Homogeneidade e reconstrução de séries climatológicas para localidades no estado de Minas Gerais

• Ano de defesa: 2012
• Autor(a) principal: Santos, Roziane Sobreira dos
• Orientador(a): Sediyama, Gilberto Chohaku
• Banca de defesa: Silva, Welliam Chaves Monteiro da , Costa, José Maria Nogueira da , Leal, Brauliro Gonçalves
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Viçosa
• Programa de Pós-Graduação: Doutorado em Meteorologia Agrícola
• Departamento: Agrometeorologia; Climatologia; Micrometeorologia
• País: BR
• Palavras-chave em Português: Climatologia; Minas Gerais; Climatologia agrícola
• Palavras-chave em Inglês: Climatology; Minas Gerais; Agricultural climatology
• Área do conhecimento CNPq: CNPQ::CIENCIAS EXATAS E DA TERRA::GEOCIENCIAS::METEOROLOGIA::CLIMATOLOGIA
• Link de acesso: http://locus.ufv.br/handle/123456789/1510

Climatic data is extremely important in the various activities undertaken by man, providing much information concerning the atmospheric environment and the impacts on it. Therefore there is a need for reliable weather information, because a failure in the time series can compromise analysis and data interpretation. To be used safely, weather data must be statistically homogeneous, since the non-homogeneity of a series leads to an erroneous interpretation of the weather conditions to be studied. The lack of historical weather data is significant, both due to the lack of weather stations and the existing gaps, as the available records are not always continuous for long periods of observations. This work aimed to analyze methodologies to evaluate the homogeneity and trend, as well as gap filling data series of maximum and minimum temperatures of air and rainfall in Minas Gerais. The homogeneity analysis was performed by three tests for the identification of discontinuity points in time series: Standard Normal Homogeneity Test (SNHT) to a single point test, Pettitt test and Buishand test. In trend analysis the Mann-Kendall and regression analysis were used. To fill gaps, two methodologies were used: the SARIMA models proposed by Box-Jenkins that uses the station's own data with gaps, and a space-time geostatistical model that considers the neighboring stations and the series with gaps, considering both the spatial and temporal dependence of the data. In the homogeneity analysis, 73% of maximum air temperature series and 71% of minimum air temperature series were considered homogeneous. The statistically significant changes occurred in the 1990s, especially around 1997. Most trends are increasing, especially in minimum air temperature. The precipitation series showed no significant trends. The SARIMA models capture well the behavior pattern of the series, with r2 values among 0.47 to 0.68 in maximum temperature and 0.85 to 0.90 for minimum temperature. The errors associated with the models were low, with a root mean square error of about 1 ° C and bias of around 0.4 º C. The space-time geostatistical model showed better results than the Box-Jenkins methodology, being more efficient, with high values of r2 and the Willmott‟s index of agreement and mean errors very close to zero. Even in a precipitation series the forecasts were well adjusted. The lowest value of r2 found explains 70% of the precipitation data variability. For maximum temperature the smallest value of r2 is 82%, much higher than the Box-Jenkins‟ model. Comparing the two methods, the space-time is model are closer to the identity, presenting a lower dispersion around the estimated straight line and coefficient of determination greater.