Rainfall erosivity in Brazil

• Ano de defesa: 2023
• Autor(a) principal: Teixeira, David Bruno de Sousa
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal de Viçosa; Meteorologia Aplicada
• Palavras-chave em Português: Solos - Erosão; Hidrologia; Meteorologia agrícola; Agrometeorologia
• Link de acesso: https://locus.ufv.br//handle/123456789/31039; https://doi.org/10.47328/ufvbbt.2023.154

The phenomenon known as rainfall erosivity (RE) expresses the ability of rainfall to cause soil erosion. Thus, the estimation of RE magnitudes is relevant for understanding how the erosive processes vary in time and space. Considering this, the present thesis explores the main aspects of RE in Brazil. In Chapter 1, an in-depth review of scientific literature on the RE assessment in Brazil is shown. It was found that the EI 30 has been the most employed erosivity index, while the use of pluviographic rainfall data and regression equations are the main methods for obtaining erosivity values. Kriging is the most widespread technique for obtaining RE maps in Brazil. Furthermore, the Southeast region accounts for the largest number of erosivity studies, while the North has a major lack of erosivity information. The advancements over the last decade are characterized by the use of synthetic series of rainfall and remote sensing products to estimate erosivity, as well as the use of machine learning techniques for its interpolation. In Chapter 2, a large national database was used to assess the RE patterns in time and space over the Brazilian territory. The results show that the mean annual RE value is 5,620 MJ mm ha -1 h -1 year -1 , with considerable spatial variation over the country. The RE values are more equitably distributed throughout the year in the southern region, while in some spots of the northeastern region, it is irregularly concentrated in specific months. Further analyses revealed that the annual RE gravity center for Brazil is in the Goiás State and that it presents a north- south migration pattern throughout the months. Complementarily, the erosivity density magnitudes allowed the identification of high-intensity rainfall spots. Additionally, the Brazilian territory was divided into eleven homogeneous regions regarding the RE patterns and for each defined region, a regression model was adjusted and validated. These models’ statistical metrics were considered satisfactory and, thus, can be used to estimate RE values for the whole country using monthly rainfall depths. In Chapter 3 machine learning techniques were applied to obtain an annual RE map for Brazil. According to the accuracy metrics analyzed, Random Forest (RF) is considered the model with the best prediction performance for mapping the annual RE. The covariates with higher importance for the predictions were the total annual rainfall, rainfall depth for August, and rainfall of the coldest quarter. Further analysis revealed that the northeastern of the country as well as the Serra do Mar mountains region are characterized as the areas with the highest uncertainties in the values mapped. The created map is considered an advancement regarding the availability of accurate RE values in the country. The present thesis shows the most complete panorama of the RE phenomenon in Brazil shown in the literature so far. Therefore, the values, maps, and analysis shown are relevant for improving the accuracy of soil loss estimates and for the establishment of soil conservation planning on a national scale. Keywords: Erosivity index. Universal Soil Loss Equation. Soil and water conservation. Erosivity density. Machine learning.