PNNL

The Science                                       

Understanding what controls rainfall in tropical regions is important for accurate climate modeling. A team examined radar-derived precipitation data from three tropical field campaigns and the corresponding large-scale environmental variables from global reanalysis data to quantify the roles of various environmental factors in the transition from a suppressed to an active rainfall regime. They identified the variability of environmental factors that explain the difference between the rainfall statistics in the studied regions. They developed a simple machine learning (ML) model that predicts the probability of a transition from a suppressed to an active precipitation regime as a function of five large-scale environmental variables. The work demonstrates a potential application of the ML model as a trigger function, a set of conditions used to determine whether convection should be activated, for climate model convection parameterizations.

The Impact

A comparative and quantitative insight into the environmental factors controlling precipitation characteristics and differences among tropical regions is essential for better understanding rainfall statistics and their representations. This will ultimately lead to improved representations in global and regional models. This work demonstrates how marginal statistical analysis can be used in tandem with ML models to obtain a quantitative and qualitative understanding of the relationship between rain clouds and the large-scale environment. The ML model captures several qualitative relationships obtained from a statistical analysis of frequencies. This work can easily be generalized to account for additional environmental variables and to include other regions.

Summary

A quantitative insight into the environmental factors that control rainfall characteristics and regional variation is needed to better understand rainfall statistics and extremes. Such information is also important for accurately representing rainfall in global and regional models. Using statistical analysis and an ML model, researchers examined the roles of different environmental variables in the transition from suppressed to active precipitation regimes. They documented the origins of differences in between environmental conditions and radar observed precipitation regime statistics across three tropical regions, the Amazon, northern Australia, and equatorial Indian Ocean. For widespread rain to start in the three tropical areas, an abrupt increase in the growth rate of precipitable water (PW, the total water in the atmospheric column) ~ 60 mm and convective inhibition (CIN, the amount of buoyant energy required to enable convection) <100J kg-1 is needed. Differences in precipitation among the three areas can be primarily explained by regional differences in CIN statistics and, to a lesser extent, PW. High CIN and low PW are comparatively common over the Amazon, while high CIN is less frequent over the equatorial Indian Ocean. Darwin, Australia has the most frequent active regimes with high PW and moderate CIN.

PNNL Contact

Jerome Fast, Pacific Northwest National Laboratory, Jerome.Fast@pnnl.gov

Funding

This study is supported by the Department of Energy Office of Science’s Biological and Environmental Research program as part of the Atmospheric System Research program area through the Integrated Cloud, Land-Surface, and Aerosol System Study scientific focus area.