Saskia Förster, Klaus Kaden, Michael Förster, Sibylle Itzerott

• Spatially explicit multi-year crop information is required for many environmental applications. The study presented here proposes a hierarchical classification approach for per-plot crop type identification that is based on spectral-temporal profiles and accounts for deviations from the average growth stage timings by incorporating agro-meteorological information in the classification process. It is based on the fact that each crop type has a distinct seasonal spectral behavior and that the weather may accelerate or delay crop development. The classification approach was applied to map 12 crop types in a 14,000 km(2) catchment area in Northeast Germany for several consecutive years. An accuracy assessment was performed and compared to those of a maximum likelihood classification. The 7.1% lower overall classification accuracy of the spectral-temporal profiles approach may be justified by its independence of ground truth data. The results suggest that the number and timing of image acquisition is crucial to distinguish crop types. TheSpatially explicit multi-year crop information is required for many environmental applications. The study presented here proposes a hierarchical classification approach for per-plot crop type identification that is based on spectral-temporal profiles and accounts for deviations from the average growth stage timings by incorporating agro-meteorological information in the classification process. It is based on the fact that each crop type has a distinct seasonal spectral behavior and that the weather may accelerate or delay crop development. The classification approach was applied to map 12 crop types in a 14,000 km(2) catchment area in Northeast Germany for several consecutive years. An accuracy assessment was performed and compared to those of a maximum likelihood classification. The 7.1% lower overall classification accuracy of the spectral-temporal profiles approach may be justified by its independence of ground truth data. The results suggest that the number and timing of image acquisition is crucial to distinguish crop types. The increasing availability of optical imagery offering a high temporal coverage and a spatial resolution suitable for per-plot crop type mapping will facilitate the continuous refining of the spectral-temporal profiles for common crop types and different agro-regions and is expected to improve the classification accuracy of crop type maps using these profiles.…