Performance-based measures are used to validate and quantify how likely the system’s results resemble that of the actual data. Its application in inundation studies is performed by comparing the extents of the predicted flood to the real event by measuring their overlap size and getting the percentage of this size to the union of both data.

In this study, performances of 1- and 2-dimensional flow models were assessed when used with different topographic data sources, rasterisation cell sizes, mesh resolution and Manning’s values with the help of Geographic Information Systems (GIS). The Generalised Likelihood Uncertainty Estimation (GLUE) was also implemented to evaluate the behaviour and the uncertainties of the Hydrologic Engineering Center-River Analysis System (HEC-RAS) steady-flow model in delineating the inundation extents when various sets of friction coefficients for floodplain and channel were utilised as inputs. Although it was not possible to perform the GLUE procedure with Telemac-2D due to the simulation time, Manning’s n performances’ effects were evaluated using ten randomly selected sets of friction for the channel and floodplain.

The LiDAR data, which had the highest resolution, performed well in all simulations, followed by Lantmäteriet data at 50 m resolution. The lowest resolution Digital Terrain Elevation Data (DTED) showed poor resemblance to the actual event and big misrepresentations of flooded areas. Rasterisation cell sizes in HEC-RAS showed minimal effect to the inundation limits when used between 1 m and 5 m, but performance started to deteriorate at 10 m (Lantmäteriet) and 20 m (LiDAR). The 10 m mesh resolution used for LiDAR behaved poorer than the 20 m mesh, which performed well in the different 2D simulations.

For HEC-RAS, =0.033 to 0.05 performed well when paired with =0.02 to 0.10. It was apparent, therefore, that the channel’s Manning’s n affected the performances of the floodplain’s . Furthermore, the study also showed that using heterogeneous roughness values corresponding to the different land use classes is not as effective as using single channel and floodplain’s Manning. The dependence of the floodplain’s roughness to the channel’s friction values had also been manifested by Telemac, even though it required lower values than the 1D simulator. = 0.007 to 0.019   and =0.01 to 0.04 gave good performance to the 2D system.

In terms of the overall model performance, HEC-RAS 1D exhibited good results for Testeboån. Even when the average distances to the actual data were estimated, the breadths were shorter compared to the most optimal output of the two-dimensional simulator, which showed more overestimated areas, despite the fact that the overlap size with the 1977 actual event was better than HEC-RAS. It could be because the measures-of-fit took into consideration the areal sizes that were over- and under-predicted aside from the overlap sizes between the observed and modelled results. This could be the same reason with the mean distances produced, wherein higher values were computed for Telemac-2D due to its bigger gap from the actual flood as brought by the enlargement in the flood extents.

But it was also made known in the study that such ambiguities in the model performance were further contributed by the characteristics of the floodplain’s topography of being flat. Testeboån’s inclination to the banks was averaged at 0.027 m/m, with the central portion at 0.002 m/m. The middle portion of the floodplain was illustrated to contain more uncertain regions, where water extents changed easily as the parameters were altered. Distances greater than 200 m were also mostly located within these inclination values or within 0.005 to 0.006 m/m. The response of distance to the floodplain’s gradient improved when the slope value became higher, and this had been particularly noticed between 0 to 50 m.