Towards integration of LCLM feedbacks within climate models: an emulator approach

Land Cover and Land Management (LCLM) changes display complex interactions with climate conditions, and can in particular modulate regional-scale extreme climate events such as heat waves and droughts. Relatedly, the potential of LCLM for not only climate mitigation but also adaptation has been underlined; it could thus play a role in achieving the corresponding goals of the Paris Agreement. It is thus essential to account for LCLM processes and their climate feedbacks within climate models, in order to inform land-use scenarios that help comply with climate and broader environmental objectives in a comprehensive manner. Emulators represent a computationally cheap but effective way of approximating climate models with an added advantage of agility in scenario exploration. Here we outline a Generalised Additive Model (GAM) based emulator approach to represent LCLM-Climate feedbacks simulated in Earth System Models (ESMs). The emulator is to be used in the LAnd MAnagement for CLImate Mitigation and Adaptation (LAMACLIMA) project, and is trained on dedicated ESM simulations which isolate the effects of de/afforestation, wood harvest and irrigation.

We showcase the emulator’s ability to represent local, monthly surface temperature responses to de/afforestation using input variables of tree cover change, longitude, latitude and orography. Spatial cross-validation is used to fit and tune the emulator, thus considering spatial autocorrelations within the training material. The resulting emulator can be used to estimate surface temperature changes over a major part of the globe and for a variety of possible tree cover changes. Such also enables us to identify the geographical areas and types of tree cover changes which are of high uncertainty within the emulator. This provides us with valuable insight into the additional ESM simulations that would be required to improve its representation of temperature responses to de/afforestation. Extending this framework to wood harvest and irrigation could then provide more clarity on the uncertainties underlying LCLM-Climate feedbacks as represented within ESMs.