Back analysis of the September 5th, 2021 rockfall near Lover's Arch at Hopewell Rocks Provincial Park, New Brunswick

Abstract

This paper presents an analysis of the shoreline cliff rockfall that occurred on September 5 th , 2021, in Hopewell Rocks Provincial Park, New Brunswick, Canada, by using Structure-from-Motion photogrammetry of the rockfall site (after failure) and meteorological data. This failure occurred near the major staircase access point to the beach for geotourists (Staircase Cove) and in an area with high pedestrian traffic. Specifically, this study presents an analysis of the roughness and curvature of the failure surface and an interpretation of how the rockmass failed. The resulting failure interpretation is comprised of multiple stages: (i) long-term tidal erosion undercut the rockmass, (ii) gravity-driven tensile cracks propagated through the rockmass, and (iii) a large precipitation event ultimately catalyzed the rockfall. RÉSUMÉ Cet article présente l'analyse d'une chute de roches d'une falaise littoral qui s'est produite le 5 septembre 2021 dans le parc provincial des Rochers Hopewell au Nouveau-Brunswick, Canada, en utilisant la technique d'imagerie photogrammétrique Structure-from-Motion sur le site de l'incident (après l'échec) et des données météorologiques. Cette rupture s'est produite à proximité du point d'accès principal à la plage pour les géo-touristes (L'Anse aux Escaliers), une zone piétonnière qui est couramment fréquentée. En détail, cette étude présente une analyse de la rugosité et de la courbure de la surface de rupture et une interprétation du mécanisme de rupture de la falaise. L'interprétation de l'effondrement est divisée en plusieurs étapes: (i) l'érosion par les marées a creusé la masse rocheuse à long terme, (ii) des fissures de traction entraînées par la gravité se sont propagées à travers la masse rocheuse, et (iii) un grande événement de précipitation a finalement catalysé la chute de roches.

Full text

Back analysis of the September 5th, 2021 rockfall
near Lover’s Arch at Hopewell Rocks Provincial
Park, New Brunswick
William R. Hoyle, Jennifer J. Day & Amanda Hyslop
Department of Geological Sciences and Geological Engineering – Queen’s
University, Kingston, ON, Canada
Kevin Snair
Hopewell Rocks Provincial Park, Department of Tourism, Heritage and Culture, New Brunswick, Canada
ABSTRACT
This paper presents an analysis of the shoreline cliff rockfall that occurred on September 5th, 2021, in Hopewell Rocks
Provincial Park, New Brunswick, Canada, by using Structure-from-Motion photogrammetry of the rockfall site (after failure)
and meteorological data. This failure occurred near the major staircase access point to the beach for geotourists (Staircase
Cove) and in an area with high pedestrian traffic. Specifically, this study presents an analysis of the roughness and
curvature of the failure surface and an interpretation of how the rockmass failed. The resulting failure interpretation is
comprised of multiple stages: (i) long-term tidal erosion undercut the rockmass, (ii) gravity-driven tensile cracks propagated
through the rockmass, and (iii) a large precipitation event ultimately catalyzed the rockfall.
RÉSUMÉ
Cet article présente l’analyse d’une chute de roches d’une falaise littoral qui s’est produite le 5 septembre 2021 dans le
parc provincial des Rochers Hopewell au Nouveau-Brunswick, Canada, en utilisant la technique d’imagerie
photogrammétrique Structure-from-Motion sur le site de l’incident (après l’échec) et des données météorologiques. Cette
rupture s’est produite à proximité du point d’accès principal à la plage pour les géo-touristes (L’Anse aux Escaliers), une
zone piétonnière qui est couramment fréquentée. En détail, cette étude présente une analyse de la rugosité et de la
courbure de la surface de rupture et une interprétation du mécanisme de rupture de la falaise. L’interprétation de
l’effondrement est divisée en plusieurs étapes: (i) l’érosion par les marées a creusé la masse rocheuse à long terme, (ii)
des fissures de traction entraînées par la gravité se sont propagées à travers la masse rocheuse, et (iii) un grande
événement de précipitation a finalement catalysé la chute de roches.
1 INTRODUCTION
Rockfalls are a common geohazard that can pose a serious
threat to downslope infrastructure and people. In the
context of this study at Hopewell Rocks Provincial Park
(Figure 1a), rockfalls primarily pose a risk to the health and
safety of geotourists and Park employees. More than
250,000 geotourists visit Hopewell Rocks Provincial Park
during the open season from May to October every year
(Bioldeau, 2019).
Rockfalls are caused by a variety of factors such as
freeze-thaw cycles, chemical dissolution of rock matrix,
heavy rainfall, and tidal erosion (e.g. Wyllie and Mah 2004;
Westoby et al. 2020). Failures can often have complex
mechanisms with multiple intersecting failure surfaces
which tend to be confined along planes of inherent
weakness in the rockmass (such as joints or bedding) or
where crack propagation has occurred from tensile failure
of the rockmass over time (Corominas et al. 2017). The
latter of these failure surfaces is especially common in
weak coastal rockmasses where rockfalls are common
above notches caused by marine action (Gong et al. 2017).
Tidal driven failures of this kind are the result of decades of
coastal erosion but may ultimately be triggered by a
specific catalyzing event (Kogure and Matsukura 2010).
One of the common failure triggers of concern in
Hopewell Rocks Provincial Park is frost jacking, which
occurs when water seeps into open joints or cracks in the
rock and freezes, causing the water to expand and
generate pressure between each section of the rock. To
mitigate this, the Park employs scaling to remove unstable
sections of rock each spring before the Park opens for the
season. Rockfalls have been noted to often occur at the
Park in the late winter to spring months. While frost jacking
is a significant factor in these seasonal rockfalls, other
factors contribute to rockfalls occurring during the open
season (May to October), such as the September 5th, 2021
cliff rockfall that is the subject of this study.
The September 5th, 2021 cliff rockfall occurred near
Lover’s Arch, which is one of the most famous sea arch
formations in the Park, located near the main beach access
staircase in Staircase Cove (Figure 1b,c). Furthermore,
debris from this rockfall covered a major footpath that
provides primary access to southern parts of the beach and
therefore sees high pedestrian traffic. The timing of this
failure (late summer and shortly after a major storm)
suggests the ultimate failure trigger(s) were related to
water infiltration along joints and within the intact rock.
This paper presents a failure analysis of the September
5th, 2021 cliff rockfall event. The data used in this study
includes 3-dimensional Structure-from-Motion (SfM)
photogrammetry models of the rockfall site (after failure),
meteorological data leading up to the failure, and historical
photographs of the rockfall site (before failure).

Figure 1. Site location: (a) maps of the Maritimes (inset) and coastal features of Hopewell Rocks Provincial Park; (b)
Google Earth satellite image of Staircase Cove showing key landmarks, location of Sept. 5, 2021 cliff rockfall, and viewpoint
of Figure 1(c); (c) panoramic photograph from UAV (taken Oct. 15, 2021) labelled with key landmarks and features.
2 GEOLOGICAL SETTING
The bedrock at Hopewell Cape is primarily composed of
arkosic sandstone and coarse red conglomerate from the
Hopewell Group, which overlies a dark grey fossiliferous
limestone from the Windsor Group (Wallace 1998). These
sedimentary successions dip toward the northeast at
approximately 35 ± 10° (Hyslop et al. 2021). The bedding
in the Hopewell Group averages 20-30 cm thick and tends
to be defined by changes in grain size (e.g. conglomerate
and sandstone facies). In the conglomerate layers, grains
are very poorly sorted and clasts range in size from 10-
15 cm in diameter and occasionally up to 30 cm. These
clasts are predominantly composed of granite, gneiss, and
volcanic lithologies. This succession is interpreted to be a
moderate to high energy alluvial fan deposit spreading
easterly from a source area in the Caledonia Mountains as
a result of lowering paleo-sea levels (Wallace 1998). Joints
are widely spaced, some undulate, and range from
moderately to extremely weathered.

Figure 2. Preliminary 3D photogrammetry model of the cliff with the failure surface highlighted in yellow. Note that this
model is a separate, lower resolution model compared to the models in Figures 3 & 4, and the yellow signs in the
foreground are 75 cm x 30 cm each for scale.
3 PHOTOGRAMMETRY MODELLING
Preliminary photographs of the September 5th, 2021
rockfall failure surface and surrounding cliff face were
taken on September 18, 2021, and used to build a 3D SfM
photogrammetry model (Figure 2). Detailed photographs of
the September 5th, 2021 rockfall failure surface and debris
pile were collected on October 13, 2021 from an
Unmanned Aerial Vehicle (UAV) platform (DJI Matrice
M300 RTK model) using a DJI Zenmuse P1 Camera
(45 MP) payload. The photographs were used to create
detailed 3-dimensional (3D) SfM photogrammetry point
cloud models of the failure surface (Figure 3) using Agisoft
Metashape software (Agisoft 2021). This detailed model
was analyzed using CloudCompare (2019) software to
assess the roughness and curvature of the failure surface
(using three methods). This analysis is important for
determining if the failure surface developed through pre-
existing discontinuities (i.e., joints), from fracture
propagation through intact rock, or a combination of both.
3.1 Failure Surface Curvature Quantification
Three methods were used to quantify the curvature of the
failure surface. The first two methods calculate (i)
roughness and (ii) surface normal rate of change in
CloudCompare. The major difference between these two
methods is that roughness is dependent on the distance
from points to their local plane of best fit within a user
defined kernel (a spherical neighbourhood of points), and
surface normal rate of change curvature measures the
change in the angle between the normal of local planes of
best fit for adjacent kernels.
The third method (iii) used to assess the curvature of
the failure surface involves analyzing the distribution of
normal orientations on a stereonet (Sturzenegger and
Stead 2009; Hyslop et al. 2021).
Figure 3. Detailed SfM 3D photogrammetry model of the
September 5th, 2021 rockfall failure surface. Note that this
view is at a different orientation to the view found in Figure
2. This was chosen as it is more oblique to the failure
surface. Surfaces C and D are difficult to see at this
orientation as they are planar surfaces that are
perpendicular to the page.
3.1.1 Failure Surface Roughness Analysis
Numerous researchers have investigated methods of
evaluating rock discontinuity roughness using SfM
photogrammetry models (e.g. Packulak et al. 2019; Salvini
A
A
B
D
C

et al. 2020; Bonneau 2021). CloudCompare (2021) has a
built-in algorithm to calculate the roughness of a surface
from a 3D point cloud. It first finds the plane of best fit for
all points within a kernel with a user defined radius (10 mm
in this study, which is the smallest possible resolution for
the image quality in the model). Next, the distance from the
plane to each point is calculated parallel to the normal of
the plane of best fit. The further the point is away from the
plane of best fit the higher value of roughness it receives.
Surfaces where almost all points along the plane of best fit
have very low roughness values represent smooth, planar
features. It is important to define a kernel radius as this will
determine the scale of the roughness being calculated.
The roughness calculation for the failure surface of the
rockfall (shown in Figure 4a) indicates the vast majority of
the failure surface has a roughness of less than 2 mm (in
the context of the 10 mm measurement kernel). There are
two distinct surfaces on the failure surface that have
different roughness characteristics and orientations. The
rougher segment (A) of the failure surface (higher
concentrations of roughness >2.5 mm) has a mean dip
direction/dip of 223/61°, which matches a joint orientation
recorded at the nearby Elephant Rock failure by Hyslop et
al. (2021) that was classed as wavy with a 1 mm gap and
a Joint Roughness Coefficient (from Barton and Choubey,
1977) of 19. This rougher segment that comprises
approximately 77% of the failure surface suggests tensile
fractures developed through intact rock to form the failure
surface; however, this segment’s similar orientation to a
local joint set suggests there may have been pre-existing
microfractures formed in this section of the rockmass
during the brittle joint-forming tectonic regime of the area,
which would have contributed to increasing connectivity of
induced tensile cracks that propagated during this rockfall.
Bedding in the area has a mean dip direction/dip of 045/37°
according to Hyslop et al. (2021). Incipient bedding through
the failure surface may have also contributed to increasing
tensile crack connectivity during the failure process.
The other dominant surface segment (B) with lower
roughness (majority less than 0.5 mm) is a smooth, planar
joint with surface staining and a mean dip direction/dip of
013/70°, which comprises approximately 13% of the failure
surface area. This joint orientation is kinematically aligned
with the cliff wall to be prone to planar sliding. The surface
staining on this segment suggests this joint provides a flow
pathway for water to infiltrate the rockmass, and its steep
inclination makes it well suited to drain precipitation down
through the rockmass from the top of the cliff.
Overall, the rougher segment (A) has a mean
roughness of 0.873 mm (standard deviation of 0.718 mm),
and the planar sliding segment (B) has a mean roughness
of 0.534 mm (standard deviation of 0.683 mm).
3.1.2 Failure Surface Normal Change Rate Curvature
The second method that quantifies the surface curvature of
this failure surface calculates the rate of change of the
normal vector of the plane of best fit across adjacent
kernels using CloudCompare (2021). Failure change rates
are measured in inverse metres and the values come from
the magnitude of the difference between the Gaussian
curvatures of adjacent kernels, essentially the square root
of Gaussian curvature. Because this method applies one
value for every set of kernels and not each kernel, the
resulting contoured image highlights larger scale
curvature, while small scale roughness is smoothed out.
The result of this method (Figure 4b) also shows the
presence of two dominant and distinct failure surface
segments (A and B), although some areas of segment A
(rougher) exhibit curvature as a low as in segment B (pre-
existing joint). The mean normal rate of change curvature
for surface segment A (rougher) is 2.98x10-2 m-1 (standard
deviation of 2.34x10-2 m-1) and for segment B (joint) is
5.1x10-5 m-1 (standard deviation of 5.2x10-5 m-1).
Figure 4. a) Roughness (m) with kernel of 1 cm, b) Normal change rate (m-1) with kernel of 1 cm. Note that values of
roughness and normal rate of change are restricted to the lowest 50% of values for definition. Note that Surfaces C and D
are difficult to see at this orientation as they are planar surfaces which are perpendicular to the page.
A
A
B
A
B
A
C
C
D
D

3.1.3 Failure Surface Normal Pole Distributions
This method was developed by Sturzenegger and Stead
(2009) and was employed in the Elephant Rock failure
analysis by Hyslop et al. (2021). The normal vectors of
every plane of best fit (kernel radius = 1 cm) are plotted on
a stereonet and the relative distribution of the poles can be
used to identify the roughness of the failure surface.
Surfaces with higher variation in their pole orientations are
rougher/undulating, while smoother/planar surfaces exhibit
a lower variation. This analysis revealed two major and two
minor pole clusters (Figure 5). The major cluster (mean
pole orientation in plunge → trend notation of 70±25° →
013±025°) corresponds to the gently undulating joint in the
failure surface (segment B). The second major pole cluster
(mean pole orientation in plunge → trend notation of
55±45° → 223±050°) corresponds to the rougher tensile
failure surface (segment A).
Both minor clusters (segments C and D) are planar
joints with low variation in normal orientation that did not
play a significant role in the failure because they are pre-
existing fractures near the top of the failure surface. It
should be noted that the tensile failure surface had a
normal variation that was 2.2x greater than the sliding
surface for dip and 2.4x greater than the sliding surface for
dip direction.
Figure 5. Stereonet of failure surface showing the major
rough surface segment A (bottom left), planar surface
segment B (top right), minor planar surface C (lower
Middle), and minor planar surface D (upper middle)
4 METEOROLOGICAL ANALYSIS
Precipitation has a significant impact on the weathering of
rocks both chemically (in the case of precipitation caused
dissolution of matrix in carbonates) and physically (in the
cases of frost jacking and wetting and drying cycles). While
the underlying cause of failure on this coastal rockface is
due to gradual weathering from wave action at high tide, it
is important to identify what event triggered the rockfall. For
many rockfalls in the Park, frost jacking (repeated cycles of
freeze-thaw, where freezing water in a rockmass expands
and induces fracture propagation) is the driving seasonal
factor; however, as this event occurred in September, frost
jacking was not the primary trigger. A single major
precipitation event leading up to failure is identified and
presented in the next section as the primary trigger for the
September 5th, 2021 rockfall.
4.1 Precipitation Intensity
Failure along discontinuities in the rock can be caused by
reduced frictional shear strength between blocks caused
by increased pore pressure. Pore pressure can increase
due to heavy precipitation, which is evidenced by rainfall
events leading up to the time of the rockfall. In this case, a
large rainstorm produced over 130 mm of rain in the Park,
as shown in Figure 6, between August 31st and September
3rd, 2021. This precipitation is directly attributed to
Hurricane Ida (National Hurricane Center 2022). It is
therefore very likely that the September 5th, 2021 rockfall
was caused by a large increase in precipitation intensity
just a few days prior. The delay in the rockfall after the
rainstorm is attributed to the time required for water to
infiltrate the rockmass and add excess mass to the block,
triggering gravity-driven failure.
Figure 6. Plot of precipitation intensity during August and
September 2021 in Hopewell Rocks Provincial Park, NB
5 CONCLUSIONS AND FUTURE WORK
The rockfall that occurred on September 5th, 2021 near
Lover’s Arch at Hopewell Rocks Provincial Park, New
Brunswick, is interpreted to be the result of long-term tidal
erosion, seasonal freeze-thaw cycles, and was ultimately
triggered by a large precipitation event.
Long-term tidal erosion caused undercutting at the toe
of the shoreline cliff face. Seasonal freeze-thaw cycles
caused frost jacking that induced fracture development and
A
B
D
C

joint opening within the rockmass. The large precipitation
trigger event increased the saturated weight in the
rockmass, as well as pore pressures along fractures, to
induce gravity-driven tensile failure through the remaining
rock bridges to release the rockfall material from the cliff
wall.
This failure interpretation was determined through
structural analysis through 3D SfM photogrammetry
modelling and analysis of meteorological records leading
up to the time of the failure.
Future work is planned to analyze the volume and
spatial distribution of failure debris, which aims to aid the
Park in designing cables to block unsafe areas from public
access for future potential rockfall events. Additional future
work aims to investigate the geomechanical properties of
the rock comparing dry and saturated states.
ACKNOWLEDGMENTS
This research was financially supported by the Queen’s
University Catalyst Fund and the Natural Sciences and
Engineering Research Council of Canada. Thank you to
Erika DeGrace and others from Hopewell Rocks Provincial
Park for their ongoing support of this research, to Paul-
Mark DiFrancesco and David Bonneau for their advice on
photogrammetry model analysis, and to the useful
comments by the anonymous reviewer.
REFERENCES
Agisoft. 2021. Metashape photogrammetry model
software, v. 1.7.6, https://www.agisoft.com/.
Barton, N. and Choubey, V. 1977. The shear strength of
rock joints in theory and in practice, Rock Mechanics
and Rock Engineering, 10, 1-54.
Bilodeau, S., 2019. Overall increase of visitors for summer
2019, News Releases Government of New Brunswick
Department of Tourism, Heritage, and Culture,
Fredericton, NB.
Bonneau, D.A. 2021. Applications of remote sensing for
characterizing debris transfer processes. PhD Thesis,
Department of Geological Sciences and Geological
Engineering, Queen’s University, Kingston, ON,
Canada, 351p.
CloudCompare. 2021. CloudCompare point cloud software
[64-bit], v. 2.11.3 (Anoia), https://www.danielgm.net/cc/.
Corominas, J., Mavrouli, O., and Ruiz-Carulla, R. 2017.
Rockfall occurrence and fragmentation. In: Sassa K.,
Mikoš M., Yin Y. (eds) Advancing Culture of Living
with Landslides. WLF 2017. Springer, Cham.
https://doi.org/10.1007/978-3-319-59469-9_4.
Gong, B., Wang, S., Sloan, S.W., Sheng, D., and Tang, C.
2018. Modelling coastal cliff recession based on the GIM-
DDD method. Rock Mechanics and Rock Engineering, 51:
1077-1095. https://doi.org/10.1007/s00603-017-1382-0.
Hyslop, A., Day, J.J., Kruse, S., and Snair, K. 2021.
Structural and geomechanical analysis of the 2016
Elephant Rock failure at Hopewell Rocks Provincial
Park, New Brunswick. In: Proc. 74th Canadian
Geotechnical Society Annual Conference, GeoNiagara
2021, Sept. 26-29, Niagara Falls, ON, Canada.
Kogure, T. and Matsukura, Y. 2010. Critical notch depths
for failure of coastal limestone cliffs: case study at Kuro-
shima Island, Okinawa, Japan. Earth Surf. Process.
Landforms, 35, 1044-1056. DOI: 10.1002/esp.1940.
National Hurricane Center. 2022. 2021 Atlantic hurricane
season: North Atlantic hurricane tracking chart.
Available from [accessed 1 April 2022]:
https://www.nhc.noaa.gov/data/tcr/index.php.
Packulak, T.R., Bonneau, D.A., Dossett, W.C.,
Rudderham, G.A., and Day, J.J. 2019. Evaluating
characterization methods for joint surface roughness.
53rd US Rock Mechanics/Geomechanics Symposium,
New York, NY, USA, 23-26 June 2019, 8p.Salvini, R.,
Vanneschi, C., Coggan, J.S., and Mastrorocco, G.
2020. Evaluation of the use of UAV photogrammetry for
rock discontinuity roughness characterization. Rock
Mechanics and Rock Engineering, 53: 3699-3720.
https://doi.org/10.1007/s00603-020-02130-2.
Sturzenegger, M. and Stead, D. 2009. Close-range
terrestrial digital photogrammetry and terrestrial laser
scanning for discontinuity characterization on rock cuts.
Engineering Geology, 106, 163-182,
http://dx.doi.org/10.1016/j.enggeo.2009.03.004.
Wallace, P. 1998. Hopewell Cape: Stratigraphy of the
Windsor and Hopewell Groups. In: Discovering Rocks,
Minerals and Fossils in Atlantic Canada; A geology field
guide to selected sites in Newfoundland, Nova Scotia,
Prince Edward Island, and New Brunswick. Atlantic
Geoscience Society, 148p.
Westoby, M., Lim, M., Hogg, M., Dunlop, L., Pound, M.,
Strzelecki, M. and Woodward, J. 2020. Decoding
complex erosion responses for the mitigation of coastal
rockfall hazards using repeat terrestrial LiDAR. Remote
Sensing, 12, 2020. DOI:10.3390/rs12162620.
Wyllie, D.C. and Mah, C.W. 2004. Rock Slope Engineering:
Civil and Mining, 4th Edition. Spon Press, Taylor &
Francis Group: London and New York, 456p.