Vplyv Priestorového Rozlísenia Satelitných Záznamov na Správnost Urcenia Výmery: Príkladová Stúdia pre Lesné Tazby - The Impact of the Spatial Resolution of Satellite Imagery on the Accuracy of Area Mapping: a Case Study for Clear Cutting of Forest

Abstract

In this study we analyse how the spatial resolution of satellite imagery affects the accuracy of area estimates for a given land cover category. Whenever this category covers only a portion of the pixel, one of the following situations occurs: (i) when the pixel is classified correspondingly a commission error is introduced, (ii) when the pixel is classified differently, an omission error is generated. Any attempt of reducing one of these errors will increase the other, representing a problem of two conflicting objectives. We use the Pareto optimality concept to solve this conflicting situation. Pareto optimal means that it is not possible to further reduce error of commission (or omission) without increasing of error of omission (or commission). We applied this concept for forest clear cut area mapping from medium spatial resolution data from MODIS (250 m) in the boreal forest in Karelia, using Landsat ETM+ data as a reference i.e. clear cut maps at 30 m spatial resolution. In order to compute the unavoidable omission and commission errors, representing the Pareto optimal solutions, the 30 m high resolution map was aggregated the coarser resolution of MODIS and the proportion of clear cut inside the coarse pixels was calculated. From the comparison of the MODIS classification with the Pareto optimal classifications we can conclude that the error is mainly influenced by the coarse spatial resolution and only to a certain extent could be improved by better performance of the classification algorithm.