Simulation of daily precipitation in the Pacific Northwest using a weather classification scheme

A weather classification scheme was coupled with a semi-Markov model to represent the coincident occurrence of rain/no rain states at a single rain gauge and classes representing regional atmospheric circulation patterns, as identified from National Meteorological Center gridded observations for a large area of the North Pacific. Weather classes were identified from daily observations of surface pressure and 850 mb pressure height at five selected ten degree latitude by ten degree longitude cells using a K-means clustering algorithm, which was applied on a month-by-month basis. The number of climate classes, K, for each month was chosen based on a preliminary analysis of the model's ability to describe statistics of observed precipitation occurrences at the Stampede Pass, Washington weather station. The length of stay distributions within each precipitation occurrence/weather class were assumed to be geometric, and the precipitation amounts for each class and season were fitted with a mixed exponential distribution. Parameters of the length of stay distributions, transition probabilities, and precipitation amounts were estimated from the period of record 1975–84.The fitted model was used to simulate a ten year sequence of daily precipitation. It was found that the semi-Markov model of climate class/wet-dry states preserved the length of wet and dry day runs reasonably well, with the exception of months with long average run lengths. Likewise, the occurrence frequencies of the climate classes were reasonably well preserved with a few exceptions. An exploratory analysis of the properties of wet and dry period runs for those classes and months whose run frequencies were poorly preserved showed that the log survivor functions and variance time curves were also poorly preserved, which suggests that more complex distributions may be required for some of the run length distributions.     

MOS based forecast of 6-hourly area precipitation

A statistical post-processing methodology for application to numerical weather prediction (NWP) model outputs for precipitation forecast is proposed. The post-processing is based on the model output statistics approach. The statistical relationships are described by the multiple linear regression model, which is complemented by an iteration procedure to further correct the regression outputs. Prognostic fields of the ALADIN/LACE (Aire Limitée Adaptation Dynamique Développement InterNational/Limited Area Modelling in Central Europe) NWP model are used for the forecast of 6-hourly areal precipitation amounts at 15 river basins. The NWP model integration starts at 00UTC and forecasts are calculated for lead times of +12, +18, +24 and +30 hours. The post-processing models are developed separately for each lead time and for separate warm (April to September) and cool (October to March) seasons. The forecasts are focused on large precipitation amounts. Using all the combinations, data from four years (1999–2002) are divided into calibration data (3 years), where the models are developed, and verification data. The models are evaluated by examining the root-mean-square error (RMSE), bias, and correlation coefficient (CC) on the verification data samples. The results show that the additional iteration procedure increases the forecast accuracy for a given range of precipitation amounts and simultaneously does not deteriorate the bias, a situation which can arise when negative regression outputs are set to zero. The post-processing method improves the forecast of the NWP model in terms of RMSE and CC. For large precipitation amounts during the summer season, the decrease of RMSE reaches 10% to 20% depending upon the applied method of verification. For the cool season, the decrease is somewhat smaller (7% to 15%).     

Statistically downscaled probabilistic multi‐model ensemble projections of precipitation change in a watershed

This paper presents the development of a probabilistic multi‐model ensemble of statistically downscaled future projections of precipitation of a watershed in New Zealand. Climate change research based on the point estimates of a single model is considered less reliable for decision making, and multiple realizations of a single model or outputs from multiple models are often preferred for such purposes. Similarly, a probabilistic approach is preferable over deterministic point estimates. In the area of statistical downscaling, no single technique is considered a universal solution. This is due to the fact that each of these techniques has some weaknesses, owing to its basic working principles. Moreover, watershed scale precipitation downscaling is quite challenging and is more prone to uncertainty issues than downscaling of other climatological variables. So, multi‐model statistical downscaling studies based on a probabilistic approach are required. In the current paper, results from the three well‐reputed statistical downscaling methods are used to develop a Bayesian weighted multi‐model ensemble. The three members of the downscaling ensemble of this study belong to the following three broad categories of statistical downscaling methods: (1) multiple linear regression, (2) multiple non‐linear regression, and (3) stochastic weather generator. The results obtained in this study show that the new strategy adopted here is promising because of many advantages it offers, e.g. it combines the outputs of multiple statistical downscaling methods, provides probabilistic downscaled climate change projections and enables the quantification of uncertainty in these projections. This will encourage any future attempts for combining the results of multiple statistical downscaling methods. Copyright © 2011 John Wiley & Sons, Ltd.     

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.     

Extreme value theory for precipitation: sensitivity analysis for climate change

Extreme value theory for the maximum of a time series of daily precipitation amount is described. A chain-dependent process is assumed as a stochastic model for daily precipitation, with the intensity distribution being the gamma. To examine how the effective return period for extreme high precipitation amounts would change as the parameters of the chain-dependent process change (i.e., probability of a wet day, shape and scale parameters of the gamma distribution), a sensitivity analysis is performed. This sensitivity analysis is guided by some results from statistical downscaling that relate patterns in large-scale atmospheric circulation to local precipitation, providing a physically plausible range of changes in the parameters. For the particular location considered in the example, the effective return period is most sensitive to the scale parameter of the intensity distribution.     

Application of adaptive fuzzy rule-based models for reconstruction of missing precipitation events

Abstract

This paper describes a fuzzy rule-based approach applied for reconstruction of missing precipitation events. The working rules are formulated from a set of past observations using an adaptive algorithm. A case study is carried out using the data from three precipitation stations in northern Italy. The study evaluates the performance of this approach compared with an artificial neural network and a traditional statistical approach. The results indicate that, within the parameter sub-space where its rules are trained, the fuzzy rule-based model provided solutions with low mean square error between observations and predictions. The problems that have yet to be addressed are overfitting and applicability outside the range of training data.     

Genetic algorithms for the classification and prediction of precipitation occurrence

Abstract

Using an approach similar to the biological processes of natural selection and evolution, the genetic algorithm (GA) is a nonconventional optimum search technique. Genetic algorithms have the ability to search large and complex decision spaces and handle nonconvexities. In this paper, the GA is applied for solving the optimum classification of rainy and non-rainy day occurrences based on vertical velocity, dewpoint depression, temperature and humidity data. The problem involves finding optimum classification based on known data, training the future prediction system and then making reliable predictions for rainfall occurrences which have significance in agricultural, transportation, water resources and tourism activities. Various statistical approaches require restrictive assumptions such as stationarity, homogeneity and normal probability distribution of the hydrological variables concerned. The GAs do not require any of these assumptions in their applications. The GA approach for the occurrence classifications and predictions is presented in steps and then the application of the methodology is shown for precipitation occurrence (non-occurrence) data. It has been shown that GAs give better results than classical approaches such as discriminant analysis. The application of the methodology is presented independently for the precipitation event occurrences and forecasting at the Lake Van station in eastern Turkey. Finally, the amounts of precipitation are predicted with a model similar to a third order Markov model whose parameters are estimated by the GA technique.     

Spillway design floods in Sweden: I. New guidelines

Abstract

The new Swedish guidelines for the estimation of design floods for dams and spillways are presented, with emphasis on high-hazard dams. The method is based on a set of regional design precipitation sequences, rescaled for basin area, season and elevation above sea level, and a full hydrological model. A reservoir operation strategy is also a fundamental component of the guidelines. The most critical combination of flood generating factors is searched by systematically inserting the design precipitation sequence into a ten year climatological record, where the initial snowpack has been replaced by a statistical 30-year snowpack. The new guidelines are applicable to single reservoir systems as well as more complex hydroelectric schemes, and cover snowmelt floods, rain floods and combinations of the two. In order to study the probabilities of the computed floods and to avoid regional inconsistencies, extensive comparisons with observed floods and frequency analyses have been carried out.     

Approaches for defining thresholds and return periods for rainfall‐triggered shallow landslides

Probabilistic thresholds for triggering shallow landslides by rainfall are developed using two approaches: a logistic regression model and Iverson's physically based model. Both approaches are applied to a 180 km² area in northern Italy. For the physically based model a Monte Carlo approach is used to obtain probabilities of slope failure associated with differing combinations of rainfall intensity and duration as well as differing topographic settings. For the logistic regression model hourly and daily rainfall data and split‐sample testing are used to explore the effect of antecedent rainfall on triggering thresholds. It is demonstrated that both the statistical and physically based models provide stochastic thresholds that express the probability of landslide triggering. The resulting thresholds are comparable, even though the two approaches are conceptually different. The physically based model also provides an estimate of the percentage of potentially unstable areas in which failure can be triggered with a certain probability. The return period of rainfall responsible for landslide triggering is studied by using a Gumbel scaling model of rainfall intensity–duration–frequency curves. It is demonstrated that antecedent rainfall must be taken into account in landslide forecasting, and a method is proposed to correct the rainfall return period by filtering the rainfall maxima with a fixed threshold of antecedent rainfall. This correction produces an increase of the return periods, especially for rainstorms of short duration. Copyright © 2009 John Wiley & Sons, Ltd.     

大别山库区降水预报性能评估及应用对策

对降水预报进行性能评估及应用对策研究可以更好地发挥降水预报在水库调度中的决策支持作用.基于大别山库区近10 a汛期(2007—2016年5月1日—9月30日)24~168 h共7个预见期降水预报和地面降水观测资料,采用正确率、TS评分、概率统计、ROC曲线以及CTS等方法评估大别山库区降水预报性能,并以响洪甸水库为重点研究区域分析降水预报在水库调度中的应用对策.结果表明:1)大别山库区各量级的降水预报都有正预报技巧;24~72 h预见期降水预报的TS评分较高且空报率、漏报率也较低,具有较高的预报性能;但96 h及以上预见期降水预报性能明显下降,中雨以上量级空报率、漏报率较大,特别是对大暴雨及其以上量级的降水预报性能显著下降.2)大别山库区预报降水量级与实况降水量级基本符合,预报降水量级大于等于实况降水量级的概率超过75%;虽然降水预报量级上呈现出过度预报的现象,但降水过程预报对水库调度仍有较好的应用价值,应用时要考虑到降水预报量级可能存在偏差.3)转折性天气预报96 h及以上预见期CTS评分较低,但72 h以内预见期的性能明显改进,尤其是24 h预见期CTS评分也提高到了38.2%;水库调度可从长预见期的降水预报获取降水过程及其可能发生转折的信息,根据短预见期的降水预报进行调度方案调整.     

Evaluation of climate reanalysis and space-borne precipitation products over Bangladesh

ABSTRACT

This study aims to quantify the spatial distribution of errors in two climate reanalysis (ERA5 and CFSR) and two satellite (TMPA-RT and TMPA-V7) precipitation products over Bangladesh. The datasets are assessed against ground-based rain gauge observations to capture the extreme rainfall accumulations at daily temporal scale over a 5-year period (January 2010–December 2014). The bias ratio scores indicate that CFSR and TMPA-RT seriously overestimate the rainfall values over much of the study area. Whilst TMPA-V7 performs better than the other precipitation products, all datasets lose their detection skills substantially for higher quantile thresholds (i.e. above 50th and 75th percentiles). With respect to rainfall detection metrics – probability of detection (POD) and volumetric hit index (VHI) – both ERA5 and CFSR show superior performance (in the range 0.9–1.0 for all the analysis grid boxes). All rainfall datasets are equally good in terms of false alarm ratio (FAR) and volumetric FAR (VFAR), even though the lowest values are associated with ERA5 for higher quantiles. All products demonstrate a decrease in skill to capture the amount of rainfall but show satisfactory results to detect the rainfall events when using higher quantile thresholds (i.e. rainfall above the 50th and 75th percentiles) to sample the data before computing product skill.     

Driving a lumped hydrological model with precipitation output from weather generators of different complexity

ABSTRACT

This paper deals with the question of whether a lumped hydrological model driven with lumped daily precipitation time series from a univariate single-site weather generator can produce equally good results compared to using a multivariate multi-site weather generator, where synthetic precipitation is first generated at multiple sites and subsequently lumped. Three different weather generators were tested: a univariate “Richardson type” model, an adapted univariate Richardson type model with an improved reproduction of the autocorrelation of precipitation amounts and a semi-parametric multi-site weather generator. The three modelling systems were evaluated in two Alpine study areas by comparing the hydrological output with respect to monthly and daily statistics as well as extreme design flows. The application of a univariate Richardson type weather generator to lumped precipitation time series requires additional attention. Established parametric distribution functions for single-site precipitation turned out to be unsuitable for lumped precipitation time series and led to a large bias in the hydrological simulations. Combining a multi-site weather generator with a hydrological model produced the least bias.     

A simulation of probabilistic wildfire risk components for the continental United States

This simulation research was conducted in order to develop a large-fire risk assessment system for the contiguous land area of the United States. The modeling system was applied to each of 134 Fire Planning Units (FPUs) to estimate burn probabilities and fire size distributions. To obtain stable estimates of these quantities, fire ignition and growth was simulated for 10,000 to 50,000 “years” of artificial weather. The fire growth simulations, when run repeatedly with different weather and ignition locations, produce burn probabilities and fire behavior distributions at each landscape location (e.g., number of times a “cell” burns at a given intensity divided by the total years). The artificial weather was generated for each land unit using (1) a fire danger rating index known as the Energy Release Component (ERC) which is a proxy for fuel moisture contents, (2) a time-series analysis of ERC to represent daily and seasonal variability, and (3) distributions of wind speed and direction from weather records. Large fire occurrence was stochastically modeled based on historical relationships to ERC. The simulations also required spatial data on fuel structure and topography which were acquired from the LANDFIRE project (). Fire suppression effects were represented by a statistical model that yields a probability of fire containment based on independent predictors of fire growth rates and fuel type. The simulated burn probabilities were comparable to observed patterns across the U.S. over the range of four orders of magnitude, generally falling within a factor of 3 or 4 of historical estimates. Close agreement between simulated and historical fire size distributions suggest that fire sizes are determined by the joint distributions of spatial opportunities for fire growth (dependent on fuels and ignition location) and the temporal opportunities produced by conducive weather sequences. The research demonstrates a practical approach to using fire simulations at very broad scales for purposes of operational planning and perhaps ecological research.     

Point process-based modeling of multiple debris flow landslides using INLA: an application to the 2009 Messina disaster

We develop a stochastic modeling approach based on spatial point processes of log-Gaussian Cox type for a collection of around 5000 landslide events provoked by a precipitation trigger in Sicily, Italy. Through the embedding into a hierarchical Bayesian estimation framework, we can use the integrated nested Laplace approximation methodology to make inference and obtain the posterior estimates of spatially distributed covariate and random effects. Several mapping units are useful to partition a given study area in landslide prediction studies. These units hierarchically subdivide the geographic space from the highest grid-based resolution to the stronger morphodynamic-oriented slope units. Here we integrate both mapping units into a single hierarchical model, by treating the landslide triggering locations as a random point pattern. This approach diverges fundamentally from the unanimously used presence–absence structure for areal units since we focus on modeling the expected landslide count jointly within the two mapping units. Predicting this landslide intensity provides more detailed and complete information as compared to the classically used susceptibility mapping approach based on relative probabilities. To illustrate the model’s versatility, we compute absolute probability maps of landslide occurrences and check their predictive power over space. While the landslide community typically produces spatial predictive models for landslides only in the sense that covariates are spatially distributed, no actual spatial dependence has been explicitly integrated so far. Our novel approach features a spatial latent effect defined at the slope unit level, allowing us to assess the spatial influence that remains unexplained by the covariates in the model. For rainfall-induced landslides in regions where the raingauge network is not sufficient to capture the spatial distribution of the triggering precipitation event, this latent effect provides valuable imaging support on the unobserved rainfall pattern.     

Spatial interpolation schemes of daily precipitation for hydrologic modeling

Distributed hydrologic models typically require spatial estimates of precipitation interpolated from sparsely located observational points to the specific grid points. We compare and contrast the performance of regression-based statistical methods for the spatial estimation of precipitation in two hydrologically different basins and confirmed that widely used regression-based estimation schemes fail to describe the realistic spatial variability of daily precipitation field. The methods assessed are: (1) inverse distance weighted average; (2) multiple linear regression (MLR); (3) climatological MLR; and (4) locally weighted polynomial regression (LWP). In order to improve the performance of the interpolations, the authors propose a two-step regression technique for effective daily precipitation estimation. In this simple two-step estimation process, precipitation occurrence is first generated via a logistic regression model before estimate the amount of precipitation separately on wet days. This process generated the precipitation occurrence, amount, and spatial correlation effectively. A distributed hydrologic model (PRMS) was used for the impact analysis in daily time step simulation. Multiple simulations suggested noticeable differences between the input alternatives generated by three different interpolation schemes. Differences are shown in overall simulation error against the observations, degree of explained variability, and seasonal volumes. Simulated streamflows also showed different characteristics in mean, maximum, minimum, and peak flows. Given the same parameter optimization technique, LWP input showed least streamflow error in Alapaha basin and CMLR input showed least error (still very close to LWP) in Animas basin. All of the two-step interpolation inputs resulted in lower streamflow error compared to the directly interpolated inputs.     

Scale-recursive assimilation of precipitation data

This paper presents a new statistical method for assimilating precipitation data from different sensors operating over a range of scales. The technique is based on a scale-recursive estimation algorithm which is computationally efficient and able to account for the nested spatial structure of precipitation fields. The version of the algorithm described here relies on a static multiplicative cascade model which relates rainrates at different scales. Bayesian estimation techniques are used to condition rainrate estimates on measurements. The conditioning process is carried out recursively in two sweeps: first from fine to coarse scales and then from coarse to fine scales. The complete estimation algorithm is similar to a fixed interval smoother although it processes data over scale rather than time. We use this algorithm to assimilate radar and satellite microwave data collected during the tropical ocean–global atmosphere coupled ocean–atmosphere response experiment (TOGA-COARE). The resulting rainrate estimates reproduce withheld radar measurements to within the level of accuracy predicted by the assimilation algorithm.     

Assessment of change in design flood frequency under climate change using a multivariate downscaling model and a precipitation-runoff model

Precipitation and runoff are key elements in the hydrologic cycle because of their important roles in water supply, flood prevention, river restoration, and ecosystem management. Global climate change, widely accepted to be happening, is anticipated to have enormous consequences on future hydrologic patterns. Studies on the potential changes in global, regional, and local hydrologic patterns under global climate change scenarios have been an intense area of research in recent years. The present study contributes to this research topic through evaluation of design flood under climate change. The study utilizes a weather state-based, stochastic multivariate model as a conditional probability model for simulating the precipitation field. An important premise of this study is that large-scale climatic patterns serve as a major driver of persistent year-to-year changes in precipitation probabilities. Since uncertainty estimation in the study of climate change is needed to examine the reliability of the outcomes, this study also applies a Bayesian Markov chain Monte Carlo scheme to the widely used SAC-SMA (Sacramento soil moisture accounting) precipitation-runoff model. A case study is also performed with the Soyang Dam watershed in South Korea as the study basin. Finally, a comprehensive discussion on design flood under climate change is made.     

Data assimilation for precipitation nowcasting using Bayesian inference

This work introduces a new variational Bayes data assimilation method for the stochastic estimation of precipitation dynamics using radar observations for short term probabilistic forecasting (nowcasting). A previously developed spatial rainfall model based on the decomposition of the observed precipitation field using a basis function expansion captures the precipitation intensity from radar images as a set of ‘rain cells’. The prior distributions for the basis function parameters are carefully chosen to have a conjugate structure for the precipitation field model to allow a novel variational Bayes method to be applied to estimate the posterior distributions in closed form, based on solving an optimisation problem, in a spirit similar to 3D VAR analysis, but seeking approximations to the posterior distribution rather than simply the most probable state. A hierarchical Kalman filter is used to estimate the advection field based on the assimilated precipitation fields at two times. The model is applied to tracking precipitation dynamics in a realistic setting, using UK Met Office radar data from both a summer convective event and a winter frontal event. The performance of the model is assessed both traditionally and using probabilistic measures of fit based on ROC curves. The model is shown to provide very good assimilation characteristics, and promising forecast skill. Improvements to the forecasting scheme are discussed.     

Simulating precipitation in the Northeast United States using a climate-informed K-nearest neighbour algorithm

Decadal prediction using climate models faces long-standing challenges. While global climate models may reproduce long-term shifts in climate due to external forcing, in the near term, they often fail to accurately simulate interannual climate variability, as well as seasonal variability, wet and dry spells, and persistence, which are essential for water resources management. We developed a new climate-informed K-nearest neighbour (K-NN)-based stochastic modelling approach to capture the long-term trend and variability while replicating intra-annual statistics. The climate-informed K-NN stochastic model utilizes historical data along with climate state information to provide improved simulations of weather for near-term regional projections. Daily precipitation and temperature simulations are based on analogue weather days that belong to years similar to the current year's climate state. The climate-informed K-NN stochastic model is tested using 53 weather stations in the Northeast United States with an evident monotonic trend in annual precipitation. The model is also compared to the original K-NN weather generator and ISIMIP-2b GFDL general circulation model bias-corrected output in a cross-validation mode. Results indicate that the climate-informed K-NN model provides improved simulations for dry and wet regimes, and better uncertainty bounds for annual average precipitation. The model also replicates the within-year rainfall statistics. For the 1961–1970 dry regime, the model captures annual average precipitation and the intra-annual coefficient of variation. For the 2005–2014 wet regime, the model replicates the monotonic trend and daily persistence in precipitation. These improved modelled precipitation time series can be used for accurately simulating near-term streamflow, which in turn can be used for short-term water resources planning and management.     

Formation of homogeneous regions for regional frequency analysis of extreme precipitation events in the Czech Republic

Extreme high precipitation amounts are among environmental events with the most disastrous consequences for human society. This paper deals with the identification of ‘homogeneous regions’ according to statistical characteristics of precipitation extremes in the Czech Republic, i.e. the basic and most important step toward the regional frequency analysis. Precipitation totals measured at 78 stations over 1961–2000 are used as an input dataset. Preliminary candidate regions are formed by the cluster analysis of site characteristics, using the average-linkage clustering and Ward’s method. Several statistical tests for regional homogeneity are utilized, based on the 10-yr event and the variation of L-moment statistics. In compliance with results of the tests, the area of the Czech Republic has been divided into four homogeneous regions. The findings are supported by simulation experiments proposed to evaluate stability of the test results. Since the regions formed reflect also climatological differences in precipitation regimes and synoptic patterns causing high precipitation amounts, their future application may not be limited to the frequency analysis of extremes.