As climate warms during the 21st century, the resultant changes in water availability are a vital issue for society, perhaps even more important than the magnitude of warming itself. Yet our climate models disagree in their forecasts of water availability, limiting our ability to plan accordingly. This thesis investigates future water availability projections from Coupled Ocean-Atmosphere General Circulation Models (GCMs), primarily using two water availability measures: soil moisture and the Supply Demand Drought Index (SDDI). Chapter One introduces methods of measuring water availability and explores some of the fundamental differences between soil moisture, SDDI and the Palmer Drought Severity Index (PDSI). SDDI and PDSI tend to predict more severe future drought conditions than soil moisture; 21st century projections of SDDI show conditions rivaling North American historic mega-droughts. We compare multiple potential evapotranspiration (EP) methods in New York using input from the GISS Model ER GCM and local station data from Rochester, NY, and find that they compare favorably with local pan evaporation measurements. We calculate SDDI and PDSI values using various EP methods, and show that changes in future projections are largest when using EP methods most sensitive to global warming, not necessarily methods producing EP values with the largest magnitudes. Chapter Two explores the characteristics and biases of the five GCMs and their 20th and 21st century climate projections. We compare atmospheric variables that drive water availability changes globally, zonally, and geographically among models. All models show increases in both dry and wet extremes for SDDI and soil moisture, but increases are largest for extreme drying conditions using SDDI. The percentage of gridboxes that agree on the sign of change of soil moisture and SDDI between models is very low, but does increase in the 21st century. Still, differences between models are smaller than differences between SDDI and soil moisture projections. Chapter Three addresses the three major differences between SDDI and soil moisture calculations that shed light on why their future projections diverge: evaporation approximations, dependence on previous months' conditions, and the inclusion of additional variables such as runoff. We implement various changes in SDDI and a GCM vegetation scheme to test the sensitivity of each measure and to evaluate which alterations increase the similarity between SDDI and soil moisture. In addition to deconstructing the differences between SDDI and soil moisture, we analyze their projections regionally in Chapter Four. In seven regions (the southwest U.S., southern Europe, eastern China, eastern Siberia, Australia, Uruguay and Colombia), we 1) assess the forecasts of future water availability changes, 2) compare the atmospheric dynamical processes that produce rainfall and drought in the real world to the way it occurs in individual GCMs, 3) determine how these processes change as global temperatures increase, and 4) identify the most likely scenarios for future regional water availability. Chapter Five summarizes key findings by chapter, enumerating this dissertation's contributions to the field. It then discusses the limitations of existing models and measures, and suggests potential solutions for overcoming their predictive shortfalls. Finally, the chapter concludes with a proposal for future research to expand upon this dissertation work. This thesis highlights the global and zonal differences between two water availability measures, SDDI and soil moisture and identifies regions where they agree and disagree in 21st century modeled scenarios. It provides an explanation for differing projections in soil moisture and SDDI and proves that it is possible to bring convergence to their future projections, which is also applicable to PDSI. Finally, a detailed analysis of climatic changes from five GCMs made it possible to present the most likely scenarios for 21st century water availability in seven regions.

M.Ing. (Electrical and Electronic Engineering Science) / Scientific workflows (SWFs) and artificial neural networks (ANNs) have attracted the attention of researchers in many fields and have been used to solve a variety of problems. Examples of these are (a) the use of scientific workflows for the sensor web in the hydrology domain and (b), the use of ANNs for the prediction of a number of water resource variables such as rainfall, flow, water level and various other water quality variables. ANNs have proved to be a powerful tool for prediction when compared with statistical methods. The aims of this research are to develop ANNs that act as predictive models for water resources and to deploy these models as predictive tools in a scientific workflow environment. While there are guidelines in the literature relating to the factors affecting network performance, there is no standard approach that is universally accepted for determining the optimum architecture of a neural network for a given problem. The parameters of a neural network and for the learning algorithm have a major effect on the performance of the neural network. We consider various recurrent and feed-forward neural network architectures for predicting changes in the water levels of dams. We explore various' hidden layer dimensions in learning the characteristics of the training data using the back propagation learning algorithm. Trained networks are deployed as predictive model in a scientific workflows environment called VisTrails. ': We review and discuss the use of SWFs and ANNs in the hydrology domain with emphasis on the development of neural network architecture that will give the best predictions for water resources. A number of architectures are employed to examine the best accurate predictive network for historical rainfall data. The findings of training experiments are promising in terms of the use of ANNs as a water resources predictive tool. Experimental results showed how the architecture of a neural network impacts on its predictive performance. This study shows that the number of hidden nodes is important factor for the improvement of the quality of the predictions.

The Caledon River Basin is located on one of the most water-scarce region on the African continent. The water resources of the Caledon River Basin play a pivotal role in socio-economic activities in both Lesotho and South Africa but the basin experiences recurrent severe droughts and frequent water shortages. The Caledon River is mostly used for commercial and subsistence agriculture, industrial and domestic supply. The resources are also important beyond the basin’s boundaries as the water is transferred to the nearby Modder River. The Caledon River is also a significant tributary to the Orange-Senqu Basin, which is shared by five southern African countries. However, the water resources in the basin are under continuous threat as a result of rapidly growing population, economic growth as well as changing climate, amongst others. It is therefore important that the hydrological regime and water resources of the basin are thoroughly evaluated and assessed so that they can be sustainably managed and utilised for maximum economic benefits. Climate change has been identified by the international community as one of the most prominent threats to peace, food security and livelihood and southern Africa as among the most vulnerable regions of the world. Water resources are perceived as a natural resource which will be affected the most by the changing climate conditions. Global warming is expected to bring more severe, prolonged droughts and exacerbate water shortages in this region. The current study is mainly focused on investigating the impacts of climate change on the water resources of the Caledon River Basin. The main objectives of the current study included assessing the past and current hydrological characteristics of the Caledon River Basin under current state of the physical environment, observed climate conditions and estimated water use; detecting any changes in the future rainfall and evaporative demands relative to present conditions and evaluating the impacts of climate on the basin’s hydrological regime and water resources availability for the future climate scenario, 2046-2065. To achieve these objectives the study used observed hydrological, meteorological data sets and the basin’s physical characteristics to establish parameters of the Pitman and WEAP hydrological models. Hydrological modelling is an integral part of hydrological investigations and evaluations. The various sources of uncertainties in the outputs of the climate and hydrological models were identified and quantified, as an integral part of the whole exercise. The 2-step approach of the uncertainty version of the model was used to estimate a range of parameters yielding behavioural natural flow ensembles. This approach uses the regional and local hydrological signals to constrain the model parameter ranges. The estimated parameters were also employed to guide the calibration process of the Water Evaluation And Planning (WEAP) model. The two models incorporated the estimated water uses within the basin to establish the present day flow simulations and they were found to sufficiently simulate the present day flows, as compared to the observed flows. There is an indication therefore, that WEAP can be successfully applied in other regions for hydrological investigations. Possible changes in future climate regime of the basin were evaluated by analysing downscaled temperature and rainfall outputs from a set of 9 climate models. The predictions are based on the A2 greenhouse gases emission scenario which assumes a continuous increase in emission rates. While the climate models agree that temperature, and hence, evapotranspiration will increase in the future, they demonstrate significant disagreement on whether rainfall will decrease or increase and by how much. The disagreement of the GCMs on projected future rainfall constitutes a major uncertainty in the prediction of water resources availability of the basin. This is to the extent that according to 7 out of 9 climate models used, the stream flow in four sub-basins (D21E, D22B, D23D and D23F) in the Caledon River Basin is projected to decrease below the present day flows, while two models (IPSL and MIUB) consistently project enhanced water resource availability in the basin in the future. The differences in the GCM projections highlight the margin of uncertainty involved predicting the future status of water resources in the basin. Such uncertainty should not be ignored and these results can be useful in aiding decision-makers to develop policies that are robust and that encompass all possibilities. In an attempt to reduce the known uncertainties, the study recommends upgrading of the hydrological monitoring network within the Caledon River Basin to facilitate improved hydrological evaluation and management. It also suggests the use of updated climate change data from the newest generation climate models, as well as integrating the findings of the current research into water resources decision making process.