Metrics for Evaluating CMIP6 Simulation of Daily Precipitation Probability Distributions
Abstract
The probability of daily precipitation is one of the most relevant climate variables for human impacts, shaping local infrastructure, everyday life and culture at large. Global climate models (GCMs) are often used to study different aspects of daily precipitation, including future projections of its change under global warming. Efforts to characterize GCM performance should go beyond the mean and variance or selected percentiles and assess the whole probability distribution of daily precipitation. The performance of thirty-five Coupled Model Intercomparison Project Phase 6 (CMIP6) models in simulating daily precipitation probabilities is evaluated in comparison to two widely used observational data sets: TRMM-3B42 and Global Precipitation Climatology Project. Observed probability density functions are characterized by a power law range with exponent τP and a near exponential cutoff-scale PL, which controls the intensity of extremes in current climate. In addition, these probability density functions have zero interior peaks ---i.e., the lowest resolvable amount is the most probable daily precipitation value on wet days. Key metrics we investigate include model representation of fraction of wet days, power law exponent, cutoff-scale, shape of the extreme probability tail and number of spurious probability peaks. We find that most models tend to simulate a probability distribution shape more complex than observed, with spurious peaks in probability in many regions. The long-standing bias of models raining too often and too lightly persists in CMIP6. These errors occur primarily over oceans and may point to overly deterministic process representations in precipitation parameterizations. We argue that stochastic parameterizations may alleviate some of these problems.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2020
- Bibcode:
- 2020AGUFMGC1040002M
- Keywords:
-
- 3305 Climate change and variability;
- ATMOSPHERIC PROCESSES;
- 3333 Model calibration;
- ATMOSPHERIC PROCESSES;
- 1622 Earth system modeling;
- GLOBAL CHANGE;
- 1626 Global climate models;
- GLOBAL CHANGE