Drought prediction in Apalachicola–Chattahoochee–Flint River Basin using a semi-Markov model

• Published in: Natural hazards (Dordrecht) 2016-05, Vol.82 (1), p.267-297
• Main Authors: Nnaji, Gideon A., Clark, Clayton J., Chan-Hilton, Amy B., Huang, Wenrui
• Format: Article
• Language: English
• Subjects: Civil Engineering; Correlation coefficient; Drought; Drought index; Droughts; Dynamical systems; Dynamics; Earth and Environmental Science; Earth Sciences; Environmental Management; Geophysics/Geodesy; Geotechnical Engineering & Applied Earth Sciences; Hydrogeology; Hydrology; Markov analysis; Markov chains; Mathematical models; Natural Hazards; Original Paper; River basins; River systems; Rivers; Stream discharge; Stream flow; Water runoff
• ISSN: 0921-030X; 1573-0840
• DOI: 10.1007/s11069-016-2201-8

The Markov models widely used in hydrology are not adequate for drought analysis because they are independent of previous processes in dealing with associated significant autocorrelations of hydrological events. Therefore, use of semi-Markov model becomes more realistic for studying droughts processes due to dynamics of the system. An embedded Markov-based model was developed to assess chances of occurrence of hydrological droughts in which waiting times of the series were defined explicitly. This presents a more global parametric method to define drought duration compared to those frequently used. This model was applied to monthly streamflow series of Apalachicola River, Chattahoochee River and Flint River in Apalachicola–Chattahoochee–Flint (ACF) River Basin, located in southeastern USA. The streamflow conditions below the mean resulting to Near Drought and Critical Drought conditions were considered crucial. Drought occurrence probabilities and corresponding flows indicate a 42 % chance of Near Drought condition and 18 % chance of Critical Drought condition, with transition time of about 1.8 months. The model results were validated using last ten-year data series. Correlation coefficient ( r ) and root-mean-square error statistics demonstrate that the model can predict Near Drought and Critical Drought conditions with high accuracy, resulting in errors less than 5 % statistical significance. The method is capable of preserving longer memory persistence of historic flow trends in the ACF River Basin, and gives an effective recursive equation of defining occurrence of droughts. The semi-Markov model developed in this work will provide valuable lead in estimating similar drought indices in other related river systems.