Sensitivity of ENVISAT ASAR image classification accuracy to spatial variability of selected environmental parameters in rural areas

摘要

Envisat images acquired in the alternating polarization mode are an important step forward in land use classification based on microwave satellite data. The application of ASAR images for crop recognition confirmed their applicability in vegetation mapping although a significant spatial variability of vegetation signatures is observed. The variability of signatures derived from ASAR image for a particular land use class is related to the inner variability of moisture and roughness. In order to specify more accurately the sources of such variability and to estimate the influence of variable parameters on the classification accuracy one needs detailed information concerning the investigated area.Valuable information concerning spatial distribution of moisture and roughness can be deduced from very high resolution (VHR) optical images acquired from IKONOS satellite. In order to investigate the influence of parameters deduced from optical images with the backscattering of the microwave signal the following images were examined: IKONOS image acquired on 12th of July, 2005 and two ASAR images acquired over the same area on 1 lth and 14th of July 2005. The analysis was conducted for a test site which is located in the western part of Poland in the traditional agriculture area. Arable land covers most of this area and at least 10 different types of crops can be found there. Other land use forms are: grasslands, forest, lakes and built-up area. The signatures of these classes were related to some spatial characteristics derived from IKONOS as for example: normalized vegetation index (NDVI), row direction in crop fields and spatial distribution of soil moisture.