La prévision d'étiage à long terme est l'une des questions émergentes en hydrologie en raison de la demande croissante en eau en période sèche. Des prévisions fiables de débits à longue échéance (quelques semaines à quelques mois à l'avance) peuvent améliorer la gestion des ressources en eau et de ce fait l'économie de la société et les conditions de vie aquatique. Les études limitées sur les étiages dans la littérature nous a conduit à traiter certaines des questions existantes sur l'hydrologie des étiages, principalement sur la simulation et la prévision des étiages. Notre objectif final de développer une approche d'ensemble pour la prévision à long terme des étiages se décline en plusieurs étapes préalables, telles que la caractérisation des étiages, l'évaluation de mesures existantes d'efficacité des simulations des modèles, le développement d'une version améliorée d'un modèle de simulation des étiages, et enfin l'intégration d'une approche de prévision d'ensemble. Un ensemble de bassins distribués partout en France avec une variété de conditions hydro-météorologiques a été utilisé pour l'évaluation des modèles. Cet échantillon de données a d'abord été analysé et les étiages ont été caractérisés en utilisant divers indices. Notre objectif de mieux évaluer les simulations des étiages par les modèles a conduit à proposer un critère basé sur le critère de Nash-Sutcliffe, calculé sur l'inverse des débits pour mettre davantage de poids sur les erreurs sur les très faibles débits. Les résultats montrent que ce critère est mieux adapté à l'évaluation des simulations des étiages que d'autres critères couramment utilisés..Une analyse de sensibilité structurelle a ensuite été menée pour développer une structure de modèle améliorée pour simuler les étiages. Des modèles couramment utilisés ont été choisis ici comme modèles de base pour commencer l'analyse de sensibilité. Le modèle développé, GR6J, atteint de meilleures performances à la fois sur les faibles et les hauts débits par rapport aux autres modèles existants testés. En raison de la complexité du processus pluie-débit et de l'incertitude liée aux conditions météorologiques futures, nous avons développé une approche d'ensemble pour émettre des prévisions et quantifier les incertitudes associées. Ainsi l'approche d'ensemble fournit une gamme de valeurs futures de débits sur la plage de prévision. Ici, la climatologie a été utilisée pour fournir les scénarios météorologiques en entrée du modèle pour réaliser les prévisions. Pour réduire le niveau d'incertitude lié au modèle hydrologique, des combinaisons variées de procédures de mise à jour et de corrections de sortie ont été testées. Une approche directe, similaire à ce qui peut être fait pour la prévision des crues, a été sélectionnée comme la plus efficace. Enfin, des essais ont été réalisés pour améliorer la qualité des prévisions sur les bassins influencés par les barrages, en tenant compte des variations de stockage dans les barrages amont. Testée sur les bassins de la Seine et de la Loire, l'approche a donné des résultats mitigés, indiquant le besoin d'analyses complémentaires. / Long-term stream low-flow forecasting is one of the emerging issues in hydrology due to the escalating demand of water in dry periods. Reliable long-lead (a few weeks to months in advance) streamflow forecasts can improve the management of water resources and thereby the economy of the society and the conditions for aquatic life. The limited studies on low flows in the literature guided us to address some of the existing issues in low-flow hydrology, mainly on low-flow simulation and forecasting. Our ultimate aim to develop an ensemble approach for long-term low-flow forecasting includes several prior steps such as characterisation of low flows, evaluation of some of the existing model's simulation efficiency measures, development of a better model version for low-flow simulation, and finally the integration of an ensemble forecasting approach. A set of catchments distributed over France with various hydrometeorological conditions are used for model evaluation. This data set was first analysed and low flows were characterized using various indices. Our objective to better evaluate the models' low-flow simulation models resulted in the proposition of a criterion based on the Nash-Sutcliffe criterion, but calculated on inverse flows to put more weight on the errors on extreme low flows. The results show that this criterion is better suited to evaluate low-flow simulations than other commonly used criteria. Then a structural sensitivity analysis was carried out to develop an improved model structure to simulate stream low flows. Some widely used models were selected here as base models to initiate the sensitivity analysis. The developed model, GR6J, reaches better performance in both low- as well as high-flow conditions compared to the other tested existing models. Due to the complexity of rainfall-runoff processes and the uncertainty linked to future meteorological conditions, we developed an ensemble modelling approach to issue forecasts and quantify their associated uncertainty. Thus the ensemble approach provides a range of future flow values over the forecasting window. Here observed (climatological) rainfall and temperature were used as meteorological scenarios fed the model to issue the forecasts. To reduce the level of uncertainty linked to the hydrological model, various combinations of simple updating procedures and output corrections were tested. A straightforward approach, similar to what can be done for flood forecasting, was selected as it proved the most efficient. Last, attempts were made to improve the forecast quality on catchments influenced by dams, by accounting for the storage variations in upstream dams. Tested on the Seine and Loire basins, the approach showed mixed results, indicating the need for further investigations.

Jazybayeva A.: Regional population forecast for the Republic of Kazakhstan 4 Regional population forecast for the Republic of Kazakhstan Abstract This dissertation has three objectives. The first objective is to present literature review about theoretical background of regional population forecast. The second objective is to analyze demographic situation with relation to past and current fertility, mortality and migration development in regions of Kazakhstan. The third objective is to demonstrate two practical implementations of regional population projections. The first example is a multiregional population projection with population horizon 2009-2029 for 16 administrative divisions of Kazakhstan using period data for the year 2008 and inferring required age-sex specific interregional transition data. The second example is a multiregional population projection for period 2004-2059 of four macroregions using period-observational plan 2004-2008 and imposing internal consistency relations. The second example follows generations of people born during period of recovering fertility when these generations will be approaching retirement ages. Keywords: multiregional population projections, internal migration, consistency restraints Regionální populační prognóza Republiky Kazachstán Shrnutí Tato disertace má tři...

Approved for public release; distribution in unlimited. / As mesoscale models increase in resolution there is a greater need to understand predictability on smaller scales. The predictability of a model is related to forecast skill. It is possible that the uncertainty of one scale of motion can affect the other scales due to the nonlinearity of the atmosphere. Some suggest that topography is one factor that can lead to an increase of forecast skill and therefore predictability. This study examines the uncertainty of a mesoscale model and attempts to characterize the predictability of the wind field. The data collected is from the summer, when the synoptic forcing is relatively benign. Mesoscale Model 5 (MM5) lagged forecasts are used to create a three-member ensemble over a 12-hour forecast cycle. The differences in these forecasts are used to determine the spread of the wind field. Results show that some mesoscale features have high uncertainty and others have low uncertainty, shedding light on the potential predictability of these features with a mesoscale model. Results indicate that topography is a large source of uncertainty. This is seen in all data sets, contrary to other studies. The ability of the model to properly forecast the diurnal cycle also impacted substantially on the character and evolution of forecast spread. The persistent mesoscale features were represented reasonably well, however the detailed structure of these features had a fair amount of uncertainty. / Lieutenant Junior Grade, United States Navy

M.A. (Clinical Psychology) / Please refer to full text to view abstract

Accurate rainfall forecasting using weather radar imagery has always been a crucial and predominant task in the field of meteorology [1], [2], [3] and [4]. Competitive Radial Basis Function Neural Networks (CRBFNN) [5] is one of the methods used for weather radar image based forecasting. Recently, an alternative CRBFNN based approach [6] was introduced to model the precipitation events. The difference between the techniques presented in [5] and [6] is in the approach used to model the rainfall image. Overall, it was shown that the modified CRBFNN approach [6] is more computationally efficient compared to the CRBFNN approach [5]. However, both techniques [5] and [6] share the same prediction stage. In this thesis, a different GRBFNN approach is presented for forecasting Gaussian envelope parameters. The proposed method investigates the concept of parameter dependency among Gaussian envelopes. Experimental results are also presented to illustrate the advantage of parameters prediction over the independent series prediction.

M.Com. (Economics) / Please refer to full text to view abstract

Background: Logistics have a vital role in supply chain management and those logistics operations are dependent on the availability of packaging material for packing goods and material to be shipped. Forecasting packaging material demand for a long period of time will help organization planning to meet the demand. Using time-series data with Deep Neural Networks for long term forecasting is proposed for research. Objectives: This study is to identify the DNN used in forecasting packaging demand and in similar problems in terms of data, data similar to the available data with the organization (Volvo). Identifying the best-practiced approach for long-term forecasting and then combining the approach with identified and selected DNN for forecasting. The end objective of the thesis is to suggest the best DNN model for packaging demand forecasting. Methods: An experiment is conducted to evaluate the DNN models selected for demand forecasting. Three models are selected by a preliminary systematic literature review. Another Systematic literature review is performed in parallel for identifying metrics to evaluate the models to measure performance. Results from the preliminary literature review were instrumental in performing the experiment. Results: Three models observed in this study are performing well with considerable forecasting values. But based on the type and amount of historical data that models were given to learn, three models have a very slight difference in performance measures in terms of forecasting performance. Comparisons are made with different measures that are selected by the literature review. For a better understanding of the batch size impact on model performance, experimented three models were developed with two different batch sizes. Conclusions: Proposed models are performing considerable forecasting of packaging demand for planning the next 52 weeks (∼ 1 Year). Results show that by adopting DNN in forecasting, reliable packaging demand can be forecasted on time series data for packaging material. The combination of CNN-LSTM is better performing than the respective individual models by a small margin. By extending the forecasting at the granule level of the supply chain (Individual suppliers and plants) will benefit the organization by controlling the inventory and avoiding excess inventory.

A research report submitted to the Faculty of Commerce, Law and Management, University of the Witwatersrand, Johannesburg, In partial fulfilment of the requirements for the degree of Master of Management (Finance and Investment Management), 2016 / The relationship between real economic activity and stock market performance is one that has been extensively researched throughout many decades, across many economies. Many issues and debates have stemmed involving this relationship, with the major ones including those of the significance of the relationship, nature of the relationship as well as causality and direction of causality within the relationship. This research paper examines this relationship within the South African context, comparing the pre and post 2008 global financial crisis periods. Results both in support of and contrary to theory were found as real economic activity had an immediate postitive response to shocks imposed on the stock index, whilst the stock index had an immediate negative response to shocks imposed on real economic activity. Through the use of granger causality testing, no causality was found in either direction. Furthermore, no major differences were noted between the pre and post crisis periods. / GR2018

Dissertation submitted in partial fulfillment of the requirements for the degree of Masters of Management in Finance and Investments (MMFI) in the Graduate School of Business Administration University of the Witwatersrand 2017. / The Western Cape Fishing industry has been a subject of discussion in numerous papers, in which the thrust has been to seek ways of sustaining the significantly fluctuating business. Common risk factors have been identified and strategies for managing the fishing business in turbulent periods have been proposed over the years. A closer examination of previous literature as well as empirical evidence indicate that the business has less to do to control or minimize the impact of most of its external factors, which include the Government imposed Total Allowable Catch (TAC) limit, the variability in natural marine populations, environmental factors and fuel price oscillations. In the interest of curbing the variability component which is borne by the internal factors, this study brings on board a quantitative dimension to the evaluation of the four commonly cited internal factors, namely; Earnings Per Share (EPS), Margin of Safety (MOS), Free Cash-Flow (FCF) and the Net-Worth (NW) on volatility of the fishing business. The performance of five large JSE-listed fishing firms: Brimstone, Oceana, Premier Fishing, Sea Harvest and Irvin & Johnson, is investigated with the view of modelling and forecasting their volatilities. Initially, the comparison of volatility forecasts from symmetric and asymmetric GARCH-family models is employed. The results of competing models are tested using cross-validation of mean error measures and the Superior Predictive Ability (SPA) and Model Confidence Set (MCS) tests. Later, a Vector Autoregressive (VAR) model is applied to assess the impact of the four commonly cited internal factors on volatility. The research analysis results reveal a generally high volatility of the Western Cape fishing sector stocks. When univariate GARCH models are applied, the asymmetric GARCH-family models (EGARCH and GJR), with fat tails, appear dominant in the sets of competing models for all stocks, which highlights evidence of the leverage effect in the sector. However, GARCH (1,1), outperformed its counterparts in modelling and forecasting Irvin & Johnson (AVI) and Oceana (OCE) stocks. In the VAR modelling process, the Granger-causality tests indicate limited causal-relationship between EPS, MOS, FCF and the company Net-worth with the companies’ volatility measures. The variance decomposition of the 10-year ahead forecast of volatility indicates that volatility lag, free cash flow and networth have the largest contribution on volatility in the long-run, followed by margin of safety. In view of the above observations, the research discusses recommendations to the Western Cape fishing business to improve business returns and sustainability. / MT2017

A dissertation submitted to the Wits Business School, Faculty of Commerce, Law and Management, in partial fulfillment of the requirements of the candidacy of the Masters of Management in Finance and Investments University of Witwatersrand April 2016 / The aim of this paper is to investigate financial contagion in African financial markets from the global financial crisis. Interest in this subject has grown exponentially in the recent past in light of expanding globalization. The empirical analysis is based on daily stock price indices of a sample of African countries in order to compute the stock returns and find the impact of correlations between them and the US market. The empirical evidence is based on correlation tests by Forbes& Rigobon (2002). The analysis suggests that the larger markets by market capitalization and number of traded stocks exhibit co-movement, whereas the smaller markets experience financial contagion. The results have implications for financial investment process and risk management in terms of globalization and the unfolding of financial liberalization in Africa. / GR2018