Using NASA remote-sensing data to constrain forest-related land-use forcings in global carbon-climate models

As part of the CMIP5 experiments, many climate models incorporated new gridded products of land-use and land-use change that were harmonized to ensure a continuous transition from the historical to the future data in a consistent format for all models (Hurtt et al. 2011). These Land-Use Harmonization (LUH) data products include all annual transitions between cropland, pasture, urban and natural land, including wood harvest, shifting cultivation, and the recovery of secondary vegetation at half-degree (fractional) spatial resolution; these transitions are then used by climate models to compute the climate forcings associated with both historical and future land-use change. The next generation of LUH products is currently under development and has been selected as an official required forcing dataset for upcoming CMIP6 experiments. At the same time, satellite remote sensing of the terrestrial biosphere has also evolved; new NASA remote-sensing-based maps of global forest extent and change are now available (Hansen et al. 2013) and can be used as an added constraint in the LUH process. Harmonizing this remote sensing data with the LUH data generates new gridded maps of land-use transitions that are based on, and consistent with, observations of actual forest cover change and which can be used in coupled carbon-climate simulations to improve the representation of land-use related climate forcings. This is a major computational challenge involving 143 billion 30m Landsat pixels, and the simulation of over 20 billion LUH unknowns, in addition to challenges arising from the differing treatment of forested area in the observed and simulated data. The incorporation of this remote sensing data has lead to the development of new algorithms to better represent anthropogenic forest disturbance in the LUH, particularly for shifting cultivation and wood harvesting which are key sources of spatial uncertainty in the LUH forest extent and change maps. This not only further improves fidelity with the remote sensing data during the contemporary time period, but also improves the LUH historical reconstructions and future projections. These algorithms will be a key feature of the new LUH datasets being produced as part of the LUMIP for the CMIP6 experiments.