Skip to main content
U.S. flag

An official website of the United States government

Evaluating hydrologic sensitivity in CMIP6: internal variability versus anthropogenic forcing

Presentation Date
Tuesday, December 8, 2020 at 7:08am



In a warming climate, global-mean precipitation increases at a rate known as “hydrologic sensitivity” (HS), i.e., the percentage increase per degree of warming. HS varies across CMIP6 models between about 1 and 2 %/K in response to anthropogenic forcing, i.e., in ScenarioMIP, 1pctCO2, and abrupt4xCO2 simulations. However, some observational studies have found that global-mean precipitation has increased at a greater rate, closer to the rate of water-vapor increase, about 7%/K. The observational record is strongly distorted by internal variability, from which HS may also be diagnosed in the absence of anthropogenic forcing. This internal variability is most easily examined in the piControl simulations in which CO2 levels are kept to pre-industrial levels for multi-centennial simulations. In these simulations, years with positive global-mean temperature anomalies also exhibit positive global-mean precipitation anomalies, and similarly for negative anomalies, implying that HS can be diagnosed from internal variability. We find that in almost all models HS is greater in the internal-variability than anthropogenic-forcing case, which explains why HS appears to be greater in the observational record. We also find that models that exhibit greater HS in the internal-variability case tend to exhibit greater HS in the anthropogenic-forcing case. This implies that the CMIP6 spread in HS in response to anthropogenic forcing may be constrained by evaluating HS in the internal-variability case against observations. However, in the internal-variability case HS is highly variable, both in a given model and in the observational record. In particular, inter-decadal variability in HS during piControl simulations exceeds the intermodel spread in HS in anthropogenic-forcing simulations.Thus, in the internal-variability case, there is a probability distribution of HS that depends on the drivers of global-mean temperature anomalies. We discuss this probability distribution, identifying models that represent it more realistically, and how this relates to HS in the anthropogenic-forcing case.

Funding Program Area(s)