Uncertainty,entropy, scaling and hydrological stochastics. 1. Marginal distributional properties of hydrological processes and state scaling / Incertitude,entropie, effet d'échelle et propriétés stochastiques hydrologiques. 1. Propriétés distributionnelles marginales des processus hydrologiques et échelle d'état

Abstract

The well-established physical and mathematical principle of maximum entropy (ME), is used to explain the distributional and autocorrelation properties of hydrological processes, including the scaling behaviour both in state and in time. In this context, maximum entropy is interpreted as maximum uncertainty. The conditions used for the maximization of entropy are as simple as possible, i.e. that hydrological processes are non-negative with specified coefficients of variation (CV) and lag one autocorrelation. In this first part of the study, the marginal distributional properties of hydrological variables and the state scaling behaviour are investigated. Application of the ME principle under these very simple conditions results in the truncated normal distribution for small values of CV and in a nonexponential type (Pareto) distribution for high values of CV. In addition, the normal and the exponential distributions appear as limiting cases of these two distributions. Testing of these theoretical results with numerous hydrological data sets on several scales validates the applicability of the ME principle, thus emphasizing the dominance of uncertainty in hydrological processes. Both theoretical and empirical results show that the state scaling is only an approximation for the high return periods, which is merely valid when processes have high variation on small time scales. In other cases the normal distributional behaviour, which does not have state scaling properties, is a more appropriate approximation. Interestingly however, as discussed in the second part of the study, the normal distribution combined with positive autocorrelation of a process, results in time scaling behaviour due to the ME principle.     

Applicability of the Nagao–Kadoya bivariate exponential distribution for modeling two correlated exponentially distributed variates

 The open literature reveals several types of bivariate exponential distributions. Of them only the Nagao–Kadoya distribution (Nagao and Kadoya, 1970, 1971) has a general form with marginals that are standard exponential distributions and the correlation coefficient being 0≤ρ<1. On the basis of the principle that if a theoretical probability distribution can represent statistical properties of sample data, then the computed probabilities from the theoretical model should provide a good fit to observed ones, numerical experiments are executed to investigate the applicability of the Nagao–Kadoya bivariate exponential distribution for modeling the joint distribution of two correlated random variables with exponential marginals. Results indicate that this model is suitable for analyzing the joint distribution of two exponentially distributed variables. The procedure for the use of this model to represent the joint statistical properties of two correlated exponentially distributed variables is also presented.     

Spatial prediction using bivariate exponential distribution

In this paper, a certain bivariate exponential distribution is used for the spatial prediction. The unobserved random variable is predicted by the projection onto the space of all linear combinations of the powers, up to degree m, of the observed random variables plus the constant 1. We obtain a solution by assuming that all the bivariate distributions follow Gumbel’s type III or logistic form of bivariate exponential. The method is implemented on two data sets and the results are presented. The predictions are compared with the original values through Mean Structural Similarity (MSSIM) index of Wang et al. (IEEE Trans Image Process 13(4):600–612, 2004). Using the MSSIM index the proposed method is also compared with Ordinary Kriging and with Simple Kriging after normal score transform.     

Uncertainty,entropy, scaling and hydrological stochastics. 2. Time dependence of hydrological processes and time scaling / Incertitude,entropie, effet d'échelle et propriétés stochastiques hydrologiques. 2. Dépendance temporelle des processus hydrologiques et échelle temporelle

Abstract

The well-established physical and mathematical principle of maximum entropy (ME), is used to explain the distributional and autocorrelation properties of hydrological processes, including the scaling behaviour both in state and in time. In this context, maximum entropy is interpreted as maximum uncertainty. The conditions used for the maximization of entropy are as simple as possible, i.e. that hydrological processes are non-negative with specified coefficients of variation and lag-one autocorrelation. In the first part of the study, the marginal distributional properties of hydrological processes and the state scaling behaviour were investigated. This second part of the study is devoted to joint distributional properties of hydrological processes. Specifically, it investigates the time dependence structure that may result from the ME principle and shows that the time scaling behaviour (or the Hurst phenomenon) may be obtained by this principle under the additional general condition that all time scales are of equal importance for the application of the ME principle. The omnipresence of the time scaling behaviour in numerous long hydrological time series examined in the literature (one of which is used here as an example), validates the applicability of the ME principle, thus emphasizing the dominance of uncertainty in hydrological processes.     

Raindrop size distribution: Fitting performance of common theoretical models

Modelling raindrop size distribution (DSD) is a fundamental issue to connect remote sensing observations with reliable precipitation products for hydrological applications. To date, various standard probability distributions have been proposed to build DSD models. Relevant questions to ask indeed are how often and how good such models fit empirical data, given that the advances in both data availability and technology used to estimate DSDs have allowed many of the deficiencies of early analyses to be mitigated. Therefore, we present a comprehensive follow-up of a previous study on the comparison of statistical fitting of three common DSD models against 2D-Video Distrometer (2DVD) data, which are unique in that the size of individual drops is determined accurately. By maximum likelihood method, we fit models based on lognormal, gamma and Weibull distributions to more than 42.000 1-minute drop-by-drop data taken from the field campaigns of the NASA Ground Validation program of the Global Precipitation Measurement (GPM) mission. In order to check the adequacy between the models and the measured data, we investigate the goodness of fit of each distribution using the Kolmogorov–Smirnov test. Then, we apply a specific model selection technique to evaluate the relative quality of each model. Results show that the gamma distribution has the lowest KS rejection rate, while the Weibull distribution is the most frequently rejected. Ranking for each minute the statistical models that pass the KS test, it can be argued that the probability distributions whose tails are exponentially bounded, i.e. light-tailed distributions, seem to be adequate to model the natural variability of DSDs. However, in line with our previous study, we also found that frequency distributions of empirical DSDs could be heavy‐tailed in a number of cases, which may result in severe uncertainty in estimating statistical moments and bulk variables.     

Return period of bivariate distributed extreme hydrological events

 Extreme hydrological events are inevitable and stochastic in nature. Characterized by multiple properties, the multivariate distribution is a better approach to represent this complex phenomenon than the univariate frequency analysis. However, it requires considerably more data and more sophisticated mathematical analysis. Therefore, a bivariate distribution is the most common method for modeling these extreme events. The return periods for a bivariate distribution can be defined using either separate single random variables or two joint random variables. In the latter case, the return periods can be defined using one random variable equaling or exceeding a certain magnitude and/or another random variable equaling or exceeding another magnitude or the conditional return periods of one random variable given another random variable equaling or exceeding a certain magnitude. In this study, the bivariate extreme value distribution with the Gumbel marginal distributions is used to model extreme flood events characterized by flood volume and flood peak. The proposed methodology is applied to the recorded daily streamflow from Ichu of the Pachang River located in Southern Taiwan. The results show a good agreement between the theoretical models and observed flood data. The author wishes to thank the two anonymous reviewers for their constructive comments that improving the quality of this work.     

On the performance of a new bivariate pseudo Pareto distribution with application to drought data

A new bivariate pseudo Pareto distribution is proposed, and its distributional characteristics are investigated. The parameters of this distribution are estimated by the moment-, the maximum likelihood- and the Bayesian method. Point estimators of the parameters are presented for different sample sizes. Asymptotic confidence intervals are constructed and the parameter modeling the dependency between two variables is checked. The performance of the different estimation methods is investigated by using the bootstrap method. A Markov Chain Monte Carlo simulation is conducted to estimate the Bayesian posterior distribution for different sample sizes. For illustrative purposes, a real set of drought data is investigated.     

Derivation and assessment of a bivariate IDF relationship using paired rainfall intensity–duration data

A procedure is presented for developing a rainfall intensity–duration–frequency (IDF) relationship that is consistent with bivariate normal distribution modeling. The Box–Cox transformation was used to derive the relation and two methods of determining the parameters of this transformation were evaluated. To assess the uncertainty of the parameters, a confidence interval was constructed and verified with the non-parametric bootstrap method. Additionally, the effect of sample size on the bivariate normality assumption was examined. Case studies, based on data from significant gauge stations in Korea, were performed. The result shows that the use of the bivariate normal model as an IDF relationship is particularly recommended when the available data size is small.     

Joint probability distribution of annual maximum storm peaks and amounts as represented by daily rainfalls

Abstract

A procedure is presented for using the bivariate normal distribution to describe the joint distribution of storm peaks (maximum rainfall intensities) and amounts which are mutually correlated. The Box-Cox transformation method is used to normalize original marginal distributions of storm peaks and amounts regardless of the original forms of these distributions. The transformation parameter is estimated using the maximum likelihood method. The joint cumulative distribution function, the conditional cumulative distribution function, and the associated return periods can be readily obtained based on the bivariate normal distribution. The method is tested and validated using two rainfall data sets from two meteorological stations that are located in different climatic regions of Japan. The theoretical distributions show a good fit to observed ones.     

Frequency analysis of droughts using the Plackett copula and parameter estimation by genetic algorithm

This study aims to model the joint probability distribution of drought duration, severity and inter-arrival time using a trivariate Plackett copula. The drought duration and inter-arrival time each follow the Weibull distribution and the drought severity follows the gamma distribution. Parameters of these univariate distributions are estimated using the method of moments (MOM), maximum likelihood method (MLM), probability weighted moments (PWM), and a genetic algorithm (GA); whereas parameters of the bivariate and trivariate Plackett copulas are estimated using the log-pseudolikelihood function method (LPLF) and GA. Streamflow data from three gaging stations, Zhuangtou, Taian and Tianyang, located in the Wei River basin, China, are employed to test the trivariate Plackett copula. The results show that the Plackett copula is capable of yielding bivariate and trivariate probability distributions of correlated drought variables.     

Modelling bivariate extreme precipitation distribution for data‐scarce regions using Gumbel–Hougaard copula with maximum entropy estimation

A new method of parameter estimation in data scarce regions is valuable for bivariate hydrological extreme frequency analysis. This paper proposes a new method of parameter estimation (maximum entropy estimation, MEE) for both Gumbel and Gumbel–Hougaard copula in situations when insufficient data are available. MEE requires only the lower and upper bounds of two hydrological variables. To test our new method, two experiments to model the joint distribution of the maximum daily precipitation at two pairs of stations on the tributaries of Heihe and Jinghe River, respectively, were performed and compared with the method of moments, correlation index estimation, and maximum likelihood estimation, which require a large amount of data. Both experiments show that for the Ye Niugou and Qilian stations, the performance of MEE is nearly identical to those of the conventional methods. For the Xifeng and Huanxian stations, MEE can capture information indicating that the maximum daily precipitation at the Xifeng and Huanxian stations has an upper tail dependence, whereas the results generated by correlation index estimation and maximum likelihood estimation are unreasonable. Moreover, MEE is proved to be generally reliable and robust by many simulations under three different situations. The Gumbel–Hougaard copula with MEE can also be applied to the bivariate frequency analysis of other extreme events in data‐scarce regions.     

BME prediction of continuous geographical properties using auxiliary variables

Using auxiliary information to improve the prediction accuracy of soil properties in a physically meaningful and technically efficient manner has been widely recognized in pedometrics. In this paper, we explored a novel technique to effectively integrate sampling data and auxiliary environmental information, including continuous and categorical variables, within the framework of the Bayesian maximum entropy (BME) theory. Soil samples and observed auxiliary variables were combined to generate probability distributions of the predicted soil variable at unsampled points. These probability distributions served as soft data of the BME theory at the unsampled locations, and, together with the hard data (sample points) were used in spatial BME prediction. To gain practical insight, the proposed approach was implemented in a real-world case study involving a dataset of soil total nitrogen (TN) contents in the Shayang County of the Hubei Province (China). Five terrain indices, soil types, and soil texture were used as auxiliary variables to generate soft data. Spatial distribution of soil total nitrogen was predicted by BME, regression kriging (RK) with auxiliary variables, and ordinary kriging (OK). The results of the prediction techniques were compared in terms of the Pearson correlation coefficient (r), mean error (ME), and root mean squared error (RMSE). These results showed that the BME predictions were less biased and more accurate than those of the kriging techniques. In sum, the present work extended the BME approach to implement certain kinds of auxiliary information in a rigorous and efficient manner. Our findings showed that the BME prediction technique involving the transformation of variables into soft data can improve prediction accuracy considerably, compared to other techniques currently in use, like RK and OK.     

Entropy-based parameter estimation for extended Burr XII distribution

Two entropy-based methods, called ordinary entropy (ENT) method and parameter space expansion method (PSEM), both based on the principle of maximum entropy, are applied for estimating parameters of the extended Burr XII distribution. With the parameters so estimated, the Burr XII distribution is applied to six peak flow datasets and quantiles (discharges) corresponding to different return periods are computed. These two entropy methods are compared with the methods of moments (MOM), probability weighted moments (PWM) and maximum likelihood estimation (MLE). It is shown that PSEM yields the same quantiles as does MLE for discrete cases, while ENT is found comparable to the MOM and PWM. For shorter return periods (<10–30 years), quantiles (discharges) estimated by these four methods are in close agreement, but the differences amongst them grow as the return period increases. The error in quantiles computed using the four methods becomes larger for return periods greater than 10–30 years.     

Assessment of modified Anderson–Darling test statistics for the generalized extreme value and generalized logistic distributions

An important problem in frequency analysis is the selection of an appropriate probability distribution for a given sample data. This selection is generally based on goodness-of-fit tests. The goodness-of-fit method is an effective means of examining how well a sample data agrees with an assumed probability distribution as its population. However, the goodness of fit test based on empirical distribution functions gives equal weight to differences between empirical and theoretical distribution functions corresponding to all observations. To overcome this drawback, the modified Anderson–Darling test was suggested by Ahmad et al. (1988b). In this study, the critical values of the modified Anderson–Darling test statistics are revised using simulation experiments with extensions of the shape parameters for the GEV and GLO distributions, and a power study is performed to test the performance of the modified Anderson–Darling test. The results of the power study show that the modified Anderson–Darling test is more powerful than traditional tests such as the χ², Kolmogorov–Smirnov, and Cramer von Mises tests. In addition, to compare the results of these goodness-of-fit tests, the modified Anderson–Darling test is applied to the annual maximum rainfall data in Korea.     

Application of bivariate frequency analysis to the derivation of rainfall–frequency curves

Bivariate distributions have been recently employed in hydrologic frequency analysis to analyze the joint probabilistic characteristics of multivariate storm events. This study aims to derive practical solutions of application for the bivariate distribution to estimate design rainfalls corresponding to the desired return periods. Using the Gumbel mixed model, this study constructed rainfall–frequency curves at sample stations in Korea which provide joint relationships between amount, duration, and frequency of storm events. Based on comparisons and analyses of the rainfall–frequency curves derived from univariate and bivariate storm frequency analyses, this study found that conditional frequency analysis provides more appropriate estimates of design rainfalls as it more accurately represents the natural relationship between storm properties than the conventional univariate storm frequency analysis.     

Bayesian comparison of different rainfall depth–duration–frequency relationships

Depth–duration–frequency curves estimate the rainfall intensity patterns for various return periods and rainfall durations. An empirical model based on the generalized extreme value distribution is presented for hourly maximum rainfall, and improved by the inclusion of daily maximum rainfall, through the extremal indexes of 24 hourly and daily rainfall data. The model is then divided into two sub-models for the short and long rainfall durations. Three likelihood formulations are proposed to model and compare independence or dependence hypotheses between the different durations. Dependence is modelled using the bivariate extreme logistic distribution. The results are calculated in a Bayesian framework with a Markov Chain Monte Carlo algorithm. The application to a data series from Marseille shows an improvement of the hourly estimations thanks to the combination between hourly and daily data in the model. Moreover, results are significantly different with or without dependence hypotheses: the dependence between 24 and 72 h durations is significant, and the quantile estimates are more severe in the dependence case.     

A new bivariate Gamma distribution generated from functional scale parameter with application to drought data

Univariate and bivariate Gamma distributions are among the most widely used distributions in hydrological statistical modeling and applications. This article presents the construction of a new bivariate Gamma distribution which is generated from the functional scale parameter. The utilization of the proposed bivariate Gamma distribution for drought modeling is described by deriving the exact distribution of the inter-arrival time and the proportion of drought along with their moments, assuming that both the lengths of drought duration (X) and non-drought duration (Y) follow this bivariate Gamma distribution. The model parameters of this distribution are estimated by maximum likelihood method and an objective Bayesian analysis using Jeffreys prior and Markov Chain Monte Carlo method. These methods are applied to a real drought dataset from the State of Colorado, USA.     

An extended Birnbaum–Saunders model and its application in the study of environmental quality in Santiago, Chile

In this article, we introduce, characterize and apply an extended version of the Birnbaum–Saunders model based on the Mudolkar–Hutson skew distribution. This model is appropriated for describing phenomena involving accumulation of some type, as is the case of environmental contamination. Specifically, we find the density, distribution function, and moments of the new model. In addition, we derive several properties and transformations related to this distribution. Furthermore, we propose an estimation method for the parameters of the model. Moreover, we conduct a study of its hazard rate focuses in environmental analysis. A computational implementation in R language of the obtained results is discussed. Finally, we present two examples with real data from environmental quality in Chile that illustrate the proposed methodology.     

The bivariate lognormal distribution to model a multivariate flood episode

Complex hydrological events such as floods always appear to be multivariate events that are characterized by a few correlated variables. A complete understanding of these events needs to investigate joint probabilistic behaviours of these correlated variables. The lognormal distribution is one of frequently selected candidates for flood‐frequency analysis. The multivariate lognormal distribution will serve as an important tool for analysing a multivariate flood episode. This article presents a procedure for using the bivariate lognormal distribution to describe the joint distributions of correlated flood peaks and volumes, and correlated flood volumes and durations. Joint distributions, conditional distributions, and the associated return periods of these random variables can be readily derived from their marginal distributions. The approach is verified using observed streamflow data from the Nord river basin, located in the Province of Quebec, Canada. The theoretical distributions show a good fit to observed ones. Copyright © 2000 John Wiley & Sons, Ltd.     

Ground motion scenarios consistent with probabilistic seismic hazard disaggregation analysis. Application to Mainland Portugal

Probabilistic seismic hazard for Mainland Portugal was re-evaluated in order to perform its disaggregation. Seismic hazard was disaggregated considering different spaces of random variables, namely, univariate conditional hazard distributions of M (magnitude), R (source-to-site distance) and ε (deviation of ground motion to the median value predicted by an attenuation model), bivariate conditional hazard distributions of M–R and X–Y (seismic source latitude and longitude) or multivariate conditional hazard distributions of M–R–ε and M–(X–Y)–ε. The main objective of the present work was achieved, as it was possible, based on the modal values of the above mentioned distributions, to characterize the scenarios that dominate some seismic hazard levels of the 278 Mainland Portuguese counties. In addition, results of 4D disaggregation analysis, in M–(X–Y)–ε, pointed out the existence of one geographic location shared by the dominant scenario of most analyzed counties, especially for hazard levels correspondent to high return periods. Those dominant scenarios are located offshore at a distance of approximately 70 km WSW of S. Vicente cape. On the other hand, the lower the return period the higher is the number of modal scenarios in the neighbourhood of the analyzed site. One may conclude that modal scenarios reproduce hazard target values in each site with great accuracy enabling the applications derived from those scenarios (e.g. loss evaluation) to be associated to a hazard level exceedance probability.