ClimateChange Assessment Guide.pdf - University of Waterloo

E-30Guide for Assessment of Hydrological Effects of Climate Change in OntarioPredictands Predictors Meaning Explained VarianceDaily Precipitationtemp Mean Temperature at 2m. 0.50 – 0.64mslpPressure at Mean Sea Level0.021 – 0.168p5_v500 hPa Meridonal Air Velocity0.028 – 0.158p5_z500 hPa Vorticity0.011 – 0.040p500500 hPa Geopotential Height0.007 – 0.050p8_v850 hPa Meridonal Air Velocity0.042 – 0.138p8_z850 hPa Vorticity0.010 – 0.148p850850 hPa Geopotential Height0.008 – 0.137p8zh850 hPa Divergence0.028 – 0.108s500500 hPa Specific Humidity0.027 – 0.105s850Specific Humidity at 850 hPa0.056 – 0.143shumSpecific Humidity at the surface 0.037 – 0.106The air temperature correlations were found to be strong while the precipitation correlations were much weaker. Assuch, various transformations of the most promising precipitation model predictors and the precipitation were madeand examined for correlation. No improvement in explained variance was found. As recommended in the SDSMmanual, correlation matrices were generated to investigate inter-variable correlations and partial correlations on aseasonal basis. This analysis confirmed that air temperature is strongly related to surface temperature (notsurprisingly) and humidity. The strongest correlations with precipitation were with the s500, shum and p8_vpredictors.The climate model was calibrated using the selected predictors on a monthly basis. Temperature calibration wasconducted using an unconditional process. This indicates that a direct link exists between the local temperature andthe predictors. The precipitation process is assumed to be conditional. The R 2 values for the temperature modelwere strong throughout the year ranging from 0.55 to 0.71 (where an R 2 of 1 is a perfect fit). Corresponding R 2 valuesfor the precipitation model were much lower at 0.03 to 0.14.The weather generator function of SDSM was used to generate future climates using the calibrated climate modelsdriven with the 2041 to 2070 predictor time series. One set of results was generated using the CGCM3-A1B scenariopredictors and one with the CGCM3-A2 scenario predictors. Various graphical comparisons between the observedtime series and simulated time series are available. These were used to obtain a qualitative assessment of thegoodness of fit. Problems were detected after close examination of the daily GCM output data downloaded from the DAI website.Application of LARS-WG to generating future scenario climates requires that the future time horizon time series ofprecipitation be analyzed for relevant statistics. This includes analysis of the length and occurrence of wet and dryday periods. The simulated daily precipitation data for the period 2041 to 2070 for the CGCM3-A2 and CGCM3-A1Bscenarios were found to have precipitation occurring in almost all days with very small amounts of rain in most days;this is referred to as the drizzle effect. Therefore the actual occurrence of dry days was very rare and short lived inthis data series. In fact there is an expectation of more frequent and long lasting droughts across much of NorthAmerica in the future (Section 2.1.2 of the Guide). This drizzle effect, or occurrence of many days of small amounts ofrain, has been noted widely in daily GCM simulation output.