Reconstrução de séries e análise geoestatística da precipitação no Estado de Minas Gerais

• Autor(a) principal: Caram, Rochane de Oliveira
• Data de Publicação: 2007
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: LOCUS Repositório Institucional da UFV
• Texto Completo: http://locus.ufv.br/handle/123456789/5284
• topic: Precipitação; Homogeneidade; Geoestatística; Precipitation; Homogeneity; Geostatistics; CNPQ::CIENCIAS EXATAS E DA TERRA::GEOCIENCIAS::METEOROLOGIA::CLIMATOLOGIA

Annual rainfalls data, provived by the Agência Nacional das Águas (ANA), were analyzed with objective of identifying climatically homogeneous data series using the statistical technique of "cumulative residues", starting from data of 78 meteorological stations network of the State of Minas Gerais. For all the climatically homogeneous areas, a reference station was chosen, and later, each reference station data series were analyzed against other rainfall stations within the same homogeneous climatic area. For all over the areas, it was chosen a reference station with the largest number of complete data series, and later, each reference data series was analyzed against the other data series of rainfall stations belonging to the same climatic area. Starting from the identification of statistically homogeneous rainfall data series, an attempt of reconstruction of the series was carried out, by filling of missing values. The "cumulative residues" technique was shown to be very efficient, even when the number of stations used for some climatically homogeneous areas was a limiting factor. After the data series reconstructions, for each station the inverse of the square of the distance method of interpolation (IDW) and the method of ordinary krigging (KG) were used, to analyze the annual precipitation in space for every year between 1942 and 2000. For the krigging algorithm, semi-variograms were adjusted, to verify the degree of space dependence (IDE), by means of spherical model. The IDE mean values found for all analyzed years were about 30.3%, which represent moderate space dependence. Both methods of interpolations employed were also evaluated based on cross validation test, in which it did supply the values of mean error (Em) and the coefficient of variation (ρf). The mean error average found in between 8.198 for the IDW method and 0.457 for the krigging. The coefficients of variation presented varied from 0.121 to 0.372 for IDW and 0.112 to 0.352 for krigging. The results showed that the geostatistic method of ordinary krigging was more efficient than the IDW method with the smallest mean errors (Em) and coefficients of variation (ρf).