Flood mapping employing local, regional and global scale modeling methods for the Uruguay river

Resumen

The delimitation of flooded areas is important information for natural disaster management, environmental impact prediction and for territorial planning. Flooding occurs when water levels overflow the limits of a river’s channel and seep into the floodplains. When they affect people, they are considered a natural disaster. The delimitation of a flooded area, in the occurrence of an extreme event, can be performed by using field techniques, such as mapping, surveying through the afflicted area or by taking aerial photographs. However, field survey is not a trivial task and not always does it allow for the attainment of the necessary data for a given project. For that purpose, there are various techniques that can be used to assess flood-spots. Some techniques are simpler, being based only on a Digital Elevation Model (DEM), and such is the case of the HAND model (Height Above Nearest Drainage), which has garnered plenty of fame recently in the process of applying delimitations to possibly flooded areas. Other techniques are more complex, like the use of hydrodynamic modeling. There are large-scale hydrologic-hydrodynamic models that simulate runoff generation and water transportation processes in the river basin from rainfall, basin, and meteorological data. This is the case of the large-scale distributed hydrological model called MGB (Modelo de Grandes Bacias), which has been developed and employed in these types of cases. However, there are no studies in the literature that compare the results produced by these multiple approaches using differing degrees of complexity and spatial scales. It was precisely to fill this gap in scientific literature that this work was developed. Featured here, for the first time, is the validation of the MGB model’s performance in two versions, one of regional scale and another of continental scale (MGB AS), in delimiting the flooded areas, compared to a satellite image of the inundation; and with two versions of the HAND model, one generated by a local study and the other by a global reference. The case study used in this validation was a major flood that occurred in the Uruguay River, near the town of Itaqui (Rio Grande do Sul), which borders Brazil and Argentina, in October 1997. The results show that the MGB model, both in the regional and the continental scales, could delimit the flooded area, with a few overestimations in the northern portion of the image. The HAND method at the local scale also allowed a mapping of the flooded area, with overestimates for more springhead-like areas in small tributaries. The HAND method at the global scale, on the other hand, proved to be inadequate for the purpose of delimiting flooded areas in this region. The leading cause is believed to be the small drainage area threshold linked to the utilized global HAND product. From the calculation of performance metrics, like hit rate, false alarm ratio and criticality index, Regional MGB was the one that most adequately expressed the extent of the flood observed in the Landsat 5 image when compared to the others, followed by MGB AS.