Accurate estimates of reference evapotranspiration (ET0) are essential for water management and agriculture, especially in data-scarce regions such as the Brazilian semi-arid. In this context, the aim of this study was to evaluate the performance of indirect empirical methods for determining ET0, Hargreaves-Samani, solar radiation, and Jensen-Haise, together with the MOD16A2 remote sensing product and the BR-DWGD gridded dataset. The methodology consisted of an analysis based on observed data from the municipalities of Garanhuns, Surubim, and Caruaru, in the Agreste region of Pernambuco, using the Penman-Monteith FAO-56 model as the standard reference. The results indicated that the solar radiation method achieved the best performance across all three municipalities, being classified as "excellent" (c > 0.85) in every case. The Jensen-Haise and Hargreaves-Samani models also performed well, with classifications ranging from "very good" to "excellent". In contrast, the MOD16A2 product showed limitations, with greater variability and lower accuracy among the evaluated sites. The BR-DWGD dataset, in turn, demonstrated strong performance, with low error margins and a high correlation with observed data. These results demonstrate that radiation-and temperature-based models are suitable for estimating ET0 in the Agreste region, while also highlighting the need to improve remote sensing ET al algorithms in climatically heterogeneous environments.