A Threshold Method for Robust and Fast Estimation of Land-Surface Phenology Using Google Earth Engine

Cloud-based platforms are changing the way of analyzing remotely sensed data by providing high computational power and rapid access to massive volumes data. Several types studies use cloud-based for global-scale analyses, but number land-surface phenology (LSP) that is low. We analyzed performance state-of-the-art LSP algorithms propose a new threshold-based method we implemented in Google Earth Engine (GEE). This method, called maximum separation (MS) applies moving window estimates ratio observations exceed given threshold before after central day. The start end growing season days year when difference between ratios day minimal maximal. MODIS metrics estimated with MS showed similar performances as traditional methods compared ground estimations derived from PhenoCam dataset, network digital cameras provides near-surface vegetation phenology. main advantage it can be directly applied daily nonsmoothed time series without any additional preprocessing steps. implementation proposed GEE allowed processing global phenological maps MODIS. distribution code allows reproducibility results scientific community.