Fidelity of CMIP5 multi-model mean in assessing Indian monsoon simulations

Considering the wide use of the multi-model mean (MMM) on the seasonal time scale, this work examines its fidelity in simulating some important characteristics of the Indian summer monsoon using Coupled Model Intercomparison Project Phase 5 (CMIP5) simulations. It is noted that the MMM captures the observed spatial pattern and annual cycle of surface air temperature to a great extent, but there are large biases in magnitude, particularly over north India. For precipitation, only the broad-scale features are captured and extreme large biases, of magnitude equal or higher than the seasonal mean precipitation, exist in the MMM. The simulation of trends in seasonal mean temperatures and precipitation is even less satisfactory than the climatological means. Several precipitation features, for example, low-to-moderate intensity precipitation events, orography-related rain bands, extreme events, are noted to improve with increasing resolution of the models, whereas, no such improvement is noted for temperatures. It is also noted that the improvement in CMIP5 MMM is marginal if compared with the best performing model from the group of models considered for analysis. There are several models that show similar skill as MMM, and therefore could be alternatively used for future projections. Moreover, using such individual models for Indian monsoon projections will also help us to understand the underlying mechanisms and processes by conducting targeted numerical experiments, which would otherwise be highly limited by approaches like MMM. Therefore, targeted efforts to improve some of these better models are required to gain more confidence in future projections of Indian monsoon. Averaging output from various climate models (the multi-model mean, MMM) fails to capture key characteristics of the Indian monsoon. The MMM—which reduces bias associated with individual models—has been used extensively to investigate observed and projected changes in monsoon rainfall, upon which millions of people rely for water resources. Saroj Mishra and colleagues from the Indian Institute of Technology Delhi use the CMIP5 ensemble to test whether the MMM is able to represent broad-scale monsoon features. Several biases exist in the simulation of both temperature and precipitation. In particular, the magnitude of MMM rainfall is incorrect, low-to-moderate precipitation events are overestimated, while extreme events are underestimated. Thus, care must be taken when using the MMM to interpret projected changes in monsoon variability, and in some instances, analysis of individual models may be preferable.