ABSTRACT
In the world, one of the most valuable resources is the water, and this importance is manifested in all the environments: conservation of the wild life and the environment, indispensable element for all the organisms, basic and indispensable input in the industry, fundamental matter of the production agricultural, component landscape, communication channel, means for the humanity's enjoyment and attraction of the tourism, energy generation, etc.
One of the most important characteristics about the water is their quality, so much for the environment where it is and it circulates, and for their uses. Currently, the legislation with respect to the water resources, in the different countries, shows the importance of conserving its natural quality, and it imposes norms and approaches that it should meet the vital liquid, for the different uses to those that is dedicated.
In this context, the reservoirs have been constituted in a very transcendent element, because they constitute in many regions of the world the most important bodies of fresh water. This can be verified when seeing that, so much in the developed nations, like in the developing countries, those that possess a significant quantity of big storages show a bigger protection against extreme events, a healthier economy,  increase of the tourism, and better conditions of life for their inhabitants, in comparison with those that possess few reservoirs.
Due to this, the objective of this work is to analyze the quality of the water in reservoirs in vertical profile, and to daily scale storm, by means of the multivariate modelling and mechanic modelling of four parameters: temperature of the water (this variable by means of both techniques), dissolved oxygen, pH and conductivity, these three last by means of multivariate methods. As case of application it was considered the Amadorio Reservoir. To make the modeling, meteorological information of the Villajoyosa station was used, the nearest to the reservoir, as well as data of the variables of state of the storage, and available information coming from a multiparameter probe installed in the dam’s facing.
The final objective of the thesis is the comparison of the methods multivariantes and mechanic, to compare its effectiveness and obtained results, and to present its advantages and inconveniences.
They were used, in the first part, several multivariate methods to process and to analyze the available data: principal components analysis, factor analysis, cluster analysis, discriminant analysis, the multiple lineal regression, canonical correlation analysis, and partial least squares regression (PLSR). With these techniques, models were obtained, with the purpose of making predictions of some variables of quality of the water.
The available information was divided in groups for calibration and validation, and descriptive statistics of the variables were obtained, analyzing its normality, as well as the transformations possible for the variables. By means of the principal components analysis the periods of calibration and validation were explored, to check the requirement of homogeneity, and anomalous data were looked for. The factor analysis was used to observe how they group the variables in possible factors. To know if the formation of conglomerates to model is an appropriate option, and to know the form in that they group the variables in them, the analysis cluster was used. The discriminant analysis was good to check the results of the cluster analysis. As exploratory technique, multiple lineal regression was used. The canonical correlation analysis allowed to model the parameters of quality of the water, also pointing out relationships between predictors variables and response variables. By means of the partial least squares regression was possible to improve the prediction of variables of quality of the water, and to get the most effective and reduced models.
Also, a multilayer mechanic model intends to explain for temperature (vertical), with the fundamental purpose of comparing the characteristics and the results of the modelling with multivariate methods, with those of a model of multiple layers with physical foundation. With the objective of knowing the pattern outlined mechanic, first the necessary data are enumerated for this model's development, then the order of the processes is described to make in the layers of the reservoir, and the physical stability of the structure of layers is analyzed. They are studied the processes of union and division of layers next, as well as the calculation of the turbulent diffusion. It continuous with the description of the algorithms for the location of the flows that enter to the reservoir, and the extraction of flows for ports and weir. Then the flows of heat are analyzed in the layers of the reservoir and their equations, besides revising like they influence in the reservoir the precipitation processes, evaporation and infiltration. Lastly, it thinks about the process for the calculation of temperature in each one of the layers of the reservoir; this is made by means of equations in finite differences, that will be resolved by means of the Thomas' Algorithm.
Finally, it has been carried out a comparative of the results obtained for both methods for the modelling of the temperature of the water in the Amadorio Reservoir (Alicante, Spain). This comparative is analyzed to show the advantages and inconveniences of both methods in front of different aspects of the modelling and their data.