Mapping of arable land in Russia using multi-year time series of MODIS data and the LAGMA classification technique

Abstract

ABSTRACTThe sustainable agriculture requires a regular country-wide update of information on the status and extension of arable land in Russia. The arable land mapping method is developed based on multi-year time series of Moderate Resolution Imaging Spectroradiometer (MODIS) data. The method exploits differences between the intra- and inter-annual changes in the spectral reflectance of arable land and the corresponding changes for other land cover types. It involves a set of satellite data-derived phenological metrics generated using a 6 years long time series of the perpendicular vegetation index (PVI). The approach utilizes the Locally Adaptive Global Mapping Algorithm (LAGMA), which is a supervised classification technique accounting for the spatial variability of intra-classes spectral properties. The method has been applied to produce a uniform time series of comparable annual arable land maps for Russia at 250 m spatial resolution for the years 2005–2013. Countrywide arable land area trends over the above time series were found to be consistent with official statistics (ROSSTAT).The mapping result has been evaluated using reference data providing F-score exceeding 80% for the most productive regions.