This paper was originally published in Australian Geomechanics in September 2018 and is reproduced by permission of AGS
ABSTRACT
Unmanned Aerial Vehicles (UAVs or drones) facilitate data collection that allows rapid and enhanced geotechnical assessment of the risks associated with landslides, slopes and structures, during normal conditions and following natural hazard events.
WSP Opus has used UAVs over the past 10 years in New Zealand to assist engineering assessments for highways, railways, and infrastructure for local authorities, government departments and private clients. The early uses were mainly for occasional geotechnical slope and site mapping purposes. However, since 2011 UAVs have been systematically used for surveying and geotechnical applications. Examples are presented in this paper of the use of UAVs for assessment of slopes and existing landslides, as well as for post disaster assessments.
UAVs were utilized following the damaging February 2011 (M6.3) Christchurch Earthquake, the 2011-2015 storm events in central New Zealand, and the November 2016 (M7.8) Kaikoura Earthquake to inspect and provide damage records of sections of highways, slopes, and river stopbanks (levees), and to facilitate rapid development of remedial options.
The data gathered from UAVs have been used with post processing of imagery to create 3D terrain models. Comparison of periodic UAV derived models or comparison with previous LiDAR models enabled the detection and monitoring of slope change in areas affected by slope failures along transportation corridors. UAVs were also used to inspect and obtain geological data to assist with the assessment of rock slope stability. As seen in the Blue Mountain’s Rail Corridor (NSW) in Australia, the use of UAVs provided a much safer alternative than having staff carry out observations directly (for example by rope access), given the hazardous nature of these locations. A comparison of geological and other data obtained from UAV inspections with those gained from conventional methods showed good correlation.
In addition to the safety benefits, the use of UAVs has enabled better and early decisions to be made to manage the risks associated with hazardous sites, and the rapid development of remedial solutions.
1. INTRODUCTION
The use of Unmanned Aerial Vehicles (UAVs) or drones, has grown rapidly in the past few years, largely through the rapid advancement and availability of affordable technology and the ability to rapidly process the captured data.
Many UAVs incorporate gimbal stabilised photography and high definition video capture with GPS geolocation of imagery. Relatively large areas can be rapidly captured and modelled accurately by post processing imagery utilising ‘structure from motion’ photogrammetric software. Software outputs typically include detailed ortho-corrected photo mosaic imagery and accurate 3D models of structures or terrain. These outputs can also be used for many CAD and GIS applications. Through these combined outputs, UAVs provide a means of rapid assessment of structures and slopes, and development of remedial options.
Continued rapid advances in the technology are anticipated with developments such as UAVs that can fly in all weather, advanced obstacle avoidance capability and the use of new sensors on UAVs such as infrared and LiDAR (Light Detection and Ranging). Such technologies are discussed by Olsen and Gillons (2015).
The current paper provides a summary of our findings regarding the usefulness and applications for UAVs in various phases of geotechnical engineering and natural hazard response over the past 10 years.
2. UAV TECHNOLOGY
2.1 Early UAV Inspections
One of the authors has used UAV technology in the Auckland Region from 2007. Initially this was using a frangible foam ‘Easystar’ model, electric powered, glider-type fixed-wing, radio controlled model aircraft carrying a remotely triggered Pentax 12 Mega Pixel compact digital camera (Figure 1). The ‘Easystar’ was largely based upon a hand launched hobbyist radio control training plane. It had a 1.4 m wingspan, and was able to remain airborne for up to 30 minutes, and operate up to 600 m away from the operator while remaining in visual control range. The camera was remotely triggered by an infra-red controller and could be positioned to take either oblique or vertical imagery of subject material. The aircraft was initially developed and utilised for remote access to difficult sites, slide inspections, site progress photography, and providing promotional aerial photos for clients and company flyers. While the ‘Easystar’ was launched and flown in left-hand circuits around the subject, most of the photography was ‘guesswork’ undertaken without any downlink or telemetry.
The first practical usage of the ‘Easystar’ for emergency response aerial support came during August 2008. Severe winter storms and extremely heavy rainfall reactivated an old deep-seated block slide on Turei Hill in Kawakawa Bay in the Auckland region. The slide threatened the closure of the coastal road and cutting off communications to a number of east coast communities. The ‘Easystar’ was utilised at Auckland City Council’s request to provide rapid response visual survey of the extent of cracking and movement of the slump block. A number of reconnaissance surveys were undertaken over a number of weeks; with the photos being then “mosaicked” and compared with previous imagery. The imagery was subsequently used to disseminate visual information to local community groups and multiple media outlets, to help the public understand the risks.
Figure 1: The “Easystar” fixed-wing radio controlled aerial camera aircraft with camera setup for vertical images (left) and oblique images (right). The UAV was developed in-house by one of the authors of this paper.
2.2 Later UAV Inspections
From July 2014, as more advanced UAV technology became commercially available, the DJI Phantom quadcopter UAV became the most common UAV (Figure 2) used by WSP Opus. The Phantom is operated by a tablet based application that incorporates gimbal stabilised 12 or 20 Mega Pixel photography or high definition video capture. While it is essentially a ‘prosumer’ level UAV, the ability of the Phantom to take-off vertically, hover in a GPS assisted stationary position (not affected by wind drift) combined with live image and video capture (some of the Phantoms can live-stream video via YouTube) have revolutionised the ability to quickly and safely carry out inspections and assess hard to access unstable terrain and structures.
The Phantom can also be operated in an autonomous mode, flying to pre-programmed waypoints or flying along gridlines for detailed surveying. Typical flight times vary between 15 and 20 minutes.
Figure 2: DJI Phantom multi-copter in use for inspection of landslides along railway 50 km north of Wanganui
The location of the central New Zealand sites highlighted in this paper are shown in Figure 3, with relevant Figure numbers indicated.
Figure 3: Location map of figures of central New Zealand sites presented in this paper
3. SURVEYING SITES WITH UAVs
We have implemented the use of structure-from-motion (photogrammetric) software in conjunction with our UAV operations. Structure from motion software has revolutionised photogrammetry as it can automatically process imagery acquired by UAV, aircraft or ground based platforms purely on image content. This process enables the conversion of images into precise, customisable digital data for a wide range of applications including for GIS and CAD (Figure 4).
Figure 4: Ortho-mosaic image showing May 2015 debris flows near Paraparaumu (north of Wellington) and corresponding 1 m contour plan of same site, from post-processing UAV imagery
The use of UAVs in conjunction with good survey control has given the ability to easily capture up-to-date site conditions and produce an accurate 3D terrain model that can be compared with subsequent captures. This has enabled:
a) The creation of base plans for use by field staff in mapping
b) Assessment of the areas and volumes of earth at affected sites
c) Preparation of cross section and plan outputs for use in the assessment
d) Rapid design of engineered remedial works, and
e) Comparison of 3D model outputs with data from previous flights, for example, for screening for slope changes as shown on Figure 5.
Figure 5: UAV ortho-photos from February 2017 and September 2017 of a 1 km length of slope south of Paraparaumu Station, and a comparison of the two UAV 3D models, with blue indicating an increase in elevation and yellow/red a decrease in elevation. Two known landslide sites are circled, as is an area of growing vegetation.
4 POST STORM EVENT UAV INSPECTIONS
A significant storm event in June 2015 resulted in extensive flooding and land sliding in the Wanganui and Taranaki regions of NZ. Helicopter inspections were used for immediate response by authorities, such as KiwiRail and the Wanganui City Council. UAVs were used in the following weeks to provide data to characterise the extent of damage and to facilitate assessment and design of remedial solutions.
The 4 km long section of railway immediately north of Wanganui known as the Westmere Bank was subject to numerous landslides both above and below the railway line. Due to the long sections of railway to be assessed, a fixed wing UAV (Figure 6 inset) was used to capture imagery of this section and also a 2 km section of damaged Council road. From the photos taken, Ortho-mosaic strips were generated within 24 to 48 hrs of the flight allowing 1:1000 scale base maps to be created and used to enable field reconnaissance and characterisation of the landslides (Figure 6). Landslides and site features were clearly visible on the Ortho-photo base maps which enabled rapid and accurate recording of site damage.
Figure 6: Base map showing ‘recent’ storm damage, generated from fixed wing survey (Fixed wing UAV shown inset, top right) with geologist’s landslide observations drawn / overlaid.
A Phantom multi-copter UAV was used during the reconnaissance to capture imagery at lower altitudes at specific sites within the Westmere Bank and a critical damage site 50 km north of Wanganui. This provided higher resolution ground models and video footage to enable better definition of the extent and nature of slope instability (Figure 7). Production of cross sections from the multi-copter’s 3D model(s) aided rapid assessment of the risk to the railway line. The use of the multi-copter UAV enabled more rapid assessment of the condition of the rail corridor to enable decisions to be made about reopening. These included UAV inspection of:
- Steep slopes below the railway line that would otherwise only have been visible by abseil but not safe to access, and
- Extensive unstable areas above the railway line, which enabled clarification of the extent and mode of failure of land sliding.
Figure 7: Ortho-photo of slide site generated from multi-copter imagery (launch site is shown in Fig.2 and is located on hazard free flat land at far right)
An earlier storm in May 2015 affected the southern part of the North Island. In parts of the Kapiti Coast, it was the heaviest recorded for more than 50 years, with peak rainfall exceeding monthly averages in a single day.
The Phantom UAV was deployed in the Wanganui and Wellington Districts to assist with the assessment of unstable terrain created by the deluge of rain and associated flooding. Two example sites are discussed below.
The Paekakariki coastal road section of State Highway 1 (SH1) and the adjacent North Island Main Trunk (NIMT) railway line, north of Wellington, were closed due to a large debris flow which originated in the steep hills above (Figures 8 and 9). The source and likelihood of further debris flows at the site, were not able to be identified immediately prior to clearance of debris and reopening. Utilising live video feed, the Phantom UAV was flown up the inaccessible stream valleys and successfully identified the source of debris which blocked the rail and road during the storm.
Figure 8: Remnants of a debris flow which came downstream (top) and blocked rail and road. UAV was flown from shore platform (foreground).
Figure 9: Close-up of May 2015 debris flow source enabled: talus and debris undermined by rapidly flowing and swollen stream originating in steep country (right).
5. POST EARTHQUAKE EVENT UAV INSPECTIONS
5.1 CANTERBURY EARTHQUAKES 2010 – 2011
The 2010 Darfield earthquake struck with a magnitude of 7.1 on 4th September 2010. Numerous damaging aftershocks followed, the strongest of which occurred on 22 February 2011 had a magnitude 6.3 (Christchurch earthquake). Because this aftershock was centred very close to Christchurch, it was much more destructive and resulted in the deaths of 185 people. This event had a maximum peak ground acceleration of 2.2g, the largest ever recorded in New Zealand. Liquefaction damage on low lying terrain was widespread and significant areas of rock fall and potential rock fall risk created significant risk to life and property in the Port Hills area of Christchurch. Identification of the extent of damage as soon as possible was a priority, which in the Port Hills area, in particular, was greatly assisted by the use of UAV technology.
5.2 UAV INSPECTIONS, POST FEBRUARY 2011 EARTHQUAKE
Following the February 2011 earthquake, and after considerable consultation with the Emergency Response Management, the ‘Easystar’ UAV was deployed on the 8th of March to obtain high quality imagery of ground damage on very steep and high bluffs in the Port Hills area of Christchurch. These bluffs were vulnerable to rock fall and landslides, both during the quake and post-quake aftershocks. The ‘Easystar’ was operated from safe, remote locations and utilised to photograph and identify the various rock fall sources and patterns to help assess the residual rock fall risk to property, roads and personnel working on remedial efforts (refer Figures 10 and 11). The imagery captured was particularly useful for determining safe access routes where site access was deemed risky for access on foot and for abseiling teams. The imagery was shared with a range of organisations involved in post-disaster response for visualisation and forward planning.
Figure 10: ‘Easystar’ imagery of earthquake effects in the Port Hills, east of Christchurch. Boulder strewn slopes and road damage near Mt. Cavendish Gondola
Figure 11: Cliff line cracking (arrowed) near Sumner Heads east of Christchurch, this area eventually failed in later aftershocks.
5.3 UAV inspections following 2016 Kaikōura Earthquakes
The M 7.8 Kaikōura earthquake of 14 November 2016 caused widespread ground and structural damage in the upper South Island and localised damage also in Wellington in the lower North Island of New Zealand. The main north – south road and rail access routes were severely damaged in the northeast transport corridor of the upper South Island. UAV imagery, in conjunction with survey of ground control points, allowed rapid production of 3D models and orthophotos to document the actual positions of tension cracks and other ground damage (Figure 12). In particular they provided valuable early documentation of damage at a number of sites prior to evidence being removed.
Figure 12: Earthquake induced landslide damage to State Highway 1 and the South Island main trunk railway line near Clarence River north of Kaikōura. Ground damage extended to near the centre line along a nearly 500 m long section of SH1 at what became known as “the sand pit” due to the sand substrate at the site
Though comprehensive aerial LiDAR survey was subsequently carried out to provide a topographic survey to allow engineering assessments and remedial works design to proceed, because it took some time to procure the LiDAR data UAVs proved very useful for providing rapid documentation to allow response and geotechnical assessment (Figure 13).
Figure 13: An accurate 3D model of the ‘sand pit’ site was produced from UAV imagery with survey of ground control points (crosses visible in Figure 12) to allow remedial works options assessment to proceed at an early stage, in advance of receiving new LiDAR data.
Capturing UAV imagery along the banks of rivers affected by lateral spreading allowed rapid documentation of the location of tension cracks and liquefaction ejecta, prior to these being obscured or repaired (Figure 14). This allowed the development of remedial measures to upgrade the damaged stop banks.
Figure 14: UAV capture of tension cracks resulting from lateral spreading of river bank slopes toward the Taylor River in Blenheim, in response to liquefaction induced from the 14 November 2016 Kaikōura Earthquake. Survey control cross just visible (circled) between track and river at top of photo
6. ROCK FACE STRUCTURAL DATA FROM UAV SURVEYS
In addition to being able to remotely carry out inspections using UAVs, the 3D models generated from UAVs can be much more comprehensive than from other sources due to the ability to ‘sight the site’ from multiple perspectives. The 3D models generated have been used to obtain rock defect (or facet discontinuity) orientations through software packages manually or automatically. This gives the potential for a much more comprehensive assessment of a site than would typically be possible from a site visit. Notwithstanding this, checks are required to validate the data, typically involving independent checks of defect orientations on site.
6.1 Mt Taranaki Rockfall
A rain induced rockfall of about 500 m3 buried the popular Mt Taranaki round the mountain walking track in February 2018. Detailed inspection of the rockfall source area was carried out using a UAV from a safe distance (Figure 15 left) with more detailed assessment carried out in the office using 3D mesh and point cloud models. The point cloud model enabled remote mapping of rock defect orientations using automatic facet picking (Figure 15 right), as well as manual measurements of facet orientations using different software. In this case the main defects controlling the stability of this rock face are the overhanging (purple) joints (providing a toppling mechanism), in conjunction with moderate angled outward dipping (green) joints providing a planar failure release.
Figure 15: UAV image of rockfall site on Mt Taranaki walking track, with UAV pilot visible (circled) just to left of stream bed. Rockfall source area is white area near top of 50 m high vertical lava bluff (left). Cloud Compare facet analysis of point cloud data for the 30 m wide and 20 m high source area is shown at right, with facets (defect) sets colour coded by orientation.
6.2 Rock Cut Stability Assessment In Blue Mountains, NSW
A multi-copter UAV was used to obtain data to assess the stability of a rock cut at a New South Wales rail tunnel portal (Figure 16). The site was dominated by a large, near vertical, approximately 15 m high rock exposure above and behind a tunnel portal. A rock block on the main exposure appeared to be free-standing and separated from the main rock exposure by a continuous sub-vertical stress relief joint. The block was previously supported by an anchored concrete buttress, however, its stability was of concern. Safe access to the site posed a number of challenges due to the potential instability, working at height concerns (cliffs), heritage-significant structures and overhead wiring. Use of a drone provided not only a greater level of access to the entire rock cut exposure, but a safer and more cost-effective alternative to rope access. Manual collection of discontinuity data was performed in areas with safer access to compare with the data obtained by the drone.
Figure 16: General site layout labelling features (left), focused image on main instability concern positioned above tunnel portal (right)
A number of survey reference points were placed on each rock exposure before capturing high-resolution geo-referenced images with the UAV. Due to the large quantity of data and images that became available through this method, the rock exposures at the site had to be divided into a number of sub-groups for processing. Subsequently, they were then combined into a single point cloud and a high-resolution 3D rendering of the site (Figure 17). The 3D model fed into a user-led and algorithmic assessment, with aid from software 3DM Analyst Suite (by Adam Technology). The assessment involved the identification and marking of rock face discontinuities and planes, then represented them by oriented coloured disks (Figure 18). Disc data, consisting of discontinuity assignment group, orientation, dip, persistence and additionally assigned geotechnical properties were then stereographically represented. This allowed for a kinematic analysis of failure mechanisms and the development of stabilisation measures. The data obtained from conventional rock-face mapping methods obtained from more accessible areas showed strong correlation with the drone obtained data.
Figure 17: Example point cloud (left) and corresponding geo-referenced image overlay, isolating the exposure above the portal (using 1 of 9 sub-groups)
Figure 18: 3D Point Cloud with Mapping Planes. Turquoise surface formed by points with yellow and red disks representing some observed discontinuity planes. Tunnel 3 portal is visible in the middle at the bottom of image.
In areas where vegetation, lighting and overhead wiring restricted image and point cloud quality, the drone was still able to capture rock features and image angles not visible from the ground. This approach was ‘safer’ than rope access methods. The nature of the digital data also allowed for an increased review of data and a higher level of input from field staff and more senior professionals.
7. RISK ASSESSMENT FOR CLIFFS IN CAMPING GROUND
An example of the effectiveness of assessing a larger site is illustrated in the assessment of cliff stability risks for the Opunake Holiday Park in Taranaki. The site has a history of rock fall and landslide events affecting the facilities (Figure 19 inset). In order to facilitate the assessment, UAV imagery was captured with survey control placed around the site to enable an accurate 3D model which allowed preparation of an accurate basemap (Figure 19) and cross sections at strategic locations (Figure 20). While the 3D model does not have full ground coverage due to the vegetated escarpment, ground coverage has been maximised by obtaining UAV shots at varying angles. Interpolation of ground levels was made manually on some profiles in areas where the ground was not visible in the model (Figure 20). The various outputs allowed a robust and rapid assessment of site characteristics and visually demonstrated clearly the options available for the local council to manage the risks.
Figure 19:Ortho-photo base map of Opunake Holiday Park from UAV imagery, showing locations of cliff cross sections generated to assess different sectors of the site.2015 landslide (inset photo) occurred on cross section 3.
Figure 20: Cross section 2 highlights the hazard areas at carpark areas above and below an 18m high cliff and cross section 7 shows the main access road and amenities building, showing risk mitigation options to reduce risks from slope failures, including cutback, anchoring and barriers.
8. CERTIFICATION FOR UAV PILOTS
UAV operators have a duty of care to the public and the owners of the land and facilities they fly over, and hence are required to be operated safely in gathering site data. UAV rules in Australia and New Zealand do differ. In New Zealand, prior to 1 August 2015, skilled but essentially untrained operators were able to use UAVs in many applications within public airspace (operating under the old Civil Aviation Authority (CAA) model aircraft Part 101 rules). More stringent CAA regulations introduced at that date require operators to gain permission from property owners and all people beneath the intended UAV flightpath prior to operating. Many authorities in New Zealand now require UAV operators to be certified under the new Part 102 rules, as the new default ‘minimum standard’ for UAV operations.
In Australia, the options through CASA are a lightweight UAV license which provides limited scope, or a full professional licence which provides greater freedom (in line with what a helicopter operator would have) than the New Zealand CAA Part 102 certification. At this stage licenses are not transferable from one country to the other.
9. DISCUSSION ON THE USE OF UAVS FOR SITE ASSESSMENT AND RESPONSE
Low cost UAVs can provide invaluable information on otherwise inaccessible sites. The appropriate use of UAVs and robotic vehicles can provide major benefits in terms of safety of field staff, and due to their small size and typically short inspection duration usually have negligible impact on immediate (emergency) response operations and subsequent operation of the facility (e.g. road).
While UAVs can provide near instantaneous coverage of the effects of a natural hazard event site, ironically sometimes it has taken days or in some cases weeks to convince authorities to understand and adopt the use of this technology. This illustrates the importance of pre-planning to ensure that authorities are aware of the full capabilities and benefits of the UAV based data capture, and the safety measures that are in place. This will enable more effective use of the UAV technology to assist with assessment and event response.
In terms of choosing appropriate UAV technology, the following comments are provided:
Fixed wing UAVs typically cover larger areas and therefore, require fewer launch areas and shorter field time than Multi-rotor UAVs, but have a disadvantage in that they require large open spaces for landing/recovery.
Multi-rotor UAVs are able to be held stationary and are ideal for close-up inspections of key features e.g. using high definition video or high density photography for production of detailed 3D models. Because of their small size, we have found that the DJI Phantom UAVs lend themselves very well to inspections adjacent to live highways, providing suitable precautions are taken to minimise driver distraction while working at quite close quarters.
Systematic photography of a site enables subsequent production of 3D models. The 3D models facilitate rapid production of survey quality base map plans and cross sections, which can be used to quickly assess current conditions and associated risks. Survey control can be either inbuilt within the UAV (e.g. RTK GPS) or introduced by way of independently surveyed control points that are visible within the images. 3D model fly-through visualisations also provide a very useful tool to allow decision makers to understand the extent of damage and other constraints at the site (e.g. proximity of overhead services).
One significant current challenge for the photogrammetry based modelling used is that it picks up the vegetation canopy rather than the ground surface, with the more costly LiDAR based system typically much better at penetrating through to the ground surface.
In some cases, the UAV operator needs to be chosen for the specific task. In many cases a UAV operator specialising in a particular discipline e.g. geotechnical or structural engineering can expedite and optimise the data collection, by focussing on the key elements of potential concern. To expedite the value of the UAV outputs, these can be made available on a shared server or website. Plan / drawing outputs should be clearly labelled with a statement of accuracy/disclaimer to prevent subsequent inappropriate use.
11. CONCLUSION
UAV technology has been increasingly used for site assessments including natural hazard event assessments by the authors, and the case studies presented illustrate the value of this tool. The use of UAV technology greatly helps facilitate the rapid collection of data for subsequent processing, and use for assessment and monitoring. Collection and the rapid dissemination of imagery and accurate base maps generated from 3D models typically can expedite event response. Subsequent repeat UAV imagery collection at sites of interest can be used to accurately determine the site changes that have occurred over time, particularly if accurate survey control is incorporated, e.g. from land movement or from site earthworks activities. The use of UAVs has revolutionised the inspection, survey and assessment of slope instability, particularly after severe storm events and earthquakes. The potential value of UAV technology and software applications for UAVs for geotechnical applications is expected to continue to increase rapidly, in parallel with the rapid development of supporting technologies. Such applications are expected to rapidly enhance the value of digital engineering in geotechnical engineering practice.
12. REFERENCES
- Follas H, Stewart DL & Lester J. (2016). Effective Post-disaster Reconnaissance using Unmanned Aerial Vehicles for Emergency Response, Recovery and Research. NZSEE 2016, Christchurch (available online)
- Olsen M.J. & Gillins. D.T. (2015). How can Geomatics Technologies benefit Geotechnical Studies? 6th International Conference on Earthquake Geotechnical Engineering (6ICEGE), November 2015, Christchurch, NZ.
