Volume 14 Number 2	Winter 1997-98

Precipitation Predictability for Climate Studies Protein Catalysts Electrochemical Electron Transfer Prediction Predicatability Seminar Synopsis Research Reports he inherently chaotic behavior of the atmosphere makes it very difficult to accurately predict atmospheric variables at all space-time scales. Despite a decade of considerable progress in atmospheric modeling, prediction of the onset, duration, location, and spatial variability of precipitation remains a challenge. Adding to the problem is the fact that regional climate models and global circulation models are run over large domains and provide information only on large grid scales. This makes it difficult to interpret the impact of climate change or climate variability on water resources management, which requires information at much smaller scales. University of Minnesota Professor Efi Foufoula-Georgiou and her co-workers, Ph.D. students Alin Carsteanu and V. Venugopal, former Ph.D. student Sanja Perica (now of the Office of Hydrology at the National Weather Service in Maryland), and Shuxia Zhang of Ohio State University, are working to adapt large grid scale atmospheric models for application to smaller-scale hydrologic problems. The group has developed a spatial rainfall downscaling scheme that can be used to disaggregate the results of regional or global models down to watershed scales for hydrologic applications. Their innovative dynamical/statistical hybrid approach is based on having unraveled a link between larger-scale dynamics of the atmosphere and smaller-scale statistics of rainfall fields. Specifically, the group found that normalized rainfall gradients (obtained via wavelet transform of rainfall intensities) exhibit self-similarity over a significant range of scales. This means that variability at one scale relates to variability at another via a simple transformation that is a function of the ratio of the two scales. They also discovered that the scale-invariant parameter of this transformation is strongly related to the convective instability of the pre-storm environment (measured by the convective available potential energy, or CAPE). These findings were confirmed by analysis of a large number of radar-monitored warm-season storms, called mesoscale convective complexes, in the Midwest. Based on these findings, a precipitation downscaling scheme was developed using an inverse wavelet transform procedure. Figure 1 shows the results of applying the developed scheme to downscale rainfall from 64 x 64 km2 to 4 x 4 km2 grid averages. The figure shows that the disaggregated (simulated) field compares well to the actual field. The only input to the model was the large-scale (64 x 64 km2) rainfall averages and the value of CAPE at the pre-storm environment. Accurate representation of the small-scale precipitation variability obviously will improve the prediction of other hydrologic variables, such as surface runoff peak and volume, which are used for water resources design, management, and operation. The question arises, however, as to whether there is a double advantage in trying to capture the small-scale precipitation variability in large-scale atmospheric models. Indeed, the group's results indicate that the inclusion of subgrid-scale rainfall variability and consideration of the small-scale land-atmosphere feedback effects can significantly affect surface temperature predictions-and even the short-term prediction of rainfall intensity itself (see figure 2). This is because redistribution of the large-scale rainfall average within the grid box (i.e., using the bottom right field of figure 1 instead of the top field) causes redistribution of soil moisture and alters the small and large scale land-atmosphere system dynamics. These findings point to the importance of understanding subgrid scale convection processes and cloud microphysics, as well as the significance of including subgrid scale precipitation parameterizations interactively, even in large-scale climate prediction models. Figure 1: A June 27, 1985 storm over Kansas and Oklahoma at 0300 UTC. The bottom figure in the left column shows the original radar data at 4 x 4 km2 resolution. From these data, rainfall fields at lower and lower resolutions were obtained by averaging up to 64 x 64 km2 fields, as shown in the top panel (upscaling). Then, using the 64 x 64 km2 field and the downscaling scheme of Perica and Foufoula-Georgiou, rainfall fields at higher resolutions were reconstructed down to the resolution of 4 x 4 km2, as shown in the bottom right panel. A good agreement was found between the rain patterns and the areas covered by rain in the simulated and original fields at all resolutions. Figure 2: Changes in ground temperature in ºC (color variation) and rainfall intensity in mm/h (black and white curves) caused by adding the feedback effect of subgrid-scale rainfall variability to the simulation of an extreme Midwestern storm on June 11, 1985 over Oklahoma. The Penn State/National Center for Atmospheric Research mesoscale modeling system with two nested domains of 36 km and 12 km, respectively, was used for this simulation. The comparison (differences in values between the run without and with subgrid scale parameterizations) is shown for a domain of 12 km resolution. The black and white curves indicate the increase and decrease of rainfall intensity with peak values of 55.2 mm/h and 66.6 mm/h, respectively.   HOME | BULLETINS | CONTACT US | PREVIOUS ARTICLE | NEXT ARTICLE | THIS ISSUE

This information is available in alternative formats upon request by individuals with disabilities. Please send email to alt-format@msi.umn.edu or call 612-624-0528.

URL: http:// This page last modified on   Website related questions or problems should be directed to webmaster@msi.umn.edu The Supercomputing Institute does not collect personal information on visitors to our website. For the University of Minnesota policy, see www.privacy.umn.edu.