Rapid Watershed Delineation Using an Automatic Outlet Relocation Algorithm

Abstract

Delineating the boundaries of numerous watersheds is essential for hydrological analysis in the big data era. Various techniques, such as ArcGIS, have been developed to delineate watershed boundaries. However, existing techniques suffer from two laborious manual processes: (a) Determine flow accumulation thresholds through trial and error to derive digital stream networks; (b) Manually relocate watershed outlets to the correct stream networks. Here, we proposed an automatic outlet relocation (AOR) algorithm for rapid watershed delineation without manual processing. The AOR algorithm relocates outlets to correct stream networks by analyzing flow accumulation gradients along the flow directions. An experiment using the AOR algorithm in 1,398 watersheds across the contiguous United States showed that 94.1% of the watersheds were correctly delineated, compared to 81.3% accuracy of the ArcGIS relocation algorithm (Snap Pour Point). In addition, the AOR algorithm took 58 min to delineate the 1,398 watersheds on a 90 m‐resolution flow direction, while the ArcGIS algorithm required 586 min to perform the same task. Thus, the AOR algorithm offers an opportunity to automatically delineate a large number of watersheds for hydrological simulations in the big data era. Software with a graphical user interface for the AOR algorithm is freely available on GitHub (https://github.com/xiejx5/watershed_delineation).

A New Automatic Hydrological Station Relocation Algorithm (ASRA) for Moving Hydrological Stations Onto a Simulated Digital River Network

“…For example, 209 GRDC hydrological stations were found to have overlapping locations, and 326 stations were missing catchment areas. This issue has also been discovered by other studies (Xie et al., 2022).…”

“…It sets an empirical search radius (e.g., 5 km) to reposition the hydrological station (Lehner, 2012), which yet requires several tests or rough statistics on the moving distance of the hydrological station. In addition, the Snap Pour Point (SPP) algorithm in ArcGIS could relocate the watershed outlet by searching the grid with the highest concentration accumulation within a certain radius (Lindsay et al., 2008; Xie et al., 2022). However, these methods fall short when it comes to relocating hydrological stations optimally because they do not specifically target the grid with the least catchment area deviation within the search perimeter.…”

Mega Forest Fires Intensify Flood Magnitudes in Southeast Australia

“…The two parameters, that is, recession constant and maximum base flow index, were automatically determined by an automatic base flow identification technique (Cheng et al., 2016) and a backward filtering operation (Collischonn & Fan, 2013), respectively. As BOM does not provide boundaries for these catchments, they were delineated by an automatic outlet relocation algorithm (Xie et al., 2022) with the flow direction data from MERIT hydro (Yamazaki et al., 2019). Catchment daily rainfall data were obtained from the SILO gridded data, which are interpolated from site observations and have a high accuracy in this data‐rich region (Jeffrey et al., 2001).…”

Clustering of causal graphs to explore drivers of river discharge

“…The reason for this is that here a lot of clusters with watershed boundaries are available. However, following the approach presented by Xie et al (2022), it is possible to infer the watershed boundaries for many more clusters across Europe making them suitable for our analysis. Moreover, the European climate has a high spatial variation that depends on far more factors than have been analyzed by us so far.…”