Intercomparison of multiple statistical downscaling methods: multi-criteria model selection for South Korea

A number of statistical downscaling methodologies have been introduced to bridge the gap in scale between outputs of climate models and climate information needed to assess potential impacts at local and regional scales. Four statistical downscaling methods [bias-correction/spatial disaggregation (BCSD), bias-correction/constructed analogue (BCCA), multivariate adaptive constructed analogs (MACA), and bias-correction/climate imprint (BCCI)] are applied to downscale the latest climate forecast system reanalysis (CFSR) data to stations for precipitation, maximum temperature, and minimum temperature over South Korea. All methods are calibrated with observational station data for 19 years from 1973 to 1991 and validated for the more recent 19-year period from 1992 to 2010. We construct a comprehensive suite of performance metrics to inter-compare methods, which is comprised of five criteria related to time-series, distribution, multi-day persistence, extremes, and spatial structure. Based on the performance metrics, we employ technique for order of preference by similarity to ideal solution (TOPSIS) and apply 10,000 different weighting combinations to the criteria of performance metrics to identify a robust statistical downscaling method and important criteria. The results show that MACA and BCSD have comparable skill in the time-series related criterion and BCSD outperforms other methods in distribution and extremes related criteria. In addition, MACA and BCCA, which incorporate spatial patterns, show higher skill in the multi-day persistence criterion for temperature, while BCSD shows the highest skill for precipitation. For the spatial structure related criterion, BCCA and MACA outperformed BCSD and BCCI. From the TOPSIS analysis, we found that MACA is the most robust method for all variables in South Korea, and BCCA and BCSD are the second for temperature and precipitation, respectively. We also found that the contribution of the multi-day persistence and spatial structure related criteria are crucial to ranking the skill of statistical downscaling methods.     

Let-It-Rain: a web application for stochastic point rainfall generation at ungaged basins and its applicability in runoff and flood modeling

We present a web application named Let-It-Rain that is able to generate a 1-h temporal resolution synthetic rainfall time series using the modified Bartlett–Lewis rectangular pulse (MBLRP) model, a type of Poisson stochastic rainfall generator. Let-It-Rain, which can be accessed through the web address http://www.LetItRain.info, adopts a web-based framework combining ArcGIS Server from server side for parameter value dissemination and JavaScript from client side to implement the MBLRP model. This enables any desktop and mobile end users with internet access and web browser to obtain the synthetic rainfall time series at any given location at which the parameter regionalization work has been completed (currently the contiguous United States and Republic of Korea) with only a few mouse clicks. Let-It-Rain shows satisfactory performance in its ability to reproduce observed rainfall mean, variance, auto-correlation, and probability of zero rainfall at hourly through daily accumulation levels. It also shows a reasonably good performance in reproducing watershed runoff depth and peak flow. We expect that Let-It-Rain can stimulate the uncertainty analysis of hydrologic variables across the world.     

Improving at-site flood frequency analysis with additional spatial information: a probabilistic regional envelope curve approach

Extreme flood events have detrimental effects on society, the economy and the environment. Widespread flooding across South East Queensland in 2011 and 2013 resulted in the loss of lives and significant cost to the economy. In this region, flood risk planning and the use of traditional flood frequency analysis (FFA) to estimate both the magnitude and frequency of the 1-in-100 year flood is severely limited by short gauging station records. On average, these records are 42 years in Eastern Australia and many have a poor representation of extreme flood events. The major aim of this study is to test the application of an alternative method to estimate flood frequency in the form of the Probabilistic Regional Envelope Curve (PREC) approach which integrates additional spatial information of extreme flood events. In order to better define and constrain a working definition of an extreme flood, an Australian Envelope Curve is also produced from available gauging station data. Results indicate that the PREC method shows significant changes to the larger recurrence intervals (≥100 years) in gauges with either too few, or too many, extreme flood events. A decision making process is provided to ascertain when this method is preferable for FFA.     

An assessment of using subsampling method in selection of a flood frequency distribution

In flood frequency analysis, a suitable probability distribution function is required in order to establish the flood magnitude-return period relationship. Goodness of fit (GOF) techniques are often employed to select a suitable distribution function in this context. But they have been often criticized for their inability to discriminate between statistical distributions for the same application. This paper investigates the potential utility of subsampling, a resampling technique with the aid of a GOF test to select the best distribution for frequency analysis. The performance of the methodology is assessed by applying the methodology to observed and simulated annual maximum (AM) discharge data series. Several AM series of different record lengths are used as case studies to determine the performance. Numerical analyses are carried out to assess the performance in terms of sample size, subsample size and statistical properties inherent in the AM data series. The proposed methodology is also compared with the standard Anderson–Darling (AD) test. It is found that the methodology is suitable for a longer data series. A better performance is obtained when the subsample size is taken around half of the underlying data sample. The methodology has also outperformed the standard AD test in terms of effectively discriminating between distributions. Overall, all results point that the subsampling technique can be a promising tool in discriminating between distributions.     

A stochastic methodology for risk assessment of a large earthquake when a long time has elapsed

We propose a stochastic methodology for risk assessment of a large earthquake when a long time has elapsed from the last large seismic event. We state an approximate probability distribution for the occurrence time of the next large earthquake, by knowing that the last large seismic event occurred a long time ago. We prove that, under reasonable conditions, such a distribution is exponential with a rate depending on the asymptotic slope of the cumulative intensity function corresponding to a nonhomogeneous Poisson process. As it is not possible to obtain an empirical cumulative distribution function of the waiting time for the next large earthquake, an estimator of its cumulative distribution function based on existing data is derived. We conduct a simulation study for detecting scenario in which the proposed methodology would perform well. Finally, a real-world data analysis is carried out to illustrate its potential applications, including a homogeneity test for the times between earthquakes.     

Frequency based imputation of precipitation

Changing climate and precipitation patterns make the estimation of precipitation, which exhibits two-dimensional and sometimes chaotic behavior, more challenging. In recent decades, numerous data-driven methods have been developed and applied to estimate precipitation; however, these methods suffer from the use of one-dimensional approaches, lack generality, require the use of neighboring stations and have low sensitivity. This paper aims to implement the first generally applicable, highly sensitive two-dimensional data-driven model of precipitation. This model, named frequency based imputation (FBI), relies on non-continuous monthly precipitation time series data. It requires no determination of input parameters and no data preprocessing, and it provides multiple estimations (from the most to the least probable) of each missing data unit utilizing the series itself. A total of 34,330 monthly total precipitation observations from 70 stations in 21 basins within Turkey were used to assess the success of the method by removing and estimating observation series in annual increments. Comparisons with the expectation maximization and multiple linear regression models illustrate that the FBI method is superior in its estimation of monthly precipitation. This paper also provides a link to the software code for the FBI method.     

Assessment of earthquake hazard for the northwestern Vietnam from geological and geophysical data using an original program package

The highest seismic activity in Vietnam is observed in the northwest of the country, hence the practical significance of more accurate assessment of the earthquake hazard for the area. The worldwide experience of seismicity, in particular, the recent Tohoku mega-earthquake (March 11, 2011, M _w = 9.0, Japan) shows that instrumental and historical data alone are insufficient to reliably estimate earthquake hazard. This is all the more relevant in relation to Vietnam where the period of instrumental observation is short and historical evidence is nearly lacking. In this connection we made an attempt to construct maps of earthquake hazard based on known seismicity data using the available geological and geophysical data and the method of G.I. Reisner and his associates for classification of areas by seismic potential. Since the question of what geological and geophysical parameters are to be used and with what weights remains unresolved, we developed a program package to estimate Mmax based on different options in the use of geological and geophysical data. In this paper we discuss the first results and the promise held by this program package.     

Quantifying differences between reservoir inflows and dam site floods using frequency and risk analysis methods

Reservoirs are the most important constructions for water resources management and flood control. Great concern has been paid to the effects of reservoir on downstream area and the differences between inflows and dam site floods due to the changes of upstream flow generation and concentration conditions after reservoir’s impoundment. These differences result in inconsistency between inflow quantiles and the reservoir design criteria derived by dam site flood series, which can be a potential risk and must be quantificationally evaluated. In this study, flood frequency analysis (FFA) and flood control risk analysis (FCRA) methods are used with the long reservoir inflow series derived from a multiple inputs and single output model and a copula-based inflow estimation model. The results of FFA and FCRA are compared and the influences on reservoir flood management are also discussed. The Three Gorges Reservoir (TGR) in China is selected as a case study. Results show that the differences between the TGR inflow and dam site floods are significant which result in changes on its flood control risk rates. The mean values of TGR’s annual maximum inflow peak discharge and 3 days flood volume have increased 5.58 and 3.85% than the dam site ones, while declined by 1.82 and 1.72% for the annual maximum 7 and 15 days flood volumes. The flood control risk rates of middle and small flood events are increased while extreme flood events are declined. It is shown that the TGR can satisfy the flood control task under current hydrologic regime and the results can offer references for better management of the TGR.     

Simulating rainfall time-series: how to account for statistical variability at multiple scales?

Daily rainfall is a complex signal exhibiting alternation of dry and wet states, seasonal fluctuations and an irregular behavior at multiple scales that cannot be preserved by stationary stochastic simulation models. In this paper, we try to investigate some of the strategies devoted to preserve these features by comparing two recent algorithms for stochastic rainfall simulation: the first one is the modified Markov model, belonging to the family of Markov-chain based techniques, which introduces non-stationarity in the chain parameters to preserve the long-term behavior of rainfall. The second technique is direct sampling, based on multiple-point statistics, which aims at simulating a complex statistical structure by reproducing the same data patterns found in a training data set. The two techniques are compared by first simulating a synthetic daily rainfall time-series showing a highly irregular alternation of two regimes and then a real rainfall data set. This comparison allows analyzing the efficiency of different elements characterizing the two techniques, such as the application of a variable time dependence, the adaptive kernel smoothing or the use of low-frequency rainfall covariates. The results suggest, under different data availability scenarios, which of these elements are more appropriate to represent the rainfall amount probability distribution at different scales, the annual seasonality, the dry-wet temporal pattern, and the persistence of the rainfall events.