Análise de preenchimento de falhas em séries históricas de precipitação mensal no Ceará

Abstract

A historical precipitation time series analysis is essential for climatological and hydrological studies, especially in areas with high spatial and temporal variability of rainfall, such as Ceará. However, the reliability of the analyses can be affected by the frequent occurrence of gaps in pluviometric records. In this scenario, the objective of this study was to analyze the effectiveness of several traditional methods for filling gaps in monthly historical precipitation time series in Ceará, considering strategies with and without monthly seasonal separation. Series of records spanning about 50 years were analyzed, obtained from the HidroWeb/ANA platform. These data were distributed across ten different climatic regions of the state, with a main station defined for each region and secondary stations located at a maximum distance of 75 km. The methods analyzed were Regional Weighting (PR), Multiple Linear Regression (RLM), Normal Ratio (MRN), and Regional Weighting Based on Linear Regressions (PRBRL). These methods were applied consistently to the series, and performance was evaluated through cross-validation and statistical error metrics. The results showed that, in most of the regions analyzed, the methods had similar performance, with small differences among the evaluated metrics. Multiple linear regression stood out as the technique with the best average performance, although with modest gains compared to the others. The analysis of monthly seasonality affected the performance of the methods, but did not yield consistent gains; in 8 out of 10 regions performance was inferior to the global scenario (positive ΔMAE), and improvements occurred in only a few regions and with small magnitude. This suggests that the selection of the gap-filling technique should take into account not only statistical performance, but also simplicity, data availability, and computational cost.