High-Frequency and Low-Frequency Variability in Stochastic Daily Weather Generator and Its Effect on Agricultural and Hydrologic Modelling

The high-frequency and low-frequency variabilities, which are often misreproduced by the daily weather generators, have a significant effect on modelling weather-dependent processes. Three modifications are suggested to improve the reproduction of the both variabilities in a four-variate daily weather generator Met&Roll: (i) inclusion of the annual cycle of lag-0 and lag-1 correlations among solar radiation, maximum temperature and minimum temperature, (ii) use of the 3rd order Markov chain to model precipitation occurrence, (iii) applying the monthly generator (based on a first-order autoregressive model) to fit the low-frequency variability. The tests are made to examine the effects of the three new features on (i) a stochastic structure of the synthetic series, and on (ii) outputs from CERES-Wheat crop model (crop yields) and SAC-SMA rainfall-runoff model (monthly streamflow characteristics, distribution of 5-day streamflow) fed by the synthetic weather series. The results are compared with those obtained with the observed weather series.Results: (i) The inclusion of the annual cycle of the correlations has rather ambiguous effect on the temporal structure of the weather characteristics simulated by the generator and only insignificant effect on the output from either simulation model. (ii) Increased order of the Markov chain improves modelling of precipitation occurrence series (especially long dry spells), and correspondingly improves reliability of the output from either simulation model. (iii) Conditioning the daily generator on monthly generator has the most positive effect, especially on the output from the hydrological model: Variability of the monthly streamflow characteristics and the frequency of extreme streamflows are better simulated. (iv) Of the two simulation models, the improvements related to the three modifications are more pronounced in the hydrological simulations. This may be also due to the fact that the crop growth simulations were less affected by the imperfections of the unmodified version of Met&Roll.     

Local-scale climate scenarios for impact studies and risk assessments: integration of early 21st century ENSEMBLES projections into the ELPIS database

We present the integration of early 21st century climate projections for Europe based on simulations carried out within the EU-FP6 ENSEMBLES project with the LARS-WG stochastic weather generator. The aim was to upgrade ELPIS, a repository of local-scale climate scenarios for use in impact studies and risk assessments that already included global projections from the CMIP3 ensemble and regional scenarios for Japan. To obtain a more reliable simulation of daily rainfall and extremes, changes in wet and dry series derived from daily ENSEMBLES outputs were taken into account. Kernel average smoothers were used to reduce noise arising from sampling artefacts. Examples of risk analyses based on 25-km climate projections from the ENSEMBLES ensemble of regional climate models illustrate the possibilities offered by the updated version of ELPIS. The results stress the importance of tailored information for local-scale impact assessments at the European level.     

USE OF A STOCHASTIC WEATHER GENERATOR IN THE DEVELOPMENT OF CLIMATE CHANGE SCENARIOS

Climate change scenarios with a high spatial and temporal resolution are required in the evaluation of the effects of climate change on agricultural potential and agricultural risk. Such scenarios should reproduce changes in mean weather characteristics as well as incorporate the changes in climate variability indicated by the global climate model (GCM) used. Recent work on the sensitivity of crop models and climatic extremes has clearly demonstrated that changes in variability can have more profound effects on crop yield and on the probability of extreme weather events than simple changes in the mean values. The construction of climate change scenarios based on spatial regression downscaling and on the use of a local stochastic weather generator is described. Regression downscaling translated the coarse resolution GCM grid-box predictions of climate change to site-specific values. These values were then used to perturb the parameters of the stochastic weather generator in order to simulate site-specific daily weather data. This approach permits the incorporation of changes in the mean and variability of climate in a consistent and computationally inexpensive way. The stochastic weather generator used in this study, LARS-WG, has been validated across Europe and has been shown to perform well in the simulation of different weather statistics, including those climatic extremes relevant to agriculture. The importance of downscaling and the incorporation of climate variability are demonstrated at two European sites where climate change scenarios were constructed using the UK Met. Office high resolution GCM equilibrium and transient experiments.     

在气候影响研究中引入随机天气发生器的方法和不确定性

通过采用不同的随机天气发生器生成一定气候背景下各种气候变率情景，许多学者在最近的研究中已经认识到气候变率对农作物生长发育影响的重要性。传统的气候影响评估方法直接以大气环流模式的模拟试验结果作为未来气候情景，这样不可能理解如上的重要性。本文着重评述将随机天气发生器应用于气候变化影响研究的一般方法框架，以及作者的具体个例研究方法。文中最后分析了目前该领域研究中还存在的一些不确定性。 在当前的气候变化影响研究中，有不同的方法用来研制一种称为WGEN的典型随机天气发生器的参数化方案及其随机试验方法。不同的研究者也有不同的参数调控方法。通常的思路是通过气候控制试验和2×CO2试验之间的气候变量平均值和方差的变化来扰动随机天气发生器的参数，以生成未来逐日气候变化情景。本文作者根据短期气候预测模式的输出产品建立了一套WGEN的参数化方案及其随机试验方法，并且在时间和空间两个尺度上检验和评估了此参数化方案下WGEN的模拟能力。另外，作者由未来降水的变化，调试随机天气发生器参数，生成了气候变率变化情景。这些参数调节可以产生各种不同类型和定性大小的气候变率变化，用于气候影响评估的敏感性分析。通过如上方法，作为一个个例，文中评估了未来气候变率变化     

MV-WG: a new multi-variable weather generator

This paper presents a new stochastic multi-variable weather generator (MV-WG) and compares its performance with LARS-WG version 4.0. Daily data of 109 meteorological stations from a North American database were used in a twofold comparison of the two generators: (1) the capability of reproducing the mean and variance of annual, seasonal and monthly values, and (2) the capability of reproducing extreme weather events were compared. Both generators did very well on imitating the mean and the variance of the monthly values of the investigated variables, but both showed a more moderate performance as far as the generation of extreme events was concerned. The three-parameter Weibull function, which is first introduced in MV-WG, was found to be a powerful tool to describe not only the distribution of the daily precipitation amounts, but also the distribution of dry and wet spell lengths, as well.     

Simulating Climate over Western North America Using Stochastic Weather Generators

The climatologies of daily precipitation and of maximum and minimum temperatures over western North America are simulated using stochastic weather generators. Two types of generator, differentiated only by their method of modeling precipitation occurrence, are investigated. A second-order Markov model, in which the probability of the occurrence of precipitation is modeled as contingent upon its occurrence on the previous two days, is compared with a spell-length model, in which mass functions of wet- and dry-spell lengths are modeled. Both models are able to reproduce the observed annual and monthly climatology in the region to a high degree of accuracy. However, there is considerable over-dispersion in annual precipitation, resulting primarily from an underestimation in the interannual variability of precipitation intensity. The interannual variability of temperatures is similarly underestimated, and is most severe for minimum temperatures. There is a severe problem in estimating minimum temperature extremes, which can be attributed to the negatively skewed distribution of daily minimum temperatures. Non-normality in the distribution of daily temperatures is shown to be a problem in simulating extreme temperature maxima as well as of minima. It is suggested that the normal distribution used in the generation of daily temperatures in the widely used Richardson (1981) generator, and its derivations, be supplanted by a more appropriate distribution that permits skewness in either direction.     

干湿持续期随机模拟

该文应用数据建模技术, 实现干湿期随机建模。主要包括:利用历史气象资料, 从中采集干湿期数据; 应用实测数据, 创建干湿期经验分布函数; 应用Monte Carlo方法和经验分布参数, 随机生成干湿期序列, 通过和Markov链模型输出的对比分析, 讨论生成序列的统计误差, 测试结果显示, 与两状态Markov链方法相比, 所建模型性能更好。     

Development of climate change projections for small watersheds using multi-model ensemble simulation and stochastic weather generation

Regional climate models (RCMs) have been increasingly used for climate change studies at the watershed scale. However, their performance is strongly dependent upon their driving conditions, internal parameterizations and domain configurations. Also, the spatial resolution of RCMs often exceeds the scales of small watersheds. This study developed a two-step downscaling method to generate climate change projections for small watersheds through combining a weighted multi-RCM ensemble and a stochastic weather generator. The ensemble was built on a set of five model performance metrics and generated regional patterns of climate change as monthly shift terms. The stochastic weather generator then incorporated these shift terms into observed climate normals and produced synthetic future weather series at the watershed scale. This method was applied to the Assiniboia area in southern Saskatchewan, Canada. The ensemble led to reduced biases in temperature and precipitation projections through properly emphasizing models with good performance. Projection of precipitation occurrence was particularly improved through introducing a weight-based probability threshold. The ensemble-derived climate change scenario was well reproduced as local daily weather series by the stochastic weather generator. The proposed combination of dynamical downscaling and statistical downscaling can improve the reliability and resolution of future climate projection for small prairie watersheds. It is also an efficient solution to produce alternative series of daily weather conditions that are important inputs for examining watershed responses to climate change and associated uncertainties.     

Sensitivity of Ceres-Maize Yields to Statistical Structure of Daily Weather Series

To study impacts of climate variations on cropproduction, the growth models are used to simulateyields in present vs. changed climate conditions.Met&Roll is a four-variate (precipitation amount,solar radiation, minimum and maximum temperatures) stochasticweather generator used to supply synthetic dailyweather series for the crop growth model CERES-Maize.Three groups of experiments were conducted in thisstudy: (1) Validation of Met&Roll reveals some discrepanciesin the statistical structure of synthetic weatherseries, e.g., (i) the frequency of occurrence of longdry spells, extreme values of daily precipitationamount and variability of monthly means areunderestimated by the generator; (ii) correlations andlag-1 correlations among weather characteristicsexhibit a significant annual cycle not assumed by themodel. On the whole, the best fit of the observed andsynthetic weather series is experienced in summermonths. (2) The Wilcoxon test was employed to comparedistributions of maize yields simulated with use ofobserved vs. synthetic weather series. As nostatistically significant differences were detected,it is assumed that the generator imperfections inreproducing the statistical structure of weatherseries negligibly affect the model yields. (3) Thesensitivity of model yields to selectedcharacteristics of the daily weather series wasexamined. Emphasis was placed on the characteristicsnot addressed by typical GCM-based climate changescenarios: daily amplitude of temperature, persistenceof the weather series, shape of the distribution ofdaily precipitation amount, and frequency ofoccurrence of wet days. The results indicate that someof these characteristics may significantly affect cropyields and should therefore be considered in thedevelopment of climate change scenarios.     

Postprocessing of simulated precipitation for impact research in West Africa. Part II: A weather generator for daily data

Data from global and regional climate models refer to grid cells and, hence, are basically different from station data. This particularly holds for variables with enhanced spatio-temporal variability like precipitation. On the other hand, many applications like for instance hydrological models require atmospheric data with the statistical characteristics of station data. Here, we present a dynamical-statistical tool to construct virtual station data based on regional climate model output for tropical West Africa. This weather generator (WEGE) incorporates daily gridded rainfall from the model, an orographic term and a stochastic term, accounting for the chaotic spatial distribution of local rain events within a model grid box. In addition, the simulated probability density function of daily precipitation is adjusted to available station data in Benin. It is also assured that the generated data are still consistent with other model parameters like cloudiness and atmospheric circulation. The resulting virtual station data are in excellent agreement with various observed characteristics which are not explicitly addressed by the WEGE algorithm. This holds for the mean daily rainfall intensity and variability, the relative number of rainless days and the scaling of precipitation in time. The data set has already been used successfully for various climate impact studies in Benin.     

Postprocessing of simulated precipitation for impact research in West Africa. Part I: model output statistics for monthly data

Rainfall represents an important factor in agriculture and food security, particularly, in the low latitudes. Climatological and hydrological studies which attempt to diagnose the hydrological cycle, require high-quality precipitation data. In West Africa, like in many parts of the world, the density of observational data is low and climate models are needed in order to perform homogeneous and complete data sets. However, climate models tend to produce systematic errors, especially, in terms of rainfall and cloud processes, which are usually approximated by physical parameterizations. In this study, a 25-year climatology of monthly precipitation in West Africa is presented, derived from a regional climate model simulation, and evaluated with respect to observational data. It is found that the model systematically underestimates the rainfall amount and variability and does not capture some details of the seasonal cycle in sub-Saharan West Africa. Thus, in its present form the precipitation climatology is not appropriate to draw a realistic picture of the hydrological cycle in West Africa nor to serve as input data for impact research. Therefore, a statistical model is developed in order to adjust the simulated rainfall data to the characteristics of observed precipitation. Assuming that the regional climate model is much more reliable in terms of atmospheric circulation and thermodynamics, model output statistics is used to correct simulated rainfall by means of other simulated parameters of the near-surface climate like temperature, sea level pressure and wind components. Monthly data is adjusted by a cross-validated multiple regression model. The resulting adjusted rainfall climatology reveals a substantial improvement in terms of the model deficiencies mentioned above. In part II of this publication, the characteristics of simulated daily precipitation is adapted to station data by applying a weather generator. Once the postprocessing approach is trained, it can be extrapolated to simulation periods, for which observational data do not exist like for instance future climate.     

随机天气模型参数化方案的研究及其模拟能力评估

文中介绍了随机天气模型 WGEN的基本结构及其模拟原理 ,并针对其中随机过程的统计结构特征和 GCMs输出要素的不同时空尺度特点 ,利用动态数据的参数化分析方法等统计学技术 ,确定了该模型参数的估计方法。同时基于蒙特卡罗数值计算原理 ,给出了 WGEN的随机试验方法 ,并通过模拟基准气候 ,从时间分布和空间场两方面对模型在中国东北地区的模拟效果及其能力进行了评估。结果表明 ,模型对于最高气温、最低气温、降水和辐射等要素均具有较好的模拟效果 ,模拟序列与观测序列的取值分布有较一致的概率特性。由此可以结合 GCMs大尺度网格上输出的月和年要素值 ,通过调控随机过程的参数 ,生成具有不同气候变率的 2× CO2 逐日气候变化情景 ,实现气候预测模式与气候影响模式的嵌套 ,进一步研究气候变率变化的可能影响。     

Statistical performances of various deterministic and stochastic models for rainfall series disaggregation

This paper compares the performance of seven disaggregation models, based on various approaches and/or concepts, for the generation of 10-min time step rainfall series from hourly rainfall series. The so-called constant disaggregation model and a linear model based on the external temporal pattern of rainfall are first considered. The other models are stochastic: the first ones are based on a given probability density function applying for the 10-min rainfall amounts of the hour to disaggregate. This probability density function is either uniform or derived from the external temporal pattern of rainfall. The other stochastic models are scaling models using canonical or microcanonical multiplicative random cascades. The comparison of the models is based on their ability to reproduce some important statistical characteristics of the observed time series: variance, skewness coefficient, wet/dry properties of 10-min rainfall amounts; rainfall amounts quantiles for different return periods; autocorrelation of 10-min rainfall amounts. A continuous hydrological simulation is next applied to produce for each generated rainfall series a continuous discharge series used afterwards for a retention design. The ability of the different disaggregation models to produce rainfall time series resulting in the same retention design than the one obtained with the observed rainfall series is finally analysed. Deterministic models as well as simpler stochastic models have rather bad performances when compared to the others. Because it is non-conservative, the model based on a microcanonical random cascade performs also very poorly. It significantly overestimates all studied statistics. Models based on microcanonical random cascades achieve the best performance. They perform reasonably well for the reproduction of rainfall statistics and almost perfectly for the reproduction of runoff and storage design variables. Results finally highlight the interest of including in the disaggregation scheme information related to the external temporal pattern of rainfall.     

A Comparison of Four Precipitation Distribution Models Used in Daily Stochastic Models

Stochastic weather generators are statistical models that produce random numbers that resemble the observed weather data on which they have been fitted; they are widely used in meteorological and hydrologi- cal simulations. For modeling daily precipitation in weather generators, first-order Markov chain-dependent exponential, gamma, mixed-exponential, and lognormal distributions can be used. To examine the perfor- mance of these four distributions for precipitation simulation, they were fitted to observed data collected at 10 stations in the watershed of Yishu River. The parameters of these models were estimated using a maximum-likelihood technique performed using genetic algorithms. Parameters for each calendar month and the Fourier series describing parameters for the whole year were estimated separately. Bayesian infor- mation criterion, simulated monthly mean, maximum daily value, and variance were tested and compared to evaluate the fitness and performance of these models. The results indicate that the lognormal and mixed-exponential distributions give smaller BICs, but their stochastic simulations have overestimation and underestimation respectively, while the gamma and exponential distributions give larger BICs, but their stochastic simulations produced monthly mean precipitation very well. When these distributions were fitted using Fourier series, they all underestimated the above statistics for the months of June, July and August.     

Downscaling daily precipitation time series using a combined circulation- and regression-based approach

The aim of this paper is to introduce a new conditional statistical model for generating daily precipitation time series. The generated daily precipitation can thus be used for climate change impact studies, e.g., crop production, rainfall–runoff, and other water-related processes. It is a stochastic model that links local rainfall events to a continuous atmospheric predictor, moisture flux, in addition to classified atmospheric circulation patterns. The coupled moisture flux is proved to be capable of capturing continuous property of climate system and providing extra information to determine rainfall probability and rainfall amount. The application was made to simultaneously downscale daily precipitation at multiple sites within the Rhine River basin. The results show that the model can well reproduce statistical properties of daily precipitation time series. Especially for extreme rainfall events, the model is thought to better reflect rainfall variability compared to the pure CP-based downscaling approach.     

MACROSCALE HYDROLOGICAL MODELING OVER THE HUAIHE RIVER BASIN

The Xin'anjiang Model is used as the basic model to develop a monthly grid-based macroscalehydrological model for the assessment of the effects of climate change on water resources.Themonthly discharge from 1953 through 1985 in the Huaihe River Basin is simulated.The sensitivityanalysis on runoff is made under assumed climatic scenarios.There is a good agreement betweenthe observed and simulated runoff.Due to the increase of time interval and decrease ofprecipitation intensity on monthly time scale,there is no monthly runoff in some model girds as themomhly hydrological model is applied to the Huaihe River Basin.Two methods of downscalingmonthly precipitation to daily resolution are validated by running the Xin'anjiang model withmonthly data at a daily time step.and the model outputs are more realistic than the monthlyhydrological model.The metbods of downscaling of monthly precipitation to daily resolution mayprovide an idea in solving the problem of the shortage of daily data.In the research of the climatechange on water resources,the daily hydrological model can be used instead of the monthly one.     

Multisite statistical downscaling model for daily precipitation combined by multivariate multiple linear regression and stochastic weather generator

This study provides a multi-site hybrid statistical downscaling procedure combining regression-based and stochastic weather generation approaches for multisite simulation of daily precipitation. In the hybrid model, the multivariate multiple linear regression (MMLR) is employed for simultaneous downscaling of deterministic series of daily precipitation occurrence and amount using large-scale reanalysis predictors over nine different observed stations in southern Québec (Canada). The multivariate normal distribution, the first-order Markov chain model, and the probability distribution mapping technique are employed for reproducing temporal variability and spatial dependency on the multisite observations of precipitation series. The regression-based MMLR model explained 16?%?~?22?% of total variance in daily precipitation occurrence series and 13?%?~?25?% of total variance in daily precipitation amount series of the nine observation sites. Moreover, it constantly over-represented the spatial dependency of daily precipitation occurrence and amount. In generating daily precipitation, the hybrid model showed good temporal reproduction ability for number of wet days, cross-site correlation, and probabilities of consecutive wet days, and maximum 3-days precipitation total amount for all observation sites. However, the reproducing ability of the hybrid model for spatio-temporal variations can be improved, i.e. to further increase the explained variance of the observed precipitation series, as for example by using regional-scale predictors in the MMLR model. However, in all downscaling precipitation results, the hybrid model benefits from the stochastic weather generator procedure with respect to the single use of deterministic component in the MMLR model.     

A comparison of downscaling techniques in the projection of local climate change and wheat yields

This study aims to evaluate the performance of two mainstream downscaling techniques: statistical and dynamical downscaling and to compare the differences in their projection of future climate change and the resultant impact on wheat crop yields for three locations across New South Wales, Australia. Bureau of Meteorology statistically- and CSIRO dynamically-downscaled climate, derived or driven by the CSIRO Mk 3.5 coupled general circulation model, were firstly evaluated against observed climate data for the period 1980–1999. Future climate projections derived from the two downscaling approaches for the period centred on 2055 were then compared. A stochastic weather generator, LARS-WG, was used in this study to derive monthly climate changes and to construct climate change scenarios. The Agricultural Production System sIMulator-Wheat model was then combined with the constructed climate change scenarios to quantify the impact of climate change on wheat grain yield. Statistical results show that (1) in terms of reproducing the past climate, statistical downscaling performed better over dynamical downscaling in most of the cases including climate variables, their mean, variance and distribution, and study locations, (2) there is significant difference between the two downscaling techniques in projected future climate change except the mean value of rainfall across the three locations for most of the months; and (3) there is significant difference in projected wheat grain yields between the two downscaling techniques at two of the three locations.     

Stochastic precipitation generator with hidden state covariates

Time series of daily weather such as precipitation, minimum temperature and maximum temperature are commonly required for various fields. Stochastic weather generators constitute one of the techniques to produce synthetic daily weather. The recently introduced approach for stochastic weather generators is based on generalized linear modeling (GLM) with covariates to account for seasonality and teleconnections (e.g., with the El Niño). In general, stochastic weather generators tend to underestimate the observed interannual variance of seasonally aggregated variables. To reduce this overdispersion, we incorporated time series of seasonal dry/wet indicators in the GLM weather generator as covariates. These seasonal time series were local (or global) decodings obtained by a hidden Markov model of seasonal total precipitation and implemented in the weather generator. The proposed method is applied to time series of daily weather from Seoul, Korea and Pergamino, Argentina. This method provides a straightforward translation of the uncertainty of the seasonal forecast to the corresponding conditional daily weather statistics.     

A comparative study of mixed exponential and Weibull distributions in a stochastic model replicating a tropical rainfall process

A stochastic rainfall model is presented for the generation of hourly rainfall data in an urban area in Malaysia. In view of the high temporal and spatial variability of rainfall within the tropical rain belt, the Spatial–Temporal Neyman–Scott Rectangular Pulse model was used. The model, which is governed by the Neyman–Scott process, employs a reasonable number of parameters to represent the physical attributes of rainfall. A common approach is to attach each attribute to a mathematical distribution. With respect to rain cell intensity, this study proposes the use of a mixed exponential distribution. The performance of the proposed model was compared to a model that employs the Weibull distribution. Hourly and daily rainfall data from four stations in the Damansara River basin in Malaysia were used as input to the models, and simulations of hourly series were performed for an independent site within the basin. The performance of the models was assessed based on how closely the statistical characteristics of the simulated series resembled the statistics of the observed series. The findings obtained based on graphical representation revealed that the statistical characteristics of the simulated series for both models compared reasonably well with the observed series. However, a further assessment using the AIC, BIC and RMSE showed that the proposed model yields better results. The results of this study indicate that for tropical climates, the proposed model, using a mixed exponential distribution, is the best choice for generation of synthetic data for ungauged sites or for sites with insufficient data within the limit of the fitted region.