Brazilian daily rainfall gridded data from a quality controlled dataset

Abstract

Accurate meteorological data are crucial for assessing the impacts of spatiotemporal variability in climate change on hydrology, agroecosystems, etc. This work addresses the significance of high-quality precipitation records in the current climate scenario, emphasizing their importance not only in the scientific and technical domains, but also for public institutions managing pluviometric networks. The primary objective of this study was to develop high-resolution grids (0.25° × 0.25°) of daily precipitation, utilizing data from more than 11,000 stations spanning 1961 to 2020. The dataset was sourced from the Brazilian National Hydrometeorological Network (RHN) was subjected to an automatic quality control procedure. The automatic quality control procedure involves two consecutive steps: Basic Quality Control and Absolute Quality Control. Monthly quality assessments categorized station quality as Very Low, Low, Acceptable, Good, or Excellent, and later as High Quality (HQ) and Low Quality (LQ). The methodology was evaluated using a visually inspected dataset from the Brazilian National Center for Monitoring and Early Warnings of Natural Disasters (CEMADEN). The results showed an accuracy of 98.4% in correctly identifying high-quality stations. Out of over 103 million daily records, approximately 1.6% were flagged as very low or low quality and subsequently discarded. The inverse distance weighting (IDW) interpolation method was employed for the remaining 101 million daily records to produce high-resolution gridded data of daily precipitation from 1961 to 2020. The cross-validation statistics of the interpolated data performed better than those of previous studies on the same dataset, and the gridded data estimations represented both reference climate normals well.