Mosaicking Landsat and Sentinel-2 Data to Enhance LandTrendr Time Series Analysis in Northern High Latitude Permafrost Regions

Permafrost is warming in the northern high latitudes, inducing highly dynamic thaw-related permafrost disturbances across the terrestrial Arctic. Monitoring and tracking of permafrost disturbances is important as they impact surrounding landscapes, ecosystems and infrastructure. Remote sensing provides the means to detect, map, and quantify these changes homogeneously across large regions and time scales. Existing Landsat-based algorithms assess different types of disturbances with similar spatiotemporal requirements. However, Landsat-based analyses are restricted in northern high latitudes due to the long repeat interval and frequent clouds, in particular at Arctic coastal sites. We therefore propose to combine Landsat and Sentinel-2 data for enhanced data coverage and present a combined annual mosaic workflow, expanding currently available algorithms, such as LandTrendr, to achieve more reliable time series analysis. We exemplary test the workflow for twelve sites across the northern high latitudes in Siberia. We assessed the number of images and cloud-free pixels, the spatial mosaic coverage and the mosaic quality with spectral comparisons. The number of available images increased steadily from 1999 to 2019 but especially from 2016 onward with the addition of Sentinel-2 images. Consequently, we have an increased number of cloud-free pixels even under challenging environmental conditions, which then serve as the input to the mosaicking process. In a comparison of annual mosaics, the Landsat+Sentinel-2 mosaics always fully covered the study areas (99.9–100 %), while Landsat-only mosaics contained data-gaps in the same years, only reaching coverage percentages of 27.2 %, 58.1 %, and 69.7 % for Sobo Sise, East Taymyr, and Kurungnakh in 2017, respectively. The spectral comparison of Landsat image, Sentinel-2 image, and Landsat+Sentinel-2 mosaic showed high correlation between the input images and mosaic bands (e.g., for Kurungnakh 0.91–0.97 between Landsat and Landsat+Sentinel-2 mosaic and 0.92–0.98 between Sentinel-2 and Landsat+Sentinel-2 mosaic) across all twelve study sites, testifying good quality mosaic results. Our results show that especially the results for northern, coastal areas was substantially improved with the Landsat+Sentinel-2 mosaics. By combining Landsat and Sentinel-2 data we accomplished to create reliably high spatial resolution input mosaics for time series analyses. Our approach allows to apply a high temporal continuous time series analysis to northern high latitude permafrost regions for the first time, overcoming substantial data gaps, and assess permafrost disturbance dynamics on an annual scale across large regions with algorithms such as LandTrendr by deriving the location, timing and progression of permafrost thaw disturbances.

Permafrost Hydrology Research Domain: Process-Based Adjustment

Permafrost hydrology is an emerging discipline, attracting increasing attention as the Arctic region is undergoing rapid change. However, the research domain of this discipline had never been explicitly formulated. Both ‘permafrost’ and ‘hydrology’ yield differing meanings across languages and scientific domains; hence, ‘permafrost hydrology’ serves as an example of cognitive linguistic relativity. From this point of view, the English and Russian usages of this term are explained. The differing views of permafrost as either an ecosystem class or a geographical region, and hydrology as a discipline concerned with either landscapes or generic water bodies, maintain a language-specific touch of the research in this field. Responding to a current lack of a unified approach, we propose a universal process-based definition of permafrost hydrology, based on a specific process assemblage, specific to permafrost regions and including: (1) Unconfined groundwater surface dynamics related to the active layer development; (2) water migration in the soil matrix, driven by phase transitions in the freezing active layer; and (3) transient water storage in both surface and subsurface compartments, redistributing runoff on various time scales. This definition fills the gap in existing scientific vocabulary. Other definitions from the field are revisited and discussed. The future of permafrost hydrology research is discussed, where the most important results would emerge at the interface between permafrost hydrology, periglacial geomorphology, and geocryology.

Comparing Spectral Characteristics of Landsat-8 and Sentinel-2 Same-Day Data for Arctic-Boreal Regions

The Arctic-Boreal regions experience strong changes of air temperature and precipitation regimes, which affect the thermal state of the permafrost. This results in widespread permafrost-thaw disturbances, some unfolding slowly and over long periods, others occurring rapidly and abruptly. Despite optical remote sensing offering a variety of techniques to assess and monitor landscape changes, a persistent cloud cover decreases the amount of usable images considerably. However, combining data from multiple platforms promises to increase the number of images drastically. We therefore assess the comparability of Landsat-8 and Sentinel-2 imagery and the possibility to use both Landsat and Sentinel-2 images together in time series analyses, achieving a temporally-dense data coverage in Arctic-Boreal regions. We determined overlapping same-day acquisitions of Landsat-8 and Sentinel-2 images for three representative study sites in Eastern Siberia. We then compared the Landsat-8 and Sentinel-2 pixel-pairs, downscaled to 60 m, of corresponding bands and derived the ordinary least squares regression for every band combination. The acquired coefficients were used for spectral bandpass adjustment between the two sensors. The spectral band comparisons showed an overall good fit between Landsat-8 and Sentinel-2 images already. The ordinary least squares regression analyses underline the generally good spectral fit with intercept values between 0.0031 and 0.056 and slope values between 0.531 and 0.877. A spectral comparison after spectral bandpass adjustment of Sentinel-2 values to Landsat-8 shows a nearly perfect alignment between the same-day images. The spectral band adjustment succeeds in adjusting Sentinel-2 spectral values to Landsat-8 very well in Eastern Siberian Arctic-Boreal landscapes. After spectral adjustment, Landsat and Sentinel-2 data can be used to create temporally-dense time series and be applied to assess permafrost landscape changes in Eastern Siberia. Remaining differences between the sensors can be attributed to several factors including heterogeneous terrain, poor cloud and cloud shadow masking, and mixed pixels.

Trajectory of the Arctic as an integrated system

Although much remains to be learned about the Arctic and its component processes, many of the most urgent scientific, engineering, and social questions can only be approached through a broader system perspective. Here, we address interactions between components of the Arctic system and assess feedbacks and the extent to which feedbacks (1) are now underway in the Arctic and (2) will shape the future trajectory of the Arctic system. We examine interdependent connections among atmospheric processes, oceanic processes, sea-ice dynamics, marine and terrestrial ecosystems, land surface stocks of carbon and water, glaciers and ice caps, and the Greenland ice sheet. Our emphasis on the interactions between components, both historical and anticipated, is targeted on the feedbacks, pathways, and processes that link these different components of the Arctic system. We present evidence that the physical components of the Arctic climate system are currently in extreme states, and that there is no indication that the system will deviate from this anomalous trajectory in the foreseeable future. The feedback for which the evidence of ongoing changes is most compelling is the surface albedo-temperature feedback, which is amplifying temperature changes over land (primarily in spring) and ocean (primarily in autumn-winter). Other feedbacks likely to emerge are those in which key processes include surface fluxes of trace gases, changes in the distribution of vegetation, changes in surface soil moisture, changes in atmospheric water vapor arising from higher temperatures and greater areas of open ocean, impacts of Arctic freshwater fluxes on the meridional overturning circulation of the ocean, and changes in Arctic clouds resulting from changes in water vapor content.

Feedbacks and Interactions: From the Arctic Cryosphere to the Climate System

Changes in the Arctic’s climate are a result of complex interactions between the cryosphere, atmosphere, ocean, and biosphere. More feedbacks from the cryosphere to climate warming are positive and result in further warming than are negative, resulting in a reduced rate of warming or cooling. Feedbacks operate at different spatial scales; many, such as those operating through albedo and evapotranspiration, will have significant local effects that together could result in global impacts. Some processes, such as changes in carbon dioxide (CO₂) emissions, are likely to have very small global effects but uncertainty is high whereas others, such as subsea methane (CH₄) emissions, could have large global effects. Some cryospheric processes in the Arctic have teleconnections with other regions and major changes in the cryosphere have been largely a result of large-scale processes, particularly atmospheric and oceanic circulation. With continued climate warming it is highly likely that the cryospheric components will play an increasingly important climatic role. However, the net effect of all the feedbacks is difficult to assess because of the variability in spatial and temporal scales over which they operate. Furthermore, general circulation models (GCMs) do not include all major feedbacks while those included may not be accurately parameterized. The lack of full coupling between surface dynamics and the atmosphere is a major gap in current GCMs.