Automatic Spatial Rainfall Estimation on Limited Coverage Areas

Autor(es)

Fava, Maria Clara ; da Silva, Roberto Fray ; Gesualdo, Gabriela Chiquito ; Benso, Marcos Roberto ; Mendiondo, Eduardo Mario ; Saraiva, Antonio Mauro ; Botazzo Delbem, Alexandre Claudio ; Padovani, Carlos Roberto ; IEEE

• Tipo de documento: Artigo Científico
• Fonte: 2021 IEEE INTERNATIONAL WORKSHOP ON METROLOGY FOR AGRICULTURE AND FORESTRY (IEEE METROAGRIFOR 2021); v. N/A, p. 6-pg., 2021-01-01.

Resumo

Providing accurate rainfall estimation at limited coverage areas is challenging, especially when considering the lack of weather stations' maintenance and the existence of missing or incorrect data. Another source of uncertainty related to in situ stations is the need to extrapolate the measures for spatial applications. The Inverse Distance Weighted (IDW) method has been widely used to interpolate rainfall data. When using this method, two hyperparameters need to be defined, the radius of influence and the power factor. However, there are no reference values for these variables in literature for different applications because these are directly related to local features. This study proposes a framework that automatically calculates the rainfall interpolation using IDW and a cross-validation method to find its optimal hyperparameters. It can be directly implemented on any rainfall dataset, regardless of: (i) the amount of data available; (ii) the quality of the area coverage (station density); (iii) the number of weather stations; and (iv) the existence of missing values. Cross-validation is performed for each timestep to consider all the available data for all stations. The method and its symmetric mean absolute percentage error (sMAPE) were evaluated in a case study for the Pantanal Region in Brazil. (AU)

• Processo FAPESP: 14/50848-9 - INCT 2014: INCT para Mudanças Climáticas (INCT-MC)
• Beneficiário: Jose Antonio Marengo Orsini
• Modalidade de apoio: Auxílio à Pesquisa - Programa de Pesquisa sobre Mudanças Climáticas Globais - Temático
• Processo FAPESP: 19/07665-4 - Centro de Inteligência Artificial
• Beneficiário: Fabio Gagliardi Cozman
• Modalidade de apoio: Auxílio à Pesquisa - Programa eScience e Data Science - Centros de Pesquisa em Engenharia
• Processo FAPESP: 20/03029-3 - Otimização multi-objetivo para o gerenciamento de problemas complexos de enchentes
• Beneficiário: Maria Clara Fava
• Modalidade de apoio: Bolsas no Brasil - Pós-Doutorado