Plain language summary
Estimates of water flow in rivers are needed to manage water resources and flood risk. However, many of the world’s rivers are not gauged, limiting hydrological understanding of river response to changing environmental conditions and storm events. Here, we demonstrate the use of satellite-acquired video to compute river discharges by mapping velocity based on the movement of surface water features from one video frame to the next. Using a video of a flood on the River Tilpa, Australia, our results agree to 0.3 - 15% of ground-based measurements. Micro- and nano- satellite configurations that are light, cheap and can be deployed to acquire video anywhere globally will contribute to measuring discharge on ungauged rivers.
Introduction
Globally, 29% of the world’s population are exposed to flood risk and insecure water supplies, yet knowledge of the river discharges upon which flood risk and water resource management are based remains inadequate (Rentschler and Salhab, 2020). Global monitoring networks for quantifying river discharge are in decline and remain logistically and politically difficult to access (King et al ., 2018; Lins, 2008; Zakharova et al ., 2020). Satellite-based remote sensing approaches to monitoring discharge are helping to alleviate these issues and expanding the availability of discharge data globally.
Previous approaches to satellite-based river discharge monitoring typically rely upon various statistical and hydraulic approximations to make indirect estimates of river discharge from space. Popular amongst these methods, satellite radar altimetry measures water elevations at virtual river cross-sections (Revel et al ., 2023; Tarpanelliet al ., 2013; Zakharova et al ., 2020) and near-simultaneous optical imagery infers water surface flow velocity from space (Kääb et al ., 2019). Other satellite approaches have relied on remote sensing of discharge (RSQ) algorithms, which retrieve hydraulic variables such as stage from remotely sensed data and then relate these quantities to river discharge (Q ) to supplement gauge networks using hydrological models and data assimilation techniques (e.g., Gleason and Durand, 2020; Riggs et al ., 2022). These techniques are limited by relatively coarse spatial resolution and the requirement for near-simultaneous satellite swath overlaps, which limits global coverage. Sichangi et al . (2016) provide a detailed review of indirect techniques to estimate river discharge including: correlations of satellite-derived water surface area with discharge rating curves (e.g., Pan et al ., 2016; Pavelsky, 2014), hydraulic estimations (e.g. Bjerklie et al ., 2018; Durga Rao et al ., 2020), and the use of remote sensing data and scaling laws based on at-many-stations hydraulic geometry (AMHG) (e.g., Barber and Gleason, 2018; Brinkerhoff et al ., 2019).
Optical satellite video sensors can record dynamic phenomena on the earth’s surface at temporal and spatial resolutions unmatched by traditional remote sensing sensors. Optical flow measurement algorithms can estimate velocity by tracking the movement of visible features between frames (e.g., Eltner et al ., 2019; Perks et al ., 2020), which supplemented with pre-existing channel bathymetry data, could enable estimation of river discharge. Currently, satellite video acquired by low earth orbiting sensors offer spatial resolutions (pixel sizes) ranging from 0.9 – 1.2 m at frame rates up to 30 Hz (e.g. SkySat (Bhushan et al ., 2021) and Jilin-1 (European Space Agency, 2022) constellations). Optical satellite video sensors offer unmatched temporal resolution for the observation of dynamic phenomena such as floods. Inference of flow velocities using satellite video has previously been demonstrated by Legleiter and Kinzel (2021), who utilized up to 17 frames of cloud-free satellite video acquired by Planet Labs SkySat constellation to estimate surface flow velocities on the Tanana River in central Alaska. Surface flow velocities matched measurements from a radar gage to within 8.65% and were further assessed using asynchronously acquired acoustic Doppler current profiler (aDcp) velocity data.
Here, we extend the estimation of space-borne velocities to introduce an application of satellite video-based velocities for estimation of discharge. We couple freely available, high-resolution topographic data with velocity estimates derived from satellite video Large-Scale Image Velocimetry (LSPIV) and some critical assumptions regarding channel hydraulics to estimate flood discharge following monsoonal rainfall in Darling River at Tilpa, Australia. Satellite video-derived velocity estimates are assessed using hydraulic model simulations while discharge estimates are compared with in-situ gauging station observations for validation and accuracy assessment, respectively.
Study Area
The River Darling at Tilpa is located within the Murray-Darling basin, with a 502,500 km2 drainage area (Matheson and Thoms, 2018; Murray-Darling Basin Authority, 2010) (Figure 1). The river basin has a strongly episodic climate, with large floods followed by lengthy dry spells due to the influence of the El-Niño-Southern Oscillation (Grimaldi et al ., 2019). Prolonged rainfall across south-eastern Australia from late February to early April 2022 led to a flood event with a 5-year return period (Q = 722 m3s-1). This location is ideally suited to testing our ability to measure river discharge using non-contact, image-based velocity calculation techniques due to the availability of: (i) cloud-free satellite video sensor overpass; (ii) high-resolution LiDAR digital elevation model (DEM) that was acquired when the river bed was dry and; (iii) gauged in-situ discharge observations at Tilpa (Station number 425900; Water New South Wales, https://realtimedata.waternsw.com.au/, last access: 6 March 2023) .