This study investigates the causes of pronounced low precipitation bias over Amazonia in the Community Atmosphere Model version 5 (CAM5), a common feature in many global climate models including DOE’s E3SM. The analysis is based on a suite of 3-day long hindcasts starting every day at 00Z from 1997 to 2012, and an AMIP simulation for the same period. The goal is to improve the understanding of the rainfall biases at a process level by linking the systematic model biases to potential deficiencies in model parameterizations.
Amazonia’s rainforest plays a major role in the global water and carbon cycles. Nevertheless, most contemporary global climate models have difficulty in simulating precipitation over Amazonia. A better understanding of the causes of precipitation biases could reduce the uncertainty in predicting future changes in global water and carbon cycles.
The Amazonia dry bias appears by the second day in the hindcasts and is very robust for all the seasons with the largest bias magnitude during the wet season (Dec-Feb). The bias pattern and magnitude do not change much during different dynamical wind regimes on sub-seasonal time scales. The diurnal cycle of precipitation near the Large-scale Biosphere-Atmosphere Experiment (LBA) sites from observations and hindcasts was further classified into three convective regimes: no precipitation, late afternoon deep convection, and nighttime deep convection. CAM5 can only simulate the late afternoon convective regime and completely fails to simulate the nighttime convection, which is mostly from propagating convective systems originating from remote locations. CAM5 mainly underestimates precipitation in the late afternoon and nighttime convective regimes, which occur during ~67% of wet season days and account for ~75% of accumulated precipitation amount in observations. The persistent warm temperature bias and slightly higher moisture below 850 mb likely trigger deep convection too frequently, resulting in an earlier but weaker rainfall peak in the diurnal cycle. Furthermore, shallow convection may not effectively transport moisture from the boundary layer to the free atmosphere, which also leads to weaker deep convection events.