One or more processors may detect that an extreme event that occurred in a first region. One or more processors may determine that the extreme event creates a drift in a probability distribution that is an output of a model associated with the first region. One or more processors may retrain the model associated with the first region using data associated with the extreme event. One or more processors may identify a second region similar to the first region according to a defined correlation threshold. One or more processors may generate data representing potential extreme scenarios for the second region based on the detected extreme event in the first region. One or more processors may retrain a model associated with the second region based on the generated data representing potential extreme scenarios.