A flood forecasting system in New Zealand using rainfall predictions has been tested since 2000. The system was evaluated by making forecasts based on past events in a controlled setting. Results show that larger catchment areas improve forecast accuracy, and adjusting forecasts to match recent flow measurements can also help. However, overall forecast accuracy is not very high, likely due to limited data for calibrating the models. These findings suggest that the system may need further improvements to enhance its effectiveness in predicting floods.