Accurate monthly temperature forecasting, as a key climatic parameter, plays a vital role in water resources management, agriculture, energy planning, and climate change adaptation. In this study, four models including AR, ARIMA, LSTM, and the hybrid ARIMA-LSTM model were developed to forecast monthly temperature at the Qazvin synoptic station for the next 5 years. For this purpose, monthly temperature data over a 25-year period (December 1999 to August 2024), comprising 297 records, were used for training 237 months (80%) and testing 60 months (20%), and their performance was evaluated using R, RMSE, NSE, WI, and PBIAS. The results showed that the hybrid ARIMA-LSTM model with R = 0.9918, RMSE = 1.1397°C, NSE = 0.9837, WI = 0.9959, and PBIAS = -0.41% had the best performance and reduced the RMSE by approximately 50.2% compared to the AR model, 52% compared to the ARIMA model, and 42.5% compared to the LSTM model. Residual analysis showed that the residual distribution was normal and satisfied statistical assumptions. The forecast results showed that the mean monthly temperature is expected to increase by approximately 0.93°C compared to the historical period (from 12.32°C to 13.25°C). This increase mainly occurs through the rise in minimum temperature (approximately 11.7°C) and the temperature of cold months (January and February, about 1.47°C), while the forecasted maximum temperature (23.72°C) is lower than the historical maximum, and the mean temperature of warm months (July and August) showed a decrease of about 1.02°C; a pattern indicating the compression of the annual thermal range and warming mainly through the moderation of winters. These findings reveal the necessity of planning for adaptation to climate change in the agriculture, water resources, and energy sectors of the Qazvin region.