Get Our e-AlertsSubmit Manuscript
Journal of Remote Sensing / 2022 / Article

Research Article | Open Access

Volume 2022 |Article ID 9765087 |

Yating Chen, Aobo Liu, Xiao Cheng, "Landsat-Based Monitoring of Landscape Dynamics in Arctic Permafrost Region", Journal of Remote Sensing, vol. 2022, Article ID 9765087, 17 pages, 2022.

Landsat-Based Monitoring of Landscape Dynamics in Arctic Permafrost Region

Received07 Dec 2021
Accepted06 Apr 2022
Published29 Apr 2022


Ice-rich permafrost thaws as a result of Arctic warming, and the land surface collapses to form characteristic thermokarst landscapes. Thermokarst landscapes can bring instability to the permafrost layer, affecting regional geomorphology, hydrology, and ecology and may further lead to permafrost degradation and greenhouse gas emissions. Field observations in permafrost regions are often limited, while satellite imagery provides a valuable record of land surface dynamics. Currently, continuous monitoring of regional-scale thermokarst landscape dynamics and disturbances remains a challenging task. In this study, we combined the Theil–Sen estimator with the LandTrendr algorithm to create a process flow for monitoring thermokarst landscape dynamics in Arctic permafrost region on the Google Earth Engine platform. A robust linear trend analysis of the Landsat Tasseled Cap index time series based on the Theil–Sen estimator and Mann–Kendall test showed the overall trends in greenness, wetness, and brightness in northern Alaska over the past 20 years. Six types of disturbances that occur in thermokarst landscape were demonstrated and highlighted, including long-term processes (thermokarst lake expansion, shoreline retreat, and river erosion) and short-term events (thermokarst lake drainage, wildfires, and abrupt vegetation change). These disturbances are widespread throughout the Arctic permafrost region and represent hotspots of abrupt permafrost thaw in a warming context, which would destabilize fragile thermokarst landscapes rich in soil organic carbon and affect the ecological carbon balance. The cases we present provide a basis for understanding and quantifying specific disturbance analyses that will facilitate the integration of thermokarst processes into climate models.

1. Introduction

The latest Arctic Monitoring and Assessment Program report [1] shows that the Arctic is warming three times faster than the global average for the period 1979-2019. Rapid Arctic warming is reducing the stability of near-surface permafrost, as model simulations indicate that every 1°C increase in temperature will result in the thawing of approximately 4.0 million km2 of permafrost [2] and the release of 14 to 19 Pg C [3] as carbon dioxide (CO2) and methane (CH4). Greenhouse gas emissions from thawing permafrost will contribute to climate feedbacks [4, 5], reduce the permissible anthropogenic carbon budget to meet the Paris Agreement targets [6], and cause socioeconomic losses in the order of trillions of dollars [3, 7]. In the latest generation of Earth system models, simulations of permafrost degradation and permafrost carbon-climate feedbacks remain highly uncertain due to poor representations of permafrost disturbances and especially thermokarst processes [8, 9].

Thermokarst is the process of land surface subsidence, collapse, and erosion due to rapid thawing of excess ground ice in permafrost [10, 11]. The consequence of thermokarst processes is the formation of three characteristic landscapes, namely, lake, wetland, and hillslope thermokarst landscapes [12]. These thermokarst landscapes, although covering only 20% of the northern permafrost region, contain approximately half of the subsurface organic carbon [12], as they were formed following the degradation of the organic-rich Yedoma deposits [13]. Thermokarst landscapes represent major sources of permafrost instability, affecting regional hydrologic and ecological conditions and potentially leading to further permafrost degradation and greenhouse gas emissions [14]. Therefore, monitoring of disturbances in thermokarst region is necessary to better estimate landscape-scale climate change impacts.

Currently, many local-scale hydrologic, ecological, and geomorphic disturbances remain unnoticed due to the remoteness and vast size of the Arctic [15]. Satellite-based monitoring strategies are becoming increasingly important for understanding the disturbances and dynamics of thermokarst landscapes, as the rapidly growing abundance and quality of satellite imagery in contrast to the paucity of ground-based data [16]. In the last decade, remote sensing has been widely applied to monitor changes in permafrost ecosystems, including surface temperature [17], seasonal thaw depth [18], snow cover [19], topography [20], surface and subsurface hydrology [21], and land cover [22]. Remote sensing techniques have enabled the detection and quantification of many disturbance processes, such as formation and drainage of thermokarst lakes and wetlands [23], retrogressive thaw slumps and active layer detachment slides on hillsides [24], thermal erosion trenches, thermokarst pits and troughs associated with degraded ice wedges [25], permafrost degradation due to coastal and fluvial erosion [26, 27], and wildfires [28]. Continuous Landsat Earth observation data have enabled the research community to gain insight into Arctic ecological and hydrological dynamics. Dynamic analysis of thermokarst lakes in four continental-scale transects indicates a 2.53% increase and 3.98% decrease in thermokarst lake area, with a net loss of 1.44% from 1999 to 2014 [16]. Quantitative analysis of the greenness of the Arctic tundra reveals that from 1985 to 2016, 37.3% of the tundra zone showed greening and 4.7% showed browning [29]. Vegetation growing on moist and nutrient-rich thermokarst lake drainage basins is about 25% greener than the peripheral areas [30] and may be an important contributor to Arctic greening [31].

With the growing archive of earth observation data and increased computational processing power, studies of local-scale disturbances in permafrost regions have shifted from single, widely spaced observations to continuous dynamic monitoring [15, 16, 23, 32]. These studies are mainly based on continuous Landsat time series data and trend analysis is performed using the Theil–Sen estimator and multispectral indices (MSI) such as Landsat Tasseled Cap (TC) index. The Theil–Sen estimator provides robust calculations of linear trends that reflect the overall trend pixel-by-pixel over the entire study period. However, for rapid or abrupt processes, such as thermokarst lake drainage events, linear trends do not reflect the onset and duration of disturbances and may lead to erroneous interpretations. To further improve the understanding and quantification of disturbances in thermokarst regions, we introduced the LandTrendr (Landsat-based detection of Trends in Disturbance and Recovery) temporal segmentation and change detection algorithm [33, 34] into permafrost disturbance studies. The LandTrendr algorithm was originally developed specifically for forestry applications, but it takes full advantage of the high-frequency multitemporal analysis capabilities of the Landsat archive and thus has also been successfully applied to disturbance detection on cropland [35], impervious surface [36], and waterlogged areas [37].

In this study, we used Landsat time-series data to continuously monitor disturbances over approximately 76,000 km2 of permafrost regions in northern Alaska for the period 2000 to 2020. The Theil–Sen estimator and the LandTrendr algorithm were combined to assess surface changes in the northern Alaska permafrost region and to help understand and quantify thermokarst landscape dynamics. Furthermore, we identified and highlighted six types of key disturbances detected at different temporal and spatial scales, including thermokarst lake expansion and drainage, fluvial erosion, coastline retreat, wildfire, and abrupt vegetation change. These disturbances were chosen because they are prevalent in Arctic thermokarst landscapes and are associated with abrupt permafrost thaw, which may help explore nonlinear transitions in the permafrost carbon feedback. All analyses were conducted based on annual image composites (2000-2020) generated by the Google Earth Engine (GEE) cloud computing platform [38].

2. Materials and Methods

2.1. Study Area

Northern Alaska (Figure 1) was chosen as the study area for two reasons: (1) it is scattered with tens of thousands of lakes and is typical of lake thermokarst landscapes [12]. (2) It has the highest carbon storage of the entire Alaska which means the occurrence of disturbances will have a greater impact on permafrost carbon release [39]. The total study area is approximately 76,000 km2, with a latitude of 69° N–71° N and a longitude of 145° W–163° W. Along the coastline to the inland, the topography of the study area gradually rises, forming a three-stage terrace (Figure 1). The entire study area can be divided into three subregions: the northernmost coastal plain is the first terrace, with elevations of less than 10 m; the foreland basin is the second terrace, with elevations of 10 to 50 m; and the southernmost terrace is the foothill of the Brooks Range, with elevations greater than 50 m.

2.2. Satellite Data

The Landsat image archive has the advantages of free access, continuous observation, and relatively high spatial and temporal resolution. The Landsat image archive is unique in that it has the advantage of high spatial resolution compared to MODIS and a much longer span of record compared to Sentinel-2. A total of 4,592 available Level-2 Landsat surface reflectance (SR) products were searched between June and September, of which 2,088 scenes met the cloud coverage filtering criteria (below 50%). Image availability in the study area has improved over time, with the number of available image tiles increasing from a minimum of 28 in 2003 to a maximum of 149 in 2019 (Figure 2). It is noted that Landsat-5 data acquisition in Alaska encountered limited mission and data relay problems during 2000-2005 [40]. In addition, Landsat-7 imagery has been affected by a scan line corrector failure (SLC-off) since 2003. For years, with limited image availability and poor data quality, the annual composite images may have gaps and be susceptible to clouds and shadows. To make the trend analysis less affected by data availability, we ensure the overall consistency of image quality by median composite and use the robust Theil–Sen regression method with the Mann–Kendall test to demonstrate long-term trends in the study area.

2.3. Method

To detect thermokarst landscape disturbances from Landsat images, the processing chain consists of several steps that can be grouped into image preprocessing, creation of annual image composites, trend analysis, disturbance detection, and validation (Figure 3).

2.3.1. Image Processing

Annual composite images (2000-2020) were generated by filtering all archived Landsat images available on the GEE platform for cloud coverage below 50% and acquisition months of June through September. Given the spectral differences between the Landsat-5 Thematic Mapper (TM), Landsat-7 Enhanced Thematic Mapper Plus (ETM+), and Landsat-8 Operational Land Imager (OLI) sensors, a statistical conversion function [41] was used to calibrate the spectral reflectance of the OLI data to reconcile it with that of the TM and ETM+. After obtaining the filtered Landsat time series dataset, noisy observations such as clouds, snow, and shadows were masked based on the quality assessment (pixel_qa) band of the Landsat SR product [42]. For data gaps due to lack of suitable observations and Landsat-7 SLC-off, we used unobscured portions of images from adjacent years to fill these gaps for complete coverage. We then used the median reducer to reduce the Landsat image collections to annual spectral-temporal image composites (2000-2020).

Based on the six common spectral bands of OLI, ETM+ and TM (blue, green, red, NIR, SWIR1, and SWIR2), we calculated six MSI for each pixel that reflect various surface characteristics such as vegetation status, soil moisture, and surface brightness, including three tasseled cap components [43, 44], normalized difference vegetation index (NDVI) [45], normalized different moisture index (NDMI) [46], and automatic water extraction index (AWEI) [47]. The six MSI are calculated as shown in equations (1) to (6). Among the three tasseled cap indices, tasseled cap greenness (TCG) reflects the vegetation spectral information; tasseled cap wetness (TCW) reflects the moisture information of the feature; and tasseled cap brightness (TCB) reflects soil spectral information [43, 44].

Among them, , , , NIR, SWIR1, and SWIR2 represent the blue, green, red, near-infrared, short-wave infrared 1, and short-wave infrared 2 bands of the Landsat images, respectively. We use AWEInsh because the study area is flat, where the subscript “nsh” indicates that the index is suitable for situations where shadows are not a major issue [47].

2.3.2. Theil–Sen Estimator

After obtaining the annual image composites and calculating the MSI, the Theil–Sen estimator and the LandTrendr algorithm were used for the trend analysis. The Theil–Sen estimator is a robust nonparametric regression method that calculates all pairwise slopes for a given MSI through time in pixel units and returns the median slope [48]. Compared with least squares regression, the Theil–Sen regression is insensitive to outliers and is therefore widely used in remote sensing image time series analysis [15, 32]. The slope and intercept of each MSI and each pixel, calculated from the Theil–Sen estimator, can reflect the overall trends in water, vegetation, soil moisture, and albedo in the study area.

2.3.3. LandTrendr Algorithm

The LandTrendr is a temporal segmentation and change detection algorithm that extracts spectral trajectories of land surface changes from annual Landsat time-series stacks and fits a point-to-point multisegment regression [34]. The Landtrendr algorithm first determines a trajectory model with maximum complexity by removing spikes, identifying potential vertices, and fitting the trajectory. This trajectory model then undergoes iterative simplification, where a series of control parameters are used to reduce overfitting due to arbitrary segmentation and to filter out noise-induced variation. The LandTrendr algorithm has the advantage of capturing both short-time abrupt changes and smoothing long-time gradual changes, and thus, the detected changes are not simply a comparison of two time periods, but a continuous process at different spatial and temporal scales. The LandTrendr algorithm does not include stripe filling processing for Landsat 7 SLC-off images, which may seriously affect the detection of disturbance signals, while the image preprocessing step incorporated in this study can improve the accuracy of the disturbance detection.

2.3.4. Disturbance Detection

The output of the Theil–Sen estimator is the slope and intercept values for each pixel based on the MSI trajectories calculated for the period 2000–2020. The main output of the LandTrendr algorithm, in addition to the fitted trajectories, includes a disturbance timing map showing the year in which the detected disturbance occurred, a disturbance magnitude map, and a map of the duration of events in years. Six types of disturbances or processes were identified and highlighted in the study area, namely, thermokarst lake expansion and drainage, fluvial erosion, coastline retreat, wildfires and abrupt vegetation change. Of these, thermokarst lake expansion, fluvial erosion, and coastline retreat are long-duration processes, while thermokarst lake drainage, wildfires and abrupt vegetation change are short-duration events. The Theil–Sen estimator was able to fit long-duration processes well, but could not accurately fit the loss–recovery process of vegetation or water bodies over a short period of time, nor could the year in which the disturbance occurred be determined [15]. In contrast, the LandTrendr algorithm was able to accurately detect sharp changes that occur over a year or years, but is less effective for slow-onset changes, which are often masked by noise in the disturbance magnitude map [37, 49]. Therefore, a combination of these two methods was used to detect and quantify disturbances in northern Alaska thermokarst region over the past 20 years.

2.3.5. Validation

For the trend analysis and disturbance detection results of the Theil–Sen estimator and LandTrendr algorithm, we used the accompanying Mann–Kendall test [50] and TimeSync tool [51] for validation. Unlike the information presented by the Theil–Sen trend slope, the Mann–Kendall statistical test for trend does not assess the magnitude of the change, but rather whether the trend of the data change is statistically significant. The Mann–Kendall test has been used for significance assessment in many Theil–Sen slope based trend analysis studies [23, 50, 52]. In this study, we filtered out the nonsignificant trends in the slopes calculated by the Theil–Sen estimator using a threshold of value 0.05.

The TimeSync tool is recommended by the authors of the LandTrendr algorithm for accuracy evaluation and is a Landsat image-based time series visualization and data collection tool that accurately detects subtle disturbances [51]. TimeSync tool provides us with image sequences and spectral trajectories of the area of interest (i.e., disturbed plot and its neighbors), thus allowing simultaneous validation of the pixel-scale boundaries of the disturbance and the year in which the disturbance occurred. The accuracy evaluation of the LandTrendr algorithm varies depending on the application target and the object of study, with some focusing on the accuracy of pixel recognition and others on the accuracy of disturbance year recognition. The pixel-scale accuracy assessment of thermokarst lake dynamics in northern Alaska by the TimeSync visual interpretation tool showed an overall accuracy of 0.95 for lake dynamics detection, 0.92 for user accuracy, and 0.82 for producer accuracy [53]. The accuracy of disturbance detection for lake drainage and subsequent vegetation growth evaluated by the TimeSync tool showed that the LandTrendr algorithm was able to accurately identify the year in which the disturbance occurred and locate the adjacent year of the disturbance even in the presence of data gaps and noise [30].

3. Results

3.1. Regional Scale Changes

The time series changes of six MSI across the study area were analyzed based on the trend slopes obtained from the Theil–Sen estimator (Figures 4 and 5). The six MSI (AWEI, NDVI, NDMI, TCW, TCG, and TCB) represent a range of surface physical attributes such as vegetation, water, soil moisture, and albedo, whose variability reflects changes at multiple scales ranging from local disturbances to regional processes.

3.1.1. Greening

As the trend slopes of the vegetation indices TCG and NDVI indicate (Table 1; Figure 4), northern Alaska shows a clear trend towards greening over the last two decades. For the overall region, the calculated decadal slope change for TCG is , compared to for NDVI (Table 1). The fitted trajectories show an average increase in NDVI of 0.07 and TCG of 0.25 from 2000 to 2020. Differences between subregions further reveal the diversity of local greening trends (Table 1; Figure 5), with the first terrace (coastal plain) showing the least increase in greening, the second terrace (foreland basin) being close to the regional average, and the third terrace (foothill) showing the strongest greening trend. The coastal plain has the highest latitude and lowest mean temperature and is dominated by perennial herbaceous and woody wetland vegetation that is influenced by the water table at or near the surface over extensive periods of time. The foreland basin is characterized by an abundance of lakes in the tundra that are wrapped by surrounding wetlands, and the extension of the wetland is grassland-lichen-moss. The foothill area is dominated by dwarf shrubs and grasslands with lichen and moss. The differences in vegetation composition between these three terraces are also reflected in the intercept values (Figure 4), with plants growing in the foothill zone being relatively lusher, taller, and, therefore, greener. Thus, the greening trend in northern Alaska may be due to the invasion of taller species.

Study area and subregionsNDVINDMIAWEITCGTCWTCB

Northern Alaska0.0410.0390.0290.046-0.0010.0710.1320.0990.0260.1230.0230.186
Coastal plain0.0360.0550.0130.068-0.0010.0860.0910.0970.0140.1470.0040.268
Foreland basin0.0440.0380.0340.0430.0020.0730.1340.0990.0310.1300.0290.171

3.1.2. Wetting

The statistics of decadal trend slopes reveal that TCW and NDMI have shown an overall upward trend over the past 20 years, while AWEI has changed very little (Figure 5). This is because AWEI is sensitive only to water, NDMI is sensitive to changes in soil and vegetation moisture, and the sensitivity of TCW is between these factors. The wetting trend reflected by NDMI is somewhat similar to the greening trend reflected by the TCG. In summary, northern Alaska has generally exhibited a wetting trend over the past 20 years, but this trend has been driven primarily by vegetation changes rather than lake dynamics. The intercept of TCW shows that the northern coastal areas are the wettest, while the surface is generally drier in the foothill area. Spatially, the wetting trend correlates well with the observed increase in greenness. The foreland basin and hillslope thermokarst regions exhibit the most significant wetting trends, with rates of change in both TCW and NDMI approaching 0.03 per decade. The mean trend slope of the foothill is close to the regional mean and has the smallest variance, indicating the overall homogeneity of wetting in this region. The coastal plain has the smallest mean trend slope but the largest variance, which may be due to the thermokarst lake expansion and drainage events that have occurred.

3.1.3. Brightening

TCB exhibits a weak positive trend over the majority of the study area between 2000 and 2020, but with a very large spatial variation compared to other MSI, resulting in a significantly greater variance than the mean values in the decadal trend slope statistics (Table 1). The intercept of TCB decreases from inland to the coast (Figure 4), with higher brightness values in areas with lush plants or steep terrain such as the foothill and hillslope thermokarst regions. Changes in brightness or albedo are mainly influenced by vegetation type, coverage, and soil moisture. Of the three subregions, the TCB at the foreland basin and foothill are close to the regional mean, and the coastal plain shows the least significant change. It is noteworthy that an area in the southeastern part of the study area experienced a substantial change in brightness due to a wildfire.

3.2. Detection of Local Disturbances

With a spatial resolution of 30 m, Landsat time series data provide excellent detection of local-scale disturbances and changes in features associated with these disturbances. In this section, six typical disturbances detected in northern Alaska thermokarst region are highlighted, including thermokarst lake expansion and drainage, river dynamics, coastal erosion, wildfires, and abrupt vegetation change (Figure 1). All detected disturbed plots were examined with the TimeSync tool [51].

3.2.1. Thermokarst Lake Expansion

The expansion of thermokarst lakes is a typical process in permafrost landscapes involving thermal and mechanical erosion of shore bluffs containing ice-rich permafrost [54], and it has been observed in most of the thermokarst lakes of all three terraces. Unlike abrupt drainage, thermokarst lake expansion is a very slow process, with typical expansion rates ranging from tens of centimeters to a few meters per year [55]. In the RGB composite map of tasseled cap index trend slopes (TCB–TCG–TCW) derived from the Theil–Sen regression, the typical thermokarst lake expansion area is shown as a blue ring around the lake (Figure 6(a)). There are also a few cases where lakes expand mainly in one direction due to wind, topography, an adjacent lake, or other impact factors. The lake (Figure 6) is located in the foreland basin, surrounded by a single vegetation type, flat terrain, and no adjacent lakes, and its expansion rate is approximately 1–3 m per year. During lake expansion, the surface cover underwent a transition from vegetated tundra surface to water, as evidenced by the trend slope changes of six MSI (Figure 6(b)). In the lake expansion area, the water-sensitive TCW and AWEI exhibit a strong positive trend, while the moisture-sensitive NDMI only shows a weak positive trend. The TCB index shows a strong negative trend due to the much lower albedo of the water body compared to that of the tundra. Moreover, the loss of vegetation causes TCG to show a strong negative trend. In the tundra surrounding the lake, all trend slopes demonstrate a spatially consistent pattern, with values close to zero.

3.2.2. Thermokarst Lake Drainage Events

A number of factors may influence the occurrence of thermokarst lake drainage events, including lake depth and density, lake vegetation, distance from shore and river, elevation, topography, and precipitation [56]. From the detected thermokarst lake drainage events, two examples were selected to represent complete drainage (Figure 7) and partial, progressive drainage (Figure 8). Remote sensing images show that a thermokarst lake (~1.5 km2) located in the foothill area was completely drained between 2010 and 2011, and the surface changed from water to bare lake bed and then to gradual vegetation cover (Figures 7(a)–7(c)). Because TCW shows a negative trend while TCB and TCG show a positive trend, the thermokarst lake drainage event appears bright yellow in the RGB composite (TCB–TCG–TCW) map of the slopes of the tasseled cap indices derived from the Theil-Sen regression (Figure 7(d)). However, for each specific lake drainage event, the Theil-Sen regression only represents rough linear changes in the overall trend and cannot be used to detect the year of thermokarst lake drainage event (Figures 7(g)–7(i)). In contrast, the segmented trajectory of the LandTrendr algorithm provides a better fit and accurately detects the year, magnitude, and duration of the disturbance. Figure 8 shows a lake that has undergone several drainage events since 2014 and remains only partially dried. The LandTrendr algorithm shows pixel-level detection capability beyond the level of visual interpretation, accurately detecting the year in which drainage occurred at the edge of the lake.

3.2.3. Coastal Erosion

Arctic coasts are some of the fastest changing coasts on Earth. Since the early 2000s, erosion rates along the Arctic permafrost coastlines have more than doubled compared to the second half of the 20th century [57], contributing to the transformation of thermokarst lakes into lagoons [58]. Northern Alaska has an extensive and varied coastline, along which coastal erosion, marine flooding, and river deltas can be commonly found. Here, the most representative coastal erosion phenomenon was selected for the analysis. The sample is located north of Lake Narukluk (Figure 9). It is the most severe area of coastal erosion, with an annual erosion rate of approximately 10–30 m. The spatial extent of the inundation zone is highly variable due to the very flat topography, with only slight undulations. The northwestern coastline of the study area retreated significantly less than the northeastern coastline, probably due to the lagoons acting as effective barriers that prevented typhoon surges from eroding and scouring the coast [59]. In the RGB composite map of the TC trend slopes, the coastal erosion zone is shown in blue because the positive trend slope of the TCW index is the dominant feature. During the transition from tundra to seawater, the water index increases, while the vegetation and brightness indices show decreasing trends, which is consistent with the process of lake expansion. Compared with the linear fitting results of the Theil–Sen regression, the trajectories fitted by the LandTrendr algorithm were divided into multiple segments that can more accurately represent the transition process from tundra to seawater. Figure 9(d) demonstrates that this transition process began in approximately 2004 and ended in 2010, after which the values of the indices stabilized.

3.2.4. River Dynamics

A variety of fluvial processes has been observed in the study area, including river erosion, accumulation and movement of sandbars, and changes in river morphology. Figure 10 shows an example of how a section of the Colville River has changed over the last 20 years, presenting both the effects of river erosion and sandy gravel accumulation. In the RGB composite map of tasseled cap trend slopes, the blue color along both sides of the river represents the transition from tundra to water under the effects of river scouring and the orange color in the center of the river represents the transition from water to sandbar due to sediment accumulation. Based on the transects of trend slopes (Figure 10(d)), the erosion and sedimentary zones of the river can be clearly identified, as well as rivers and riverbanks that have not changed significantly. In addition, the fitted temporal profiles (Figures 10(e) and 10(f)) illustrate that TCW and TCB show significant decreasing and increasing trends in the sandbar deposition zone, respectively, while the opposite but weaker trend is observed in the river erosion zone.

3.2.5. Wildfire Events

Under the influence of climate change, wildfires are occurring with unprecedented frequency in the Arctic [60]. Wildfires may greatly accelerate thermokarst processes, propel thermokarst landscape reorganization, and exacerbate regional carbon release [61]. Between 2000 and 2020, in the southeastern part of the study area, significant changes in brightness and vegetation indices were detected in the area near the Anaktuvuk River, indicating the occurrence of wildfire events. It was found from the disturbance maps developed by the LandTrendr (Figure 11) that the fires occurred between 2007 and 2008. Magnitude of disturbance (Figure 11(e)) shows the uneven distribution of fire intensity, with a weaker fire in wetlands around lakes and more intense in shrub tundra. Most areas did not regain their prefire vegetation levels until five years later (Figure 11(f)). Severe burning removed vegetation and overlying soil organic matter, and this loss of insulation layer resulted in higher ground temperatures, increased active layer thickness, and accelerated development of the thermokarst landscapes.

3.2.6. Abrupt Vegetation Changes

Vegetation cover plays an important role in protecting permafrost from degradation [61]. The Theil-Sen slopes captured the overall regional greening trend, while the LandTrendr detection showed that greening hotspots were primarily located in the thermokarst lake drainage basins and postfire areas. Here, we found that the abrupt vegetation changes in northern Alaska were primarily due to vegetation loss and recovery from wildfire (Figure 11) and rapid vegetation growth following thermokarst lake drainage (Figure 12).

In the fire area, rapid vegetation loss and recovery was observed, with NDVI trajectories fitted by the LandTrendr showing roughly four stages of change (Figure 11(g)). Before the fire in 2007, vegetation in the absence of disturbance remained stable. The vegetation index declined sharply after the fire but started to recover rapidly in the following year until it reached a steady state again. It is noteworthy that the region reached higher levels of greenness after 2013 than before the fires. Similar trends can be seen in the trajectories of other MSI. In the thermokarst lake drainage basin, NDVI trends fitted by the LandTrendr algorithm (Figure 12) showed that tundra vegetation invaded the lake basin and grew luxuriantly over time after the lake was completely drained in 2014. It was also noted that the vegetation grew more luxuriantly in the drained lake basin than the surrounding area (Figure 12). This may be due to the relatively moist and nutrient-rich sediments on the lake bottom that favor the development of highly productive plant communities [30]. In addition, the exposed lake bed was colonized by plants, which helped to stabilize the permafrost and accelerate soil recovery [30].

4. Discussion and Conclusion

In this study, we combined the Theil–Sen estimator and the LandTrendr algorithm to detect and analyze landscape dynamics in a thermokarst lake area of approximately 76,000 km2 in northern Alaska. Compared to existing studies on trend analysis and disturbance detection based on Landsat time series images, this study demonstrates the complementary advantages of the Theil–Sen estimator, which provides detection of long-term disturbances, and the LandTrendr algorithm, which provides specific timing, magnitude, and duration of short-term disturbances. For relatively slow processes such as localized erosion occurring near lakes, rivers, and shorelines, the LandTrendr algorithm may provide detection results with errors due to image quality, whereas the noise-resistant Theil-Sen estimator provides robust, median-unbiased results. For abrupt events such as thermokarst lake drainage and wildfires, the Theil–Sen estimator can only provide a rough fit, while the LandTrendr algorithm can determine the timing and course of the event. Robust linear trend analyses of several MSI (tasseled cap indices, NDVI, NDMI, and AWEI) showed a significant greening trend and weaker wetting and brightening trends in northern Alaska during the last 20 years. Differences in greenness, wetness, and brightness trends were observed in different subregions and were related to coastal distance, vegetation type, topography, and climatic conditions. In addition, we analyzed six typical disturbances prevalent in the thermokarst region, including thermokarst lake expansion and drainage, shoreline retreat, river erosion, wildfire, and abrupt vegetation change. Among them, wildfire, shoreline retreat, and river erosion lead to a decrease in permafrost stability and trigger further development of the thermokarst landscape, promoting thermokarst lake expansion and drainage [16, 57, 58, 62]. During the formation and expansion of thermokarst lakes, anaerobic microorganisms at the bottom of the lake decompose thawed soil organic matter, releasing methane bubbles into the atmosphere and exacerbating climate change [14]. Lake drainage occurs as a result of overtopping of the lake shore or melting of the ice wedge, creating a lateral or bottom drainage channel [55]. Following thermokarst lake drainage, the organic-rich lake bed sediments are exposed and refreeze and are subsequently covered by vegetation. This process will contribute to the recovery of permafrost, stabilize soil carbon, and partially offset the release of carbon from thawing permafrost due to the biomass accumulated by vegetation [30, 31]. Overall, these disturbances represent important anomalies in the permafrost region and provide valuable information for regional studies of geomorphology, hydrology, ecology, and carbon balance, as well as for the selection of field sites for costly fieldwork and validation activities.

Abrupt thawing occurs in less than 20% of the permafrost region, yet may increase carbon release from permafrost soil by about 50% [12]. Permafrost degradation in existing models is usually modeled as a top-down gradual process, ignoring the nonlinear abrupt thawing process caused by disturbances, thus largely underestimating the carbon emissions from permafrost degradation [63]. The current research community has limited knowledge of the expansion and drainage processes of thermokarst lakes and the dynamics of vegetation after lake drainage [14]. The remote sensing monitoring approach demonstrated in this study will help to obtain observations and parameters for modeling thermokarst processes, and its superiority has been demonstrated in monitoring thermokarst lake drainage [53] and vegetation dynamics [30]. This study presents regional-scale analysis based on Landsat time-series imagery, but our approach is highly automated and scalable and can be applied to other regions and remote sensing data sources as needed. For example, the detection of landscape disturbances of thermokarst lakes across the circum-Arctic region will reveal spatial heterogeneity in thermokarst lake dynamics, allowing for deeper mechanistic analysis. Quantifying the vegetation dynamics of drainage lake basins in the Arctic will help explain key issues in Arctic greening and carbon cycle studies.

It should be noted that much work remains to be done to improve the accuracy of detection of complex permafrost disturbances, including improving the robustness of trend analysis algorithms and combining multisource remote sensing data to improve the availability of image data. The Theil-Sen estimator can accurately respond to changes in vegetation and moisture in a region but is also susceptible to high-frequency changes when detecting long-term disturbances. Despite powerful temporal segmentation and change detection capabilities, LandTrendr’s detection may have extra errors for the first and last year of the study period [64], which is a common problem with many time series analysis methods. In addition, when disturbances occur in years with missing data, the LandTrendr detects disturbances a year later than they actually are, thereby reducing the number of disturbances in years with gaps and increasing the number of disturbances in subsequent years [65]. The continuity and availability of Landsat imagery is a key factor to consider for the analysis of specific disturbance events prior to the launch of the Landsat-8 satellite in 2013. In contrast, with the launch of Landsat-9 satellite in 2021, the temporal density of Landsat Earth observation data increases, which is expected to reveal more details of the disturbance process [66]. In addition, the Copernicus Sentinel-2 mission provides comparable multispectral data with higher spatial and temporal resolution, which is expected to further improve the observation density and monitoring capabilities in high-latitude permafrost regions as its image archive continues to accumulate [67].

Conflicts of Interest

The authors declare no competing interests.

Authors’ Contributions

Y.C. and X.C. conceived and designed the experiment. Y.C and A.L. performed the experiments and data analysis. All authors contributed to the discussion and writing of the manuscript.


This research was funded by the National Outstanding Youth Grant (#41925027) and the National Key Research and Development Program of China (Grant 2019YFC1509104).


  1. AMAP, Arctic Climate Change Update 2021: Key Trends and Impacts. Summary for Policy-Makers, Arctic Monitoring and Assessment Programme (AMAP), 2021.
  2. S. E. Chadburn, E. J. Burke, P. M. Cox, P. Friedlingstein, G. Hugelius, and S. Westermann, “An observation-based constraint on permafrost loss as a function of global warming,” Nature Climate Change, vol. 7, no. 5, pp. 340–344, 2017. View at: Publisher Site | Google Scholar
  3. Y. Chen, A. Liu, and J. C. Moore, “Mitigation of Arctic permafrost carbon loss through stratospheric aerosol geoengineering,” Nature Communications, vol. 11, no. 1, 2020. View at: Publisher Site | Google Scholar
  4. E. A. G. Schuur, A. D. McGuire, C. Schädel et al., “Climate change and the permafrost carbon feedback,” Nature, vol. 520, no. 7546, pp. 171–179, 2015. View at: Publisher Site | Google Scholar
  5. C. Mu, B. W. Abbott, A. J. Norris et al., “The status and stability of permafrost carbon on the Tibetan Plateau,” Earth Science Reviews, vol. 211, article 103433, 2020. View at: Publisher Site | Google Scholar
  6. E. Comyn-Platt, G. Hayman, C. Huntingford et al., “Carbon budgets for 1.5 and 2 °C targets lowered by natural wetland and permafrost feedbacks,” Nature Geoscience, vol. 11, no. 8, pp. 568–573, 2018. View at: Publisher Site | Google Scholar
  7. Y. Chen, A. Liu, Z. Zhang, C. Hope, M. James, and C. Crabbe, “Economic losses of carbon emissions from circum-Arctic permafrost regions under RCP-SSP scenarios,” Science of the Total Environment, vol. 658, pp. 1064–1068, 2019. View at: Publisher Site | Google Scholar
  8. E. J. Burke, Y. Zhang, and G. Krinner, “Evaluating permafrost physics in the Coupled Model Intercomparison Project 6 (CMIP6) models and their sensitivity to climate change,” The Cryosphere, vol. 14, no. 9, pp. 3155–3174, 2020. View at: Publisher Site | Google Scholar
  9. J. C. Moore, I. Mettiäinen, M. Wolovick et al., “Targeted geoengineering: local interventions with global implications,” Global Policy, vol. 12, Supplement 1, pp. 108–118, 2021. View at: Publisher Site | Google Scholar
  10. S. V. Kokelj and M. T. Jorgenson, “Advances in thermokarst research,” Permafrost and Periglacial Processes, vol. 24, no. 2, pp. 108–119, 2013. View at: Publisher Site | Google Scholar
  11. H. Michiel, in 't Zandt, S. Liebner, and C. U. Welte, “Roles of thermokarst lakes in a warming world,” Trends in Microbiology, vol. 28, no. 9, pp. 769–779, 2020. View at: Publisher Site | Google Scholar
  12. D. Olefeldt, S. Goswami, G. Grosse et al., “Circumpolar distribution and carbon storage of thermokarst landscapes,” Nature Communications, vol. 7, no. 1, 2016. View at: Publisher Site | Google Scholar
  13. T. Schneider Von Deimling, G. Grosse, J. Strauss et al., “Observation-based modelling of permafrost carbon fluxes with accounting for deep carbon deposits and thermokarst activity,” Biogeosciences, vol. 12, no. 11, pp. 3469–3488, 2015. View at: Publisher Site | Google Scholar
  14. K. W. Anthony, T. S. von Deimling, I. Nitze et al., “21st-century modeled permafrost carbon emissions accelerated by abrupt thaw beneath lakes,” Nature Communications, vol. 9, no. 1, 2018. View at: Publisher Site | Google Scholar
  15. I. Nitze and G. Grosse, “Detection of landscape dynamics in the Arctic Lena Delta with temporally dense Landsat time-series stacks,” Remote Sensing of Environment, vol. 181, pp. 27–41, 2016. View at: Publisher Site | Google Scholar
  16. I. Nitze, G. Grosse, B. M. Jones, V. E. Romanovsky, and J. Boike, “Remote sensing quantifies widespread abundance of permafrost region disturbances across the Arctic and Subarctic,” Nature Communications, vol. 9, no. 1, 2018. View at: Publisher Site | Google Scholar
  17. M. Langer, S. Westermann, M. Heikenfeld, W. Dorn, and J. Boike, “Satellite-based modeling of permafrost temperatures in a tundra lowland landscape,” Remote Sensing of Environment, vol. 135, pp. 12–24, 2013. View at: Publisher Site | Google Scholar
  18. J. E. Anderson, T. A. Douglas, R. A. Barbato, S. Saari, J. D. Edwards, and R. M. Jones, “Linking vegetation cover and seasonal thaw depths in interior Alaska permafrost terrains using remote sensing,” Remote Sensing of Environment, vol. 233, article 111363, 2019. View at: Publisher Site | Google Scholar
  19. H. Bergstedt, S. Zwieback, A. Bartsch, and M. Leibman, “Dependence of C-band backscatter on ground temperature, air temperature and snow depth in Arctic permafrost regions,” Remote Sensing, vol. 10, no. 1, p. 142, 2018. View at: Publisher Site | Google Scholar
  20. J. van der Sluijs, S. V. Kokelj, R. H. Fraser, J. Tunnicliffe, and D. Lacelle, “Permafrost terrain dynamics and infrastructure impacts revealed by UAV photogrammetry and thermal imaging,” Remote Sensing, vol. 10, no. 11, p. 1734, 2018. View at: Publisher Site | Google Scholar
  21. D. M. Rey, M. Walvoord, B. Minsley, J. Rover, and K. Singha, “Investigating lake-area dynamics across a permafrost-thaw spectrum using airborne electromagnetic surveys and remote sensing time-series data in Yukon Flats, Alaska,” Environmental Research Letters, vol. 14, no. 2, article 025001, 2019. View at: Publisher Site | Google Scholar
  22. L. Wang, P. Marzahn, M. Bernier, and R. Ludwig, “Mapping permafrost landscape features using object-based image classification of multi-temporal SAR images,” ISPRS Journal of Photogrammetry and Remote Sensing, vol. 141, pp. 10–29, 2018. View at: Publisher Site | Google Scholar
  23. I. Olthof, R. H. Fraser, and C. Schmitt, “Landsat-based mapping of thermokarst lake dynamics on the Tuktoyaktuk Coastal Plain, Northwest Territories, Canada since 1985,” Remote Sensing of Environment, vol. 168, pp. 194–204, 2015. View at: Publisher Site | Google Scholar
  24. L. Huang, J. Luo, Z. Lin, F. Niu, and L. Liu, “Using deep learning to map retrogressive thaw slumps in the Beiluhe region (Tibetan Plateau) from CubeSat images,” Remote Sensing of Environment, vol. 237, article 111534, 2020. View at: Publisher Site | Google Scholar
  25. H. Gao, N. Nie, W. Zhang, and H. Chen, “Monitoring the spatial distribution and changes in permafrost with passive microwave remote sensing,” ISPRS Journal of Photogrammetry and Remote Sensing, vol. 170, pp. 142–155, 2020. View at: Publisher Site | Google Scholar
  26. F. Günther, P. P. Overduin, A. V. Sandakov, G. Grosse, and M. N. Grigoriev, “Short- and long-term thermo-erosion of ice-rich permafrost coasts in the Laptev Sea region,” Biogeosciences, vol. 10, no. 6, pp. 4297–4318, 2013. View at: Publisher Site | Google Scholar
  27. M. Fuchs, I. Nitze, J. Strauss et al., “Rapid fluvio-thermal erosion of a Yedoma permafrost cliff in the Lena River Delta,” Frontiers in Earth Science, vol. 8, 2020. View at: Publisher Site | Google Scholar
  28. C. M. Gibson, L. E. Chasmer, D. K. Thompson, W. L. Quinton, M. D. Flannigan, and D. Olefeldt, “Wildfire as a major driver of recent permafrost thaw in boreal peatlands,” Nature Communications, vol. 9, no. 1, 2018. View at: Publisher Site | Google Scholar
  29. R. Massey, P. Jantz, B. C. Forbes et al., “Summer warming explains widespread but not uniform greening in the Arctic tundra biome,” Nature Communications, vol. 11, no. 1, p. 4621, 2020. View at: Publisher Site | Google Scholar
  30. Y. Chen, A. Liu, and X. Cheng, “Vegetation grows more luxuriantly in Arctic permafrost drained lake basins,” Global Change Biology, vol. 27, no. 22, pp. 5865–5876, 2021. View at: Publisher Site | Google Scholar
  31. M. R. Turetsky, B. W. Abbott, M. C. Jones et al., “Carbon release through abrupt permafrost thaw,” Nature Geoscience, vol. 13, no. 2, pp. 138–143, 2020. View at: Publisher Site | Google Scholar
  32. T. Hermosilla, M. A. Wulder, J. C. White, N. C. Coops, P. D. Pickell, and D. K. Bolton, “Impact of time on interpretations of forest fragmentation: three-decades of fragmentation dynamics over Canada,” Remote Sensing of Environment, vol. 222, pp. 65–77, 2019. View at: Publisher Site | Google Scholar
  33. R. E. Kennedy, Z. Yang, and W. B. Cohen, “Detecting trends in forest disturbance and recovery using yearly Landsat time series: 1. LandTrendr -- temporal segmentation algorithms,” Remote Sensing of Environment, vol. 114, no. 12, pp. 2897–2910, 2010. View at: Publisher Site | Google Scholar
  34. R. E. Kennedy, Z. Yang, N. Gorelick et al., “Implementation of the LandTrendr algorithm on Google Earth Engine,” Remote Sensing, vol. 10, no. 5, p. 691, 2018. View at: Publisher Site | Google Scholar
  35. L. Zhu, X. Liu, W. Ling, Y. Tang, and Y. Meng, “Long-Term monitoring of cropland change near Dongting Lake, China, using the Landtrendr algorithm with Landsat imagery,” Remote Sensing, vol. 11, no. 10, p. 1234, 2019. View at: Publisher Site | Google Scholar
  36. B. Wang, A. Zhenjie Chen, X. Zhu, Y. Hao, and X. Changqing, “Multi-level classification based on trajectory features of time series for monitoring impervious surface expansions,” Remote Sensing, vol. 11, no. 6, p. 640, 2019. View at: Publisher Site | Google Scholar
  37. T. He, W. Xiao, Y. L. Zhao, X. Deng, and H. Zhenqi, “Identification of waterlogging in Eastern China induced by mining subsidence: a case study of Google Earth Engine time-series analysis applied to the Huainan coal field,” Remote Sensing of Environment, vol. 242, article 111742, 2020. View at: Publisher Site | Google Scholar
  38. N. Gorelick, M. Hancher, M. Dixon, S. Ilyushchenko, D. Thau, and R. Moore, “Google Earth Engine: planetary-scale geospatial analysis for everyone,” Remote Sensing of Environment, vol. 202, pp. 18–27, 2017. View at: Publisher Site | Google Scholar
  39. G. Hugelius, J. Strauss, S. Zubrzycki et al., “Estimated stocks of circumpolar permafrost carbon with quantified uncertainty ranges and identified data gaps,” Biogeosciences, vol. 11, no. 23, pp. 6573–6593, 2014. View at: Publisher Site | Google Scholar
  40. S. Goward, T. Arvidson, D. Williams, J. Faundeen, J. Irons, and S. Franks, “Historical record of Landsat global Coverage,” Photogrammetric Engineering and Remote Sensing, vol. 72, no. 10, pp. 1155–1169, 2006. View at: Publisher Site | Google Scholar
  41. D. P. Roy, V. Kovalskyy, H. K. Zhang et al., “Characterization of Landsat-7 to Landsat-8 reflective wavelength and normalized difference vegetation index continuity,” Remote Sensing of Environment, vol. 185, pp. 57–70, 2016. View at: Publisher Site | Google Scholar
  42. S. Foga, P. L. Scaramuzza, S. Guo et al., “Cloud detection algorithm comparison and validation for operational Landsat data products,” Remote Sensing of Environment, vol. 194, pp. 379–390, 2017. View at: Publisher Site | Google Scholar
  43. E. P. Crist, “A TM tasseled cap equivalent transformation for reflectance factor data,” Remote Sensing of Environment, vol. 17, no. 3, pp. 301–306, 1985. View at: Publisher Site | Google Scholar
  44. C. Huang, B. Wylie, L. Yang, C. Homer, and G. Zylstra, “Derivation of a tasselled cap transformation based on Landsat 7 at-satellite reflectance,” International Journal of Remote Sensing, vol. 23, no. 8, pp. 1741–1748, 2002. View at: Publisher Site | Google Scholar
  45. J. Jr. Rouse, R. H. Haas, J. A. Schell, and D. W. Deering, Monitoring vegetation systems in the great plains with ERTS, NASA Special Publication, 1974.
  46. E. H. Wilson and S. A. Sader, “Detection of forest harvest type using multiple dates of Landsat TM imagery,” Remote Sensing of Environment, vol. 80, no. 3, pp. 385–396, 2002. View at: Publisher Site | Google Scholar
  47. G. L. Feyisa, H. Meilby, R. Fensholt, and S. R. Proud, “Automated Water Extraction Index: a new technique for surface water mapping using Landsat imagery,” Remote Sensing of Environment, vol. 140, pp. 23–35, 2014. View at: Publisher Site | Google Scholar
  48. R. Fernandes and S. G. Leblanc, “Parametric (modified least squares) and non-parametric (Theil-Sen) linear regressions for predicting biophysical parameters in the presence of measurement errors,” Remote Sensing of Environment, vol. 95, no. 3, pp. 303–316, 2005. View at: Publisher Site | Google Scholar
  49. I. Nitze, S. W. Cooley, C. R. Duguay, B. M. Jones, and G. Grosse, “The catastrophic Thermokarst Lake drainage events of 2018 in northwestern Alaska: fast-forward into the future,” The Cryosphere, vol. 14, no. 12, pp. 4279–4297, 2020. View at: Publisher Site | Google Scholar
  50. M. Gocic and S. Trajkovic, “Analysis of changes in meteorological variables using Mann-Kendall and Sen's slope estimator statistical tests in Serbia,” Global and Planetary Change, vol. 100, pp. 172–182, 2013. View at: Publisher Site | Google Scholar
  51. W. B. Cohen, Z. Yang, and R. Kennedy, “Detecting trends in forest disturbance and recovery using yearly Landsat time series: 2. TimeSync -- tools for calibration and validation,” Remote Sensing of Environment, vol. 114, no. 12, pp. 2911–2924, 2010. View at: Publisher Site | Google Scholar
  52. H. Liu, P. Gong, J. Wang, X. Wang, G. Ning, and X. Bing, “Production of global daily seamless data cubes and quantification of global land cover change from 1985 to 2020 - iMap World 1.0,” Remote Sensing of Environment, vol. 258, article 112364, 2021. View at: Publisher Site | Google Scholar
  53. Y. Chen, A. Liu, and X. Cheng, “Detection of thermokarst lake drainage events in the northern Alaska permafrost region,” Science of the Total Environment, vol. 807, Part 2, article 150828, 2022. View at: Publisher Site | Google Scholar
  54. L. J. Plug and J. J. West, “Thaw lake expansion in a two-dimensional coupled model of heat transfer, thaw subsidence, and mass movement,” Journal of Geophysical Research - Earth Surface, vol. 114, no. F1, 2009. View at: Publisher Site | Google Scholar
  55. B. M. Jones, G. Grosse, C. D. Arp, M. C. Jones, K. M. Walter Anthony, and V. E. Romanovsky, “Modern Thermokarst Lake dynamics in the continuous permafrost zone, northern Seward Peninsula, Alaska,” Journal of Geophysical Research – Biogeosciences, vol. 116, no. G2, 2011. View at: Publisher Site | Google Scholar
  56. K. M. Hinkel, B. M. Jones, W. R. Eisner, C. J. Cuomo, R. A. Beck, and R. Frohn, “Methods to assess natural and anthropogenic thaw lake drainage on the western Arctic coastal plain of northern Alaska,” Journal of Geophysical Research - Earth Surface, vol. 112, no. F2, 2007. View at: Publisher Site | Google Scholar
  57. A. M. Irrgang, M. Bendixen, L. M. Farquharson et al., “Drivers, dynamics and impacts of changing Arctic coasts,” Nature Reviews Earth & Environment, vol. 3, no. 1, pp. 39–54, 2022. View at: Publisher Site | Google Scholar
  58. M. Angelopoulos, P. P. Overduin, S. Westermann et al., “Thermokarst lake to lagoon transitions in eastern Siberia: do submerged taliks refreeze?” Journal of Geophysical Research: Earth Surface, vol. 125, no. 10, 2020. View at: Publisher Site | Google Scholar
  59. R. Canavesio, “Distant swells and their impacts on atolls and tropical coastlines. The example of submersions produced by lagoon water filling and flushing currents in French Polynesia during 1996 and 2011 mega swells,” Global and Planetary Change, vol. 177, pp. 116–126, 2019. View at: Publisher Site | Google Scholar
  60. A. Witze, “The Arctic is burning like never before — and that’s bad news for climate change,” Nature, vol. 585, no. 7825, pp. 336-337, 2020. View at: Publisher Site | Google Scholar
  61. D. Blok, M. M. P. D. Heijmans, G. Schaepman-Strub, A. V. Kononov, T. C. Maximov, and F. Berendse, “Shrub expansion may reduce summer permafrost thaw in Siberian tundra,” Global Change Biology, vol. 16, no. 4, pp. 1296–1305, 2010. View at: Publisher Site | Google Scholar
  62. S. V. Kokelj, T. C. Lantz, J. Tunnicliffe, R. Segal, and D. Lacelle, “Climate-driven thaw of permafrost preserved glacial landscapes, northwestern Canada,” Geology, vol. 45, no. 4, pp. 371–374, 2017. View at: Publisher Site | Google Scholar
  63. S. M. Natali, J. P. Holdren, B. M. Rogers et al., “Permafrost carbon feedbacks threaten global climate goals,” Proceedings of the National Academy of Sciences of the United States of America, vol. 118, no. 21, 2021. View at: Publisher Site | Google Scholar
  64. J. Yan, L. Wang, W. Song, Y. Chen, X. Chen, and Z. Deng, “A time-series classification approach based on change detection for rapid land cover mapping,” ISPRS Journal of Photogrammetry and Remote Sensing, vol. 158, pp. 249–262, 2019. View at: Publisher Site | Google Scholar
  65. T. Mugiraneza, A. Nascetti, and Y. Ban, “Continuous monitoring of urban land cover change trajectories with Landsat time series and Landtrendr-Google Earth Engine cloud computing,” Remote Sensing, vol. 12, no. 18, p. 2883, 2020. View at: Publisher Site | Google Scholar
  66. J. G. Masek, M. A. Wulder, B. Markham et al., “Landsat 9: empowering open science and applications through continuity,” Remote Sensing of Environment, vol. 248, article 111968, 2020. View at: Publisher Site | Google Scholar
  67. A. Runge and G. Grosse, “Mosaicking Landsat and Sentinel-2 data to enhance LandTrendr time series analysis in northern high latitude permafrost regions,” Remote Sensing, vol. 12, no. 15, p. 2471, 2020. View at: Publisher Site | Google Scholar

Copyright © 2022 Yating Chen et al. Exclusive Licensee Aerospace Information Research Institute, Chinese Academy of Sciences. Distributed under a Creative Commons Attribution License (CC BY 4.0).

 PDF Download Citation Citation
Altmetric Score