A seasonally blended and regionally integrated drought index using Bayesian network theory

Abstract

Among the list of all‐natural hazards, the unique characteristic of drought is its multiplex nature. Besides, the inherited regional characteristics and seasonal variation of drought make it more complicated. Conventional drought indices are inadequate to integrate seasonal and local elements. Therefore, to integrate seasonal components and regional factors, the present study emphasizes the following features: the existence of multiple drought monitoring indicators, the regional broadcasting of drought‐related statistics, and seasonal fluctuations. The seasonally blended regionally integrated drought index (SBRIDI) based on two‐stage Bayesian network theory comprehensively captures seasonal and regional variability embedded in individual drought indicators and regionally scattered homogeneous meteorological stations. It extracts information from a widely accepted standardized precipitation index and other multiscalar variants (standardized precipitation evapotranspiration index, standardized precipitation temperature index). It offers a better estimate of drought severity and pattern in the whole study region. The application of the SBRIDI is based on six meteorological stations located in the Gilgit Baltistan region of Pakistan. Among 216 time series datasets, the stage I Bayesian network selects 72 seasonally relevant standardized drought indices datasets, and the stage II Bayesian network selects 12 of the seasonally most prominent meteorological station datasets. Temporal graphical evidence shows that the SBRIDI can assess the drought risk and depicts the spatial extent of drought conditions at the regional level. The SBRIDI can highlight hotspots of drought‐prone regions, which could help freshwater management agencies and stakeholders.

A new spatiotemporal two-stage standardized weighted procedure for regional drought analysis

“…These selected drought indices provide the standardized values for the given climate indicators (precipitation and temperature) in the selected stations. The appropriate probability distributions according to time scales and stations are selected for the standardization ( Niaz et al, 2020 ; Ali et al, 2020 ; Niaz et al, 2021 ; Raza et al, 2021 ). The BIC criteria are used to select these probability distributions.…”

A New Bayesian Network-Based Generalized Weighting Scheme for the Amalgamation of Multiple Drought Indices

“…Ten, the posterior probability of any hypothesis of interest can be computed by averaging all networks. For a detailed description of Bayesian learning, the Bayesian model average approach, and marginal posterior of features (edges), see [66].…”

“…Te Bayesian information criterion (BIC) has been used to determine appropriate distribution using the propagate R Package (Spies, 2014). Detailed calculation procedures of these SDIs at these selected meteorological stations can be seen in [66]. Afterward, the datasets of these SDIs are further seasonally (monthly defned) segregated to integrate seasonal components.…”

Application of Extreme Learning Machine Algorithm for Drought Forecasting

“…Existing conventional stochastic models are inadequate for accurate drought predictions [31]. e recently developed machine learning (ML) models have extensive application in climatology including Naïve Bayes classifier, Bayesian networks [32], support vector machine (SVM), wavelet gene expression programming [33], maximum entropy, and artificial neural networks (ANNs). Results of several studies affirmed that the ML models perform comparatively better than conventional stochastic and dynamic models for drought estimation [34,35].…”