Sensitivity of systematic biases in South Asian summer monsoon simulations to regional climate model domain size and implications for downscaled regional process studies

Global climate models (GCMs) have good skill in simulating climate at the global scale yet they show significant systematic errors at regional scale. For example, many GCMs exhibit significant biases in South Asian summer monsoon (SASM) simulations. Those errors not only limit application of such GCM output in driving regional climate models (RCMs) over these regions but also raise questions on the usefulness of RCMs derived from those GCMs. We focus on process studies where the RCM is driven by realistic lateral boundary conditions from atmospheric re-analysis which prevents remote systematic errors from influencing the regional simulation. In this context it is pertinent to investigate whether RCMs also suffer from similar errors when run over regions where their parent models show large systematic errors. Furthermore, the general sensitivity of the RCM simulation to domain size is informative in understanding remote drivers of systematic errors in the GCM and in choosing a suitable RCM domain that minimizes those errors. We investigate Met Office Unified Model systematic errors in SASM by comparing global and regional model simulations with targeted changes to the domain and forced with atmospheric reanalysis. We show that excluding remote drivers of systematic errors from the direct area of interest allows the application of RCMs for process studies of the SASM, despite the large errors in the parent global model. The findings in this study are also relevant to other models, many of which suffer from a similar pattern of systematic errors in global model simulations of the SASM.