Remotely-sensed quantification of covered areas and spatial resolution

Abstract

The basic remotely-sensed signal is a parameter defined on the soil surface. This signal, mathematically represented by a function of two space variables, is measured by sensors, at different resolutions. Many applications consist in counting pixels where the measured signal exceeds a given threshold and calculate the percentage of these pixels. This percentage is used to quantify the parameter in question. For example, the quantification of vegetation is often made in this way, the surface parameter being the NDVI. By doing so, one implicitly assumes that this percentage corresponds to the quantification of the parameter at the soil level, and, does not then depend on the sensor resolution. This paper shows theoretically and on synthetic images that this procedure can be completely wrong.