers, something never previously accomplished, defining features such as the tendency of weeds to survive from one planting to the next despite efforts by farmers to destroy them. We also produced maps of areas with significantly elevated (or reduced) risks of having serious problems with individual weed species using procedures that protected confidentiality of the original data while displaying the tendency of weeds to cluster together across the landscape.

Technical Abstract: Data from 10 years of field inspections conducted by the Oregon Seed Certification Service (OSCS) were shared with us for analysis of spatial and temporal trends in weed severity in Oregon's grass seed industry under the provision that no confidential business information (CBI) would be released to the public. Preserving confidentiality while imparting the essence of our findings was, and still remains, challenge #1. In the OSCS database, only field sizes and nearest township/range/section (TRS) locations were known. We used 10 years of late summer Landsat images to differentiate among crop species/landuse patterns, USDA-FSA Common Land Unit (CLU) polygons to identify likely boundaries for fields, and descriptive information from within the OSCS field inspections to georeference 20% of the entire dataset, approximately 10,000 single-year reports representing all certified grass seed production in Linn County, Oregon, from 1994 to 2003. Georeferencing to convert a non-spatial database into a GIS was challenge #2. Once the 10,643 field inspection reports from harvests on 2779 distinct fields in Linn County had been georeferenced, we began searching for ways to summarize this massive volume of data into smaller, more comprehendible slices. Knowledge of the spatial location of fields allowed us to extract crop rotation patterns and compare differences between old and new stands in weed severity, choice of crop grown, and stand duration. A wide range of agronomic conclusions from these analyses have been submitted to the journal "Weed Science" with the approval of the OSCS. Developing and statistically testing these summary tables was challenge #3. Many of the weed species displayed unique spatial patterns across the landscape, but simply showing those on maps risked releasing CBI. We examined a wide variety of GIS tools in our search for ways to communicate the spatial properties of these weeds without violating CBI expectations. On one extreme, maps of field polygons color-coded to represent severity of individual weeds within single years would clearly compromise growers' expectations of privacy. Our first step from that extreme was to pool data over years to display only the worst case (or average case) within a field over the decade, providing some shielding of the CBI contained in the OSCS records. These 10-year maximum severity data were then converted from polygons to centroids to allow use of interpolation tools such as IDW and kriging to produce rasters extending across space and blurring the edges of fields. How well each of the various options in IDW and kriging succeeded in smearing the data and protecting CBI was debatable, but such procedures have inherent tendencies to display values close to, if not identical with, the real values at the field centroids. Kriging probability maps avoided many of our concerns over release of CBI, but failed to characterize the clustering of weeds as clearly as could be achieved with global and local Moran's I and Getis-Ord G statistics. Successful implementation of these clustering algorithms was challenge #4.