Weekly Water Stress Monitoring in a Savannah Environment using a new Data Fusion Drought Index

Abstract:

Due to the increasing pressure on water resources, water stress monitoring has become one of the most significant issues in water resources planning and management, especially during periods of extreme climate conditions. The present study compares the performance of four currently used data fusion based drought indices (DFDIs) to evaluate the weekly water stress at the Howard Springs OzFlux Tower in Northern Australia, covering a 3-year period from January 2011 to December 2013. In addition, a new DFDI has been developed and applied to address the individual shortcomings of the traditional indices. The proposed DFDI comprehensively considers all types of drought through a selection of indices and proxies associated with each drought type (water, vegetation etc). Here, weekly data from three different data sources (OzFlux Network, Asia-Pacific Water Monitor, and MODIS-Terra satellite) were utilized for the evaluations.

To derive the new DFDI, an appropriate set of individual standardized drought indices (SDIs) was derived, that are categorized through an advanced clustering method. For two groups in which the clustered SDIs best reflected the water availability and vegetation conditions, the variables are aggregated based on an averaging between the standardized first principal components of three different multivariate methods of PCA, FA and ICA. Then, considering those aggregated indices as well as the classifications of months into dry/wet and active/non-active, the time series of the proposed DFDI is finalized. A comparison, employing the Spearman correlation coefficient, between the proposed index and the traditional data fusion based indices shows a range of correlations from 0.46 to 0.85. The results underline that the proposed index can be more reliable in compare to the previous indices, due to simultaneously relating hydro-meteorological and ecological concepts to define the actual water stress throughout the study area.