Improved Land Surface Phenology Detection in China's Drylands and Associated Spatiotemporal Trends

This dataset is produced on the Google Earth Engine cloud computing platform using all valid annual MODIS observations (MCD43A4) after quality control. Start of Season (SOS) and Peak of Season (POS) are extracted using the NDPI vegetation index, combined with polynomial function fitting, winter masking, and the 10% relative threshold method. End of Season (EOS) is derived based on the EVI2 vegetation index, using self-weighting double logistic function fitting, winter masking, and the 70% relative threshold method. SOS, POS, and EOS values are finally generated for each pixel. A threshold of 0.0078 is used to mask non-vegetation background. Annual EVI amplitude data from 2001 to 2024 are masked to retain only pixels with EVI amplitude above 0.0078 in all years. This procedure ensures the accuracy and ecological representativeness of vegetation phenology extraction in arid regions and avoids omitting key ecological areas due to over-selection.