**4.1 Alongshore variability**

Alongshore, the eastern extent of the BB survey recorded the highest magnitude topographic change. This is particularly evident in areas that experienced erosion of the full-frontal section of the foredune (**Figure 7A** and **B**). In these locations, the scarp line forms at or behind the former crest line, leading to slope failure,

#### **Figure 7.**

*A topographic change profile sampled from Brackley Beach (BB) east was measured from the pre-storm Dorian (SD) LiDAR DTM and and post-SD UAV DSM (A) BB east experienced significant scarping that eroded the frontal section of the foredune and lead to large volumetric losses (A and B). Aerial UAV images from pre-SD (C) and post-SD (D) show the removal of the stoss slope and formation of a steep scarp at the former crestline. Images from pre-SD (C) and post-SD (D).*

significant downcutting and volumetric loss (**Figure 7A** and **B**)*.* Only a small proportion of this eroded sediment was deposited on the upper beach surface suggesting that the majority of the sediment was moved further seaward, beyond the detectable range of our SfM DSM. Pre-SD, UAV images from July 2019 (**Figure 7C**) show a sparsely vegetated stoss slope extending 10 m seaward of the dune crest. Post-SD, UAV images from September 2019 show the stoss slope has been entirely eroded leaving behind a steep scarp face formed at the crest line (**Figure 7B** and **D**). The foredunes in this area consistently displayed stoss slopes exceeding the angle of repose for dry sand >34° and appeared to be temporarily maintained by surface moisture post-SD. Additional slumping of the surface is expected and may lead to further instability and reduction of dune height at the eastern extent of BB East.

The mid and western sections of BB recorded a continuous 1–2 m scarp at the base of the stoss slope, resulting in lower volumetric losses from the foredune in these areas (**Figure 8A** and **B**). At BB West, the transition between erosion and deposition occurs much closer to the dune toe, compared to BB east. Also, sediment deposition in this section of BB (**Figure 8A** and **B**) contained a high proportion of the volume eroded from the foredune (**Figure 8A**). In July 2019, prior to SD, the stoss slope of the foredune typically extended 13 m or more seaward of the crest (**Figure 8C**). Sediment accumulation was aided by the increased seaward extent of

> the vegetation during the summer months (**Figure 8C**) and promoted the intermittent growth of embryo dunes seaward of the established foredune (e.g., **Figure 8A** and **C**). Following SD in September 2019, the scarp line has eroded into the lower stoss slope and has removed any alongshore embryo dunes present pre-SD

> *Alongshore variability of antecedent (i.e., prior to the storm event) morphometry, measured form the LiDAR bathymetric and topographic survey at Brackley Beach (BB; A) east and BB west (B). BB east (A) displayed a higher sloping near shore, narrower beach (Bw) and dune width (Dw*Þ*, lower dune height (Dh*Þ*, higher slope* ð Þ *Ds , and lower D*ð Þ*<sup>v</sup> than BB west (B). These antecedent beach-dune metrics indicate that BB east was more*

*vulnerable to storm erosion that resulted in higher magnitude scarping in this area of BB.*

*Monitoring Storm Impacts on Sandy Coastlines with UAVs*

*DOI: http://dx.doi.org/10.5772/intechopen.91459*

The variability of storm impacts observed at BB were likely controlled by the antecedent morphology that existed prior to SD. From the LiDAR DTM, BB east was observed to have a higher nearshore slope and closer inner bar structure, in comparison to BB west (**Figure 9A** and **B**). This could have led to higher wave energy and erosive potential at BB East during SD, and stronger counter current moving sediment further seaward post-SD (e.g., **Figure 7A** and **8A**). Beach width ð Þ *Bw* at BB East was also significantly narrower than BB West and likely would have increased the exposure of the foredune to wave run up, despite a lower storm surge level relative to the dune toe (**Figure 9A** and **B**). In comparison to BB West, a combination of foredune metrics including narrower dune width ð Þ *Dw* , shorter dune heights ð Þ *Dh* , higher dune slopes ð Þ *Ds* , and lower dune volumes ð Þ *Dv* indicate that BB East was more vulnerable to storm induced erosion (**Figure 9A** and **B**). Foredunes that displayed a narrower *Dw* and high *Ds* at BB East experienced the highest magnitude of erosion resulting from major slope failures. While the foredune at BB West appeared to only lose a small proportion of the frontal *Dv*, a significant reduction of *Dv* at BB East could lead to further lowering of the foredune

making this area of BB increasingly vulnerable to future storm events.

UAV monitoring has allowed for the rapid assessment of Brackley Beach (BB), 11 days following Storm Dorian (SD). Projected 7–8 m significant wave heights and 1.2 m storm surge levels associated with SD resulted in a highly erosive post storm

(**Figure 8D**).

**Figure 9.**

**5. Discussion**

**79**

#### **Figure 8.**

*A topographic change profile sampled from Brackley Beach (BB) west was measured from the pre-storm Dorian (SD) LiDAR DTM and and post-SD UAV DSM (A). BB west experienced scarping at the base of the foredune and removal of embryo dunes (A and B). Aerial UAV images from pre-SD (C) and post-SD (D) show the removal of seaward dune deposits leaving behind a low continuous scarp alongshore.*

*Monitoring Storm Impacts on Sandy Coastlines with UAVs DOI: http://dx.doi.org/10.5772/intechopen.91459*

#### **Figure 9.**

significant downcutting and volumetric loss (**Figure 7A** and **B**)*.* Only a small pro-

(**Figure 7C**) show a sparsely vegetated stoss slope extending 10 m seaward of the dune crest. Post-SD, UAV images from September 2019 show the stoss slope has been entirely eroded leaving behind a steep scarp face formed at the crest line (**Figure 7B** and **D**). The foredunes in this area consistently displayed stoss slopes exceeding the angle of repose for dry sand >34° and appeared to be temporarily maintained by surface moisture post-SD. Additional slumping of the surface is expected and may lead to further instability and reduction of dune height at the

The mid and western sections of BB recorded a continuous 1–2 m scarp at the base of the stoss slope, resulting in lower volumetric losses from the foredune in these areas (**Figure 8A** and **B**). At BB West, the transition between erosion and deposition occurs much closer to the dune toe, compared to BB east. Also, sediment deposition in this section of BB (**Figure 8A** and **B**) contained a high proportion of the volume eroded from the foredune (**Figure 8A**). In July 2019, prior to SD, the stoss slope of the foredune typically extended 13 m or more seaward of the crest (**Figure 8C**). Sediment accumulation was aided by the increased seaward extent of

*A topographic change profile sampled from Brackley Beach (BB) west was measured from the pre-storm Dorian (SD) LiDAR DTM and and post-SD UAV DSM (A). BB west experienced scarping at the base of the foredune and removal of embryo dunes (A and B). Aerial UAV images from pre-SD (C) and post-SD (D) show the*

*removal of seaward dune deposits leaving behind a low continuous scarp alongshore.*

portion of this eroded sediment was deposited on the upper beach surface suggesting that the majority of the sediment was moved further seaward, beyond the detectable range of our SfM DSM. Pre-SD, UAV images from July 2019

*Spatial Variability in Environmental Science - Patterns, Processes, and Analyses*

eastern extent of BB East.

**Figure 8.**

**78**

*Alongshore variability of antecedent (i.e., prior to the storm event) morphometry, measured form the LiDAR bathymetric and topographic survey at Brackley Beach (BB; A) east and BB west (B). BB east (A) displayed a higher sloping near shore, narrower beach (Bw) and dune width (Dw*Þ*, lower dune height (Dh*Þ*, higher slope* ð Þ *Ds , and lower D*ð Þ*<sup>v</sup> than BB west (B). These antecedent beach-dune metrics indicate that BB east was more vulnerable to storm erosion that resulted in higher magnitude scarping in this area of BB.*

the vegetation during the summer months (**Figure 8C**) and promoted the intermittent growth of embryo dunes seaward of the established foredune (e.g., **Figure 8A** and **C**). Following SD in September 2019, the scarp line has eroded into the lower stoss slope and has removed any alongshore embryo dunes present pre-SD (**Figure 8D**).

The variability of storm impacts observed at BB were likely controlled by the antecedent morphology that existed prior to SD. From the LiDAR DTM, BB east was observed to have a higher nearshore slope and closer inner bar structure, in comparison to BB west (**Figure 9A** and **B**). This could have led to higher wave energy and erosive potential at BB East during SD, and stronger counter current moving sediment further seaward post-SD (e.g., **Figure 7A** and **8A**). Beach width ð Þ *Bw* at BB East was also significantly narrower than BB West and likely would have increased the exposure of the foredune to wave run up, despite a lower storm surge level relative to the dune toe (**Figure 9A** and **B**). In comparison to BB West, a combination of foredune metrics including narrower dune width ð Þ *Dw* , shorter dune heights ð Þ *Dh* , higher dune slopes ð Þ *Ds* , and lower dune volumes ð Þ *Dv* indicate that BB East was more vulnerable to storm induced erosion (**Figure 9A** and **B**). Foredunes that displayed a narrower *Dw* and high *Ds* at BB East experienced the highest magnitude of erosion resulting from major slope failures. While the foredune at BB West appeared to only lose a small proportion of the frontal *Dv*, a significant reduction of *Dv* at BB East could lead to further lowering of the foredune making this area of BB increasingly vulnerable to future storm events.

### **5. Discussion**

UAV monitoring has allowed for the rapid assessment of Brackley Beach (BB), 11 days following Storm Dorian (SD). Projected 7–8 m significant wave heights and 1.2 m storm surge levels associated with SD resulted in a highly erosive post storm

surface, the majority of which occurred at the beach-dune boundary (**Figure 6A** and **<sup>B</sup>**). A total of 11,000 m<sup>3</sup> of erosion was recorded between a pre-SD LiDAR and post-SD UAV survey (**Table 1**). This included the complete removal of embryo dune deposits and the formation of a continuous foredune scarp in the middle and western sections of BB (**Figure 8A** and **B**). At BB East, significant slope failure and high magnitude erosion of the frontal section of the foredune was observed (**Figure 7A** and **B**).

artifacts (e.g., vegetation). This chapter has attempted to address some of these common challenges using basic quality control measures and handling of data uncertainties. In doing so, we were able to confidently quantify alongshore topographic and volumetric changes resulting from a storm event. Survey uncertainties in our post-storm measurements were mitigated due the implementation of quality controls to produce high accuracy (i.e., low RMSE) surface models from UAV surveys. Furthermore, environmental uncertainties were reduced in large part by the wave induced removal of vegetation from the beach dune boundary post-SD. While vegetation uncertainty did not significantly constrain our post storm measurements of the dune scarp, subsequent dune recovery studies will need to develop new strategies to handle uncertainty introduced by vegetation recolonization of the

The quality controls measures, discussed in Section 3, partially include data processing techniques that are unique to the Pix4D software. It is currently beyond the scope of this chapter and expertise of the authors to provide a comprehensive review of different SfM software. Therefore, the remainder of this discussion will

UAV SfM surveys have proven to be highly accurate with RMSE values, typically <1o cm [36, 38–40, 43]. This level of accuracy is capable of monitoring topographic adjustments associated with storm impacts. However, the uncertainty of topographic change measurements should be reported especially when changes approach the same magnitude of the survey error. While the RMSE values are commonly reported, interpretation of how these values influence uncertainty in topographic change measurements are not. To address survey error, a previous study used RMSE as a threshold to monitor topographic change [44]. In this study, the RMSE value was 16 cm and only positive or negative values in exceedance of �16 cm were reported. This method is effective at filtering data that could represent no change, but there still could be a high level of uncertainty associated with low magnitude changes. For instance, 17 cm of deposition recorded from a 1 m � 1 m pixel would result in a volume change (*ΔV*) and volume change uncertainty (*ΔVU*)

the range of uncertainty. This indicates that interpreting the significance of low magnitude topographic changes should be done carefully and only after survey

Results presented in this chapter limited survey uncertainty by applying a topographic change threshold ð Þ *CL*95% that approximates to a minimum detectable range of 1.96 times the propagated RMSE value. Applying *CL*95% effectively increased the confidence in *ΔV* measurements by filtering out small magnitude changes that are disproportionately responsible for high levels of *ΔVU*. For example, 42% of the total observable area of post-SD BB was classified as a zone of elevational uncertainty but, once removed, this only resulted in a change of total *ΔV* reported by 3% (**Table 1**). Furthermore, the ratio between *ΔV* and *ΔVU* dropped from 49 to 23% prior to- and after the*CL*95% threshold was applied. It is important to note that *ΔV* below the *CL*95% threshold, or within the zone of uncertainty, may represent actual change; however, to increase confidence in reported *ΔV* values it is important to systematically address survey uncertainty. This chapter has demonstrated that thresholding can be an effective approach to reduce topographic change uncertainties, but the most effective approach remains the quality of repeatable survey strategies and data processing (e.g., discussed in Section 3). For coastal studies, maintaining a high level of survey accuracy limits the areal extent and vertical range

uncertainty has been reported and adequately addressed.

, respectively. Or in other words, 94% of the *ΔV* would be within

focus on addressing survey and environmental uncertainties.

**5.1 Survey and environmental uncertainties**

*Monitoring Storm Impacts on Sandy Coastlines with UAVs*

*DOI: http://dx.doi.org/10.5772/intechopen.91459*

backshore zone.

0.17 m<sup>3</sup> � 0.16 m<sup>3</sup>

**81**

Despite a highly erosive post storm surface, a follow up field investigation in November 2019, or 2 months post-SD, observed the that the initial stages of dune recovery were already taking place. Evidence of the widespread remobilization of beach sediment by eolian processes was observed in the accumulation of dry ripple laden sediment deposits at beach-dune boundary, both across and alongshore (**Figure 10A**). This has resulted in the initial deposition of sediment into the dune scarp (**Figure 10B** and **C**) and ramps forming in areas of higher magnitude deposition (**Figure 10D**). Monitoring storm impacts can reveal initial topographic adjustments resulting from a single event; however, subsequent beach-dune responses can provide a broader understanding on the resiliency of sandy coastlines. Given the low cost, rapidity, and high resolution of UAV SfM surveys, researchers will increasingly have access to high resolution geo-spatial data sets to continuously monitor both short- and longer-term coastal dynamics.

Although UAV SfM research has increased significantly over the last few years, studies that have reported topographic or volumetric change from beach-dune systems are still limited [37, 38, 40, 44, 45]. In part, this has been due to the inherent difficulties in data collection, post-processing, and handling of survey

#### **Figure 10.**

*Initial dune recovery was observed during a follow up field site assessment in November 2019, or 2 months following storm Dorian (SD). Alonshore, dry sediment has begun to accumulate at the beach-dune boundary (A). This has resulted in the initial accumulation of sediment at the dune scarp (B and C) and ramp development in areas of higher magnitude deposition (D).*

*Monitoring Storm Impacts on Sandy Coastlines with UAVs DOI: http://dx.doi.org/10.5772/intechopen.91459*

surface, the majority of which occurred at the beach-dune boundary (**Figure 6A** and **<sup>B</sup>**). A total of 11,000 m<sup>3</sup> of erosion was recorded between a pre-SD LiDAR and post-SD UAV survey (**Table 1**). This included the complete removal of embryo dune deposits and the formation of a continuous foredune scarp in the middle and western sections of BB (**Figure 8A** and **B**). At BB East, significant slope failure and

*Spatial Variability in Environmental Science - Patterns, Processes, and Analyses*

high magnitude erosion of the frontal section of the foredune was observed

monitor both short- and longer-term coastal dynamics.

Despite a highly erosive post storm surface, a follow up field investigation in November 2019, or 2 months post-SD, observed the that the initial stages of dune recovery were already taking place. Evidence of the widespread remobilization of beach sediment by eolian processes was observed in the accumulation of dry ripple laden sediment deposits at beach-dune boundary, both across and alongshore (**Figure 10A**). This has resulted in the initial deposition of sediment into the dune scarp (**Figure 10B** and **C**) and ramps forming in areas of higher magnitude deposition (**Figure 10D**). Monitoring storm impacts can reveal initial topographic adjustments resulting from a single event; however, subsequent beach-dune responses can provide a broader understanding on the resiliency of sandy coastlines. Given the low cost, rapidity, and high resolution of UAV SfM surveys, researchers will increasingly have access to high resolution geo-spatial data sets to continuously

Although UAV SfM research has increased significantly over the last few years,

studies that have reported topographic or volumetric change from beach-dune systems are still limited [37, 38, 40, 44, 45]. In part, this has been due to the inherent difficulties in data collection, post-processing, and handling of survey

*Initial dune recovery was observed during a follow up field site assessment in November 2019, or 2 months following storm Dorian (SD). Alonshore, dry sediment has begun to accumulate at the beach-dune boundary (A). This has resulted in the initial accumulation of sediment at the dune scarp (B and C) and ramp*

(**Figure 7A** and **B**).

**Figure 10.**

**80**

*development in areas of higher magnitude deposition (D).*

artifacts (e.g., vegetation). This chapter has attempted to address some of these common challenges using basic quality control measures and handling of data uncertainties. In doing so, we were able to confidently quantify alongshore topographic and volumetric changes resulting from a storm event. Survey uncertainties in our post-storm measurements were mitigated due the implementation of quality controls to produce high accuracy (i.e., low RMSE) surface models from UAV surveys. Furthermore, environmental uncertainties were reduced in large part by the wave induced removal of vegetation from the beach dune boundary post-SD. While vegetation uncertainty did not significantly constrain our post storm measurements of the dune scarp, subsequent dune recovery studies will need to develop new strategies to handle uncertainty introduced by vegetation recolonization of the backshore zone.

The quality controls measures, discussed in Section 3, partially include data processing techniques that are unique to the Pix4D software. It is currently beyond the scope of this chapter and expertise of the authors to provide a comprehensive review of different SfM software. Therefore, the remainder of this discussion will focus on addressing survey and environmental uncertainties.

#### **5.1 Survey and environmental uncertainties**

UAV SfM surveys have proven to be highly accurate with RMSE values, typically <1o cm [36, 38–40, 43]. This level of accuracy is capable of monitoring topographic adjustments associated with storm impacts. However, the uncertainty of topographic change measurements should be reported especially when changes approach the same magnitude of the survey error. While the RMSE values are commonly reported, interpretation of how these values influence uncertainty in topographic change measurements are not. To address survey error, a previous study used RMSE as a threshold to monitor topographic change [44]. In this study, the RMSE value was 16 cm and only positive or negative values in exceedance of �16 cm were reported. This method is effective at filtering data that could represent no change, but there still could be a high level of uncertainty associated with low magnitude changes. For instance, 17 cm of deposition recorded from a 1 m � 1 m pixel would result in a volume change (*ΔV*) and volume change uncertainty (*ΔVU*) 0.17 m<sup>3</sup> � 0.16 m<sup>3</sup> , respectively. Or in other words, 94% of the *ΔV* would be within the range of uncertainty. This indicates that interpreting the significance of low magnitude topographic changes should be done carefully and only after survey uncertainty has been reported and adequately addressed.

Results presented in this chapter limited survey uncertainty by applying a topographic change threshold ð Þ *CL*95% that approximates to a minimum detectable range of 1.96 times the propagated RMSE value. Applying *CL*95% effectively increased the confidence in *ΔV* measurements by filtering out small magnitude changes that are disproportionately responsible for high levels of *ΔVU*. For example, 42% of the total observable area of post-SD BB was classified as a zone of elevational uncertainty but, once removed, this only resulted in a change of total *ΔV* reported by 3% (**Table 1**). Furthermore, the ratio between *ΔV* and *ΔVU* dropped from 49 to 23% prior to- and after the*CL*95% threshold was applied. It is important to note that *ΔV* below the *CL*95% threshold, or within the zone of uncertainty, may represent actual change; however, to increase confidence in reported *ΔV* values it is important to systematically address survey uncertainty. This chapter has demonstrated that thresholding can be an effective approach to reduce topographic change uncertainties, but the most effective approach remains the quality of repeatable survey strategies and data processing (e.g., discussed in Section 3). For coastal studies, maintaining a high level of survey accuracy limits the areal extent and vertical range of elevational uncertainty and will allow for low magnitude and morphologically significant topographic adjustments (e.g., dune recovery) to be observed through time.

Environmental uncertainties within UAV SfM data also pose a significant challenge as they are space and time dependent. Images taken at the water boundary often captures breaking waves, wave run up, or low contrast moist surfaces in the nearshore and foreshore zones. This can lead to insufficiencies in ATP generation during SfM processing [36, 37, 41] and inconsistencies on the seaward extent in which repeat topographic change measurements are taken. Previously, mean sea level (MSL; [45]) and the wet-dry line [44] have been identified as the observable extent of coastal DSMs. These provide a reasonable estimate of the average annual water line or the water line at a particular time but can become inconsistent when monitoring spatially and/or temporally extensive surveys.

For instance, 2.5 km of post-SD BB was surveyed over a two-day period with approximately 4 h of surveying time per day. Considering the 6 h semi-diurnal (i.e., two high and low) tidal cycle at BB, variability of the across shore extent of the wetdry line was captured during our alongshore survey. Beyond daily cycles, monthly and annual cycles effect the magnitude of the tidal range and can result in significant variability of the water line boundary during repeat surveys. This may introduce uncertainty when quantifying multi-temporal sediment budgets (e.g., *ΔV*) and beach metrics (e.g., beach width) as these values may be over- or underestimated depending on unique spatiotemporal tidal patterns captured while surveying. Alternatively, this chapter used the HHW line as consistent elevation that is typically above the intertidal foreshore zone regardless of tide cycles and ranges. While this approach is conservative in constricting the seaward survey extent, it provides a repeatable methodology to limit environmental uncertainty at the water line boundary.

At the beach-dune boundary, perhaps the biggest challenge in producing repeat UAV surveys in sandy coastal systems is the presence and handling of vegetation. Vegetation can directly lead to over-estimation of surface elevation values in SfM DSMs or point clouds [41]. Furthermore, any topographic change values that have not removed vegetation would be inaccurate as they could represent either a change in vegetation density or canopy height at the seasonal scale or loss of vegetation following a disturbance [38]. Post-SD, elevated storm surge removed vegetation seaward of the scarp along BB and did not have a significant influence our ability to measure topographic change; however, following storm events or seasons, vegetation recolonization could constrain subsequent monitoring surveys during periods of high growth rates.

the point cloud was completed in Pix4D and consisted of classifying and removing supra-surficial features and produce a 'bare earth' DTM. The standard Pix4D vegetation class, 'high vegetation', may be suitable for other prominent vegetation types (e.g., trees) but was not designed for and, therefore, unable to adequately classify and remove low profile vegetation that typically grows on the backshore (e.g., *Ab*;

*An across profile and vegetation density, measured from aerial UAV photos taken at Brackley Beach (BB) in July 2019, show variable density at the seaward extent followed by rapid increase in density from the mid stoss slope to dune crest (A). The UAV SfM point cloud generated for this same section of BB, show variability in bare sand and vegetated points captured in the point cloud as you move landward from the sparsely vegetated*

Vegetation filtering can be improved by applying algorithms that have been specifically designed for removing beach grasses. For example, a recent study [41] applied a vegetation filtering method originally designed to filter *Ab* from a TLS derived point cloud [51]. Results of this study show that sparsely vegetated

foredunes can be effectively filtered from relatively low density SfM derived point

vegetation densities, relative to higher density point clouds (i.e., 2000 points per

) produced from a TLS sensor. Vegetation is often intermittent at the beach-dune boundary, suggesting that point cloud filtering approaches, especially when applied to higher density SfM point clouds (e.g., **Figure 11B**), could accurately monitor low magnitude topographic changes associated dune building and recovery. In this regard, further studies are needed to test the capability of different vegetation filtering algorithms to remove variant patterns of backshore vegetation growth. Handling of vegetation remains problematic and, without using a suitable filtering method or quantifying the additional uncertainty that it may introduce, should not

), but do not perform as well in areas of higher

**Figure 11B**).

**Figure 11.**

m2

**83**

clouds (i.e., 100 points per m<sup>2</sup>

*beach-dune boundary to highly vegetated back dune zone (B).*

*Monitoring Storm Impacts on Sandy Coastlines with UAVs*

*DOI: http://dx.doi.org/10.5772/intechopen.91459*

be included in topographic change measurements.

A UAV vegetation monitoring study at BB from July 2019 sampled vegetation density, almost entirely low profile (i.e., 50–60 cm) *Ammophila breviligulata* (*Ab*)*,* both across- and alongshore during the peak growing season [53]. Across shore (i.e., from the shoreline to the back dune zone), vegetation density tended to fluctuate between the vegetation line and dune toe, followed by a rapid increase in density up the stoss slope, before reaching 100% density near the crest (**Figure 11A**). Alongshore, the overall density between the shoreline and dune crest, sampled at every 10 m along BB, ranged from 10 to 60% (**Figure 11B**). Taking into account uncertainty introduced with the fluctuating water line, only a narrow band landward of the foreshore and seaward of the vegetation line would consistently be observable without error introduced by spatially and seasonally variant vegetation density patterns.

A high resolution SfM derived point cloud of 2000 points per m<sup>2</sup> was produced from a preliminary UAV survey during July 2019 (**Figure 11B**). The high vegetation density during the peak growing season is evident in the backshore; however, a large number of 'bare earth' points are also visible (**Figure 11B**). An attempt to filter *Monitoring Storm Impacts on Sandy Coastlines with UAVs DOI: http://dx.doi.org/10.5772/intechopen.91459*

#### **Figure 11.**

of elevational uncertainty and will allow for low magnitude and morphologically significant topographic adjustments (e.g., dune recovery) to be observed through time. Environmental uncertainties within UAV SfM data also pose a significant challenge as they are space and time dependent. Images taken at the water boundary often captures breaking waves, wave run up, or low contrast moist surfaces in the nearshore and foreshore zones. This can lead to insufficiencies in ATP generation during SfM processing [36, 37, 41] and inconsistencies on the seaward extent in which repeat topographic change measurements are taken. Previously, mean sea level (MSL; [45]) and the wet-dry line [44] have been identified as the observable extent of coastal DSMs. These provide a reasonable estimate of the average annual water line or the water line at a particular time but can become inconsistent when

*Spatial Variability in Environmental Science - Patterns, Processes, and Analyses*

For instance, 2.5 km of post-SD BB was surveyed over a two-day period with approximately 4 h of surveying time per day. Considering the 6 h semi-diurnal (i.e., two high and low) tidal cycle at BB, variability of the across shore extent of the wetdry line was captured during our alongshore survey. Beyond daily cycles, monthly and annual cycles effect the magnitude of the tidal range and can result in significant variability of the water line boundary during repeat surveys. This may introduce uncertainty when quantifying multi-temporal sediment budgets (e.g., *ΔV*) and beach metrics (e.g., beach width) as these values may be over- or underestimated depending on unique spatiotemporal tidal patterns captured while surveying. Alternatively, this chapter used the HHW line as consistent elevation that is typically above the intertidal foreshore zone regardless of tide cycles and ranges. While this approach is conservative in constricting the seaward survey extent, it provides a repeatable methodology to limit environmental uncertainty at the water

At the beach-dune boundary, perhaps the biggest challenge in producing repeat UAV surveys in sandy coastal systems is the presence and handling of vegetation. Vegetation can directly lead to over-estimation of surface elevation values in SfM DSMs or point clouds [41]. Furthermore, any topographic change values that have not removed vegetation would be inaccurate as they could represent either a change in vegetation density or canopy height at the seasonal scale or loss of vegetation following a disturbance [38]. Post-SD, elevated storm surge removed vegetation seaward of the scarp along BB and did not have a significant influence our ability to measure topographic change; however, following storm events or seasons, vegetation recolonization could constrain subsequent monitoring surveys during periods

A UAV vegetation monitoring study at BB from July 2019 sampled vegetation density, almost entirely low profile (i.e., 50–60 cm) *Ammophila breviligulata* (*Ab*)*,* both across- and alongshore during the peak growing season [53]. Across shore (i.e., from the shoreline to the back dune zone), vegetation density tended to fluctuate between the vegetation line and dune toe, followed by a rapid increase in

(**Figure 11A**). Alongshore, the overall density between the shoreline and dune crest, sampled at every 10 m along BB, ranged from 10 to 60% (**Figure 11B**). Taking into account uncertainty introduced with the fluctuating water line, only a narrow band landward of the foreshore and seaward of the vegetation line would consistently be observable without error introduced by spatially and seasonally variant vegetation

A high resolution SfM derived point cloud of 2000 points per m<sup>2</sup> was produced from a preliminary UAV survey during July 2019 (**Figure 11B**). The high vegetation density during the peak growing season is evident in the backshore; however, a large number of 'bare earth' points are also visible (**Figure 11B**). An attempt to filter

density up the stoss slope, before reaching 100% density near the crest

monitoring spatially and/or temporally extensive surveys.

line boundary.

of high growth rates.

density patterns.

**82**

*An across profile and vegetation density, measured from aerial UAV photos taken at Brackley Beach (BB) in July 2019, show variable density at the seaward extent followed by rapid increase in density from the mid stoss slope to dune crest (A). The UAV SfM point cloud generated for this same section of BB, show variability in bare sand and vegetated points captured in the point cloud as you move landward from the sparsely vegetated beach-dune boundary to highly vegetated back dune zone (B).*

the point cloud was completed in Pix4D and consisted of classifying and removing supra-surficial features and produce a 'bare earth' DTM. The standard Pix4D vegetation class, 'high vegetation', may be suitable for other prominent vegetation types (e.g., trees) but was not designed for and, therefore, unable to adequately classify and remove low profile vegetation that typically grows on the backshore (e.g., *Ab*; **Figure 11B**).

Vegetation filtering can be improved by applying algorithms that have been specifically designed for removing beach grasses. For example, a recent study [41] applied a vegetation filtering method originally designed to filter *Ab* from a TLS derived point cloud [51]. Results of this study show that sparsely vegetated foredunes can be effectively filtered from relatively low density SfM derived point clouds (i.e., 100 points per m<sup>2</sup> ), but do not perform as well in areas of higher vegetation densities, relative to higher density point clouds (i.e., 2000 points per m2 ) produced from a TLS sensor. Vegetation is often intermittent at the beach-dune boundary, suggesting that point cloud filtering approaches, especially when applied to higher density SfM point clouds (e.g., **Figure 11B**), could accurately monitor low magnitude topographic changes associated dune building and recovery. In this regard, further studies are needed to test the capability of different vegetation filtering algorithms to remove variant patterns of backshore vegetation growth. Handling of vegetation remains problematic and, without using a suitable filtering method or quantifying the additional uncertainty that it may introduce, should not be included in topographic change measurements.

The density and variety of vegetation increases at BB moving from the crest landward. As a result, a significant reduction in the 'bare earth' points become problematic in these areas (**Figure 11B**). Vegetation filtering then becomes a less viable option. Although, these locations are less prone to low magnitude topographic changes, quantifying dune metrics (e.g., *Dh* or *Dv*) measured from UAV surveys would likely be inflated. Highly vegetated back dune areas are typically stable as they are protected from the wave, tidal, and wind processes that are actively shaping the seaward zones. These locations are likely to remain a 'zone of uncertainty' in UAV SfM surveys without significant vegetation removal or burial through blowout development, breaching, or overwash events.

**Acknowledgements**

**Author details**

**85**

Alex Smith\*, Brianna Lunardi, Elizabeth George and Chris Houser

\*Address all correspondence to: absmith9@uwindsor.ca

provided the original work is properly cited.

School of the Environment, University of Windsor, Windsor, Ontario, Canada

© 2020 The Author(s). Licensee IntechOpen. This chapter is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/ by/3.0), which permits unrestricted use, distribution, and reproduction in any medium,

The authors would like to thank Prince Edward Island National Park for permitting our field work at Brackley Beach, logistical assistance, and supplying of equipment. We would also like to thank CBCL Limited for supplying the baseline LiDAR survey used in this chapter. Finally, a special thank you to Phillipe Wernette

and Jacob Lehner for their invaluable field assistance.

*Monitoring Storm Impacts on Sandy Coastlines with UAVs*

*DOI: http://dx.doi.org/10.5772/intechopen.91459*

Therefore, current UAV SfM applications for monitoring repeat topographic changes occurring in coastal systems are likely to be constrained to the backshore. The accuracy of UAV SfM data on unvegetated surfaces has been demonstrated and there is significant potential to increase the accuracy of sparsely vegetated foredune slopes by applying filtering algorithms (e.g., [41]). It is evident that challenges still remain in resolving water and vegetation boundaries; however, the benefits of UAVs are also clear as they provide an accessible cost-effective method to produce high resolution and spatially extensive surveys of sandy coastal systems. The number of UAV SfM monitoring studies in coastal systems are likely to increase in the coming years, but in order for these studies to confidently report topographic adjustments between the beach-dune boundary, addressing data uncertainties and improvements in vegetation filtering methods are needed.
