HR: 0800h
AN: GC11A-0142    [Abstracts]
TI: On the Verification and Comparison of Extreme Rainfall Indices From Climate Models
AU: * Chen, C
EM: chen@rain.geos.ntnu.edu.tw
AF: National Taiwan Normal University, Department of Earth Sciences and Institute of Marine Environmental Science and Technology, 88, Sec. 4, Ting-Chou Rd., Taipei, 116, Taiwan
AU: Knutson, T
EM: tom.knutson@noaa.gov
AF: Geophysical Fluid Dynamics Laboratory, 201 Forrestal Road, Princeton, NJ 08540-6649, United States
AB: The interpretation of model precipitation output (e.g., as a grid-point estimate vs. as an areal mean) has a large impact on the evaluation and comparison of simulated daily extreme rainfall indices from climate models. We first argue that interpretation as a grid-point estimate (i.e., corresponding to station data) is incorrect. We then illustrate impacts of this interpretation vs. the areal-mean interpretation in the context of rainfall extremes. A high resolution (0.25°x0.25° grid) daily observed precipitation dataset for the U.S. (from CPC) is used as idealized perfect model gridded data. Both 30-year return levels of daily precipitation (P₃₀) and a simple daily intensity index are substantially reduced in this data when estimated at coarser resolution compared to the estimation at finer resolution. The reduction of P₃₀ averaged over the conterminous US is about 9, 15, 28, 33, and 43% when the data were first interpolated to 0.5°x0.5°, 1°x1°, 2°x2°, 3°x3° and 4°x4° grid boxes, respectively, before the calculation of extremes. The differences resulting from the point estimate vs. areal mean interpretation are sensitive to both the data grid size and to the particular extreme rainfall index analyzed. The differences are not as sensitive to the magnitude and regional distribution of the indices. Almost all IPCC AR4 models underestimate US mean P₃₀ if it is compared directly with P₃₀ estimated from the high resolution CPC daily rainfall observation. On the other hand, if CPC daily data is first interpolated to various model resolutions before calculating the P₃₀ (a more correct procedure in our view), about half of the models show good agreement with observations while most of the remaining models tend to overestimate the mean intensity of heavy rainfall events. A further implication of interpreting model precipitation output as an areal mean is that use of either simple multimodel ensemble averages of extreme rainfall or of inter-model variability measures of extreme rainfall to assess the common characteristics and range of uncertainties in current climate models is not appropriate if simulated extreme rainfall is analyzed at a model?|s native resolution. Owing to the large sensitivity to the assumption used, we recommend that for analysis of precipitation extremes, investigators interpret model precipitation output as an area average as opposed to a point estimate, and then ensure that various analysis steps remain consistent with that interpretation.
DE: 1626 Global climate models (3337, 4928)
DE: 1637 Regional climate change
DE: 1817 Extreme events
DE: 3354 Precipitation (1854)
SC: Global Environmental Change [GC]
MN: 2007 Fall Meeting