Toward global mapping of river discharge using satellite images and at-many-stations hydraulic geometry

Abstract

Rivers provide critical water supply for many human societies and ecosystems, yet global knowledge of their flow rates is poor. We show that useful estimates of absolute river discharge (in cubic meters per second) may be derived solely from satellite images, with no ground-based or a priori information whatsoever. The approach works owing to discovery of a characteristic scaling law uniquely fundamental to natural rivers, here termed a river's at-many-stations hydraulic geometry. A first demonstration using Landsat Thematic Mapper images over three rivers in the United States, Canada, and China yields absolute discharges agreeing to within 20-30% of traditional in situ gauging station measurements and good tracking of flow changes over time. Within such accuracies, the door appears open for quantifying river resources globally with repeat imaging, both retroactively and henceforth into the future, with strong implications for water resource management, food security, ecosystem studies, flood forecasting, and geopolitics. S ome 80% of the world's population and 65% of its river ecosystems are threatened by insecure water supply, yet global knowledge of the river discharges upon which these depend is surprisingly poor (1, 2). For much of the world, river gauge measurements are rare, nonexistent, or proprietary. Even wellmonitored countries have sparsely distributed networks, thus limiting current understanding of water losses along river courses, habitat changes, and flood risk (3, 4). Satellites, in contrast, provide spatially dense coverage globally, attracting calls for a global river discharge mapping capacity from space (5-10). However, previous efforts to estimate river discharge from remotely sensed observations have all required inclusion of some form of ancillary ground-based information, such as gauge measurements, bathymetric surveys, and/or calibrated hydrology models that are simply unavailable for most of the planet (11-18). To remove this dependence on ground-based information, we show that useful estimates of absolute river discharge (i.e., in units of cubic meters per second) may be derived solely from multiple satellite images of a river, with no ground-based or a priori information whatsoever, through use of a characteristic scaling law, here termed a river's atmany-stations hydraulic geometry (AMHG). As will be shown in this paper, AMHG effectively halves the number of parameters required by traditional hydraulic geometry, thus paving the way for remote estimation of a single remaining parameter-and thus river discharge-through repeated satellite image observations along a river course. The presence of AMHG is verified in 12 of 12 rivers examined, using 88 in situ gauging stations, three fieldcalibrated hydrodynamic models incorporating 772 field-surveyed bathymetric cross-sections, and 42 Landsat Thematic Mapper (TM) satellite images (SI Text, section S1, Materials and Methods and Tables S1 and S2). Following a description of width AMHG, an innovative satellite discharge estimation approach i...

Improving the representation of hydrologic processes in Earth System Models

“…The key areas are (1) improve simulations of the storage and transmission of water in the soil matrix, obtained through (a) implementing the mixed form of Richards' equation [Celia et al, 1990;Maxwell and Miller, 2005] and (b) explicitly representing macropore flow [Beven and Germann, 1982;Weiler, 2005;Nimmo, 2010;Yu et al, 2014]; (2) improve representation of hydraulic gradients throughout the soil-plantatmosphere continuum to improve simulations of root water uptake and evapotranspiration [Baldocchi and Meyers, 1998;Mackay et al, 2003;Bonan et al, 2014]; (3) improve representation of groundwater dynamics across a hierarchy of spatial scales, including improving ''among grid'' and ''within grid'' groundwater representations [Famiglietti and Wood, 1994;Troch et al, 2003;Miguez-Macho et al, 2007]; and (4) improve simulations of streamflow, by explicitly representing stream-aquifer interactions and improving parameterizations of channel/floodplain routing [Qu and Duffy, 2007;Shen and Phanikumar, 2010;MiguezMacho and Fan, 2012a;Pappenberger et al, 2012]. Underpinning all of these areas is the need to improve data sets on geophysical attributes, especially data on bedrock depth and permeability [Tesfa et al, 2009;Fan et al, 2015] and data sets on the physical characteristics of rivers [Getirana et al, 2013;Mersel et al, 2013;Gleason and Smith, 2014].…”

“…Globally-available, 1/88 resolution river flow direction, accumulation and slope hydrography datasets, e.g., HydroSHEDS [Lehner et al, 2008] and DRT [Wu et al, 2012], have emerged. Remote-sensing data sets and the application of scaling laws provides global estimates of recharge [Gleason and Smith, 2014] and water depths [Mersel et al, 2013]. The next generation remote-sensing products (e.g., the Surface Water and Ocean Topography, SWOT, expected to launch in 2020) will further expand our ability to characterize stream geometries and discharges [e.g., Getirana et al, 2013;Pavelsky et al, 2014;Pedinotti et al, 2014;Allen and Pavelsky, 2015].…”

Modelling stream flow with a discrete rainfall–runoff model and 37 GHz PDBT microwave observations: the Xiangjiang River basin case study

“…WSS and inundated area) and even stream flow in the future, i.e. by SWOT (Bates et al, 2014;Gleason and Smith, 2014;Paiva et al, 2015), can be applied to determine the weights w i using Eq. (3).…”

Metric-Resolution 2D River Modeling at the Macroscale: Computational Methods and Applications in a Braided River

“…Leopold and Maddock (1953) present two interpretations of HG including at-a-station hydraulic geometry (AHG) which describes how channel form adjusts to changes in the instantaneous discharge at a channel cross-section, and downstream hydraulic geometry (DHG) which describes how channel form varies in the along-stream direction with discharge of a common frequency such as the mean annual discharge. Gleason (2015) provides a review of HG studies and notes that several "Non-Leopoldian" HG relationships have also been developed including At-Many-Station HG (AMHG) introduced by Gleason and Smith (2014) which characterizes the longitudinal variability in AHG exponents and coefficients, and reach-average hydraulic geometry (RHG) first introduced by Jowett (1998) which uses reach-averaged flow attributes instead of cross-sectional data. For this study, Platte hydraulic model outputs are utilized to study the relationship between BR channel form with a changing hydrograph of the September 2013 highflow event.…”