A global time series dataset to facilitate forest greenhouse gas reporting

Abstract

We have developed a version of the Continuous Change Detection and Classification algorithm within the Google Earth Engine environment. It has been used with 20 years of Landsat data (1999-2019) to produce a new, publicly available global dataset of pre-computed time series break points and harmonic coefficients. We present results from regional use cases demonstrating classification and change detection with this new dataset and compare them to other temporal compositing techniques. Our results demonstrate that gains in overall accuracy using CCDC may be small on a yearly basis, but they are consistent, and improvements in temporal coherence—correctly detecting land use transitions and temporal trends—can be significant. These improvements can translate into better estimates of land use change activity and reduce the uncertainty in the greenhouse gas emissions estimates in REDD+ reporting.