Master Thesis

The Coello river basin has been studied in the past with conceptual models that show the high importance of the spatial representation for water availability and contamination in the central region of Colombia. The stateof-the-art modelling software comprise a large gamma of open source and commercial software. However, still spatial distributed systems are limited in their application and their flexibility to represent regional interactions. For this purpose and to be able to analyze the Coello river basin, this work uses the concept of the distributed Tracer-Aided Catchment model (TACD). The concept of Tracer aid (TAC) uses a gridded system that comprises upper zones that can be homogenous and groundwater systems that have interactions. This was an important component due to the future research in the mining industry in the region. For now, the adaptation from an old format in a PCRASTER library has been updated to the latest Python. This updated version will be referred as TACD2. To evaluate the distributed hydrological modelling tool a model for the Coello was build. The model parameters were calibrated to minimize the Nash-Sutcliffe error between observed and estimated flow data. However, water balance and the RMSE were evaluated. The distributed model allowed us to identify the hydrological sectorization of processes. The basin shows that the high region represents almost half of the flow that arrives to the Payandé station. Moreover, topographic variability of the basin might increase its vulnerability to extreme events.

PDF Thesis

Water management is a complex problem that relates to human and physical variables that are very hard to predict, there is a cultural and social and human driven variables (Agricultural practices, Cities growth and others) that are hard to capture into the complex interactions of the water allocation and the hydrological system. This research presents a new novel yet simple system that integrates information from river basin social and hydrological variables in an online system for decision support. The concept developed so far represents the Magdalena river system in Colombia, and allows to simulate and to share ideas. A communication bar includes a score of goodness provided by decision makers such that a level of agreement between actors can be achieved. Aside of this, the system is made open source such that other river basin can be set up and even other hydrological modelling systems can be plugged in

PDF Thesis

Hydraulic structures are criticized since it is assumed that they carry an important negative environmental effects that are summarized in the degradation of freshwater ecosystems. Colombia is categorized as the country with the second greatest hydroelectric potential in LA. Currently, it has 26 dams in operation and 30 more are projected. In this way, it is necessary to implement methodologies for the estimation of environmental flow from a holistic approach and thus achieve a reduction in damage to freshwater ecosystems. In this research, different holistic methodologies like DRIFT or BBM, analysed as reference, and the “Ecological Limits of Hydrologic Alteration (ELOHA)” is compared to the recent methodology developed by the Colombian Ministry of Environment and Sustainable Development, called “Methodological Guide for the Estimation of Environmental Flows in Colombia (GMECAC)”. ELOHA is claimed to have the scientific basis required for the consensus-building in the decisionmaking process of a basin.The ELOHA is used to assess the environmental flow required at the regional level by determining the relationships between the alteration of the flow regime and the ecological response of the ecosystem. On the other side, GMECAC is a methodology recently developed that seeks to standardize the process to estimate the required environmental flows in Colombia from a holistic perspective, considering the magnitude, duration and intensity of such flows. The comparative analysis was carried out based on three aspects: technical, social and management, according to the level 1 of the framework proposed by Opperman (2018). Within the technical component, four elements were considered; hydrology, ecology, the environmental impact assessment and flow regime. According to the results of the study, it was found that both methodologies use the same principle of hydrological characterization (high, typical and low flows), but GMECAC does not take into account the ecological processes associated with the hydrological regime. Furthermore, the GMECAC criteria to determine accepted impact does not consider the response of the ecosystem to alterations, as the criteria is based on statistical tests that only correlate the natural and modified hydrological condition. From another point of view, although ELOHA does not directly contemplate the social processes in relation to the hydrological regime, it encourages spaces for negotiation in reference to the objectives of environmental flow, while GMECAC does not regard the social dimension within the asses of the environmental flow regime. Finally, in both methodologies there is the absence of a guide to evaluate their implementation, beyond the use of some indicators.

This study presents a new statistical downscaling method called Chaotic Statistical Downscaling (CSD). The method is based on three main steps: Phase space reconstruction for different time steps, identification of deterministic chaos and a general synchronization predictive model. The Bogotá river basin was used to test the methodology. Two sources of climatic information are downscaled: the first corresponds to 47 rainfall gauges stations (1970-2016, daily) and the second is derived from the information of the global climate model MPI-ESM-MR with a resolution of 1,875° x 1,875° daily resolution. These time series were used to reconstruct the phase space using the Method of Time-Delay. The Time-Delay method allows us to find the appropriate values of the time delay and the embedding dimension to capture the dynamics of the attractor. This information was used to calculate the exponents of Lyapunov, which shows the existence of deterministic chaos. Subsequently, a predictive model is created based on the general synchronization of two dynamical systems. Finally, the results obtained are compared with other statistical downscaling models for the Bogota River basin using different measures of error which show that the proposed method is able to reproduce reliable rainfall values (RMSE=73.37).

PDF Thesis

Particle tracking is very important for the appropriate management of water resources. Morphological heterogeneities of rivers make the prediction of the particle motions difficult due to the complex numerical and physical variations in the mathematical formulation. Data availability in recent years have allowed to extend dimensionality of the problem and even use coupled models for a better understanding of those patterns. Aside from this, the hydrogeomorphic characteristics of Mountain Rivers are poorly studied around the world. In certain cases, like the river la Miel in Colombia, there are strong dynamic associated with external variables like the operation of a reservoir. The environmental conditions of the operation and the transport of particles are important to determine environmental impacts of the operation. In this research, a hydrodynamic modeling exercise coupled with particle tracking was developed to determine transport patterns. The development of this model was carried out using the Delft 3D software. Information about the hydrophysical recognition in "La Miel" river downstream of "La Miel" hydroelectric complex located in Caldas -Colombia was gathered in a campaign on 21 and 27 of July 2019. The bathymetries were collected using a ECHOMA 54v, and velocities of the river obtained with and ADCP River Ray, for a 10 km length. Data correction have been done so the digital elevation model was made and the topographic conditions for the construction of the two dimensional hydrodynamic modeling system fitted a logical representation. Permanent flow was assumed, because the variation of the areas and hydraulic conditions that are only influenced by the rules of Hydroelectric operation. Finally, the hydrodynamic model coupling was performed with the "following-up" model of particles to determine transport patterns. The main result of this research is still to follow in a project that aims to describe the movement and behavior of small marine species, the travel trajectory of a pollutant and other local uses such as forensic investigation in rivers. Results will also be used to study the dynamics of high mountain rivers.

PDF Thesis

This study presents a new statistical downscaling method called Chaotic Statistical

Este estudio aborda el uso de herramientas computacionales con el fin de obtener reglas de operación óptimas para el embalse de Hatillo (República Dominicana), considerando principalmente el propósito de disminuir inundaciones aguas abajo de la presa sin afectar los demás usos del mismo (generación hidroeléctrica y el riego para la agricultura). Debido a que es un embalse multipropósito, el problema es planteado bajo un enfoque multiobjetivo, donde se explora el uso de algoritmos evolucionarios (optimizadores), junto con funciones de aproximación de Redes Neuronales Artificiales, Redes de Base Radial y Funciones lineales (Modelos de operación paramétricos) para la búsqueda directa de las reglas de operación obtenidas de los Frentes de Pareto generados. Los modelos de operación propuestos se desarrollaron para la información disponible que comprende un periodo de 10 años (2009-2019), a nivel diario, las descargas controladas del embalse fueron definidas a partir de las funciones de aproximación, las cuales reciben como entradas las variables de estado del sistema (nivel del embalse, caudales de entrada, descargas previas), así mismo, se utilizan componentes físicos propios del sistema para definir las restricciones de las descargas controladas, límites de operación del embalse, y para definir las descargas no controladas (rebose por el vertedero). Sobre los modelos de operación se aplicaron los algoritmos de optimización para la obtención de las reglas de operación óptimas, siendo los parámetros de las funciones de aproximación las variables de decisión de cada modelo. Los algoritmos de optimización utilizados fueron el Non-dominated Sorting Genetic Algorithm II (NSGA-II) y Multiobjective Evolutionary Algorithm Based on Decomposition (MOEA/D), el proceso anterior fue realizado en JMETALPY un entorno de optimización multiobjetivo desarrollado en Python. Los resultados obtenidos muestran que es posible reducir los picos de los hidrogramas de descarga del embalse y por ende la magnitud de las inundaciones aguas abajo aplicando las reglas de operación obtenidas de la optimización de los modelos de operación. Para este caso en particular se encontró que las funciones de aproximación de Redes Neuronales y Redes de Base Radial permiten parametrizar adecuadamente las reglas de operación del embalse ya que pueden generar patrones o formas complejas que normalmente no pueden ser construidas por otras funciones, como, por ejemplo, funciones lineales. Los resultados de la optimización demuestran que las Redes Neuronales artificiales se ajustan mejor respecto a los otros métodos para este caso de estudio, siendo el NSGAII el algoritmo de optimización que mejor desempeño tiene en términos de tiempo computacional y resultados de optimización.

PDF Thesis