A multivariate conditional model for streamflow prediction and spatial precipitation refinement

Abstract

The effective prediction and estimation of hydrometeorological variables are important for water resources planning and management. In this study, we propose a multivariate conditional model for streamflow prediction and the refinement of spatial precipitation estimates. This model consists of high dimensional vine copulas, conditional bivariate copula simulations, and a quantile‐copula function. The vine copula is employed because of its flexibility in modeling the high dimensional joint distribution of multivariate data by building a hierarchy of conditional bivariate copulas. We investigate two cases to evaluate the performance and applicability of the proposed approach. In the first case, we generate one month ahead streamflow forecasts that incorporate multiple predictors including antecedent precipitation and streamflow records in a basin located in South China. The prediction accuracy of the vine‐based model is compared with that of traditional data‐driven models such as the support vector regression (SVR) and the adaptive neuro‐fuzzy inference system (ANFIS). The results indicate that the proposed model produces more skillful forecasts than SVR and ANFIS. Moreover, this probabilistic model yields additional information concerning the predictive uncertainty. The second case involves refining spatial precipitation estimates derived from the tropical rainfall measuring mission precipitationproduct for the Yangtze River basin by incorporating remotely sensed soil moisture data and the observed precipitation from meteorological gauges over the basin. The validation results indicate that the proposed model successfully refines the spatial precipitation estimates. Although this model is tested for specific cases, it can be extended to other hydrometeorological variables for predictions and spatial estimations.

Seasonal Drought Prediction: Advances, Challenges, and Future Prospects

“…The joint distribution has been commonly used to characterize the joint behavior of multiple drought variables, from which the conditional distribution can be constructed to model the predictand conditioned on predictors. A variety of studies have been conducted for the prediction of hydroclimatic variables based on the joint distribution (Khedun et al, ; Liu et al, ; Wang, Robertson, & Chiew, ; Wu et al, ; Yan et al, ). For example, streamflow prediction can be achieved by constructing a joint distribution of streamflow of the target period and several predictors, such as precipitation, streamflow, and climate indices in previous time steps (Liu et al, ; Souza Filho & Lall, ; Sun et al, ).…”

“…A variety of studies have been conducted for the prediction of hydroclimatic variables based on the joint distribution (Khedun et al, ; Liu et al, ; Wang, Robertson, & Chiew, ; Wu et al, ; Yan et al, ). For example, streamflow prediction can be achieved by constructing a joint distribution of streamflow of the target period and several predictors, such as precipitation, streamflow, and climate indices in previous time steps (Liu et al, ; Souza Filho & Lall, ; Sun et al, ). The multivariate distribution function can be used for statistical drought prediction by establishing the conditional distribution of the variable of interest with other predictors that may provide predictive information (Cancelliere et al, ; Hao, Hao, Singh, Sun, et al, ; Madadgar & Moradkhani, ).…”

Monthly Streamflow Simulation for the Headwater Catchment of the Yellow River Basin With a Hybrid Statistical‐Dynamical Model

“…The concurrent flow of HCYRB is recharged by precipitation, snowmelt, and groundwater, and the travel time and concentration are influenced by the underlying surface of the catchment (X Liu & Chang, 2005). The complex process can be implicitly detected by analyzing the relation of streamflow with corresponding climate drivers.…”

Multivariate statistical modelling of compound events via pair-copula constructions: analysis of floods in Ravenna (Italy)

“…to Brechmann et al (2013) and Liu et al (2015) as other works where conditional joint pdfs decomposed as C-vines were used for statistical modelling.…”