Modeling Uncertainty Resulting from Multiple Downscaling Methods in Assessing Hydrological Impacts of Climate Change

Climate Change Impacts Assessment using Statistical Downscaling is observed to be characterized by uncertainties resulting from multiple downscaling methods, which may perform similar during training, but differs in projections when applied to GCM outputs of future scenarios. The common wisdom in statistical downscaling, for selection of downscaling algorithms, is to select the model with the best overall system performance measure for observed period (training and testing). However, this does not guarantee that such selection will work best for any rainfall states, viz., low rainfall, or extreme rainfall. In the present study, for Assam and Meghalaya meteorological subdivision, India, three downscaling methods, Linear Regression (LR), Artificial Neural Network (ANN) and Support Vector Machine (SVM) are used for simulating rainfall with reanalysis data and similar training and testing performances are obtained for observed period. When the developed relationships are applied to GCM output for future (21st century), differences are observed in downscaled projections for extreme rainfalls. ANN shows decrease in extreme rainfall, SVM shows increase in extreme rainfall and LR shows first decrease and then increase in extreme rainfall. Such results motivate further investigation, which reveals that, although, the overall performances of training and testing for all the transfer functions are similar, there are significant differences, when the performance measures are computed separately for low, medium, and high rainfall states. To model such uncertainty resulting from multiple downscaling methods, different transfer functions (LR, ANN and SVM) are used for different rainfall states (viz., low, medium and high), where they perform best. The rainfall states are predicted from large scale climate variables using Classification and Regression Tree (CART). As muti-model averaging (with equal weights or performance based weights) is commonly used in climatic sciences, the resulting output are also compared with the average of multiple downscaling model output. Rainfall is projected for Assam and Meghalaya meteorological subdivision, using this logic, with multiple GCMs. GCM uncertainty, resulting from the use of multiple GCMs, is further modeled using reliability ensemble averaging. The resultant Cumulative Distribution Function (CDF) of projected rainfall shows an increasing trend of rainfall in Assam and Meghalaya meteorological subdivision.     

Downscaling Monsoon Rainfall over River Godavari Basin under Different Climate-Change Scenarios

Evaluating the impact of climate change at river basin level has become essential for proper management of the water resources. In the present study, Godavari River basin in India is taken as study area to project the monthly monsoon precipitation using statistical downscaling. The downscaling method used is a regression based downscaling termed as fuzzy clustering with multiple regression. Among the atmospheric variables simulated by global circulation/climate model (GCM) mean sea level pressure, specific humidity and 500 hPa geopotential height are used as predictors. 1^o × 1^o gridded rainfall data over Godavari river basin are collected from India Meteorological Department (IMD). A statistical relationship is established between the predictors and predictand (monsoon rainfall) to project the monsoon rainfall for the future using the Canadian Earth System Model (CanESM2) over IMD grid points under the Representative Concentration Pathways 2.6, 4.5 and 8.5 (RCP 2.6, 4.5, 8.5) scenarios of Fifth Coupled Model Inter-Comparison Project (CMIP 5). Downscaling procedure is applied to all 25 IMD grid points over the basin to find out the spatial distribution of monsoon rainfall for the future scenarios. For 2.6 and 4.5 scenarios results show an increasing trend. For scenario 8.5 rainfall showed a mixed trend with rainfall decreasing in the first thirty years of prediction and then increasing gradually over the next sixty years.     

Examination of Various Feature Selection Approaches for Daily Precipitation Downscaling in Different Climates

To turn General Circulation Models (GCMs) projection toward better assessment, it is crucial to employ a downscaling process to get more reliability of their outputs. The data-driven based downscaling techniques recently have been used widely, and predictor selection is usually considered as the main challenge in these methods. Hence, this study aims to examine the most common approaches of feature selection in the downscaling of daily rainfall in two different climates in Iran. So, the measured daily rainfall and National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) predictors were collected, and Support Vector Machine (SVM) was considered as downscaling methods. Also, a complete set of comparative tests considering all dimensions was employed to identify the best subset of predictors. Results indicated that the skill of various selection methods in different tests is significantly different. Despite a few partial superiorities viewed between selection models, they not presented an obvious distinction. However, regarding all related factors, it may be deduced that the Stepwise Regression Analysis (SRA) and Bayesian Model Averaging (BMA) are better than others. Also, the finding of this study showed that there are some weaknesses in the interpretation of SRA, so concerning this issue, it may be concluded that BMA has more reliable performance. Furthermore, results indicated that generally, the downscaling procedure has more accuracy in arid climate than cold-semi arid climate.

     

The Correlation Between Statistically Downscaled Precipitation Data and Groundwater Level Records in North-Western Turkey

Downscaling of atmospheric climate parameters is a sophisticated tool to develop statistical relationships between large-scale atmospheric variables and local-scale meteorological variables. In this study, the variables selected from the National Centre for Environmental Prediction and National Centre for Atmospheric Research (NCEP/NCAR) reanalysis data set were used as predictors for the downscaling of monthly precipitation in a watershed located in north-western Turkey where station records terminated two decades ago. An Artificial Neural Network (ANN) based approach was used to downscale global climate predictors that are positively correlated to the existing time frame of precipitation data in the basin. The downscaled precipitation information were used to extend the non-existing data from the meteorological station, which were later correlated with groundwater level data obtained from automatic pressure transducers that continuously record depth to groundwater. The results of the study showed that, among a large set of NCEP/NCAR parameters, surface precipitation data recorded at the meteorological station was strongly correlated with precipitation rate, air temperature and relative humidity at surface and air temperature at 850, 500, and 200 hPa pressure levels, and geopotential heights at 850 and 200 hPa pressure levels. The gaps in station data were then filled with the correlations obtained from NCEP/NCAR parameters and a complete precipitation data set was obtained that extended to current time line. This extended precipitation time series was later correlated with the existing groundwater level data from an alluvial plain in order to develop a general relationship that can be used in basin-wide water budget estimations. The proposed methodology is believed to serve the needs of engineers and basin planners who try to create a link between related hydrological variables under data-limited conditions.     

Statistical Downscaling of River Runoff in a Semi Arid Catchment

Linear and non-linear statistical ‘downscaling’ study is applied to relate large-scale climate information from a general circulation model (GCM) to local-scale river flows in west Iran. This study aims to investigate and evaluate the more promising downscaling techniques, and provides a through inter comparison study using Karkheh catchment as an experimental site in a semi arid region for the years of 2040 to 2069. A hybrid conceptual hydrological model was used in conjunction with modeled outcomes from a General Circulation Model (GCM), HadCM3, along with two downscaling techniques, Statistical Downscaling Model (SDSM) and Artificial Neural Network (ANN), to determine how future streamflow may change in a semi arid catchment. The results show that the choice of a downscaling algorithm having a significant impact on the streamflow estimations for a semi-arid catchment, which are mainly, influenced, respectively, by atmospheric precipitation and temperature projections. According to the SDSM and ANN projections, daily temperature will increase up to +0.58 0C (+3.90 %) and +0.48 0C (+3.48 %), and daily precipitation will decrease up to ?0.1 mm (?2.56 %) and ?0.4 mm (?2.82 %) respectively. Moreover streamflow changes corresponding to downscaled future projections presented a reduction in mean annual flow of ?3.7 m^3/s and ?9.47 m^3/s using SDSM and ANN outputs respectively. The results suggest a significant reduction of streamflow in both downscaling projections, particularly in winter. The discussion considers the performance of each statistical method for downscaling future flow at catchment scale as well as the relationship between atmospheric processes and flow variability and changes.     

Probing Norwalk-like virus presence in shellfish, using artificial neural networks.

A database was examined using artificial neural network (ANN) models to investigate the efficacy of predicting PCR-identified Norwalk-like virus presence and absence in shellfish. The relative importance of variables in the model and the predictive power obtained by application of ANN modelling methods were compared with previously developed logistic regression models. In addition, two country-specific datasets were analysed separately with ANN models to determine if the relative importance of the input variables was similar for geographically diverse regions. The results of this analysis found that ANN models predicted Norwalk-like virus presence and absence in shellfish with equivalent, and better, precision than logistic regression models. For overall classification performance, ANN modelling had a rate of 93%, vs 75% for the logistic regression. ANN models were able to illuminate the site-specific relationships between indicators and pathogens.     

Uncertainty Analysis of Climate Change Impacts on Flood Frequency by Using Hybrid Machine Learning Methods

In the present study, for the first time, a new framework is used by combining metaheuristic algorithms, decomposition and machine learning for flood frequency analysis under climate-change conditions and application of HadCM3 (A2 and B2 scenarios), CGCM3 (A2 and A1B scenarios) and CanESM2 (RCP2.6, RCP4.5 and RCP8.5 scenarios) in global climate models (GCM). In the proposed framework, Multivariate Adaptive Regression Splines (MARS) and M5 Model tree are used for classification of precipitation (wet and dry days), whale optimization algorithm (WOA) is considered for training least square support vector machine (LSSVM), wavelet transform (WT) is used for decomposition of precipitation and temperature, LSSVM-WOA, LSSVM, K nearest neighbor (KNN) and artificial neural network (ANN) are performed for downscaling precipitation and temperature, and discharge is simulated under present period (1972–2000), near future (2020–2040) and far future (2070–2100). Log normal distribution is used for flood frequency analysis. Furthermore, analysis of variance (ANOVA) and fuzzy method are employed for uncertainty analysis. Karun3 Basin, in southwest of Iran, is considered as a case study. Results indicated that MARS performed better than M5 model tree. In downscaling, ANN and LSSVM_WOA slightly outperformed other machine learning algorithms. Results of simulating the discharge showed superiority of LSSVM_WOA_WT algorithm (Nash-Sutcliffe efficiency (NSE)?=?0.911). Results of flood frequency analysis revealed that 200-year discharge decreases for all scenarios, except CanESM2 RCP2.6 scenario, in the near future. In the near and far future periods, it is obvious from ANOVA uncertainty analysis that hydrological models are one of the most important sources of uncertainty. Based on the fuzzy uncertainty analysis, HadCM3 model has lower uncertainty in higher return periods (up to 60% lower than other models in 1000-year return period).

     

日雨量随机解集模式研究

全球气候模式(GCMs)预测的气候变化情景，必须经解集模式得出小尺度上未来气候变化时空分布资料，才能满足评估气候变化对资源、环境和社会经济等影响的需要。本文提出由随机天气生成器和统计参数尺度转换关系组成的随机解集模式，应用17个站32年实测日降雨资料，对随机解集模式进行了分析和验证。首先利用随机天气生成器，通过对站点和GCM尺度面平均降雨系列的模拟，确定模型参数，验证模型模拟历史降雨过程的可靠性。然后，建立模型参数从大尺度向站点转换的关系，并从历史降雨系列中抽出某一日雨量系列，假设为未来气候变化情形，对降雨系列在不同尺度间的转换关系进行了验证。在此基础上，对GCMs预测结果的时空解集方法进行了探讨。     

统计降尺度方法对黄河上游流域气象要素模拟分析

将CMIP5模式的输出作为降尺度的输入来预估区域性气候的研究较少,本文使用CMIP5中精度较高的Can ESM2模式下的RCP4.5情景(中等温室气体排放)对黄河上游流域未来气象要素进行预估。利用黄河上游流域(上诠站以上)14个气象站点1967-2010年的逐月降水、气温和NCEP再分析资料,选取拟合度、均值相对误差、标准差相对误差作为评价指标,利用逐步回归算法筛选22个预报因子,建立了月资料序列的统计降尺度模型,并将模型应用于CMIP5中Can ESM2模式下RCP4.5情景,产生了未来气候要素的变化情景。结果表明:该模型对降水的模拟效果好于对气温的模拟。     

Uncertainty Modeling of Statistical Downscaling to Assess Climate Change Impacts on Temperature and Precipitation

Consideration of different Statistical Downscaling (SD) models and multi-sources global climate models’ (GCMs) data can provide a better range of uncertainty for climatic and statistical indices. In this study, results of two SD models, ASD (Automated Statistical Downscaling) and SDSM (Statistical Downscaling Model), were used for uncertainty analysis of temperature and precipitation prediction under climate change impacts for two meteorological stations in Iran. Uncertainty analysis was performed based on application of two GCMs and climate scenarios (A2, A1B, A2a and B2a) for 2011–2040, 2041–2070 and 2071–2100 future time slices. A new technique based on fuzzy logic was proposed and only used to describe uncertainties associated with downscaling methods in temperature and precipitation predictions. In this technique, different membership functions were defined to fuzzify results. Based on these functions width, precipitation had higher uncertainty in comparison with the temperature which could be attributed to the complexity of temporal and local distribution of rainfall. Moreover, little width of membership functions for temperatures in both stations indicated less uncertainty in cold months, whereas the results showed more uncertainty for summer. The results of this study highlight the significance of incorporating uncertainty associated with two downscaling approaches and outputs of GCMs (CGCM3 and HadCM3) under emission scenarios A2, A1B, A2a and B2a in hydrologic modeling and future predictions.     

Improving Hydrological Process Modeling Using Optimized Threshold-Based Wavelet De-Noising Technique

The extent of the noise on hydrological data is inevitable, which reduces the efficiency of Data-Driven Models (DDMs). Despite of this fact that the DDMs such as Artificial Neural Network (ANN) are capable of nonlinear functional mapping between a set of input and output variables, but refining of the time series through data pre-processing methods can provide with the possibility to increase the performance of these set of models. The main objective of this study is to propose a new method called Optimized Threshold-Based Wavelet De-noising technique (OTWD) to de-noise hydrological time series and improve the prediction accuracy while the DDM is being used. For this purpose, in the first step, Wavelet-ANN (WNN) model was developed for identifying suitable wavelet function and maximum decomposition level. Afterward, sub-signals of original precipitation time series which were determined in the first step were de-noised by using of OTWD technique. Therefore, these clean sub-signals of precipitation time series were imposed as input data to the ANN to predict the precipitation one time step ahead. The results showed that OTWD technique could improve the efficiency of WNN model dramatically; this outcome was reported by the different efficiency criterions such as Nash-Sutcliffe Efficiency (NSE?=?0.92), Root Mean Squared Error (RMSE?=?0.0103), coefficient correlation of linear regression (R?=?0.93), Peak Value Criterion (PVC?=?0.021) and Low Value Criterion (LVC?=?0.026). The best fitted WNN model in comparison by proposed model showed weaker performance by the NSE, RMSE, R, PVC and LVC values of 0.86, 0.043, 0.87, 0.034 and 0.045, respectively.     

Ensembling Downscaling Techniques and Multiple GCMs to Improve Climate Change Predictions in Cryosphere Scarcely-Gauged Catchment

Future projections of climate variables are the key for the development of mitigation and adaptation strategy to changing climate. However, such projections are often subjected to large uncertainties which make implementation of climate change strategies on water resources system a challenging job. Major uncertainty sources are General Circulation models (GCMs), post-processing and climate heterogeneity based on catchment characteristics (e.g. scares data and high-altitude). Here we presents the comparisons between different GCMs, statistical downscaling and bias correction approaches and finally climate projections, with the integration of gridded and converted (monthly to daily) data for a high-altitude, scarcely-gauged Jhelum River basin, Pakistan. Current study relies on climate projections obtained from factorial combination of 5-GCMs, 2 statistical downscaling and 2 bias correction methods. In addition, we applied bias corrected APHRODITE, converted daily data using MODAWEC model and observed data. Further, five GCMs (CGCM3, HadCM3, CCSM3, ECHAM5 and CSIRO-MK3.5) were tested to scrutinize two suitable GCMs integrated with Statistical Downscaling Model (SDSM) and Smooth Support Vector Machine (SSVM). Results illustrate that the CGCM3 and HadCM3 were suitable GCMs for selected study basin. Both downscaling techniques are able to simulate precipitation, however, SSVM performed slightly better than SDSM. We found that the integration of CGCM3 with SSVM (SSVM-CGCM3) generates precipitation and temperature better than the CGCM3 (SDSM-CGCM3) and HadCM3 (SDSM-HadCM3) with SDSM. Furthermore, the low elevation stations were influenced by monsoon, significantly prone to rise in precipitation and temperature, while high-altitude stations were influenced by westerlies circulations, less prone to climate change. The projections indicated rise in basin-wide annual precipitation by 25.51, 36.76 and 45.52 mm and temperature by 0.64, 1.47 and 2.79 °C, during 2030s, 2060s and 2090s, respectively. The methods and results of this study can be adopted to evaluate climate change implications in the catchments of characteristics similar to Jhelum River basin.     

统计降尺度方法研究进展综述

统计降尺度方法是将大气环流模式GCMs输出的低分辨率的气象资料转换为流域尺度的主要方法之一,现已发展成为气候学中较为完善的领域。简要介绍了统计降尺度方法的基本原理,包括基本假设条件及主要步骤和关键点;重点介绍统计降尺度方法,大致分为转换函数法、天气分析技术和天气发生器这三类,并对几种方法的国内外应用进展做了阐述;对统计降尺度方法的不确定性研究做了简要介绍。指出未来研究应重点研究统计降尺度模型的适用条件及范围、提高降水模拟的精度;统计降尺度与动力降尺度两种降尺度结合的方法将是降尺度主要发展方向之一。     

An Uncertainty-Based Regional Comparative Analysis on the Performance of Different Bias Correction Methods in Statistical Downscaling of Precipitation

Statistical downscaling of General Circulation Models (GCM) simulations is widely used for projecting precipitation at different spatiotemporal scales. However, the downscaling process is linked with different source of uncertainty including structural/parametric uncertainty of the model and output uncertainty. This research proposes a novel framework to assess the parametric uncertainty of downscaling model, and used this framework to assess the performance of different bias correction methods linked to the regression-based statistical downscaling model. The used downscaling framework in the current paper is Statistical Downscaling Model (SDSM). The conventional bias correction method linked with SDSM is the Variance InFlation method (VIF), this paper substitutes this method with three different bias correction methods including Local Intensity Scaling (LOCI), Power Transformation (PT), and Quantile Mapping (QM) to assess the associated parametric and global uncertainty of each method in different climate by using a new approach. The proposed method is applied to six different stations located in Iran and United States with different climate status. Average Relative Interval Length (ARIL), P-level, and Normalized Uncertainty Efficiency (NUE) are used as uncertainty indicators to evaluate the results. Results represent that in every assessed climate class, LOCI, and PT, work better than conventional VIF in both amount and occurrence modules of SDSM framework. More precisely, LOCI works better in station that has wet summer, while PT performs well in the stations where there is no or very limited precipitation in summer. Substituting LOCI with VIF, result in increasing the value of NUE by at least factor of 3 in occurrence and amount model which means the significant reduction in structural uncertainty. Also applying PT in arid regions improves the NUE indicator at least by factor 2 in occurrence and amount model and by factor 3 in output uncertainty assessment, and results in less parametric and output uncertainty. Results illustrate the important role of bias correction approaches in reducing structural, and output uncertainty and improving the statistical efficiency of the downscaling model.

     

Lake Superior water level fluctuation and climatic factors: A dynamic linear model analysis

We use Dynamic Linear Models (DLM) to analyze the time series of annual average Lake Superior water levels from 1860 to 2007, as well as annual averages of climate drivers including precipitation (1900–2007), evaporation and net precipitation (1951–2007). Our results indicate strong evidence favoring the presence of a systematic trend over a random walk for Lake Superior water levels, and this trend has been negative in recent decades. We then show decisive evidence, in terms of improved predictive performance, favoring a model in which the trend component is replaced with regression components consisting of climatic drivers as predictor variables. Because these models use lagged values of precipitation or net precipitation as predictors, the models can be used to forecast water levels, with the associated uncertainty, several years into the future. We use several of the best fit models and compare one (2008) and two step-ahead (2009) forecasts. The 2008 forecasts compare very well with the observed 2008 water level; the two step-ahead 2009 forecasts are offered as testable hypotheses. The Bayesian context in which these models are developed provides a rigorous framework for data assimilation and regular model updating.     

基于ATAK的MSWEP数据空间降尺度及对降水融合的影响研究

全球性降水数据为获取大范围降水空间分布提供了新途径,但其空间分辨率不高一直是制约其应用于流域或区域尺度上的重要因素之一,因此研究全球性降水数据的空间降尺度方法具有重要的理论和实用价值。本文采用从区域到区域的Kriging（Area to Area Kriging,ATAK）和反距离权重（Inverse Distance Weighted,IDW）两种方法,不考虑地面雨量资料及影响雨量的有关辅助信息,在汉江流域将全球性降水数据MSWEP的空间分辨率由0.1°×0.1°提高至0.02°×0.02°。结果发现ATAK降尺度得到的月雨量场虽然在统计精度上与IDW无明显差异,但提高了对月降水量局部空间变异特征的描述能力,在一定程度上克服了IDW的平滑效应。进一步以ATAK、IDW降尺度处理后的MSWEP数据以及不作空间降尺度处理的原始MSWEP数据为背景场,采用GWR方法分别与雨量站网降水数据融合,发现3种情况下得到的月降水融合数据在空间基本格局上相同,精度统计结果也较为接近,但雨量场的空间连续性及细节特征仍有一定差异。在地表雨量站网密度较高的情况下,背景场差异对MSWEP和站点降水融合结果的影响不能完全消除,甚至在局部可能放大。因此,对于MSWEP等全球性降水数据与站网降水资料的融合而言,选择适当的空间降尺度方法是必要的。本文的结论和认识为全球性降水数据的空间降尺度和雨量场精细化估计提供了重要参考。     

多GCM模型集合下开都河流域未来气候变化预估

基于逐步聚类分析的统计降尺度模型(SCADS模型),在多GCM模型集合的9个大尺度气象变量与开都河流域6个气象变量之间,建立统计降尺度关系,并进行开都河流域未来气候变化的预估。结果表明,SCADS模型生成的开都河流域各气象变量的模拟值与实测值拟合较好。各气象变量在率定期(1961年-1990年)和验证期(1991年-1999年)的NSE系数均大于0.55,精度较高。此外,利用SCADS模型进行开都河流域各气象变量的预估。发现在三个不同时期内(2011年-2040年,2041年-2070年和2071年-2100年),月均气温升高,月均蒸发量、降水量、日照时数增加,月均相对湿度升高。     

A statistical approach to multisite downscaling of daily extreme temperature series: A case study using data in Bangladesh

Downscaling techniques are required to describe the linkages between Global Climate Model outputs at coarse-grid resolutions to surface hydrologic variables at relevant finer scales for climate change impact and adaptation studies. In particular, several statistical methods have been proposed in many previous studies for downscaling of extreme temperature series for a single local site without taking into account the observed spatial dependence of these series between different locations. The present study proposes therefore an improved statistical approach to downscaling of daily maximum (Tmax) and minimum (Tmin) temperature series located at many different sites concurrently. This new multisite multivariate statistical downscaling (MMSD) method was based on a combination of the modeling of the linkages between local daily temperature extremes and global climate predictors by a multiple linear regression model; and the modeling of its stochastic components by the combined singular value decomposition and multivariate autoregressive (SVD-MAR) model to represent more effectively and more accurately the space-time variabilities of these extreme daily temperature series. Results of an illustrative application using daily extreme temperature data from a network of four weather stations in Bangladesh and two different NCEP/NCAR reanalysis datasets have indicated the effectiveness and accuracy of the proposed approach. In particular, this new approach was found to be able to reproduce accurately the basic statistical properties of the Tmax and Tmin at a single site as well as the spatial variability of temperature extremes between different locations. In addition, it has been demonstrated that the proposed method can produce better results than those given by the widely-used single-site downscaling SDSM procedure, especially in preserving the observed inter-site correlations.     

Performance Enhancement of Rainfall Pattern – Water Level Prediction Model Utilizing Self-Organizing-Map Clustering Method

During recent two decades, Artificial Neural Network (ANN) has become one of the most widely used methods in hydrology. One solution for better capturing the existing non-linear and complex nature of data is to develop new hybrid approaches. These hybrid models can be developed in a way that two or more techniques are combined in order to benefit from the advantages of these available approaches and eliminate their limitations. The main scope of this paper is to improve the performance of rainfall-water level modeling by combining ANN with Self Organizing Map (SOM) as an unsupervised clustering method. The proposed method in this study consists of two phases. In the first phase, with the aim of reducing the complexity and dimensionality of input data, a two-step clustering using SOM technique is carried out. Then, in the second phase, separate ANN models are used to model each cluster of data, and final results are obtained by combining the outputs of all models. The proposed new hybrid approach is evaluated using real hydrological data of Johor River. The results of the study indicate that the new proposed SOM-ANN hybrid model has a better performance in daily rainfall-water level forecasting compared to ANN model alone.