In Amazon region, a factor which prevents the most comprehensive knowledge of water resources is the lack of hydrological data (flow and precipitation) of small and medium-sized watersheds. This is mainly due to size of the region, which increases the costs of implementation and operation of the network. In this context, this work aims to develop a model of regionalization and estimated rainfall for the state Pará For this, we applied a methodology for delineation of homogeneous regions of precipitation through the cluster analysis was then determined probability of rain for some point rainfall homogeneous region obtained with the cluster analysis by applying probability functions, and finally was given estimates of rainfall heights, using multiple. For every step we used annual and monthly averages precipitation of a time series of 31 years (period 1960-1990), obtained at the Center for Climatic Research, Department of Geography, University of Delaware, Newark site, DE, USA. Among the analyzed years, years were selected with the occurrence of El Niño and La Niña. Using the agglomerative hierarchical Ward method, having as similarity measure the Euclidean distance for annual and monthly rainfall averages six homogeneous regions of precipitation were found, except for monthly averages for rainfall series with the occurrence of El Niño and La Niña, who has four and five homogeneous regions, respectively. After the definition of homogeneous regions, probability models (Normal, Gumbel and Exponential) were fitted to determine the heights of the three sequences of rainfall time series, applied the chi-square test for this check. After the calibration step to annual rainfall, it was found that the model is best fit normal distribution the probability of exceedance observed, since average monthly precipitation for the Gumbel distribution model got better grip frequencies of exceedance. The above models were validated using the rainfall series of 12 stations of the Agência Nacional de Água (ANA), considered as target stations. At this stage, it was observed that to mean annual rainfall occurred adherence of the data to all the rainfall stations targeted because they presented the results of applying the chi-square test less than 3.84 (for normal distribution functions). And it was also found that for average monthly rainfall, there was adherence of the data to all the rainfall stations target. To simulate rainfall heights were tested for calibration models of power, according to Power and Linear model by means of multiple regression. As a criterion of performance models, the percentage relative error was used. For time series containing series every year and with the occurrence of La Niña, the model showed a lower relative. As for series with the occurrence of El Niño, the model of power had minor errors. As for the probabilistic models, the calibration results of the multiple regression models were validated with the use of rainfall stations of the ANA. In the validation step for series containing every year the percentage errors ranging from 0.2 to 39.2%, as when used in El Niño years there has been an increase in error ranging from 1.9 to 54.8%, and La Niña years from 8.5 to 55.9%. Although some estimates have had considerable errors, above 50%. The results of applying this methodology are important for a better understanding of rainfall in the state of Pará and the Amazon, and can serve as a tool for better planning and management of water resources in the region.
|
