Journal Article

Observational Constraints on Basin‐Scale Runoff: A Request for Both Improved ESMs and Streamflow Reconstructions

Douville, H.

2024

Overview

Efforts to predict long‐term changes in continental runoff at both global and basin scales generally remain ambiguous. Here we use a global runoff reconstruction and a Bayesian statistical method to narrow uncertainties in runoff projections from the latest generation of global climate models. Three representative tropical river basins are used to illustrate the application and showcase the potential for substantial reduction in modeling uncertainty. Yet, results are fairly sensitive to the selected reconstruction thus highlighting the need for reliable and homogeneized gridded runoff data sets or river discharge measurements. Moreover, climate models do not account for water withdrawals, whose effect on observed runoff should also be removed in order to detect and attribute the hydrological effect of climate change. Finally, and more importantly, most models fail at capturing the observed recent decrease in runoff ratio, which may highlight either model deficiencies or increasing water derivation over the selected river basins. Plain Language Summary The response of river discharge under the effect of climate change generally remains very uncertain. Bayesian statistical tools and global runoff reconstructions, constrained by flow observations, can nevertheless be used to evaluate the capacity of climate models to simulate the historical runoff and, thus, constrain its future changes at the basin scale. Three representative tropical river basins help illustrate the method. The results are nevertheless sensitive to the choice of the runoff reconstruction, thus emphasizing the need to have good quality flow data, if possible corrected for the direct effects of water withdrawals from the rivers or the aquifers which supply them. Worryingly, the latest generation of global climate models show a systematic underestimation of the downward evolution of the ratio between runoff and precipitation, which could reflect the increasing importance of these withdrawals or the inability of models to capture the rapidly increasing land surface evapotranspiration under climate change. Key Points Global 21st century projections of basin‐scale runoff remain highly model‐dependent within the latest generation of Earth System Models Gridded runoff reconstructions and precipitation observations can however be used to constrain the projections through Bayesian statistics A model‐observation mismatch is underlined regarding recent changes in the annual runoff to precipitation ratio over specific river basins

• Publisher: John Wiley & Sons, Inc,American Geophysical Union,Wiley
• Subject: Aquifers / Basins / Bayesian analysis / Bayesian statistics / Bayesian theory / Climate change / Climate effects / Climate models / Climate prediction / Constraints / Environmental Sciences / Evapotranspiration / Global climate / Global climate models / Hydrologic data / intermodel spread / Mathematical models / observational constraint / Physics / Probability theory / projections / Reconstruction / River basins / River discharge / River flow / Rivers / Runoff / runoff to precipitation ratio / Scale (ratio) / Statistical analysis / Statistical methods / Statistical models / Stream discharge / Stream flow / Uncertainty / Water discharge