A probabilistic model for flood forecasting based on hydrological data in the state of Maranhão, Brazil

Investigating critical precipitation phenomena is essential for predicting flood events. In this study, data from 8 hydrological stations in the state of Maranhão in northeastern Brazil were analyzed. The first step involved collecting valid data from the time series of 4 municipalities (Açailândia, Arame, Buriticupu, and Santa Luzia), which were affected by floods due to heavy rainfall in the region. The aim was to use a probabilistic model to predict the occurrence of new floods based on rainfall and river discharge patterns. Atypical precipitation points were identified in boxplots, confirming the correlation between heavy rainfall and floods in 2023. The exponential equations-based model estimated the river discharge during the floods. The Flow Duration Curves (FDC) indicated the probability of events of equal or greater magnitude occurring as follows: Açailândia (10%), Arame (15%), Buriticupu (32%), and Santa Luzia (<5%). Finally, significant trends in the monthly precipitation series were investigated using the Mann-Kendall test.