drones and digital photogrammetry: from classifications to …eprints.worc.ac.uk/5481/1/woodget et...

Focus Article

Drones and digitalphotogrammetry: fromclassifications to continuums formonitoring river habitat andhydromorphologyAmy S. Woodget,1* Robbie Austrums,2 Ian P. Maddock1 and Evelyn Habit3

Recently, we have gained the opportunity to obtain very high-resolution imageryand topographic data of rivers using drones and novel digital photogrammetricprocessing techniques. The high-resolution outputs from this method areunprecedented, and provide the opportunity to move beyond river habitat classi-fication systems, and work directly with spatially explicit continuums of data.Traditionally, classification systems have formed the backbone of physical riverhabitat monitoring for their ease of use, rapidity, cost efficiency, and direct com-parability. Yet such classifications fail to characterize the detailed heterogeneityof habitat, especially those features which are small or marginal. Drones anddigital photogrammetry now provide an alternative approach for monitoringriver habitat and hydromorphology, which we review here using two case stud-ies. First, we demonstrate the classification of river habitat using drone imageryacquired in 2012 of a 120 m section of the San Pedro River in Chile, which was atthe technological limits of what could be achieved at that time. Second, wereview how continuums of data can be acquired, using drone imagery acquiredin 2016 from the River Teme in Herefordshire, England. We investigate the preci-sion and accuracy of these data continuums, highlight key current challenges,and review current best practices of data collection, processing, and manage-ment. We encourage further quantitative testing and field applications. If currentdifficulties can be overcome, these continuums of geomorphic and hydraulicinformation hold great potential for providing new opportunities for understand-ing river systems to the benefit of both river science and management. 2017 TheAuthors. WIREs Water published by Wiley Periodicals, Inc.

How to cite this article:WIREs Water 2017, e1222. doi: 10.1002/wat2.1222

INTRODUCTION

Monitoring the spatial and temporal variation inphysical river parameters is important forunderstanding and improving habitat quality and dis-tribution, especially with respect to the potentialimpacts of climate change.13 Remote sensing basedmethods have long played a role in surveying and

*Correspondence to: [email protected] of Science and Environment, University of Worcester,Worcester, UK2Independent Geospatial Consultant, Worcester, UK3Faculty of Environmental Science and Eula Centre, Universidadde Concepcin, Concepcin, Chile

Conflict of interest: The authors have declared no conflicts of inter-est for this article.

1 of 20 2017 The Authors. WIREs Water published by Wiley Periodicals, Inc.This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution andreproduction in any medium, provided the original work is properly cited and is not used for commercial purposes.

http://creativecommons.org/licenses/by-nc/4.0/

monitoring physical river habitat and hydromorphol-ogy.4,5 A growing body of literature demonstratesthe use of digital photogrammetry6,7 and spectral-depth correlations810 for quantifying fluvial topog-raphy and flow depth, the computation of image tex-tural variables and roughness of terrestrial laserscanner point clouds for quantifying fluvial substratesize,1115 and the use of multispectral imagery formapping hydrogeomorphic units.16 These develop-ments have made important contributions to ourabilities to map and measure physical river habitatparameters. However, few of these approaches arecapable of quantifying a number of physical habitatparameters simultaneously (e.g., topography, waterdepth, substrate size, and hydraulic variables) using asingle dataset at the spatial resolutions most appro-priate for habitat evaluation.

Recent developments in the capabilities andavailability of drones [otherwise known as smallunmanned aerial systems (sUAS), unmanned aerialvehicles (UAVs), or remotely piloted aircraft systems(RPAS)], alongside parallel advances in digital photo-grammetry mean that a relatively new, alternativeapproach for the remote sensing of rivers is now pos-sible. Interest in drones for image acquisition andassociated structure-from-motion (SfM) digital pho-togrammetric image processing has seen a dramaticexpansion over the last few years, within both aca-demic and commercial spheres. Web of Sciencereturns only eight articles featuring the words UAVand river (topic field) between 1950 and 2010, a fig-ure which increases almost eightfold for the period2011 to present (search conducted September9, 2016). Published studies suggest that this novelapproach is capable of rapid, flexible, bespoke dataacquisition, which is relatively inexpensive and capa-ble of exceptionally high spatial resolutions.17,18 Assuch, the combined use of drones and digital photo-grammetry has been heralded as the route to democ-ratization of data acquisition within the geosciences,and in this capacity might enable the quantificationof physical river habitat parameters at the mesoscale.

Within this article, we review the importance ofmonitoring physical river habitat before describingthe key elements of the drone and SfM-basedphotogrammetry method. We present two case stud-ies to demonstrate the contributions made usingdrones and SfM in fluvial settings to date. The firstrepresents an early example for replicating traditionalhabitat classification schemes, which was conductedwith a view to characterize habitat rapidly inresponse to planned channel engineering works on alarge river in Chile. The second case study provides amore recent example of the continuums of physical

habitat data which were obtained using the drone-SfM method on a highly mobile lowland river in theUK, as informed by best practice at the time of writ-ing. Here, we also provide an assessment of dataaccuracy and precision. We consider how technologi-cal and methodological developments might permitus to move away from classification schemes andtoward these data continuums for monitoring andmapping physical habitat in a way which has thepotential to precipitate a fundamental shift in ourunderstanding of river science and our managementof river systems.

The Importance of Physical River HabitatMonitoringRiver systems form important habitats for a range offauna and flora. To monitor and assess habitat qual-ity, we often collect data on the water quality andenergy budget, as well as the physical conditions ofthe river i.e., the geomorphology and hydrology orhydromorphology.1921 These river habitat para-meters vary in space and in time with changes inunderlying geology and tectonics, climate and weatherpatterns and as a result of anthropogenic modifica-tions, including channel engineering works(i.e., modification to the geomorphology) and/or regu-lation of river discharge (i.e., modification to thehydrology). Such alterations modify the basic physicaltemplate of rivers and therefore modify the qualityand availability of fluvial habitat.1,21,22 Monitoringthis spatial and temporal variation in physical riverparameters is important therefore for understandingand improving habitat quality and distribution, espe-cially in light of the potential impacts of climatechange.13 Physical habitat monitoring within riversforms a key aspect of the European Unions WaterFramework Directive and a variety of other applica-tions within both science and management, includingriver restoration works, fisheries development, andenvironmental flow assessments.19,23,24

The methods by which we retrieve physical hab-itat information vary widely and are often influencedby our prevailing conceptualization of the river as asystem. The river continuum concept (RCC) advo-cates a smoothly changing river system in which thephysical environment and its inhabitant biota varygradually and predictably downstream.25 Acceptanceof this concept has permitted broad scale mapping orquantification of physical habitat parameters at dis-crete sampling locations, and the subsequent assump-tion that interpolation between these locationsprovides an adequate representation of the spatial dis-tribution of physical conditions within the system.4

Focus Article wires.wiley.com/water

2 of 20 2017 The Authors. WIREs Water published by Wiley Periodicals, Inc.

Alternatively, Ref 26 proposed a spatially nested, hier-archical classification framework for organizing ourunderstanding of the spatial and temporal variationwithin river systems (Figure 1). This frameworkallows for greater spatial heterogeneity within riversystems than the RCC, and its hierarchical structurefacilitates integration of different data types at differ-ent resolutions, assisting scientists and managers tochoose the level most applicable to their work.26

More recently, some of these established theories havebeen questioned,4,27,28 by those who argue that established research and management concepts oftenfail to fully recognise the crucial roles played by habi-tat heterogeneity (Ref 27, p. 36). Instead, a river-scape type approach has been proposed, which shiftsour understanding of rivers from gradually changinglongitudinal elements to those characterized by highlevels of spatial and temporal diversity. Such a changein our conceptualization of rivers precipitates a needfor different ways in which physical habitat can bemeasured and classified. Broad scale mapping and dis-crete point or transect sampling are no longer suffi-cient for characterizing the detailed spatial structureof fluvial habitat heterogeneity.21

Ref 4 suggests that a riverscape approachshould involve characterization of physical habitat ina spatially continuous (rather than sampling pointsor lines) and spatially explicit way (i.e., fully geore-ferenced to absolute or relative co-ordinate systems),to allow precise quantification of physical habitatvariables and improve our understanding of the size,distribution, and connectivity of different habitat

types.20 Such an approach should facilitate the inte-gration of multiple spatial datasets, consider spatialscale and context, enable the establishment of base-line conditions, and provide an opportunity forexploring temporal variability within physical habitatparameters, which is currently rarely considered. Ref29 (p. 199) argues that Real contributions fromresearch to sustainable management of river sys-tems need to match a sophistication of conceptswith a direct practicality (without which applicationsare unlikely). Thus, an ideal approach for quantify-ing physical habitat parameters should also be practi-cal, logistically feasible, cost effective as well asobjective and repeatable. The challenges associatedwith an approach which can fulfill all these criteriaare not to be underestimated, but it is suggested thatstudies at the mesohabitat scale have potential toact as a fulcrum between scientific detail and applieduniversality (Ref 29, p. 199). We typically define themesoscale as the active channel width, and channellengths, which are small multiples of channelwidth.29 Studies over these extents comprise bothmeso- and microscale features and typically requiredata of hyperspatial resolutions (

reconnaissance and target acquisition,30,31 since the1990s new application opportunities have opened upwithin civilian research and the commercial sector.This has included application in fields such asarcheology,32 landslide and hazard mapping,33 map-ping of glacial landforms,34 monitoring of crops andvegetation,3538 geomorphological mapping,39 and inriver science.17,18,40

A drone system usually consists of an aircraftplatform mounted with one or more sensors com-bined with a ground-based control station fromwhere it is operated. The sensor typically comprisesan inexpensive, nonmetric, consumer-grade digitalcamera, from which small format, overlappingimages are acquired for photogrammetric purposes.However, an increasing variety of drone-mountablesensing systems, including miniaturized laser scannersand multispectral cameras, are now becoming availa-ble.41 Drone platforms themselves are classified intofixed- and rotary-winged systems and by weight andendurance capabilities.31,42 These classifications varybetween countries, but are important in determiningthe legal regulations for flying and data acquisitionfor both commercial and research purposes. Themarket provision of drones for civilian applicationsis expanding very rapidly, as is platform and sensortechnology. Consequently, drones and the reviewswhich describe them quickly become out-of-date.Within this review, we consider primarily the use ofRGB imagery collected from small drones (lessthan 20 kg).

Digital Photogrammetry (Structure-from-Motion)In parallel with the rapid development of drones forcivilian research, advances in computer vision andimage analysis have led to an increased availability ofsoftware packages offering image processing chainscapable of producing both orthophotos and raster-ized digital elevation models (DEMs) from droneimagery. SfM provides an automated method formodeling the relative 3D geometry of a scene byimage matching a series of overlapping 2D images,which may then be georeferenced to map co-ordinates.

In contrast to traditional photogrammetry, SfMalgorithms allow the reconstruction of the 3D scenewithout prior knowledge of camera positions or theuse of ground control points (GCPs),4346 althoughthe latter are usually required for scaling and geore-ferencing. The process is multistage, starting with theidentification of common features across sets ofphotographs that are overlapping and convergent

(i.e., which show the same subject from differentangles). The scale invariant feature transform func-tion developed by Ref 47 is one of the image match-ing algorithms frequently used as part of the SfMprocess and is a powerful method capable of recog-nizing conjugate (matching) points in overlappingimages, regardless of changes in image scale, viewangle, or orientation.46,48 This is performed usingpatterns of image brightness and color gradients(i.e., variations in image texture) which can be identi-fied at various different scales. The kernel- or area-based approaches used in traditionalphotogrammetry require constant image resolutionand the acquisition of imagery at nadir, which can besignificantly more difficult to obtain using dronesthan manned aircraft.44,46,48 As a result, the SfMphotogrammetric process, which actually providesoptimal results using imagery which has beenacquired off-nadir, convergent and from multiplealtitudes,49,50 represents a dramatic step forward forprocessing drone imagery. Recent research suggeststhat even smartphone imagery can be used success-fully within an SfM workflow.51

Following the identification of prominentmatching points between the convergent, overlappingimages, collinearity equations are computed. Theseequations describe the geometric relationshipbetween the co-ordinates of each point in the two-dimensional plane of the sensor/camera, to the three-dimensional co-ordinates of the same point on thescene/object below. Successive bundle adjustmentsare then performed on these equations. This processuses a nonlinear least-squares optimization approachto estimate simultaneously the camera parameters,relative camera positions, and 3D scene geometrybased on all input images.52 The result is an align-ment of all the input images and a 3D point cloudwhich describes the scene geometry as a set of sparsedata points at the locations of matched image fea-tures.53,54 During this phase, automated camera lenscalibrations are conducted to reduce the impact oflens distortion on the model output. The next steptakes the information gained by the bundle adjust-ment to subject the original images to multiviewstereo techniques.55 This is a computationally inten-sive process that adds many more 3D coordinates tothe scene geometry model to create a dense pointcloud.

Finally, the model is georeferenced from anarbitrary co-ordinate system to an absolute co-ordinate system so that it can be used for quantitativemeasurement and comparison with other spatialdatasets.44,48,54,56 Nondirect georeferencing typicallycomprises the use of GCPs, at least three of which



are required to perform a three-dimensional, seven-parameter transformation of the model.45,56 Thistransformation is linear and any errors are carriedthrough to the final georeferenced output.46 Oncetransformed, the model can be exported as (1) apoint cloud, (2) a rasterized DEM, and (3) texturedusing the original imagery to produce an orthophoto.Textured triangle-mesh outputs are also possible,though these are typically used for visualizationrather than analytical purposes. For further detail onthe SfM process, readers are referred to recentreviews by Refs 57 and 58.

The SfM-photogrammetry method has beenintegrated into a number of software packages,including the commercial PhotoScan Pro (AgisoftLLC, St. Petersburg, Russia) and Pix4Dmapper Pro(Pix4D, Lausanne, Switzerland), and the open sourceVisualSFM.59 The SfM processing chains are usuallylargely automated and can be performed easily bynonexperts. In recent years, SfM has seen applicationin various fields, including archeology,56

glaciology,60 and geomorphology,45,46,51,54,6163

using various camera and platform setups. Withinthis review, we focus specifically on the use of SfMfor river habitat applications, in conjunction withdrone imagery.

CLASSIFICATIONS OF PHYSICALRIVER HABITAT

Much of the work using drones and digital photo-grammetry for mapping and monitoring physicalriver habitat has focused on replicating the well-usedclassification systems which are typically implemen-ted during walk-over assessments or large scaleremote sensing surveys. Examples can be broadlycategorized into those which conduct manual map-ping on the drone-SfM outputs, and those whichexperiment with automated classification proceduresfor feature detection. Our first case study provides anillustration of the former.

Case Study 1: San Pedro River, ChileFor this case study, we provide an example of tradi-tional geomorphological mapping conducted usingoutputs from the drone-SfM process. The imagerywas collected from a site on the right bank of the SanPedro River in south-central Chile in May 2012, andprocessed using SfM-photogrammetry. The site islocated just downstream of the outlet from LakeRiihue, where the San Pedro River is deep, gentlysinuous with very clear water. Here, the right bank iscomposed of consolidated clay bedrock, overlain by

areas of gravel and cobbles with some boulders andlarge woody debris. The focus of this case study isthis area of clay bedrock, which hugs parts of thebank. Here, water depth is relatively shallow (up toc. 2 m).

BackgroundIn order to meet growing energy demands in Chile,this site has been selected for the construction of alarge (56 m) hydropower dam and reservoir, as oneof a number to be installed on high gradient, power-ful Andean rivers. Significant concern has been raisedabout the impact the dams will have on natural riverhabitats64 as channel engineering works will modifythe natural hydromorphology.65 Such changes havesignificant implications for habitat quality and availa-bility, and subsequent impacts on unique native Chil-ean fish populations are inevitable.66 Ongoingresearch has been aiming to characterize the currentphysical river habitat, prior to dam construction, inorder to better understand the requirements of the lit-tle studied native fish.6769 It is hoped that this willhelp inform planning on how to maintain or restoresimilar habitats in future, in the knowledge that thisparticular section of the San Pedro River has a highspecies richness and thus is representative of thephysical conditions suitable for many fish species.Drone-based work on the San Pedro in May 2012was aimed at providing high-resolution remote sen-sing data about the physical habitat (specifically sub-strate size), in a way, which is faster, less laborious,more objective, and more spatially continuous thantraditional habitat surveying. The need for such datais especially acute given the relative lack of nationalaerial remote sensing surveys in countries such asChile.

Application of Drones and DigitalPhotogrammetryTo obtain the imagery of the San Pedro site, wemounted a small, RGB, consumer-grade, digital cam-era (Panasonic Lumix DMC-LX3, Osaka, Japan)onto a small, lightweight, rotary-winged drone(Draganflyer X6, Draganfly Innovations Inc., Saska-toon, Canada) and flew over the extensive shelfsection on the right bank of the river (c. 170 mlong c. 20 m wide). A study area of this size was atthe technological limits of what could be achievedusing drones at the time of survey in 2012, but is rel-atively small in comparison to the areas which can besurveyed with ease by contemporary drone systemstoday. We positioned the drone at c. 25 m aboveground level and set the camera focal length manu-ally to 5 mm, to allow production of hyperspatial

WIREs Water Drones and digital photogrammetry

2017 The Authors. WIREs Water published by Wiley Periodicals, Inc. 5 of 20

resolution outputs. We collected images with a highlevel of overlap (>80%) to allow for subsequent pro-cessing using SfM-photogrammetry. We distributedartificial GCPs across the site and recorded their posi-tion using a Spectra Precision EPOCH 50 global nav-igation satellite system (GNSS), capable ofsubcentimeter accuracy. Using this drone, about adays work was required in total for site setup anddata collection of c. 210 images. Significant advancesin drone technology since 2012 mean that an equiva-lent amount of imagery can now be collected by adrone such as the DJI Inspire 1 (Da-Jiang Innova-tions Science and Technology Co. Ltd., Shenzhen,China) in about 15 min, with some additional timefor GCP layout and surveying. We imported thedrone images into Agisofts PhotoScan Pro softwarefor SfM photogrammetric processing to create a0.01 m resolution orthophoto (Figure 2) and a0.02 m resolution DEM. As shown in Figure 3, thesevery high-resolution SfM products permit the follow-ing, all of which were achieved by mapping at a scaleof 1:100 using the measuring tools available within aGIS environment (ArcMap, ESRI Ltd, Red-lands, USA):

1. Detailed mapping of key breaks of slope fromthe DEM, in accordance with geomorphologi-cal classification system presented by Ref 71.

2. Quantitative measurement of slope anglesbetween breaks of slope, using the DEM.

3. Manual measurement and classification of sub-strate size according to the Wentworth Scale72

using the orthophoto.

4. Mapping of the position of large woody debris,as shown in the orthophoto.

These outputs allow, for the first time, a spa-tially continuous and explicit characterization of thephysical habitat at this location. Data such as thishas the potential to contribute to a wide range of sci-ence and management scenarios, including sedimentbudgeting, enabling the development of conceptualmodels of habitat distribution and spatial connectiv-ity, aiding decision-making and research designs forfurther investigation of the effects of engineeringworks (such as this proposed hydropower facility)and for inferring habitat preferences or informationon fish swimming performance when combined withfish sampling surveys.68,73

In recent years, other work has demonstratedhow a range of physical habitat indicators can beidentified and classified using drone imagery and dig-ital photogrammetry. For instance, surface flow types

have been mapped manually from tethered balloonsand rotary-winged drone imagery,74,75 hydromor-phological mapping has been conducted using super-vised classifications of drone images,76 mesohabitatmapping has been conducted using a combination ofdrone imagery and side-scan sonar data,17 locatingthe presence of intermittent streams has been per-formed using supervised classifications of fixed-wingimagery,77 the position of large woody debris hasbeen mapped both manually78 and using supervisedclassification techniques,79 various mapping methodshave been used to characterize riparian and in-channel vegetation and algae from droneimagery,35,8082 and the impact of species reintroduc-tion has been assessed using DEMs andorthophotos.83

As with traditional habitat mapping, decisionsare made within all these examples about which clas-sification scheme to use, the scale, coverage, and fre-quency of sampling, and in some instances, themethod of interpolation or extrapolation. Thesechoices are not always explicit or fully justified, butare typically led by the nature of the application,existing practices, or protocols within certain geo-graphic regions or academic disciplines, and theavailability of time, funds, and resources. As a result,the habitat classification methods we use, whethertraditional or drone-based and including ourexample on the San Pedro River, can be subjective,scale-dependent, nontransferable, nonquantitative,inconsistent, and/or based on inference. Furthermore,assumptions are made about the accuracy and relia-bility of these classifications, which are not substan-tiated by quantitative evidence. However, significanttechnological advances in drones and associated dig-ital photogrammetric processing techniques nowoffer us the opportunity to move away from some ofthese restraints by providing an approach which isobjective, rapid, quantitative, spatially continuous,and of exceptionally high spatial and temporal reso-lutions.78,84,85 The luxury of such a method has beenpreviously unattainable within physical habitat stud-ies, but now encourages us to move away from broadclassifications of river habitat in favor of detailedcontinuums of quantitative data.

CONTINUUMS OF PHYSICAL RIVERHABITAT

A continuum is defined as a coherent whole, charac-terized as a collection, sequence, or progression ofvalues varying by minute degrees.86 The physicalhabitat within river systems includes numerous



continuums, such as grain size, water depth, topo-graphic elevation, and flow velocity. The concept isthat if we can capture these continuums, we maythen be able to provide two- or three-dimensional,hyperspatial resolution, quantitative measurementsof physical river habitat parameters over entire rivercatchments. In theory, these continuums can then beused over a range of spatial or temporal scales,dependent on the requirements of a given

application. In practice, while recent technologicaldevelopments in drones and digital photogrammetryare beginning to convert this theory into reality, thecurrent challenge is to evaluate fully and realisticallythe application of this approach to real-world riverscience and management scenarios. We are facedwith a number of key questions (Box 1).

In the last few years, research aimed specifi-cally at answering these questions has started to

FIGURE 2 | Orthophoto of the Piedra Blanca site on the San Pedro River, Chile, obtained using the drone-structure-from-motion (SfM)method.70



FIGURE3

|Geomorphologicalmapping

ofthePiedra

Blanca

siteon

theSanPedroRiver,Ch

ile.70



emerge within the academic literature, most fre-quently addressing Q1. Much of the work withinriver science has focused on the continuums of topo-graphic data in particular, provided by DEMsderived directly from the drone-SfMmethod.18,46,78,85,87 Numerous examples haveappeared within the more general geomorphologyliterature too.33,34,40,54,58,84,88,89

Case Study 2: River Teme, UK

BackgroundThe River Teme, at our selected site in Herefordshire(England), is a meandering, highly mobile, gravel bedriver with a width of 313 m, which is intimately con-nected to its floodplain. In very recent times (since2014) it has changed course dramatically in responseto major winter storm events, shifting its position bymore than 100 m in some places. Clearly such signifi-cant geomorphic change will impact heavily on thedistribution of river habitat. We are unable to studypast change in habitat availability in great detail dueto a lack of earlier imagery at hyperspatial resolutions.However, research is ongoing to monitor futurechanges in channel geomorphology and quantify howaccurately we are able to measure these changes usingdrone and digital photogrammetry based approaches.

Application of Drones and DigitalPhotogrammetryFigure 4 provides an example of an orthophoto andDEM generated using drone imagery and SfM for

this stretch of the River Teme. These data form ourbaseline survey, against which future change will becompared. We obtained the imagery using a Panaso-nic Lumix DMC-LX3 camera mounted on a Dra-ganflyer X6 drone, flown at a height of c. 25 mabove ground level, in August 2016. We used sevenartificial GCPs surveyed using a total station and sixfixed markers surveyed using a Trimble R8 RTKGNSS to give scale to our model (British NationalGrid). The resulting spatial resolution of the DEMand orthophoto were 0.02 and 0.01 m respectively.Given the clear water and low flow levels during oursurvey we are able to model the submerged topogra-phy in most parts of the site, with the exception ofsome parts of the main channel and an area of tur-bid, ponded water at the downstream end of thepoint bar (as shown by the holes in the data inFigure 4(b) and (d)). Figure 4 clearly demonstratesthat detailed topographic heterogeneity exists bothwithin the wetted channel itself as well as the sur-rounding area (Q1proof of concept).

Data AnalysisElevation accuracy was computed using 490 valida-tion points collected during a total station topo-graphic survey, also conducted during August 2016(Figure 5, Table 1). The equation intercept valuesin Figure 5 indicate a trend toward elevation over-estimation within all parts of the site, with astronger correlation between observed and pre-dicted values for exposed (slope = 0.9878) thansubmerged areas (slope = 0.6825). Mean elevationerror in exposed parts of the site is 0.052 m, butdegrades to 0.101 m in submerged areas due to therefraction of light at the airwater interface(Table 1). Large overestimations in elevation inexposed areas (>0.2 m) are associated with notable,vegetated breaks of slope. Here, any small errors inthe horizontal dimension result in significant errorsin the vertical. Where validation points in theselocations are excluded, mean elevation error forexposed areas is improved to 0.037 m. These errorassessment results reflect other findings reported todate where a similar approach isused.18,40,46,54,61,85,87 In submerged areas, overpre-diction increases with water depth (i.e., withdecreasing elevation in Figure 5(b)). A relativelysimple refraction correction procedure has beenshown to reduce the magnitude of elevation overes-timation in submerged areas with clear water.18

Very recently, a development of this correction pro-cedure, which iteratively corrects the effect ofrefraction for each image in the SfM process, hasbeen able to reduce elevation error further still, to

BOX 1

FOCUS AREAS FOR CONVERTINGTECHNOLOGICAL DEVELOPMENTS INTOVALUABLE METHODS FOR RIVERSCIENCE AND MANAGEMENT

Q1. Can drones and digital photogrammetricapproaches actually provide continuums ofphysical river habitat data in practice? And,if so;

Q2. How accurate and precise are thesecontinuums?

Q3. Do these continuums confirm the validityof (1) our prevailing conceptualizations of theriver system, and (2) the practices we use tomanage river habitats?



FIGURE 4 | River Teme orthophoto (a and c) and digital elevation model (DEM) (b and d) generated using drone imagery processed usingstructure-from-motion (SfM)-photogrammetry. Flow is from the left to right side of the image in (a) and (b) and from bottom to top in (c) and (d).The validation points shown in (a) and (c) denote the location of survey points collected with a total station and subsequently used to assessquantitatively the accuracy and precision of the drone-SfM derived DEM.

124.5(a) (b)

122.6

122.4

122.2

122.0

121.8

121.6

121.4

121.2124.5

x = y

x = y

124.0

124.0

123.5

123.5

123.0

123.0

y = 0.9878x + 1.5319R2 = 0.9873

y = 0.6825x + 38.822R2 = 0.8844

Validation elevation (m asl) Validation elevation (m asl)

122.5

122.5

DE

M e

leva

tion

(m a

sl)

DE

M e

leva

tion

(m a

sl)

122.0122.0 121.2 121.4 121.6 121.8 122.0 122.2 122.4 122.6

FIGURE 5 | Observed versus predicted elevation data for (a) exposed and (b) submerged parts of the River Teme field site.



levels which are commensurate with results typi-cally obtained in dry, exposed terrain.90. However,accurate and reliable quantification of submergedbed topography remains challenging in deep(>1.4 m), turbid, and turbulent waters (Q2).87,90,91

Further research is needed to quantify the exactdepths, levels of turbidity and turbulence at whichthe method begins to fail.

Evidence of our ability to provide spatial conti-nuums of other physical river habitat parameters(Q1) is also beginning to emerge within the academicliterature. For example, image texture approaches,previously used with traditional aerial imagery toproduce continuous maps of fluvial grain size11 haverecently seen application to drone imagery. Ref 78generated high-resolution orthophotos (0.05 m) for a1 km section of the Elbow River in Canada usingimagery acquired from an Aeryon Scout quadcopter(Aeryon Labs Inc., Waterloo, Canada) and processedusing the EndoMOSAIC digital photogrammetrysoftware (MosaicMill Ltd, Finland). A strong predic-tive relationship (R2 = 0.82) was developed betweenclose-range photo-sieved grain size data and a meas-ure of drone image texture, to provide site-wide grainsize predictions (Q1). Unfortunately, however, thiswas not accompanied by any quantitative assessmentof error (Q2) and the spatial resolution of grain sizepredictions (1 m) does not improve on the resolutionof earlier work using manned aircraft.11,78 Recentconference proceedings indicate that otherapproaches to continuous grain size quantificationusing drones and digital photogrammetry are also

being explored.9294 Our ability to measure waterdepth and quantify the roughness of topographicpoint clouds within the wetted channel (Q1) also sug-gests that drone derived products may present anopportunity to assess the hydraulic environmentwithin relatively shallow, clear-flowing rivers.75

Additional developments have made use of SfMapplied to imagery derived from manned aircraft too,including riverscape mapping over longer reaches(e.g., 32 km).63 While much of this research remainsin its infancy, surveying physical river habitat para-meters at these fine scales has the potential to providenew opportunities for knowing rivers as continuums,including those smaller and more marginal features,which may be overlooked by broader scaleclassifications.

EVALUATION

Table 2 offers an overview of how far research intothe use of drones and digital photogrammetry forriver habitat assessment has progressed to date. Proofof concept work (Q1) which has emerged within thelast few years is now beginning to pave the way fordetailed assessments of accuracy and precision ofthese continuums (Q2). Continued research and sys-tematic experimentation for the acquisition of awider range of physical habitat variables, in a rangeof contrasting fluvial environments, is clearly criticalhowever if drones and digital photogrammetry are toprovide continuums which actually allow us toimprove our understanding of the science of riversystems and refine our methods of management(Q3). The potential of this method to provide rele-vant data at the critical scale for habitat assessmentsmeans that it may have a key role to play in enablingreliable, quantitative monitoring of river habitats,informing successful and sustainable climate changeadaptation strategies and subsequently reducing therisk of species decline and loss in certain settings. Atpresent, this remains an ambition rather than a real-ity, and further contributions to Q1, Q2, and Q3 aretherefore required.

TABLE 1 | Accuracy and Precision Measures for the River TemeDigital Elevation Model

Area of Site

Accuracy PrecisionMean

Error (m)Standard

Deviation (m)

Exposed areas 0.0515 0.0917

Exposed areas exc. breaksof slope

0.0369 0.0510

Submerged areas 0.1007 0.8170

TABLE 2 | An Overview of the State of Drone and Digital Photogrammetry Based Research for Quantifying Physical River Habitat Parameters

Physical Habitat Variable

Data Continuums

Q1. Proof of Concept Q2. Error Assessment Q3. New Knowledge/Practices

Topography/water depth Refs 40, 46, 70, 78, and 95 Refs 18, 70, and 87; Ref 85

Grain size Ref 78; Ref 92 Refs 70, 93 and 94

Flow velocity Ref 96

Text in roman indicates peer-reviewed publications.Text in italics indicates conference abstracts and proceedings, and unpublished findings where these are publically available.Note that this table does not include studies which do not use both drone imagery and digital photogrammetry.



In terms of the routine operational use ofdrones and digital photogrammetry, we are able toidentify a number of broad challenges, which mustbe addressed to help realize this ambition. These aregiven in no particular order:

Data Acquisition Image blur. The low flying altitude and variable

effectiveness of camera gimbals mean thatdrone imagery can suffer from blur. Image blurincreases noise within the SfM point cloudwhich subsequently degrades the quality of top-ographic products.97 Methods to identify,reduce, or prevent blur within UAS imagery98

require further attention.

Need for GCPs. The use of GCPs providingadequate representation of the survey area inthree dimensions has been essential for accuratespatial positioning and scaling of image andtopography data to date. However, the distribu-tion and survey of GCPs increases significantlythe time needed in the field and necessitates theuse of expensive survey equipment. Technologi-cal developments in miniaturizing higher gradeGNSS and increasing the payload capacity ofsmall drones may reduce the need for GCPs.Initial work suggests that while horizontal posi-tioning can be adequately obtained with the useof an on-board RTK GNSS, that vertical errorsare greater than those obtained with the use ofGCPs.99 Other research in this vein suggeststhat acceptable positioning accuracies can beobtained with relatively small numbers ofGCPs100 and that direct georeferencing usingequifinality relationships within the SfM work-flow may obviate the need for GCPsaltogether.50

Radiometric and spectral resolution. The major-ity of research to date makes use of 8-bit RGBimagery acquired from small-format, consumer-grade digital cameras because they are lightenough to be carried by small drones. The lim-ited radiometric resolution of this imagery tendsto hinder the restitution of topography in dar-ker parts of the site (including in areas of sha-dowing and deeper water), due to a reductionin the image texture which is required to iden-tify matching points between images. Futureresearch should give greater consideration toradiometric resolution41 by making use of cam-eras capable of acquiring RAW format images,which offer higher bit depths. Multispectral

sensors will also benefit habitat mapping infuture, especially as the market for drone-mountable sensors continues to expand anddiversify.

Weather conditions. Poor scene illumination isthought to hinder the success of SfM imagematching processes101 and windy conditionscan result in significant amounts of image blur.Recent research suggests that techniques forreducing the problematic impacts of sun glinton remotely sensed imagery are possible,102 butto our knowledge these have not yet been spe-cifically applied to imagery acquired using adrone-SfM approach. Careful flight planning istherefore required to ensure high quality imagecapture.

Battery life. Single flight times vary in accord-ance with the power supply and demand of dif-ferent drones. This can mean that multipleflights are necessary to provide adequate cover-age of field sites. Combined with the regulatoryrequirement in many countries to maintain vis-ual line-of-sight at all times and the need forsurveying GCPs, battery life is one of the majorfactors in determining field time and coveragepotential. In our experience, the Draganflyer X6drone used within this article can realisticallymanage single flight times of c. 6 min. Newerdrones are capable of longer flight durationsthough. For example, the DJI Inspire 1 rotary-winged system (Da-Jiang Innovations Scienceand Technology Co. Ltd., Shenzhen, China) canfly for up to 20 min on a single battery and thusachieve much greater coverage within a singleflight. Fixed wing drones are typically capableof even longer flight times of 30 min1 h, withthe Bramor ppX drone (C-Astral AerospaceLtd, Ajdovscina, Slovenia) advertised to deliverindustry-leading flight times of up to 3.5 h.Newer lithium-sulfur (Li-S) batteries are alsoadvocated to increase possible flight times.

Flight regulations. The UK regulatory environ-ment is currently favorable for noncommercialdrone-based research, provided that Articles 94-95 and 241 of the Air Navigation Order 2016are adhered to.103 Drone use in congestedareas of the UK does require Civil AviationAuthority permission however, and should befactored into the use of drones for monitoringhabitats within urban waterways. The regula-tory landscape for drone flying is rapidly evol-ving both within the UK and elsewhere(e.g., new Part 107 of Federal Aviation



Authority regulations in the USA104 andEuropean Aviation Safety Agencys PrototypeCommission Regulation on Unmanned AircraftOperations105), therefore researchers and prac-titioners should consult the local civil aviationauthority guidance for up-to-date advice on fly-ing regulations and for obtaining relevantpermissions.

Water quality. In rivers and other aquatic envir-onments, the success of the drone-SfMapproach is reliant on good visibility and thuswater quality is a critical limiting factor. Themethod is not applicable in areas of very turbid(i.e., high sediment concentrations) or whitewater (i.e., high water surface roughness).Investigations by the lead author are currentlyexploring the thresholds at which turbidity andturbulence prohibit the use of this method influvial settings.

Spatial coverage. At present, the areal extent ofdrone surveys is limited. Coverage of entireriver systems is not possible without a notableloss in spatial resolution or a significant increasein survey time and effort. This is a consequenceof the well-known trade-off between spatialcoverage and spatial resolution, limited dronebattery life and widespread regulations concern-ing maximum flight altitudes, distance from thepilot and line-of-sight flying requirements.While ongoing technological developments mayhelp to improve this situation in future, we sug-gest that regulatory requirements are unlikely tobe relaxed significantly. Therefore, difficulties incovering larger spatial extents may always be alimitation of drone-based surveying methods.

Data Processing Reliance on image texture. The SfM process is

heavily dependent on image texture to success-fully match points between overlapping images.Where this is lacking, the approach is compro-mised.106,107 As a result, surveys conductedover large expanses of smooth, opaque watersurfaces or areas of homogeneous grasslandshould not be expected to produce useful resultsfor physical river habitat assessments.

Geometric restitution is only possible for visiblefeatures. Unlike laser scanning approaches, theSfM technique is only able to reconstruct sur-faces that are clearly visible from the position ofthe drone. As a result, dense or overhangingvegetation will obscure any underlying features

of interest and elevation models will eitherappear to overestimate the true position of theground surface45,62 or confuse the image match-ing process as its appearance varies betweenview angles.108 This can make the use of dronesfor river assessment impossible on narrowstreams with dense riparian tree cover. Simi-larly, for the SfM process to work in submergedareas, a clear view of the channel bed is neces-sary, thereby excluding rivers with highly turbidand turbulent flows.

Processing power. The SfM process is computa-tionally demanding and time consuming, partic-ularly for large image datasets. The River Temesurvey described here is relatively small, com-prising 134 nongeotagged images which coveran area of approximately 8800 m2. Theseimages took approximately one working day toprocess to completion within Agisofts Photo-Scan Pro v.1.2.4.2399 (Agisoft LLP) using anIntel Core i7-4790 32 GB RAM 64-bit com-puter with a 2 GB Dual Nvidia Quadro K620video card. The use of geotagged imagery andthe use of higher performance computing willinevitably enable much shorter processingtimes.

Large data volumes. Drone image datasets andtheir resultant SfM outputs typically requirelarge digital storage volumes. The River Temedataset presented here (photographs and pro-cessed data) occupies c. 3 GB of disk space.While a handful of surveys of this size are rela-tively manageable, if this approach is to becomeroutine then researchers and management agen-cies may be faced with difficulties whenattempting to share datasets for perhaps hun-dreds of site surveys. Furthermore, continuedresearch is needed into methods of data proces-sing and analysis, which exploit effectively thehigh spatial and temporal frequency of thisdata. Some progress has been made in recentyears, including the increase in availability ofopen source point cloud management software(e.g., CloudCompare) and processingroutines.14,109

Errors within the SfM process. The black-boxnature of SfM software packages, includingPhotoScan Pro (Agisoft LLP), means that isolat-ing exact sources of error is challenging. Sys-tematic doming errors have been observedwithin SfM-generated DEMs18,62 and arethought to result from inadequate self-calibrations of camera lens models.110,111



Recent research suggests that careful flight pla-nning can reduce these systematic errors by theaddition of imagery acquired at convergentview angles.49 Ref 58 provides a thorough over-view of SfM error sources.

Despite these current limitations, drones and digitalphotogrammetry have undoubtedly and irreversiblyrevolutionized our data collection toolbox for riverhabitat surveying. The research reviewed here evi-dences numerous advantages over traditional meth-ods; namely the hyperspatial resolution, spatiallyexplicit and spatially continuous data which can beobtained more quickly, more easily, more flexibly,and more frequently than using other approaches.These advantages permit new possibilities for survey-ing quantitatively the heterogeneity of the physicalriver habitat. Drone platforms capable of the kind ofanalysis presented here can currently be purchasedfor c. 1000 (e.g., DJI Phantom 4, Da-Jiang Innova-tions Science and Technology Co. Ltd., Shenzhen,China)), and SfM processing workflows are user-friendly and easily learnt by nonexperts. This democ-ratization of data collection places control with theuser, thereby making question-driven, high resolu-tion research considerably more feasible than it hasbeen previously (Ref 112, p. 71). Results of initialpublications lend support to the concept of a spa-tially heterogeneous riverscape63,75,85 as advocatedby Ref 4 and others in the early 2000s.27,28 Goingforward, further examination of this heterogeneitymay permit new and valuable insights into the eco-logical importance of physical habitat, its spatial andtemporal significance, and the mechanisms by whichit is regulated . In Box 2, we provide some specificguidance for those considering using drones and dig-ital photogrammetry for monitoring physical riverhabitat and hydromorphology.

CONCLUSION

Within this article, we have reviewed the contributionof drones and digital photogrammetry for monitoringphysical river habitat and hydromorphology to date.The monitoring of physical habitat parameters con-tinues to play an important role in how we manageand understand river systems, and traditional methodsfall short of providing the mesoscale, spatially contin-uous and explicit datasets required under the river-scape paradigm.4 A growing number of proof-of-concept studies indicate that drones and digital photo-grammetry provide a promising alternative method,including the ability to produce orthoimagery and

BOX 2

IMPORTANT CONSIDERATIONS FOR THEPRACTICAL APPLICATION OF DRONESAND DIGITAL PHOTOGRAMMETRY FORPHYSICAL RIVER HABITAT MONITORING

What is your budget?The budget should be sufficient to cover theacquisition and maintenance of the drone(s),the sensor(s), the digital photogrammetry soft-ware, the computer, and video card to be usedfor SfM processing, the digital storagefacility(s), data collection accessories (e.g., GCPs,additional batteries, additional controllers, andsurveying equipment), and for insurance (dronecover plus public liability). Additional costs maybe incurred for flight training and obtainingpermission to conduct aerial work(i.e., commercial work103).

What is your application area?Your intended application will influence thetype of drone and sensor you need, and carefulconsideration should be given to factors suchas; the size and accessibility of the area to besurveyed, the required spatial resolution of theoutput orthophotos and topographic datasets,the required spectral and radiometric resolutionof the sensor, the stability and typical flightduration of the drone, the capabilities of SfM-photogrammetry, the need for GCPs, and anawareness of how to minimize output errors.57

What are your responsibilities?Drone pilots are responsible for safe and legalflying, in accordance with the regulations andrecommendations, which apply in the locationof flight. This should include an awareness offlight limiting factors such as the presence oftrees and other substantial vegetation, power-lines, pedestrians not under the pilots control,and the proximity of urban areas andrestricted airspace. This list is not exhaustiveand it is the responsibility of the drone pilotto conduct a thorough risk assessment inadvance of flights. Special permissions may berequired in some countries and therefore it isimportant to obtain advice from the local avia-tion authority, and approval from the relevantlandowners. Reckless and noncompliant flyingposes a danger to life and property and threa-tens to jeopardize the future use of drones forresearch and management within ourdiscipline.



topographic data at exceptionally high spatial(

rivers using airborne digital imagery. Water ResourRes 2004, 40:W07202. https://doi.org/10.1029/2003WR002759.

12. Heritage GL, Milan DJ. Terrestrial laser scanning ofgrain roughness in a gravel-bed river. Geomorphol-ogy 2009, 113:411. https://doi.org/10.1016/j.geomorph.2009.03.021.

13. Hodge R, Brasington J, Richards K. In situ character-ization of grain-scale fluvial morphology using terres-trial laser scanning. Earth Surf Process Landf 2009,34:954968. https://doi.org/10.1002/esp.1780.

14. Brasington J, Vericat D, Rychov I. Modeling riverbed morphology, roughness, and surface sedimentol-ogy using high resolution terrestrial laser scanning.Water Resour Res 2012, 48:W11519. https://doi.org/10.1029/2012WR012223.

15. Rychov I, Brasington J, Vericat D. Computationaland methodological aspects of terrestrial surface anal-ysis based on point clouds. Comput Geosci 2012,42:6470. https://doi.org/10.1016/j.cageo.2012.02.011.

16. Wright A, Marcus WA, Aspinall R. Evaluation ofmultispectral, fine scale digital imagery as a tool formapping stream morphology. Geomorphology 2000,33:107120. https://doi.org/10.1016/S0169-555X(99)00117-8.

17. Cheek BD, Grabowski TB, Bean PT, Groeschel JR,Magnelia S. Evaluating habitat associations of a fishassemblage at multiple spatial scales in a minimallydisturbed stream using low-cost remote sensing.Aquat Conserv Mar Freshw Ecosyst 2015, 26:2034.https://doi.org/10.1002/aqc.2569.

18. Woodget AS, Carbonneau PE, Visser F, Maddock I.Quantifying submerged fluvial topography usinghyperspatial resolution UAS imagery and structurefrom motion photogrammetry. Earth Surf ProcessLandf 2015, 40:4764. https://doi.org/10.1002/esp.3613.

19. Maddock I. The importance of physical habitatassessment for evaluating river health. Freshw Biol1999, 41:373391. https://doi.org/10.1046/j.1365-2427.1999.00437.x.

20. Orr HG, Large ARG, Newson MD, Walsh CL. Apredictive typology for characterising hydromorphol-ogy. Geomorphology 2008, 100:3240. https://doi.org/10.1016/j.geomorph.2007.10.022.

21. Vaughan IP, Diamond M, Gurnell AM, Hall KA,Jenkins A, Milner NJ, Naylor LA, Sear DA,Woodward G, Ormerod SJ. Integrating ecology withhydromorphology: a priority for river science andmanagement. Aquat Conserv Mar Freshw Ecosyst2009, 19:113125. https://doi.org/10.1002/aqc.895.

22. Nilsson C, Berggren K. Alterations of riparian ecosys-tems caused by river regulation. BioScience 2000,50:783792. https://doi.org/10.1641/0006-3568(2000)050[0783:AORECB]2.0.CO;2.

23. Raven PJ, Fox P, Everard M, Holmes NTH,Dawson FH. River habitat survey: a new system forclassifying rivers according to their habitat quality.In: Boon PJ, Howell DL, eds. Freshwater Quality:Defining the Undefinable. Edinburgh: HMSO; 1998,215234.

24. Maddock I, Thoms M, Jonson K, Dyer F,Lintermans M. Identifying the influence of channelmorphology on physical habitat availability for nativefish: application to the two-spined blackfish (Gadop-sis bispinosus) in the Cotter River, Australia. MarFreshw Res 2004, 55:173184. https://doi.org/10.1071/MF03114.

25. Vannote RL, Minshall GW, Cummins KW, Sedell JR,Cushing CE. The river continuum concept. Can J FishAquat Sci 1980, 37:130137.

26. Frissell CA, Liss WJ, Warren CE, Hurley MD. A hier-archical framework for stream habitat classification:viewing streams in a watershed context. EnvironManag 1986, 10:199214. https://doi.org/10.1007/BF01867358.

27. Ward JV, Malard F, Tockner K. Landscape ecology:a framework for integrating pattern and process inriver corridors. Landsc Ecol 2002, 17:3545. https://doi.org/10.1023/A:1015277626224.

28. Wiens JA. Riverine landscapes: taking landscape ecol-ogy into the water. Freshw Biol 2002, 47:501515.https://doi.org/10.1046/j.1365-2427.2002.00887.x.

29. Newson MD, Newson CL. Geomorphology, ecologyand river channel habitat: mesoscale approaches tobasin-scale challenges. Progr Phys Geogr 2000,24:195217. https://doi.org/10.1177/03091333000240020.

30. Hardin PJ, Jensen RR. Small-scale unmanned aerialvehicles in environmental remote sensing: challengesand opportunities. GISci Remote Sens 2011,48:99111. https://doi.org/10.2747/1548-1603.48.1.99.

31. Watts AC, Ambrosia VG, Hinkley EA. Unmannedaircraft systems in remote sensing and scientificresearch: classification and considerations of use.Remote Sens (Basel) 2012, 4:16711692. https://doi.org/10.3390/rs4061671.

32. Eisenbeiss H, Lambers K, Sauerbier M, Li Z. Photo-grammetric documentation of an archaeological site(Palpa, Peru) using an autonomous model helicopter.In: CIPA 2005 XX International Symposium, Torino,Italy, 26 September1 October, 2005.

33. Niethammer U, James MR, Rothmund S,Travelletti J, Joswig M. UAV-based remote sensing ofthe Super-Sauze landslide: evaluation and results. EngGeol 2012, 128:211. https://doi.org/10.1016/j.enggeo.2011.03.012.

34. Smith MJ, Chandler J, Rose J. High spatial resolutiondata acquisition for the geosciences: kite aerial



https://doi.org/10.1029/2003WR002759https://doi.org/10.1029/2003WR002759https://doi.org/10.1016/j.geomorph.2009.03.021https://doi.org/10.1016/j.geomorph.2009.03.021https://doi.org/10.1002/esp.1780https://doi.org/10.1029/2012WR012223https://doi.org/10.1029/2012WR012223https://doi.org/10.1016/j.cageo.2012.02.011https://doi.org/10.1016/j.cageo.2012.02.011https://doi.org/10.1016/S0169-555X(99)00117-8https://doi.org/10.1016/S0169-555X(99)00117-8https://doi.org/10.1002/aqc.2569https://doi.org/10.1002/esp.3613https://doi.org/10.1002/esp.3613https://doi.org/10.1046/j.1365-2427.1999.00437.xhttps://doi.org/10.1046/j.1365-2427.1999.00437.xhttps://doi.org/10.1016/j.geomorph.2007.10.022https://doi.org/10.1016/j.geomorph.2007.10.022https://doi.org/10.1002/aqc.895https://doi.org/10.1641/0006-3568(2000)050[0783:AORECB]2.0.CO;2https://doi.org/10.1641/0006-3568(2000)050[0783:AORECB]2.0.CO;2https://doi.org/10.1071/MF03114https://doi.org/10.1071/MF03114https://doi.org/10.1007/BF01867358https://doi.org/10.1007/BF01867358https://doi.org/10.1023/A:1015277626224https://doi.org/10.1023/A:1015277626224https://doi.org/10.1046/j.1365-2427.2002.00887.xhttps://doi.org/10.1177/03091333000240020https://doi.org/10.1177/03091333000240020https://doi.org/10.2747/1548-1603.48.1.99https://doi.org/10.2747/1548-1603.48.1.99https://doi.org/10.3390/rs4061671https://doi.org/10.3390/rs4061671https://doi.org/10.1016/j.enggeo.2011.03.012https://doi.org/10.1016/j.enggeo.2011.03.012

photography. Earth Surf Process Landf 2009,34:155161. https://doi.org/10.1002/esp.1702.

35. Dunford R, Michel K, Gagnage M, Piegay H,Tremelo M-L. Potential constraints of unmanned aer-ial vehicle technology for the characterization ofMediterranean riparian forest. Int J Remote Sens2009, 30:49154935. https://doi.org/10.1080/01431160903023025.

36. Lalibert AS, Rango A. Texture and scale in object-based analysis of subdecimeter resolution unmannedaerial vehicle (UAV) imagery. IEEE Trans GeosciRemote Sens 2009, 47:761770. https://doi.org/10.1109/TGRS.2008.2009355.

37. Hunt ER, Hively WD, Fujikawa SJ, Linden DS,Daughtry CST, McCarty GW. Acquisition of NIR-green-blue digital photographs from unmanned air-craft for crop monitoring. Remote Sens (Basel) 2010,2:290305. https://doi.org/10.3390/rs2010290.

38. Lalibert AS, Winters C, Rango A. UAS remote sen-sing missions for rangeland applications. GeocartoInt 2011, 26:141156. https://doi.org/10.1080/10106049.2010.534557.

39. Hugenholtz C, Whitehead K, Brown OW,Barchyn TE, Moorman BJ, LeClair A, Riddell K,Hamilton T. Geomorphological mapping with a smallunmanned aircraft system (sUAS): feature detectionand accuracy assessment of a photogrammetricallyderived digital terrain model. Geomorphology 2013,194:1624. https://doi.org/10.1016/j.geomorph.2013.03.023.

40. Lejot J, Delacourt C, Piegay H, Fournier T,Tremelo M-L, Allemand P. Very high spatial resolu-tion imagery for channel bathymetry and topographyfrom an unmanned mapping controlled platform.Earth Surf Process Landf 2007, 32:17051725.https://doi.org/10.1002/esp.1595.

41. DeBell L, Anderson K, Brazier RE, King N, Jones L.Water resources management at catchment scalesusing lightweight UAVs: current capabilities andfuture perspectives. J Unmanned Vehicle Syst 2016,4:730. https://doi.org/10.1139/juvs-2015-0026.

42. Petrie G. Commercial operation of lightweight UAVsfor aerial imaging and mapping. GeoInformatics2013, 16:2839.

43. Lane SN, Richards KS, Chandler JH. Developmentsin photogrammetry; the geomorphological potential.Progr Phys Geogr 1993, 17:306328. https://doi.org/10.1177/030913339301700302.

44. Rosnell T, Honkavaara E. Point cloud generationfrom aerial image data acquired by a quadrocoptertype micro unmanned aerial vehicle and a digital stillcamera. Sensors 2012, 12:453480. https://doi.org/10.3390/s120100453.

45. Westoby MJ, Brasington J, Glasser NF,Hambrey MJ, Reynolds JM. Structure-from-motionphotogrammetry: a low cost, effective tool for

geoscience applications. Geomorphology 2012,179:300314. https://doi.org/10.1016/j.geomorph.2012.08.021.

46. Fonstad MA, Dietrich JT, Courville BC, Jensen JL,Carbonneau PE. Topographic structure from motion:a new development in photogrammetric measure-ment. Earth Surf Process Landf 2013, 38:421430.https://doi.org/10.1002/esp.3366.

47. Lowe DG. Distinctive image features from scale-invariant keypoints. Int J Comput Vis 2004,60:91110. https://doi.org/10.1023/B:VISI.0000029664.99615.94.

48. Turner D, Lucieer A, Watson C. An automated tech-nique for generating georectified mosaics from ultra-high resolution unmanned aerial vehicle (UAV)imagery, based on structure from motion (SfM) pointclouds. Remote Sens (Basel) 2012, 4:13921410.https://doi.org/10.3390/rs4051392.

49. James MR, Robson S. Mitigating systematic error intopographic models derived from UAV and ground-based image networks. Earth Surf Process Landf2014, 39:14131420. https://doi.org/10.1002/esp.3609.

50. Carbonneau PE, Dietrich JT. Cost-effective non-metric photogrammetry from consumer-grade sUAS:implications for direct georeferencing of structurefrom motion photogrammetry. Earth Surf ProcessLandf 2016, 42:473489. https://doi.org/10.1002/esp.4012.

51. Micheletti N, Chandler JH, Lane SN. Investigatingthe geomorphological potential of freely availableand accessible structure-from-motion photogramme-try using a smartphone. Earth Surf Process Landf2015, 40:473486. https://doi.org/10.1002/esp.3648.

52. Snavely N, Seitz S, Szeliski R. Photo tourism: explor-ing image collections in 3D. ACM Trans Graph(Proceedings of SIGGRAPH 2006) 2006, 25:835846.https://doi.org/10.1145/1141911.1141964.

53. Neitzel F, Klonowski J. 2011. Mobile 3D mappingwith a low-cost UAV system. In: InternationalArchives of the Photogrammetry, Remote Sensingand Spatial Information Sciences Vol. XXXVIII-I/C22, Conference on UAVs in Geomatics, Zurich,Switzerland.

54. Harwin S, Lucieer A. Assessing the accuracy of geore-ferenced point clouds produced via multi-view stere-opsis from unmanned aerial vehicle (UAV) imagery.Remote Sens (Basel) 2012, 4:15731599. https://doi.org/10.3390/rs4061573.

55. Furukawa Y, Ponce J. Accurate, dense, and robustmultiview stereopsis. IEEE Trans Pattern Anal MachIntell 2010, 32:13621376. https://doi.org/10.1109/TPAMI.2009.161.

56. Verhoeven G, Doneus M, Briese C, Vermeulen F.Mapping by matching: a computer vision-basedapproach to fast and accurate georeferencing of



https://doi.org/10.1002/esp.1702https://doi.org/10.1080/01431160903023025https://doi.org/10.1080/01431160903023025https://doi.org/10.1109/TGRS.2008.2009355https://doi.org/10.1109/TGRS.2008.2009355https://doi.org/10.3390/rs2010290https://doi.org/10.1080/10106049.2010.534557https://doi.org/10.1080/10106049.2010.534557https://doi.org/10.1016/j.geomorph.2013.03.023https://doi.org/10.1016/j.geomorph.2013.03.023https://doi.org/10.1002/esp.1595https://doi.org/10.1139/juvs-2015-0026https://doi.org/10.1177/030913339301700302https://doi.org/10.1177/030913339301700302https://doi.org/10.3390/s120100453https://doi.org/10.3390/s120100453https://doi.org/10.1016/j.geomorph.2012.08.021https://doi.org/10.1016/j.geomorph.2012.08.021https://doi.org/10.1002/esp.3366https://doi.org/10.1023/B:VISI.0000029664.99615.94https://doi.org/10.1023/B:VISI.0000029664.99615.94https://doi.org/10.3390/rs4051392https://doi.org/10.1002/esp.3609https://doi.org/10.1002/esp.3609https://doi.org/10.1002/esp.4012https://doi.org/10.1002/esp.4012https://doi.org/10.1002/esp.3648https://doi.org/10.1145/1141911.1141964https://doi.org/10.3390/rs4061573https://doi.org/10.3390/rs4061573https://doi.org/10.1109/TPAMI.2009.161https://doi.org/10.1109/TPAMI.2009.161

archaeological aerial photographs. J Archaeol Sci2012, 39:20602070. https://doi.org/10.1016/j.jas.2012.02.022.

57. Smith MW, Carrick JL, Quincey DJ. Structure frommotion photogrammetry in physical geography. ProgrPhys Geogr 2015, 40:247275. https://doi.org/10.1177/0309133315615805.

58. Eltner A, Kaiser A, Castillo C, Rock G, Neugirg F,Abelln A. Image-based reconstruction in geomor-phometry merits, limits and developments. EarthSurf Dyn 2016, 4:359389. https://doi.org/10.5194/esurf-4-359-2016.

59. Wu C. Visual SfM: a visual structure from motionsystem. Available at: http://ccwu.me/vsfm/ (AccessedSeptember 29, 2016).

60. Ryan JC, Hubbard AL, Box JE, Todd J,Christoffersen P, Carr JR, Holt TO, Snooke N. UAVphotogrammetry and structure from motion to assesscalving dynamics at Store Glacier, a large outletdraining the Greenland ice sheet. Cryosphere 2015,9:111. https://doi.org/10.5194/tc-9-1-2015.

61. James MR, Robson S. Straightforward reconstructionof 3D surfaces and topography with a camera: accu-racy and geosciences application. J Geophys Res2012, 117:F03017. https://doi.org/10.1029/2011JF002289.

62. Javernick L, Brasington J, Caruso B. Modelling thetopography of shallow braided rivers using structure-from-motion photogrammetry. Geomorphology2014, 213:166182. https://doi.org/10.1016/j.geomorph.2014.01.006.

63. Dietrich JT. Riverscape mapping with helicopter-based structure-from-motion photogrammetry. Geo-morphology 2016, 252:144157. https://doi.org/10.1016/j.geomorph.2015.05.008.

64. Goodwin P, Jorde K, Meier C, Parra O. Minimizingenvironmental impacts of hydropower development:transferring lessons from past projects to a proposedstrategy for Chile. J Hydroinform 2006, 8:253270.

65. Garcia A, Jorde K, Habit E, Caamano D, Parra O.Downstream environmental effects of dam opera-tions: changes in habitat quality for native fish spe-cies. River Res Appl 2011, 27:312327. https://doi.org/10.1002/rra.1358.

66. Habit E, Belk MC, Parra O. Response of the riverinefish community to the construction and operation ofa diversion hydropower plant in central Chile. AquatConserv Mar Freshw Ecosyst 2007, 17:3749.https://doi.org/10.1002/aqc.774.

67. Colin N, Piedra P, Habit E. Variaciones espaciales ytemporales de las comunidades ribereas de peces enun sistema fluvial no intervenido: ro San Pedro,cuenca del ro Valdivia (Chile). Gayana 2012,76:3644.

68. Link O, Habit E. Requirements and boundary condi-tions for fish passes of non-sport fish species based on

Chilean experiences. Rev Environ Sci Bio/Technol2015, 14:921. https://doi.org/10.1007/s11157-014-9357-z.

69. Wilkes MA, Maddock I, Link O, Habit E. Acommunity-level, mesoscale analysis of fish assem-blage structure in shoreline habitats of a large riverusing multivariate regression trees. River Res Appl2015, 32:652665. https://doi.org/10.1002/rra.2879.

70. Woodget AS. Quantifying physical river habitat para-meters using hyperspatial resolution UAS imageryand SfM-photogrammetry. Unpublished PhD Thesis,University of Worcester, UK, 2015.

71. Cooke RU, Doornkamp JC. Geomorphology in Envi-ronmental Management: A New Introduction. 2nded. Oxford: Oxford University Press; 1990.

72. Wentworth CK. A scale of grade and class terms forclastic sediments. J Geol 1922, 30:377392.

73. Laborde A, Gonzlez A, Sanhueza C, Arriagada P,Wilkes M, Habit E, Link O. Hydropower develop-ment, riverine connectivity and nonsport fish species:criteria for hydraulic design of fishways. River ResAppl 2016, 32:19491957. https://doi.org/10.1002/rra.3040.

74. Reid MA, Thoms MC. Mapping stream surface flowtypes by balloon: an inexpensive high resolutionremote sensing solution to rapid assessment of streamhabitat heterogeneity? In IAHS Publication 328 Eco-hydrology of Surface and Groundwater DependentSystems: Concepts, Methods and Recent Develop-ments. In: Proceedings of JS.1 at the Joint IAHS andIAH Convention, Hyderabad, India, September 2009.

75. Woodget AS, Visser F, Maddock I, Carbonneau PE.The accuracy and reliability of traditional surfaceflow type mapping: is it time for a new method ofcharacterizing physical river habitat? River Res Appl2016, 32:19021914. https://doi.org/10.1002/rra.3047.

76. Rivas-Casado M, Ballesteros Gonzalez R,Kriechbaumer T, Veal A. Automated identification ofriver hydromorphological features using UAV highresolution aerial imagery. Sensors 2015,15:2796927989. https://doi.org/10.3390/s151127969.

77. Spence C, Mengistu S. Deployment of an unmannedaerial system to assist in mapping an intermittentstream. Hydrol Process 2015, 30:493500. https://doi.org/10.1002/hyp.10597.

78. Tamminga A, Hugenholtz C, Eaton B, Lapointe M.Hyperspatial remote sensing of channel reach mor-phology and hydraulic fish habitat using anunmanned aerial vehicle (UAV): a first assessment inthe context of river research and management. RiverRes Appl 2015, 31:379391. https://doi.org/10.1002/rra.2743.

79. MacVicar BJ, Pigay H, Anderson A, Oberlin C,Comiti F, Pecorari E. Quantifying the temporaldynamics of wood in large rivers: new techniques for



https://doi.org/10.1016/j.jas.2012.02.022https://doi.org/10.1016/j.jas.2012.02.022https://doi.org/10.1177/0309133315615805https://doi.org/10.1177/0309133315615805https://doi.org/10.5194/esurf-4-359-2016https://doi.org/10.5194/esurf-4-359-2016http://ccwu.me/vsfm/https://doi.org/10.5194/tc-9-1-2015https://doi.org/10.1029/2011JF002289https://doi.org/10.1029/2011JF002289https://doi.org/10.1016/j.geomorph.2014.01.006https://doi.org/10.1016/j.geomorph.2014.01.006https://doi.org/10.1016/j.geomorph.2015.05.008https://doi.org/10.1016/j.geomorph.2015.05.008https://doi.org/10.1002/rra.1358https://doi.org/10.1002/rra.1358https://doi.org/10.1002/aqc.774https://doi.org/10.1007/s11157-014-9357-zhttps://doi.org/10.1007/s11157-014-9357-zhttps://doi.org/10.1002/rra.2879https://doi.org/10.1002/rra.3040https://doi.org/10.1002/rra.3040https://doi.org/10.1002/rra.3047https://doi.org/10.1002/rra.3047https://doi.org/10.3390/s151127969https://doi.org/10.1002/hyp.10597https://doi.org/10.1002/hyp.10597https://doi.org/10.1002/rra.2743https://doi.org/10.1002/rra.2743

wood tracking, monitoring, surveying, and dating.Earth Surf Process Landf 2009, 34:20312046.https://doi.org/10.1002/esp.1888.

80. Flynn KF, Chapra SC. Remote sensing of submergedaquatic vegetation in a shallow non-turbid river usingan unmanned aerial vehicle. Remote Sens (Basel)2014, 6:1281512836. https://doi.org/10.3390/rs61212815.

81. Husson E, Hagner O, Ecke F. Unmanned aircraft sys-tems help to map aquatic vegetation. Appl Veg Sci2014, 17:567577. https://doi.org/10.1111/avsc.12072.

82. Michez A, Piegay H, Jonathan L, Claessens H,Lejeune P. Mapping of riparian invasive species withsupervised classification of unmanned aerial system(UAS) imagery. Int J Appl Earth Observ Geoinform2016, 44:8894. https://doi.org/10.1016/j.jag.2015.06.014.

83. Puttock AK, Cunliffe AM, Anderson K, Brazier RE.Aerial photography collected with a multirotor dronereveals impact of Eurasian beaver reintroduction onecosystem structure. J Unmanned Vehicle Syst 2015,3:123130. https://doi.org/10.1139/juvs-2015-0005.

84. Flener C, Vaaja M, Jaakkola A, Krooks A,Kaartinen H, Kukko A, Kasvi E, Hyypp H,Hyypp J, Alho P. Seamless mapping of river chan-nels at high resolution using mobile LiDAR andUAV-photography. Remote Sens (Basel) 2013,5:63826407. https://doi.org/10.3390/rs5126382.

85. Bagheri O, Ghodsian M, Saadatseresht M. Reachscale application of UAV + SfM methods in shallowrivers: hyperspatial bathymetry. In: The InternationalArchives of the Photogrammetry, Remote Sensingand Spatial Information Sciences, Volume XL-1/W5,International Conference on Sensors and Models inRemote Sensing and Photogrammetry, Kish Island,Iran, 2325 November, 2015. https://doi.org/10.5194/isprsarchives-XL-1-W5-77-2015

86. Merriam-Webster. Available at: http://www.merriam-webster.com/dictionary/continuum (AccessedSeptember 29, 2016).

87. Vezza P, Astegiano L, Fukuda S, Lingua A,Comoglio C, Palau-Salvador G. Using structure frommotion techniques to describe and evaluate instreamphysical habitat. In: Extended Abstract, 11th Interna-tional Symposium on Ecohydraulics, Melbourne,Australia, 712 February, 2016.

88. Rippin DM, Pomfret A, King N. High resolutionmapping of supra-glacial drainage pathways revealslink between micro-channel drainage density, surfaceroughness and surface reflectance. Earth Surf ProcessLandf 2015, 40:12791290. https://doi.org/10.1002/esp.3719.

89. Clapuyt F, Vanacker V, Van Oost K. Reproducibilityof UAV-based earth topography reconstructionsbased on structure-from-motion algorithms.

Geomorphology 2016, 260:415. https://doi.org/10.1016/j.geomorph.2015.05.011.

90. Dietrich JT. Bathymetric structure-from-motion:extracting shallow stream bathymetry from multi-view stereo photogrammetry. Earth Surf ProcessLandf 2017, 42:355364. https://doi.org/10.1002/esp.4060.

91. Vericat D, Batalla RJ. Morfodinmica Fluvial. In:Batalla RJ, Tena A, eds. Procesos Hidro-sedimentarios Fluviales. Lleida: Editorial Mile-nio; 2016.

92. Hedger R, Foldvik A, Gjelland K. Drone remote sen-sing of river bed substrate size. In: Session VI, SmallUAS for Environmental Research Conference, Uni-versity of Worcester, UK, 2829 June, 2016.

93. Woodget AS, Visser F, Maddock IP, Carbonneau P.Quantifying fluvial substrate size using hyperspatialresolution UAS imagery and SfM-photogrammetry.In: Extended Abstract, 11th International Symposiumon Ecohydraulics, Melbourne, Australia, 712February, 2016.

94. Woodget AS, Austrums R. (in press) Subaerial gravelsize measurement using topographic data derivedfrom a UAV-SfM approach. Earth Surface Processesand Landforms doi: 10.1002/esp.4139

95. Feurer D, Bailly J-S, Puech C, Le Coarer Y, Viau AA.Very-high-resolution mapping of river-immersedtopography by remote sensing. Progr Phys Geogr2008, 32:403419.

96. Detert M, Weitbrecht V. A vision for contactlesshydraulic measurements. In: Session III, Small UASfor Environmental Research Conference, Universityof Worcester, UK, 2829 June, 2016.

97. De Haas T, Ventra D, Carbonneau P, Kleinhans MG.Debris flow dominance of alluvial fans masked byrunoff reworking and weathering. Geomorphology2014, 217:165181. https://doi.org/10.1016/j.geomorph.2014.04.028.

98. Sieberth T, Wackrow R, Chandler JH. 2013. Auto-mation isolation of blurred images from UAV imagesequences. In: International Archives of Photogram-metry, Remote Sensing and Spatial InformationSciences, vol. XL-1/W2, 361-366

99. Hugenholtz C, Brown O, Walker J, Barchyn TE,Nesbit P, Kucharczyk M, Myshak S. Spatial accuracyof UAV-derived orthoimagery and topography: com-paring photogrammetric models processed with directgeo-referencing and ground control points. Geoma-tica 2016, 70:2130. https://doi.org/10.5623/cig2016-102.

100. Tonkin TN, Midgley NG. Ground-control networksfor image based surface reconstruction: an investiga-tion of optimum survey designs using UAV derivedimagery and structure-from-motion photogrammetry.Remote Sens (Basel) 2016, 8:786. https://doi.org/10.3390/rs8090786.



https://doi.org/10.1002/esp.1888https://doi.org/10.3390/rs61212815https://doi.org/10.3390/rs61212815https://doi.org/10.1111/avsc.12072https://doi.org/10.1111/avsc.12072https://doi.org/10.1016/j.jag.2015.06.014https://doi.org/10.1016/j.jag.2015.06.014https://doi.org/10.1139/juvs-2015-0005https://doi.org/10.3390/rs5126382https://doi.org/10.5194/isprsarchives-XL-1-W5-77-2015https://doi.org/10.5194/isprsarchives-XL-1-W5-77-2015http://www.merriam-webster.com/dictionary/continuumhttp://www.merriam-webster.com/dictionary/continuumhttps://doi.org/10.1002/esp.3719https://doi.org/10.1002/esp.3719https://doi.org/10.1016/j.geomorph.2015.05.011https://doi.org/10.1016/j.geomorph.2015.05.011https://doi.org/10.1002/esp.4060https://doi.org/10.1002/esp.4060info:doi/10.1002/esp.4139https://doi.org/10.1016/j.geomorph.2014.04.028https://doi.org/10.1016/j.geomorph.2014.04.028https://doi.org/10.5623/cig2016-102https://doi.org/10.5623/cig2016-102https://doi.org/10.3390/rs8090786https://doi.org/10.3390/rs8090786

101. Eltner A, Mulsow C, Maas H-G. Quantitative meas-urement of soil erosion from TLS and UAV data. In:International Archives of the Photogrammetry,Remote Sensing and Spatial Information Sciences,Volume XL-1/W2, UAV-g2013, Rostock, Germany,46 September, 2013.

102. Overstreet BT, Legleiter CJ. Removing sun glint fromoptical remote sensing images of shallow rivers. EarthSurf Process Landf 2017, 42:318333. https://doi.org/10.1002/esp.4063.

103. Civil Aviation Authority. Available at: https://www.caa.co.uk/Commercial-industry/Aircraft/Unmanned-aircraft/Unmanned-Aircraft/ (Accessed March29, 2016).

104. Federal Aviation Authority. Available at: https://www.faa.gov/uas/ (Accessed January 26, 2017).

105. European Aviation Safety Agency. Prototype Com-mission Regulation on Unmanned Aircraft Opera-tions. Available at: https://www.easa.europa.eu/easa-and-you/civil-drones-rpas (Accessed January26, 2017).

106. Mancini F, Dubbini M, Gattelli M, Stecchi F,Fabbri S, Gabianelli G. Using unmanned aerial vehi-cles (UAV) for high-resolution reconstruction oftopography: the structure from motion approach oncoastal environments. Remote Sens (Basel) 2013,5:68806898. https://doi.org/10.3390/rs5126880.

107. Gmez-Gutirrez A, de Sanjos-Blasco JJ, de Matas-Bejarano J, Berenguer-Sempere F. Comparing twophoto-reconstruction methods to produce high

density point clouds and DEMs in the Corral delVeleta Rock Glacier (Sierra Nevada, Spain). RemoteSens (Basel) 2014, 6:54075427. https://doi.org/10.3390/rs6065407.

108. Eltner A, Baumgart P, Maas H-G, Faust D. Multi-temporal UAV data for automatic measurement of rilland interrill erosion on loess soil. Earth Surf ProcessLandf 2015, 40:741755. https://doi.org/10.1002/esp.3673.

109. Lague D, Brodu N, Leroux J. Accurate 3D compari-son of complex topography with terrestrial laser scan-ner: application to the Rangitikei canyon (N-Z).ISPRS J Photogramm Remote Sens 2013, 82:1026.https://doi.org/10.1016/j.isprsjprs.2013.04.009.

110. Chandler JH, Fryer JG, Jack A. Metric capabilities oflow-cost digital cameras for close range surface meas-urement. Photogramm Rec 2005, 20:1226.

111. Wackrow R, Chandler JH. Minimising systematicerror surfaces in digital elevation models usingoblique convergent imagery. Photogramm Rec 2011,26:1631. https://doi.org/10.1111.j.1477-9730.2011.00623.x.

112. Hervouet A, Dunford R, Piegay H, Belletti B,Tremelo M-L. Analysis of post-flood recruitment pat-terns in braided-channel rivers at multiple scalesbased on an image series collected by unmanned aer-ial vehicles, ultra-light aerial vehicles, and satellites.GISci Remote Sens 2011, 48:5073. https://doi.org/10.2747/1548-1603.48.1.50.



https://doi.org/10.1002/esp.4063https://doi.org/10.1002/esp.4063https://www.caa.co.uk/Commercial-industry/Aircraft/Unmanned-aircraft/Unmanned-Aircraft/https://www.caa.co.uk/Commercial-industry/Aircraft/Unmanned-aircraft/Unmanned-Aircraft/https://www.caa.co.uk/Commercial-industry/Aircraft/Unmanned-aircraft/Unmanned-Aircraft/https://www.faa.gov/uas/https://www.faa.gov/uas/https://www.easa.europa.eu/easa-and-you/civil-drones-rpashttps://www.easa.europa.eu/easa-and-you/civil-drones-rpashttps://doi.org/10.3390/rs5126880https://doi.org/10.3390/rs6065407https://doi.org/10.3390/rs6065407https://doi.org/10.1002/esp.3673https://doi.org/10.1002/esp.3673https://doi.org/10.1016/j.isprsjprs.2013.04.009https://doi.org/10.1111.j.1477-9730.2011.00623.xhttps://doi.org/10.1111.j.1477-9730.2011.00623.xhttps://doi.org/10.2747/1548-1603.48.1.50https://doi.org/10.2747/1548-1603.48.1.50

Drones and digital photogrammetry: from classifications to continuums for monitoring river habitat and hydromorphologyINTRODUCTIONThe Importance of Physical River Habitat MonitoringDrones (UAVs, UAS, and RPAS)Digital Photogrammetry (Structure-from-Motion)

CLASSIFICATIONS OF PHYSICAL RIVER HABITATCase Study 1: San Pedro River, ChileBackgroundApplication of Drones and Digital Photogrammetry

CONTINUUMS OF PHYSICAL RIVER HABITATCase Study 2: River Teme, UKBackgroundApplication of Drones and Digital PhotogrammetryData Analysis

EVALUATIONData AcquisitionData Processing

CONCLUSIONACKNOWLEDGMENTSREFERENCES

drones and digital photogrammetry: from classifications to …eprints.worc.ac.uk/5481/1/woodget et...

Documents