Global land cover dynamics and their drives observed at 30m resolution over 1985-2022

Land cover products record spatiotemporal dynamic information of the Earth's surface. With advancements in satellite remote sensing technology and rapid increases in computer storage and computational power, global land cover mapping has evolved from low-resolution to high-resolution products, and from single-epoch to long-term time-series products. However, most existing global land cover products exhibit considerable inconsistencies in mapping results between different epochs, leading to a substantial number of spurious changes. Based on the Google Earth Engine platform, we recently developed GLC_FCS30D, a global land cover dynamics monitoring product with fine classification at 30m dynamic based on the continuous change detection (CCD) algorithm and all available Landsat satellite imagery. GLC_FCS30D comprises 35 land-cover subcategories with 26 epochs, updated every five years prior to 2000 and annually thereafter. The GLC_FCS30D is validated using 84,526 globally distributed validation samples in 2020, and achieves an overall accuracy of 80.88% (±0.27%) at 10 major land-cover types. The global land-cover dynamics were clearly recorded in GLC_FCS30D, namely, the forest and cropland variations dominate global land cover change over past 37 years, and net loss of forests reaches about 2.5 million km2 and net gain in cropland area is approximately 1.3 million km2. Human activities have been dominating global land cover changes. Taking forest loss as an example, global forest loss was the primary drivers, accounting for approximately 79.99% of the total loss, while natural drivers contributed to the remaining 20.01%. Therefore, the GLC_FCS30D dataset represents a pioneering advancement in global land-cover monitoring, offering comprehensive insights into land cover dynamics at a 30-meter resolution spanning the period from 1985 to 2022.