Arctic permafrost-affected soils store large amounts of carbon, and high-quality maps of these soils are needed to model climate feedbacks from permafrost thaw. Ice-rich polygonal tundra is one landscape type that is widespread in the Arctic and rich in carbon. These environments are especially susceptible to climate change as thawing of ground ice causes the irreversible degradation of these landforms. The thawing processes open pathways for release of carbon that has been preserved under frozen conditions over long timescales. This release can occur through gradual thickening of the active layer, which is the upper ground layer that thaws seasonally, but also through abrupt thaw processes, such as thermokarst formation following thaw of ice-rich ground. To better project the future trajectory of permafrost carbon at local to regional scales we need high-resolution information on soil and landscape properties.

This thesis aims to combine field sampling and spatial modeling to investigate the soils and landforms of permafrost landscapes along the Canadian Yukon coast of the Beaufort Sea coast. A major focus of this thesis is on mapping the variability of the landscape on different scales, as most pan-Arctic studies have a coarse resolution and do not capture local variability. It utilizes advanced machine learning methods for digital soil mapping to analyze soil organic carbon and nitrogen stock distributions across multiple scales, while also assessing the associated uncertainties. The availability of high- and medium-resolution (here defined as <10 m and ≥10 m pixels resolution) satellite imagery enables detailed landcover mapping, and this thesis explores various pattern recognition methods for landcover classification.

The results show that parallel analyses at multiple scales is necessary to understand carbon storage and landscape dynamics. For studies beyond the local scale medium-resolution data has the advantage of capturing differences at the landform level, while also being more widely available and accessible compared to high-resolution data. Lower spatial resolution fails to detect local variability and masks subpixel heterogeneity, whereas high-resolution mapping uncovers this variability, revealing distinctions between landforms and regions with varied landscape histories.

Ice wedge polygon landscapes are heterogeneous and carbon storage as well as lateral fluxes are determined by polygon type (high center polygon, low center polygon), but also their sublandform types (troughs, rims, centers). The object-based landcover mapping approach shows that spectral properties allow the differentiation of ice wedge polygon type, but scale properties are important to distinguish between centers, troughs and rims.

This thesis emphasizes that properties and spatial distribution of sampling sites are critical for accurate mapping results; high mapping accuracy requires that available field sites effectively capture the full range of the landscape's variability. This poses significant challenges for synthesis studies that utilize existing soil data. This thesis further highlights that an integrated view on soils and hydrological systems is necessary to understand carbon storage and potential release from ice wedge polygon landscapes.

Ice-wedge polygon landscapes make up a substantial part of high-latitude permafrost landscapes. The hydrological conditions shape how these landscapes store and release organic carbon. However, their coupled water‑carbon dynamics are poorly understood as field measurements are sparse in smaller catchments and coupled hydrology-dissolved organic carbon (DOC) models are not tailored for these landscapes. Here we present a model that simulates the hydrology and associated DOC export of high-centered and low-centered ice-wedge polygons and apply the model to a small catchment with abundant polygon coverage along the Yukon Coast, Canada. The modeled seasonal pattern of water and carbon fluxes aligns with sparse field data. These modeled seasonal patterns indicate that early-season runoff is mostly surficial and generated by low-centered polygons and snow trapped in troughs of high-centered polygons. High-centered polygons show potential for deeper subsurface flow under future climate conditions. This suggests that high-centered polygons will be responsible for an increasing proportion of annual DOC export compared to low-centered polygons. Warming likely shifts low-centered polygons to high-centered polygons, and our model shows that this shift will cause a deepening of the active layer and a lengthening of the thawing season. This, in turn, intensifies seasonal runoff and DOC flux, mainly through its duration. Our model provides a physical hypothesis that can be used to further quantify and refine our understanding of hydrology and DOC export of arctic ice-wedge polygon terrain. 

Soil organic carbon (SOC) in Arctic coastal polygonal tundra is vulnerable to climate change, especially in soils with occurrence of large amounts of ground ice. Pan-arctic studies of mapping SOC exist, yet they fail to describe the high spatial variability of SOC storage in permafrost landscapes. An important factor is the landscape history which determines landform development and consequently the spatial variability of SOC. Our aim was to map SOC stocks, and which environmental variables that determine SOC, in two adjacent coastal areas along Canadian Beaufort Sea coast with different glacial history. We used the machine learning technique random forest and environmental variables to map the spatial distribution of SOC stocks down to 1 m depth at a spatial resolution of 2 m for depth increments of 0-5, 5-15, 15-30, 30-60 and 60-100 cm. The results show that the two study areas had large differences in SOC stocks in the depth 60-100 cm due to high amounts of ground ice in one of the study areas. There are also differences in variable importance of the explanatory variables between the two areas. The area low in ground ice content had with 66.6 kg C/m(-2) more stored SOC than the area rich in ground ice content with 40.0 kg C/m(-2). However, this SOC stock could be potentially more vulnerable to climate change if ground ice melts and the ground subsides. The average N stock of the area low in ground ice is 3.77 kg m(-2) and of the area rich in ground ice is 3.83 kg m(-2). These findings support that there is a strong correlation between ground ice and SOC, with less SOC in ice-rich layers on a small scale. In addition to small scale studies of SOC mapping, detailed maps of ground ice content and distribution are needed for a validation of large-scale quantifications of SOC stocks and transferability of models.

Ongoing climate warming in the western Canadian Arctic is leading to thawing of permafrost soils and subsequent mobilization of its organic matter pool. Part of this mobilized terrestrial organic matter enters the aquatic system as dissolved organic matter (DOM) and is laterally transported from land to sea. Mobilized organic matter is an important source of nutrients for ecosystems, as it is available for microbial breakdown, and thus a source of greenhouse gases. We are beginning to understand spatial controls on the release of DOM as well as the quantities and fate of this material in large Arctic rivers. Yet, these processes remain systematically understudied in small, high-Arctic watersheds, despite the fact that these watersheds experience the strongest warming rates in comparison. Here, we sampled soil (active layer and permafrost) and water (porewater and stream water) from a small ice wedge polygon (IWP) catchment along the Yukon coast, Canada, during the summer of 2018. We assessed the organic carbon (OC) quantity (using dissolved (DOC) and particulate OC (POC) concentrations and soil OC content), quality (δ¹³C DOC, optical properties and source apportionment) and bioavailability (incubations; optical indices such as slope ratio, S_r; and humification index, HIX) along with stream water properties (temperature, T; pH; electrical conductivity, EC; and water isotopes). We classify and compare different landscape units and their soil horizons that differ in microtopography and hydrological connectivity, giving rise to differences in drainage capacity. Our results show that porewater DOC concentrations and yield reflect drainage patterns and waterlogged conditions in the watershed. DOC yield (in mg DOC g⁻¹ soil OC) generally increases with depth but shows a large variability near the transition zone (around the permafrost table). Active-layer porewater DOC generally is more labile than permafrost DOC, due to various reasons (heterogeneity, presence of a paleo-active-layer and sampling strategies). Despite these differences, the very long transport times of porewater DOC indicate that substantial processing occurs in soils prior to release into streams. Within the stream, DOC strongly dominates over POC, illustrated by DOC/POC ratios around 50, yet storm events decrease that ratio to around 5. Source apportionment of stream DOC suggests a contribution of around 50 % from permafrost/deep-active-layer OC, which contrasts with patterns observed in large Arctic rivers (12 ± 8 %; Wild et al., 2019). Our 10 d monitoring period demonstrated temporal DOC patterns on multiple scales (i.e., diurnal patterns, storm events and longer-term trends), underlining the need for high-resolution long-term monitoring. First estimates of Black Creek annual DOC (8.2 ± 6.4 t DOC yr⁻¹) and POC (0.21 ± 0.20 t yr⁻¹) export allowed us to make a rough upscaling towards the entire Yukon Coastal Plain (34.51 ± 2.7 kt DOC yr⁻¹ and 8.93 ± 8.5 kt POC yr⁻¹). Rising Arctic temperatures, increases in runoff, soil organic matter (OM) leaching, permafrost thawing and primary production are likely to increase the net lateral OC flux. Consequently, altered lateral fluxes may have strong impacts on Arctic aquatic ecosystems and Arctic carbon cycling.

Aboveground vegetation biomass in northern treeless landscapes - peatlands and Arctic tundra - has been modelled with spectral information derived from optical remote sensing in several studies. However, synthesized overviews of biomass patterns across circumpolar sites have been limited. Based on data from eight study sites in Europe, Siberia and Canada, we ask (1) how biomass is divided between plant functional types (PFTs) and (2) how well biomass patterns can be detected with widely available, moderate spatial resolution (3-10 m) satellite imagery and topographic data. We explain biomass patterns using random forest regressions with the predictors being spectral bands and indices calculated from multi-temporal Sentinel-2 and PlanetScope imagery and topographic information calculated from ArcticDEM data. Our results indicate that there are notable differences in vegetation composition between northern landscapes with mosses, graminoids and deciduous shrubs being the most dominant PFTs. Remote sensing data detects biomass patterns, but regression performance varies between sites (explained variance 36-70%, normalized root mean square error 9-19%). There is also variability between sites whether Sentinel-2 or PlanetScope data is more suitable to detect biomass patterns and which the most important predictors are. Topographic information has a minor or negligible importance in most of the sites. Our results suggest that there is no easily generalizable relationship between satellite-derived vegetation greenness and biomass.

Arctic tundra landscapes are highly complex and are rapidly changing due to the warming climate. Datasets that document the spatial and temporal variability of the landscape are needed to monitor the rapid changes. Synthetic Aperture Radar (SAR) imagery is specifically suitable for monitoring the Arctic, as SAR, unlike optical remote sensing, can provide time series regardless of weather and illumination conditions. This study examines the potential of seasonal backscatter mechanisms in Arctic tundra environments for improving land cover classification purposes by using a time series of HH/HV TerraSAR-X (TSX) imagery. A Random Forest (RF) classification was applied on multi-temporal Sigma Nought intensity and multi-temporal Kennaugh matrix element data. The backscatter analysis revealed clear differences in the polarimetric response of water, soil, and vegetation, while backscatter signal variations within different vegetation classes were more nuanced. The RF models showed that land cover classes could be distinguished with 92.4% accuracy for the Kennaugh element data, compared to 57.7% accuracy for the Sigma Nought intensity data. Texture predictors, while improving the classification accuracy on the one hand, degraded the spatial resolution of the land cover product. The Kennaugh elements derived from TSX winter acquisitions were most important for the RF model, followed by the Kennaugh elements derived from summer and autumn acquisitions. The results of this study demonstrate that multi-temporal Kennaugh elements derived from dual-polarized X-band imagery are a powerful tool for Arctic tundra land cover mapping.

Although many studies exist on the estimation and monitoring of above-ground biomass (AGB) of forest ecosystems by methods of remote sensing, very little research has been carried out for ecosystems of low primary production, such as grasslands, steppes, or savannas. Our study intends to approach this gap and investigates the correlation between space-borne radar information and AGB at the scale of 10 tons per hectare and below. Additionally, we introduce the integration of passive brightness temperature as an additional covariate for biomass estimation, based on the hypothesis that it contains information complementary to microwave backscatter of the active sensors. Our findings show that large-scale estimates of AGB can be conducted for grasslands and savannas at high accuracy (R-2 up to 0.52). Additionally, we found that the integration of passive radar can increase the quality of AGB estimates in terms of explained variance for selected cases. We hope that these indications are a starting point for more integrated approaches toward biomass estimations based on Earth observation methods.