| dc.description.abstract | In recent years, the frequency of droughts have been increasing due to climate anomalies mostly attributed to climate variability and high water demand brought about by changing lifestyle and urban migration. Although several drought indices have been identified to characterise droughts in various literature and some even recommended by the World Meteorological Organisation (WMO), a more comprehensive index is the one known as Standardised Precipitation and Evapotranspiration Index (SPEI), as an extension of the well-known Standardised Precipitation Index (SPI) to include evapotranspiration in the water balance. Droughts have been described in this research using SPEI values to provide overall reasonable characteristics of different categories of drought at various timescales of 1-, 3-, 6-, 12-, 18- and 24-months. Further to make water resources management meaningful, it requires exploration of the association between SPEI and hydrological droughts represented by Standardised Flow Index (SFI). Climate anomalies leading to climate variability and frequent droughts have been associated with El Niño Southern Oscillation (ENSO) across southern Africa. These anomalies also are attributed to shifts in hydroclimatological factors. Hence it is necessary to carry out in-depth analysis of the onset and cessation of rainfall which otherwise impact agricultural production and food security of the region. These climate anomalies are being felt across the globe but vary to a large extent from one geographical location to another. For this reason, impacts from one location cannot be extrapolated to another. It is therefore necessary to develop location specific mitigation measures towards impacts of climate variability and change. It is on this premise that Botswana, located in the semiarid subtropics and relying mainly on surface water stored in reservoirs, is selected as a case study. Moreover 80% of Botswana’s population is engaged in rainfed agriculture as a source of livelihood. In order to arrive at futuristic status of rain onset and cessation dates, Standardised Flow Index (SFI) and cereal yields, an attempt has been made to use artificial intelligence framework in the form of Nonlinear Autoregressive with Exogenous input (NARX) Neural Network (NN) modelto make five (5) years ahead forecasts of the above climatic factors. This research therefore attempts to study the following using meteorological data from 14 synoptic stations for a period 1960-2016: (i) Climatic behavior under ENSO influence including trends in meteorological variables, (ii) determination of onset and cessation of rainfall including dry spell frequency analysis, (iii) determination of SPEI and SFI for climatic and hydrological drought characterisation respectively (iv) determination of the association between climatic indices and cereal yields and (v) providing predictions of stream flow drought in the form of SFI, cereal yields, onset and cessation of rain dates.
In order to achieve the above objectives, this research carried out homogeneity tests and trend analysis using four (4) absolute tests (Standard Normal Homogeneity test, Buishand test, Pettit test, Von Nuemann test) and Man-Kendall test. Results indicate that rainfall time series are fully homogeneous with minimum and maximum temperature showing some years of intervention (i.e. change points). These years of intervention closely followed patterns of ENSO events. Trends also indicate a regionalised decrease in rainfall by 14.7% coupled with 2% increase in temperature across Botswana. The influence of ENSO on local climate was also investigated by determining the degree of association between climatic variables and Sea Surface Temperatures (SSTs) on one hand and Southern Oscillation Index (SOI) on the other hand. Results show negative correlations between SST and rainfall while positive correlations were observed between rainfall and SOI following the ENSO phenomenon. To further understand the influence of summer rainfall on agriculture, seasonal characteristics in the form of dry spell frequency, number of rainy days, onset and cessation of rain were investigated. From these investigations, results show the earliest onset occurring on 30th/November in the northeast while the latest on 11th/January in the southwest. The cessation of rain dates took an opposite spatial trend with shorter rain season observed in the southwest. Shakawe, Pandamatenga and Kasane were found to have length of the rain season (LRS) of more than 100 days making them suitable for medium maturing cereals. Since droughts have been identified as a feature of climate variability, their spatiotemporal variability was explored. Botswana was found to be more vulnerable to moderate droughts which showed a high degree of association with ENSO during summer season. A common timescale of 15-months was identified to be suitable for drought monitoring across the study area. It was also observed that climatic droughts take 6 and 7 months to propagate into hydrological droughts in the Okavango and Limpopo river basins respectively. The implication of climate variability on agriculture was further investigated by determining the influence of climatic indices, LRS and ENSO on cereal yields of maize and sorghum. ENSO was found to have the greatest influence on cereal yield accounting for 85% and 78% in maize and sorghum yield variations respectively. Predictions from the three NARX-NN models reveal a likelihood of a shift in the rain season and higher variability in SFI in the near future. Cereal yield projections for the next 5 years reveal a possible yield decline in both maize and sorghum by 52 kg/ha and 126 kg/ha respectively. This research has demonstrated that Botswana’s climate is closely associated with ENSO leading to more uncertainty in rainfall and rainfed agriculture. It is envisaged that information generated from this research will enhance agriculture and water resources planning and management especially in semiarid regions where adaptation to climate variability and change is still a challenge. | en_US |
