Date of this Version
Journal of Climate Volume 18 (2005) pp. 1541-1550
Drought is a complex natural hazard that is best characterized by multiple climatological and hydrological parameters. Improving our understanding of the relationships between these parameters is necessary to reduce the impacts of drought. Data mining is a recently developed technique that can be used to interact with large databases and assist in the discovery of associations between drought and oceanic data by extracting information from massive and multiple data archives. In this study, a new data-mining algorithm [i.e., Minimal Occurrences With Constraints and Time Lags (MOWCATL)] has been used to identify the relationships between oceanic parameters and drought indices. Rather than using traditional global statistical associations, the algorithm identifies drought episodes separate from normal and wet conditions and then uses drought episodes to find time-lagged relationships with oceanic parameters. As with all association-based data-mining algorithms, MOWCATL is used to find existing relationships in the data, and is not by itself a prediction tool. Using the MOWCATL algorithm, the analyses of the rules generated for selected stations and state averaged data for Nebraska from 1950 to 1999 indicate that most occurrences of drought are preceded by positive values of the Southern Oscillation index (SOI), negative values of the multivariate ENSO index (MEI), negative values of the Pacific–North American (PNA) index, negative values of the Pacific decadal oscillation (PDO), and negative values of the North Atlantic Oscillation (NAO). The frequency and confidence of the time-lagged relationships between oceanic indices and droughts at the selected stations in Nebraska indicate that oceanic parameters can be used as indicators of drought in Nebraska.