Recent remote sensing applications for hydro and morphodynamic monitoring and modelling

It is not new to recognise that data from remote sensing platforms is transforming the way we characterise and analyse our environment. The ability to collect continuous data spanning spatial scales now allows geomorphological research in a data rich environment and this special issue (coming just 7 years after the 2010 special issue of ESPL associated with the remote sensing of rivers) highlights the considerable research effort being made to exploit this information, into new understanding of geomorphic form and process. The 2010 special issue on the remote sensing of rivers noted that fluvial remote sensing papers made up some 14% of the total river related papers in ESPL. A similar review of the papers up to 2017 reveals that this figure has increased to around 25% with a recent proliferation of articles utilising satellite based data and structure from motion derived data. It is interesting to note, however that many studies published to date are proof of concept, concentrating on confirming the accuracy of the remotely sensed data at the expense of generating new insights and ideas on fluvial form and function. Data is becoming ever more accurate and researchers should now be concentrating on analysing these early data sets to develop increased geomorphic insight challenging paradigms and moving the science forward. The prospect of this occurring is increased by the fact that many of the new remote sensed platforms allow accurate spatial data to be collected cheaply and efficiently. This is providing the individual researcher or small research grouping with tremendous opportunity to move the science of fluvial geomorphology forward unconstrained to a large degree of the need to secure substantial research funding. Fluvial geomorphologists have never before been in such a liberated position! As techniques and analytical skills continue to improve it is inevitable that Marcus and Fondstad's (2010) prediction that remotely sensed data will revolutionising our understanding of geomorphological form and process will prove true, altering our ideas on the very nature of system functioning in the process.


Introduction
Remote sensing is leading to a radical transformation in the way we map and analyse our environment (Marcus and Fonstad, 2010) with new instrumentation and software allowing for the collection and analysis of continuous data spanning spatial scales previously unachievable using earlier technologies. We can now research in a data rich environment and are beginning to exploit this transformation in available integrated scales of information, into new understanding of geomorphic form and process, helping to truly cover the range of fluvial structure and function advocated by Lane et al. (1994).
The 2010 special issue of Earth Surface Processes and Landforms (ESPL) associated with the remote sensing of rivers contained 12 seminal articles on the emerging use of remote sensing platforms in geomorphology covering system structure, evolution and process measurement. In their review article the editors concluded that 'the time for more widespread application of river remote sensing techniques is now' Fonstad, 2010, p. 1867). Then Marcus and Fonstad (2010) noted that fluvial remote sensing articles made up some 14% of the total river related articles in ESPL. Seven years have passed during which time geomorphological research utilizing remotely sensed data has increased further with improvements in data acquisition and processing facilitating the analysis of increasingly detailed and complex environmental data sets. Table I illustrates the publication trend since 2010 reviewing the type of remote sensing employed. What is immediately striking is that around a quarter of those papers published in the journal make use of remote sensing of river environments, a significant increase over only seven years. Also of note are the recent proliferation of articles utilizing unmanned aerial platforms for data collection and structure from motion photogrammetry.
Since the ESPL review article by Marcus and Fonstad (2010), the general uptake of remote sensing methods in physical geography in general has been reviewed by Tarolli (2014) and for fluvial systems by Carbonneau and Piégay (2012). Gilvear and Bryant (2016) also provide a useful review of satellite-based remotely sensed data. This article leads the special issue on 'Remote sensing applications for hydro and morphodynamic monitoring and modelling' that brings together a series of articles presented at the 11th International Symposium on Ecohydraulics, held in Melbourne, February 2016. It is abundantly clear that Marcus and Fonstad (2010) were correct in their prediction, remotely sensed data are revolutionizing our understanding of geomorphological form and process altering our ideas on the very nature of system functioning as it does so in the process.

Early work
The use of remote sensed data is far from new. What has changed is the type of remote sensed data being used and the manner in which these data have been collected. Some of the earliest use of remote sensed data centred around oblique and aerial photography (see review by Gilvear and Bryant, 2003). Notable early work included elucidating information on alpine glacier retreat (LaChapelle, 1962). In a fluvial context aerial photography has been used to study floodplain geometry (Lewin and Manton, 1975), bank retreat (Williams et al., 1979), historical meander development (Hooke, 1984), and barform change (Warburton et al., 1993). Spaceborne imagery has also proved usable on large rivers with Salo et al. (1986) mapping planform movement on the Amazon system.
Since these early studies we have seen the emergence of optical image analysis utilizing parameters such as contrast and spectral signatures to derive a variety of fluvial variables. For example, Hicks et al. (2000) used spectral analysis to derive flow depth and Hardy et al. (1994) and Winterbottom and Gilvear (1997) used the approach to map hydraulic habitat.
Passive and active laser-based instruments have emerged as a remote sensing tool with light detection and ranging (LiDAR)-based studies (Heritage and Large, 2009). Photogrammetric approaches have evolved, with vertical imagery used to study areas over 1 km 2 Smith et al., 2016), and oblique photogrammetry used to map gravel bar surfaces (Heritage et al., 1998).

Recent Research
Articles published in the 2010 ESPL special issue on fluvial remote sensing reflects well the remote sensing techniques of the time with aerial and terrestrial LiDAR articles dominating and the use of aerial imagery being largely restricted to analysis of optical properties. Table I illustrates clearly how other techniques, most notably those based on photogrammetry have emerged to dominate the literature assisted by the development of small aerial platforms to facilitate the data collection phase.
Optical approaches Legleiter (2013) mapped river depth from publicly available US National Agricultural Imagery Programme aerial images but found issues with geo-referencing errors and a coarser spatial resolution. Optical approaches have moved towards using image texture and reflectance spectra as a discriminator to differentiate and map fluvial character and Belletti et al. (2015) review optical approaches to classify river hydromorphology. A detailed  investigation of bed sedimentology has been achieved at the catchment scale using image texture analysis to generate a map of median grain size for submerged and dry fluvial surfaces along an 80 km length of the Sainte-Marguerite River, Québec, Canada (Carbonneau et al., 2004). Flood inundation area has been mapped using Spaceborne imagery (Pierdicca et al., 2013) including Landsat (Feyisa et al., 2014), although issues have been shown to arise with emergent vegetation (Silva et al., 2008). Multispectral imaging has been exploited by Wright et al. (2000) to map morphological units and Legleiter (2012) utilized similar data to derive channel bathymetry as part of a combined study with LiDAR, reporting that pool depths were generally underestimated with turbidity strongly influencing the data quality. Multi feature classification has also been attempted for river hydromorphology using high-resolution small unmanned air vehicle (sUAV) imagery (Casado et al., 2015), whilst Bentley et al. (2016) utilized Google Earth and Bing aerial imagery to generate long-reach digital elevation models (DEMs) for twodimensional (2D) hydraulic habitat modelling.
Imagery has also been used to infer process. Pioneering early work by Mertes et al. (1993) estimated suspended sediment concentrations for the Amazon floodplain after reflectance correction and Gomez et al. (1995) derived suspended sediment characteristics for the Mississippi using satellite data. More recently Bywater-Reyes et al. (2017) quantified the impact of forestry operations on suspended sediment yields from five headwater sub-catchments of the Trask River, Oregon.

LiDAR
Aerial LiDAR has increased our understanding of earth systems most commonly over landscape scales achieving decimetre scale vertical elevation accuracies (Smith and Vericat, 2014). The application of aerial LiDAR to fluvial systems saw a rapid increase from around 2000 due to the wider availability of hardware, and the introduction of terrestrial LiDAR systems (for reviews see Hohenthal et al., 2011;Milan and Heritage, 2012;Heritage and Large, 2009). Aerial LiDAR has the advantage over photography in that it can penetrate vegetation to provide information below the canopy (Glennie et al., 2013), although penetration can fail where vegetation is too dense (Malinowski et al., 2016). Aerial LiDAR data is being used in a range of fluvial applications including (1) hydrogeomorphic assessment at the reach (Charlton et al., 2003) and catchment-scale (Biron et al., 2013), (2) detection and quantification of logjams in lowland rivers (Abalharth et al., 2015), (3) parameterization of spatial vegetation roughness on floodplains for 2D hydraulic modelling (e.g. Bertoldi et al., 2011;Straatsma and Baptist, 2008;Antonarakis et al., 2008;Abu-Aly et al., 2014), (4) delineation of water surface and flood innundation extent based on laser pulse reflectance values (Crasto et al., 2015;Malinowski, 2016), and (5) provision of boundary conditions for modelling the geomorphic impacts of catastrophic events (Thompson and Croke, 2013).
Sequential aerial LiDAR surveys are also increasingly being used to (6) detect morphological changes in rivers (e.g. Lallias-Tacon et al., 2017;Nelson and Dubé, 2016;Thompson et al., 2016;Milan et al., 2018). Work has also integrated historical aerial photographs and LiDAR to extend the historic DEM time-series, allowing a much longer record of historical spatial patterns of morphological change to be assessed (De Rose and Basher, 2011).
Full waveform LiDAR has been tested (Schofield et al., 2016) and used extensively in forest remote sensing (Wulder et al., 2012). In a fluvial context, Kinzel et al. (2007) surveyed a shallow, braided, sand-bedded river system using an experimental terrestrial algorithm to successfully approximate the position of the river bed. Mean signed errors of 0.18 m were reported across on exposed sand for two surveys and 0.18 m and 0.24 m on submerged sand. Pan et al. (2015) have used high resolution airborne full waveform LiDAR to estimate shallow river bathymetry, and Höfle et al. (2009) used the characteristics of full waveform LiDAR to distinguish and map areas of water.
Terrestrial LiDAR systems (TLSs) are more restricted in their coverage but achieve greater accuracy (Smith and Vericat, 2014;Williams et al., 2014), negate the need for high cost platforms such as aircraft, and can be used to retrieve data with a higher temporal frequency. Milan et al. (2007) have demonstrated the application of TLSs for morphological change detection in rivers, mapping rapid feature-scale change across a 6 km 2 area of the proglacial zone of Glacier du Ferpècle and Mont Miné, Switzerland. In addition, Brasington et al. (2012) achieved accurate morphologic unit and bed roughness mapping over a 1 km reach of the braided River Feshie, Scotland. Mobile platforms have been shown to increase the spatial range of TLSs in fluvial settings, with Alho et al. (2009), Lotsari et al. (2014 and Leyland et al. (2017) using boat-based survey, Vaaja et al. (2011) using cart-based survey, and Williams et al. (2013) using an amphibious all-terrain vehicle.
A number of studies have also been fusing techniques, with Leyland et al. (2017) integrating mobile laser scanning and multibeam echo sounding with acoustic Doppler current profiling to directly measure changes in river bank and bed, and Williams et al. (2013) and Lotsari et al. (2014) coupling high density acoustic doppler profiling with TLS to provide information on flow and sediment transport fields.
Terrestrial laser scanning has been used to characterize spatial patterns of grain size and roughness for bar surfaces Hodge et al., 2009aHodge et al., , 2009b and to characterize pebble clusters (Entwistle et al., 2007), through the application of point cloud analysis algorithms. Milan et al. (2010) have used first return red-wavelength TLS to map instream habitat based upon water surface roughness characterization. Smith et al. (2012) have also demonstrated limited bathymetric capabilities of red-wavelength terrestrial laser scanning in calm shallow water. Terrestrial laser scanning has also been used to characterize riparian vegetation to assist in spatial roughness parameterization for 2D hydraulic modelling, which offers potential improvements to flood prediction (Antonarakis et al., 2009(Antonarakis et al., , 2010Manners et al., 2013).
Green-wavelength, bathymetric LiDAR has been shown to be capable of whole system fluvial mapping, capturing both terrestrial and bathymetric surfaces as McKean et al. (2009) and Kinzel et al. (2013) have demonstrated using EAARL (the experimental advanced airborne research LiDAR) narrowbeam aquatic-terrestrial LiDAR. Bathymetric LiDAR has gained a strong presence in the literature around estuaries, with Valle et al. (2011) successfully modelling estuarine habitats using airborne bathymetric data. In fluvial systems bathymetric LiDAR however appear less effective for shallow water environments which are often the focus of geomorphic river investigations (e.g. gravel-bed rivers), with studies generally restricted to depths greater than 0.5 m Bailly et al., 2010;Milan and Heritage, 2012). Pan et al. (2015) suggest that full waveform LiDAR may offer improved shallow water bathymetry estimates.
Photogrammetry and structure-from-motion (SfM) methods Terrestrial photogrammetry continues to be used in fluvial studies with sub-centimetre scale resolution (Lane, 2000). Errors on survey data are reported by Smith and Vericat (2015 (2010) has achieved mapping over 32 km of river using this technique. Bird et al. (2010) recognized the limited areal coverage achieved using conventional oblique photography extending their coverage through the use of a camera mounted on a long pole to map channel bed elevation linked to pool-riffle sequences.
Structure-from-motion (SfM) algorithms can automatically generate three-dimensional (3D) data from a set of photographs of the area interest. The algorithms retrieve the various camera positions and orientations used for a reconstruction. This technique coupled with the rapid development of image acquisition utilizing sUAVs allow rapid construction of accurate orthophotographs and DEMs from the images captured and has led to a recent significant increase in geomorphological research utilizing this combined approach (Table I). Westoby et al. (2012) highlight the effectiveness of SfM for geoscience applications essentially focusing on the low-cost, user-friendly photogrammetric technique for obtaining high-resolution datasets at a range of scales. Smith et al. (2016) note that many of the publications relating to SfM in physical geography are at present proof of concept studies, with many using the orthophotograph produced as part of the SfM process to detect features (e.g. Entwistle and Heritage, 2017).
Derivation of accurate DEMs, using photogrammetric methods, across fluvial environments has proved challenging due to their complexity and the variable presence of vegetation and water. Exposed surfaces have been successfully mapped by many researchers (e.g. Fonstad et al., 2013;Smith and Vericat, 2014;Entwistle and Heritage, 2017), with the data becoming increasing used for subsequent, multi scale analysis (Smith and Vericat, 2014). Javernick et al. (2014) used SfM to carry out a detailed survey and subsequent error analysis of sub-meter resolution DEMs from two contiguous reaches of the braided Ahuriri River, New Zealand; mapping 3.3 km of watercourse in total. Imagery was acquired from high altitude (600-800 m from a helicopter), achieving mean vertical surface errors of ±0.1 m across non-vegetation areas.
Roughness metrics have also been derived from SfM photogrammetry; across large exposed (Leon et al., 2015) and submerged areas (Entwistle and Heritage, 2017). Most recently Carbonneau and Dietrich (2017) have employed direct georeferencing techniques commonly used to position airborne LiDAR to generate ground elevation models with average mean residual errors of only ±0.06 m (equivalent to 1% of the flying altitude) utilising an sUAV with GPS position accuracy of 2.5 m. This success opens the way for DEM generation without the use of ground control points; further increasing the efficiency and ease of data collection using sUAV platforms. Woodget et al. (2015), Woodget and Austrums (2017), Dietrich (2017) and Entwistle and Heritage (2017) evaluate and expand on approaches to accurately survey bathymetric data using SfM.
Issues with data acquisition and processing have been reported, with a useful summary provided by Woodget et al. (2015). Sun glint issues are known to occur with remote sensing images (Legleiter, 2013) and UAV imagery Zeng et al., 2017) and this may be reduced by flying when the sun is at its highest (Dietrich, 2017) and through glint removal algorithms (Overstreet and Legleiter, 2017).
Other remote sensing methods Fausch et al. (2002) demonstrated the value of satellite remote sensing in characterizing the riverscape, whilst Gilvear and Bryant (2016) provide an excellent and up to date review of spaceborne remote sensing techniques. They conclude that remotely sensed data has wide application in detecting and mapping landforms, quantifying temporal change in fluvial landforms and elucidating on controlling processes. Bizzi et al. (2016) also review the advances in the use of remote sensed data to characterize fluvial hydromorphology concluding that such data could benefit the Water Framework Directive assessment process across Europe. Ecohydraulic information has also been derived from satellite survey of channel bathymetry and sedimentology and subsequent modelling using these data have facilitated fish habitat assessment at the river scale (Bergeron and Carbonneau, 2012). Of particular note is the increasing availability of Landsat data and Google Earth engine imagery which is now providing unprecedented information on land-surface changes. For example, Yousefi et al. (2016) succeeded in mapping and quantifying morphometric change over 20 meandering reaches of the Karoon River in Iran between 1989 and 2008.

Derivation of fluvial metrics and Point-cloud Modelling
Change detection and morphological characterization of fluvial systems is increasingly being dominated by the use of the DEM or digital terrain model (DTM) (Williams, 2012). Remote sensing methodologies (e.g. LiDAR and SfM) are used to collect point cloud data that can be imported into spatial analyses software, where the data can be interpolated to produce a modelled surface. A number of factors can introduce error into the DEM including survey point quality, sampling strategy, surface composition, topographic complexity and interpolation methods (Milan et al., 2011). In fluvial geomorphology a common approach has been to re-survey rivers following events that may have changed the topography, such as large floods (Milan, 2012), and subtract the successive DEMs from one another to produce DEMs of Difference (DoDs) (Williams, 2012), that require error filtering for each of the DEMs involved. This approach has led to a step-change in our ability to understand spatial patterns of morphological response to forcing factors within fluvial systems (e.g. Pasternack and Wyrick, 2017). Some key developments have been made towards error assessment when using DEMs for fluvial studies in the last decade. Milan et al. (2011), have shown how error is spatially variable across a DEM, and often linked to local topographic variability (form roughness). These provide a workflow that allow 'Levels of Detection' to be applied to DEMs, that can be used to filter error in a spatially distributed manner.  also show the importance of considering survey strategy when collecting data, particularly when using more traditional field data retrieval approaches [total station and differential GPS (dGPS)]. A further widely used approach is that of Wheaton et al. (2010), who use fuzzy set theory coupled with a method for discriminating DoD uncertainty on the basis of the spatial coherence of erosion and deposition using Bayes Theorem. These workers suggest that various components of elevation uncertainty are collinear variables and do not exhibit a single monotic relationship to elevation uncertainty, and therefore apply a heuristic approach to the problem. More recent advances in point cloud analyses involve cloud-to-cloud comparisons rather than DEM differencing (Lague et al., 2013), which has proved particularly useful in change detection analyses of vertical surfaces such as river banks (e.g. Leyland et al., 2017).
Advances have also been made in the spatial interrogation of point cloud data at a range of scales. Cavalli et al. (2008) have also interrogated LiDAR DEMs to delineate channel types, through analysis of the residuals of elevations orthogonal to the regression line drawn along the channel profile, and the 2286 N. ENTWISTLE ET AL.
standard deviation of the local slope, and has proved particularly useful in delineation step-pool and riffle-pool reaches. The dense and highly accurate point clouds produced from terrestrial LiDAR have allowed detailed analysis at the grain-scale, including grain roughness and structure (Entwistle and Fuller, 2009;Hodge et al., 2009aHodge et al., , 2009b. Hodge et al. (2009a) analysed the distribution of surface elevations (1 mm spacing on small patches of gravel) and surface slope and aspect to provide information on grain packing and the role of grain size in determining surface structure, and grain orientation and imbrication.  and Entwistle and Fuller (2009) found strong relationships between twice the standard deviation of the surface elevations in a small moving window over the point cloud, and grain size; most closely with the c-axis of gravel clasts, which are most closely associated with particle protrusion. These approaches are permitting full spatial description of the grain size and roughness of bar surfaces and have the potential to improve roughness characterization for hydraulic modelling (Milan, 2009), and potentially improve flood flow prediction. Demarchi et al. (2017) concentrate on the Piedmont region of the Italian Apennines, in particular the Po River. Remote sensing data (near red orthophotographs and LiDAR) was obtained for the entire region (25 000 km 2 ) of the land surface with a resolution of 0.4 m. The authors used a combination of very high resolution near infrared aerial imagery and low-resolution LiDAR to provide a hydromorphological characterization of rivers at a regional scale, offering an approach that may be used to answer basin-scale questions. Data were interrogated at the pixel-level, applying a 'fluvial corridor' toolbox (Roux et al., 2015) to the detrended DTM, to delineate morphological features, and an object-based classification approach to identify and delineate 'riverscape' units from the near infrared imagery. The work allowed 1700 km 2 of floodplains to be mapped and delineated into geomorphological meaningful units, and allowed the production of a database (HyMo DB), where hierarchical clustering was used to classify river reaches from the database. Sun glint (specular reflection of sun light from the water surface) can pose significant problems with regards to extracting river habitat metrics from remotely-sensed images of rivers, as sun glint often results in unusually bright pixels, and subsequent loss of data. Glint removal has been carried out in marine environments, however these techniques do not work well for shallow rivers. Overstreet and Legleiter (2017) detail the development, application and testing of a method for removing sun glint from shallow areas of the bed, from remotely sensed imagery. The technique overcomes over-correction (removal of too much reflectance) inherent in previous approaches, through accounting for non-neglible water-leaving near infrared radiance. The new approach develops a depth-assisted method for sun glint removal, requiring field measurements of depth and imagery that includes at least one near infrared band. Example data for the gravel-bed Snake River, showed improved r 2 values for depth prediction (from 0.66 to 0.76), when compared to previous approaches that fail to work well in shallow water environments such as gravel-bed rivers. Pasternack and Wyrick (2017) provide a detailed assessment of topographic change along a 37 km reach of the coarsegrained, alluvial Lower Yuba River; a regulated river also influenced by mining, and an extreme flood over the period of the investigation (1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008). The study compared point cloud DEMs created using recent ground-and boat-based LiDAR, with earlier contour-based DEMs created using photogrammetry (0.6 m resolution). The morphological budget presented is for the full 37 km long reach, following the application of state-of-the-art error filtering (Wheaton et al., 2010). The analysis of morphological change across multiple spatial scales provides useful insight into the morphodynamics of the river. In particular, the results that emerge from analysing how scour/fill varies along the river, and how morphological change volumes and mean vertical changes vary significantly for different morphological units, demonstrating how geomorphic change detection can be used to gain insight into morphological evolution. The longevity of remnant mining sediment was estimated based on a multi-scalar approach to quantifying geomorphic and associated volumetric change. Marteau et al. (2017) use SfM photogrammetry to produce DEMs and assess geomorphic changes associated with a river restoration project in north west UK. The approach uses a low cost GoPro camera with a fish-eye lens attached to a drone. SfM photogrammetry is becoming a major technique in geomorphology in the capture of morphological change data, and for the first time this article demonstrates its' potential as a tool in monitoring channel response following river restoration; in this case an artificial channel created to restore the connection between two rivers. The authors use the SfM data to produce DEMs from a point cloud, and subtract these over a series of surveys to identify spatial patterns of scour and fill and report volumetric change, and carefully account for error. Although the use of fisheye lenses for photogrammetry has previously been criticized, the authors results are of high quality. Thumser et al. (2017) introduce a new method for remotely sensing surface velocity in rivers in real time. The approach uses UAV-based particle tracking (RAPTR-UAV), using a combination of floating, infrared light-emitting particles and a programmable embedded colour vision sensor attached to a UAV to simultaneously detect and track the positions of objects. The approach can rapidly collect and process position data in real-time, and has the potential to improve hydraulic model validation, and increase understanding of process and form within river channels. Wheaton et al. (2018) present a framework for the application of ecohydraulic fish habitat models at a range of scales, using salmon populations in the Columbia River basin (900+ sites in 12 watersheds) as an example. Readily available remotely-sensed data such as 10 m DEMs, geology and Landsat-derived vegetation layers, satellite and aerial imagery, coupled with reach-scale remotely-sensed data and at-a-site validation using more basic survey techniques (e.g. total stations), permit full geomorphic assessment and habitat modelling that transcends scale boundaries. The approach links habitat (defined by geomorphology, hydraulics, and water temperature dynamics), with population and life cycle modelling of the species in question. A conceptual and methodological toolbox is developed that operationalizes the notion that fish habitat should be studied at a landscape scale.

Conclusions
At the outset of this article we noted the value of remote sensing to generate data-sets rich in spatial information. The subsequent review of research has certainly shown this to be the case with sensors able to capture detailed datasets across often large areas allowing integrated analysis of form and process across a variety of spatial scales. Our investigation of the proliferation of remote sensing techniques in river research (Table I) Marcus and Fonstad (2010) noted the drivers for remote sensing data collection were research needs, through investigation of new technologies and greater engagement of fluvial geomorphologists with spatial data. It would appear from the subsequent seven years that the development of new technologies is exerting the greatest influence; Smith and Vericat (2014) conclude that many studies published to date are proof of concept, concentrating on confirming the accuracy of the new remote sensing approaches rather than using the data more fully to generate initial new insights and ideas on geomorphic form and function. This is understandable but also disappointing. It is important to recognize the value of the data sets collected so far in the context of the error that they contain in relation to natural surface variability. Entwistle and Heritage (2017) note that part of the remotely sensed error in water depth is likely due to the inherent variability of the river bed being studied which is also not being picked up by the reference survey technique (for example their theodolite survey). A practical recognition of the impact of such error dependent on the character of the environment being studied is more important. This is in no way an excuse for poor surveying but it will allow researchers greater flexibility in analysing these early data sets and in developing increased geomorphic insight. Marcus and Fonstad (2010) also concluded that the rise in remotely sensed fluvial research was serendipitous. Whilst this may be true to some degree, there appears to have been a genuine research driven desire to begin to exploit new remote sensed technologies with considerable research effort expended across many institutions and that individual desire appears to be continuing unabated assisted by the ability to utilize new remote sensed platforms cheaply and easily to obtain detailed and accurate spatial data. This is providing the individual researcher or small research grouping with tremendous opportunity to move the science of fluvial geomorphology forward unconstrained to a large degree of the need to secure substantial research funding. Fluvial geomorphologists have never before been in such a liberated position.
Many studies are also now developing sophisticated landscape models based on integrated survey techniques with the ability to work to common reference coordinates greatly aiding this process. In such a way the connectivity between terrestrial and aquatic zones is becoming increasingly understood (see for example Leyland et al., 2017). The advent of newer and more sophisticated techniques in remote sensing should, however, not preclude the use of more traditional approaches, especially given the proliferation of data available for these techniques too. For instance, Lisle (2006) acknowledged the great value of the historic imagery availability from Google Earth allowing change mapping based on increasingly frequent images and other satellite-based sources such as Landsat data providing unprecedented temporal change information to investigate land surface changes.
It is exciting also to view the possibility of increasing use of remote sensed data to drive further studies with the DEMs with the reviews of Fausch et al. (2002) and Gilvear and Bryant (2016) amply demonstrating the proliferation of available data from multiple satellite platforms. Importantly these data are proving of great value in process modelling as Javernick et al. (2014) demonstrated using SfM derived DEMs to conduct 2D hydraulic modelling.