Statistical downscaling uses equations to convert global-scale output to regional-scale conditions. Instead of maintaining a dynamic climate model at the higher resolution of a region, this approach uses a series of equations to relate climate fluctuations in global climate models (GCMs) to the finer scale at the regional level.
Because the statistical approach requires less computational effort than dynamical downscaling, it offers the opportunity for testing scenarios for many decades or even centuries rather than the brief “time slices” of most dynamical downscaling efforts.
Advantages of statistical downscaling also include the opportunity to use “ensemble” GCM results. Modelers recommend using results from an ensemble of numerous GCMs rather than a single model because ensemble simulations tend to match overall observations better than results from any individual model. However, because ensemble results average the projections from many models, the variability of ensemble projections is dampened compared to that of individual models.
With statistical downscaling, the ensemble average for a region can be applied using equations that relate the larger-scale observations to regional climate parameters. So, for instance, a northern shift in the subtropical jet stream at the global scale may translate into more winter precipitation in one region and less in another. Equations based on past observations of how the jet stream position affects local precipitation then provides a basis for interpreting how the shift might affect the snow on specific mountain ranges and perhaps even the amount of water likely to reach local rivers.
For instance, Niklas Christensen and Dennis Lettenmaier (2006) used statistical downscaling to consider how Colorado River streamflow might change with climate. Their detailed analysis considered which high-elevation sites might remain cold enough to retain mountain snowpack, despite the projected warming, which in turn required a downscaling of the projected temperature rise to a scale fine enough to distinguish how temperatures drop as one ascends a mountain. From here, the researchers were able to analyze the sensitivity of river flow to the timing of precipitation changes. When their input shifted the timing of precipitation slightly from winter to summer, their model showed a 16 percent decline in river flow on average this century. But when they used models that showed a slight shift from summer to winter precipitation, the decline amounted to about half of that, with an 8 to 11 percent decrease by the end of the century.
Many hydrologic models require input of climate variables at daily time steps or even 6-hourly time steps, so the need for downscaling in time also arises. So-called “weather generators” (see Gangopadhyay et al. 2005) use historical weather data that are resampled into daily or hourly values according to the conditions projected by the climate model. Similar resampling techniques can be applied to other climate topics of interest, such as river flow that is consistent with both the historical variability (Prairie et al. 2006) and the climate model average projections.
Statistical downscaling often involves “bias removal,” the correction of factors inaccurately modeled by GCMs. Many models overestimate precipitation in the western United States, for instance, on the order of a millimeter a day, or roughly an inch a month (Bader et al. 2008). Statistical downscaling techniques would typically correct that bias before modeling future rainfall (Salathé 2005).
Adapted for eXtension.org by Melanie Lenart, University of Arizona
Bader, D.C., C. Covey, W.J. Gutowski, I.M. Held, K.E. Kunkel, R.L. Miller, R.T. Tokmakian, and M.H. Zhang, 2008. Climate Models: An Assessment of Strengths and Limitations. A Report by the U.S. Climate Change Science Program and the Subcommittee on Global Change Research. Department of Energy, Office of Biological and Environmental Research, Washington, D.C. Available at: http://www.sc.doe.gov/ober/sap3-1-final-all.pdf
Christensen, N., and D.P. Lettenmaier, 2006. A multimodel ensemble approach to assessment of climate change impacts on the hydrology and water resources of the Colorado River Basin. Hydrology and Earth System Science. Discussions 11(4):1417-1434.
Gangopadhyay, S., M. Clark, and B. Rajagopalan, 2005. Statistical downscaling using K-nearest neighbors. Water Resources Research. 41: W02024.
Prairie, J.R., B. Rajagopalan, T.J. Fulp, and E.A. Zagona, 2006. Modified K-NN model for stochastic streamflow simulation. Journal of Hydrologic Engineering. 11:371-378.
Salathé, E.P., 2005. Downscaling simulations of future global climate with application to hydrologic modeling. International Journal of Climatology. 25:419-36.
For more on Downscaling Techniques: