The challenge to downscale present and future climates John Horel University of Utah John.horel@utah.edu Observations Defining Mountain Climates i t a t S a it c s o lM de li n g Dynamical Modeling Observations Defining Mountain Climates Data Assimilation Dynamical Modeling Creating an Analysis • Models or observations cannot independently define mountain climates and climate processes effectively • Data assimilation methods – Horizontally/vertically interpolate coarse resolution model output to finer scale grid – Use observations to adjust downscaled model guidance at every gridpoint Caveats • Number of conventional observations much less than number of gridpoints • Models and observations contain errors that must be taken into account (either specified or predicted) Recognition of Sources of Errors Smooth terrain Inaccurate ICs Representative Incomplete Physics Dynamical Model Errors Instrumental Analysis Errors Observational Errors Need for balance… Spatial & Temporal Continuity Dynamical Model Climates Specificity Analysis Climates Observational Climates Analyses of Record (AOR) • There is considerable demand for AORs: Best realtime and retrospective analyses at high spatial and temporal resolution (Horel and Colman, BAMS, 2005) • Planning led by Lee Anderson, NWS Office of Science and Technology • Clear applications: – Contribute to generation of NWS operational gridded forecasts and verification of those products – Support efforts in regional and local climate modeling and prediction • Initial states for climate predictions • Verification of climate models • Process studies of past and ongoing climate NWS Proposed AOR Program • Phase I: Real Time Mesoscale Analysis • Analyses to be produced hourly within 30 minutes after valid time • Phase II – Ongoing Analysis of Record – Use state-of-the-art data assimilation methods to obtain best analysis possible a day or so after valid time • Phase III – Reanalysis – Apply mature AOR retrospectively – 30 year time history of AORs THE REAL TIME MESOSCALE ANALYSIS (RTMA): Progress towards a National Mesoscale Analysis of Record Manuel Pondeca, Geoff Manikin, David Parrish, Jim Purser, Wan-Shu Wu, Geoff DiMego, John Derber, Stan Benjamin, John Horel, Lee Anderson, Brad Colman, Stephen Jascourt Environmental Modeling Center National Centers for Environmental Prediction http://www.met.utah.edu/mesowest http://www.met.utah.edu/mesowest Moving Beyond the RTMA to Analyses of Record • RTMA 2D variational assimilation approach is insufficient for Analyses of Record • Requires 4 dimensional data assimilation system • Ensemble (vs. single deterministic) approaches are as relevant to analyses as model forecasts • Resources not available yet to support R&D • Significant opportunity to advocate improving data assimilation methods appropriate for mountain climate applications Observations Downscaling Future Mountain Climates Statistical downscaling i t a t S a it c s o lM de li n g Dynamical Modeling Premise • Resolution of global general circulation models (and regional models) is too coarse to provide detailed information on response of mountain precipitation or surface temperature to increasing greenhouse gas emissions • Relate in physically-based manner observed mountain precipitation to year-to-year variations in regional tropospheric circulation features likely simulated by GCMs with greater fidelity • Assume relationships between present-day circulation features & mountain precipitation continue during next 100 years GFDL GCM 700 hPa winter (DJF) temperature response** in northern Utah to A1B greenhouse gas emission scenario ** bias removed Separating the Inseparable… • Nonlinear feedbacks are critical in the climate system • These feedbacks make it difficult to separate the many physical mechanisms responsible for future changes in snow pack • One of the many ways precipitation may be affected by increased greenhouse gas emissions: – Increased elevation of rain/snow line during winter as a result of increased tropospheric temperature Pilot Study Area: East Slope Transect near Utah’s Ben Lomond Peak Approach • Estimate during each storm the variability in elevation of rain/snow line over past 27 winters using simplified physically-based approach • Relate year-to-year variability in fraction of precipitation above snow line during winter to variability in 700 hPa winter temperature from nearby rawinsonde • Assume present relationship between 700 hPa temperature & snow fraction will continue during next 100 years • Assess sensitivity of snow fraction to evolving 700 hPa temperature estimated by GCMs Sensitivity to Tw: for each 1oC increase, snow level rises 166 m ~3000 m Tw observed ~ -5oC 700 hPa Ben Lomond Peak 2970 m BLP BLT Rawinsonde Salt Lake City Airport ~65 km SSW 1288 m Tw=2oC Ogden Valley ~1500 m Sensitivity to Tw: for each 1oC increase, snow level rises 166 m ~3000 m Tw observed ~ -4oC 700 hPa Ben Lomond Peak 2970 m BLP BLT Rawinsonde Salt Lake City Airport ~65 km SSW 1288 m Tw=2oC Ogden Valley ~1500 m Regression: Winter SLC 700 hPa Temperature vs. Ben Lomond Transect Winter Snow Fraction DJF Seasonal Snow Fraction R= -0.65 Slope= -7.1% per oC DJF 700 hPa Temperature (oC) at SLC Snowpack Temperature Sensitivity • Sierras – Howat and Tulaczyk (2005) – ~6-10% decrease in SWE per oC • Cascades – Casola et al. (2008) – ~20% loss in snowpack per oC • Ben Lomond Transect – ~7% decrease in snow fraction per oC DJF Seasonal Snow Fraction Estimated Winter Snow Fraction Based on GFDL A1B Simulation Summary • Analysis of Record proposed program provides framework for improving downscaling for present climate • Funding for that program not clear and support from broader community would be of great help • Downscaling for present day or future climates requires using entire suite of tools – – – – Observations Surface state information Statistical modeling Dynamical modeling