Unmanned aerial vehicles (UAVs) have the potential to capture information
about the earth's surface in dangerous and previously inaccessible locations.
Through image acquisition of flash flood events and subsequent object-based
analysis, highly dynamic and oft-immeasurable hydraulic phenomena may be
quantified at previously unattainable spatial and temporal resolutions. The
potential for this approach to provide valuable information about the
hydraulic conditions present during dynamic, high-energy flash floods has
until now not been explored. In this paper we adopt a novel approach,
utilizing the Kande–Lucas–Tomasi (KLT) algorithm to track features present
on the water surface which are related to the free-surface velocity.
Following the successful tracking of features, a method analogous to the
vector correction method has enabled accurate geometric rectification of
velocity vectors. Uncertainties associated with the rectification process
induced by unsteady camera movements are subsequently explored.
Geo-registration errors are relatively stable and occur as a result of
persistent residual distortion effects following image correction. The
apparent ground movement of immobile control points between measurement
intervals ranges from 0.05 to 0.13 m. The application of this approach to
assess the hydraulic conditions present in the Alyth Burn, Scotland, during a
The occurrence of flash flooding from intense rainfall in western Europe is predicted to increase throughout the first half of the 21st century (Beniston, 2009; Rojas et al., 2012). These events pose severe risks to society, transform communities, and under extreme conditions can permanently alter the state of the river system (Doocy et al., 2013; Milner et al., 2013). Flash floods in fluvial systems pose high risks to communities, especially when they occur in small, upland catchments where orographic effects can enhance precipitation intensity with runoff being concentrated rapidly along narrow and steep flow pathways (Bracken and Croke, 2007; Sangati et al., 2009; Garambois et al., 2014). Despite a substantial body of work on physical flood processes observed in research catchments (e.g. Quinn and Beven, 1993; Soulsby et al., 2000; Mayes et al., 2006), there is currently a paucity of data describing the antecedent and concurrent processes associated with extreme flash flood events. This is mainly due to conventional monitoring networks often failing to adequately sample small, responsive catchments (Borga et al., 2008; Gaume and Borga, 2008; Soulsby et al., 2008; Braud et al., 2014). Measurement and monitoring of these events is therefore largely responsive rather than active, opportunistic rather than strategic, and hindered by practical difficulties (Borga et al., 2008; Tauro et al., 2015b). Making observations of peak flood discharge in real time remains a significant practical challenge.
Given current operational constraints, favourable sources of process data during flash floods in ungauged catchments often rely on post hoc analyses of air- and space-borne earth observation sensors (e.g. visible, near-infrared, and multispectral imaging and synthetic aperture radar). Increasing availability of these remotely sensed data has furthered our understanding of floodplain inundation processes (e.g. Wright et al., 2008); enabled hydraulic properties such as roughness (Simeonov et al., 2013), river stage, and discharge (Liu et al., 2015) to be successfully modelled; provided justification for the incorporation of spatially and temporally varied roughness values (Mason et al., 2003; Schumann et al., 2007); and enabled calibration and validation of hydrodynamic models (e.g. Martinis et al., 2009; Refice et al., 2014). Various contributions have been enabled by the fortuitous availability of archived satellite and aerial records (e.g. Chen and Mied, 2013; Kääb and Leprince, 2014). However, the highly transient temporal and spatial domains of flash floods, combined with the significant lead times required to mobilize monitoring resources, have up until now limited the use of satellite and aerial records to larger, more slowly responding catchments (e.g. Fujita and Kunita, 2011; Wong et al., 2015).
The widespread availability of unmanned aerial vehicles (UAVs) has, in recent years, increased our ability to monitor and quantify higher magnitude, and lower frequency environmental phenomena (e.g. Niethammer et al., 2012; Ryan et al., 2015), whilst at the same time reducing operational costs of traditional environmental monitoring (Fekete et al., 2015). The potential for the use of UAVs for non-contact flow measurement has been identified (Kääb and Leprince, 2014), leading to proof-of-concept studies utilizing UAVs for monitoring of low-flow conditions (e.g. Pagano et al., 2014; Detert and Weitbrecht, 2015; Patalano et al., 2015; Tauro et al., 2015a, b). However, the potential for this approach to provide valuable information about the hydraulic conditions present during dynamic, high-energy flash floods has yet to be realized.
Image-based non-contact methods of flow estimation utilize algorithms (e.g. Large-Scale Particle Tracking Velocimetry, LSPTV, and Large-Scale Particle Image Velocimetry, LSPIV) designed to track optically visible features of the free surface to determine the rate of fluid flow in artificial or natural open channels (Jodeau et al., 2008; Kim et al., 2008; Le Coz et al., 2010; Sun et al., 2010; Dramais et al., 2011; Puleo et al., 2012; Pentari et al., 2014; Le Boursicaud et al., 2015). The rate at which naturally occurring features (e.g. foam, seeds, woody debris, and turbulent structures) or artificially introduced tracers (e.g. Ecofoam chips, fluorescent micro-spheres) are displaced downstream can be used to estimate the free-surface velocity, which may be related to depth-averaged flow characteristics (e.g. Jodeau et al., 2008; Dramais et al., 2011; Fujita and Kunita, 2011; Simeonov et al., 2013; Le Boursicaud et al., 2015). Conceptually, terrestrial and airborne tracking of surface water features is similar; however, the uncertainties associated with rectification of captured images to account for perspective, radial, and tangential distortions are compounded when using a UAV for image acquisition. This is due to unsteady camera movement, which must be accounted for if accurate geometric rectification of velocity vectors or oblique images is to be achieved (Kantoush et al., 2008; Kim et al., 2008). A second source of uncertainty is introduced in situations where low seeding densities prevail, resulting in a lack of stable and identifiable surface features (Lewis and Rhoads, 2015). However, in the case of flash floods, coherent flow structures at the free surface and the presence of washed-in floating material may produce favourable seeding conditions (Jodeau et al., 2008; Dramais et al., 2011).
A schematic of the proposed methodology for tracking surface water features from UAVs and their conversion to velocities.
This paper presents a novel methodology for the derivation of key hydraulic data during flash floods using imagery captured by a low-cost, commercially available UAV platform. Our approach overcomes uncertainties associated with image rectification, transformation, and feature tracking to determine river surface velocity during flash floods. Our approach yields fundamental process data, invaluable for flash flood reconstruction in ungauged river catchments. The adoption of this technique has the potential to significantly advance our understanding of high-flow stage processes during flash floods.
The materials presented in the following section describe the entire workflow for the extraction of surface water velocities from a UAV through the utilization of image-based non-contact methods. This method is organized in five sub-sections, which are presented sequentially: (i) primary data collection; (ii) development of an initial camera model, and (iii) updated camera models for projective transformations; (iv) assessment of transformation accuracy and apparent movement of GCPs; and, finally, (v) surface velocity calculation. A schematic overview of this method is provided in Fig. 1, wherein each heading corresponds to the homonymous section within the main text.
On 17 July 2015, the Alyth Burn, Perthshire, Scotland (324600, 748600 OS
BNG), breached its banks as a result of a prolonged period of rainfall over
the catchment. While rainfall totals were not in themselves extreme (41 mm
over a 6 h period), the prolonged nature of the precipitation event, coupled
with the particular catchment configuration upstream of the town, resulted in
over 70 properties being flooded and four footbridges in the town centre
being destroyed (Perth and Kinross Council et al., 2015). During this flood
event, a Phantom Vision 2 UAV equipped with a FC200 camera unit was deployed
over the Alyth Burn in manual flight mode by a member of the public at
Ground control points (GCPs) for the area of interest were required to
convert the image (px) co-ordinates into geographical co-ordinates (OS BNG
m). The deployment of a Leica MS50 multi-station shortly after the flood
event enabled the generation of a detailed point cloud with an average point
spacing of < 0.002 m from which GCPs could be accurately identified
(Fig. 1, Sect. 2.1). These GCPs represented immobile objects that were
present during the recording, and which persisted following the clean-up
operation (e.g. lamp-posts and wall corners). Individual point clouds were
joined using CloudCompare (2015), resulting in an
internal error (RMS) of 0.04 m. This point cloud was rectified to real-world
co-ordinates through comparison with control point measurements (
Due to the lack of available navigation data for the UAV, its starting
position was modelled using an a priori assumption about its approximate
location [
Following generation of the master camera model for the first frame of the
video, an updated camera model solution based on updated GCP co-ordinates was
generated for each subsequent frame (Messerli and Grinsted, 2015). This
enabled UAV movement and changes in view direction to be accounted for. The
updated camera model was obtained by randomly generating 1000 new camera
positions proximal to the co-ordinates of the optimized camera model for the
previous frame (
Every
As with the GCPs, between the
Optimized parameters of the distorted camera models.
Using the 20 000 potential solutions, the optimized master camera model was
selected based on the minimum square projection error of the GCPs (RMSE –
root-mean-square error). The minimum RMSE of the 20 000 solutions was
11.4 px (
Box plots showing how projection residuals
Analysis indicates that the precision of the geometric projection
Box plots showing how the apparent movement
Whilst accurate geometric projection is essential for observed velocities to
be assigned an appropriate spatial reference, the precision of the
transformation over time is of greatest importance. Unacceptable apparent
ground velocities as a result of unstable transformation over time would
undermine the value of tracking surface features. This error
Following the analysis of the 5.2 s video, and the filtering of features
tracked from within inaccurately projected regions of the image, a total of
2644 velocity vectors were compiled within a 624 m
Images showing
Application of feature tracking in open channels is dominated by methods operating in the Eulerian frame of reference (e.g. LSPIV). These methods have been widely successful in the characterization of instantaneous and time-averaged velocities for the determination of flood discharges, with deviations from acoustically derived measurements of < 10 % (Jodeau et al., 2008; Muste et al., 2008; Dramais et al., 2011). Measurements made in the Lagrangian frame of reference (e.g. LSPTV), where the paths of individual particles are assessed, have been less widely adopted for monitoring high-magnitude events. This is despite LSPTV replicating hydraulics accurately with improved performance close to boundaries and in areas experiencing high velocity gradients (Admiraal et al., 2004). Enhanced spatial resolution of measurements may also be possible with lower seeding densities (Detert and Weitbrecht, 2015). Our implementation of the KLT algorithm has demonstrated its potential to generate large volumes of temporally consistent data at a distance of up to 50 m. However, feature tracking from non-stationary platforms poses additional challenges in accounting for errors related to sensor movement and orientation. These challenges, which must be addressed for this approach to be beneficial for monitoring flood flows, are discussed in the following sections.
Transformation from pixel to world co-ordinates is one of the greatest
challenges in generating accurate velocity estimates, even when measurements
are conducted in controlled conditions from sensors of known, fixed locations
(Lewis and Rhoads, 2015). Specific error associated with rectification can be
controlled by ensuring the camera lens is (i) orthogonal to the water surface
(e.g. Lewis and Rhoads, 2015); (ii) corrected for distortion (e.g. Le
Boursicaud et al., 2015); and (iii) accurately calibrated using stable GCPs
throughout the field-of-view (e.g. Dramais et al., 2011). Unfortunately it is
not always possible to maintain the camera lens orthogonal to the water
surface whilst capturing flow processes at the scale of interest, which often
necessitates oblique image capture. Such oblique image capture may pose
methodological difficulties due to far-field objects being poorly resolved
relative to those in near-field. Secondly, lens distortion must be removed
prior to the implementation of traditional plan-to-plan perspective
projection (Le Boursicaud et al., 2015). This can be achieved based on the
manufacturer's specifications (e.g. Detert and Weitbrecht, 2015), or through
manual calibration (e.g. Tauro et al., 2015a); however, residual distortion
may persist close to image boundaries. Finally, following internal camera
calibration, the success of the transformation depends on the
three-dimensional distribution of GCPs. Distribution of at least four GCPs is
required for a two-dimensional transformation (Fujita et al., 1998; Fujita
and Kunita, 2011), or a minimum of six GCPs distributed across the region of
interest for a three-dimensional plan-to-plan perspective projection (Jodeau
et al., 2008; Muste et al., 2008). For accurate transformation, elevation
errors can be minimized by ensuring GCPs are similar to or located parallel
to the water surface elevation (Jodeau et al., 2008; Fujita and Kunita,
2011). However, an implicit assumption of this approach is that the planar
free surface is horizontal and that free-surface undulations are negligible
across the frame. Due to the often negligible water surface slopes across the
area of interest, errors are typically assumed to be insignificant (Hauet et
al., 2008), with previous research indicating that water level errors
of
In addition to oblique image capture, camera motion can greatly diminish the precision of any calibration and transformation process. In the case of monitoring fluvial flash floods from UAV platforms, camera motion is inevitable (Tauro et al., 2015a, b), and this movement should be corrected for on a frame-by-frame basis. This may be achieved through the utilization of on-board GPS systems (e.g. Bolognesi et al., 2016) or fixed reference points (e.g. Lewis and Rhoads, 2015). In the approach reported on here, we adopt a methodology to account for these uncertainties and their impacts on subsequent velocity measurements whereby fixed control points are manually selected and automatically tracked between frames using the KLT algorithm. Automatic tracking of GCPs is enabled by the distinct image textures of the water surface and the built environment, enabling the precision of the rectification process to be quantified and uncertainty in velocity measurements to be established. Whilst this procedure requires some supervision, in future deployments, purpose-built GCPs will be installed across the area of interest with distinct optical characteristics so that (semi-)automatic registration would be possible. However, in areas where naturally existing GCP features do not exist, or where installation of purpose-built GCPs would be problematic, a different approach would be required. Therefore, future research should seek to assess the potential for on-board GPS systems, ranging tools (e.g. lasers), and calibrated cameras to enable UAVs to be utilized. This will also open up the possibility of real-time capture of hydraulic properties of flow.
Due to the responsive nature of this survey of the July 2015 Alyth flood event, the distribution of GCPs was not pre-determined, so despite a total of 15 linear structures within the urban landscape that intersected the water surface being identified as GCPs, spatial coverage is incomplete and availability is temporally variable. While rapid response deployment of UAVs during floods may therefore introduce errors in the projection that would otherwise be accounted for in planned deployments, the majority of surveys at high discharge will naturally be “unplanned” and the result of rapid field deployment. Despite this, and the technical challenges of flying surveys during periods of heavy rainfall associated with floods, the relatively stable transformations achieved throughout the duration of the July 2015 Alyth video presented here demonstrate the utility of the approach.
UAVs have the potential to capture information about dynamics at the earth's surface in hazardous and previously inaccessible locations. Highly transient and oft-immeasurable hydraulic phenomena may be quantified at previously unattainable spatial and temporal resolutions using image acquisition of flash floods and subsequent object-based analysis. The potential for this approach to provide valuable information about the hydraulic conditions present during dynamic, high-energy flash floods has until now not been explored.
This paper adopts a novel approach, utilizing the KLT algorithm to track features present on the water surface which are related to the free-surface velocity. Following the successful tracking of features, a method analogous to the vector correction method has enabled accurate geometric rectification of velocity vectors. We subsequently explored uncertainties associated with the rectification process induced by unsteady camera movements. The maximum geolocation error is 1.0 m, which provides an indication of the minimum spatial scale over which measurements should be averaged and reported. Significant spatial variability in geo-registration error values is observed, with median individual GCP error values ranging from 0.27 to 1.0 m. Our analysis eliminates the potential for significant errors being a function of reduced pixel density per unit area as GCP distance increases. Geo-registration errors are relatively stable and occur as a result of persistent residual distortion effects following image correction, especially close to the image boundaries, due to the specified transformation parameters being sub-optimal. Future approaches should seek to use a camera with minimal lens distortion, for which the internal properties of the camera are calibrated, rather than adopting manufacturer lens specifications. The apparent ground velocities of the 15 GCPs range from 0.05 to 0.13 m, with no apparent relationship between the distance of the GCP and observed ground velocity. These findings illustrate the relative spatial and temporal stability of the geometric transformation.
The application of this approach to assess the hydraulic conditions present
in the Alyth Burn during a
Comprehensive and innovative monitoring programmes (e.g. Ip et al., 2006; Quevauviller et al., 2012; Smith et al., 2014) have previously improved understanding of transient, rate limiting processes and catchment dynamics during extreme flash floods (Zanon et al., 2010), Similarly, we anticipate that this methodology will be of great use in quantifying highly transient flood flows within ungauged rivers across a wide range of fluvial environments.
Datasets utilized in the production of this research article are available for download in the Supplement.
This work was funded by NERC grant NE/K008781/1 “Susceptibility of
catchments to INTense RAinfall and flooding (SINATRA)”. The authors wish to
thank Flavia Tauro, an anonymous reviewer, and the handling Editor for their
detailed comments which greatly improved the quality of this paper. Thanks
also to Angus Forbes of Angus Forbes Photography