The Effects of Soils and Time on the Accuracy of Crop Identification in S.W. Ontario from 1974 Satellite Imagery

Abstract

ERTS-1 imagery (June 2, June 20 and July 8, 1974), supplemented with airborne photography was used to identify com, winter wheat, soybeans, oats and fieldbeans in Southwestern Ontario. Sample areas were selected within Chatham - Harwich Townships, Kent County, Ontario according to land use capability classes. Following the stratification, five test sites were established over eight soil types within the two Townships (Figure 2.3). Two types of classifiers based on the maximum likelihood and minimum distance decision rules were used to analyze successfully the data derived from the imagery. Using the maximum likelihood decision rule the highest identification accuracy for c o m (100%) was achieved on June 2 over the Brookston Sandy loam. The corresponding accuracy for soybeans (100%) was achieved on June 20 and July 8 over the clyde clay/clyde loam and the Brookston silt loam; fieldbeans (100%) was achieved on June 2, June 20 and July 8 over the Beverly loam; and winter wheat (86%) was achieved on June 2, June 20 and July 8 over the Brookston silt loam. Using the minimum distance decision rule, the highest identification accuracy for c o m (82%) was achieved on July 8 over the Beverly loam. The corresponding accuracy for soybeans (80%) was achieved on June 20 and July 8 over the Brookston silt loam; fieldbeans (83%) was achieved on July 8 over the Beverly loam and the Brookston silt loam; winter wheat (100%) was achieved on June 2 and June 20 over the Beverly loam and the Brookston Sandy loam; and oats (100%) on June 2 and July 8 over the Brookston clay and the Beverly loam. The results of classification indicated possible significant time and soil type influences on the accuracy level of crop identification achieved using ERTS-1 digital data.