Using predictions based on geostatistics to monitor trends in Aspergillus flavus strain composition

Aspergillus flavus is a soil-inhabiting fungus that frequently produces afl atoxins, potent carcinogens, in cottonseed and other seed crops. A. flavus 8 strain isolates, characterized on the basis of sclerotial morphology, are highly toxigenic. Spatial and temporal characteristics of the percentage o f the A. flavus isolates that are S strain (S strain incidence) were used t o predict patterns across areas of more than 30 km(2). Spatial autocorrelat ion in S strain incidence in Yuma County, AZ, was shown to extend beyond fi eld boundaries to adjacent fields. Variograms revealed both short-range (2 to 6 km) and long-range (20 to 30 km) spatial structure in 8 strain inciden ce. S strain incidence at 36 locations sampled in July 1997 was predicted w ith a high correlation between expected and observed values (R = 0.85, P = 0.0001) by kriging data from July 1995 and July 1996. 8 strain incidence at locations sampled in October 1997 and March 1998 was markedly less than pr edicted by kriging data from the same months in prior years. Temporal analy sis of four locations repeatedly sampled from April 1995 through July 1998 also indicated a major reduction in 8 strain incidence in the Texas Hill ar ea after July 1997. Surface maps generated by kriging point data indicated a similarity in the spatial pattern of S strain incidence among all samplin g dates despite temporal changes in the overall 8 strain incidence. Geostat istics provided useful descriptions of variability in S strain incidence ov er space and time.