Abstract
How areas of land are allocated for different uses, such as forests, urbanareas, and agriculture, has a large effect on the terrestrial carbon balance,and therefore climate change. Based on available historical data on land-usechanges and a simulation of the associated carbon emissions and removals, asurrogate model can be learned that makes it possible to evaluate the differentoptions available to decision-makers efficiently. An evolutionary searchprocess can then be used to discover effective land-use policies for specificlocations. Such a system was built on the Project Resilience platform andevaluated with the Land-Use Harmonization dataset LUH2 and the bookkeepingmodel BLUE. It generates Pareto fronts that trade off carbon impact and amountof land-use change customized to different locations, thus providing apotentially useful tool for land-use planning.