Floods are random and natural occurrences that are brought on by heavy rain, flash floods, storms, broken dams, and glacial lake eruptions. Numerous studies in the literature demonstrate that the use of artificial intelligence (AI) in modeling methodologies yields outcomes for linear, non-linear, and other systems that are close to the real data. In this paper, we analyzed the state-of-the-art advancements in artificial intelligence modeling for four different types of water management variables: precipitation, streamflow, temperature, and relative humidity. We will develop several Machine Learning models in this study, including a variety of membership functions, optimizing approaches, and data set resources for training and testing. This research aims to compare different models of Artificial Intelligence specifically Deep Learning techniques such as Long-Short term memory (LSTM), and Seasonal autoregressive integrated moving average (SARIMA) in forethought extreme climatic devastation events in the Indus basin. The correspondence coefficient, neat medium inaccuracy, and core average square inaccuracy will be used as pursuance measures to assess and compare the models given. Using the model that will be most accurate, we will estimate future flows using the data of CMIP6 Models. AI systems are nearing or surpassing human performance on a growing number of demanding tasks, thanks to increased database availability and recent advancements in deep learning approaches. Photo rating, emotional analysis, sound interpretation, and strategic gameplay are all examples of this progress. These very successful deep learning models are generally deployed in a black-box method, which implies no information is supplied on how they obtain their predictions, owing to their unique non-linear structure. The current study will help in the forecasting of high storms for effective water resources management.