Emulation of high-resolution land surface models using sparse Gaussian processes with application to JULES

Land surface models are typically integrated into global climate projections, but as their spatial resolution increases the prospect of using them to aid in local policy decisions becomes more appealing. If these complex models are to be used to make local decisions, then a full quantification of uncertainty is necessary, but the computational cost of running just one full simulation at high resolution can hinder proper analysis. Statistical emulation is an increasingly common technique for developing fast approximate models in a way that maintains accuracy but also provides comprehensive uncertainty bounds for the approximation. In this work, we developed a statistical emulation framework for land surface models, enabling fast predictions at a high resolution. To do so, our emulation framework acknowledges, and makes use of, the multitude of contextual data that are often fed into land surface models (sometimes called forcing data, or driving data), such as air temperature or various soil properties. We use The Joint UK Land Environment Simulator (JULES) as a case study for this methodology, and perform initial sensitivity analysis and parameter tuning to showcase its capabilities. The JULES is perhaps one of the most complex land surface models and so our success here suggests incredible gains can be made for all types of land surface model.