
Percentage change in # days with >.5 in of precipitation, 2050s A2 scenario relative to 1980s baseline
The climate team, led by Kushnir and Horton, has created downscaled projections of means and extremes of temperature, preciptation, and sea level for the three focus cities of Boston, New York, and Philadelphia. Projection techniques are described below.
Statistical downscaling
The large number of available GCMs makes possible model-based probabilistic assessment of future climate projections across a range of climate sensitivities. Although GCMs are the primary tool used for long-range climate prediction (IPCC, 2007), the spatial scale of climate model output is still too coarse for most impacts studies and decision-support purposes at the urban scale. Statistical downscaling, which involve statistically relating large-scale climate features to fine-scale climate for a region of interest, can address this spatial resolution gap. The BCSD approach, developed for hydrologic impact studies (Wood et al. 2004), is well tested, automated, and computationally efficient enough to be easily applied to ensembles of projections, and able to produce spatially continuous, fine-scaled gridded output of precipitation and temperature for regional and local impacts analysis. The BCSD technique, available as the World Climate Research Programme's (WCRP's) Coupled Model Intercomparison Project phase 3 (CMIP3), entails statistical downscaling spatially continuous fields and is unique in that it can produce 1/8th degree gridded time series of precipitation and surface air temperature over a large spatial domain. While all downscaling techniques have potential shortcomings, most notably the assumption that the historical relationship between the large and fine scale features will remain unchanged in the future, the BCSD method has been shown to provide downscaling capabilities comparable to other statistical and dynamical methods (Wood et al. 2004). Figure 1 shows our use of BCSD to make projections for both the regional scale (left) and the urban scale (right).
Model based probabilities will be created by sampling a range of GCMs and emissions scenarios. A valuable contribution to the NCA will be comparing different ways of integrating the CMIP 3 and CMIP 5 analyses, and exploring how different the model based distributions based on CMIP 3 and CMIP 5 are.
The microclimate and heat impacts on health analyses of investigators Quattrochi and Gaffin will require daily maximum and minimum temperatures. The team will use validated historical temperature data from multiple stations within urban areas to perform a second downscaling step to create daily projections for stations both within urban areas and in the surrounding regions using the monthly change fields generated for temperature using BCSD. The climate team is currently testing whether historical daily anomaly temperatures about the mean for the same month can be randomly applied to the BCSD fields or whether correlations between monthly temperature and intra-monthly temperature variance make it necessary to user other techniques such as weather generators.
Drs. Rosenzweig and Horton have extensive experience with developing climate projections for sector-based impact and adaptation assessments (e.g., Horton and Rosenzweig 2010, Rosenzweig et al. 2011, and Rosenzweig and Solecki 2010).
Figure 1: On the left, the map shows the temperature change (°F) across the Northeast for the 2050s relative to the 1980s base period, for the A2 scenario averaged across the 16 GCMs to form an ensemble mean. On the right, combined observed (black line) and projected temperature changes for New York City. Projected model changes through time are applied to the observed historical data. The three thick lines (green, red, blue) show the average for each emissions scenario across the 16 GCMs. Shading shows the central range. The bottom and top lines, respectively, show each year's minimum and maximum projections across the suite of simulations. A ten-year filter has been applied to the observed data and model output. The dotted area between 2005 and and 2015 represents the period that is not covered due to the smoothing procedure.
Regional climate model simulations
The climate team has also conducted analysis of the Northeast using the outputs of the North American Regional Climate Change Assessment Program. Dr. Horton is assessing how six RCMs perform (as defined by stakeholder relevant metrics such as days over 90F, days below 32F, and days with over .5 inches of preciptation) compared to observations when driven by the best available boundary conditions. Additional analysis is focused on relative biases is RCMs and the driving GCMs required for climate projections. Future work will focus on NARCCAP projections, once bias-correction has been completed.
Dr. Horton has presented preliminary NARRCAP results for the Northeast in numerous settings, including the American Meteorological Society Annual meeting (Seattle), the CORDEX meeting (Trieste, Italy), Lamont Doherty Earth Observatory, the University of Massachusetts, and a CCRUN webinar.