Disturbances in North American boreal forest and Arctic tundra: impacts, interactions, and responses

Ecosystems in the North American Arctic-Boreal Zone (ABZ) experience a diverse set of disturbances associated with wildfire, permafrost dynamics, geomorphic processes, insect outbreaks and pathogens, extreme weather events, and human activity. Climate warming in the ABZ is occurring at over twice the rate of the global average, and as a result the extent, frequency, and severity of these disturbances are increasing rapidly. Disturbances in the ABZ span a wide gradient of spatiotemporal scales and have varying impacts on ecosystem properties and function. However, many ABZ disturbances are relatively understudied and have different sensitivities to climate and trajectories of recovery, resulting in considerable uncertainty in the impacts of climate warming and human land use on ABZ vegetation dynamics and in the interactions between disturbance types. Here we review the current knowledge of ABZ disturbances and their precursors, ecosystem impacts, temporal frequencies, spatial extents, and severity. We also summarize current knowledge of interactions and feedbacks among ABZ disturbances and characterize typical trajectories of vegetation loss and recovery in response to ecosystem disturbance using satellite time-series. We conclude with a summary of critical data and knowledge gaps and identify priorities for future study.


Introduction
In the North American Arctic-Boreal Zone (ABZ), climate change and human activity are rapidly and extensively reshaping vegetation dynamics via a range of disturbance processes, resulting in considerable uncertainty in the fate of these ecosystems (Shaw et al 2021). Many disturbances (i.e. an event that alters ecosystem composition, structure, function, or the physical environment, Pickett and White 1985) trigger a transient reduction and gradual recovery of vegetation cover and ecosystem function (Liu et al 2011, Li et al 2021, although there is high variability in the nature and pace of these changes depending on the type and severity of disturbance (Jorgenson et al 2015, Gaglioti et al 2021 (figure 1). Climate warming is occurring in the ABZ at more than twice the global average rate (Price et al 2013, Smith et al 2019, Chylek et al 2022, Rantanen et al 2022, and many disturbance processes are highly sensitive to climate. Consequently, the impact of climate change via disturbance on ABZ vegetation dynamics is expected to increase over the next century (Price et al 2013, Gauthier et al 2015, Bush and Lemmen 2019, Smith et al 2019.
Disturbance-driven loss and subsequent recovery of vegetation partly explain widespread trends in satellite-observed vegetation indices (i.e. 'greening' and 'browning') within the North American ABZ (Ju and Masek 2016, Sulla-Menashe et al 2018, Wang and Friedl 2019. Large-scale greening trends across the ABZ are complex (Myers-Smith et al 2020), but have generally been interpreted as an increase in ecosystem productivity driven by climatic warming and recovery from disturbance (Bhatt et al 2010, Berner et al 2020. Meanwhile, areas of browning are generally attributed to vegetation stress from disturbances such as fires, insect outbreaks, warminginduced drought, and increased surface water associated with permafrost degradation (Goetz et al 2005, Shur and Jorgenson 2007, Verbyla 2011, Berner and Goetz 2022. Many of these disturbances are increasing in their extent, frequency, and/or severity because of climatic changes and increasing anthropogenic pressures (Jorgenson et al 2006, Baltzer et al 2021. Understanding the net impact of climate change and its effects on different disturbance regimes is critical for forecasting future ABZ composition, dynamics, ecosystem services, and potential management responses.
As in many other ecosystems, fires have dramatic and extensive impacts on vegetation cover and carbon dynamics in the ABZ, and exceptional warming in this region is intensifying fire regimes (Soja et al 2007, Kasischke et al 2010, Veraverbeke et al 2017, McCarty et al 2021, Whitman et al 2022. However, the unique characteristics of ABZ ecosystems result in additional types of disturbances that lack analogs in tropical and temperate ecosystems. The wide extent of permafrost (i.e. perennially frozen ground; Gruber 2012) that underlies large parts of the northern high-latitudes makes these ecosystems vulnerable to a unique set of other disturbances (Shur and Jorgenson 2007). For example, thawing permafrost causes ground surface subsidence that can induce persistent changes in hydrology, vegetation, and microtopography in ABZ landscapes with high groundice contents (Grosse et al 2011, Jones et al 2015, Carpino et al 2018, Farquharson et al 2019, Swanson 2021. Exceptional warming in the ABZ also makes high-latitude forests vulnerable to increasing incidences of drought and insect outbreaks (Volney and Fleming 2000, Hogg et al 2008, Kurz et al 2008. Natural resource development activities such as oil and gas well exploration and production and logging introduce additional complexity to disturbance regimes (Gauthier et al 2015, Shaw et al 2021 in various parts of the region (Williams et al 2013, Pasher et al 2013, Raynolds et al 2014.
Fire is a key driver of the carbon balance of boreal ecosystems (Harden et al 2000, Bond-Lamberty et al 2007, Wang et al 2021, but the relative importance and impacts of other disturbance types have been less studied (Shaw et al 2021). Thus, it remains unclear how much these other disturbance types and their interactions (Buma 2015) impact ABZ ecosystems. In this review, we summarize the existing state of knowledge of major disturbance types in North American ABZ ecosystems and use case studies of Landsat satellite-derived time series of vegetation greenness and moisture indices to illustrate the distinct spatiotemporal characteristics of vegetation loss and recovery associated with each disturbance type. Additionally, we review interactions between disturbances, which are likely to intensify in the future (Buma 2015, Seidl et al 2017. In this review, we focus on 'pulse' disturbances, characterized as generally abrupt, relatively discrete events that rapidly alter ecosystem structure, resources, or the physical environment (Pickett and White 1985). We do not address 'press' disturbances which impact ecosystems slowly over decades and centuries (e.g. long-term warming; Grosse et al 2011). We divide major ABZ disturbances into several categories: (a) fire; (b) insects and pathogens; (c) permafrost-related disturbances; (d) anthropogenic disturbances; (e) weather-related disturbances; (f) riverine processes; and (g) ungulate and grazer activity. These disturbance types are not meant to be an exhaustive list of all known disturbances within the North American ABZ, but rather a characterization and discussion of the major climatesensitive and anthropogenic disturbances within the region that impact vegetation processes. We do not, for example, include coastal erosion, alpine landscapes (e.g. avalanches), or localized geologic settings (e.g. volcanism). By considering a range of major disturbance types, we seek to answer a set of interrelated questions: What are the distinct causes of each disturbance type, and how are disturbance regimes (i.e. extent, frequency, and severity) sensitive to climate change and human activity? How does each disturbance type impact vegetation composition, structure, and recovery? How do different disturbance regimes interact with each other? In doing so, we aim to provide context, identify data and knowledge gaps, and lay the groundwork for future studies that analyze how the full suite of disturbance agents are reshaping the vegetation dynamics of ABZ ecosystems.

Methods
This paper discusses the background, outstanding science questions, and data relevant to each of the seven broad disturbance categories. We also introduce case studies showcasing typical vegetation loss and recovery in response to select disturbances evident from remote sensing data.

Literature survey
Articles referenced in the background (section 3), spatiotemporal characteristics (section 4), and interactions (section 5) sections were selected based on a thematic literature review as well as our own bibliographic lists derived from our active research in these fields. We searched the peer-reviewed literature using terms related to each disturbance category and type and biome (e.g. 'boreal forest windthrow' , 'cryoturbation' , 'ice-jam flooding'). We emphasized recent (since 2014) papers and studies published on the North American boreal and Arctic ecosystems; however, we included studies from Eurasia to supplement topics where North American studies are lacking and to expand the global relevancy of this review.

Case studies and datasets
To evaluate patterns of vegetation loss and recovery after different disturbance types we compiled a set of locations (n = 397) of known disturbances within the North American ABZ to serve as case studies (figure 2). We compiled locations of known disturbance occurrences based on expert knowledge and field work of the authors as well as published locations in the literature and existing disturbance databases (table S1). For each case study, we analyzed vegetation greenness and moisture changes during and following disturbance using time series of surface reflectance data from the Landsat series of satellites Wulder et al 2019).
For case study locations derived from individual latitude and longitude points, we extracted Landsat time series within a 100 m buffer surrounding each site to mitigate issues with geospatial accuracy of the case study locations. For case study locations derived from polygons, polygons were first filtered to only include 'severe' impacts (if known), as well as disturbances that occurred between 2001 and 2016 to ensure adequate temporal coverage of pre-and post-disturbance vegetation greenness and wetness. The selected polygons were then randomly sampled (n = 25 per disturbance type), and 30 m Landsat pixels were randomly selected within each sampled polygon (n = 50 per polygon). For fire disturbance, in order to ensure broad coverage of diverse ecological conditions present within the North American ABZ, ten random points were sampled within each of five random fire polygons per Level II Ecoregion (US EPA 2015).

Case study analysis
We calculated spectral indices representing land surface greenness (the Normalized Difference Vegetation Index-NDVI; Rouse et al 1974, Tucker 1979 and wetness (the Normalized Difference Moisture Index-NDMI; Gao 1996). NDVI is a widely used index that is sensitive to leaf chlorophyll content and is generally correlated with vegetative cover and photosynthetic productivity. However, NDVI is less sensitive to changes in the state of evergreen forests (Jin et al 2017), which are the dominant forest type in the ABZ (Gauthier et al 2015). NDMI is an index that is sensitive to leaf water content and may reflect more subtle changes in vegetative stress in evergreen trees (Goulden and Bales 2019). While more specific and fine-scale indices may lend more information about, for example, species composition changes following disturbance, the use of NDVI and NDMI allows for broad coverage of the impact of different disturbances on vegetative cover and condition. Changes in NDVI and NDMI thus are interpreted as vegetation loss (e.g. declining NDVI or NDMI) and recovery (e.g. increasing NDVI or NDMI) in response to disturbance.
We developed time series of annual summer maximum greenness and wetness for the case study sites (table S1). For each sampled location, we extracted all available Landsat 5, 7, and 8 surface reflectance data acquired each summer (day-of-year 151-242; 31 May-31 August) from 1985 to 2020 for a total of ∼11 000 000 multi-band measurements tallied across all pixels. These data were retrieved from the Landsat Collection 2 surface reflectance dataset (Masek et al 2006, USGS 2021, accessed using Google Earth Engine (Gorelick et al 2017) and functions provided by the lsatTS package (Berner et al 2021, accepted) in R (R Core Team 2021). We quality-screened these surface reflectance measurements based on pixel-and scene-criteria (i.e. scene-wide cloud cover <80%, geometric uncertainty <30 m, and solar zenith angle <60 • ) and further cross-calibrated them among Landsat sensors using the lsatTS package. Crosssensor calibration is necessary to avoid spurious trends in NDVI and other spectral indices that arise from systematic differences in spectral bands among Landsat sensors (Sulla-Menashe et al 2016, Berner et al 2021. We calculated annual summer maximum surface greenness (NDVI) and wetness (NDMI) at each sampled location as the maximum summer NDVI or NDMI. Overall, we developed 14 709 annual time series of surface greenness and wetness for recently disturbed pixels across the study domain (table S1).
Because some case study locations were approximate or derived from large aerial survey polygons, not all pixels were located over an actual disturbed area. Therefore, to focus our analyses on pixels that captured disturbance events, we filtered pixels to those that included detectable disturbance impacts on NDVI and NDMI within five years of the known disturbance event, except for cryoturbation and icewedge degradation, which occur within landscape mosaics and do not correspond to a single 'event' . Aside from cryoturbation and ice-wedge degradation, disturbances were identified using visual interpretation of each time series and via the Breaks For Additive Season and Trend (BFAST) algorithm in the bfast package (Verbesselt and Herold 2012) in R (figure S1). BFAST iteratively estimates abrupt changes (or 'breakpoints') within time series and can be used to analyze seasonal and annual time series of satellite-observed reflectance to detect statistically significant temporal changes (Verbesselt et al 2010, Verbesselt andHerold 2012).
Following breakpoint detection, each time series with detected breakpoints was smoothed using the R function smooth (Tukey 1977), and inflection points were identified in the smoothed time series. The series was first smoothed to identify 'true' changes in the vegetation index trajectory, rather than those simply due to noise or interannual variability. The inflection point with the minimum (or maximum, for NDVI of lake drainage) spectral index value was identified as the year of full effect from the disturbance on land surface greenness and wetness. The time series before the breakpoint and following any breakpoints detected earlier in the series (e.g. between 1994 and 2014 in figure S1) was used to calculate an average pre-disturbance mean NDVI and NDMI. Each time series was then normalized by its predisturbance mean (NDVI norm = NDVI/NDVI mean , figure 3). We normalized the time series to better compare within and between disturbances, which occurred in different biomes and bioclimatic regions.
These normalized time series were used as our case study trajectories to evaluate the impact of each disturbance on vegetation as well as the magnitude, direction, and speed of recovery following each disturbance (see section 3).

Disturbance characteristics and interactions
The major ABZ disturbance types have distinct spatial, temporal, and severity characteristics. To compare the spatial and temporal dynamics among disturbances, we developed several spatiotemporal metrics. Spatial grain describes the average extent of an individual disturbance event (e.g. for a wildfire it would be the size of a polygon associated with the outer perimeter of the burn scar, but for insect infestation it might be a single tree or forest stand). Return interval refers to the average length of time for the disturbance to reoccur in the same location. Occurrence timeline describes the average length of time a disturbance event lasts from initiation to completion (e.g. for wildfire: from ignition to extinction). Recovery timeline refers to the average length of time it takes for the vegetation/ecosystem to return to pre-disturbance conditions. Finally, intensity/impact refers to the average effect on vegetation and the ecosystem, from vegetation stress to complete vegetation mortality. We determined qualitative values for each of these categories and disturbance types using scientific literature and expert knowledge (see section 4). The metrics were converted into relative numerical scales (table S2) and applied to a principal component analysis (PCA) to understand how the different metrics correlate with one another across the different disturbance types. The PCA was conducted using the R function prcomp, with the categorical metrics scaled and centered within the PCA.
The degree to which different disturbance types interact with each other is complex and not well understood, and critical feedbacks between disturbances make their potential impacts difficult to analyze and predict. Therefore, we developed a disturbance interaction matrix based on our literature survey and expert knowledge. This matrix describes the impact (strong/weak positive, strong/weak negative, both, none, or unknown) of a 'driver' disturbance on potential subsequent 'response' disturbances (see section 5). We distinguish 'strong' and 'weak' interactions by their relative effect on ecosystem structure and function, the ubiquity and likelihood of this impact occurring, and the ability of the ecosystem to resist or recover from subsequent response disturbances. For example, we classify the impact of boreal windthrow on subsequent insect and pathogen disturbance as 'strong positive' (figure 20), because this interaction is a well-documented and impactful phenomenon within forested ecosystems (e.g. Malmstrom and Raffa 2000). In contrast, we classify the impact of logging on subsequent windthrow events as 'weak positive' (figure 20), because while forest fragmentation, such as that created by forest harvest, does impart higher susceptibility to windthrow (Meilby et al 2001, Peterson 2004, the low probability of windthrow in boreal North America (Bouchard et al 2009) reduces the overall impact of this interaction. See section 5 for a further discussion of these interactions.

Disturbance agents in North American
Arctic and boreal ecosystems 3.1. Fire

Background
Wildfire is the most well-studied disturbance agent in forests of boreal North America, as fires have substantial impacts on human settlements (Kent 2017), subsistence resources (Nelson et al 2008), and air quality (Trainor et al 2009), in addition to climate (Randerson et al 2006, Potter et al 2020b and vegetation (Rogers et al 2013, Foster et al 2022. Fires in boreal North America are generally high-intensity crown fires that kill most affected trees and consume substantial belowground carbon stocks, in contrast to those in boreal Eurasia or more temperate ecosystems which include a high fraction of lower-severity surface fires that result in relatively low tree mortality (Stocks and Kaufmann 1997, de Groot et al 2013, Rogers et al 2015. Fire is less common in Arctic tundra but has been increasing in frequency and severity (Hu et al 2015, McCarty et al 2021, especially in the Beringian region (Racine et al 1985, Rocha et al 2012, Masrur et al 2018, Gaglioti et al 2021. Recent increases in boreal and Arctic wildfire activity may indicate fundamental shifts in the causes and impacts of the underlying fire regime, including overwintering fires that smolder during winter months and reappear the following year (Scholten et al 2021, Xu et al 2022 Precursors to fire in boreal ecosystems are well understood-an adequate amount of fuel and fuel dryness are required for fires to ignite and spread, in addition to ignition sources such as lightning strikes and anthropogenic activities (Veraverbeke et al 2017, Archibald et al 2018, Rogers et al 2020. In the boreal zone, fires are generally limited by fuel dryness and ignition sources because the characteristically deep organic and moss layers provide ample fuel. Both species composition and litter moisture are influenced by site drainage conditions, with organic-rich soils dominated by fire-prone and flammable species such as black spruce (Picea mariana). Conversely, Jack pine (Pinus banksiana) and less flammable deciduous species typically occur in well-drained locations with thinner, drier soils (Walker et al 2018(Walker et al , 2020b. Lightning strikes ignite most fires in the North American ABZ. Lightning ignitions have increased since the mid-20th century due to a warmer and more convective atmosphere (Veraverbeke et al 2017, Chen et al 2021c. More severe fire weather is also prolonging fire seasons and increasing fire intensity and annual area burned. For example, Kasischke et al (2010) found the mean annual area burned in Alaska during the 2000s was 50% greater than any previous decade since the start of the record in 1940, resulting in increased ground-layer combustion and net carbon emissions to the atmosphere (Turetsky et al 2011).
Within the North American boreal region, fires create lasting legacies on vegetation, driving changes in soil characteristics, regeneration patterns, and successional trajectories (Johnstone et al 2010, Gaglioti et al 2021, Mack et al 2021. High-severity forest fires that remove much of the organic layer favor regeneration by deciduous and fast-growing pine species, which may maintain dominance under a warming climate (Johnstone et al 2011). Field data have also suggested that increased warming and fires may be altering the ability of typically resilient black spruce forests to recover following large fires, leading potentially to a tipping point for boreal vegetation-shifting from evergreen to deciduous or non-forested land cover types (Baltzer et al 2021). Alterations to phenological metrics from time series of NDVI and other greenness metrics observed in burned areas in Alaska may also indicate long-term shifts in vegetation cover type and photosynthetic activity at regional scales (Potter 2020a, Madani et al 2021.
In the Arctic tundra, our understanding of the drivers of the wildfire regimes is less thorough, due to a combination of factors including lower fire frequency, remoteness, and limited in-situ observations. It is commonly believed that lightning (Chen et al 2021c, He et al 2022), summer temperature, and precipitation (Hu et al 2015, Vachula et al 2022 are among the primary factors controlling the wildfire regimes in Alaskan tundra. Fire usually favors the recruitment and growth of deciduous shrubs in the tundra. It is therefore an important mechanism for Arctic shrubification (Lantz et al 2010b, Jones et al 2013, Frost et al 2020. Following fire, net ecosystem productivity (NEP) declines because of reduced vegetation productivity and increased ecosystem respiration, with forest ecosystems becoming a carbon source for roughly one to two decades (Amiro et al 2010, Liu et al 2011, Kurz et al 2013). In the tundra, vegetation productivity recovers more quickly, in as little as three years post-fire, though longer term impacts on NEP remain less clear (Gaglioti et al 2021). As vegetation and soils recover, NEP increases up to a maximum and then decreases to a steady state, at which point the ecosystem is again carbon neutral or a carbon sink , Song et al 2018. Climate change, however, may alter the postfire NEP response in the future due to species composition shifts, productivity changes, and permafrost thaw (Rocha et al 2012, Gibson et al 2018, Foster et al 2019, Mekonnen et al 2019, Baltzer et al 2021. Vegetation responses to fire disturbance can be seen in Landsat-derived trajectories of greenness (NDVI) and wetness (NDMI), as showcased in the average across all fire trajectories (n = 32) as well six individual fires (figure 4). The average trajectory shows a rapid decline in normalized NDVI and NDMI immediately following fire, with a moderate recovery rate in the following years (approximately 10 years for NDVI and 15 years for NDMI). Tundra NDVI recovers more rapidly, with NDVI values reaching the pre-disturbance mean within a decade following fire. The NDMI response following fire is more varied for the tundra locations, a pattern which highlights the cascading effects of wildfire on accelerated permafrost thaw and associated changes in soil thermal and moisture regimes, and variability arising from local differences in fire severity and ground ice conditions (Jones et al 2015).

Limitations, data needs, and unknowns
Large fire databases are crucial for understanding fire precursors, effects, trends, and dynamics in boreal and Arctic ecosystems. In Alaska and Canada, existing fire history databases provide fire perimeter polygons beginning in the 1940s and 1960s, respectively, and are maintained and updated annually. These databases are some of the longest and most complete large-scale historical fire records available anywhere on the planet (Kasischke et al 2002, Stocks et al 2002 and they are foundational datasets for investigating regional impacts of post-fire vegetation succession (Rogers et al 2013, Potter et al 2020b. Despite this, due to the great challenges in mapping wildfires in the high latitudes (e.g. limited availability of Landsat observations during a short growing season and persistent cloud cover; Chen et al 2021a, 2021b), omissions of large wildfire events by these wildfire history records still exist, particularly in the tundra (Jones et al 2013). Moreover, the fire perimeters themselves become less accurate further back in time, and often contain substantial patches of unburned vegetation (Kasischke et al 2002, Walker et al 2018, Potter et al 2020b. Advances in remote sensing tools enable fires and their impacts to be mapped and tracked at increasingly finer spatiotemporal resolutions (  . However, additional combustion estimates are needed to better understand the interactions between fire weather, fire spread and intensity, and combustion (Walker et al 2020b).
Further data are required to elucidate the interactions between wildfire, vegetation, and permafrost in the context of changing climate (Gibson et al 2018, Treharne et al 2022. Increasing temperatures, changing precipitation, and increases in fire activity will impact vegetation composition and structure, hydrology, and carbon fluxes. Future researchers could utilize a combination of active radar and subsidence data, high spatial and spectral resolution imagery, digital elevation models, and airborne LiDAR and other remote sensing data to observe and analyze these changes. It is also unclear how these changes to vegetation and fuels will interact with future fire regimes. Predicted increases in deciduous fraction and declines in organic layer and other fuels (Foster et al 2019, Mekonnen et al 2019) may lead to decreasing fire frequency and severity, even as fire weather and fuel drying increases (Parks et al 2015). Further, if young stands re-burn following fire, it is unknown how and which species may be able to regenerate as seed banks become depleted and soils become less conducive to seedling establishment (Baltzer et al 2021).
From a societal perspective, the increasing frequency of large fires, and necessary increased investments in fire-fighting activities at the wildlandurban interface, will strain the existing fire management budgets and governance structures (Rogers et al 2020). More studies are needed linking the influence of management on fire regimes, both historically and in the future, to quantify these relationships and make predictions for the efficacy and costs of fire management efforts (

Background
Biotic disturbances, such as fungal pathogens (e.g. root rots and needle rusts) and insect outbreaks (e.g. bark beetles and defoliators/leaf miners) can cause extensive tree mortality during outbreaks (Holsten et al 2008. Fungal pathogens often kill individuals slowly by disrupting water and nutrient transport (Holsten et al 1985) and reducing growth. In contrast, episodic insect outbreaks can cause major growth reductions and spatially widespread tree mortality over a few years, at times eclipsing that due to fire. For example, annual forest volume lost due to productivity reduction and mortality from pests and pathogens in Canada was estimated to be 106 million m 3 yr −1 between 1982 and 1987, which was three times that lost annually to fire and 70% of volume harvested in Canada nationally during that period (Hall and Moody 1994, Malmstrom and Raffa 2000, Volney and Fleming 2000, Price et al 2013. In the 1990s in Alaska, insects cumulatively damaged 1.6-2 million hectares of forest, which was 30% more area than burned during that period (Malmstrom and Raffa 2000).
Bark beetles, such as the mountain pine beetle (Dendroctonus ponderosae) and spruce beetle (Dendroctonus rufipennis), kill host trees outright by feeding on the cambium and phloem (figures 5(a) and (c)) and disrupting water transport (Malmstrom and Raffa 2000, Bentz et al 2010. These beetles attack trees through 'mass attacks' of many beetles, attracted via massing pheromones released by the beetles (Raffa et al 2008). Bark beetle populations typically exist at relatively low levels, punctuated by occurrences of high, epidemic levels due to climate-, disturbance-, or forest structure-related triggers (DeRose et al 2013, Seidl et al 2016). Young, healthy trees can often defend against low levels of attacking beetles by exuding resin and allelochemicals. However, stressed trees and those experiencing a large number of attacking beetles are more likely to succumb to infestation (DeRose and Long 2012). Thus, conditions that lead to vegetation stress, such as drought, often lead to outbreak events (Sherriff et al 2011, Seidl et al 2016. Defoliators and leaf miners feed on the leaves and needles of host plants. In the North American ABZ, these guilds include, for instance, eastern and western spruce budworms (Choristoneura spp.), Jack pine budworm (Choristoneura pinus), aspen leaf miner (Phyllocnistis populliella) (figure 6(a)), and large aspen tortrix (Choristoneura conflictana). Outbreaks of these defoliators and miners cause significant tree growth reduction and potentially tree mortality. Removal or damage to needles and leaves disrupts water transport and interferes with photosynthesis, which can kill trees directly or cause physiological stress that predisposes them to death from other factors, such as drought (Malmstrom and Raffa 2000). Recovery from major defoliation and mining depends on the extent of damage and the amount of carbon reserves held in other tissues (Boyd et al 2021). Deciduous species generally are more able to re-foliate from leaf damage than evergreen species, even in the same year as defoliation (Krause andRaffa 1996, Holsten et al 2008). Evergreen species, however, often have a high rate of mortality following successive years of intense defoliation, potentially leading to species composition shifts post-outbreak.
The most common pathogens in the North American ABZ include root rot (e.g. Inonotus tomentosus), heart rot fungi (e.g. Fomitopsis pinicola), and needle rusts (e.g. Chrysomyxa ledicola; figure 6(c)) (Armstrong and Ives 1995, Holsten et al 2008. These pathogens can cause hydraulic impairment by damaging vascular systems, reduce productivity through impacts on needles and leaves, and ultimately lead to plant mortality. Recently, an outbreak of the novel aspen running canker (Neodothiopora populina) caused widespread mortality of quaking aspen (Populus tremuloides) in interior Alaska (figure 6(b)). Aspen mortality from these infections was exacerbated by ongoing drought as well as an outbreak of aspen leaf miner (Ruess et al 2021).
While pathogens frequently affect a wide range of species, insects are often species-or genus-specific in their host requirements (Armstrong andIves 1995, Holsten et al 2008). Hosts that are larger, older, or stressed are generally more susceptible to bark beetles. Thus, areas with high numbers of susceptible hosts are most vulnerable to insect outbreak, with mature, host-dominated stands being the most susceptible (Raffa et al 2008, Chapman et al 2012, DeRose et al 2013, Hart et al 2015. These homogenous stands provide a high quality habitat for insects, allowing for self-sustaining populations and sources of large-scale outbreaks (Malmstrom and Raffa 2000, Seidl et al 2016. The relatively low biodiversity in ABZ forests thus makes them particularly vulnerable to insect and pathogen outbreaks (Jactel et al 2005, Campbell et al 2008, Senf et al 2017a. Increasing temperatures and drought are thus generally expected to increase the impacts of insects and pathogens in the North American ABZ. Insect and pathogen outbreak dynamics are affected and compounded by climate and weather by influencing the range and population size of insects and pathogens and altering the vulnerability of plants.  warmer conditions can accelerate larval growth, causing a shift to a one-year (univoltine) life cycle (Hansen et al 2011). More beetles with univoltine life cycles drives faster population growth and more severe outbreaks, such as occurred with the expansion of bark beetle outbreaks in British Columbia in the 1970s and 1980s (Bentz et al 2010). Host plants also interact with climate through host stress levels. Drought predisposes trees to disease and infestation (Raffa et al 2008, McKenzie et al 2009, Boyd et al 2021, Ruess et al 2021, and can be a secondary cause of mortality following defoliation stress (Malmstrom and Raffa 2000). Climate change is predicted to result in range expansion of insect species (de la Giroday et al 2012) and increases in outbreak severity and frequency (Raffa et al 2008). In Alaska, drought, high vapor pressure deficit, and high temperatures are key contributors to mortality linked with aspen leaf miner and aspen canker (Boyd et al 2021, Ruess et al 2021).
Because bark beetles tend to affect one or only a few host tree species and preferentially attack larger trees, their outbreaks often result in a shift towards smaller size classes and non-host species (Veblen et al 1991, Campbell and Antos 2015, Zeppenfeld et al 2015. Productivity often increases in these subsequent stands as non-infested trees are released from suppression (Campbell et al 2019). In more homogenous stands, species composition can shift towards early successional species after an outbreak. These impacts can be seen in trajectories of NDVI and NDMI before and during outbreaks (figure 7). Defoliators also tend to impact one or a few species-the eastern spruce budworm (C. fumiferana) mostly infests balsam fir (Abies balsamea) and white spruce (Picea glauca), and infestation-caused mortality often leads to release of seedlings and saplings of host species (Boulanger and Arsenault 2004). Changes in NDVI are generally subtle as outbreaks build, sometimes asynchronously, within individual trees (figure 5(b)), and are usually only visible in moderate-resolution satellites when large areas are impacted (DeRose and Long 2012). This subtle NDVI pattern (figure 7) is especially characteristic of spruce beetle outbreaks, which do not exhibit a characteristic 'red-stage' attack as do pine species infested with mountain pine beetle (Coops et al 2006). However, NDMI often does decline (figure 7), due to decreases in transpiration and increases in foliar water stress during and following bark beetle outbreaks (Foster et al 2017). In contrast, trajectories of NDVI and NDMI during and following spruce budworm infestation in the Northwest Territories (figure 8) have a clearer signal, with some variability across the individual sites, highlighting the impact of infestation severity on the spectral signal. Sites which have a lower infestation severity (e.g. percent defoliation) will have a more subtle signal than sites which had more complete defoliation from spruce budworm infestation (Senf et al 2016). The response of NDVI to aspen running canker is also clear, with limited recovery following the drop in NDVI due to infestation (figure 8). NDMI response is less clear, with some decline following infestation.

Limitations, data needs, and unknowns
Past insect outbreaks are often identified through dendrochronology and pollen records (Anderson et al 2010, Sherriff et al 2011). However, these are limited to specific locations, usually where an outbreak is known, resulting in biases in our understanding of their extent and occurrence. Aerial detection surveys that produce polygons of infestation extent and severity are valuable for determining the regional and national impacts of forest pests. However, these polygons are often at a coarse spatial scale with potentially low positional accuracy (Wulder et al 2006. Detection of recent or ongoing outbreaks using moderate resolution satellite sensors is possible, especially for large, severe outbreaks (Meddens and Hicke 2014, Hall et al 2016, Senf et al 2016, 2017a. Specialized methods are generally required for each insect type (e.g. bark beetles vs. defoliators). Foliar color changes of conifers infested with bark beetles often progress from green, sometimes to red, and to gray as needles lose moisture and are ultimately shed from the tree. The red and gray stages are easily detectable in multispectral imagery (Coops et al 2006), however the green stage is more subtle, making early detection difficult (DeRose et al 2011). Despite this difficulty, some studies have had success in using the water-sensitive shortwave infrared wavelengths to detect early moisture stress from green-stage infestations (Foster et al 2017).
Accurate and temporally and spatially consistent datasets of infestation/infection status and extent across jurisdictions are crucial for determining the extent and severity of past and ongoing outbreaks, and for predicting future outbreaks , Senf et al 2017b. Such large-scale datasets would also aid in generalizing detection methods across wider regions and disturbance agents. Because some major limitations to detecting insect and pathogen disturbance from remotely sensed data include accurately discriminating between these disturbances and other vegetation stressors, due to the exhibition of similar spectral signals (Senf et al 2017b), field observations of infestation status that are coincident with remote sensing observations will assist in developing more accurate algorithms for multi-stage detection efforts (Cessna et al 2021). Increased availability of different types of remote sensing data, particularly hyperspectral and radar imagery, have the potential to identify changes in forest moisture related to insect and pathogen outbreaks at regional scales and with high spatial detail.
Studies have shown that insects and pathogens are expanding their ranges poleward with increasing temperatures, increasing the area of forest vulnerable to outbreak (de la Giroday et al 2012, Pureswaran et al 2018). Insects are also beginning to infest novel host species (NRC 2018), and it is unclear how such host species will respond. Such range expansion highlights the need for increased detection and monitoring of outbreaks, as well as the need for predictions of future infestation vulnerability.

Permafrost-related disturbances
Throughout much of the northern high-latitudes, ecosystems are underlain by permafrost, or soil that remains frozen for more than two years (Gruber 2012). However, with climate change, permafrost ground temperatures are increasing (Biskaborn et al 2019) and the active layer-the upper layer of soil that thaws in the summer-is becoming deeper across large areas (Smith et al 2022). In addition to the active layer, the physical structure of these soils is being altered across many landscapes in the ABZ due to extensive changes to permafrost status due to warming, and permafrost thaw is expected to increase further in the future, both linearly and abruptly . These changes in physical structure can dramatically alter the topography, hydrology, and vegetation, resulting in heterogeneous topography and thermokarst features, especially in ice-rich locations. In this section, we describe several unique disturbances in the ABZ and their associated permafrost-related processes, including cryoturbation, ice-wedge degradation, cryogenic landslides, and lake drainage.

Cryoturbation
Permafrost soils often exhibit warped or broken soil horizons that result from cryoturbation, the frost-based movements of seasonally frozen materials (Bockheim and Tarnocai 1998). Cryoturbation can also create distinctive surficial disturbance features that generate fine-scale spatial heterogeneity in ground conditions and serve as foci for ecological change (Walker et al 2011, Frost et al 2013, Aalto et al 2017. Frost circles are a common form of patterned ground. They occur as approximately circular patches (∼0.5-3 m diameter) of mineral soil that often form geometric mosaics of vegetated and unvegetated microsites at uniform spacing of ∼1-3 m (figure 1(c)).
Frost circles are common in permafrost regions, particularly where surface organic material is lacking and the soil profile is dominated by fine-textured silt or clay (Bockheim et al 1998, Peterson andKrantz 2003). The intensity of cryoturbation is strongly affected by soil moisture, soil texture, changes in seasonal temperature, and snow cover (Daanen et al 2007, Aalto et al 2017. In general, climate warming and increased snow cover dampen cryoturbation by reducing differential frost-heave. Climate warming can also dampen cryoturbation indirectly by promoting vegetation colonization, which stabilizes the soil and results in organic matter accumulation on cryoturbated soils. Species that are fast-growing and/or tolerant of ground surface disturbances are best able to colonize cryoturbated surfaces (Kade et al 2005, Sutton et al 2006, Frost et al 2013. Once cryoturbation is reduced or no longer occurring, the increase in biomass is abrupt and persistent; however, cryoturbation can be renewed if vegetation and organic material are removed by other disturbances (chiefly wildfire).
Cryoturbation can have nonlinear responses to climate change with respect to vegetation cover and biomass, which can be detected in multi-decadal NDVI time series (Frost et al 2014). Furthermore, cryoturbation has distinctive spatiotemporal properties as a disturbance agent, because features usually occur as a multitude of 1-3 m microsites within a broad landscape mosaic, and the disturbance acts annually and is not episodic. At our case study locations, both NDVI and NDMI increased over the 30 year Landsat record (figure 9). With respect to NDVI, the warming climate could allow vegetation to colonize previously bare frost circles in cryoturbated landscapes, which would reduce further cryoturbation (Frost et al 2013). The NDVI increase could additionally reflect a general background greening (i.e. vegetation increase) of the landscape, as only a fraction is cryoturbated, and the remainder can have nearly complete vegetation cover. For NDMI, the increase in moisture is likely in part due to the moisture content of the colonizing vegetation, but also increased soil moisture beneath the vegetation cover (figure 9).

Ice-wedge degradation
Polygonal ground, encompassing mosaics of icewedge polygons (∼5-15 m wide) formed by contraction cracking followed by annual cycles of thawing and refreezing, is widespread and conspicuous in permafrost landscapes (Liljedahl et al 2016). Wedgeshaped masses of ice underlie the edges of each polygon (figure 10). Ice-wedge degradation occurs when the uppermost portions of ice wedges thaw, which triggers local ground subsidence, ponding, and persistent changes to vegetation and hydrologic connectivity across the landscape (figure 10). Ice-wedge degradation often results in substantial micro-topographic changes, such as the transition from low-centered to high-centered polygonal landforms.
Polygonal ground is most common in tundra with continuous permafrost, especially areas with fine-textured soils, and patterned landscapes can cover areas as large tens of square kilometers or larger (Lachenbruch 1962). However, ice wedges are also common in discontinuous permafrost regions well into the boreal forest (Kokelj et al 2014, Swanson 2016. Extreme warm and wet summers initiate ice-wedge degradation . Average trajectories for NDVI and NDMI (left, not normalized) and distributions of trends (right) for NDVI and NDMI at cryoturbation case study pixels (n = 2129, across 47 sites) in northern Alaska. Trends were calculated as the slope from linear models fit at each Landsat pixel for the vegetation index. Note that these trajectories were not normalized because cryoturbation is an ongoing disturbance, rather than a single event. wedges, so the terrain affected by ice-wedge degradation has historically supported 'climax' vegetation communities-usually tussock tundra or needleleaf Figure 11. Average trajectories for NDVI and NDMI (left, not normalized) and distributions of trends (right) for NDVI and NDMI at ice-wedge degradation case study pixels (n = 679, across 15 sites) in northern Alaska. Trends were calculated as the slope from linear models fit at each Landsat pixel for the vegetation index. Note that these trajectories were not normalized because ice-wedge degradation is an ongoing disturbance, rather than a single event.
woodlands in boreal forest settings (Billings and Peterson 1980). Local and regional variability in the timing and extent of ice-wedge degradation arises from differences in surficial materials and ground-ice content, disturbance history (natural and anthropogenic), regional climate gradients, and regional differences in the timing and magnitude of recent extreme warm summers (Raynolds et al 2014, Kanevskiy et al 2017, Frost et al 2018a, Farquharson et al 2019. This variability in ice-wedge degradation contributes to variability in patterns of tundra vegetation change (e.g. tundra greening or browning). Once thaw begins, the resultant subsidence forms small, flooded pits and troughs along the polygon margins. These pits and troughs pock-mark the landscape, kill existing vegetation that is adapted to mesic conditions (i.e. a mechanism for tundra browning) (Lara et al 2018), and support the colonization of hydrophytic vegetation (e.g. wetland sedges and mosses). Secondary impacts can affect large areas because the generation of pits and troughs creates new hydrologic flowpaths that alter soil hydrology and the distribution of surface water (Koch et al 2018). Over time (usually a matter of years to a decade), most pits and troughs become colonized by wetland vegetation, and surface water extent declines due to the development of an organic mat (i.e. a mechanism for tundra greening) (Wolter et al 2016).
Successional processes after ice-wedge degradation could explain in part the increasing NDVI trajectories in ice-wedge polygon landscapes (figure 11). However, this increase is likely also being driven by a general background greening of the tundra landscape in response to climate warming (Berner et al 2020, Myers-Smith et al 2020), as the affected microsites comprise only a fraction of the broader polygonal landscape. The distribution of NDVI dynamics includes numerous pixels with strong 'browning' signals, probably due to extensive ice wedge degradation and increasing surface water (Jorgenson et al 2022). The increasing NDMI (figure 11) over time in these landscapes is likely being driven by the increasing surface water due to the development of pits and troughs.

Cryogenic landslides
Climate-induced thawing of permafrost-affected hillslopes can trigger a variety of abrupt and gradual disturbances involving the mass movement of soils, collectively termed 'cryogenic landslides.' These landslides can result in losses of vegetation, followed by the development of successional vegetation on retransported materials. Different forms of cryogenic landslides vary with respect to their spatial extent and temporal characteristics, and thus the pattern and rate of ecological succession after disturbance. These subtypes include (a) active-layer detachments, (b) frozen debris lobes, and (c) retrogressive thaw slumps. Active-layer detachment slides are relatively small, local slope failures that develop after warm, wet summers, such that saturated active-layer soils slide abruptly over the permafrost table (Leibman 1995, Ermokhina and Myalo 2012, Verdonen et al 2020. Future climate warming and associated permafrost degradation, as well as increases in triggers such as extreme warm summer periods, increases in rainfall, and forest fires, could increase their frequency (Lewkowicz and Harris 2005). Frozen debris lobes are slow-moving, lobate permafrost features consisting of soil, rock, organic material, and ice that move down permafrost-affected slopes via shear along their bases (Darrow et al 2015, 2016, Simpson et al 2016. The distribution and dynamics of frozen debris lobes are comparatively poorly known. Retrogressive thaw slumps are thermokarst slope disturbances that contribute large volumes of materials downslope to lakes, drainage networks, and coastal zones Lewkowicz 1990, Lantuit andPollard 2008). Initiation of retrogressive thaw slumps depends on local geomorphological conditions and meteorology. Fluvial erosion along riverbanks or coastal zones can initiate slope failures, promoted by extended warm and wet conditions (Burn and Lewkowicz 1990). Following an initial slope failure, exposure of ice-rich permafrost enables thaw slump development, which can persist for many years while the areal size of the thaw slumps can expand to tens of hectares. For example, the thaw slump shown in figures 12(b) and (c) expanded from 0.63 ha immediately after the detachment failure in 2016 to 1.04 hectares three years later (Turner et al 2021).
The frequency and size of retrogressive thaw slumps can be highly variable within and among landscapes. The largest thaw slumps in North America have been observed in the Richardson Mountains and Peel Plateau regions, NWT, Canada . This region, which includes the Mackenzie Delta, has experienced an increase in occurrences of thaw slumps in response to wet summer conditions (Lantz and Kokelj 2008. Zwieback et al (2018) also found an increase in thaw slumps on the Tuktoyaktuk Peninsula, northwest of the Mackenzie Delta, Canada, and the Bykovsky Peninsula, Russia, associated with available energy and late-season rainfall. Many coastal areas have seen an increase in thaw slump activity, including Banks Island (Lewkowicz and Way 2019). Interactions with marine environments, including thermo-abrasion from waves and ice, can have a strong influence on thaw slump activity along coastlines (Günther et al 2013).
Cryogenic landslides impact terrestrial and aquatic ecosystems and atmospheric feedbacks. Within lake and river aquatic environments, biogeochemical cycling can be impacted by the liberated sediment and solutes, which are typically rich in nutrients and ions. However, the downstream impacts, on nutrient concentrations, for example, can be highly variable (Frey and McClelland 2009, Harms et al 2014, Lafrenière et al 2017, Mu et al 2017 and depend on local geomorphic conditions including relief, ice content, permafrost extent, and parent material (Tank et al 2020). These complex relations present uncertainties for associated impacts on local and downstream ecology. Vegetation can efficiently colonize stabilized areas of cryogenic landslides (Turner et al 2021). Habitat characteristics associated with landslide age and vegetation composition also have an influence on wildlife (Cray and

Lake drainage
Thermokarst lakes are formed when permafrost degradation results in subsidence of the land, which subsequently fills with water. These features are abundant across ice-rich permafrost terrain and are highly sensitive to climate conditions (Jones et al 2022). Though thermokarst lakes may remain stable for centuries, the shorelines are highly susceptible to erosion and expansion, the rate of which can be strongly influenced by dominant fetch and shoreline ground-ice content as well as climate (Roy-Léveillée and Burn 2010). When shoreline expansion progresses into low-lying areas or invades the boundaries of other thermokarst lakes, they can drain and experience near-complete water loss within days. Additional mechanisms that trigger drainage events can include drainage across an ice-wedge network, headward erosion along adjacent streams or coastal boundaries, and bank overflow when established outflow channels are blocked by snow and ice (Mackay 1981, 1988, Brewer et al 1993, Marsh and Neumann 2001, Hinkel et al 2007, Wolfe and Turner 2008, Jones and Arp 2015. Drainage can also occur incrementally through partial tapping by a stream and the development of open talik systems beneath the lake (Yoshikawa and Hinzman 2003).
Thermokarst lake drainage events represent drastic landscape transitions. Newly exposed lacustrine deposits serve as seedbeds for colonizing vegetation and can quickly develop continuous vegetation cover (e.g. Eriophorum russeolum, Carex aquatilis, and Senecio congestus) within the first few years following drainage depending on local conditions (Ovenden 1986, Mackay and Burn 2002, Shur and Jorgenson 2007, Lantz 2017. For example, willow (Salix spp.) encroached within 30.8% of the former 12 km 2 lakebed of Zelma Lake in Old Crow Flats, Yukon . After a lake drainage event, the aquatic environment of the remaining water body can become highly dynamic for several years following drainage as biogeochemical properties are strongly influenced by weather and pluvial runoff across the exposed lakebed (Tondu et al 2017). Lake water biogeochemical properties stabilize as shrubs encroach, which enhances snowpack depth and snowmelt input . The increasing NDVI in our lake drainage case study trajectories (figure 13) suggests encroachment of shrub vegetation. NDMI likely does not change because encroaching vegetation at these point locations are inundated with water.
Catchment hydrologic and vegetation characteristics typically do not return to pre-drainage conditions (Bandara et al 2020) and can thus exert long-term influence on carbon cycling. Drained lake basins can effectively sequester atmospheric carbon as peat accumulates (Fuchs et  . While the overall surface water area has remained stable in many of these regions, the spatial redistribution of water bodies suggests that these lake-rich landscapes are in a state of climate-driven transition (Rowland et al 2010, Pastick et al 2019. Ongoing research and monitoring will build our understanding of the short and long-term consequences for ecology, hydrology, and carbon cycling.

Limitations, data needs, and unknowns
Broadly, the study of permafrost-related disturbances would benefit from remote sensing studies which leverage higher-resolution sensors. Many of these disturbances at the individual scale can be quite small (e.g. frost circles, <3 m in diameter), and thus medium resolution satellites such as Landsat or Sentinel may miss small-scale changes in surface geology and vegetation driven by thermokarst processes. Additionally, more studies are needed to understand vegetation colonization and succession on newly available land created by permafrost-related disturbances.
Although frost circles are common across the entire Arctic climate gradient, the small size of individual features makes them difficult to detect, even in imagery with submeter spatial resolution. As a result, their distribution has not been mapped or constrained except at local scales. Such mapping of cryoturbated surfaces would be highly desirable, especially in the Low Arctic, where they are at risk of becoming less active (Aalto et al 2017). At present, areas that support dense frost circles can only be predicted based on coarse-scale maps of surficial geology and generalized soil texture. Whereas individual features may be challenging to identify, it may be possible to distinguish cryoturbated landscapes based on landscape-scale average spatial features.
There are numerous unknowns regarding the dynamics of ice-wedge degradation and potential restabilization, and the extent to which they are occurring. Ice wedges are generally insulated by a mat of vegetation and accumulating snow in the winter, and it is still unclear what weather conditions induce ice wedge melting and what might drive heterogeneity in degradation among ice wedges. It is also unclear what factors drive vegetation succession following ice-wedge degradation and the development of surface water ponds and troughs. One factor could be the availability of nutrients such as nitrogen and phosphorus (Beermann et al 2015, Herndon et al 2020), however only a few studies have attempted to address changes in nutrient concentrations following ice-wedge degradation (Norby et al 2019). Finally, the rates of accumulation of organic matter in degraded ice wedges and their potential for stabilization are still poorly understood. Field studies of ice wedge dynamics utilize space-for-time substitution, examining ice wedges at different stages of degradation (Jorgenson et al accepted) as opposed to assessing the dynamics of individual ice wedges over time.
There has been substantial progress on our capacity to gauge the extent of ice-wedge degradation utilizing high-resolution remote sensing and machine learning techniques (Witharana et al 2020(Witharana et al , 2021. Whereas these studies and associated applications can map ice-wedge polygon networks across extensive areas of land, and even potentially estimate the fraction of land that contains ice wedges versus polygon centers, there is still work to be done to distinguish among the different stages of degradation. There have been many studies that have documented the detection of cryogenic landslides (e.g. Barnhart and Crosby 2013, Balser et al 2014, Swanson and Nolan 2018, however, detection of the frequency of relatively small landslides may be difficult using medium resolution imagery (e.g. Landsat). Thus, large-scale mapping of these disturbances is difficult because the size of individual thaw slumps can be characteristically different depending on the region, and because high-resolution imagery at large scales is both cost prohibitive and difficult to work with.
Lake drainage events and associated impacts are complex and require additional research, especially where drainage frequency is increasing. Our ability to identify where and when thermokarst lake drainage will occur in the future must be refined. Existing data archives (e.g. Landsat 5-8, Sentinel-2) provide resources needed for identifying locations of past drainage and associated changes in land cover of larger lakes. While many studies have successfully utilized products from these sensors, the availability of scenes can be limited for any given year according to the timing of cloud-free conditions and the spatial resolution may not be adequate for detection of small-scale surface area change (e.g. <30 m resolution) or for smaller water bodies. Broader coverage of high-resolution (optical, radar and elevation) products will improve these analyses and enhance detection of landscape responses to drainage and geomorphological characteristics (e.g. the proximity of lakes to low-lying areas) that make lakes vulnerable to drainage.

Anthropogenic disturbances
The North American ABZ has experienced extensive industrial activity and development in the last half-century (Schneider 2002, Pasher et al 2013. These disturbances include flooding for hydroelectricity, timber harvest, and other natural resource development (e.g. mining, oil, and natural gas), including associated infrastructure such as pipelines, roads, and seismic lines for resource exploration. Additional highly localized disturbances in this region include landfills and dumps for disposal of domestic and industrial waste. These disturbances do not always fully remove or eliminate vegetation and soil, but often result in highly fragmented landscapes, leading to significant changes in ecosystem composition, structure, and function (Pasher et al 2013). As climate continues to change, northward expansion of agricultural areas is expected in southern regions of the ABZ, resulting in lasting removals of natural vegetative cover (King et al 2018). Although the cumulative area disturbed by the combined activities is vast, the impact of past and present natural resource development on ABZ ecosystem function (e.g. For example, the Canadian forest products industry harvested over 710 000 hectares (∼143 million m 3 ) of forest in 2020 (National Forestry Database 2020). In these higher productivity and more easily accessible southern and eastern regions, coniferous evergreen species (e.g. spruce, fir and pine) and aspen dominate the landscape and are utilized for lumber, pulp, and paper (Burton et al 2003, Venier et al 2014. In comparison, timber harvest is less extensive in the Alaskan boreal forest (Potapov et al 2008), where managed forests are generally concentrated in areas with high-value sawtimber species, adequate road access, and proximity to milling facilities-mostly in southeastern Alaska (Morimoto and Juday 2018) and episodically within the interior.
Clear-cutting is the most common silvicultural method used in the boreal forest (Haggstrom and Kelleyhouse 1996, Burton et al 2003, Cyr et al 2009. It was initially justified as an adequate replication of stand-replacing natural wildfire (Bergeron et al  2002); however, post-treatment belowground conditions (e.g. soil depth, nutrient content) can substantially differ from those following wildfire (Simard et al 2001), ultimately impacting post-disturbance successional trajectories in unique ways (Nguyen-Xuan et al 2000). Additionally, the coarse woody debris left after wildfire generates habitat for songbirds and other species, but is largely absent from post-harvest landscapes (Morissette et al 2002). Finally, post-treatment planting can increase regrowth compared to postfire regrowth (Dieleman et al 2020). This is evident in the NDVI and NDMI time series for our logging case studies (figure 14), which in general show a faster initial recovery than those for fire (figure 4), and aligns with prior research (White et al 2017). However, it should be noted that other work has found the opposite result wherein post-fire forests recover slightly more quickly than harvested areas (Bartels et al 2016).
Traditional clear-cutting results in even-aged forest stands, as all trees are either harvested or disturbed due to harvesting activity, with only a small fraction left to stand as a seed source. Consequently, intensively managed landscapes often yield an even distribution of tree ages across the managed area, with no or few stands older than the harvest rotation time (Bergeron et al 2002). When the rotation time is shorter than the fire frequency, the resulting stands will be less diverse in terms of stand structure and species composition than stands that grow for longer periods and allow tree replacement or fires to kill a population of trees. Long fire intervals (e.g. 200+ years) allow for shifts in canopy dominance and forest age structure as a result of forest successional processes (Bergeron et al 2002). Thus, biodiversity concerns for highly managed areas have arisen, particularly in southern and eastern Canada (Boucher et al 2009, Venier et al 2014. While clear-cut or group selection harvests predominate areas with high value stands or in areas where managers are attempting to mimic fire, approaches such as partial harvest and individual tree selection in mixed or deciduous stands often allow for individuals with a range of ages to coexist and the promotion of certain forest types (Gauthier et al 2009). Such harvesting practices can help increase diverse forest structural attributes, particularly in stands that are even-aged following prior harvest practices (Bose et al 2015). Comparatively, selective harvest is less impactful on total stand biomass than even-aged selection or fire, and thus has a more nuanced signal

Oil and gas well production
Oil and gas well production in the North American ABZ can be traced back a century to the still-active Norman Wells drilled in the 1920s in the Northwest Territories (Bone and Mahnic 1984). In British Columbia, the first commercial gas well was drilled along the Peace River in 1947 and the first discovery of oil in Alaska occurred in 1957. The density of wells in the ABZ is typically less than 1 km −2 , however some locations can be as dense as 3 km −2 or higher (Warrack et al 2021).
Over its long history, oil and gas exploration and the associated production technology, practices, and regulations have evolved King 2013, Kang et al 2016). Depending on the type of well (e.g. conventional oil, unconventional gas), intent of the well (e.g. production, exploration, injection), geology (including depth and formation properties), and other factors, the resulting disturbance to the surrounding vegetation can be highly variable in terms of size, shape, and form, with the area of influence ranging from tens to hundreds of square meters. The disturbance also varies temporally throughout the life cycle of the well from site preparation to plug and abandonment (Burnham et al 2012, Allen et al 2013).
For both exploratory and development (or production) wells, well site preparation includes constructing a well pad and access roads. The lengths of new access roads for well sites in the Wayne National Forest in Ohio are 8-30 km (USFS 2004), but the lengths of new access roads needed in the ABZ may be much longer (Pasher et al 2013, Wilkinson et al 2021. The well pad involves clearing land so that the drill rigs can be brought in. Wells meant for producing oil and gas are first cased with steel piping and cemented, and then the inside of the innermost casing is connected with the host rock containing oil and/or gas. These activities remove vegetation, degrade soils, result in loss of seed and bud stores (Pickell et al 2015), and  Northrup and Wittemyer 2013). Due to the impacts of well production on soil nutrients, hydrology, and seed sources, regeneration on well sites is slower than that following fire or forest harvest (Osko and MacFarlane 2001). Forest succession and regrowth and overall landscape recovery can thus take decades following oil and gas activity (Powers et al 2015, Chowdhury et al 2017. NDVI and NDMI responses to oil and gas wells (figure 15) show a clear decline in both vegetation moisture and greenness, with recovery lasting longer than ten years.
The production life of a well is highly variable, with some wells remaining in production for decades and others being abandoned after only a few years. Nevertheless, all wells are eventually abandoned and, according to modern regulations, must be plugged and the well site restored (Kang et al 2019). In Alberta and British Columbia, site restoration involves removing surface infrastructure and revegetating the land to pre-development conditions . However, some wells have not been plugged and abandoned according to these regulations and have not had the surface restored.

Seismic lines
The largest anthropogenic disturbance across much of boreal and Arctic North America are seismic lines (Jorgensen et al 2010, Strack et al 2019), which are long linear clearings cut across forests and wetlands for oil and gas exploration (figures 1(d) and 16(a), (b)). Seismic exploration for underground sources of oil and natural gas involves drilling a series of holes 6-20 m deep along the lines and analyzing the reflection of sound waves generated from either explosives detonated at the site or truck-mounted surface vibrators (EMR 2006). Originally, these lines (previously known as legacy or 2D lines) were cleared using heavy machinery to cut through heavily for- These changes in soil characteristics and hydrological conditions can alter understory vegetation composition, including shifts from feather moss-shrub dominated understories to complete cover by sedges (e.g. Carex aquatilis) in fen peatlands, or sphagnum moss (Sphagnum spp.) in bog peatlands (Deane et al 2020, Davidson et al 2021). In recent decades, there has been a move towards a method called 'low-impact' seismic lines, created using lighter-weight machinery and by hand and allowing for minimal disturbance to the ground-surface (Dabros et al 2018). These lines are narrower (1-5 m) than legacy lines but they are far more abundant on the landscape, creating a dense grid-like network of disturbances and can still create substantial changes to both tree cover (van Rensen et al 2015) and understory vegetation communities (Davidson et al 2021).
Although the creation of some seismic lines occurred almost 40-50 years ago, tree recovery and regeneration in many of these locations is slow and often fails. For example, Lee and Boutin (2006) estimated that after 35 years, approximately 65% of seismic lines crossing forests in Canada's boreal plains remained free of woody vegetation. Yet, our mechanistic understanding of how seismic testing influences vegetation recovery is limited. For example, in wetland locations, mechanical flattening of localized topography can result in a water table closer to the ground surface, leading to unfavorable conditions for black spruce (Picea mariana) seedlings to regenerate (Lee and Boutin 2006, Caners and Lieffers 2014). Furthermore, the post-disturbance understory vegetation communities, often dominated by hydrophilic species such as sedges and sphagnum mosses, may outcompete slow growing tree saplings (Davidson et al 2020b). In addition to the initial disturbance, continued use of these linear features for hunting, recreational sports, and further resource extraction activities can hinder tree recovery (van Rensen et al 2015). This poor recovery can be seen in our Landsat case studies of vegetation response to seismic lines ( figure 16). There is a substantial drop in NDVI and NDMI at both upland and peatland seismic line sites following disturbance given trees are actively removed, and NDMI recovery is slow.

Limitations, data needs, and unknowns
Recent progress has been made to identify and map annual forest disturbance from logging across the North American ABZ based on the Landsat data archive spanning 1984-2014 (Zhang et al 2022). Between 1987 and 2012, 10.8% of the Alaska and western Canada experienced disturbance, with 1.4% attributed to logging. However, state and provincial forestry records are still an essential data source for understanding the scale and impact of logging and validating satellite detection of forest management, especially for lower-impact forestry practices that may be challenging for remote sensing approaches to detect. Such long-term data (e.g. polygons dating back to the 1960s, and GeoPDFs dating back to the 1800s in Saskatchewan, Canada) are crucial for studies of the impact of forest management on the North American boreal forest. However, many of these records are difficult to obtain. Similarly, data and records of seismic lines are not readily available across all Canadian provinces.
There are limited studies on land disturbances caused by oil and gas well production and exploration with only a few recent studies that are based in the contiguous U.S. (Raynolds et al 2014, Nallur et al 2020, Chomphosy et al 2021. Though databases with information on wells (i.e. intent, type, age, etc) are developed and maintained by numerous state, provincial, and territorial governments as well as the U.S. Bureau of Land Management for wells on federal lands, they can be incomplete (e.g. completely missing wells or incomplete information on well depth and age, etc). Nevertheless, these databases have been compiled for Canada and the U.S. to understand oil and gas well distribution, methane emissions, and other environmental impacts . Commercial databases are also available (e.g. GeoScout), however, they are not likely to contain information on the size of well pads and land disturbances. There is research on using machine learning and high-resolution imagery to detect active oil and gas well pads, which may provide data on well pad sizes and shapes (Bartsch et al 2020). Overall, there is a need for improved oil and gas well databases and information on well pads to understand the full extent of impacts.

Weather-related disturbances
Though anthropogenic-driven climate change is likely to have longer-term 'press' disturbance effects on ABZ vegetation, a handful of weather-related disturbances can affect vegetation markedly in the shortterm, including rain-on-snow events, heat waves and extreme drought, and windthrow. Such disturbances can impact boreal and tundra vegetation, nutrient, and hydrology dynamics.

Rain-on-snow
Rain-on-snow events, or more broadly wet surface snow conditions (Pan et al 2018), are driven by a range of physical processes, though most often are caused by wintertime rain events that result in a wet snow surface (Singh et al 2000). Wet snow conditions can cause flooding and paludification in ABZ ecosystems, accelerate permafrost thawing, and decrease vegetation productivity , Bjerke et al 2014, Jeong and Sishama 2018. In mountainous regions rain-on-snow can destabilize the snowpack and trigger avalanches (Conway and Benedict 1994).
Most notably, re-freezing of melted snow creates ice barriers between the soil surface and the snowpack, making it difficult for ungulates such as caribou (Rangifer tarandus) and musk oxen (Ovibos moschatus) to forage for lichen during the winter . These water and ice layers also facilitate the growth of toxic fungi, which can spoil lichens, further lowering wintertime food sources for ungulates, increasing foraging efforts and negatively impacting fat and protein reserves. In some cases, this can lead to movement of herds outside of their normal ranges, or even starvation and death, as occurred in 2003 on Banks Island,  While an individual rain-on-snow event is generally short-lived-on the order of days-the subsequent impacts on soil hydrologic and thermal conditions can last months. The frequency of rain-on-snow is predicted to increase in the future in the ABZ due to rising temperatures (Ye et al 2008, Jeong and Sishama 2018, Pan et al 2018, with potential cascading impacts on hydrology, thermal conditions, ecosystem function, and ecosystem services. . This plant damage can decrease productivity and lead to mortality. For example, an experimental manipulation study of a sub-Arctic heathland found a 50% reduction in gross primary production (GPP) after multiple extreme winter warming events (Bokhorst et al 2011). Such extreme warming events are predicted to increase in the future as temperatures rise (Meehl and Tebaldi 2004). In Utqiagvik, Alaska, the number of winter days with maximum temperatures above freezing has steadily increased since 1960, and several record-high days occurred in 2020 (as compared to the previous 20 years) (figure 17). Such temperature anomalies will continue to impact ABZ vegetation, potentially leading to plant damage and decreased productivity if wintertime extremes continue to increase (Richardson et al 2018).

Windthrow
Windthrow, or tree blowdown events from high wind, are important disturbance agents within the North American boreal zone that act primarily at the standscale (Ruel 2000, Bouchard et al 2009. While extreme wind events resulting in stand-replacement are rare in the boreal zone, partial windthrow where some individuals survive is more common, with return intervals ranging from 40 to 450 years in eastern Canada (Ruel 2000, De Grandpré et al 2018. Damage to trees depends on individual tree and stand factors, including tree size, species, canopy position, and previous stem damage, as well as soil depth and moisture, stand density, fragmentation, and angle with respect to wind direction (Peterson 2004). Tree size and species are the most reliable predictor of windthrow survival-some tree species are more 'wind firm' than others, and damage susceptibility increases with increasing tree size (Peterson 2004, Rich et al 2007. Because of the differential impact of partial windthrow on tree size and species, these events can cause shifts in the species composition and stand structure of impacted stands (Veblen et al 2001, Girard et al 2014. Windthrow can also act as a trigger for subsequent bark beetle outbreaks, as beetle populations are able to colonize and grow within downed stems (Wichmann and Ravn 2001).

Limitations, data needs, and unknowns
Some of the main challenges of studying extreme weather events like rain-on-snow, winter warming, and windthrow include the sparsity of weather stations in northern regions, the lack of routinely deployed weather equipment , and the unpredictable occurrence of events such as severe blowdown (Bouchard et al 2009). Detection of rain-on-snow events with satellite measurements is possible using radar, microwave, and multispectral imagery (Serreze et al 2000, Bartsch et al 2010, Pan et al 2018. Accurate detection of windthrow depends on the spatial resolution of remotely sensed measurements compared to the scale of the blowdown (Schwarz et al 2003). An enhanced monitoring network of weather conditions and snowpack, such as those present in the SNOTEL network (Schaefer and Paetzold 2001) would help better characterize and identify the occurrence of these events.
With respect to extreme drought and heat waves, while the physiological mechanisms underlying plant drought response and vulnerability are well established and emerging remote sensing techniques offer promise , it is still difficult to predict which individuals will die from such drought stress (Trugman et al 2021). Critical needs include further understanding of plant physiological and site characteristics that influence drought exposure and susceptibility and better information about how biotic agents interact with drought to cause plant mortality (Trugman et al 2021).

Background
Despite their relatively small footprint in ABZ landscapes, riparian zones are disproportionately important for ecological disturbance (Scrimgeour et al 1994), hydrological processes (Ploum et Beltaos 2002, Beltaos et al 2006), biogeochemistry and water quality (Tiwari et al 2022), and beaver colonization (Tape et al , 2022. In addition, there have been widespread changes observed in permafrost extent both on floodplains and within their catchments (St. Jacques and Sauchyn 2009, Jones and Rinehart 2010, Quinton et al 2011, Tananaev and Lotsari 2022. It has been hypothesized that these processes will lead to a reduction in the areal extent of active floodplains in ABZ landscapes due to increased river channelization, smaller peak flows, and reduced riparian disturbance intensity (Ström et al 2011, Nilsson et al 2013, Jansson et al 2019.
Streams and rivers in the ABZ are strongly influenced by geology and topographic relief as well as hydroclimate, ice cover, and the permafrost regime (Ashmore and Church 2001, Rokaya et al 2018), with high variability in river morphology (Nilsson et al 2015). Streamflow rates can range from slow-moving tundra streams to large flowing rivers that span Arctic-boreal ecotones (e.g. the Mackenzie and Yukon rivers) (Nilsson et al 2015). Riparian ecosystems are especially dynamic because they experience frequent erosion, flooding, and sedimentation (Wiens 2002). Channel migration and flooding can be seen as similar to fire disturbance, both creating shortterm destruction to vegetation with the capacity for regeneration following the event (Rood et al 2007). Channel migration in particular can 'reset' vegetation succession at any successional stage through floodplain erosion and simultaneous sedimentation and creation of new land for vegetation establishment (Walker and Chapin 1986, Viereck et al 1993, Van Cleve et al 1996, Helm and Collins 1997, Lininger et al 2017.
In addition to channel migration, ice-jam flooding is also an important disturbance in ABZ riparian zones. Ice-jams occur when ice floes in rivers are impeded by stationary ice covers, bridges, islands, or river width constrictions, leading to flooding (Rokaya et al 2018). Ice-jam flooding can occur during any river ice freeze-up or breakup period but are most common during the spring breakup period ( Beavers are important ecosystem engineers in the North American ABZ through their dam-building and hydrologic engineering of rivers, streams, sloughs, and lakes. Previously considered only a sub-Arctic species, recent observations show beaver colonization into low arctic tundra regions of Alaska and Canada in recent decades , Jones et al 2020b due to climate-change driven landscape change as well as population recovery from historical over-trapping . Beaver dams trap water on the landscape, turning streams and sloughs into connected ponds, widening riparian zones and altering groundwater flow (Westbrook et al 2006, Tape et al 2022. Jones et al (2020b) found that beavers preferentially targeted thermokarst landforms in their dam-building activities within the Baldwin Peninsula, Alaska, accounting for 60% of the increase in surface water in the region between 2002 and 2019. Increases in surface and groundwater due to beaver dams transfers additional heat to the ground and thaws permafrost surrounding and beneath beaver ponds (Tape et al 2022). In permafrost-affected regions, beavers have the capacity to initiate and affect lake formation and drainage, ice-wedge degradation, cryogenic landslides, and other thermokarst events (Jones et al , 2021. These physical changes to waterways and the surrounding permafrost effectively create warmer patches of mixed aquatic and terrestrial ecosystems that likely act as oases.

Limitations, data needs, and unknowns
Given the role of climate and extreme events on floodplains, spatiotemporal properties of disturbance, succession, and floodplain evolution are likely to be influenced by recent climatic warming at high latitudes, leading to important changes in the structure and function of riparian ecosystems in the ABZ. However, most ecosystem change studies to date have focused on upland and lowland ecosystems, whereas the observational record for riparian zones is comparatively sparse. There is thus substantial uncertainty concerning recent changes and future trajectories on floodplains across gradients of climate, stream order, catchment size, and floodplain morphology. For example, the pace of vegetation succession may increase in a warming climate due to longer, more productive growing seasons and changes in permafrost properties on or near riparian zones, particularly in forest-tundra ecotones (Wilmking and Juday 2005, Kharuk et al 2006, Beck et al 2011, while altered flow-regimes may influence the frequency and intensity of disturbance regimes. In Alaska, several studies have documented conspicuous, long-term increases in the extent and canopy height of tall shrublands in subarctic and Arctic riparian zones (Tape et al 2011, Brodie et al 2019, Liljedahl et al 2020. Understanding the interactions between biological and physical processes in the context of climate warning is important for assessing long-term impacts of continued warming on ABZ floodplains. Beaver activity may be an important disturbance within permafrost regions, potentially causing widespread changes to the hydrologic and biotic environment, and initiating permafrost degradation (Tape et al 2022). Current research is exploring how these newly constructed oases affect carbon cycling, aquatic and terrestrial biodiversity, fish, and other ecosystem attributes. Further investigation is needed to understand the spatial extent and implications of beaver activity within the North American and circumpolar ABZ (Tape et al 2022).

Mammalian herbivore activity 3.7.1. Background
Mammalian herbivores like moose (Alces alces), caribou (Rangifer tarandus), and snowshoe and arctic hares (Lepus americanus, L. arcticus) impact ABZ ecosystems through coupled herbivore-vegetation feedbacks. For example, selective foraging, trampling, and inputs of excreta, urine, and decomposing carcasses can directly alter plant community composition or indirectly affect ecosystem properties through changes to soil characteristics and nutrient cycling (Olofsson et al 2004, Väisänen et al 2014, Schmitz et al 2018, Leroux et al 2020. These species are also a crucial subsistence resource for indigenous communities (Rexstad and Kielland 2006). Caribou in particular occur in high abundance across much of the North American ABZ, numbering in the millions, and are one of the Arctic's most ecologically, culturally, and economically important species (Hummel and Ray 2008, Parlee et al 2018, Gagnon et al 2020. These large herbivores also make some of the longest terrestrial animal migrations in the world, with some herds traveling over 1000 km from boreal wintering grounds to Arctic tundra breeding grounds (Gurarie et al 2019, Joly et al 2019. During calving and migratory periods, caribou herds aggregate in dense groups and can alter landscapes as they pass through, impacting vegetation cover and structure, soils, and ecosystem carbon storage (Olofsson and Post 2018).
The distribution and intensity of caribou impacts are driven primarily by grazing and trampling associated with fluctuations in population sizes, which occur on a multi-decadal basis (Gunn 2003, Vors and Boyce 2009, Joly et al 2011. These fluctuations are influenced by snow conditions and forage availability (Post and Forchhammer 2002, Gunn 2003, Joly et al 2011. A meta-analysis of caribou impacts on vegetation cover across the Eurasian and North American ABZ showed a clear negative effect on lichen (Bernes et al 2015). Because lichens are slow to recover from disturbance, this impact is both acute and long-lasting (Suominen and Olofsson 2000, Joly et al 2009, Macander et al 2020. Reductions in lichens in turn drive density-dependent feedbacks on caribou, causing population declines and influencing population cycles (Manseau et al 1996, Gunn 2003. Impacts of caribou trampling and grazing on vegetation can also include transitions to graminoid dominated communities (van der Wal 2006), and constraints on deciduous shrub expansion (Olofsson et al 2009, Christie et al 2015, Bråthen et al 2017 or treeline advance (Munier et al 2010, Bryant et al 2014. Caribou impacts are most pronounced in arctic environments where population densities are highest. In the boreal zone, low caribou density likely minimizes impacts. In contrast, herbivores like hares and moose in the boreal forest can shift the age distribution of the foraged species towards younger age classes (Butler 2003. Selective feeding can also shift species composition. For example, moose herbivory can cause a shift from palatable deciduous species towards unpalatable evergreen species (Pastor et al 1988. Recent work suggests that moose alter their behavior to favor dense canopy areas during increased summer temperatures, suggesting shifts in areas vulnerable to browsing under warmer conditions (Jennewein et al 2020). Whereas moose generally avoid evergreen species like white spruce (Picea glauca), snowshoe hares browse heavily on white spruce seedlings, especially during periods of high hare abundance (Rexstad and Kielland 2006, Angell and Kielland 2009, Sharam and Turkington 2009, Hollingsworth et al 2010. Snowshoe hare populations in Alaska and Canada exhibit cyclic dynamics, driven by predator population size and herbivorevegetation feedbacks (Krebs et al 2018). During peaks that occur about every ten years, snowshoe hare browsing can alter vegetation composition and plant chemical defenses (Fox and Bryant 1984), suppress the succession of white spruce (Olnes and Kielland 2016), and curb treeline advance (Olnes et al 2018).

Limitations, data needs, and unknowns
Most studies of herbivore impacts on vegetation use exclosures to assess what happens when herbivores are removed from a system. However, responses of vegetation to increasing vs. decreasing grazing pressure are not equal (Olofsson 2006). For example, studies that examine the impact of increasing caribou herd size (typically observational) often report stronger impacts than experiments that exclude caribou and examine the impact of decreasing herd size (typically manipulative) (Olofsson 2006). Geographic disparities in research can also influence conclusions. For example, studies of caribou impacts on vegetation primarily come from Fennoscandia (Soininen et al 2021). This raises issues of transferability of results because ecological conditions are different. Most caribou in Fennoscandia are managed in domesticated or semi-domesticated herds that often occur at higher densities than wild herds in North America (Bernes et al 2015).
Results from remote sensing and modeling studies which attempt to capture the relationship between caribou population density and vegetation productivity have produced mixed results, with some studies reporting significant negative relationships (Rickbeil et  animal tracking datasets that can be used to analyze climate-driven variation in animal movement and foraging activity. As remote sensing technologies improve, increasing spectral and spatial resolution of satellite imagery might bolster the ability to quantify herbivore impacts across space and time.

Temporal and spatial scale of disturbances
Disturbances in the North American ABZ notably occur across a wide range of spatial and temporal scales (table 1). The spatial grain of individual disturbance events ranges from on the order of meters for individual patterned-ground features such as frost circles (Frost et al 2013) to 1000s of square kilometers for large boreal 'megafires' (Stephens et al 2014). Temporally, ABZ disturbances occur over the course of hours or days, such as windthrow, or over years, such as with drought (Michaelian et al 2011). Their return frequency for the same location also varies from a general one-time event, such as with lake drainage (Shur and Jorgenson 2007), to an annual occurrence, such as with cryoturbation ( The temporal and spatial scale of disturbance occurrence and recovery as well as the overall intensity of impact can also vary within disturbance and landscape types. For example, high severity boreal wildfires tend to be stand-replacing large-scale events lasting weeks or months (Sedano and Randerson 2014, Rogers et al 2015, Veraverbeke et al 2017, in contrast to smoldering fires, which can burn yearround and survive the winter (Scholten et al 2021). Spatially, the resolution of individual disturbance events can be quite small but can cover large extents in their overall scale of impact. For example, insect infestations occur at the individual tree scale, but can then spread to whole stands and landscapes (Raffa et al 2008). Similarly, though individual seismic lines cover only a few meters in area, their combined extent is vast across the North American ABZ (Jorgensen et al 2010).
We compiled these spatiotemporal characteristics across disturbance types (table 1) and analyzed how they vary using a PCA. The results from our PCA analysis ( figure 19) indicate the broad spread in the spatiotemporal characteristics associated with ABZ disturbances. The loadings for frequency and intensity and size and occurrence/recovery timeline are opposite one another, indicating negative correlation. In general, high-intensity events occur at a lower frequency than lower severity disturbances which only impact productivity (but not necessarily mortality) (table 1; figure 19). Some of the overarching groups are clustered together in the PCA (e.g. anthropogenic, pests and pathogens, weather), whereas the permafrost-related disturbances span the entire range of the first two principal components.
Understanding spatiotemporal differences is crucial when detecting and studying these disturbances via remote sensing, or when including them in process-based models. Advances in Earth observation sensor resolution have improved the capability to characterize and monitor disturbances and their interactions. However, in the context of detection and monitoring of multi-disturbance landscapes, an integrative approach is necessary to extend knowledge about disturbance (or multi-disturbance) recovery processes across high-latitude landscapes. Integration with remote sensing typically implies validation against pre-and post-disturbance in situ data across whole landscapes, and often involves crosssensor harmonization to extend temporal or spatial ranges. Synthesis of disturbance-related studies toward understanding disturbance processes and their interactions across such a broad and heterogeneous domain requires bridging of temporal and spatial scales across scientific disciplines (i.e. ecology, geology, hydrology, etc) (Cavender-Bares et al 2022). The disturbance spatial grain and extent are particularly important, and should match the spatial resolution of the sensor (Senf et al 2017b, Duncan et al 2020. Sensor pixel size is known to affect the measurement magnitude, location, and geospatial congruence of disturbance hotspots and the characterization of the effects of disturbances on ecosystems (Cavender-Bares et al 2022). While some of the midresolution sensors like Landsat have long records and are capable of tracking trajectories, they may be limited to tracking only larger-scale disturbances because their pixel size (e.g. 30 m) is large relative to the subpixel of disturbances such as cryoturbation (∼1-5 m) or the early stages of insect outbreaks.
Scale is also crucial for the prediction of future disturbance effects, interactions, and feedbacks using process-based modeling. Models that do not consider individual plant species, such as many global climate models, will not fully capture species-specific effects of biotic disturbances, herbivory, and windthrow, or accurately capture successional dynamics following disturbances (Foster et al 2019, Shugart et al 2020). Ecosystem demographics represented in a modeling framework should interact with vegetation dynamically and be represented at scales that correspond to the frequency and extent of the disturbances that the model framework includes (Seidl et al 2011, Albrich et al 2020. For example, fine temporal scales (e.g. daily, table 1) may be required to accurately model the disturbance interactions of a wildfire leading to a cryogenic landslide. It is also crucial to consider gridcell-to-gridcell spread of 'contagious' disturbances like fire or insect infestation, as well as the temporal and spatial scales at which this spread occurs (Johnstone et al 2011). Representing the spatial and temporal complexities of multi-disturbance interactions in these systems accurately is an emerging area of high-resolution forest and tundra modeling. As remote sensing and modeling technologies improve, and more accurate and spatially continuous occurrence data are acquired, we will be better able to detect and predict ongoing ABZ disturbances, as well as their future trajectories.

Disturbance interactions
Disturbances within the ABZ can interact with one another, often with positive feedbacks that amplify the impact of subsequent events, such as wildfire and subsequent abrupt permafrost thaw (Gibson et al 2018). Other interactions may have a negative or dampening effect on subsequent disturbances, such as cryogenic landslides and subsequent reduction in wildfire potential (figures 20 and 21). Broadly, disturbances may interact by altering the resistance of an ecosystem to subsequent disturbances, altering the probability of future disturbances, or by altering an ecosystem's resilience, or its ability to recover from a subsequent disturbance and its overall impact  (table 1). Qualitative characteristics were modified to numeric scalar values (table S2). and severity (Buma 2015). As most of these disturbances are predicted to increase in frequency, severity, and/or extent with climate change (Chen et al 2016, Veraverbeke et al 2017, Pan et al 2018, Pureswaran et al 2018, Berner and Goetz 2022, the opportunity for interactions among these disturbances will likewise increase, leading to potentially nonlinear and cascading impacts on ABZ ecosystems and vegetation (Buma 2015, Seidl et al 2017. Typically, studies of disturbances, in the ABZ or otherwise, only focus on a single disturbance type, and thus do not capture the true potential impact of a disturbance that includes its downstream effects on other disturbance regimes (Seidl and Turner 2022). Here, we discuss some of the interactions between ABZ disturbances and present our findings in figures 20 and 21 but note that there are many complex interactions which are still the subject of further study.
Due to the ubiquitous nature of wildfire across the North American ABZ, fire interacts with most other disturbances within these regions (figures 20 and 21). Drought and wildfire are often linked, with low moisture conditions increasing fuel flammability (i.e. decreasing resistance), and post-fire impacts on soil conditions often leading to moisture stress (i.e. decreasing resilience) (Whitman et al 2019, Baltzer et al 2021. In general, fire probability increases in the initial stages following bark beetle outbreaks as needles dry and thus become more flammable (Jenkins et al 2012(Jenkins et al , 2014. However, once the needles fall, the ground-to-canopy continuity is lost, thus lowering the probability of high severity crown fires. Low severity fires that damage trees but do not kill them can increase susceptibility to insect and pathogen attack and subsequent mortality (Hood and Bentz 2007), however stand-replacing wildfire removes host availability and thus decreases the risk for outbreak (Veblen et al 1994). Fire in permafrost areas can lead to thermokarst features, permafrost degradation, and changes to hydrology (Holloway et al 2020). Research indicates that tundra fires are becoming more frequent (French et al 2015, Hu et al 2015 and that post-fire deciduous shrub expansion may, in turn, further facilitate fire (Higuera et al 2008, Lantz et al 2010a, Bret-Harte et al 2013, Gaglioti et al 2021. However, herbivory and trampling of expanding deciduous shrubs has the potential to provide a negative feedback effect that lengthens fire return intervals in the Arctic (Olofsson et al 2009, Christie et al 2015, Bråthen et al 2017. Beavers have also been shown to prevent fire spread and provide fire refugia (Fairfax and Whittle 2020).
Aside from fire, windthrow is often a precursor to bark beetle infestation through facilitation of beetle population growth within downed Figure 20. Interactions between disturbances in the North American boreal forest. Driver (x-axis) disturbances are the initiating disturbance, whereas response disturbances (y-axis) are the potential subsequent disturbances. Negative interactions correspond to a dampening effect of the driver on the response disturbance. Positive interactions correspond to an enhancing effect of the driver on the response disturbance.

Figure 21.
Interactions between disturbances in the North American Arctic tundra. Driver (x-axis) disturbances are the initiating disturbance, whereas response disturbances (y-axis) are the potential subsequent disturbances. Negative interactions correspond to a dampening effect of the driver on the response disturbance. Positive interactions correspond to an enhancing effect of the driver on the response disturbance.
logs (Christiansen et al 1987, Malmstrom andRaffa 2000). Defoliators and bark beetles influence one another, where defoliators can weaken hosts and increase susceptibility to subsequent attacks by bark beetles (Cole et al 2022). Likewise, drought and biotic disturbances can enhance one another through decreased vegetation resilience (Malmstrom and Raffa 2000, Boyd et al 2021, Ruess et al 2021. Many disturbances are linked with cryoturbation, ice wedge degradation, and cryogenic landslides (figures 20 and 21). For example, a physical disturbance to the landscape, such as a fire or seismic line placement, can reactivate cryoturbation features and local permafrost degradation by removing live vegetation and surface organic material (Frost et al 2013). Thaw slumps can also trigger catastrophic drainage of adjacent thermokarst lakes (Marsh et al 2009).
Anthropogenic features such as roads, seismic lines, and logging affect the landscape and can result in additional disturbance; roads can lead to additional wildfires by opening access to human ignitions. Across Canada, the majority of human-caused ignitions are within 10 km of communities . These fires then have the potential to destroy human infrastructure. However, roads and infrastructure can also act as fire breaks and prevent fire spread (Narayanaraj andWimberly 2011, Cochrane et al 2012). Some salvage logging can take place after a fire event, but fires can also destroy stands designated for harvest, or previously harvested stands. Insect outbreaks and pathogens have destroyed merchantable timber across Canada (Volney andFleming 2000, Hennigar et al 2007), reducing the area available for harvest. The large network of seismic lines associated with oil and gas exploration has also negatively impacted habitat quality for boreal woodland caribou across Canada, with many populations in decline (Hebblewhite 2017, Nagy-Reis et al 2021. This type of habitat fragmentation has been shown to alter animal behavior and reduce mammalian movements globally (Finnegan et al 2018, Tucker et al 2018. Many of the disturbances have no or unknown interactions (figures 20 and 21), either because of lack of study (e.g. insect outbreaks and pathogens and subsequent thaw slumps) or because the disturbances are not generally co-located (e.g. cryoturbation and logging). These unknowns present both an opportunity and need for further study as well as the potential for previously geographically separate disturbances to interact as climate change continues to modify their extent and range. Disturbance interactions in particular should be a priority for further field, remote sensing, and modeling studies in the ABZ.

Conclusions
Present in all these disturbances is the amplifying effect of climate change, as this region is warming much faster than other areas of the globe (Price et al 2013, Smith et al 2019, Chylek et al 2022. The direction and magnitude of precipitation change is of growing concern, and this shift will feed back to changes in disturbance trajectories-a drier landscape will lead to larger and more severe wildfires, whereas abrupt permafrost thaw may increase in a wetter environment that dampens wildfire risk. Ultimately, disturbances are pivotal in creating local hotspots of change against the backdrop of long-term climate change. These disturbances create the potential for persistent shifts in vegetation composition (e.g. shift towards deciduous dominance post-fire) and biomass and extent (e.g. tall shrub and tree migration at treeline) (Mack et al 2021, Maher et al 2021, Foster et al 2022.
Disturbances also have the capacity to increase colonization and spread of non-native and invasive plant species (Sanderson et al 2012, Kent et al 2018, Kelly et al 2020. Previously, boreal and Arctic ecosystems were seen as too hostile and remote to facilitate invasion of non-native species (Sanderson et al 2012), however increasing temperatures and longer growing seasons are facilitating the northward migration of species in response to climate change (Chen et al 2011). Many studies have begun to document non-native and invasive plant species within the ABZ (Kent et al 2018, Wasowicz et al 2020, Leostrin and Pergl 2021, and show increasing establishment of these species following disturbances like fire (e.g. narrowleaf hawksbeard, Crepis tectorum, Carlson et al 2008) or harvest (e.g. bull thistle, Cirisum vulgare, Randall and Rejmánek 1993). Increasing anthropogenic presence and activities such as oil and gas exploration and production will also increase invasion of non-native plants, particularly in the Arctic (Wasowicz et al 2020). Through rapid growth, shading, and altered nutrient cycling (especially for N 2 -fixing species) invasive plants can reduce growth of native plants, potentially leading to cascading impacts on biogeochemical cycling (Carlson et al 2008, Sanderson et al 2012. Though non-native and invasive species are gaining more attention in the ABZ, further studies are still needed to determine the potential pace of future colonization as well as how these species will interact with native flora in conjunction with climate change. Disturbances also interact with human society in fundamental and profound ways. Smoke from large fires in the ABZ can substantially reduce air quality (Trainor et al 2009, Johnson et al 2021, and fires themselves cause significant destruction of human property and resources (de Groot et al 2013, Thomas et al 2017. Many disturbances (e.g. insects, pathogens, windthrow, drought) reduce timber resources (Volney and Fleming 2000, Hennigar et al 2007, Anderegg et al 2012, Boucher et al 2018. Permafrost thaw and subsequent ground subsidence is hazardous for travel and can damage critical infrastructure (e.g. roads, airports, homes), with impacts across Alaska  . Though most of these disturbances are natural and integral components of the ABZ system, anthropogenic climate change is pushing their extent, frequency, and severity outside of historical regimes. Continued study and data acquisition is crucial for projecting the future magnitude and direction of these disturbance trajectories and how they may interact (table 2).

Data availability statement
The data that support the findings of this study are available upon reasonable request from the authors. We thank Kelcy Kent for providing the drawing of ice wedge degradation in figure 10