Predicting the Areal Extent of Land-Cover Types Using Classified Imagery and Geostatistics

Abstract

Remote sensing is an efficient means of obtaining large-area land-cover data. Yet, remotely sensed data are not error-free. This paper presents a geostatistical method to model spatial uncertainty in estimates of the areal extent of land-cover types. The area estimates are based on exhaustive but uncertain (soft) remotely sensed data and a sample of reference (hard) data. The method requires a set of mutually exclusive and exhaustive land-cover classes. Land-cover regions should be larger than the pixels' ground resolution cells. Using sequential indicator simulation, a set of equally probable maps are generated from which uncertainties regarding land-cover patterns are inferred. Collocated indicator cokriging, the geostatistical estimation method employed, explicitly accounts for the spatial cross-correlation between hard and soft data using a simplified model of coregionalization. The method is illustrated using a case study from southern Spain. Demonstrated uncertainties concern the areal extent of a contiguous olive region and the proportion of olive vegetation within large pixel blocks. As the image-derived olive data were not very informative, conditioning on hard data had a considerable effect on the area estimates and their uncertainties. For example, the expected areal extent of the contiguous olive region increased from 65 ha to 217 ha when conditioning on the reference sample.