Gobal Forest Cover Mapping using Landsat and Google Earth Engine cloud computing

Abstract

Nowadays, the access of Landsat data-sets and the ever-lowering costs of computing make it feasible to monitor the Earth’s land cover at Landsat resolutions of 30 meter. However, the rapid forest-covered products on a large scale, such as intercontinental or global, is still challenging. By utilizing the huge catalog of satellite imagery as well as the high-performance computing capacity of Google Earth Engine, we proposed an automated pipeline for generating 30-meter resolution global-scale forest map from time-series of Landsat images, and a novel 30-meter resolution global forest map of 2018 is released. In this paper, we describe the methods to create products of forest cover at Landsat resolutions. First, we partitioned the landscapes into sub-regions of similar forest type and spatial continuity, thus maximizing spectral differentiation, simplifying classifier model and improving classification accuracy. Then, with the existing forest cover, which come from a variety of sources, a multi-source forest/non-forest sample set was established for machine algorithm learning training. Finally, a machine learning algorithm was used to obtain samples automatically, extract the characteristics of satellite images and establish the forest / non-forest classifier models. Taking the Landsat8 images in 2018 as a case, selecting satellite image features based on the study of forest reflectance, including onboard reflectivity, the index of forest vegetation and the texture features of each band, using established forest eco-zoning and multi-source forest / non-forest sample points, we realized automated learning and classification of forest cover for three initial zones. The accuracy verification of forest cover products in the three region was carried on two aspects: collecting verification points on high resolution satellite imagery (e.g. google earth), and cross-validating the current globally disclosed forest cover products. These two methods will illustrate the accuracy of the forest cover product.