Modeling Drought Impact Occurrence Based on Climatological Drought Indices for Europe
Monday, 15 December 2014
Meteorological drought indices are often assumed to accurately characterize the severity of a drought event; however, these indices do not necessarily reflect the likelihood or severity of a particular type of drought impact experienced on the ground. In previous research, this link between index and impact was often estimated based on thresholds found by experience, measured using composite indices with assumed weighting schemes, or defined based on very narrow impact measures, using either a narrow spatial extent or very specific impacts. This study expands on earlier work by demonstrating the feasibility of relating user-provided impact reports to the climatological drought indices SPI and SPEI by logistic regression. The user-provided drought impact reports are based on the European Drought Impact Inventory (EDII, www.geo.uio.no/edc/droughtdb/), a newly developed online database that allows both public report submission and querying the more than 4,000 reported impacts spanning 33 European countries. This new tool is used to quantify the link between meteorological drought indices and impacts focusing on four primary impact types, spanning agriculture, energy and industry, public water supply, and freshwater ecosystem across five European countries. Statistically significant climate indices are retained as predictors using step-wise regression and used to compare the most relevant drought indices and accumulation periods for different impact types and regions. Agricultural impacts are explained best by 2-12 month anomalies, with 2-3 month anomalies found in predominantly rain-fed agricultural regions, and anomalies greater than 3 months related to agricultural management practices. Energy and industry impacts, related to hydropower and energy cooling water in these countries, respond to longer accumulated precipitation anomalies (6-12 months). Public water supply and freshwater ecosystem impacts are explained by a more complex combination of short (1-3 month) and seasonal (6-12 month) anomalies. A mean of 47.0% (22.4-71.6%) impact deviance is explained by the resulting models, highlighting the feasibility of using such statistical techniques and drought impact databases to model drought impact likelihood based on relatively easily calculated meteorological drought indices.