Characterisation of land surface phenology and land cover based on moderate resolution satellite data in cloud prone areas - A novel product for the Mekong Basin

- A phenological-specific EVI time series was processed for 11 years of MODIS data.
- Spatial continuous phenological characteristics of the land surface were derived.
- A regional-tuned classification was applied on spectral and temporal information.
- First regional-specific land cover map for the entire Mekong Basin was generated.
- Derived land cover map shows higher thematic accuracy to existing products.

Information on vegetation phenology and land cover for the area of the Mekong Basin are derived for the year 2010, based on optical satellite data from the Moderate Resolution Imaging Spectroradiometer (MODIS). On account of almost persistent cloud cover in the rainy season, the application of optical remote sensing data presents a challenging exercise and demands special data processing and classification methods. An Enhanced Vegetation Index (EVI) time series is produced based on data from both platforms AQUA and TERRA, with dropouts and noise being effectively reduced by applying an adaptive Savitzky-Golay filter. Thereby, different parameterisation is applied dependent on the number of harvest cycles, which are defined by an initial harmonic analysis, based on an EVI time series of 5 years. Information on land cover is derived by a multistep unsupervised classification approach optimised for regions with frequent cloud cover, based on multispectral monthly and seasonal composites, amplitude and phase information from a 11-year EVI time series (2001-2011), and phenological metrics from 2010. Moreover, the environmental heterogenic conditions in the region are addressed by a regionally tuned clustering approach based on physiographic subregions and the use of auxiliary geodata. The results obtained demonstrate that the adapted approach performs satisfactory in terms of accuracy under given conditions.