Temperature forecasting is critical for urban planning, environmental monitoring, and climate resilience strategies. Each model offers unique approaches to handling seasonality, trend components, and external regressors, providing varied strengths in forecasting accuracy and computational efficiency. Prophet, developed by Facebook, is known for its flexibility and robustness in capturing seasonality. AutoRegressive Integrated Moving Average (ARIMA) and its seasonal variant Seasonal ARIMA (SARIMA) are classical models widely utilized for their statistical rigor and interpretability. The Seasonal Naïve model, while simplistic, serves as a baseline for its ease of implementation and surprising effectiveness in many cases. This research evaluates these models on a dataset comprising historical urban temperature records, assessing their predictive performance through error metrics such as Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE). Our findings reveal insights into the suitability of each model for long-term temperature forecasting, highlighting their strengths and limitations. In this research, we deployed a real-world time series dataset to test our time series model. Delhi, a city known for its diverse climate patterns, serves as an ideal setting for experimenting with our weather prediction models. This research emphasize on strengthening of weather prediction for dynamic atmospheric condition. The results aim to guide researchers and practitioners in selecting appropriate models for enhancing urban climate prediction frameworks.