Precipitation-buoyancy relationships in CMIP6 Models

Abstract:

Differing treatments of convection are among the largest contributors of spread and uncertainty among climate models. To better understand the source of these uncertainties we present process-oriented diagnostics based on the empirical precipitation-buoyancy relationship for tropical rainfall. This relationship combines the observed temperature and moisture influences on precipitation within a single metric that measures the lower-tropospheric buoyancy of a bulk plume. Physically meaningful decomposition of this buoyancy metric yields two components: a measure of the lower-tropospheric Convectively Available Potential Energy (CAPE) and a lower tropospheric subsaturation measure (SUBSAT). Rainfall observations from the Tropical Rainfall Measuring Mission (TRMM) show a strong pickup when the buoyancy measure—as computed from reanalyses—exceeds a critical value. When viewed in CAPE-SUBSAT space, the 2D precipitation surface shows a pickup when the critical buoyancy value in crossed from either the CAPE or SUBSAT directions. Corresponding precipitation-buoyancy relationships are constructed for a suite of models from the Coupled Model Intercomparison Project Phase 6 (CMIP 6). When compared to the observational baseline, CMIP6 models exhibit considerable diversity in their precipitation-buoyancy statistics. The best performing models show a pickup in both buoyancy and CAPE-SUBSAT spaces, while the less well performing models do not show a pickup in either spaces. A scalar metric is used to condense information about the sensitivity of model precipitation to changes in CAPE versus SUBSAT. This scalar metric is also shown to capture information about the variability of CAPE and SUBSAT, with models that exhibit excessive moisture sensitivity showing diminished variability in the CAPE direction. The buoyancy-based diagnostics presented here show promise in evaluating the set of assumptions that link model convection to its environment, diagnosing convection-related biases in the large-scale thermodynamic fields, and making recommendations for targeted parameter/structural changes in model convection schemes.