Abordagem metodológica em relação ao ajuste dos parâmetros do Modelo Neyman-Scott do Pulso Retangular para simulação da chuva em escala sub-horária.

Abstract

The existence of precipitation time series in a given location has great importance for projects related to water resources, being necessary to know the precipitation series discretized over short durations (5mim). However, there are many difficulties in obtaining long series for precipitation data over short time durations. An alternative is using models allow stochastic simulation for precipitation data, which are calibrated based on statistics from historical precipitation data. One of the models that has been commonly applied in the precipitation simulation is the Neyman-Scott of the Rectangular Pulse, which is generally used in the modeling of the precipitation of each month of the year separately and practically the behavior of this one when applied annually or seasonal. Thus, the objective of this work was to quantify the influence of the set of parameters of the Neyman-Scott model of the Rectangular Pulse (NSPR) in relation to the simulated series of rain in the sub-hourly duration (5min), for the winter and summer months comparatively to historical series. The precipitation data used comprise the period from 1982 to 2015. From the pluviographic records, the statistics of the winter and summer precipitation series were obtained for the durations of 5, 10, 15, 20, 30, 60, 120, 360 , 720 and 1440min. The estimated statistics were the average, variance, covariance and proportion of dry periods. Based on these, eighteen statistical combinations (A to R) were constituted, with four repetitions each, referring to the weights attributed to the NSPR model statistics and restrictions applied to the five parameters during their estimation. The restrictions were obtained from the analysis of fifteen storms from the winter and summer time series. Thus, the four repetitions studied were merged with those used by Cowpertwait et al., (1996) and those developed by this study. Thus, seventy-two statistical combinations were analyzed for each period, however not all combinations were able to estimate the parameters. The sets of parameters adjusted for winter (47) and summer (62) were used to simulate rain through the model, which was performed 1000 times for each set of parameters. The simulated series were compared with the historical ones of the winter and summer period, using the Mean Square Relative Error (RMSE), selecting two best statistical combinations for each analyzed period, comparing the historical and simulated statistics, as well as visually. through historical and simulated series in durations of 5 and 1440min, comparing their values of maximum precipitation. Through the simulation of the model, it was verified that the statistical combinations constituted with the weights and restrictions elaborated by this study presented better results, M2 and F2 for the winter and N4 and M4 for the summer. The NSPR model was unable to represent the statistical characteristics of the time series of historical precipitation in the city of Pelotas seasonally, representing only the average.