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

Abstract

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

Keywords

• Cloud computing
• global monitoring
• Google Earth Engine (GEE)
• land surface phenology (LSP)
• threshold method