<p>Flooding is one of the most frequently occurring natural disasters causing loss of life property damage and degradation of ecological systems. Therefore a reliable flood forecasting and warning system is crucial for mitigating the devastating impacts of floods. However the accuracy of flood forecasts can be significantly compromised by inadequate rain gauge networks and the lack of real-time rainfall data particularly in remote and inaccessible regions. The increasing availability of real-time satellite-based rainfall products offers a promising solution to these limitations. By integrating these satellite products with advanced data-driven models the accuracy and reliability of real-time flood forecasting (RTFF) can be greatly enhanced. This study aims to assess the efficacy of satellite-based rainfall products specifically the integrated multi-satellite retrievals for global precipitation measurement (IMERG) dataset and to compare them with observed rain gauge data from the India Meteorological Department for use as input in developing the RTFF model. Secondly, it proposes and evaluates a data-driven model the nonlinear autoregressive with exogenous inputs (NARX) model and compares it with the wavelet integrated NARX (W-NARX) model to predict water level at multiple lead times. The contingency test results showed that the IMERG datasets exhibited the closest agreement with the observed daily data with a probability of detection of 78.03%. For forecasting extreme flood events at longer lead times the W-NARX model performed better than the NARX model demonstrating higher R and NSE values and lower RMSE values.</p>