Skip to main content
U.S. flag

An official website of the United States government

Constraining CMIP runoff projections over global river basins

Presentation Date
Friday, December 15, 2023 at 5:10pm - Friday, December 15, 2023 at 5:20pm
MC - 3003 - West



The runoff projections from Earth System Models (ESMs) are increasingly being utilized for future water resource risk assessments. However, such projections for mid-21st century (2030-2070) based on the CMIP models are still highly uncertain, in part due to differences in the land process representation among ESMs and the generally scant validation of runoff in coupled ESM development. The land processes related to runoff projections can be statistically emulated by quantifying the observed annual sensitivity of runoff to temperature (temperature sensitivity) and precipitation (precipitation sensitivity) using multiple linear regression. The future runoff changes can be then predicted from the temperature and precipitation projections by the linear regression models with corresponding sensitivities. By substituting the observational sensitivities for the predictions while retaining temperature and precipitation projections from ESMs, the ESM runoff projections can be constrained. Here, we present observational constraints on runoff projections across a large set of global river basins. The possible impacts from the internal variability, non-stationarity, model generation, and observational uncertainty on the runoff sensitivities are incorporated to obtain statistically robust constraints. As a result, runoff projections in 36 among 81 global river basins are constrained for mid-21st century (2030-2070) both in CMIP5 (RCP45) and CMIP6 (SSP245) ensembles. Most of the largest deviations in ESM projections versus the observational constraints are related to model biases in temperature sensitivity, possibly due to issues with evapotranspiration processes. We will discuss the possible causes of temperature sensitivity biases in representative regions. The observational constraint and associated metrics are slated to be implemented in NOAA’s Model Diagnostics Task Force metrics package.

Acknowledgement: This work is supported by NOAA MAPP award NA21OAR4310349.

Funding Program Area(s)