USE OF GROUNDTRUTH DATA TO CORRECT LAND-COVER AREA ESTIMATES FROM REMOTELY-SENSED DATA

Estimates of area by land cover category from photo interpretation are subject to bias arising from interpretation error. Quantification of this error by groundtruth survey allows corrections to be made to the estimates. The precision of estimates of interpretation error for a fixed cost depends on the survey design; simple survey designs are inefficient for this purpose. The penalty for using more complex designs is that a sophisticated correction method is required. We note the assumptions of existing correction methods, develop a more general method, and show how the precision of corrected estimates of area may be quantified. We demonstrate the effect of 'correcting' area estimates under different methods, and provide an example in which use of an inappropriate method leads to substantial bias (up to 40 per cent).