Climate information based streamflow and rainfall forecasts for Huai River basin using hierarchical Bayesian modeling

Abstract

Abstract.A Hierarchal Bayesian model is presented for one season-ahead forecasts of summer rainfall and streamflow using exogenous climate variables for east central China. The model provides estimates of the posterior forecasted probability distribution for 12 rainfall and 2 streamflow stations considering parameter uncertainty, and cross-site correlation. The model has a multi-level structure with regression coefficients modeled from a common multi-variate normal distribution resulting in partial pooling of information across multiple stations and better representation of parameter and posterior distribution uncertainty. Covariance structure of the residuals across stations is explicitly modeled. Model performance is tested under leave-10-out cross-validation. Frequentist and Bayesian performance metrics used include receiver operating characteristic, reduction of error, coefficient of efficiency, rank probability skill scores, and coverage by posterior credible intervals. The ability of the model to reliably forecast season-ahead regional summer rainfall and streamflow offers potential for developing adaptive water risk management strategies.

China’s socioeconomic risk from extreme events in a changing climate: a hierarchical Bayesian model

“…At the second level of the model, we assess the spread of covariate effects across provinces. A multivariate Normal distribution is considered for the regression coefficients β s and b s , respectively (Chen et al 2014;Devineni et al 2013;Kwon et al 2011). The corresponding equations are expressed as…”

“…A hierarchical Bayesian approach can help quantify model and parameter uncertainties, and provides an opportunity for uncertainty reduction through partial pooling of the common information from different regions while considering heterogeneity (Gelman and Hill 2007). Such methods have been employed to flexibly construct statistical relationships in some fields (Chen et al 2014;Devineni et al 2013;Sun et al 2015). For climate change impact analysis, a hierarchical Bayesian model could help provide reasonable ranges of potential damages.…”

Streamflow Simulation Using Bayesian Regression with Multivariate Linear Spline to Estimate Future Changes

“…Three-fold validation eliminates the possibility of overfitting, includes out-of-sample data, and exhibits the performance of the model under different climate conditions [34]. The dataset containing the entire observed period (i.e., 1951-1999) is divided into three equal blocks.…”

“…RE can be calculated for the calibration and validation period; however CE can only be calculated during the validation. The current study has modified the definition of RE and CE from the traditional definition [34] which changes the range of these two cross-validation statistics. The change is made to ensure that the definition of RE or CE can be extended to compare performances between hindcast and historical runs (shown in Section 4.2).…”

Local and regional flood frequency analysis based on hierarchical Bayesian model: application to annual maximum streamflow for the Huaihe River basin

“…Many regional flood frequency analysis models have incorporated temporal or/and spatial covariates for the mapping function (Chen et al, 2014;Lima and Lall, 2010;Renard and Lall, 2014;. The temporal covariates are often identified as time or climatic indices (e.g.…”