Detecting and correcting logically inconsistent crop rotations and other land-use sequences

Abstract

ABSTRACTConverting multi-year land-use data into crop rotation history is relatively simple in the absence of classification errors, but severely compromised in their presence. Several classification errors can in theory be detected with a matrix of logically forbidden (or extremely unlikely) year-to-year land-use transitions. We categorized 730 of 3249 potential year-to-year transitions among 57 land-use classes in western Oregon as being logically permissible, with the remaining 77.5% forbidden. Applying these restrictions to eight consecutive years of land-use data revealed that an average of 26.7% of apparent year-to-year transitions among agricultural classes were illogical, in contrast to only 2.5% and 0.6% of urban development and forest transitions. The most useful correction applied to the data involved replacement of original majority-rule values (generated during prior pixel- to object-based data conversion) with second-place classification categories for fields with inconsistent land-uses identified as occurring at the beginning, middle, or ending year of multi-year sequences. This approach reduced year-to-year land-use inconsistency to 20.1% of the agricultural area. A lengthy series of additional iterations involving substitution of either majority-rule or second-place classification categories (randomized within years and counties) in cases lacking obvious ways to determine which year in pairs of years was in error in the previous iteration stabilized at 17.4% inconsistency by iteration 128. Our corrections improved the measurement of perennial crop stand ages in the complex landscape of the Pacific Northwest and similar approaches should be useful in other complex landscapes.