Optimising disaster response: opportunities and challenges with Uncrewed Aircraft System (UAS) technology in response to the 2020 Labour Day wildfires in Oregon, USA

Earthquake

Following significant earthquake events, Rathje and Franke (2016) used UAS to collect 2D orthophotos, video logs, or 3D geometric information to document and quantify damage patterns throughout an impacted city, capture detailed geometric information on failed infrastructure for forensic information, and map and measure ground movements to evaluate stability and further risk. Although often not available, high-quality baseline data from pre-disaster UAS flights can significantly improve quantitative damage analysis after disasters (Restas 2015). In many cases, researchers and disaster responders used lower-spatial resolution satellite data for baseline reference, given its wider availability. These data are essential to capture the initial damage from the actual event, while repeat flights can efficiently monitor changes to the ground surface, slopes, or infrastructure after the earthquake or associated aftershocks. Several recent events demonstrated these applications. For example, following the 2016 Kumamoto, Japan earthquake, Yamazaki et al. (2017) used UAS data to develop 3D models based on the Structure-from-Motion (SfM) reconstruction technique to support identification and quantification of damage throughout the affected area. Montgomery et al. (2021) also used UAS technology for reconnaissance to obtain high resolution imagery of the significant damage observed throughout Palu City and the surrounding Central Sulawesi region of Indonesia following the 2018 Palu-Donggala earthquake. The use of UAS for this application enabled rapid capture of information across a large flowside area efficiently, particularly for areas that are difficult, if not impossible, to access directly by humans to perform conventional surveying and mapping.

Tsunami

UAS have been used to collect data on the inundation extents following tsunamis as well as to generate detailed digital elevation model topographic information to simulate the substantial fluctuations in water levels when a tsunami occurs. For planning purposes, Marfai et al. (2019) used high-resolution UAS imagery and GIS software for developing disaster response and mitigation plans as well as evacuation plans by estimating areas likely to be affected by tsunamis based on scenario events. Similar to earthquakes, UAS technology is also used for damage assessment following a tsunami. For example, De Oliveira et al. (2018) used a UAS equipped with an infrared camera to detect faults in large-scale Photovoltaic plants in tsunami-affected areas. The UAS was very beneficial in terms of cost-efficiency and its ability to detect post-tsunami damage rapidly. UAS also has the potential to be used in direct lifesaving applications. Katayama et al. (2018) explored the usage of UAS technology to provide immediate evacuation guidance to persons in tsunami hazard zones. In this application, multiple UAS coordinate and continuously share information such as their location, evacuation guidance routes, and the number of people each system has guided.

Hurricane

UAS played an important role in supporting reconnaissance of hurricane-impacted areas by enabling quick map generation of heavily impacted areas. Yuan and Liu (2018) devised a framework that combines UAS with social media to quickly deploy UAS to affected areas in order to obtain timely information for response planning. Hurricane affected areas are generally extensive; hence, UAS can prove valuable to determine the extent of damage and plan the recovery efforts compared with deploying humans to the impacted sites. When compared to reliable baseline data (e.g. orthophotos; lidar digital elevation models, DEMs), remotely sensed data from UAS can serve as a low-cost tool to accurately quantify topographical changes resulting from hurricanes. For example, after Hurricane Maria, Schaefer et al. (2020) investigated changes within a damaged area by comparing UAS imagery to aerial photographic data collected before the storm. Notably, given that the UAS is relatively vulnerable to wind and rain, it has primarily been used for post-disaster damage assessment rather than directly during the event (Yeom et al. 2019). UAS usage is limited during hurricane events and in the immediate aftermath for SAR purposes until environmental conditions permit safe UAS operation.

Wildfire

UAS are being widely used and studied as a potential tool to detect wildfires. Early wildfire detection is very important to help prevent large-scale fires, which can grow rapidly. Current fire detection technologies based on satellite imaging or remote cameras tend to be slow to detect fires and have lower accuracy (Bushnaq et al. 2021). Kanand et al. (2020) used UAS equipped with high-resolution RGB and thermal cameras and a modern 5G mobile network infrastructure. They strategically deployed this for early detection of wildfires, enhancing both the speed and accuracy of detection. UAS have demonstrated remarkable capabilities, extending beyond the early detection of wildfires to include identifying and mapping burned areas. Samiappan et al. (2019) compared the UAS-based classification produced from the normalised difference vegetation index (NDVI) and a digital surface model (DSM) with the Landsat-based Burned Area Reflectance Classification (BARC). The authors noted that a UAS platform with a multispectral sensor could provide more timely data for mapping the extent of the burned area than an RGB photogrammetric camera alone.

UAS technology proves effective for supporting a variety of monitoring and detection tasks, most of which are performed manually by human observers watching video feeds from the UAS (Kanand et al. 2020; Tang et al. 2020). Researchers are currently exploring deep learning methods using UAS imagery to rapidly detect and monitor wildfires (Tang et al. 2020, Kang et al. 2023). Although capable of very efficient and promising data analysis, machine learning, more specifically deep learning, require extensive image data sets that are manually annotated (e.g. fire) in order to train a model to identify similar features in other images (Zhao et al. 2018). Hence, a hindrance to this approach will be the robust collection and systematic annotation of the very large dataset necessary to detect wildfires accurately. However, once sufficient datasets are available, deep learning models can automate the detection of fires in imagery, enabling responders to use their time more effectively in responding to priority locations.

During the event, UAS also transport and drop fire suppression materials (e.g. fire retardant, water, or other extinguishing agents) at strategic locations (Saikin et al. 2020). The UAS enables the fire to be extinguished in a safer method than the conventional crewed helicopter method, which puts people directly in harm’s way by flying over the flames and smoke from the fire. By incorporating infra-red camera technology on UAS, wildfire management teams can detect thermal energy and produce images highlighting heat signatures. This capability is critical for identifying hotspots along the perimeter of a fire zone, enhancing the safety and strategic deployment of firefighting crews (Fagen et al. 2021).

Study area

This study explores UAS usage in preparation for, during, and in response to the 2020 Labour Day Fires in Oregon, USA (Fig. 1), which initiated on 7 September 2020. These fires were some of the most catastrophic on record in the State of Oregon, killing at least 11 people, burning more than 1 million acres of land, and causing more than 40,000 people to flee their homes (Oregon Office of Emergency Management 2020). Additionally, the fires had significant economic impacts, including a USD5.9 billion loss for Oregon’s forest-dependent industries and businesses as well as recovery costs estimated at USD1.15 billion for the state (Oregon Office of Emergency Management 2021; Tillamook Headlight Herald 2021).

Fig. 1.

Map of the extent of major fires occurring during the 2020 Labour Day Fires in Oregon, USA along with approximate numbers of damaged structures and acreage. (Data sources: NWS Billings, National Institution Fire Centre (NIFC) geodata, and Incident Information system’s (Inciweb) daily fire reports.)

Interview research methodology

Fig. 2 shows an overview of the interview research methodology and how we synthesised the results. We used semi-structured interviews to collect information on UAS usage by institutions in response to the 2020 Labour Day fires. The general goal of a semi-structured interview is to gather systematic information about a set of central topics while also allowing flexibility for exploration when new issues or topics emerge (Longhurst 2009; Wilson 2014). Considering that the purpose of the interviews in this study is to collect information on the use of UAS in wildfire response and associated challenges derived from the situations and experiences of each organisation and expert, the flexibility of the semi-structured interview approach is well-suited for this study given the limited amount of information currently available.

Fig. 2.

Overview of the research methodology for the application of UAS for disaster response. The topics in the centre column were used to guide the semi- structured interviews.

We generated a questionnaire to guide the conversation and address the four aforementioned research objectives. The questionnaire consisted of open-ended questions with prompts for responses when needed. The literature review explored UAS utilisation in disaster questions (Table 1), which enabled the creation of categories for common UAS use methods (Search and Rescue, SAR; Reconnaissance and Mapping, RM; Damage Assessment, DA; Monitoring and Detection, MD). These categories then informed the questionnaire process, enabling the development of more specific questions tailored to the situations in which UAS was used. Synthesising the limitations and further research mentioned in studies, we formulated specific questions to identify the challenges of using UAS in disaster situations, which led to in-depth discussions with respondents. Another important outcome of the literature review was the identification of possible gaps between research and real-world challenges regarding using UAS in disaster situations. Through these discussions with experts, we generated questions about operational and political challenges as well as data management that were not addressed in much of the literature. Given the limited potential number of respondents with UAS experience, we used snowball sampling for participant recruitment (Goodman 1961). After each interview, we revised the questionnaire to tailor questions to the target audience based on input received from the prior interviewee(s). Oregon State University’s Institutional Review Board (IRB) reviewed the questionnaire and associated interview methodology and rendered an exempt decision based on the determination that it was not classified as human subject research.

We employed several methods to identify potential interviewees consisting of government officials and UAS experts involved in wildfire response. First, we identified potential interviewees through the collaboration networks of the authors and their colleagues. The research team then extensively searched media articles to identify key personnel at agencies involved in the response. Second, we explored public agency websites to identify additional potential respondents. If no clear candidate was identified on the webpage, we contacted the heads of relevant divisions to identify suitable personnel to participate in the interviews. Lastly, we asked interviewees to provide the names of other potential respondents.

We conducted each interview as a web-based video conference lasting between 40 and 60 min. The interview questions generally followed the order are in Fig. 2, but were adapted based on the conversation’s flow. We first asked questions about the use of UAS, in general, within the respondent’s institution and their use in wildfire situations. Through this question, the authors tried to understand each institution’s current UAS uses and applications. The authors then asked if there were any problems or challenges that could be improved in the processes for using UAS. This question aimed to identify reasons limiting UAS usage. Finally, we discussed UAS-related policies, training resources, as well as UAS data management and sharing. Some respondents provided follow-up materials, such as presentations and documents, to be reviewed by the research team. Following the interviews, we transcribed key information based on interview notes and recordings. At least two research team members participated in the interviews to reduce potential bias in transcription and interpretation. We also employed a process where multiple team members carefully reviewed the transcriptions to ensure consistency and objectivity in reporting key information.

We applied qualitative analysis methodology to the semi-structured interview results in order to explore how they used UAS before and after wildfire disasters, what were the major barriers to their use, and how they managed the collected UAS data. We integrated each respondent’s response to each question. Lastly, we reclassified the integrated data according to the three pre-determined themes: (1) application; (2) challenges; (3) and data management), with three sub-themes: (1) timeline; (2) operational and political; (3) and data lifecycle. This process identifies common information mentioned by several institutions and information unique to a particular respondent.

UAS applications

Respondents highlighted UAS deployments that occurred before, during, and after the disaster situation. Table 2 shows the utilisation of UAS in wildfires by each institution organised within the context of the event timeline. Agency A was the only institution that indicated the use of UAS before wildfires to identify vulnerable areas and gather image data on these areas. They used these baseline data to prepare for wildfires through strategic mitigation practices.

During wildfire response, the usage of UAS was diverse. Many institutions mentioned that UAS may replace conventional helicopter operations as UAS can perform missions more safely and efficiently. Agency C used UAS for aerial ignition purposes when firefighter safety on the ground was compromised or when large land areas required ignition to slow and help contain the wildfire. Aerial ignition requires flying low to the ground and at a slow speed while igniting the fire. This method of flying is risky and can cause accidents using a helicopter. UAS can substantially reduce the risk. UAS paired with infrared cameras can capture high-resolution photographs during a wildfire. Due to safe flying altitudes, most crewed aircraft will obtain photos with smoke; however, a UAS can fly at significantly lower altitudes above ground and obtain the necessary photos to understand where burning and combustion are occurring. Agencies B and C used UAS for response management through real-time videos and mapping.

After the fires were contained, there was considerable UAS use across the different institutions. All institutions used UAS to collect imagery to evaluate the extent and type of damage resulting from the wildfires. They used vulnerable regions (communities and forested land), the pre-fire imagery for comparison purposes. Agency A used UAS post-wildfire to monitor emissions and community activity to determine which regions are safe for on-the-ground crews to commence the forest restoration processes. Agency D used UAS after a fire to determine where trees threaten the right of way on roadways. UAS imagery showed where trees were significantly compromised and which trees needed to be removed, such that transportation through wildfire-impacted regions is not compromised. Agency E was also conducting an ongoing study using multispectral imagery obtained by UAS to categorise the conditions of trees. Currently, these data are acquired using ground-based lidar; however, a UAS with kinematic lidar is expected to expedite this work and protect crew members. During this evaluation of trees for removal, Agency D also used UAS to monitor and rescue osprey baby birds living in the burn zone. A nest had been constructed on top of a burned, which needed to be removed for safety reasons given its proximity to the highway. The UAS minimised disturbing the osprey by obtaining nest imagery, counting how many baby ospreys were in the nest, and safely removing the nest from the unstable tree.

In addition to government institutions, Agency E, whose primary role is to assist the natural hazards and disaster research communities in collecting perishable data in post-natural disaster reconnaissance. Agency E used UAS to quickly investigate the extent of the wildfire impacted area and produce detailed orthomosaic maps for future disaster monitoring, change detection, debris removal work, and geotechnical assessment of fire-affected structures and landscapes. Note that Agency E is not a first responder performing direct lifesaving or fire suppression tasks. Instead, they used the UAS to obtain research-related data for post-fire damage assessment and disaster recovery.

Challenges in using UAS

As mentioned in the UAS utilisation section, most institutions noted many positive roles for UAS in wildfire situations. However, there were also challenges cited by each institution when using UAS during and after the 2020 Labour Day fires. We classified these challenges into two categories: operational and political.

UAS data management

UAS technology can produce various data types for applications in wildfire response, from basic images to videos to purpose-built maps. The efficiency of a UAS also tends to result in substantial data volumes. Interviewees discussed how they handle UAS data in each institution, particularly from a data-sharing perspective. Notably, most institutions faced similar concerns about data management and sharing, including the difficulty of sharing across institution boundaries and refreshing data with the most up-to-date information. These concerns are summarised in the data lifecycle (Fig. 3), which refers to the sequence of processes from data creation, storage, use, sharing, archival, and destruction.

Fig. 3.

Stages of the UAS Data lifecycle. Adapted from Rahul and Banyal (2020).

Most interviewees mentioned the critical need for standards and specific criteria throughout the data life cycle. For example, Agency C noted the need for protocols to distinguish which data are classified as federal data and which are public data. This determination can be ambiguous. Agency C also noted challenges in determining which data to archive as a formal record and which data to dispose of whose purpose was to inform the immediate situation and was soon outdated. Attempting to archive all data are a substantial burden on personnel to provide all necessary metadata to comply with federal data sharing standards. It also creates bottlenecks and confusion in the reuse of the data in the future, given the large data volume users must sift through to locate the data of interest, ultimately causing the most important information to be lost in the vast data repositories.

A consistent data storage format is necessary for effective (and, in some cases, legal) data sharing. However, some institutions, such as Agencies A and C, do not have standards in place. One interviewee noted that this lack of standardisation poses difficulty in sharing data within Agency C. Agency B said that they are working on building an improved data storage system to address these issues.