Characteristics of wet and dry spells in Tanzania during the rainy seasons

Abstract

Rainfall extremes which cause floods or droughts are known to have capacity to destroy many years of national economic investments on roads, industries, energy, agriculture and other social economic infrastructures. Efforts have thus been carried out in this study to address the space-time . characteristics of wet and dry spells, which causes extremes to minimize The study narrowed down to a number of specific objectives. They various risks that are associated with negative impacts of rainfall and climate extremes.included examination of the characteristics of the wet/dry spells in terms of frequency/probability distribution,. modeling occurrence of rain and daily amounts during the major rainfall seasons. Others were examination of trend and cyclical characteristics of the weUdry spells to detect any significant inter annual changes. The thitd involved examination of the differences in the patterns of dry/wet spells during some years with unique atmospheric circulation like during the\periods with strong Indian Ocean Dipole and EI Nino Southern Oscillation. The last was on potential to derive projection scenarios of the weUdry spells based on Indian Ocean Dipole and EI Nino Southern Oscillation evolutions. The datasets used in this study included daily rainfall from 22 locations in Tanzania, sea surface temperature over the tropical Indian Ocean from which Indian Ocean Dipole indices were derived and Southern Oscillation Indices from the Pacific Ocean. The period of data was for the period 1960 - 2000. Methods of analyses included residual mass curves for data quality; frequency/probability analyses; fitting of Markov chain models to occurrence of rain; fitting of gamma models to daily rainfall amount; composite analyses; trend, spectral and wavelet analyses together with correlation and regression methods. Residual mass curves indicated that data used was of good quality for sound statistical inferences. Mean duration of seasons for the bimodal locations observed to last for about 73 and 61 days during March-May and October-December seasons respectively and for about 168 days on unimodal stations during October-May season. As for the frequency of wet/dry spells, the longest run of 28 days of wet spells was observed on the highland stations while the longest run of dry spells of 249 days was observed in 1999 in central Tanzania. Markov chain models simulated wet/dry spells using first order Markov chain models quite realistically and compared well with the observed especially in the wetter parts of Tanzania. Gamma distribution on the other hand, fitted the deily amount of rainfall best with the shape parameter (k) for the' fitted gamma distri~ution rangihg from 0.59 to 0.98 an indication of large coefficient of variations that increased from wetter high grounds to the semi arid parts of Tanzania. Rainfall trends were noted to be localized over most parts of Tanzania. Positive trends were observed over parts of Lake Victoria basin during March, April and May and October, November and December seasons and south eastern highland~of Tanzania during the October to May single season. Negative trends were common over the western parts of southern Tanzania around Lake Tanganyika and north eastern highlands of the country. Spectral and wavelet analyses indicated spectral bands of 5 to 7 days, 10 to 15 days and 20 to 30 days at most locations in Tanzania. Dominance of wavelet bands of more than 20 days over bimodal locations in northern Tanzania during October -December wet season may indicate sensitivity of the phases of MJO in the region. Composite analyses showed that the number of wet (dry) spells was greatly enhanced at most of locations during the years of strong positive (negative) ENSO and 100 events. It was also shown that some simple time lagged regression equations based on 100 and SOl predictors could be used to build some scenarios of the expected wet/dry spells. The study has hence identified some unique space-time characteristics of the wet/dry spells in Tanzania that are quite useful in the monitoring, building expectation scenarios, planning and management of many weather/climate sensitive activities including agriculture and water resources. The findings could also be useful for the policy makers to address some of the climate risk management challenges to various climate sensitive sectors and enhance the national disaster management, poverty reduction and national sustainable development efforts