NEAR-REAL TIME FLOOD EXTENT MONITORING IN MARIKINA RIVER PHILIPPINES: MODEL PARAMETERISATION USING REMOTELY-SENSED DATA AND FIELD MEASUREMENTS

摘要

Floods are a persistent problem in the Philippines that need to be addressed in a more scientific way in order to mitigate its costly impacts to human lives and properties. The September 2009 floods caused by Typhoon Ketsana that devastated Metro Manila and its surroundings exemplified the need for an accurate and reliable flood monitoring tool for determining the extents of floods and for assessing the risks due to this disaster. In this study, we developed and parameterized a near-real time flood extent monitoring model for Marikina River, Philippines using the Hydrologic Engineering Center - River Analysis System (HEC RAS) program. River bathymetric surveys and cross-section measurements were conducted to generate the geometry of the Marikina River required by HEC RAS. Flood plain surface roughness coefficients needed to parameterize the model were derived from multispectral classification of a 10-m spatial resolution ALOS AVNIR-2 of the study area. The HEC RAS model was configured to accept real time 10-minute water level data from the Enhanced Flood Control and Operation Warning System (EFCOS) monitoring stations along the Marikina River as model boundary conditions. Automation scripts were used to convert the time series of water level data from the stations to a format that could be readily used by the HEC RAS model. An automation script was also developed for running the HEC RAS modeling workflow without human intervention such as assigning of model initial and boundary conditions, setting of simulation time window, model computation and generation of flood extent and flood depths. The outputs generated are then uploaded automatically to the Project NOAH (Nationwide Operational Assessment of Hazards) website where the public could view in near real-time the flooding extent along Marikina River. This near-real time generation of flood maps could be useful in providing information to the public as to the possible extent and depth of flooding in the Marikina River that could then assist in preparation for evacuation. Validation of the model using time series of water level data for two rainfall events showed an average error of -16 cm and root mean square error of 25 cm. The model has a Nash-Sutcliffe Coefficient of Model Efficiency (E) of 0.88 which signifies satisfactory model performance but requires further improvement and calibration. The study proves the usefulness of remote sensing and GIS technologies in model preparation and parameterizations, as well as in providing near-real time outputs for viewing by the public.