The Cheiamaz software has been developed for flood forecasting in Manaus based on statistical-type modelling of the stage time series recorded at the main river gauges in the Brazilian Amazon basin. Forecast equations have been established for several lead times from 10 to 60 days, using a forward stepwise regression method, coupled with a procedure that gradually eliminates variables with missing data. The mean square errors, both for calibration and validation, vary from less than 10 cm to about 40 cm for the 10-day and 60-day lead times respectively. A neural method has also been tested to forecast the peak flood level. A posteriori comparisons of previous forecasts with actual stage observations, once available, are made possible with the Cheiamaz software.