A probabilistic prediction network for hydrological drought identification and environmental flow assessment. Issue 8 (16th August 2016)

• Record Type: Journal Article
• Authors: Liu, Zhiyong; Törnros, Tobias; Menzel, Lucas
• Is Part Of: Water resources research. Volume 52:Issue 8(2016:Aug.)
• Page Start: 6243
• Page End: 6262
• Publication Date: 2016-08-16
• Subjects: streamflow -- hydrological drought -- multivariate -- probabilistic prediction network -- environmental flow assessment; Hydrology -- Periodicals; 333.91
• DOI: 10.1002/2016WR019106
• Languages: English
• ISSNs: 0043-1397

Abstract

Abstract: A general probabilistic prediction network is proposed for hydrological drought examination and environmental flow assessment. This network consists of three major components. First, we present the joint streamflow drought indicator (JSDI) to describe the hydrological dryness/wetness conditions. The JSDI is established based on a high‐dimensional multivariate probabilistic model. In the second part, a drought‐based environmental flow assessment method is introduced, which provides dynamic risk‐based information about how much flow (the environmental flow target) is required for drought recovery and its likelihood under different hydrological drought initial situations. The final part involves estimating the conditional probability of achieving the required environmental flow under different precipitation scenarios according to the joint dependence structure between streamflow and precipitation. Three watersheds from different countries (Germany, China, and the United States) with varying sizes from small to large were used to examine the usefulness of this network. The results show that the JSDI can provide an assessment of overall hydrological dryness/wetness conditions and performs well in identifying both drought onset and persistence. This network also allows quantitative prediction of targeted environmental flow required for hydrological drought recovery and estimation of the corresponding likelihood. Moreover, the results confirm that the general network can estimate the conditional probability associated with the required flow under different precipitation scenarios. The presented methodology offers a promising tool for water supply planning and management and for drought‐based environmental flow assessment. The network has no restrictions that would prevent it from being applied to other basins worldwide. Key Points: A probabilistic prediction network is presented for hydrological drought identification This general network can also allow for drought‐based environmental flow assessment It enables estimation of the conditional dependence of rainfall and streamflow