How large is the Upper Indus Basin? The pitfalls of auto-delineation using DEMs

• Published in: Journal of hydrology (Amsterdam) Vol. 509; pp. 442 - 453
• Main Authors: Khan, Asif, Richards, Keith S., Parker, Geoffrey T., McRobie, Allan, Mukhopadhyay, Biswajit
• Format: Journal Article
• Language: English
• Published: Kidlington Elsevier B.V 13-02-2014; Elsevier
• Subjects: Automated; Automated basin delineation; Basins; Closed lake basins; Delineation; Digital elevation models; Discrete element method; Drainage basins; Earth sciences; Earth, ocean, space; Estimates; Exact sciences and technology; Geographic information systems; Hydrology; Hydrology. Hydrogeology; Upper Indus Basin; Watersheds; Asia; Indian Peninsula; Pakistan; China, People's Rep., Xizang, Tibetan Plateau; Automated basin delineation; Closed lake basins; Upper Indus Basin; Digital elevation models; environmental analysis; geomorphology; software; digital elevation models; geographic information systems; drainage patterns; extraction; quality; drainage basins; drainage; lakes; information systems; data bases; Pixel
• ISSN: 0022-1694; 1879-2707
• DOI: 10.1016/j.jhydrol.2013.11.028

•Auto pit-filling applied to DEMs can result in significant errors in catchment areas.•Past estimates of the Upper Indus Basin area are ranging between 160,000 and 266,000km2.•The best estimate of the Upper Indus Basin (UIB) area is 164,867km2, using SRTM DEM.•The Pangong Tso catchment is not part of the UIB, for at least past thousand years.•Past studies based on the over-estimated watershed areas of the UIB need revision.
Extraction of watershed areas from Digital Elevation Models (DEMs) is increasingly required in a variety of environmental analyses. It is facilitated by the availability of DEMs based on remotely sensed data, and by Geographical Information System (GIS) software. However, accurate delineation depends on the quality of the DEM and the methodology adopted. This paper considers automated and supervised delineation in a case study of the Upper Indus Basin (UIB), Pakistan, for which published estimates of the basin area show significant disagreement, ranging from 166,000 to 266,000km2. Automated delineation used ArcGIS Archydro and hydrology tools applied to three good quality DEMs (two from SRTM data with 90m resolution, and one from 30m resolution ASTER data). Automatic delineation defined a basin area of c.440,000km2 for the UIB, but included a large area of internal drainage in the western Tibetan Plateau. It is shown that discrepancies between different estimates reflect differences in the initial extent of the DEM used for watershed delineation, and the unchecked effect of iterative pit-filling of the DEM (going beyond the filling of erroneous pixels to filling entire closed basins). For the UIB we have identified critical points where spurious addition of catchment area has arisen, and use Google Earth to examine the geomorphology adjacent to these points, and also examine the basin boundary data provided by the HydroSHEDS database. We show that the Pangong Tso watershed and some other areas in the western Tibetan plateau are not part of the UIB, but are areas of internal drainage. Our best estimate of the area of the Upper Indus Basin (at Besham Qila) is 164,867km2 based on the SRTM DEM, and 164,853km2 using the ASTER DEM). This matches the catchment area measured by WAPDA SWHP. An important lesson from this investigation is that one should not rely on automated delineation, as iterative pit-filling can produce spurious drainage networks and basins, when there are areas of internal drainage nearby.