Snow model intercomparisons – relevance to alpine research Richard Essery, University of Edinburgh INARCH Workshop, 23 October 2015 Evaluation, intercomparison and benchmarking • Evaluation – comparing model outputs with observations • (Inter)comparison – comparing models with each other Pomeroy definition: “A tedious and thankless task of dubious scientific merit” • Benchmarking – comparing models with an a priori standard Abramowitz, 2012. Towards a public, standardized, diagnostic benchmarking system for land surface models. Geosci. Model Dev., 5, 819-827 2 Intercomparison projects with snow components Global Regional Local Coupled Uncoupled AMIP CMIP ESM-SnowMIP GSWP WaterMIP ESM-SnowMIP PILPS2e Rhône-Agg PILPS2d SnowMIP ESM-SnowMIP http://www.climate-cryosphere.org/activities/targeted/esm-snowmip Drive snow models by: • full coupling with global Earth System Models (CMIP6) • bias-corrected global reanalyses (GSWP3) • in-situ meteorological data from well-instrumented reference sites • global reanalyses downscaled from 0.5º grids to reference sites 3 SnowMIP and SnowMIP2 ● SnowMIP (Etchevers et al. 2004) ● SnowMIP2 (Essery et al. 2009) Col de Porte Goose Bay Sleepers River Weissfluhjoch Alptal BERMS Fraser Hitsujigaoka Hyytiälä (45.3ºN, 5.8ºE) (53.3ºN, 60.4ºW) (44.5ºN, 72.2ºW) (46.8ºN, 9.8ºE) (47.3ºN, 8.7ºE) (53.6ºN, 104.4ºW) (39.5ºN, 105.5ºW) (42.6ºN, 141.2ºE) (61.5ºN, 24.2ºE) Only 1 or 2 winters at most sites 4 ESM-SnowMIP reference sites • • • • • Bayelva, Svalbard BERMS, Saskatchewan Col de Porte, France * Imnavait Creek, Alaska Marmot Creek, Alberta * • Reynolds Creek, Idaho * • Sodankylä, Finland • Swamp Angel, Colorado * • Weissfluhjoch, Switzerland * 7 – 25 years at each site * high-elevation sites 5 Snow study sites and synoptic stations 6 In situ met data and reanalyses 7 In situ met data and reanalyses Sodankylä 2009-2010 daily ݎൌ 0.989 ݎൌ 0.996 ݎൌ 0.719 ݎൌ 0.628 8 In situ met data and reanalyses Col de Porte 2009-2010 daily ݎൌ 0.978 ݎൌ 0.971 ݎൌ 0.552 ݎൌ 0.713 9 Snow mass simulations with in situ driving 10 Simulations driven with reanalyses 11 Simulations with bias-corrected reanalyses 12 Single variables replaced with reanalyses 13 Temperature sensitivity and energy partition Weissfluhjoch in situ Weissfluhjoch reanalyses 14 Conclusions • snow accumulation and melt can be modelled well at (sheltered) mountain sites with an energy balance model and good-quality in situ meteorological data • large-scale meteorological reanalyses tend to have positive temperature and negative snowfall biases at high elevations • adjusting reanalyses to site averages of driving variables (if known) will improve simulations • total precipitation should be downscaled first and then partitioned into rain and snow according to downscaled temperature • models differ in their sensitivities to temperature biases • snow mass balance alone does not contain enough information to constrain snow energy balance 15 CMIP5 snow headlines "models reproduced the observed snow cover extent very well, but the significant trend towards reduced spring snow cover extent over the 1979-2005 period was underestimated" Brutel-Vuilmet et al. (2012), doi:10.5194/tcd-6-3317-2012 “The observed decreases in June SCE were found to diverge markedly from the rate of snow cover loss projected by CMIP5 climate model simulations” Derksen and Brown (2012), doi:10.1029/2012GL053387 "The spread in snow albedo feedback is very similar to that found in CMIP3 models, and it accounts for much of the spread in the 21st century warming of Northern Hemisphere land masses in the CMIP5 ensemble” Qu and Hall (2013), doi:10.1007/s00382-013-1774-0 “Much of the disagreement in modeled mean soil temperatures can be traced to the representation of thermal connection between the air and land surface and, in particular, its mediation by snow in winter" Koven et al. (2013), doi:10.1175/JCLI-D-12-00228.1 16 GSWP-2, Rhône-Agg and WaterMIP snow headlines “Most of the imbalances in both energy and water can be attributed to shortcomings in the handling of snow” “there is mediocre agreement among the models for most of the snowrelated variables, suggesting a potential area of continuing weakness” Dirmeyer et al. (2009), doi:10.1175/BAMS-87-10-1381 “The need for a better treatment of snow-cover dependence on spatial scale is warranted” Boone et al. (2004), J. Climate, 17 “Significant simulation differences between land surface and global hydrological models are found to be caused by the snow scheme employed” Haddeland (2011), doi:10.1175/2011JHM1324.1 17 PILPS and SnowMIP snow headlines Mid-latitudes: “early season ablation events are a significant source of model scatter” Slater et al. (2001), J. Hydrometeorol., 2 High latitudes: “The greatest among-model differences in energy and moisture fluxes in these high-latitude environments occur during the spring snowmelt period, reflecting different model parameterizations of snow processes” Nijssen et al. (2003), doi:10.1016/S0921-8181(03)00004-3 “models generally predict the duration of snow cover quite well but show broad ranges in their simulations of maximum snow accumulation, particularly at warmer sites and during warmer winters, and differences between open and forested plots are underestimated on average” Essery et al. (2009), doi:10.1175/2009BAMS2629.1 18