Water resources are getting scarce in each year. Thus, water resource management is becoming more crucial, especially for Turkey which is a water stress country having, approximately 1,500 m3 water per capita. Most of the water is used by the irrigation sector in Turkey. Therefore, to determine an appropriate rainfall distribution with minimum number of estimated parameters is an essential step. There are many models suggested in the literature. Williams (1952) suggests one parameter model log series distribution whereas Gabriel and Neumann (1957) suggests geometric distribution to model dry (wet) spells in Harpenden and Tel Aviv, respectively. Besides one parameter models, many researchers found mixture models are also suitable. Racsko et al. (1991) and Deni and Jemain (2008) found mixture of two geometric distributions fits quite well for the dry (wet) spells in Hungary and Peninsular Malaysia, respectively. Dobi-Wantuch et al. (2000) use the mixture of geometric and Poisson distribution for Hungary. Srinivasan (1958) suggests mixture of log series and geometric series where the estimation of parameters requires solving quadratic equations. Deni et al. (2009) try mixture of log series distribution with Poisson, truncated Poisson and geometric distribution and introduced two more alternative models; mixture of two log series distributions, and geometric distribution mixed with truncated Poisson distribution for Peninsular Malaysia.
In this study, the dry (wet) spells of Central Anatolia Region of Turkey will be considered. The dry (wet) spells data will be modeled by different probability models and/or their mixtures. The most appropriate one will be selected according to Akaike's Information Criterion. The findings will constitute a base possibly for dam construction, ground water management projects and so on.
Deni, S.M., and Jemain, A.A. (2008), Probability models for dry spells in Peninsular Malaysia, Asia-Pasific, J. Atmos. Sci. 44(1), 37-47.
Deni, S.M., Jemain, A.A., and Ibrahim, K. (2009), Mixed probability models for dry and wet spells, Statistical Methodology 6, 290-303.
Dobi-Wantuch, I., Mika, J., and Szeidl, L. (2000), Modeling wet and dry spells with mixture distributions, Meteorol. Atmos. Phys. 73, 245-256.
Gabriel, K.R., and Neumann, J. (1957), On the distribution of weather cycles by length, Quart. J. Roy. Meteor. Soc. 83, 375-379
Racsko, P., Szeidl, L., and Semenov, M. (1991), A serial approach to local stochastic weather models, Ecol. Model. 57, 27-41.
Srinivasan, T.R. (1958), A statistical model suitable for describing weather persistence, Indian J. Meteorol. Geophys. 10, 321-324.
Williams, C.B. (1952), Sequences of wet and of dry days considered in relation to the logarithmic series, Quart. J. Roy. Meteor. Soc. 78, 511-516.
Keywords: Rainfall; Dry and wet spells; Mixture models; Akaike's Information Criterion
Biography: Dr. Özlem Türker Bayrak is an Asst. Prof. Dr. at Çankaya University since 2009 and was an instructor at the same university between 2003-2009. She was a research assistant in Statistics Department of Middle East Technical University (METU) between 1996-2003. She holds Ph.D., MSc. and B.Sc. in statistics from METU. Her areas of research include time series analysis, statistical inference, external quality control schemes and data analysis. She supervises the proficiency testing scheme for fuel sector in Turkey at Petroleum Research Center at METU. She had joined many national and international projects including analysis of social effects of establishment of Hydro-Electric Power Plant. She has supervised MSc. theses and still supervising both MSc. and PhD. theses.