V. V. Srinivas

Regional frequency analysis of precipitation using large-scale atmospheric variables

Effective estimates of the magnitude and frequency of precipitation are necessary for hydrological designs. However, often the available data at target site are inadequate to arrive at reliable estimates. Practicing hydrometeorologists overcome this impediment by pooling information at target site with that from other locations depicting similar characteristics of precipitation. To facilitate pooling of information, hydrometeorologists use regionalization approaches for partitioning sites in the study region into groups having similar precipitation characteristics. The conventional approaches to regionalization are based on statistics computed from observed precipitation, rather than attributes affecting hydrometeorology in a region. Therefore independent validation of the delineated regions for homogeneity in precipitation was not possible. To address this issue, a new approach is proposed. Large-scale atmospheric variables affecting the precipitation in study region and location attributes are suggested as features for regionalization by K-means cluster analysis. This allows independent validation of the identified regions for homogeneity using statistics computed from the observed precipitation. The summer monsoon rainfall (SMR) regions that are currently in use by India Meteorological Department (IMD) are shown to be heterogeneous. Subsequently the effectiveness of the proposed approach to regionalization is illustrated through application to India for delineating new SMR regions. Frequency distributions are identified to fit rainfall in the regions using L-moment–based goodness-of-fit test. Error in rainfall quantile estimates for the new regions is found to be significantly less than that estimated for the IMD SMR regions. The results show that the proposed approach to regional frequency analysis of precipitation is promising.

Post-blackening approach for modeling dependent annual streamflows

The post-blackening (PB) approach is introduced for modeling annual streamflows that exhibit significant dependence. This is a hybrid approach that blends a simple low-order, linear parametric model with the moving block resampling scheme. Empirical simulations performed using known hypothetical nonlinear parametric models, show that the hybrid model gains significantly by utilizing the merits of both the parametric model and the moving block resampling scheme (nonparametric). Following this, the performance of the PB model is tested with four annual streamflow records with complex dependence, drawn from different parts of the world. The results from these examples show that the PB approach exhibits a better performance in terms of preservation of summary statistics, dependence structure, marginal distribution, and drought characteristics of historical streamflows, compared to low-order linear parametric models and model based resampling schemes (nonparametric model). Furthermore, it offers flexibility to the modeler and is also simple to implement on a personal computer. This hybrid approach seems to offer considerable scope for improvement in hydrologic time series modeling and its applications to water resources planning.

Post-blackening approach for modeling periodic streamflows

Abstract The post-blackening (PB) approach introduced by the authors for modeling annual streamflows in an earlier work is extended to model periodic streamflows. This is basically a semi-parametric approach that blends a simple low-order, linear periodic parametric model with the moving block resampling scheme. The first part of the paper demonstrates the hybrid character of the PB model through Monte-Carlo simulations performed on hypothetical data sets drawn from a known population. Following this, the PB model is used for stochastic simulation of periodic streamflows of Beaver and Weber rivers in the US. The results show that the PB model is more consistent in reproducing a wide variety of statistics of periodic streamflows, compared to low-order linear periodic parametric models (Box–Jenkins type) and the periodic k -nearest-neighbor bootstrap (nonparametric) method. In addition, the PB model is able to preserve cross-year serial correlations as well as the month-to-year cross-correlations. This hybrid approach seems to offer considerable scope for improvement in hydrologic time series modeling.

A hybrid stochastic model for multiseason streamflow simulation

A hybrid model is presented for stochastic simulation of multiseason streamflows. This involves partial prewhitening of the streamflows using a parsimonious linear periodic parametric model, followed by resampling the resulting residuals using moving block bootstrap to obtain innovations and subsequently postblackening these innovations to generate synthetic replicates. This model is simple and is efficient in reproducing both linear and nonlinear dependence inherent in the observed streamflows. The first part of this paper demonstrates the hybrid character of the model through stochastic simulations performed using monthly streamflows of Weber River (Utah) that exhibit a complex dependence structure. In the latter part of the paper the hybrid model is shown to be efficient in simulating multiseason streamflows, through an example of the San Juan River (New Mexico). This model ensures annual-to-monthly consistency without the need for any adjustment procedures. Furthermore, the hybrid model is able to preserve both within-year and cross-year monthly serial correlations for multiple lags. Also, it is seen to be consistent in predicting the reservoir storage (validation) statistic at low as well as high demand levels.

Regional flood frequency analysis using kernel‐based fuzzy clustering approach

Regionalization approaches are widely used in water resources engineering to identify hydrologically homogeneous groups of watersheds that are referred to as regions. Pooled information from sites (depicting watersheds) in a region forms the basis to estimate quantiles associated with hydrological extreme events at ungauged/sparsely gauged sites in the region. Conventional regionalization approaches can be effective when watersheds (data points) corresponding to different regions can be separated using straight lines or linear planes in the space of watershed related attributes. In this paper, a kernel-based Fuzzy c-means (KFCM) clustering approach is presented for use in situations where such linear separation of regions cannot be accomplished. The approach uses kernel-based functions to map the data points from the attribute space to a higher-dimensional space where they can be separated into regions by linear planes. A procedure to determine optimal number of regions with the KFCM approach is suggested. Further, formulations to estimate flood quantiles at ungauged sites with the approach are developed. Effectiveness of the approach is demonstrated through Monte-Carlo simulation experiments and a case study on watersheds in United States. Comparison of results with those based on conventional Fuzzy c-means clustering, Region-of-influence approach and a prior study indicate that KFCM approach outperforms the other approaches in forming regions that are closer to being statistically homogeneous and in estimating flood quantiles at ungauged sites. Key Points Kernel-based regionalization approach is presented for flood frequency analysis Kernel procedure to estimate flood quantiles at ungauged sites is developed A set of fuzzy regions is delineated in Ohio, USA

Daily relative humidity projections in an Indian river basin for IPCC SRES scenarios

A two-stage methodology is developed to obtain future projections of daily relative humidity in a river basin for climate change scenarios. In the first stage, Support Vector Machine (SVM) models are developed to downscale nine sets of predictor variables (large-scale atmospheric variables) for Intergovernmental Panel on Climate Change Special Report on Emissions Scenarios (SRES) (A1B, A2, B1, and COMMIT) to RH in a river basin at monthly scale. Uncertainty in the future projections of RH is studied for combinations of SRES scenarios, and predictors selected. Subsequently, in the second stage, the monthly sequences of RH are disaggregated to daily scale using k-nearest neighbor method. The effectiveness of the developed methodology is demonstrated through application to the catchment of Malaprabha reservoir in India. For downscaling, the probable predictor variables are extracted from the (1) National Centers for Environmental Prediction reanalysis data set for the period 1978–2000 and (2) simulations of the third-generation Canadian Coupled Global Climate Model for the period 1978–2100. The performance of the downscaling and disaggregation models is evaluated by split sample validation. Results show that among the SVM models, the model developed using predictors pertaining to only land location performed better. The RH is projected to increase in the future for A1B and A2 scenarios, while no trend is discerned for B1 and COMMIT.

Bivariate frequency analysis of floods using a diffusion based kernel density estimator

Recent focus of flood frequency analysis (FFA) studies has been on development of methods to model joint distributions of variables such as peak flow, volume, and duration that characterize a flood event, as comprehensive knowledge of flood event is often necessary in hydrological applications. Diffusion process based adaptive kernel (D-kernel) is suggested in this paper for this purpose. It is data driven, flexible and unlike most kernel density estimators, always yields a bona fide probability density function. It overcomes shortcomings associated with the use of conventional kernel density estimators in FFA, such as boundary leakage problem and normal reference rule. The potential of the D-kernel is demonstrated by application to synthetic samples of various sizes drawn from known unimodal and bimodal populations, and five typical peak flow records from different parts of the world. It is shown to be effective when compared to conventional Gaussian kernel and the best of seven commonly used copulas (Gumbel-Hougaard, Frank, Clayton, Joe, Normal, Plackett, and Students T) in estimating joint distribution of peak flow characteristics and extrapolating beyond historical maxima. Selection of optimum number of bins is found to be critical in modeling with D-kernel.

A fuzzy approach to reliability based design of storm water drain network

This paper proposes an approach to estimate reliability of a storm water drain (SWD) network in fuzzy framework. It involves: (i) use of proposed fuzzy Monte-Carlo simulation (FMCS) methodology to estimate fuzzy reliability of conduits in the network, (ii) construction of a reliability block diagram (RBD) for the network (system) using suggested guidelines, and (iii) use of the RBD and reliability estimates of the conduits in the network to compute system reliability based on a proposed procedure. In addition, a system reliability based methodology is proposed for design/retrofitting of SWD network by optimization of its conduit dimensions. Conventionally used reliability analysis approaches assume that the cumulative distribution function (CDF) of performance function (marginal safety) of conduits follows Gaussian distribution, which cannot be ensured in the real world scenario. The proposed approach alleviates the need for making such assumptions and can account for linguistic ambiguity in variables defining the performance function. Effectiveness of the proposed approach is demonstrated on a hypothetical SWD network and a real network in Bangalore, India. Comparison of the results obtained from the proposed approach with those from conventional Monte-Carlo simulation (MCS) reliability assessment approach indicated that the estimate of system reliability and conduit reliability are higher with FMCS approach. Consequently, conduit dimensions required to attain required system (network) reliability could be expected to be lower when FMCS approach is used for designing or retrofitting a system.

Support Vector Machine Approach to Downscale Precipitation in Climate Change Scenarios

Concern over changes in global climate has increased in recent years with improvement in understanding of atmospheric dynamics and growth in evidence of climate link to long‐term variability in hydrologic records. Climate impact studies rely on climate change information at fine spatial resolution. Towards this, the past decade has witnessed significant progress in development of downscaling models to cascade the climate information provided by General Circulation Models (GCMs) at coarse spatial resolution to the scale relevant for hydrologic studies. While a plethora of downscaling models have been applied successfully to mid‐latitude regions, a few studies are available on tropical regions where the atmosphere is known to have more complex behavior. In this paper, a support vector machine (SVM) approach is proposed for statistical downscaling to interpret climate change signals provided by GCMs over tropical regions of India. Climate variables affecting spatio‐temporal variation of precipitation at each meteorological sub‐division of India are identified. Following this, cluster analysis is applied on climate data to identify the wet and dry seasons in each year. The data pertaining to climate variables and precipitation of each meteorological sub‐division is then used to develop SVM based downscaling model for each season. Subsequently, the SVM based downscaling model is applied to future climate predictions from the second generation Coupled Global Climate Model (CGCM2) to assess the impact of climate change on hydrological inputs to the meteorological sub‐divisions. The results obtained from the SVM downscaling model are then analyzed to assess the impact of climate change on precipitation over India.

Evaluation of the Index-Flood Approach Related Regional Frequency Analysis Procedures

AbstractIndex-flood related regional frequency analysis (RFA) procedures are in use by hydrologists to estimate design quantiles of hydrological extreme events at data sparse/ungauged locations in river basins. There is a dearth of attempts to establish which among those procedures is better for RFA in the L-moment framework. This paper evaluates the performance of the conventional index flood (CIF), the logarithmic index flood (LIF), and two variants of the population index flood (PIF) procedures in estimating flood quantiles for ungauged locations by Monte Carlo simulation experiments and a case study on watersheds in Indiana in the U.S. To evaluate the PIF procedure, L-moment formulations are developed for implementing the procedure in situations where the regional frequency distribution (RFD) is the generalized logistic (GLO), generalized Pareto (GPA), generalized normal (GNO) or Pearson type III (PE3), as those formulations are unavailable. Results indicate that one of the variants of the PIF procedu...