<p>The process of evaporation is crucial in facilitating the exchange of moisture between the earth and the air. The understanding of the evaporation trend plays a pivotal role in revealing the status of actual evaporation, proving to be highly beneficial in the allocation of regional water resources. This study investigates the estimation of evaporation rates in the southeast region of Iran using temperature, solar radiation, and wind speed data. The Long short-term memory (LSTM) and attention-mechanism based dynamic-inner partial least squares long short-term memory (ADiPLS-LSTM) models are applied in 14 distinct scenarios. The research findings indicate that the adipls LSTM model significantly enhances the predictive accuracy of the model. Specifically, in scenarios 11 to 14, this model achieved an R² value exceeding 0.9. The findings indicate that the model demonstrates a robust capacity for predicting values with a substantial number of inputs. The results suggest that the performance of LSTM and adip in scenarios characterized by a limited number of inputs is nearly equivalent. Specifically, in scenarios 1 to 4, the discrepancies in the error metrics calculated by the model are 3.7%, 4%, 6.7%, and 2.7%, respectively. The findings of this research indicate that when the temperature parameter is incorporated as an input, the model demonstrates significantly enhanced predictive capabilities compared to other scenarios. The adipls model exhibits optimal performance in scenario 11, where it effectively identifies noise within the time series, thereby improving prediction accuracy. In this scenario, the model achieved increases in the values of the CC, NSE, and R² indices by 38.91%, 104.95%, and 80.4%, respectively, while simultaneously reducing the values of the RMC and MA indices by 49.88% and 49.33%, respectively. The results indicate that the proposed model effectively predicts the daily pan evaporation of the study area.</p>