Calibration of 2D Hydraulic Inundation Models in the Floodplain Region of the Lower Tagus River
Pestana, R.¹; Matias, M.¹; Canelas, R.²; Roque, D.³; Araújo, A.²; Van Zeller, E.⁴; Trigo-Teixeira, A.²; Ferreira, R.²; Oliveira, R.²; Heleno, S.^5
1ICIST, IST, UTL, PORTUGAL; ²CEHIDRO, IST, UTL, PORTUGAL; ³LabImagem, LNEC, PORTUGAL; ⁴ARH do Tejo, PORTUGAL; ⁵CERENA, IST, UTL, PORTUGAL

Floods are one of the most deadly natural hazards worldwide, and the deadliest in Portugal in the last 100 years. In terms of inundated area, the largest floods in Portugal occur in the Lower Tagus (LT) River. On average, the river overflows every 2.5 years, at times blocking roads and causing important agricultural damages. The economical relevance of the area and the high frequency of the relevant flood events make the LT floodplain a good pilot region to conduct a data-driven, systematic calibration work of flood hydraulic models.

This paper focus on the calibration of 2D-horizontal flood simulation models for the floods of 1996, 1997, 2001, 2006 and 2010 on a 60-km stretch of the LT River, between Tramagal and Omnias, using the commercial code Tuflow. This computational engine provides 2D solutions based on the Stelling finite-difference, alternating direction implicit (ADI) scheme that solves the full 2D free surface shallow-water flow equations and allowed the introduction of structures that constrain water flow, such as dykes.

The models were based on a digital terrain model (DTM) acquired in 2008 by radar techniques (5m of spatial resolution) and on in situ measurements of water elevation in Omnias (downstream boundary condition) and discharge in Tramagal (upstream boundary condition). Due to the relevancy of several dykes on this stretch of the LT River, non-existent on the available DTM, five of them were introduced in the models. All models have the same boundaries and were simulated using steady-state flow initial conditions. The resolution of the 2D grid mesh was 30m. Land cover data for the study area was retrieved from Corine Land Cover 2006 (CO-ordination of INformation on the Environment) with spatial resolution of 100m, and combined with estimated manning coefficients obtained in literature for the different land cover classes.

Flood extent maps, derived from satellite-born Synthetic Aperture Radar (SAR), namely ERS SAR and ENVISAT ASAR imagery, provided the spatially distributed data needed for the calibration of the hydraulic models for the several floods. The flood extent maps obtained for each simulation were then compared with the flood extent maps derived from SAR imagery for each flood and the roughness coefficients changed accordingly. The models were also calibrated in terms of the stage at the gauging station Almourol, located 12km downriver from Tramagal. The combination of the calibration results for the several past floods provided a 100 meters resolution manning coefficient map of the study area.