Sensitivity of the water cycle over the Indian Ocean and Maritime Continent to parameterized physics in a regional model

A regional model was used to simulate the water cycle over the Indian Ocean (IO) and Maritime Continent (MC). Sixteen 92 day simulations were performed using different combinations of eight cumulus parameterization schemes and three planetary boundary-layer (PBL) parameterization schemes. The strength of the water cycle in the IO and MC, measured by its domain mean precipitation and precipitable water, differs substantially among the simulations. The large spread of water cycle strength is mainly toward dry biases in comparison to global data assimilation products. The simulated water cycle, its spread, and biases differ between the IO and MC. Influences of PBL schemes can penetrate into the upper troposphere and those by cumulus schemes into the boundary layer. Dry biases in the simulations are produced mainly because of feedbacks among erroneously low diabatic heating peaks, shallow moisture convergence layers, dry lower troposphere, and weak surface evaporation. There is no single type of parameterization scheme that can be identified to be the main sources of the dry biases. It is the combination of errors from three types of parameterization schemes, namely, cumulus, PBL, and microphysics, that makes the simulated water cycle unrealistic. The lesson learned is that the tropical water cycle can be better simulated only by improving parameterization schemes of different processes all together as a package.