*2.1. Database*

The Brazilian territory is the fifth largest country in the world with 8,547,403 km<sup>2</sup> of territorial extension, of which 50% are constituted by the Amazon region. The database used was obtained from the project "Meteorological Data Set for Teaching and Research" of INMET (www.inmet.gov.br, accessed on 30 September 2020) and consists of rainfall collected by 411 automatic weather stations (Figure 1) with 1-hour sampling during the period from 1 January 2008 to 31 December 2020.

**Figure 1.** Geographic location of Brazil with spatial distribution of automatic stations used in this study.

### *2.2. Clustering Method*

The diurnal precipitation cycle was analyzed from a perspective of homogeneous regions. In order to separate homogeneous groups, we applied a multivariate statistical technique of cluster analysis. This method joins elements with similar characteristics, thus increasing the similarity between the groups and the difference between the groups [14,15]. The first step of the process is to estimate the degree of similarity or dissimilarity among the data and, in the present study, the Euclidean distance that according to [16] was used, which is frequently used for this purpose in climatic studies in Brazil [17,18]. Additionally, we choose the Ward method, which is a hierarchical clustering [19] based on the minimum variance within a cluster [20]. For the grouping was performed in four, five, and six groups.
