Global climate model scenarios downscaled for Canada David Price Northern Forestry Centre, Edmonton CIF e-lecture, 14 March 2012
Acknowledgements Slide 2 Dan McKenney Marty Siltanen Pia Papadopol Kevin Lawrence John Pedlar Kathy Campbell Yonghe Wang (all the above are with CFS) Mike Hutchinson (Australian National University, Canberra) Linda Joyce (USDA Forest Service, Ft Collins CO) Dave Coulson (USDA Forest Service, Ft Collins CO) Canadian Centre for Climate Modelling and Analysis (CCCMA) US National Center for Atmospheric Research (NCAR) G. Strand (UCAR) Australia: Commonwealth Scientific and Industrial Research Organisation (CSIRO) M. Collier, M. Dix, and T. Hirst (CSIRO Marine and Atmospheric Research Division) Japan: Center for Climate System Research, University of Tokyo, National Institute for Environmental Studies, Frontier Research Center for Global Change Many reviewers in Canada and the USA of two reports published in 2011.
Outline Slide 3 Techniques: Downscaling and ANUSPLIN Selecting GCMs and GHG scenarios Results Maps Graphs Sample applications Testing the climate models Biophysical variables and moisture indices Concluding remarks Publications & data distribution
Outline Slide 4 Techniques: Downscaling and ANUSPLIN Selecting GCMs and GHG scenarios Results Maps Graphs Sample applications Testing the climate models Biophysical variables and moisture indices Concluding remarks Publications & data distribution
What is Downscaling? Slide 5 Downscaling is using coarse spatial resolution data to generate information that is more useful at smaller scales Dynamical: physically consistent simulations of weather at GCM timesteps but at higher spatial resolution. RCMs are the prime example. Statistical: many methods to relate GCM results to observed weather and climate. Diagram from: D. Viner, CRU, University of East Anglia Spatial interpolation: relatively simple and largely empirical, but robust!
About ANUSPLIN Slide 6 Mike Hutchinson at Australian National University (Canberra) FORTRAN program for application of multi-variate thin-plate splines (typically up to 4 independent variables, more covariates) Most applications routinely incorporate a spatially (and temporally) varying dependence on elevation A key element is that ANUSPLIN minimizes General Cross Validation (GCV) statistic an objective method to select the best interpolation models, optimize data smoothing, and provide estimates of predictive error Diagnostics help to identify data errors ANUSPLIN continually being updated Check out http://fennerschool.anu.edu.au/publications/software/anusplin.php Numerous applications worldwide, many independent of Hutchinson s group Lots of references available: see CFS GLFC web site for a list and links: http://cfs.nrcan.gc.ca/subsite/glfc-climate
Normalizing the GCM data Slide 7 Ideally, we want to capture the climate change signal generated by the GCM, but corrected for the GCM s inaccuracies in representing reality We use the Delta method, based on a reference period for which we also have observed data (e.g., 1961-1990) Interpolated long-term monthly means (30-year normals ) provide a reference data set with spatial detail We then add the change signal (i.e., the temperature difference) to the observed climate normals for the reference period, interpolated to the same coordinates. Requires processing of GCM data to convert them to deltas WRT the same period in the simulation.
Correcting for GCM inacuracy Slide 8 Step 1: Determine mean of observations for reference period.
Correcting for GCM inacuracy Slide 9 Step 1: Determine mean of observations for reference period. Step 2: Determine mean of GCM projection for reference period.
Correcting for GCM inacuracy Slide 10 Step 1: Determine mean of observations for reference period. Step 2: Determine mean of GCM projection for reference period. Step 3: Calculate delta values by subtracting (or dividing by) the GCM mean from Step 2
Correcting for GCM inacuracy Slide 11 Step 1: Determine mean of observations for reference period. Step 2: Determine mean of GCM projection for reference period. Step 3: Calculate delta values by subtracting (or dividing by) the GCM mean from Step 2 Step 4: Calculate corrected GCM data by adding (or multiplying by) the observed mean from Step 1
Selected GCM Scenarios Slide 12 Projection data generated by GCMs from CCCma (Canada), CSIRO (Australia), NCAR (USA) and NIES (Japan). [Data also available from IPCC 3 rd Assessment (TAR, 2001) (CCCma, CSIRO, Hadley Centre (UK) and NCAR.] SRES A2: increasing population, little technological change, greater deforestation, pollution and CO 2 emissions SRES B1: as A2, but rapid global shift towards resource-efficient technologies and reduced GHG emissions SRES B2: as B1, but more local efforts to increase resource efficiency and reduce emissions SRES A1B: higher population growth than A2, with balance of energy from fossil and renewable sources Monthly time series extending from 1961 to 2100, gridded to 5 arcminute (1/12 degree lat/lon) resolution about 10 km. 20 data sets in total. Lots of ways to use these data! Nakićenović et al. 2000. IPCC Special Report on Emissions Scenarios.
IPCC AR4 GHG Scenarios Slide 13
GCM spatial resolutions vary Slide 14
GCM input data sets Slide 15 GCM 1 IPCC AR4 scenario(s) Monthly variable(s) 2 Source 3 Time period CGCM31MR 20C3M, A1B, A2, B1 pr, tas, rsds, hur, huss, psl, uas, vas CMIP3 1961 2100 CGCM31MR 20C3M, A1B, A2, B1 tasmin, tasmax CCCma 1961 2100 CSIROMK35 20C3M, A1B, A2, B1 pr, tas, tasmin, tasmax, rsds, psl, uas, vas, hur, huss (except B1) CMIP3 1961 2100 CSIROMK35 B1 huss CSIRO 2001 2100 MIROC32MR 20C3M, A1B, A2, B1 pr, tas, tasmin, tasmax, rsds, hur, huss, psl, uas, vas CMIP3 1961 2100 NCARCCSM3 20C3M, A1B, B1 pr, tas, tasmin, tasmax, rsds, hur, huss, psl, uas, vas CMIP3 1961 2099 NCARCCSM3 A2 pr, tas, tasmin, tasmax, rsds, hur, huss, psl, uas, vas ESG 1961 2099
Outline Slide 16 Techniques: Downscaling and ANUSPLIN Selecting GCMs and GHG scenarios Results Maps Graphs Applications Testing the climate models Biophysical variables and moisture indices Concluding remarks Publications & data distribution
Projected changes in T max Slide 17 These maps show absolute temperatures. It is very hard to see the changes over 100+ simulated years!
Projected changes in Precip Slide 18 These maps show absolute total annual precipitation. Again, it is very hard to see the changes over 100+ simulated years! It would be just as hard to see differences among different GCMs when forced by the same GHG emissions scenario.
Changes in annual means (SRES A2, 2080s) Slide 19 CGCM3.1 - Canada CSIRO Mk 3.5 - Australia MIROC3.2 - Japan NCAR CCSM3 - USA Temperature Increase ( C) Precipitation Change (ratio)
Analysis by Canadian ecozones Slide 20
Annual Mean Daily T min Prairies subhumid ecozone (Parkland) Slide 21 Temperature ( C) (Historical data ~45 years)
Winter (DJF) Mean Daily T min Boreal Plains ecozone Slide 22 Temperature ( C)
Summer (JJA) Mean Daily T max Atlantic Maritime ecozone Slide 23 Temperature ( C)
Summer (JJA) Mean Vapour Pressure Atlantic Maritime ecozone Slide 24 Vapour pressure (kpa) (No historical record)
Fall (SON) Total Precipitation Mixedwood Plains ecozone Slide 25 Total Precipitation (mm)
Winter (DJF) Mean Daily Solar Radn Boreal Shield W ecozone Slide 26 MJ m -2 day -1 (No historical record) M
Winter (DJF) Mean Daily Solar Radn Boreal Shield W ecozone Slide 27 MJ m -2 day -1 10-year moving averages
Which GCM scenarios are best for a regional study? It depends. Slide 28 Pacific Maritime Montane Cordillera Atlantic Maritime Prairies Aspen parkland Mixedwood Plains 2090 2050
Which GCM scenarios are best for a regional study? It depends. Slide 29 Taiga Plains Boreal Shield W Boreal Shield E 2050 2090 Boreal Cordillera Boreal Plains Hudson Plains
IPCC AR4 GHG Scenarios Slide 30
Temperature trends by ecozone 1960 to 2100 Slide 31 Annual Mean Daily Tmin ( C) Taiga Plains Aspen Parkland Hudson Plains Boreal Plains Shield E
Temperature trends by ecozone 1960 to 2100 Slide 32 Annual Mean Daily Tmin ( C) Taiga Plains Aspen Parkland Hudson Plains Boreal Plains Shield E
Precipitation trends by ecozone 1960 to 2100 Slide 33 Annual Total Precipitation (mm) Taiga Plains Aspen Parkland Hudson Plains Boreal Plains Shield E
Outline Slide 34 Techniques: Downscaling and ANUSPLIN Selecting GCMs and GHG scenarios Results Maps Graphs Applications Testing the climate models Biophysical variables and moisture indices Concluding remarks Publications & data distribution
Do GCMs really work? Slide 35 www.globalwarmingart.com Data from: Meehl et al. 2004. J. Clim. 17: 3721-3727. Jones and Moberg. 2003. J. Clim. 16: 206-223.
Slide 36 GCM validation (15 years of observed data vs. simulated) Underestimating observed Tmin it was warmer in 1990-2005 than the GCMs predicted! Overestimating Tmax Ensemble_a2_MaxT_July Ensemble_a2_MinT_January
Biophysical variables Slide 37 No. Variable Description 1 Annual Mean Temperature Annual mean of monthly mean temperatures 2 Mean Diurnal Range Annual mean of monthly mean daily temperature ranges 3 Isothermality (2) / (7) 4 Temperature Seasonality Standard deviation of monthly mean temperature estimates expressed as a percentage of their mean 5 Max Temperature of Warmest Period Highest monthly maximum temperature 6 Min Temperature of Coldest Period Lowest monthly minimum temperature 7 Temperature Annual Range (5) (6) 8 Mean Temperature of Wettest Mean temperature of three wettest months Quarter 9 Mean Temperature of Driest Quarter Mean temperature of three driest months 10 Mean Temperature of Warmest Mean temperature of three warmest months Quarter 11 Mean Temperature of Coldest Mean temperature of three coldest months Quarter 12 Annual Precipitation Sum of monthly precipitation values 13 Precipitation of Wettest Period Precipitation of the wettest month 14 Precipitation of Driest Period Precipitation of the driest month 15 Precipitation Seasonality Standard deviation of monthly precipitation estimates expressed as a percentage of their mean 16 Precipitation of Wettest Quarter Total precipitation of three wettest months 17 Precipitation of Driest Quarter Total precipitation of three driest months 18 Precipitation of Warmest Quarter Total precipitation of three warmest months 19 Precipitation of Coldest Quarter Total precipitation of three coldest months 20 Start of Growing Season Date when daily mean temperature first meets or exceeds 5 C for five consecutive days in spring 21 End of Growing Season Date when daily minimum temperature first falls below -2 C after 1 August 22 Growing Season Length (21) (20) 23 Total Precipitation for Period 1 Total precipitation of three months prior to (20) 24 Total Precipitation for Period 3 Total precipitation during (22) 25 Growing Degree Days for Period 3 Total degree days during (22), accumulated for all days where mean temperature exceeds 5 C. 26 Annual Minimum Temperature Annual mean of monthly minimum temperatures 27 Annual Maximum Temperature Annual mean of monthly maximum temperatures 28 Mean Temperature for Period 3 Mean temperature during (22) 29 Temperature Range for Period 3 Highest minus lowest temperature during (22)
Projecting Future Drought Slide 38 Range of approaches to calculating balance of annual precipitation and evapotranspiration, the latter a function of temperature and radiation. Climate Moisture Index (Hogg 1994, 1997) PDSI (Palmer 1965) Considerable interest in projecting how climate change will affect water supplies, availability, and hence ecosystems and communities E.g., Dai (2011) We are investigating implications for Canadian forest regions using our own data
The future global context? Slide 39 Dai 2011. WIRES Climate Change 2: 45-66. doi: 10.1002/wcc.81 1955 1980 2005 2035 2065 2095-20 -10-6 -3-1 0 +1 +3 +6 +10 +20 10-yr average Palmer Drought Severity Index from IPCC AR4 (A1B scenario, 22 models,)
Projections of Hogg CMI Slide 40 1961-1990 2011-2040 SRES A2 averages of four GCMs 2041-2070 2071-2100 Maps prepared by Y. Wang, NRCan
Outline Slide 41 Techniques: Downscaling and ANUSPLIN Selecting GCMs and GHG scenarios Results Maps Graphs Applications Testing the climate models Biophysical variables and moisture indices Concluding remarks Publications & data distribution
Data distribution - GLFC Slide 42 http://cfs.nrcan.gc.ca/projects/3/3
Data distribution - GLFC Slide 43 http://cfs.nrcan.gc.ca/projects/3/3
Concluding Remarks Slide 44 We have created a suite of several nationalscale climate scenarios which allow us to explore a range of potential impacts We chose a simple downscaling approach that is easy to understand and provides robust data for application over large regions It is not clear that other methods give results that are more meaningful, especially considering all the assumptions and errors in current GCMs. (Not suggesting there is anything wrong with other downscaling methods!) Data are freely available to anyone on request!
Publications Slide 45 http://cfs.nrcan.gc.ca/publications?id=32971 http://www.fs.fed.us/rm/pubs/rmrs_gtr263.pdf