Water benefits life, but it can also lead to disaster in the form of floods and landslides, and can generate conflicts if water is decreasing precisely at the time when water demand is increasing. This happens naturally and is emphasized in areas where the difference between the amount of water in the rainy and that in the dry season is very high. Like other areas in Indonesia, Surabaya experiences two seasons, that is, the rainy and dry season. In the rainy season the run off generated from rainfall is high enough but in dry season there is considerably less water. This condition means that effort to sustain water resources have very strategic role, so that floods management has to pay attention to sustaining water resources by storage water upstream, beside letting excess flow go to the sea. Surabaya River Management is aimed at achieving the target that fulfils the amount of water required from the river in dry season by controlling floods in the rainy season. The river structure operation is relied on to maintain minimum water levels as specified but not to cause overflows when a flood occurs. Therefore information on floods is required as early as possible in order to make necessary control preparation for controlling. Various flood forecasting technique have been developed using mathematical modelling where models are based on the description of the behaviour of a phenomenon or system under study (physically based model) or based on the connection of input and output variable (Data Driven Model). There is a difference in the structure operation during the rainy and the dry seasons. At the dry season, the structure operation is conducted to increase, water level of the river to fulfill the amount of water required for irrigation and drinking water, while during rainy season care is needed to control floods so that overtopping of water from the river can be prevented. Introducing operation rules for the structure in the model can make model si ate high and low discharge and can maintain water level requirements upstream of the structures. But the iteration method that is used in model causes high frequency in the change of gate opening. Flood forecasting model based on Mike 11 can forecast Gunungsari discharges 6 hours ahead with good enough result. The performance of the model increases with a decreasing error after the updating process is finished. Predicting and forecasting discharge at Wonokromo can be done using data measurements at Gunungsari or Perning. There are linear relations between the data time series of Wonokromo and Gunungsari or Perning. The time series at Wonokromo for the discharge was extracted from Mike 11 model. The input of the model is the discharge at Perning and the lateral inflow and outflow are assumed constant. M5 Model Tree and Artificial Neural Network models for forecasting Gunungsari and Perning at 1 and 3 hours ahead have very good performance, while the model for predict and forecast Gunungsari discharge at 6 hours ahead has a performance that underestimated compared with the discharge observed. Reducing the number of input attributes can reduce the complexity of the model, but the performance of the model does not deteriorate enough. Performance of the model also can be improved using data transformation where the objective of the transformation is making the data distributed more uniform. Keywords: physically based modelling, data driven modelling, Mike 11, M5 Model Tree, Artificial Neural Network, flood, discharge, predicting, forecasting.
Back to the list of MSc abstracts