Hybrid CNN-LSTM Model for Accurate Long-Term and Short-Term Temperature Prediction: A Case Study for Bingöl and Tunceli

Abstract

Extreme and sudden weather events experienced with global warming and climate change reveal the importance of accurate air temperature prediction. For this reason, it can be used to optimize decision-making processes for a wide range of applications from health and agricultural planning to energy consumption strategies. Artificial intelligence methods are successfully applied in many application areas due to their flexibility and efficiency. Traditional weather forecasting models are inefficient in detecting sudden fluctuations and complex, irregular patterns in data. Artificial in-telligence methods overcome these limitations thanks to their ability to process big data and capture long-term temporal dependencies. In this study, the aim is to predict future temperature changes more accurately by capturing patterns in past data with the developed CNN-LSTM hybrid model. The developed hybrid model is compared in detail with RF, SVM, CNN, and LSTM. The compared models were tested using daily average temperature data between 1961-2024 and hourly temperature data between 2020-2024. Experiments have shown that CNN-LSTM outperforms the compared models with R2 value above 0.97 in all scenarios.