Performance comparison of three predictor selection methods for statistical downscaling of daily precipitation

Predictor selection is a critical factor affecting the statistical downscaling of daily precipitation. This study provides a general comparison between uncertainties in downscaled results from three commonly used predictor selection methods (correlation analysis, partial correlation analysis, and stepwise regression analysis). Uncertainty is analyzed by comparing statistical indices, including the mean, variance, and the distribution of monthly mean daily precipitation, wet spell length, and the number of wet days. The downscaled results are produced by the artificial neural network (ANN) statistical downscaling model and 50 years (1961–2010) of observed daily precipitation together with reanalysis predictors. Although results show little difference between downscaling methods, stepwise regression analysis is generally the best method for selecting predictors for the ANN statistical downscaling model of daily precipitation, followed by partial correlation analysis and then correlation analysis.     

Statistical Downscaling of Pattern Projection Using Multi-Model Output Variables as Predictors

A pattern projection downscaling method is employed to predict monthly station precipitation. The predictand is the monthly precipitation at 1 station in China, 60 stations in Korea, and 8 stations in Thailand. The predictors are multiple variables from the output of operational dynamical models. The hindcast datasets span a period of 21 yr from 1983 to 2003. A downscaled prediction is made for each model separately within a leave-one-out cross-validation framework. The pattern projection method uses a moving window, which scans globally, in order to seek the most optimal predictor for each station. The final forecast is the average of the model downscaled precipitation forecasts using the best predictors and is referred to as DMME. It is found that DMME significantly improves the prediction skill by correcting the erroneous signs of the rainfall anomalies in coarse resolution predictions of general circulation models. The correlation coefficient between the prediction of DMME and the observation in Beijing of China reaches 0.71; the skill is improved to 0.75 for Korea and 0.61 for Thailand. The improvement of the prediction skills for the first two cases is attributed to three steps: coupled pattern selection, optimal predictor selection, and multi-model downscaled precipitation ensemble. For Thailand, we use the single-predictor prediction, which results in a lower prediction skill than the other two cases. This study indicates that the large-scale circulation variables, which are predicted by the current operational dynamical models, if selected well, can be used to make skillful predictions of local precipitation by means of appropriate statistical downscaling.     

多模式集合预报及其降尺度技术在东亚夏季降水预测中的应用

利用动力季节模式输出的匹配域投影技术和多模式集合预报技术对多个国家和城市的站点月平均降水进行预报。预报变量是北京1个站、韩国60个站和曼谷地区8个站点的月平均降水,预报因子是从多个业务动力季节预报模式输出的多个大尺度变量。模式回报数据和站点观测降水数据时段是1983—2003年。降尺度预报降水的技巧是在交叉验证的框架下进行的。匹配域投影方法是设定一个可以活动的窗口在全球范围内大尺度场上进行扫描,寻求与目标站点降水最优化的因子和最相关的区域,目标站点的降水变率就是由该匹配域上大尺度环流场信息决定的。最终预报是用多个降尺度模式预报结果的集合预报(DMME)。多个降尺度模式预报结果的集合预报能显著地提高站点降水的预报技巧。北京站,多个降尺度模式预报结果的集合预报的预报和观测降水的相关系数可以提高到0.71;韩国地区,多个降尺度模式预报结果的集合预报平均技巧提高到0.75;泰国,多个降尺度模式预报结果的集合预报技巧是0.61。     

Evaluation of statistical downscaling in short range precipitation forecasting

The objective of this study is to compare several statistical downscaling methods for the development of an operational short-term forecast of precipitation in the area of Bilbao (Spain). The ability of statistical downscaling methods nested inside numerical simulations run by both coarse and regional model simulations is tested with several selections of predictors and domain sizes. The selection of predictors is performed both in terms of sound physical mechanisms and also by means of “blind” criteria, such as “give the statistical downscaling methods all the information they can process”.Results show that the use of statistical downscaling methods improves the ability of the mesoscale and coarse resolution models to provide quantitative precipitation forecasts. The selection of predictors in terms of sound physical principles does not necessarily improve the ability of the statistical downscaling method to select the most relevant inputs to feed the precipitation forecasting model, due to the fact that the numerical models do not always fulfil conservation laws or because precipitation events do not reflect simple phenomenological laws. Coarse resolution models are able to provide information usable in combination with a statistical downscaling method to achieve a quantitative precipitation forecast skill comparable to that obtained by other systems currently in use.     

Uncertainty analysis of statistical downscaling models using Hadley Centre Coupled Model

Regression-based statistical downscaling is a method broadly used to resolve the coarse spatial resolution of general circulation models. Nevertheless, the assessment of uncertainties linked with climatic variables is essential to climate impact studies. This study presents a procedure to characterize the uncertainty in regression-based statistical downscaling of daily precipitation and temperature over a highly vulnerable area (semiarid catchment) in the west of Iran, based on two downscaling models: a statistical downscaling model (SDSM) and an artificial neural network (ANN) model. Biases in mean, variance, and wet/dry spells are estimated for downscaled data using vigorous statistical tests for 30 years of observed and downscaled daily precipitation and temperature data taken from the National Center for Environmental Prediction reanalysis predictors for the years of 1961 to 1990. In the case of daily temperature, uncertainty is estimated by comparing monthly mean and variance of downscaled and observed daily data at a 95 % confidence level. In daily precipitation, downscaling uncertainties were evaluated from comparing monthly mean dry and wet spell lengths and their confidence intervals, cumulative frequency distributions of monthly mean of daily precipitation, and the distributions of monthly wet and dry days for observed and modeled daily precipitation. Results showed that uncertainty in downscaled precipitation is high, but simulation of daily temperature can reproduce extreme events accurately. Finally, this study shows that the SDSM is the most proficient model at reproducing various statistical characteristics of observed data at a 95 % confidence level, while the ANN model is the least capable in this respect. This study attempts to test uncertainties of regression-based statistical downscaling techniques in a semiarid area and therefore contributes to an improvement of the quality of predictions of climate change impact assessment in regions of this type.     

华北汛期降水分离时间尺度降尺度预测模型的改进

本文采用偏相关预报因子挑选法和条件降尺度法,对已有的华北汛期(7~8月)降水时间尺度分离(TSD)降尺度模型进行了改进.利用偏相关法,找到一个新的影响华北汛期降水年际分量的前期预报因子,即6月北大西洋—欧亚遥相关(AEAT).该因子将扰动信号储存于北大西洋三极子结构,并在7~8月释放出来影响下游贝加尔湖低压系统的发展,从而影响华北汛期降水.利用6月Ni?o3指数和AEAT指数,本文建立了条件TSD统计降尺度模型,即按照预报因子的强度进行逐年分类,对于每个分类设计相应的预报模型,从而避免信息较弱因子的干扰.条件TSD降尺度方法显著改善了华北汛期降水的预测技巧,在独立检验阶段,预报降水与观测降水的相关系数由原模型的0.61提高到0.77,符号一致率从70%提高到87%.     

Evaluating climate change effects on runoff by statistical downscaling and hydrological model GR2M

The main purpose of this study is to evaluate the impacts of climate change on Izmir-Tahtali freshwater basin, which is located in the Aegean Region of Turkey. For this purpose, a developed strategy involving statistical downscaling and hydrological modeling is illustrated through its application to the basin. Prior to statistical downscaling of precipitation and temperature, the explanatory variables are obtained from National Centers for Environmental Prediction/National Center for Atmospheric Research reanalysis data set. All possible regression approach is used to establish the most parsimonious relationship between precipitation, temperature, and climatic variables. Selected predictors have been used in training of artificial neural networks-based downscaling models and the trained models with the obtained relationships have been operated to produce scenario precipitation and temperature from the simulations of third Generation Coupled Climate Model. Biases from downscaled outputs have been reduced after downscaling process. Finally, the corrected downscaled outputs have been transformed to runoff by means of a monthly parametric hydrological model GR2M to assess the probable impacts of temperature and precipitation changes on runoff. According to the A1B climate scenario results, statistically significant trends are foreseen for precipitation, temperature, and runoff in the study basin.     

统计降尺度法对华北地区未来区域气温变化情景的预估

迄今为止,大部分海气耦合气候模式（AOGCM）的空间分辨率还较低,很难对区域尺度的气候变化情景做合理的预测。降尺度法已广泛用于弥补AOGCM在这方面的不足。作者采用统计降尺度方法对1月和7月华北地区49个气象观测站的未来月平均温度变化情景进行预估。采用的统计降尺度方法是主分量分析与逐步回归分析相结合的多元线性回归模型。首先,采用1961～2000年的 NCEP再分析资料和49个台站的观测资料建立月平均温度的统计降尺度模型,然后把建立的统计降尺度模型应用于HadCM3 SRES A2 和 B2 两种排放情景, 从而生成各个台站1950～2099年1月份和7月份温度变化情景。结果表明:在当前气候条件下,无论1月还是7月,统计降尺度方法模拟的温度与观测的温度有很好的一致性,而且在大多数台站,统计降尺度模拟气温与观测值相比略微偏低。对于未来气候情景的预估方面,无论1月还是7月,也无论是HadCM3 SRES A2 还是B2排放情景驱动统计模型,结果表明大多数的站点都存在温度的明显上升趋势,同时7月的上升趋势与1月相比偏低。     

Statistical Downscaling Based on Dynamically DownscaledPredictors： Application to Monthly Precipitation in Sweden

A prerequisite of a successful statistical downscaling is that large-scale predictors simulated by the General Circulation Model (GCM) must be realistic. It is assumed here that features smaller than the GCM resolution are important in determining the realism of the large-scale predictors. It is tested whether a three-step method can improve conventional one-step statistical downscaling. The method uses predictors that are upscaled from a dynamical downscaling instead of predictors taken directly from a GCM simulation. The method is applied to downscaling of monthly precipitation in Sweden. The statistical model used is a multiple regression model that uses indices of large-scale atmospheric circulation and 850-hPa specific humidity as predictors. Data from two GCMs (HadCM2 and ECHAM4) and two RCM experiments of the Rossby Centre model (RCA1) driven by the GCMs are used. It is found that upscaled RCA1 predictors capture the seasonal cycle better than those from the GCMs, and hence increase the reliability of the downscaled precipitation. However, there are only slight improvements in the simulation of the seasonal cycle of downscaled precipitation. Due to the cost of the method and the limited improvements in the downscaling results, the three-step method is not justified to replace the one-step method for downscaling of Swedish precipitation.     

DOWNSCALING FORECAST OF MONTHLY PRECIPITATION OVER GUANGXI BASED ON BP NEURAL NETWORK MODEL

Variables fields such as enstrophy, meridional-wind and zonal-wind variables are derived from monthly 500 hPa geopotential height anomalous fields. In this work, we select original predictors from monthly 500-hPa geopotential height anomalous fields and their variables in June of 1958 - 2001, and determine comprehensive predictors by conducting empirical orthogonal function (EOF) respectively with the original predictors. A downscaling forecast model based on the back propagation (BP) neural network is built by use of the comprehensive predictors to predict the monthly precipitation in June over Guangxi with the monthly dynamic extended range forecast products. For comparison, we also build another BP neural network model with the same predictands by using the former comprehensive predictors selected from 500-hPa geopotential height anomalous fields in May to December of 1957 - 2000 and January to April of 1958 - 2001. The two models are tested and results show that the precision of superposition of the downscaling model is better than that of the one based on former comprehensive predictors, but the prediction accuracy of the downscaling model depends on the output of monthly dynamic extended range forecast.     

CGCM3 predictors used for daily temperature and precipitation downscaling in Southern Québec, Canada

This study provides some guidance on the choice of predictor variables from both reanalysis products and the third version of the Canadian Coupled Global Climate Model (CGCM3) outputs for regression-based statistical downscaling models (SDMs) for climate change application in southern Québec (Canada). Twenty CGCM3 grid points and four surface observation sites in the study area were employed. Twenty-five deseasonalized predictors and four deseasonalized predictands (daily maximum and minimum temperatures, precipitation occurrence and wet day precipitation amount) were used to investigate correlation coefficients among predictors and to evaluate their predictive ability when used in a multiple linear regression (MLR) downscaling model. The basic statistical characteristics of vorticity at 1,000-, 850- and 500-hPa levels, U-component of velocity at 1,000-hPa level, temperature at 2?m (T ₂) and wind direction at 1,000- and 500-hPa level of CGCM3 showed a larger difference with those of the NCEP reanalysis data. Therefore, those seven variables require high caution to be included as predictors in statistical downscaling models. Specific humidity at 1,000-, 850- and 500-hPa levels, geopotential height at 850- and 500-hPa levels and T ₂ were the most sensitive predictors for future climate conditions (i.e. A1B and A2 emission scenarios). Specific humidity and geopotential height at different levels and T ₂ were important explainable predictors for the daily temperatures. Mean sea level pressure, specific humidity, U and V components and divergence showed potential as predictors for daily precipitation. Spatial explained variance of MLRs between predictors of every different CGCM3 grid points and the four predictands showed large values at the CGCM3 grid points located near the observation sites, whereas relatively small values were shown at the CGCM3 grid points located more than 400?km from the sites. The explained variance of the downscaled predictands by predictors of three or four CGCM3 grid points located near the observation site produced 2–5% larger R-squares than those by predictors of the nearest grid point. The results illustrated that the use of predictors from more than one AOGCM grid points located near the observation site can increase the skill of the MLR downscaling models.     

Comparison of regional climate model and statistical downscaling simulations of different winter precipitation change scenarios over Romania

Summary Regional climate model and statistical downscaling procedures are used to generate winter precipitation changes over Romania for the period 2071–2100 (compared to 1961–1990), under the IPCC A2 and B2 emission scenarios. For this purpose, the ICTP regional climate model RegCM is nested within the Hadley Centre global atmospheric model HadAM3H. The statistical downscaling method is based on the use of canonical correlation analysis (CCA) to construct climate change scenarios for winter precipitation over Romania from two predictors, sea level pressure and specific humidity (either used individually or together). A technique to select the most skillful model separately for each station is proposed to optimise the statistical downscaling signal. Climate fields from the A2 and B2 scenario simulations with the HadAM3H and RegCM models are used as input to the statistical downscaling model. First, the capability of the climate models to reproduce the observed link between winter precipitation over Romania and atmospheric circulation at the European scale is analysed, showing that the RegCM is more accurate than HadAM3H in the simulation of Romanian precipitation variability and its connection with large-scale circulations. Both models overestimate winter precipitation in the eastern regions of Romania due to an overestimation of the intensity and frequency of cyclonic systems over Europe. Climate changes derived directly from the RegCM and HadAM3H show an increase of precipitation during the 2071–2100 period compared to 1961–1990, especially over northwest and northeast Romania. Similar climate change patterns are obtained through the statistical downscaling method when the technique of optimum model selected separately for each station is used. This adds confidence to the simulated climate change signal over this region. The uncertainty of results is higher for the eastern and southeastern regions of Romania due to the lower HadAM3H and RegCM performance in simulating winter precipitation variability there as well as the reduced skill of the statistical downscaling model.     

Coupling statistical and dynamical methods for spatial downscaling of precipitation

The resolution of General Circulation Models (GCMs) is too coarse for climate change impact studies at the catchment or site-specific scales. To overcome this problem, both dynamical and statistical downscaling methods have been developed. Each downscaling method has its advantages and drawbacks, which have been described in great detail in the literature. This paper evaluates the improvement in statistical downscaling (SD) predictive power when using predictors from a Regional Climate Model (RCM) over a GCM for downscaling site-specific precipitation. Our approach uses mixed downscaling, combining both dynamic and statistical methods. Precipitation, a critical element of hydrology studies that is also much more difficult to downscale than temperature, is the only variable evaluated in this study. The SD method selected here uses a stepwise linear regression approach for precipitation quantity and occurrence (similar to the well-known Statistical Downscaling Model (SDSM) and called SDSM-like herein). In addition, a discriminant analysis (DA) was tested to generate precipitation occurrence, and a weather typing approach was used to derive statistical relationships based on weather types, and not only on a seasonal basis as is usually done. The existing data record was separated into a calibration and validation periods. To compare the relative efficiency of the SD approaches, relationships were derived at the same sites using the same predictors at a 300km scale (the National Center for Environmental Prediction (NCEP) reanalysis) and at a 45km scale with data from the limited-area Canadian Regional Climate Model (CRCM) driven by NCEP data at its boundaries. Predictably, using CRCM variables as predictors rather than NCEP data resulted in a much-improved explained variance for precipitation, although it was always less than 50?% overall. For precipitation occurrence, the SDSM-like model slightly overestimated the frequencies of wet and dry periods, while these were well-replicated by the DA-based model. Both the SDSM-like and DA-based models reproduced the percentage of wet days, but the wet and dry statuses for each day were poorly downscaled by both approaches. Overall, precipitation occurrence downscaled by the DA-based model was much better than that predicted by the SDSM-like model. Despite the added complexity, the weather typing approach was not better at downscaling precipitation than approaches without classification. Overall, despite significant improvements in precipitation occurrence prediction by the DA scheme, and even going to finer scales predictors, the SD approach tested here still explained less than 50?% of the total precipitation variance. While going to even smaller scale predictors (10–15?km) might improve results even more, such smaller scales would basically transform the direct outputs of climate models into impact models, thus negating the need for statistical downscaling approaches.     

N–PLS regression as empirical downscaling tool in climate change studies

Summary An N–PLS regression technique was tested as an empirical downscaling method. Average monthly near-ground air temperature (t), specific humidity (q), and sea-level pressure (p) fields across Central and Western Europe were used as predictors for average monthly air temperature (T), dew temperature (D), and precipitation amount (P) at 4 locations in Slovenia. The empirical downscaling models (EM) were developed by means of available predictand data from the ARSO archive and predictor data from the NCEP/NCAR reanalysis project, for the period 1951–2002, separately for single months. Using the combination of t and p as predictors, the EM for T explained from 73% to 95% of predictand variability, for D from 74% to 97% of predictand variability, and for P from 31% to 76% of predictand variability. The use of q as an additional predictor did not improve the quality of the EM considerably. Developed EM using p and t as predictors were applied to the results of 5 general circulation models (GCM): CSIRO/Mk2, CCC/CGCM2, UKMO/HadCM3, DOE-NCAR/PCM, and MPI-DMI/ECHAM4-OPYC3. Only the simulations based on SRES A2 and B2 emission scenarios were considered in our calculations. Available mean monthly values of predictors for the period 1951–2100 were used. All the projections of GCM results indicate an increase in T and D and decrease in P in the 21^st century at all 4 locations. The expected range of changes in T, D, and P is wide due to the different response of GCM to identical changes in the atmospheric composition, and represents a source of uncertainty in empirical downscaling results. Another important source of uncertainty in empirical downscaling studies, especially when temperature dependent predictors are used, is the problem of extrapolation. By using the proper mathematical approach for EM development we only reduce a part of the uncertainty related to the quality of empirical models that also strongly depend on the quality of input data and predictor selection. The N–PLS regression seems to be a suitable choice of mathematical method, as the feature selection from a large number of predictor time series is not predictand independent. Finally, any climate change and impact studies for the future are affected by many other uncertainties that we have to be fully aware of, while interpreting their results.     

An intercomparison of observational precipitation data sets over Northwest India during winter

Development of downscaling models for each calendar month using the data of predictors specifically selected for each calendar month may assists in better capturing the time-varying nature of the predictor-predictand relationships. Such approach will not allow the explicit modelling of the persistence of the predictand (e.g. lag-1 correlation). However, downscaling at an annual time step and subsequent disaggregation to monthly values can explicitly consider the modelling of the persistence of the predictand. This study investigated the potential of annual downscaling of a predictand and subsequent disaggregation of annual values to monthly values, in comparison to the potential of downscaling models separately developed for each calendar month. In the case study, annual and monthly downscaling models were developed for precipitation and evaporation at two stations located in Victoria, Australia. The output of the annual downscaling models was then disaggregated into monthly values using four different methods based on the method of fragments. It was found that the annual to monthly disaggregation methods and monthly downscaling models are able to reproduce the average of monthly observations with relatively higher accuracy in comparison to their ability in reproducing standard deviation, skewness and lag-1 serial correlation. Downscaling models separately developed for each calendar month were able to show relatively smaller root mean square errors for their time series indicating better overall agreement with observations in comparison to their counterpart annual to monthly disaggregation methods. Furthermore, it was found that not only the bias in the output of an annual downscaling model but also the presence of annual totals in the records of observations of a predictand that are very similar in magnitude, but having significantly different sets of fragments, can largely contribute to the poor performance of an annual to monthly disaggregation method.     

中国气温的降尺度研究

Monthly mean temperatures at 562 stations in China are estimated using a statistical downscaling technique. The technique used is multiple linear regressions （MLRs） of principal components （PCs）. A stepwise screening procedure is used for selecting the skilful PCs as predictors used in the regression equation. The predictors include temperature at 850 hPa （7）, the combination of sea-level pressure and temperature at 850 hPa （P＋T） and the combination of geo-potential height and temperature at 850 hPa （H＋T）. The downscaling procedure is tested with the three predictors over three predictor domains. The optimum statistical model is obtained for each station and month by finding the predictor and predictor domain corresponding to the highest correlation. Finally, the optimum statistical downscaling models are applied to the Hadley Centre Coupled Model, version 3 （HadCM3） outputs under the Special Report on Emission Scenarios （SRES） A2 and B2 scenarios to construct local future temperature change scenarios for each station and month, The results show that （1） statistical downscaling produces less warming than the HadCM3 output itself; （2） the downscaled annual cycles of temperature differ from the HadCM3 output, but are similar to the observation; （3） the downscaled temperature scenarios show more warming in the north than in the south; （4） the downscaled temperature scenarios vary with emission scenarios, and the A2 scenario produces more warming than the B2, especially in the north of China.     

Comparison of statistically downscaled precipitation in terms of future climate indices and daily variability for southern Ontario and Quebec,Canada

Given the coarse resolution of global climate models, downscaling techniques are often needed to generate finer scale projections of variables affected by local-scale processes such as precipitation. However, classical statistical downscaling experiments for future climate rely on the time-invariance assumption as one cannot know the true change in the variable of interest, nor validate the models with data not yet observed. Our experimental setup involves using the Canadian regional climate model (CRCM) outputs as pseudo-observations to estimate model performance in the context of future climate projections by replacing historical and future observations with model simulations from the CRCM, nested within the domain of the Canadian global climate model (CGCM). In particular, we evaluated statistically downscaled daily precipitation time series in terms of the Peirce skill score, mean absolute errors, and climate indices. Specifically, we used a variety of linear and nonlinear methods such as artificial neural networks (ANN), decision trees and ensembles, multiple linear regression, and k-nearest neighbors to generate present and future daily precipitation occurrences and amounts. We obtained the predictors from the CGCM 3.1 20C3M (1971–2000) and A2 (2041–2070) simulations, and precipitation outputs from the CRCM 4.2 (forced with the CGCM 3.1 boundary conditions) as predictands. Overall, ANN models and tree ensembles outscored the linear models and simple nonlinear models in terms of precipitation occurrences, without performance deteriorating in future climate. In contrast, for the precipitation amounts and related climate indices, the performance of downscaling models deteriorated in future climate.     

Using second-order approximation to incorporate GCM uncertainty in climate change impact assessments

This study presents a method to incorporate uncertainty of climate variables in climate change impact assessments, where the uncertainty being considered refers to the divergence of general circulation model (GCM) projections. The framework assesses how much bias occurs when the uncertainties of climate variables are ignored. The proposed method is based on the second-order expansion of Taylor series, called second-order approximation (SOA). SOA addresses the bias which occurs by assuming the expected value of a function is equal to the function of the expected value of the predictors. This assumption is not valid for nonlinear systems, such as in the case of the relationship of climate variables to streamflow. To investigate the value of SOA in the climate change context, statistical downscaling models for monthly streamflow were set up for six hydrologic reference stations in Australia which cover contrasting hydro-climate regions. It is shown that in all locations SOA makes the largest difference for low flows and changes the overall mean flow by 1–3%. Another advantage of the SOA approach is that the individual contribution of each climate variable to the total difference can be estimated. It is found that geopotential height and specific humidity cause more bias than wind speeds in the downscaling models considered here.     

BP-CCA方法用于四川盆地夏季日降水量的可预报性研究

基于BP-CCA方法,首先讨论了多个因子对四川盆地夏季降水降尺度模型的可预报性,然后选取最佳预报因子并进行集合,最终基于T639模式建立最优多因子降尺度预报模型.结果表明,分别以东亚夏季10m纬向风、700hPa纬向风和700hPa相对湿度为预报因子的降尺度模型对四川盆地夏季降水的预报技巧较高,而将三个因子集合的多因子降尺度预报模型具有更好的预报能力.进一步将该方法应用于T639模式预报的预报因子场,发现多因子降尺度模型对降水的预报效果要优于T639模式直接输出的结果.     

Statistical downscaling in the arid central Andes: uncertainty analysis of multi-model simulated temperature and precipitation

Statistical downscaling is a technique widely used to overcome the spatial resolution problem of General Circulation Models (GCMs). Nevertheless, the evaluation of uncertainties linked with downscaled temperature and precipitation variables is essential to climate impact studies. This paper shows the potential of a statistical downscaling technique (in this case SDSM) using predictors from three different GCMs (GCGM3, GFDL and MRI) over a highly heterogeneous area in the central Andes. Biases in median and variance are estimated for downscaled temperature and precipitation using robust statistical tests, respectively Mann?CWhitney and Brown?CForsythe's tests. In addition, the ability of the downscaled variables to reproduce extreme events is tested using a frequency analysis. Results show that uncertainties in downscaled precipitations are high and that simulated precipitation variables failed to reproduce extreme events accurately. Nevertheless, a greater confidence remains in downscaled temperatures variables for the area. GCMs performed differently for temperature and precipitation as well as for the different test. In general, this study shows that statistical downscaling is able to simulate with accuracy temperature variables. More inhomogeneities are detected for precipitation variables. This first attempt to test uncertainties of statistical downscaling techniques in the heterogeneous arid central Andes contributes therefore to an improvement of the quality of predictions of climate impact studies in this area.