Uncertainties in land use data

摘要

This paper deals with the description and assessment of uncertainties inland use data derived from Remote Sensing observations, in the context ofhydrological studies. Land use is a categorical regionalised variablereporting the main socio-economic role each location has, where the role isinferred from the pattern of occupation of land. The properties of thispattern that are relevant to hydrological processes have to be known withsome accuracy in order to obtain reliable results; hence, uncertainty inland use data may lead to uncertainty in model predictions. There are twomain uncertainties surrounding land use data, positional and categorical. Thefirst one is briefly addressed and the second one is explored in more depth,including the factors that influence it. We (1) argue that the conventionalmethod used to assess categorical uncertainty, the confusion matrix, isinsufficient to propagate uncertainty through distributed hydrologic models;(2) report some alternative methods to tackle this and otherinsufficiencies; (3) stress the role of metadata as a more reliable means toassess the degree of distrust with which these data should be used; and (4)suggest some practical recommendations.