WATER ALLOCATION OPTIMIZATION UNDER CONFLICTING DEMAND AND SUPPLY AT A WATERSHED SCALE: THE CASE IN GIDABO DAM AND GIDABO WATERSHED, ETHIOPIA

Abstract

Water allocation under conflicting demand and supply have increased in frequency and geographic coverage areas in recent decades as results of rising population demand, urbanization, industrialization, and climate change impacts, necessitating contemporary approaches to water management. Rivers and lakes that cross political boundaries have the potential to cause conflict because states might restrict access to the water resource via over-extraction or pollution. These kinds of conflicts may arise on all geographic scales from local, national, international to global levels and they are intertwined. As a result, conflicts at one level have an impact on water allocation at other levels. This dissertation proposes interdisciplinary approaches to water allocation optimization under conflicting demand and supply using operations research (OR), systems analysis (SA), artificial intelligence (AI) and computer applications from an integrated analysis of an international basin, an internal basin, an agro-process water allocation model and an irrigation reservoir operation model. Interdisciplinary methods to water allocation are treated as numerous models in SA and the system analyst runs these models in series or parallel to generate a number of different solutions. In order to collect supportive evidences to water development in Blue Nile (Abay) in Ethiopia, cooperative and non-cooperative water allocation under conflicting demand and supply in Nile River Basin were compared from chosen mathematical modeling approaches that focus on social equality, economic efficiency, and environment protection. The dissertation analyzes the impacts of this international level conflict on an internal basin by taking the case study of Gidabo Watershed of Central Rift Valley Basin (CRVB) in Ethiopia. Due to the international level conflicts, Ethiopia has exhaustively utilized its internal basins and it necessitates to look more and more on its international basins. To shift to irrigation technologies, Ethiopia is increasingly investing in irrigation sector in order to exploit the agricultural production potential of the country so as to achieve food self sufficiency at the national level, to generate foreign currency from export earnings and to satisfy the raw material demand of local industries. To help the planning of irrigation agriculture and reservoir operation the dissertation proposes a Bi-Level Neuro-Fuzzy System (BL-NFS) Soft Computing Methodology with a feed forward water conservation goal and ixa back propagation flood control goal that were compromised through a third neural network correction algorithm. The forward operation remembers the previous states of the reservoir and with given weights and bias terms it computes the output levels for reservoir diversion, release and spillover for the given input levels of inflow, storage level and irrigation demand. On the other hand the back propagation long short time memory (LSTM) predicts the flood risk and decides on spillover. The model was trained with data from Gidabo Irrigation Dam (GID), a newly inaugurated irrigation project in Gidabo Watershed of Central Rift Valley Basin (CRVB) in Ethiopia and it showed good practical applicability. More recently, Ethiopia is transforming from agricultural economy to agricultural development lead industrialization (ADLI). There are several agro-processing opportunities for food processing, fruit punching, oil refinery, sugar production etc., but the country imports processed outputs at export market. Several agroprocessing industries are emerging with the objectives of import substitute to the unfair trade. To overcome the problem of agro-processing sector the dissertation proposes an agro-processing water allocation model. The original contribution uses two-stage production dynamic inventory control optimization (TSP-DICO) to serve the purposes of balancing production and consumption with water availability and other goals and constraints of economic and environment concerns. Any excess production beyond consumption/market demand would be removed from both agricultural and processing stages. Furthermore, the two-stage productions were also scheduled based on waste control capacities in addition to water availability constraints. Thus, the water allocation problem takes the form of hierarchical decision making model also known as a bi-level fuzzy goal programming (BL-FGP). The model was trained with sampled data collected from 178 coffee wet-processing plants from Sidama and Gedeo sub-watersheds in upper and middle reaches of Gidabo Watershed. The model showed high practical applicability measured in reliability and resilient coefficients.