Are model-data differences in the Indo-Asian monsoon due to model or data biases?

Abstract:

Precipitation biases over Southern Asia persist in atmospheric general circulation models (AGCM) when compared to various observational data sets. It has been hypothesized that high-resolution modeling is key to reducing these biases. Here we compared a suite of high to low-resolution simulations (~0.25 to ~ 2.0) using Community Earth System Model version 1.0 (CESM1) to show that these features are consistent across the range of resolutions.

We find that the general spatial distribution of precipitation over the Arabian Sea, the Bay of Bengal, and Central Asia stay relatively consistent as we increase model resolution. For instance, orographic precipitation over the Himalayan front and the exceedingly strong signal from the Somali Jet are the two most prominent and robust features we find in the model. Furthermore, we used the CESM’s recent modified AGCM, Community Atmospheric Model version 5 (CAM5) with two model resolutions (1.0 and 2.0). Despite the improved formation of cloud condensation nuclei through the new parameterization of direct and indirect aerosol interactions, we find that the features mentioned above continually persist throughout both simulations.

Both orographic precipitation over the Himalayan range and the Somali jet are slightly improved in our ~0.25 simulation, although the zonal average profile of precipitation over Asia consistently shows two maximum peaks one near the tropics and another in the subtropics that are not seen in both observational data sets Global Precipitation Climatology Project (GPCP) and Tropical Rainfall Measuring Mission (TRMM). We also find the magnitude of these peaks to be greater than the observations. However, hi-resolution observations also show different precipitation patterns especially over the Arabian Sea and Central Asia. To evaluate the sensitivity of the model-data mismatch, we will utilize hi-resolution rain gauge (APHRODITE) and satellite observation (CMAP) data sets, and discuss the implication of these different data sets for the interpretation of model accuracy.