Comparison with TLS

Research Question 3

How does the DEM obtained using the UAS-SfM approach compare with a DEM obtained using TLS?

Chapter 3 – Topography 137 Comparison with TLS at Coledale Beck

Figure 3.27 shows the DEM of difference for the UAS-SfM and TLS DEMs of Coledale Beck.

Whilst this appears to highlight the slight doming present in the UAS-SfM data, its usefulness may be limited by a possible spatial misalignment between the two datasets. This is highlighted in the difference patterns around larger clasts which occurs throughout the scene and therefore is not thought to be related to TLS scan angle. Mean planimetric residual errors are low (<1cm) in both datasets and therefore it is not clear why this apparent offset is observed. Nevertheless, it should be noted that the following observations may have been affected by this spatial misalignment.

A comparison of DEM elevations with the independent topographic validation data suggests that the accuracy and precision of the UAS-SfM DEM is better than the TLS DEM in nearly all parts of the scene (Table 3.11), and the coverage is more continuous and complete. The spatial distribution of TLS DEM error (Figure 3.26) shows a trend towards over-prediction of elevations in nearly all parts of the scene. Given the use of TLS measurements as validation data for UAS-SfM/SfM outputs elsewhere (e.g. Westoby et al., 2012), higher accuracy and precision values might have been expected.

In submerged areas, it is suggested that this overestimation can be largely attributed to the strongly oblique viewing angle of the TLS. This results in a greater angle of refraction at the air-water interface in submerged areas and subsequent over-prediction of channel bed elevations, as also observed in the UAS-SfM data. Refraction correction of TLS surveys in submerged fluvial settings has not been widely used to date, but has recently seen some success in field settings over small areas (Smith et al., 2012, Smith and Vericat 2014), albeit at relatively coarse spatial resolutions (e.g. 1m, Smith and Vericat 2014). However, time, software and data access constraints meant that refraction correction of TLS was not attempted within this study so further comparison is not possible. In deeper parts of the channel (> c. 0.3m), there is a lack of TLS returns altogether resulting in large holes within the DEM. This problem is not observed with the UAS-SfM approach, though it is noted that success of the UAS-SfM approach in these areas is contingent on the presence of clear water and has only been demonstrated up to a maximum depth of c. 0.7m. This is roughly in line with reported maximum depths achieved using digital photogrammetry, bathymetric LiDAR and the spectral-depth method (Table 3.2).

Chapter 3 – Topography 138 The oblique viewing angle of the TLS also means that in areas of dense vegetation the TLS is more likely to ‘see’ the upper elevations of vegetation than an approach looking vertically downwards. Thus larger over-predictions are also observed in these areas when compared to the UAS-SfM results. Under-estimations observed along overhanging banks probably also result from the oblique viewing angle of the TLS, allowing it to ‘see’ beneath the overhang, whilst the independent validation data represents the top of the bank. In contrast, in exposed areas where there is little or no vegetation cover, the TLS survey performs much better. Spatial coverage is improved (i.e. fewer holes in the DEM) and the TLS DEM mean error is not only notably lower than in other parts of the scene, but is slightly better than the equivalent value for the UAS-SfM DEM.

These findings suggest certain advantages of the aerial viewpoint of the UAS-SfM approach over the oblique viewpoint of the TLS. However, they also indicate that where a view of the terrain is not impacted by vegetation or the presence of water, the TLS provides slightly higher resolution and slightly more accurate DEMs. This comparison is not without its trade-offs though. At the current time, TLS systems remain significantly more expensive to acquire (c.

£50,000 for a Leica ScanStation C10, Leica Geosystems Ltd 2014, pers. comm.) than a UAS-SfM set-up equivalent to the one used here (c. £2000 for a DJI F550 UAS including camera, P.

Carbonneau 2014, pers. comm.). Furthermore, in the current example, the collection of TLS data took more than double the fieldwork time when compared to the UAS-SfM approach.

Future work should look to explore and if possible, reduce the spatial offset between datasets, so that stronger conclusions may be drawn from the comparison of UAS-SfM and TLS DEMs for fluvial topography.

3.6.3 Summary

The choice of a method for quantifying topography, within both fluvial and other settings, will be determined by the specific requirements of the intended application in terms of scale and accuracy, as well as the availability of resources, time and funds. However, the research presented within this chapter has demonstrated the potential of a UAS-SfM approach for quantifying the topography of fluvial environments at the mesoscale with hyperspatial resolutions (0.02m). This approach provides a single surveying technique for generating complete, accurate and precise DEMs for exposed areas of the fluvial environment, and within submerged areas for depths up to 0.7m providing the water is clear, there is limited water

Chapter 3 – Topography 139 surface roughness (e.g. white water) and refraction correction is implemented. As such, it represents an alternative to hybrid approaches and has potential as a tool for characterising the topographic heterogeneity at the mesoscale within a ‘riverscape’ style framework (Fausch et al., 2002), albeit within the specified constraints.

Whilst these initial results are encouraging, further targeted research in the following areas would be of benefit:

 The effects of varying the level of overlap between UAS images. This has not been addressed within this chapter, but if the level of overlap could be reduced without negatively impacting on data quality, then flight times could be reduced and/or larger areas covered within each flight. Currently flight times are limited by battery life, so any opportunities to expand the amount of data collected per flight would be beneficial.

 The potential of alternative refraction correction procedures.This study is one of the first to quantitatively assess the ability of the UAS-SfM approach for producing topographic datasets for shallow submerged fluvial environments. The use of a simple refraction correction procedure has given results which seem to reflect those found in similar studies using digital photogrammetry. However, since the SfM technique builds geometry within submerged areas from multiple images taken from different view angles, it is unclear whether this simple correction procedure represents the most appropriate method for reducing the effects of refraction within data produced from PhotoScan. Further research on this topic would require a greater insight into the largely unknown PhotoScan processing algorithms. Testing of different correction techniques in different fluvial settings would also be of benefit here.

 Comparison of refraction corrected UAS-SfM DEMs with refraction corrected TLS DEMs for submerged areas. The results of this research suggest that the TLS DEM suffers strongly from the effects of refraction and lack of returns in deeper waters.

However, a direct comparison of the refraction corrected UAS-SfM DEM with an equivalent TLS product was not possible, but should be conducted in future.

 The ability of repeat surveys for detecting geomorphic change. Very few studies concerning the use of UAS-SfM surveys for change detection in fluvial settings have been published to date (the exception being Flener et al., 2013). The repeat surveys of the River Arrow presented here were not at sufficient time intervals so as to detect

Chapter 3 – Topography 140 any meaningful geomorphic change. Therefore, detailed quantitative assessments are required and should consider both the exposed and submerged environments.

 Use of UAS-SfM point clouds. Recently, some have suggested a move away from the use of the 2.5D DEM for topographic assessments and a new focus on the analysis of the 3D point clouds themselves (e.g. Lague et al., 2013). The kind of 3D point cloud analysis presented by Lague et al., (2013)’s M3C2 algorithm for TLS data could equally be implemented on UAS-SfM point clouds. Such methods hold clear advantages for geomorphic change detection of complex scenes, including both horizontal and vertical surfaces, such as the steep banks observed at Coledale Beck. Time constraints and limitations in computer processing power prevented the use of the M3C2 algorithm within this research, but future work in this area would be of interest.

3.7 Conclusions

This chapter has provided a quantitative assessment of the use of UAS imagery, processed within a SfM-photogrammetry workflow, to generate hyperspatial resolution topographic datasets at the mesoscale for both the exposed and submerged parts of the fluvial environment. This approach was tested at three contrasting field sites and within a flat sports hall setting with a view to exploring the accuracy, precision and repeatability of the method in contrast to existing remote sensing approaches.

Within exposed areas DEM accuracy values are in the range 4-44mm and approaching those typically obtained using TLS for non-vegetated surfaces. DEM quality was found to be poorer within submerged areas, with an apparent scaling of error with increasing water depth. The use of a simple refraction correction procedure improved mean errors in submerged areas by 8-53mm for sites where there was an existing correlation between error and water depth.

Multiple surveys acquired from the River Arrow site gave consistently high quality results, indicating the repeatability of the approach. Variability in DEM accuracy and precision of differing magnitudes is observed both within and between surveys. However, it is suggested that this variability can be attributed primarily to the presence of water and vegetation, the arrangement of GCPs, view angle of the camera and quality of input UAS imagery. For example, the results from Coledale Beck and the sports hall experiments show a slight central doming of the DEMs, which is thought to relate to the self-calibration of the camera lens model within the SfM software combined with the acquisition of imagery predominantly at

Chapter 3 – Topography 141 nadir. Results from the San Pedro River and the sports hall tests demonstrate that a central configuration of GCPs can result in a significant tilting of the DEM and introduce large systematic errors. It is also thought that the dense vegetation coverage at Coledale Beck and poor scene illumination conditions during the August 2013 survey of the River Arrow lead to increased DEM errors. Whilst the errors introduced by these factors are typically small in absolute terms and the UAS-SfM outputs often visually impressive, these results highlight the importance of conducting thorough quantitative error assessments, especially where the approach is to be applied for geomorphic change detection.

The selection of an approach for quantifying fluvial topography (and/or flow depth) will ultimately be determined by the specific requirements of a given application. Typically, a compromise will be required in terms of scale, spatial coverage, accuracy, precision, data acquisition and processing times, and cost. This research has demonstrated that a UAS-SfM approach offers a valid alternative to existing remote sensing methods for quantifying the topography of fluvial environments within both exposed and shallow submerged areas simultaneously, albeit within certain constraints. Comparison of results against TLS datasets collected concurrently, and other remote sensing approaches reported within the wider literature, suggests that the approach is well suited to studies of the mesoscale, where hyperspatial resolution datasets covering up to a few hundred metres of channel are required within a day’s fieldwork. This is discussed further in Chapter 6.

Key areas which would benefit from further targeted research include; the effects of varying GCP densities; the effects of varying the level of image overlap; the potential of alternative refraction correction procedures; direct comparisons with refraction corrected TLS data in submerged environments; testing the ability of repeat surveys for detecting geomorphic change and; exploring the use of UAS-SfM point clouds rather than 2.5D DEMs.

Chapter 3 – Topography 142

3.8 References

Bailly, J-S., Le Coarer, Y., Languille, P., Stigermark, C. and Allouis, T. (2010) Geostatistical estimation of bathymetric LiDAR errors on rivers. Earth Surface Processes and Landforms 35: 1199-1210.

Bailly, J-S., Kinzel, P.J., Allouis, T., Feurer, D. and Le Coarer, Y. (2012) Airborne LiDAR methods applied to riverine environments. In Fluvial Remote Sensing for Science and Management, Carbonneau, P.E. and Piegay, H. (Eds). Wiley-Blackwell, Chichester.

Bangen, S.G., Wheaton, J.M., Bouwes, N., Bouwes, B. and Jordan, C. (2014) A methodological intercomparison of topographic surveying techniques for characterizing wadeable streams and rivers.

Geomorphology 206: 343-361.

Bergeron, N. and Carbonneau, P.E. (2012) Geosalar: Innovative remote sensing methods for spatially continuous mapping of fluvial habitat at riverscape scale. In Fluvial Remote Sensing for Science and Management, Carbonneau, P.E. and Piegay, H. (Eds). Wiley-Blackwell, Chichester.

Butler, J.B., Lane, S.N., Chandler, J.H. and Porfiri, E. (2002) Through-water close range digital photogrammetry in flume and field environments. Photogrammetric Record 17 (99): 419-439.

Carbonneau, P.E., Lane, S.N. and Bergeron, N. (2003) Cost-effective non-metric close-range digital photogrammetry and its application to a study of coarse gravel river beds. International Journal of Remote Sensing 24: 2837-2854.

Carbonneau, P.E., Lane, S.N. and Bergeron, N. (2006) Feature based image processing methods applied to bathymetric measurements from airborne remote sensing in fluvial environments. Earth Surface Processes and Landforms 31: 1413-1423.

Carbonneau, P.E., Fonstad, M.A., Marcus, W.A. and Dugdale, S.J. (2012) Making riverscapes real.

Geomorphology 123: 74-86.

Cavalli, M., Tarolli, P., Marchi, L. and Fontana, G.D. (2008) The effectiveness of airborne LiDAR data in the recognition of channel-bed morphology. Catena 73: 249-260.

Chandler, J.H. (1999) Effective application of automated digital photogrammetry for geomorphological research. Earth Surface Processes and Landforms 24: 51-63.

Chandler, J.H., Buffin-Bélanger, T., Rice, S., Reid, I. and Graham, D.J. (2003) The accuracy of a river bed moulding/casting system and the effectiveness of a low-cost digital camera for recording river bed fabric. Photogrammetric Record 18 (103): 209-223.

Chandler, J.H., Fryer, J.G. and Jack, A. (2005) Metric capabilities of low-cost digital cameras for close range surface measurement. The Photogrammetric Record 20 (109): 12-26.

Costa, J.E., Spicer, K.R., Cheng, R.T., Haeni, P., Melcher, N.B., Thurman, E.M., Plant, W.J. and Keller, W.C.

(2000) Measuring stream discharge by non-contact methods: a proof-of-concept experiment.

Geophysical Research Letters 27 (4): 553-556.

Dugdale, S.J. (2007) An evaluation of imagery from an unmanned aerial vehicle (UAV) for the mapping of intertidal macroalgae on Seal Sands, Tees Estuary, UK. Unpublished MSc thesis, Durham University.

Dunford, R., Michel, K., Gagnage, M., Piegay, H. and Tremelo, M-L. (2009) Potential constraints of unmanned aerial vehicle technology for the characterization of Mediterranean riparian forest.

International Journal of Remote Sensing 30 (19): 4915-4935.

Chapter 3 – Topography 143 Dunning, S.A., Massey, C.I. and Rosser, N.J. (2009) Structural and geomorphological features of landslides in the Bhutan Himalaya derived from Terrestrial Laser Scanning. Geomorphology 103: 17-29.

Dyer, F.J. and Thoms, M.C. (2006) Managing river flows for hydraulic diversity: an example of an upland regulated gravel-bed river. River Research and Applications 22: 257-267.

Eisenbeiss, H., Lambers, K., Sauerbier, M. and Li, Z. (2005) Photogrammetric documentation of an archaeological site (Palpa, Peru) using an autonomous model helicopter. CIPA 2005 XX International Symposium, 26 September-1 October, Torino, Italy.

Entwistle, N.S. and Fuller, I.C. (2009) Terrestrial laser scanning to derive the surface grain size facies character of gravel bars. In Heritage, G.L. and Large, A.R.G. (Eds) Laser Scanning for the Environmental Sciences, Wiley-Blackwell, London.

Fausch, K.D., Torgersen, C.E., Baxter, C.V. and Hiram, L.W. (2002) Landscapes to riverscapes: bridging the gap between research and conservation of stream fishes. BioScience 52 (6): 483-498.

Feurer, D., Bailly, J-S., Puech, C., Le Coarer, Y. and Viau, A.A. (2008) Very-high-resolution mapping of river-immersed topography by remote sensing. Progress in Physical Geography 32(4): 403-419.

Flener, C., Vaaja, M., Jaakkola, A., Krooks, A., Kaartinen, H., Kukko, A., Kasvi, E., Hyyppä, H., Hyyppä, J.

and Alho, P. (2013) Seamless mapping of river channels at high resolution using mobile LiDAR and UAV-photography. Remote Sensing 5: 6382-6407.

Fonstad, M.A., Dietrich, J.T., Courville, B.C., Jensen, J.L. and Carbonneau, P.E. (2013) Topographic structure from motion: a new development in photogrammetric measurement. Earth Surface Processes and Landforms 38 (4): 421-430.

Frissell, C.A., Liss, W.J., Warren, C.E. and Hurley, M.D. (1986) A hierarchical framework for stream habitat classification: viewing streams in a watershed context. Environmental Management 10 (2): 199-214.

Frothingham, K.M., Rhoads, B.L. and Herricks, E.E. (2002) A multiscale conceptual framework for integrated ecogeomorphological research to support stream naturalisation in the agricultural Midwest.

Environmental Management 29 (1): 16-33.

Fryer, J.G. (1983) A simple system for photogrammetric mapping in shallow water. Photogrammetric Record 11 (62): 203-208.

Fryer, J.G. and Kniest, H.T. (1985) Errors in depth determination caused by waves in through-water photogrammetry. Photogrammetric Record 11 (66): 745-753.

Harwin, S. and Lucieer, A. (2012) Assessing the accuracy of georeferenced point clouds produced via multi-view stereopsis from unmanned aerial vehicle (UAV) imagery. Remote Sensing 4: 1573-1599.

Hetherington, D., Heritage, G.L. and Milan, D.J. (2005) Reach scale sub-bar dynamics elucidated through oblique lidar survey. in International Association of Hydrological Scientists Red Book Publication, IAHS Publication 291: 278-284.

Heritage, G. and Hetherington, D. (2007) Towards a protocol for laser scanning in fluvial geomorphology.

Earth Surface Processes and Landforms 32: 66-74.

Heritage, G.L. and Milan, D.J. (2009) Terrestrial laser scanning of grain roughness in a gravel-bed river.

Geomorphology 113: 4-11.

Chapter 3 – Topography 144 Hicks, D.M. (2012) Remotely sensed topographic change in gravel riverbeds with flowing channels. In Gravel-bed Rivers: Processes, Tools, Environments, Church, M., Biron, P. and Roy, A. (Eds) Wiley-Blackwell, Chichester.

James, M.R. and Robson, S. (2012) Straightforward reconstruction of 3D surfaces and topography with a camera: accuracy and geoscience application. Journal of Geophysical Research 117, F03017, doi:

10.1029/2011JF002289.

James, M.R. and Robson, S. (2014) Mitigating systematic error in topographic models derived from UAV and ground-based image networks. Earth Surface Processes and Landforms 39: 1413-1420.

Javernick, L., Brasington, J. and Caruso, B. (2014) Modelling the topography of shallow braided rivers using Structure-from-Motion photogrammetry. Geomorphology 213: 166-182.

Jerlov, N.G. (1976) Marine Optics. Elsevier, Amsterdam.

Kinzel, P.J., Wright, C.W., Nelson, J.M. and Burman, A.R. (2007) Evaluation of an experimental LiDAR for surveying a shallow, sand-bedded river. Journal of Hydraulic Engineering 133 (7): 838-842.

Kinzel, P.J., Legleiter, C.J. and Nelson, J.M. (2013) Mapping river bathymetry with a small footprint green lidar: applications and challenges. Journal of the American Water Resources Association 49 (1): 183-204.

Kraus, K. and Pfeifer, N. (2001) Advanced DTM generation from LIDAR data. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences XXXIV 3/W4 III Land Surface Mapping and Characterisation using Laser Altimetry Workshop, 22-24 October 2001, Annapolis, Maryland, USA.

Laliberté, A.S. and Rango, A. (2009) Texture and scale in object-based analysis of subdecimeter resolution unmanned aerial vehicle (UAV) imagery. IEEE Transactions on Geoscience and Remote Sensing 47 (3): 761-770.

Lague, D., Brodu, N. and Leroux, J. (2013) Accurate 3D comparison of complex topography with terrestrial laser scanner: Application to the Rangitikei canyon (N-Z). ISPRS Journal of Photogrammetry and Remote Sensing 82: 10-26.

Lane, S.N. (2000) The measurement of river channel morphology using digital photogrammetry.

Photogrammetric Record 16 (96): 937-957.

Lane, S.N., Richards, K.S. and Chandler, J.H. (1993) Developments in photogrammetry; the geomorphological potential. Progress in Physical Geography 17 (3): 306-328.

Lane, S.N., Chandler, J.H. and Richards, K.S. (1994) Developments in monitoring and modelling small-scale river bed topography. Earth Surface Processes and Landforms 19: 349-368.

Lane, S.N., James, T.D. and Crowell, M.D. (2000) Application of digital photogrammetry to complex topography for geomorphological research. Photogrammetric Record 16 (95): 793-821.

Lane, S.N. and Carbonneau, P.E. (2007) High resolution remote sensing for understanding instream habitat’ in Hydroecology and Ecohydrology, Wood, P.J., Hannah, D.M. and Sadler, J.P. (Eds), John Wiley and Sons, Chichester.

Lane, S.N., Widdison, P.E., Thomas, R.E., Ashworth, P.J., Best, J.L., Lunt, I.A., Sambrook Smith, G.H. and Simpson, C.J., (2010) Quantification of braided river channel change using archival digital image analysis.

Earth Surface Processes and Landforms 35(8), 971-985.

Chapter 3 – Topography 145 Legleiter, C.J. (2012) Remote measurement of river morphology via fusion of LiDAR topography and spectrally based bathymetry. Earth Surface Processes and Landforms 37: 499-518.

Legleiter, C.J. (2013) Mapping river depth from publicly available aerial images. River Research and Applications 29 (6): 760-780.

Legleiter, C.J., Roberts, D.A., Marcus, W.A. and Fonstad, M.A. (2004) Passive optical remote sensing of river channel morphology and instream habitat: physical basis and feasibility. Remote Sensing of the Environment 93: 493-510.

Legleiter, C.J., Roberts, D.A. and Lawrence, R.L. (2009) Spectrally based remote sensing of river bathymetry. Earth Surface Processes and Landforms 34: 1039-1059.

Lejot, J., Delacourt, C., Piegay, H., Fournier, T., Tremelo, M-L. and Allemand, P. (2007) Very high spatial resolution imagery for channel bathymetry and topography from an unmanned mapping controlled platform. Earth Surface Processes and Landforms 32: 1705-1725.

Lowe, D.G. (2004) Distinctive image features from scale-invariant keypoints. International Journal of Computer Vision 60: 91-110.

MacVicar, B.J., Piégay, H., Anderson, A., Oberlin, C., Comiti, F. and Pecorari, E. (2009) Quantifying the temporal dynamics of wood in large rivers: new techniques for wood tracking, monitoring, surveying,

MacVicar, B.J., Piégay, H., Anderson, A., Oberlin, C., Comiti, F. and Pecorari, E. (2009) Quantifying the temporal dynamics of wood in large rivers: new techniques for wood tracking, monitoring, surveying,