Identifying increasing risks of hazards for northern land-users caused by permafrost thaw: integrating scientific and community-based research approaches

Worldwide, communities are increasingly coping with altered environmental conditions due to climate change (IPCC 2018). Identifying the causes and consequences of these environmental changes is a research priority as this information is needed by decision-makers to help support adaptation planning in an uncertain future. This need is particularly evident in northern regions where rapid climate warming is causing widespread permafrost thaw (IPCC 2018). Given that people and nature are inextricably linked, overcoming environmental challenges, such as thawing permafrost, will inevitably require the use of both social and ecological sciences (Liu et al 2007, Milner-Gulland 2012, Fischer et al 2015).

Social-ecological systems (SES) research has been widely accepted and touted as the direction research and funding agencies are headed (Chapin et al 2016). Many approaches to SES (see review by Guerrero et al 2018) focus heavily on the integration of disciplines but do not require or incorporate insight from local communities, public interest groups, or non-scientist communities that may be affected by problems and attempted solutions (Fischer et al 2015, Turner et al 2016). By linking scientific expertise and sophisticated large-scale modeling and data capabilities with community-based local knowledge of finer-scale change and adaptation histories, power, experience, and wisdom can be shared in a bidirectional way to enhance understanding and outcomes (Chapin et al 2016, figure 1). Here we use a case study from interior Alaska that couples modeling of permafrost vulnerability with community-derived data of landscape hazard experiences by land-users to demonstrate how enhanced understanding of the permafrost land-user system can occur through bidirectional knowledge sharing.

Zoom In Zoom Out Reset image size

Traditional harvest resources are the natural and cultural resources used by households to support their livelihoods and include the traditional use of wild resources for food shelter, fuel, clothing, tools, transportation, handicraft articles, customary trade, barter, and sharing (Huntington and Fox 2005, Ford and Furgal 2009). For many northern residents, traditional harvest plays a vital role in social, economic, nutritional, and spiritual wellbeing (Wolfe and Walker 1987, Brinkman et al 2016). Traditional harvests in the Arctic-Boreal region of Alaska have declined by 30%–50% in the past 30 years (Wolfe and Walker 1987, Fall and Kostick 2018). While this decline is likely due to a combination of complex interactions amongst social, economic, regulatory, and environmental factors (Ford and Smit 2004), land-users have routinely identified environmental change as a key factor impacting their ability to efficiently and safely access traditional harvest resources (Brinkman et al 2016, Cold et al 2020).

While climate change is causing widespread and diverse types of environmental changes, one change that is particularly relevant to northern environments is the increasing rate of thawing of permafrost (Camill 2005, Jorgenson et al 2006, Sniderhan and Baltzer 2016, Gibson et al 2018, IPCC 2018, Schuur and Mack 2018, Biskaborn et al 2019, Lewkowicz and Way 2019). The area of permafrost in the northern hemisphere is expected to decline by 20%–35% by the mid-21st century (IPCC 2018). There is mounting evidence that permafrost thaw is driving widespread landscape change (IPCC 2018 and references therein), and anecdotal evidence that permafrost thaw creates hazards for land-users when out on the land. There is limited quantitative data to test the relationship between permafrost thaw and the generation of landscape hazards for land-users, posing a barrier to northern government's and communities' ability to plan and adapt to the impacts of permafrost thaw.

Land-users can observe and immediately attribute some environmental changes to permafrost thaw. We consider those changes to be direct impacts. These can include land slumping, riverbank erosion, falling over or 'drunken' trees, and changes to infrastructure such as damage to roads or buildings (Kokelj et al 2013, Baltzer et al 2014). Land-users may also experience, but not immediately attribute some changes to permafrost thaw, though more in-depth investigations may reveal the linkage to changes in the permafrost. We defined those as indirect impacts. Indirect impacts can include changes in river flow, altered vegetation communities (Schuur et al 2007), fluctuations in wildlife abundance and distribution (Chin et al 2016), or changes in water quality (Tank et al 2016).

Past research has explored how permafrost thaw either directly or indirectly affects land-based ecosystem services, food security (Chapin et al 2006), archeological sites (Andrews et al 2016), community relocations (Maldonado et al 2013), use of traditional travel routes (Andrews et al 2016, Proverbs et al 2020), and wildlife distribution (Tape et al 2016). While these studies provide insight into how land-users may be affected, our understanding remains largely qualitative, anecdotal, or conceptual with limited research quantifying the magnitude and frequency of specific impacts. Being able to quantify the extent of permafrost-driven impacts for land-users is important as it will allow us to couple this understanding with large-scale climate models to predict future impacts for communities, ultimately allowing for more generalizable, proactive, and effective adaptation planning.

Effective adaptation planning will require quantifying the extent to which permafrost thaw creates hazards for land-users as well as the severity of the consequences for human access to those resources. A key component of ecosystem service 'availability' is the safe and reliable access to a given resource (Brinkman et al 2016). For example, residents of northern communities perceived that climate-related changes in the environment are having significantly greater impacts on access to resources than the abundance and distribution of those same resources (Brinkman et al 2016). Alaskan communities have identified safety considerations while on the land (Brubaker et al 2011, Clark et al 2016, Cold et al 2020) and unpredictable conditions in the physical environment (Brinkman et al 2016) as key concerns regarding current and future harvesting activities. For example, wildfire has been shown to restrict access, topple traplines, and burn shelters that are critical to safe travel (Chapin et al 2006). These represent 'hazards'; any environmental condition that negatively affects a land-user's travel, access to a given area, or their safety. Permafrost thaw can likely create new hazards and exacerbate ones that are already affecting land-users.

To quantify the extent to which permafrost thaw generates hazards for land-users, we apply an integrated framework (figure 1) to address three core objectives. Firstly, we quantified the extent to which permafrost thaw was a source of hazards within a dataset of hazards identified by land-users. Secondly, using data derived from interviews with land-users, we then aimed to understand how these permafrost-driven hazards affect land-user's activities, access to the land, and their safety. Finally, by combining local-level land-user data with large-scale models of permafrost thaw vulnerability we quantified the extent and the potential for permafrost thaw to generate hazards for land-users.

The Yukon River basin of interior Alaska (figure 2) was used as the study area to quantify the extent to which permafrost thaw generates hazards for land-users (supplemental 1 (available online at stacks.iop.org/ERL/16/064047/mmedia)). As part of a NASA ABoVE project (ID # NNX15AT72A), an intensive year-long partnership was developed with nine rural communities that are representative of a range of social and ecological conditions within the study area (Cold et al 2020, figure 2, table S1). The data collected by the communities provide a personal photo, written, and oral documentation of experiences with landscape hazards in the study area (i.e. community-based knowledge, figure 1). All interviewees and participants gave their informed written consent to participate in this study in accordance with the University of Alaska Fairbanks Human Institutional Review Board (IRB # 700936). Only participants who consented to have their data points presented have been included visually in figures in this study. Therefore, we will be reporting statistics on all data collected, but only visually displaying the geospatial data points and photos that have been approved by the participants to share with public audiences (figures 2, 4 and 6).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The population size across the nine communities as of 2010/2011 was between 46 and 1239 residents, and are a mix of European descent and Native Alaskan (Koyukon, Holikachuk, Deg Hit'an Athabascan, and Gwich'in). Traditional harvest practices are a way of life for all partner communities and residents use a series of roads, trails, and navigable waterways to access local wild resources. Travel along these routes consists of passenger vehicles, snowmobiles, ATVs, and boats. The nine communities vary in their connectivity to roads, and per capita consumption of wild food is known to increase as the distance from the road system (Burnsilver et al 2016). Given this, we report results at both the regional level (pooling all communities) and for road-connected and non-road-connected (remote) communities. Tok, Delta Junction, and Healy are connected via a road network and have greater access to commercial resources such as fuel and groceries, and the cost of commercial goods is less compared to the remote communities (Goldsmith 2007) of Nulato, Grayling, Holy Cross, Lake Minchumina, Beaver and Venetie.

Land-users that were actively participating in traditional harvest activities and that had experience and knowledge of the traditional areas around their community received a camera-equipped GPS. Using their camera, they collected photos and spatial coordinates of any environmental condition that affected their travel and access to a given area on the land, referred to herein as 'hazard' data (supplemental 2). Land-users then filled out a form about each photo that explained their interpretation of what was pictured, how the conditions pictured influenced their travel or access, how frequent this environmental condition was observed in other places around their community, and to what extent the condition affected their travel safety. Eighteen individuals recorded environmental conditions affecting their land use for 17 months from March 2016 to July 2017.

Land-users' hazard data that were missing GPS coordinates or photos (n = 34) were removed , resulting in a total of 479 hazard locations for subsequent analyses. To determine if permafrost thaw directly generated hazards for land-users, an expert assessment was used to code the photos based on the likelihood that the hazard was related to a permafrost thaw event (supplemental 3).

To understand how permafrost-driven hazards affect land-user's activities, access to the land, and their safety we used the local participant's description of the effect of each observation on travel and access to resources. Quotations from the participant narratives were used to provide important context relevant to the hazards and the situations land-users faced. Descriptions were organized into three broad categories: impacts to equipment (e.g. damage to a motor), access (e.g. the blockage of a trail), and access to areas used historically (e.g. previous traditional harvest locations). For this data, only conditions from hazards that were 'highly likely' or 'likely' to be caused by permafrost were reported, as other conditions could be caused by other forms of environmental change.

To quantify the extent to which permafrost thaw directly created hazards for land-users we used descriptive and summary statistics for the number hazards 'highly likely' or 'likely' caused by permafrost thaw. Due to differences in the reporting frequency and engagement by communities (table S1), we reported on the proportion of permafrost hazards from a given community type (i.e. road connected or non-road connected) or region, as opposed to the raw number of all hazards observed.

To quantify the extent to which permafrost thaw may indirectly create hazards for land-users it was recognized that there is no one best permafrost vulnerability assessment data product. Therefore, we first assessed three geospatial permafrost vulnerability data products (Pastick et al 2015, Olefeldt et al 2016, Hjort et al 2018) to determine which one was most suited for this study. Hazard locations that were classified as either highly likely or likely to be caused by permafrost thaw (n = 144, considered to act as proxy ground truth points) were overlaid on each spatial product, and the proportion of these hazards that fell within each permafrost vulnerability category was summarized (figure 3). We found that the Olefeldt et al (2016) dataset best represented the 'ground truth' points and thus was used in all subsequent analyses. This dataset assesses predisposition to abrupt thaw. Many of the spatial proxies used to predict abrupt thaw can also be used as proxies to predict gradual thaw (for example ground ice content). Given this, we consider this dataset to provide a coarse-scale estimate of landscapes that are predisposed to permafrost-related change.

Using the Olefeldt et al (2016) dataset, we estimated the area of none, low, moderate, and high predisposition to permafrost-related change within each community's resource harvesting range. A spatial join was completed in ArcGIS (Version 10.6) to assign a predisposition to permafrost-related change to each hazard point. For this analysis, we assumed that all hazards located in high-thaw vulnerability areas are either directly or indirectly caused by permafrost thaw. We used the Neufeld et al (2019) dataset that provides a modeled traditional harvest—area for each community and assume that the availability of the use area is equal among land-users from that community (supplemental 4). A chi-square test was then used to determine if the proportion of hazards in a given thaw vulnerability class are greater than would be expected based on the areas of that vulnerability class within their harvesting range. For example, a significant chi-square statistic (α < 0.05) would indicate that when a land-user traverses a higher predisposition to permafrost-related change area, the chances of encountering a hazard are disproportionately higher compared with traversing a lower thaw probability area.

Over the one-year period (2016–2017), 18 land-users from nine partner communities collected GPS data associated with 479 hazard locations (figures 2 and 4), including a total of 441 with GPS locations and photos. Using the expert assessment of the photos and descriptions collected by the land-users to determine if permafrost thaw generated hazards for land-users, it was revealed that there were numerous instances (n = 142/441 hazards) and ways in which permafrost thaw created hazards (figure 4).

The written and oral descriptions documented by land-users provided insight into how permafrost-driven hazards affected land-user's activities, access to the land, and their safety. Land-users described how permafrost-driven hazards affected them in these different ways:

Impacts to equipment

'[It] adds debris to the water. This can ruin your day, or your whole fishing summer in fact ([boat] propeller, lower unit). Makes for tougher gathering and fishing. Fills your wheel or net with stuff. Have to pull net and wheel'—66.34° N, −147.59° W.

Impacts to access

'Debris in water makes boating more dangerous and makes it harder to dock the boat to access the area'—66.33° N, −147.59° W.

Impacts to historical areas

'Changes navigable channels. Need to find new fishing spots. Trees falling on you, rolling waves. Changes to historic fishing area for the first time'—66.33° N, −147.59° W.

Of the hazards that were 'highly likely' or 'likely' attributed to permafrost thaw (n = 157), land-users reported 61% had a strong or moderate effect on their safety. Land-users reported that of the hazards located within high or moderate thaw probability area (n = 256), 68% had a strong or moderate effect on their safety. One common hazard that was reported and had a strong effect on land-user safety was the impact of thaw and subsidence on ATV trails. The conversion of solid ground to wet, muddy areas makes it extremely difficult, time-consuming, and potentially dangerous to traverse as ATVs can tip over. These impact travel routes as it does not support travel using any form of motorized vehicle (e.g. truck, quad/ATV). When this occurred, land-users had to break new trails to find alternate routes, seek out new harvesting areas or change the timing of access to the resource (i.e. wait for drier or colder conditions). These hazards increased the amount of time a land-user was out on the land and reduced harvesting efficiency. Land-users described how permafrost thaw generated these conditions and the impacts it might have on their safety:

'This causes great difficulty as new trails must be cut around washouts or dangerous riding occurs'—63.84° N, −145.21° W.

'Trail is very wet and muddy—got stuck several times, other years it has been dry in September'—64.12° N, −141.89° W.

'Need to use ATV instead of vehicles, takes longer when you are harvesting'—67.03° N, −146.48° W.

The extent to which permafrost thaw directly created hazards for land-users throughout the entire region (i.e. the frequency of 'highly likely' or 'likely' caused by permafrost thaw) was 33% (n = 153) across the entire region (figure 5(a)). This frequency varied greatly between road-connected and non-road-connected communities. In general, road-connected communities had a lower frequency of 'highly likely' or 'likely' compared to rural communities.

The extent to which permafrost thaw both directly and indirectly created hazards for land-users throughout the entire region (i.e. the proportion of hazards located in high thaw vulnerability areas) was 52% (figure 5(b)). For context, high thaw vulnerability areas covered only 21% of the study region and 27% of the combined harvesting area of each community (table S2, figure 4). In nearly all the remote communities, there were more hazards located in higher thaw vulnerability areas than would be expected based on the area of high thaw vulnerable permafrost in a community's harvesting range (chi-squared test, table S2). The proportion of high thaw probability within communities harvesting areas ranged from 2% to 61% (table S2). Road-connected communities had lower proportions of their harvesting ranges composed of high thaw probability areas (mean ± SD = 3.6 ± 1.5%) compared to remote communities (mean ± SD = 45.5 ± 12.8%) (figure 6). This phenomenon was particularly evident in Beaver where all the hazards occurred in high thaw probability areas despite high thaw probability areas only occupying 51% ofphenomenon was more muted in road-connected communities with a higher proportion of hazards occurring in low probability areas compared to no-thaw probability areas (chi-squared test, table S2), but with relatively few hazards occurring in the limited moderate and high thaw probability areas.