Avaliação do uso do GPM (Global Precipitation Measurement) para determinação de eventos chuvosos e suas propriedades no Brasil: uma análise na escala subdiária

• Autor(a) principal: Freitas, Emerson da Silva
• Data de Publicação: 2019
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da UFPB
• Texto Completo: https://repositorio.ufpb.br/jspui/handle/123456789/18947
• topic: Intervalo Mínimo entre Eventos (MIT); Análise espaço-temporal; Análise de Componentes Principais (PCA); Análise de agrupamento; Minimum Inter-event Time (MIT); Spatial-temporal analysis; Principal Component Analysis (PCA); Cluster analysis; CNPQ::ENGENHARIAS::ENGENHARIA CIVIL

Rainfall, among the variables of the hydrological cycle of hydrographic basins, is the most important for many applications. Nevertheless, understanding the spatial-temporal properties of rains is still a challenge in developing countries due to the scarce monitoring network. This Master thesis aimed to analyze and compare spatial-temporal precipitation and its properties based on rainfall data monitored by CEMADEN (National Center for Monitoring and Alert of Natural Disaster) and estimated by the Global Precipitation Measurement (GPM) mission. For this, we used data from about 3,000 rain stations distributed throughout the country, with a high temporal resolution of 10 minutes when it rains. The study period was from 2015 to 2017. The methodology consisted of a few steps, initially a qualitative-quantitative analysis of the observed data was carried out. Rainfall events and their respective average properties (precipitated blade, duration, intensity, dry time and intermittency) were determined considering the Minimum inter-event time (MIT) criteria for gauge (grounded base data) and each pixel (GPM data) and each year. Principal components analyses and cluster analyses were applied to identify regions that have similar characteristics. It was needed to identify groups to the data comparation and analyses. The main results indicated that only 6 principal components from 44 variables are responsible for representing more than 90% of the data variability, and then 5 regions with cluster analysis were identified. With respect to the number of events and the mean rainfall depth there is a good agreement between the data, having an absolute relative error of at most 50% and having correlations of up to 0.8 for these characteristics. Regarding duration and intensity, for the first there is an overestimation and for the second an underestimation. The absolute relative errors are higher, reaching up to 160% and low correlation coefficients. Regarding precipitation systems, it was observed that the frontal precipitations were the best represented and the convective systems were the ones that presented the worst results. The main conclusion of this work is that GPM cannot be used to estimate the precipitation properties without considering an error of at least 50% in the data.