McGill University Montreal, Quebec, Canada Brace Centre for Water Resources Management Global Environmental and Climate Change Centre Department of Civil Engineering and Applied Mechanics School of Environment DOWNSCALING METHODS FOR CLIMATE RELATED IMPACT ASSESSMENT STUDIES Van-Thanh-Van Nguyen (and Students) Endowed Brace Professor Chair in Civil Engineering 1 OUTLINE INTRODUCTION What a hydrologic engineer needs from an atmospheric (climate) scientist? Extreme Precipitation Process (Extreme Temperature Process) The “scale” problem Climate variability and climate change OBJECTIVES DOWNSCALING METHODS Spatial Downscaling Issues APPLICATIONS SDSM and LARS-WG Some Current Developments CONCLUSIONS December 17, 2007, Singapore Climate Change Symposium 2 INTRODUCTION Information on rainfall characteristics is essential for planning, design, and management of various hydraulic structures (flood protection works, urban sewers, etc.) Rainfall records by raingages or radar are usually limited (< 50 years) and are not sufficient for assessing reliability of hydraulic structure design. Stochastic simulation of rainfall processes is needed to generate many long rainfall series. Several rainfall samples of adequate record length are needed to be able to determine how different system designs and operating policies might perform. the variability and the range of future system performance are better understood, and better system designs and policies could be selected. Extreme storms and floods account for more losses than any other natural disaster (both in terms of loss of lives and economic costs). Damages due to Saguenay flood in Quebec (Canada) in 1996: $800 million dollars. Average annual flood damages in the U.S. are US$2.1 billion dollars. Design Rainfall = maximum amount of precipitation at a given site for a specified duration and return period. December 17, 2007, Singapore Climate Change Symposium 3 … The choice of an estimation method depends on the availability of historical data: Sites Sufficient long historical records (> 20 years?) At-site Methods. Partially-Gaged Sites Limited data records Regionalization Methods. Ungaged Sites Data are not available Regionalization Methods. Gaged December 17, 2007, Singapore Climate Change Symposium 4 Extreme Rainfall Estimation Methods At-site Frequency Analysis of Precipitation Current practice: Annual maximum series (AMS) using 2parameter Gumbel/Ordinary moments method, or using 3parameter GEV/ L-moments method. Problem: Uncertainties in Data, Model and Estimation Method Regional Frequency Analysis of Precipitation Current practice: GEV/Index-flood method. Problem: How to define similarity (or homogeneity) of sites? 1 2 3 (WMO Guides to Hydrological Practices: 1st Edition 1965 → 6th Edition: Section 5.7) 4 Geographically contiguous fixed regions December 17, 2007, Singapore Climate Change Symposium Geographically non contiguous fixed regions Hydrologic neighborhood type regions 5 Rainfall Estimation Issues (1) THE “SCALE” PROBLEM The properties of a variable depend on the scale of measurement or observation. Are there scale-invariance properties? And how to determine these scaling properties? Existing methods are limited to the specific time scale associated with the data used. Existing methods cannot take into account the properties of the physical process over different scales. December 17, 2007, Singapore Climate Change Symposium 6 What are the impacts due to the scale problem? On SAMPLING and MEASUREMENT Low resolution High resolution ↓ Accuracy ↓ Noise ↓ Costs Optimum resolution? ↑ ↑ ↑ On DATA ANALYSIS TECHNIQUE Artifacts due to scale of measurement or computation. Scale-invariance properties? New techniques? December 17, 2007, Singapore Climate Change Symposium 7 ... On MODELLING TECHNIQUES Scale-invariance models? The SCALE problem has PRACTICAL and THEORETICAL implications. Scale-Invariance (or Scaling) Methods are developed in research ⇒ Engineering Practice? December 17, 2007, Singapore Climate Change Symposium 8 Rainfall Estimation Issues (2) Climate Variability and Change will have important impacts on the hydrologic cycle, and in particular the precipitation process! How to quantify Climate Change? General Circulation Models (GCMs): A credible simulation of the “average” “large-scale” seasonal distribution of atmospheric pressure, temperature, and circulation. (AMIP 1 Project, 31 modeling groups) Climate change simulations from GCMs are “inadequate” for impact studies on regional scales: Spatial resolution ~ 50,000 km2 Temporal resolution ~ (daily), month, seasonal Reliability of some GCM output variables (such as cloudiness precipitation)? December 17, 2007, Singapore Climate Change Symposium 9 … How to develop Climate Change scenarios for impacts studies in hydrology? Spatial scale ~ a few km2 to several 1000 km2 Temporal scale ~ minutes to years A scale mismatch between the information that GCM can confidently provide and scales required by impacts studies. “Downscaling methods” are necessary!!! GCM Climate Simulations Precipitation at a Local Site December 17, 2007, Singapore Climate Change Symposium 10 OBJECTIVES To review recent progress in downscaling methods from both theoretical and practical viewpoints. To assess the performance of statistical downscaling methods to find the “best” method in the simulation of daily precipitation (and extreme temperature) time series for climate change impact studies. To demonstrate the importance of scaling consideration in the estimation of daily and sub-daily extreme precipitations. December 17, 2007, Singapore Climate Change Symposium 11 DOWNSCALING METHODS Scenarios RCM or LAM (Dynamic Downscaling) Stochastic Weather Generators GCM Statistical Models (Statistical Downscaling) Weather Typing or Classification Impact Models (Hydrologic Models) Regression Models low resolution ~ 300 km month, season, year December 17, 2007, Singapore Climate Change Symposium high resolution 1 km day, hour, minute 12 (SPATIAL) DYNAMIC DOWNSCALING METHODS Coarse GCM + High resolution AGCM Variable resolution GCM (high resolution over the area of interest) GCM + RCM or LAM (Nested Modeling Approach) More accurate downscaled results as compared to the use of GCM outputs alone. Spatial scales for RCM results ~ 20 to 50 km still larges for many hydrologic models. Considerable computing resource requirement. December 17, 2007, Singapore Climate Change Symposium 13 (SPATIAL) STATISTICAL DOWNSCALING METHODS Weather Typing or Classification Generation daily weather series at a local site. Classification schemes are somewhat subjective. Stochastic Weather Generators Generation of realistic statistical properties of daily weather series at a local site. Inexpensive computing resources Climate change scenarios based on results predicted by GCM (unreliable for precipitation) Regression-Based Approaches Generation daily weather series at a local site. Results limited to local climatic conditions. Long series of historical data needed. Large-scale and local-scale parameter relations remain valid for future climate conditions. Simple computational requirements. December 17, 2007, Singapore Climate Change Symposium 14 APPLICATIONS LARS-WG Stochastic Weather Generator (Semenov et al., 1998) Generation of synthetic series of daily weather data at a local site (daily precipitation, maximum and minimum temperature, and daily solar radiation) Procedure: Use semi-empirical probability distributions to describe the state of a day (wet or dry). Use semi-empirical distributions for precipitation amounts (parameters estimated for each month). Use normal distributions for daily minimum and maximum temperatures. These distributions are conditioned on the wet/dry status of the day. Constant Lag-1 autocorrelation and cross-correlation are assumed. Use semi-empirical distribution for daily solar radiation. This distribution is conditioned on the wet/dry status of the day. Constant Lag-1 autocorrelation is assumed. December 17, 2007, Singapore Climate Change Symposium 15 Statistical Downscaling Model (SDSM) (Wilby et al., 2001) Generation of synthetic series of daily weather data at a local site based on empirical relationships between local-scale predictands (daily temperature and precipitation) and largescale predictors (atmospheric variables) Procedure: Identify large-scale predictors (X) that could control the local parameters (Y). Find a statistical relationship between X and Y. Validate the relationship with independent data. Generate Y using values of X from GCM data. December 17, 2007, Singapore Climate Change Symposium 16 Some Current Developments The Markov Chain, Mixed Exponential (MCME) Model for Daily Rainfall: Daily rainfall occurrences (First-Order Two-State Markov Chain) pij ,n P X ,n j | X ,n 1 i for n 1 p00,k a00,k a10,k p10,k a00,k a01,k a10,k a11,k Daily rainfall amounts (Mixed exponential distribution) f ( x) p 1 e ( x 1 ) (1 p) December 17, 2007, Singapore Climate Change Symposium 2 e ( x 2 ) 17 AN MCME-BASED DOWNSCALING METHOD n Minimize Z w1AMPi meanMCME w2 AMPi meandownscaledGCM AMPi observed i 1 AMPs by MCME Downscaled-GCM AMPs by SDSM method w 1 + w2 = 1 December 17, 2007, Singapore Climate Change Symposium 18 A STATISTICAL DOWNSCALING METHOD USING PRINCIPAL COMPONENT REGRESSION n n Oi 0 j ij Ai 0 j 1 j ij j 1 Oi = precipitation occurrence on day i Ai = precipitation amount on day i Pij = principal components of predictor climate variables α , β = parameters ε = residual December 17, 2007, Singapore Climate Change Symposium 19 DATA: Observed daily precipitation and temperature extremes at four sites in the Greater Montreal Region (Quebec, Canada) for the 1961-1990 period. NCEP re-analysis daily data for the 1961-1990 period. Calibration: 1961-1975; validation: 1976-1990. Variable Mean sea level pressure Airflow strength Zonal velocity Meridional velocity Vorticity Wind direction Divergence Specific humidity Geopotential height Level of measurement surface surface surface surface surface surface near surface December 17, 2007, Singapore Climate Change Symposium 850 hPa 850 hPa 850 hPa 850 hPa 850 hPa 850 hPa 850 hPa 850 hPa 500 hPa 500 hPa 500 hPa 500 hPa 500 hPa 500 hPa 500 hPa 500 hPa 20 EVALUATION INDICES December 17, 2007, Singapore Climate Change Symposium 21 Geographical locations of sites under study. Geographical coordinates of the stations Station Dorval Drummondville Maniwaki Montreal McGill December 17, 2007, Singapore Climate Change Symposium Lat (o) 45o28’05” 45o52’34” 46o18’11” 45o30’00” Long (o) -73o44’31” -72o28’29” -76o00’36” -73o34’19” Alt (m) 35.7 76.0 192.0 56.9 22 The mean of daily precipitation for the period of 1961-1975 (mm) 14 Dorval 12 10 8 6 Dorval 4 OBS SDSM LARS 2 (mm) 0 J F M A M J J A S O N D OBSERVED vs. SDSM-GENERATED MEAN (mm) 16 14 12 10 8 6 4 2 0 Dorval 15 10 5 0 -5 J F M A M J J A S O N D BIAS J F M A M J J A S O N D OBSERVED vs. LARS-WG-GENERATED MEAN December 17, 2007, Singapore Climate Change Symposium 23 The mean of daily precipitation for the period of 1976-1990 (mm) 14 Dorval 12 10 8 6 Dorval 4 OBS SDSM LARS 2 0 (mm) J F M A M J J A S O N D OBSERVED vs. SDSM-GENERATED MEAN (mm) 14 Dorval 12 15 10 5 0 -5 J F M A M J J A S O N D 10 8 BIAS 6 4 2 0 J F M A M J J A S O N D OBSERVED vs. LARS-WG-GENERATED MEAN December 17, 2007, Singapore Climate Change Symposium 24 The 90th percentile of daily precipitation for the period of 1976-1990 (mm) 35 Dorval 30 25 20 15 Dorval 10 OBS SDSM LARS 5 0 (mm) J F M A M J J A S O N D OBSERVED vs. SDSM-GENERATED 90th PERCENTILE (mm) 40 35 30 25 20 15 10 5 0 J 40 20 0 Dorval -20 J F M A M J J A S O N D BIAS F M A M J J A S O N D OBSERVED vs. LARS-WG-GENERATED 90th PERCENTILE December 17, 2007, Singapore Climate Change Symposium 25 The mean of daily tmax for the period of 1976-1990 (oC) 30 McGill 20 10 0 -10 J F M A M J J A S O N D OBSERVED vs. SDSM- AND LARS-WG-MEAN OF TMAX McGill (oC) BIAS OBS SDSM LARS 30 20 10 0 -10 J F M A M December 17, 2007, Singapore Climate Change Symposium J J A S O N D 26 The 90th percentile of daily tmax for the period of 1976-1990 (oC) 35 30 25 20 15 10 5 0 McGill J F M A M J J A S O N D OBSERVED vs. SDSM- AND LARS-WG-90th PERCENTILE OF TMAX McGill (oC) BIAS OBS SDSM LARS 40 30 20 10 0 -10 J F M A M J December 17, 2007, Singapore Climate Change Symposium J A S O N D 27 The mean of daily tmin for the period of 1976-1990 (oC) 20 15 10 5 0 -5 -10 -15 -20 Drummondville J F M A M J J A S O N D OBSERVED vs. SDSM- AND LARS-WG-MEAN OF TMIN Drummondville (oC) BIAS OBS SDSM LARS 20 10 0 -10 -20 J F M A M December 17, 2007, Singapore Climate Change Symposium J J A S O N D 28 The 10th percentile of daily tmin for the period of 1976-1990 (oC) 20 Drummondville 10 0 -10 -20 -30 -40 J F M A M J J A S O N D OBSERVED vs. SDSM- AND LARS-WG-10th PERCENTILE OF TMIN Drummondville (oC) BIAS OBS SDSM LARS 20 0 -20 -40 J F M A M December 17, 2007, Singapore Climate Change Symposium J J A S O N D 29 GCM and Downscaling Results (Daily Temperature Extremes ) 1- Observed 2- SDSM [CGCM1] 3- SDSM [HADCM3] 4- CGCM1-Raw data 5- HADCM3-Raw data From CCAF Project Report by Gachon et al. (2005) December 17, 2007, Singapore Climate Change Symposium 30 GCM and Downscaling Results (Precipitation Extremes ) 1- Observed 2- SDSM [CGCM1] 3- SDSM [HADCM3] 4- CGCM1-Raw data 5- HADCM3-Raw data From CCAF Project Report by Gachon et al. (2005) December 17, 2007, Singapore Climate Change Symposium 31 SUMMARY Downscaling is necessary!!! LARS-WG and SDSM models could describe well basic statistical properties of the daily temperature extremes at a local site, but both models were unable to reproduce accurately the observed statistics of daily precipitation. GCM-Simulated Daily Precipitation Series Is it feasible? Daily Extreme Precipitations December 17, 2007, Singapore Climate Change Symposium 32 APPLICATION OF MCME-BASED DOWNSCALING METHOD Sooke Reservoir (1500 mm) Dorval (897 mm) Roxas City (2029 mm) DATA: • 30-year daily rainfall record at Dorval Airport (Quebec), Sooke Reservoir (BC), and Roxas City (Philippines) for the 1961-1990 period. • Calibration: 1961-1980 • Validation: 1981-1990 December 17, 2007, Singapore Climate Change Symposium 33 Estimation of MCME Model Parameters MCME Model Parameters: p00 , p10 , p, 1 , 2 Seasonal Variability: Parameters estimated for each month. Observed frequencies of daily rainfall occurrences for estimation of p00 and p10 Maximum likelihood method for estimation of p, μ1, and μ 2. Multi-start (MSX) procedure using the local simplex technique (Nelder and Mead, 1965): A good guess of initial value; otherwise, no convergence to optimal solution. Shuffled Complex Evolution (SCE) method (Duan et al., 1993): Random search + local search, more accurate and more robust. December 17, 2007, Singapore Climate Change Symposium 34 Mixed Exponential Model for Daily Rainfall Amounts Dorval Roxas City December 17, 2007, Singapore Climate Change Symposium 35 Dorval Roxas City Transition Probabilities December 17, 2007, Singapore Climate Change Symposium 36 Dorval: Mean Standard deviation Roxas City: Mean Standard deviation December 17, 2007, Singapore Climate Change Symposium 37 Dorval Physical Properties 1: Observed 2: MCME Model (100 simulations for June-July-August) Roxas City PRECIPITATION CHARACTERISTIC INDEX DEFINITION UNIT Frequency Prcp1 Percentage of wet days (Threshold of 1 mm) % Intensity SDII Simple daily intensity index: sum of daily precipitation divided by the number of wet days mm/number of wet days CDD Maximum number of consecutive dry days (<1mm) days R3days Maximum 3-day precipitation total mm Prec90p 90th percentile of rainy amount mm/day R90N Number of days precipitation exceeds the 90th percentile days Extremes: Magnitude and Occurrence December 17, 2007, Singapore Climate Change Symposium 38 MCME CGCM HadCM3 December 17, 2007, Singapore Climate Change Symposium Calibration: 1961-1980 39 CGCM HadCM3 Model 1 Validation: 1981-1990 Calibration ('61-'80) Validation ('81-'90) MAE RMSE MAE RMSE MCME 3.00 3.77 4.35 4.08 HadCM3 3.52 4.94 5.66 4.69 MCME+HadCM3 2.83 3.40 4.23 3.56 Model 2 Calibration ('61-'80) Validation ('81-'90) MAE RMSE MAE RMSE MCME 3.00 3.77 4.35 4.08 CGCM 3.16 3.71 3.72 3.01 MCME+CGCM 3.09 3.63 3.81 3.23 December 17, 2007, Singapore Climate Change Symposium 40 APPLICATION OF DOWNSCALING USNG PRINCIPAL COMPONENT REGRESSION December 17, 2007, Singapore Climate Change Symposium 41 1: Annual PC; 2: Seasonal PC; 3: Stepwise; and 4: SDSM December 17, 2007, Singapore Climate Change Symposium (1976-1990) 42 1: Annual PC; 2: Seasonal PC; 3 Stepwise; and 4: SDSM December 17, 2007, Singapore Climate Change Symposium 43 1: Annual PC; 2: Seasonal PC; 3 Stepwise; and 4: SDSM December 17, 2007, Singapore Climate Change Symposium 44 Daily AMPs estimated from GCMs versus observed daily AMPs at Dorval. Calibration period: 1961-1975 Dist. of AM Daily Precip. before and after adjustment,1961-1975, Dorval 100 90 90 80 80 AM Daily Precipitation (mm) AM Daily Precipitation (mm) Dist. of AM Daily Precip. before and after adjustment,1961-1975, Dorval 100 70 60 50 Observed 40 70 60 50 Observed 40 CGCM2A2 HadCM3A2 Adj-CGCM2A2 30 0 10 1 10 Adj-HadCM3A2 2 10 Return period (years) CGCMA2 December 17, 2007, Singapore Climate Change Symposium 30 0 10 1 10 2 10 Return period (years) HadCM3A2 45 Residual = Daily AMP (GCM) - Observed daily AMP (local) Calibration period: 1961-1975 HadCM3A2 estimates vs Residuals, 1961-1975 CGCM2A2 estimates vs Residuals, 1961-1975 25 16 14 20 12 10 Residuals Residuals 15 8 6 10 4 2 5 0 Residuals Residuals Fitted curve Fitted curve -2 30 35 40 45 50 55 60 65 70 75 80 CGCM2A2 estimates CGCMA2 December 17, 2007, Singapore Climate Change Symposium 0 30 35 40 45 50 55 60 65 70 75 HadCM3A2 estimates HadCM3A2 46 Daily AMPs estimated from GCMs versus observed daily AMPs at Dorval. Validation period: 1976-1990 Dist. of AM Daily Precip. before and after adjustment,1961-1975, Dorval 100 90 90 80 80 AM Daily Precipitation (mm) AM Daily Precipitation (mm) Dist. of AM Daily Precip. before and after adjustment,1961-1975, Dorval 100 70 60 50 Observed 40 70 60 50 Observed 40 CGCM2A2 HadCM3A2 Adj-CGCM2A2 30 0 10 1 10 Adj-HadCM3A2 2 10 Return period (years) CGCMA2 30 0 10 1 10 2 10 Return period (years) HadCM3A2 Adjusted Daily AMP (GCM) = Daily AMP (GCM) + Residual December 17, 2007, Singapore Climate Change Symposium 47 CONCLUSIONS (1) Significant advances have been achieved regarding the global climate modeling. However, GCM outputs are still not appropriate for assessing climate change impacts on the hydrologic cycle. Downscaling methods provide useful tools for this assessment. Calibration of the SDSM suggested that: precipitation was mainly related to zonal velocities, meridional velocities, specific humidities, geopotential height, and vorticity; tmax and tmin were strongly related to geopotential heights and specific humidities at all levels. LARS-WG and SDSM models could describe well basic statistical properties of the daily temperature extremes at a local site, but both models could provide “good” but “biased” estimates of the observed statistical properties of the daily precipitation process. The MCME model could describe from good to excellent many important (statistical and physical) properties of daily rainfall time series. It is feasible to link local-scale MCME rainfall extreme simulations with large-scale climate variable simulations. December 17, 2007, Singapore Climate Change Symposium 48 CONCLUSIONS (2) The proposed PC regression models outperform the SDSM and the stepwise model in the prediction of the mean and standard deviation of the observed series. The PC regression models are more accurate than the SDSM in reproducing the SDII, R3days and Prec90p for the winter, spring and autumn seasons, and has comparable performance for the summer season and for other indices. The principal component analysis created statistically and physically meaningful groupings of the NCEP predictor variables. It is feasible to link daily GCM-simulated AMPs with observed daily AMPs at a local site using a second-order nonlinear biascorrection function. Hence, the impacts of climate change for different scenarios on daily AMPs could be described. Choice of the “best” downscaling method requires rigorous evaluation (study objectives and region of interest). December 17, 2007, Singapore Climate Change Symposium 49 ... Thank You! December 17, 2007, Singapore Climate Change Symposium 50 Validation of GCMs for Current Period (1961-1990) Winter Temperature (°C) Model mean =all flux & non-flux corrected results (vs NCEP/NCAR dataset) December 17, 2007, Singapore Climate Change Symposium [Source: IPCC TAR, 2001, chap. 8] 51 Climate Scenario development need: from coarse to high resolution A mismatch of scales between what climate models can supply and what environmental impact models require. December 17, 2007, Singapore Climate Change Symposium Point GCMs or RCMs supply... 1m 10km 50km 300km Impact models require ... P. Gachon 52 I (mm/hr) True image time (hr) I (mm/hr) time (hr) December 17, 2007, Singapore Climate Change Symposium 53 December 17, 2007, Singapore Climate Change Symposium 54 UNKNOWN TRUE IMAGE A A1 A2 Α1 Α2 Α December 17, 2007, Singapore Climate Change Symposium 55 Climate Trends and Variability 1950-1998 Maximum and minimum temperatures have increased at similar rate Warming in the south and west, and cooling in the northeast (winter & spring) Trends in Winter Mean Temp (°C / 49 years) Trends in Spring Mean Temp (°C / 49 years) Trends in Summer Mean Temp (°C / 49 years) Trends in Fall Mean Temp (°C / 49 years) From X. Zhang, L. Vincent, B. Hogg and A. Niitsoo, Atmosphere-Ocean, 2000 December 17, 2007, Singapore Climate Change Symposium 56