Evaluating crop phenology retrieving accuracies based on ground observations

Crop phenological information is an important parameter for crop growth monitoring, grain yield prediction, crop model simulation and crop’s response to climate change. Improving the accuracy of the retrieved crop phenology parameters contributes to researches about climate change, global carbon balance, etc. This paper focuses on assessing the retrieval accuracy of crop SOS and EOS by remote sensing based on the dynamic threshold model. Ground observations of crop growth and development records from China Meteorological Administration (CMA) and Chinese Ecosystem Research Network (CERN) in 2015 and 2016 were used as reference data. Firstly, we improved the dynamic threshold model to ensure the 100% retrieval rate for detecting SOS and EOS. Then, we retrieved the SOS and EOS of different crops under different thresholds by the improved dynamic threshold model from the Normalized Difference Vegetation Index (NDVI) time series derived from MODerate-resolution Imaging Spectroradiometer (MODIS). Accuracy assessment indicated that the mostly used 20% or 50% threshold is not the optimal threshold for retrieving all crops’ SOS and EOS. In additional, it is inappropriate to use the same threshold to retrieve SOS and EOS. There is a big difference between the optimal thresholds for retrieving SOS and EOS of different crops.