Efeito da resolução espacial na simulação hidrológica com o modelo HIDROPIXEL-DLR

Abstract

The use of hydrological models has been driven by the increasing availability of data from remote sensing in the most diverse spatial and temporal resolutions. As a result, efforts have been made to incorporate distributed approaches to these models so that the characteristics of the basins can be properly represented in the models. However, the choice of spatial resolution is a fundamental phase of the modeling process, as the use of coarse resolutions can impair the spatial representation of basin characteristics, while the use of finer resolutions entails large amounts of data to be processed and stored, as well as the acquisition of this data can increase the cost of the process. Given this situation, several studies have sought to investigate the effect of using coarser resolutions in hydrological modeling. This work aims to evaluate the effect of spatial resolution in the distributed rainfall-runoff hydrological simulation using the Hidropixel-DLR model. Hidropixel is a distributed rainfall-runoff hydrological model that uses a pixel-level approach to estimate the runoff hydrograph. Excess rain is estimated in each of these pixels based on the SCS-CN method, and the flow generated is translated to the outlet based on the travel time of the analyzed pixel. The target study area of this research was the Upper Medway watershed (250 km²), located in the south east of England. From previous researches, carried out using the spatial resolution of 10m, the information planes were resampled to the target resolutions of this study: 30, 50, 90, 250, 500 and 1000m. Altogether 15 precipitation events were simulated in the Upper Medway basin using the Hidropixel-DLR using the calibration parameters developed in previous research using the resolution of 10m. The event response hydrographs were analyzed based on characteristic parameters, such as total volume drained, peak flow and peak time. The metrics used to evaluate the simulated hydrographs were the Nash-Sutcliffe coefficient (NSE) and the percentage of bias (PBIAS). An analysis of the effect of spatial resolution on the hydrological simulation was also carried out from other aspects, such as the assessment of changes in the topographic characteristics of the basin, evaluation of the route and length of rivers, quality of drainage networks and basin delimitation. Finally, a specific calibration process was also carried out for each resolution worked on in order to evaluate their performance in representing the hydrological response of events. The results showed that the model performed well when simulating precipitation events for the finer resolutions (30, 50 and 90m), obtaining average NSE values for these resolutions of 0.85, 0.74 and 0.58, respectively. On the other hand, the coarser resolutions showed unsatisfactory results for all events. In general aspects, the increase in the size of the pixel used by the model promoted a shortening of the drainage flow networks, resulting in a reduction in the travel times of the pixels belonging to the basin. Consequently, the hydrographs had the peak time parameter anticipated, as well as the peak flows increased. Specific calibration has greatly improved the NSE associated with hydrographs, indicating that the process can be used to enable the use of even coarser resolutions for the hydrological modeling process.