Global tree-canopy cover over three decades (Q12210)

Losses and gains in canopy cover of the worlds tree canopies affect carbon stocks, species habitats, water cycles, and human livelihoods. Consistent and multi-decadal global data on tree-canopy cover dynamics are needed for modelling climate scenarios, tracking progress towards restoration targets, and diverse other research, management and policy applications. However, most data only map binary forest/non forest distinctions that are regionally restricted or biassed by data gaps, and those mapping tree-canopy cover are limited to the 21st century. Here, we present an annual and global time-series of tree-canopy cover between 1992 and 2018. To develop these data, we integrated complementary products, using their respective strengths to compensate for weaknesses, and exploiting path dependencies in change processes to derive predictions into the data-sparse 1990s. Our model validation indicates we can accurately map tree-canopy cover (r2=0.95 [0.01], RMSE=6.75% [0.08], F1-score=0.97 [0.0]) and our time-series agree well with national forest statistics (r2=0.94 [0.0]). This repository contains the Global Tree-Canopy Cover Change dataset (GTCCC), which consists of a global time-series on per-pixel tree-canopy covers estimated at a 300-m resolution between 1992 and 2018. The repository contains the following: GTCCC_canopyDensity.tar.gz - Annual GeoTiffs on per-piel tree-canopy cover estimates (EPSG:4326) GTCCC_uncertainty.tar.gz - Annual GeotiFFs with per-pixel estimates of the 95% confidence interval of the the predictions of each RFReg decision tree. GTCCC_change.tar.gz - Multiple outputs describing changes in tree-canopy cover between 1992 and 2018. The contents are described in a README.txt file found within. modelling_infrastructure.tar.gz - Infrastructure to generate the GTCCC dataset, including code, and some intermediary outputs, such as reference samples and the predictive model. The contents are described in a README.txt file found within. The predictors used to generate the GTCCC are provided separately given large volume, and can be reached by clicking here.