Resumen de Estimación de datos faltantes de temperatura combinando IDW y una serie truncada de Fourier.

Carolina Martínez, Jorge Paredes Tavares, Sergio Ivvan Valdez.

• español

  El estado de Jalisco presenta una alta diversidad climática debido a su ubicación geográfica y a la heterogeneidad de su relieve. Los registros desde 1961 a la fecha presentan diversos problemas, siendo el más importante la ausencia de registros. Para solventar lo anterior, se propone un método de dos etapas para la estimación de datos faltantes; la primera etapa utiliza un algoritmo no supervisado, por lo que no requiere de datos históricos para hacer una primera aproximación, la segunda etapa utiliza una interpolación supervisada para corregir el error de la primera.

• English

  The state of Jalisco presents a high climatic diversity due to its geographical location and its varied landscape. The recorded data, from 1961 to present date, have several problems, the most important is the large amount of missing data. To circumvent this issue, a two-stage method for estimating missing data is proposed; the first uses an unsupervised algorithm; hence it does not require historical data. The second uses supervised interpolation for correcting the error from the first one. The unsupervised algorithm is selected as the best performed among ordinary kriging (KO), nearest neighbor (NN), inverse distance weighting (IDW), and modified inverse distance weighting (MIDW). These methods use spatial neighbors that can be affected by missing data; thus, we propose a methodology for considering variations in the amount of data. The optimal number of neighbors minimizes the median of the MAE. The estimations, from the first stage, are significantly improved in the second stage, using a truncated Fourier Series. The combined method estimates missing data with a MAE of 0.5273°C.