A global land cover training dataset from 1984 to 2020.

State-of-the-art cloud computing platforms such as Google Earth Engine (GEE) enable regional-to-global land cover and land cover change mapping with machine learning algorithms. However, collection of high-quality training data, which is necessary for accurate land cover mapping, remains costly and labor-intensive. To address this need, we created a global database of nearly 2 million training units spanning the period from 1984 to 2020 for seven primary and nine secondary land cover classes. Our training data collection approach leveraged GEE and machine learning algorithms to ensure data quality and biogeographic representation. We sampled the spectral-temporal feature space from Landsat imagery to efficiently allocate training data across global ecoregions and incorporated publicly available and collaborator-provided datasets to our database. To reflect the underlying regional class distribution and post-disturbance landscapes, we strategically augmented the database. We used a machine learning-based cross-validation procedure to remove potentially mis-labeled training units. Our training database is relevant for a wide array of studies such as land cover change, agriculture, forestry, hydrology, urban development, among many others.

Spatiotemporal tracking of carbon emissions and uptake using time series analysis of Landsat data: A spatially explicit carbon bookkeeping model.

Reducing terrestrial carbon emissions to the atmosphere requires accurate measuring, reporting and verification of land surface activities that emit or sequester carbon. Many carbon accounting practices in use today provide only regionally aggregated estimates and neglect the spatial variation of pre-disturbance forest conditions and post-disturbance land cover dynamics. Here, we present a spatially explicit carbon bookkeeping model that utilizes a high-resolution map of aboveground biomass and land cover dynamics derived from time series analysis of Landsat data. The model produces estimates of carbon emissions/uptake with model uncertainty at Landsat resolution. In a case study of the Colombian Amazon, an area of 47 million ha, the model estimated total emissions of 3.97 ± 0.71 Tg C yr-1 and uptake by regenerating forests of 1.11 ± 0.23 Tg C yr-1 2001-2015, with an additional 45.1 ± 7.99 Tg of carbon remaining in the form of woody products, decomposing slash and charcoal at the end of 2015 (estimates provided with ±95% confidence intervals). Total emissions attributed to the study period (including carbon not yet released) is 6.97 ± 1.24 Tg C yr-1. The presented model is based on recent technological advancements in the field of remote sensing to achieve spatially explicit modeling of carbon emissions and uptake associated with land surface changes and post-disturbance landscapes that is compliant with international reporting criteria.