Biological and Environmental Research - Earth and Environmental System Sciences
Earth and Environmental System Modeling
15 May 2017

Regional Downscaling of Subseasonal-to-Seasonal (S2S) Prediction: Past Lessons and Future Prospects

Scientists identified an opportunity for improving S2S prediction of summer precipitation through regional, convection-permitting modeling of land-atmosphere interactions.


Small errors in representing tropical convection in global climate models can quickly escalate to large errors in extra-tropical (mid-latitude) weather patterns. Even with dynamical downscaling of global subseasonal-to-seasonal (S2S) predictions, forecasting summer precipitation remains challenging because of model limitations in representing mesoscale convection associated with large, slow-moving storms. Scientists at the U.S. Department of Energy’s Pacific Northwest National Laboratory found that regional, convection-permitting models that better simulate the different effects of isolated-versus-mesoscale organized convection on soil moisture and its feedbacks on clouds and precipitation may improve S2S prediction of summer precipitation in the United States.


A brief survey of previous S2S prediction efforts found particularly limited skill for summer precipitation. With high-performance computing, regional, convection-permitting modeling—that better captures mesoscale organized convection and land-atmosphere interactions—is more feasible. This type of modeling also offers an opportunity to improve S2S prediction skill. Because precipitation has an important influence on soil moisture and streamflow, better S2S precipitation prediction has significant value for resource planning and management.


In the Northern Hemisphere winter, global climate models have demonstrated better prediction of large-scale circulation. This factor motivated the MultiRCM Ensemble Downscaling (MRED) project to explore the value of dynamical downscaling of seasonal climate forecasting over the United States using multiple regional climate models (RCMs) for the cold season. Comparison of the global and regional forecasts showed some improved skill in the regional forecast. Researchers did not attempt a warm-season experiment in the MRED project, partly because RCM simulation skill for warm-season precipitation is sensitive to model resolutions and physics representations. Scientists performed modeling experiments using a regional, convection-permitting model in which convection is explicitly resolved. They found that the convection-permitting model better simulated isolated and mesoscale organized convection and their diurnal (sub-daily) variability compared with RCMs. This method allowed the impacts of precipitation on soil moisture and the subsequent feedbacks on clouds and precipitation to be more realistically captured in the convection-permitting simulations. Because soil moisture has a longer memory beyond the weather timescale and it influences precipitation, improved simulation of soil moisture may advance the S2S prediction skill for summer precipitation. Better simulations of soil moisture also could lead to more useful forecasts for resource planning and management.

L. Ruby Leung
Pacific Northwest National Laboratory (PNNL)
Leung, L, and Y Gao.  2016.  "Regional Downscaling of S2S Prediction: Past Lessons and Future Prospects."  US CLIVAR Variations 14(4): 13-18.