University of Washington Center for Science in the Earth System Applied Climate Dynamics Progress 2006-2007 For ARCs Council Meeting Tallahassee, FL October 21, 2007
University of Washington west-wide hydrologic forecast system Almost 250 forecast points Monthly (biweekly in winter) updates to 12 month max lead Based on ESP, CPC “official”, and CFS ensembles
UW national surface water monitor Daily updates for soil moisture and runoff Used in NRCS NWCC weekly snowpack and drought monitor report, and CPC input to drought monitor Spatial resolution currently ½ degree, increase to 1/8 degree and merger with Princeton University 1/8 degree product for eastern U.S. underway
Drought research – multimodel comparison – July 1934 soil moisture as percentiles
Multi-Model Ensemble Process Flow: Break 40-year time series into 40 Januaries, 40 Februaries, etc. Examine interannual variation of each month’s time series For each month: Apply bias correction to all 3 models Find PCs of the set of 3 models Mult. Lin. Reg. of obs vs PCs Ensemble mean = weighted avg of PCs, with weights given by coefs in step 3 Conclusions: Different models are best at different times Ensemble is better than the best model at all times Better correlation with interannual signal → better long-lead-time forecast skill Better prediction of summer low flows Month Month
December-January-February Evaluation of 15-day ensemble forecast accuracy from Hamill-Whitaker reforecast data set Input: “reforecast” 15-day 500mb geopotential height anomaly forecast and verification maps Output: canonical correlation pairs of forecast and verification maps that are patterns of variability for which the forecast has the most skill December-January-February (DJF): Leading verification map (top panel) is a mix of Pacific North American and Northern Annular Modes (map shows typical anomalies, contour interval 10m) Associated forecast map (not shown) similar and weaker Temporal amplitudes of verification and forecast maps correlated at 0.67 October-November-December: Leading verification map similar to that for DJF, albeit weaker and more of a Pacific / Polar emphasis Temporal correlation 0.52 Future work: There are more floods in Cold and neutral ENSO: stratify reforecast skill Certain MJO phases: examine observations (reforecast poor at MJO) October-November-December
Regional Climate Modeling Current work uses MM5 Weather Model Transition to Weather Research and Forecasting (WRF) Model underway Nested grids 15-45-135 km Outermost grid by forcing global model ECHAM5 - Max Planck, Hamburg, model CCSM3 - NCAR Community Model
Fall Precipitation in MM5 forced by ECHAM5 Shift in storm track winds and increased moisture flux increases precipitation along N-S ridges MM5 forced By ECHAM5 Increased Westerly Flow Change in Sep-Oct-Nov Precip (mm/day) 1990s to 2050s Contours: Change in 500-mb heights
Winter Warming in MM5 forced by ECHAM5 Snow-albedo feedback amplifies warming at mid-elevation along terrain 1990s to 2050s DJF Temperature Change Difference between MM5 and ECHAM5 Less warming In MM5 More warming In MM5 Change in Winter Temperature (degrees C) Change in Winter Temperature (degrees C)
Statistical Downscaling Empirically relate large-scale spatial pattern to observed 1/8-degree patterns Fast but lacks potential feedbacks Applied to 20 IPCC models and 2 scenarios Method: Bias-correct global climate model simulation Spatially downscale to 1/8-degree Wood et al. (2004), Salathé et al (2007)
DJF Results for 2050s Temperature Precipitation Composite of 18 models Change from 1950-1999 to 2050-2059 Temperature Precipitation °C mm/day
DJF Results for 2050s Temperature Precipitation Composite of 18 models Change from 1950-1999 to 2050-2059 Temperature Precipitation °C Percent