Validating the data fusion‐based drought index across Queensland, Australia, and investigating interdependencies with remote drivers

Abstract

Drought monitoring and assessments are important tasks requiring comprehensive, validated indices. Exceptional circumstances (EC) data provided by the Queensland Government are used here as "ground truth" to validate the data fusion-based drought index (DFDI) and to recalibrate the regional drought thresholds for the purpose of increasing the predictive accuracy. To achieve this, Queensland is regionalized into relatively homogeneous regions following existing climate and land use classifications, followed by a recalibration of the index's drought thresholds over each subregion. As a benchmark, the rainfall percentile ranking (RPR) and standardized precipitation index (SPI), currently employed for drought monitoring by the Australian Bureau of Meteorology, have been compared with the EC maps and the DFDI, showing an underperformance in reliably detecting droughtaffected areas by the precipitation-based indices. The results suggest that the DFDI's true-positive detection rate of water stress conditions closely resembles the EC data in 84.48, 74.23, and 76.31% of the cases across desert, grassland, and tropical/subtropical regions. In addition, correlations along with copulas have been applied to reveal teleconnections and interdependencies between DFDI and effective remote drivers across the whole state, as well as each subregion. Here, comparisons of the 6-month moving average of the DFDI and the equivalent average of the Southern Oscillation index (SOI) (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016) showed that the SOI has the highest impact on drought conditions as SOI thresholds of <−4 associated with the whole state, while the climate subregions had thresholds of <−2.2, <−3.4, and <−4 related to desert, grassland, and tropical/subtropical regions, respectively.

Application of Artificial Neural Network and Climate Indices to Drought Forecasting in South-Central Vietnam

“…However, in this study the SOI index was a possible new input variable for drought prediction in Vietnam. In fact, previous studies showed that its role in drought conditions in Australia was claimed [37]. The distance from the region's calculated climate indices to the case study might affect their appearances in the drought prediction models.…”

Machine learning information fusion in Earth observation: A comprehensive review of methods, applications and data sources

“…The fusion of the multiple data sources, e.g., satellite imaging, radar data, laser point clouds, UAV images, weather stations, crowdsourcing data, social media, and GIS have shown their advantages to greatly enhance the robustness and performance of hazard detection and avoidance systems, leading to a safer planetary anytime, anywhere. [144] Validating the data fusion-based drought index (DFDI) and recalibrating the regional drought thresholds for increasing the predictive accuracy. Among the variety of geohazards, flood and drought modeling and prediction is regarded as a very complex phenomenon which is known to be among the least understood natural hazards due to its multiple causing reasons or contributing factors operating at different temporal and spatial scales.…”