The Physiology of Betula glandusa on Two Sunny Summer Days in the Arctic and Linkages with Optical Imagery

Abstract

Controls on Arctic vegetation physiology have been linked to microscale (1–100 m) topography and landscape position, yet drivers may change under future climates as temperature, active-layer thickness, and nutrient limitations are removed or altered. Focusing on the cosmopolitan dwarf birch (Betula glandusa), physiological metrics were measured over two field campaigns at Trail Valley Creek, NWT, Canada, and linked to tasked and archived multispectral imagery to investigate drivers. Relative humidity was ~31.1% on 25 June 2023, and increased to 45.6% on 29 June 2023, which corresponded to heightened physiological activity of stomatal conductance and light-adapted fluorescence (gsm: 0.118 vs. 0.165 μmol m⁻² s⁻¹, Fs: 129.29 vs. 178.42). Normalized difference vegetation index of AVIRIS, Sentinel 2, and SkySat were negligibly correlated to dwarf birch physiological activity, but moderately correlated to dwarf birch height and active-layer thickness. Random forest variable importance revealed that environmental factors and field-measured active-layer thickness ranked higher than remote sensing metrics in explaining physiological activity regardless of the field campaign. Overall, these findings suggest that microscale variation can influence dwarf birch physiological activity, yet microscale effects are overwritten by environmental conditions that may hinder fine-scale space-based monitoring of Arctic vegetation physiological dynamics.

1. Introduction

Arctic vegetation has considerable microscale (1–100 m) heterogeneity in community composition and the morphology plus biomass of the dominant species related to topography and landscape position. The prominent hill slopes across the landscape have considerable effects on shrub distribution, with shrub expansion and increases in shrub cover primarily occurring in wetter sites including drainage channels and concave hillslopes [1]. Transects averaging 116.2 m along hillslopes of alder (Alnus alnobetula) at Trail Valley Creek, NWT, Canada, had canopy volumes ranging between 11.8 and 22.2 m³ with the lowest volumes mid-slope and highest volumes in wetter bottom channels [2]. The degree of morphological range is not uncommon in Arctic shrubs. At the Toolik Lake field station, Alaska, USA, three shrub species across 28 survey points had maximum heights ranging between 19 and 202 cm and maximum crown diameters between 40 and 398 cm [3], and along a two-kilometer stretch north to south on the eastern hills of Toolik Lake, three 25 m radius plots with eight sub-plots had leaf areas for dwarf birch (Betula glandusa) ranging ~8 times from ~15 to 120 cm² [4]. Although microscale heterogeneity in shrubs has been well documented, attribution to drivers remains challenging in the Arctic as logistical challenges and costs hinder the deployment of monitoring infrastructure at sufficient spatial–temporal resolutions. Remote sensing remains a viable long-term monitoring approach, with next-generation satellites offering improvements in both spectral and spatial resolution. However, there is a paucity of baseline data on shrub physiology at microscales and shrub response to warming-induced stress to investigate linkages.

Understanding shrub physiological activity and drivers across space and time in the Arctic is mainly limited to snapshot measurements, and there is a paucity of field data on the thermostability and photostability of shrub photosynthetic activity at microscales. High solar radiation is often associated with higher leaf temperatures that induce stress, leading to photoinhibition of the PSII system and closure of the stomata [5]. Arctic cloud cover in the summer may not increase under climate change [6]. Thus, the region could be subjected to a combination of high solar radiation and high temperatures, which is likely to incur physiological stress that co-varies with microscale abiotic conditions. Variation in the dark-adapted ratio of variable to maximum fluorescence (Fv/Fm) was unchanged in Betula glandusa as a function of hill slope position, but the maximum electron transport rate at light saturation did vary with hill slope position for several shrub species [7]. On an upslope-to-downslope transect, green alder (Alnus Alnobetula) stem water potential decreased downslope, while linear mixed-effect models of cumulative daily sap flow in 2015 and 2016 found that topography was a poor predictor and active-layer thickness ranked as the top predictor in both years [2]. However, variability in productivity-related functional traits was found between higher resource channels and drier upslopes where alder employed conservative resource usage strategies resulting in increased leaf mass per area and lower leaf area, resulting in lower transpirational surface area. Leaf nitrogen content further increased downslope and peaked in channels, which may explain the higher channel biomass as several studies have argued that tundra shrub productivity is nutrient- rather than water-limited [8]. As nitrogen fixation may increase under a warmer climate [9], climate change may exacerbate microscale physiological heterogeneity.

Conditions in the Inuvik, NWT, Canada, region in the summer of 2023 were unusually warm and vegetation responded positively with shrub patches unusually green and abundant. These climatic anomalies were hypothesized to exacerbate active-layer thickness, microclimatic, and plant biomass by topography and landscape position that would illicit stress responses in dwarf birch and that patch level differences would be associated with remote-sensing-derived metrics of plant biomass and leaf area index (e.g., normalized difference vegetation index). Utilizing a snapshot approach to acquire environmental conditions and plant physiology within a short temporal window, this study investigated the distribution of the physiological activity at the microscale and the factors that affect its activity, and the potential linkages to remotely sensed imagery. We employed a porometer/fluorometer to examine the physiological activity of the ubiquitous dwarf shrub (Betula glandusa) at the individual patch scale in the Trail Valley Creek watershed, NWT.

2. Materials and Methods

Trail Valley Creek (TVC: latitude: 68°44′N, longitude: 133°30′W) is located approximately 50 km North of Inuvik, NWT, Canada, and is a long-term Arctic research site within a 57 km² watershed at the forest–tundra ecotone (Figure 1). The region is characterized by rolling hills, incised river valleys, and small lakes underlain by continuous permafrost [10] with an active layer that varies between 0.5 and 0.8 m in thickness [11]. Soils are cryosols consisting of an upper peat layer (0.5–50 cm) underlain by silty clay [12]. The region is typically covered by snow for eight months of the year from October to May. Mean annual temperature and precipitation in Inuvik between 1991 and 2020 were −7.0 °C and 249.8 mm, respectively (Environment and Climate Change Canada 2023). Summertime (June/July) normals have daily average temperatures of 11.6/14.2 °C, yet the temperatures in 2023 were 13.4/20.5 °C. Monthly precipitation was also lower, 19.6/18.1 mm versus normal of 23.5/40.1 mm.

Vegetation at TVC varies between “open tundra”, which is mostly shrub-free, to shrub-dominated patches of either alder or dwarf birch, with individual alders usually separate by more than a meter and interspersed with other vegetation, while dwarf birch patches form a nearly closed and uniform canopy where individuals are closely packed and hard to discern (Figure 2). Vegetation in the open tundra is of a low height (5–25 cm) and is predominantly covered with reindeer lichen (Cladonia rangiferina L.), Sphagnum moss (Sphagnum L.), tussock and non-tussock sedges (e.g., Carex L.), and Labrador tea (Rhododendron groenlandicum). Dwarf birch occurrence is predominantly as small individual plants with considerably less canopy volume (Figure 2A). The open tundra primarily occurs on the flatter hilltop summits that feature gradual elevation changes and wetter conditions, with Sphagnum occurring in slightly deeper flow melt paths. Upland, individual small dwarf birch individuals are common, and patches of alder and birch occur on these flat lands, albeit at lower density and reduced size. Transitioning from open tundra, low-growing woody vegetation (e.g., Vaccinium uliginosum, Rhododendron tomentosum, R. lapponicum, Betula glandulosa) increase in abundance with these species closely interspersed (Figure 2B) until alder or dwarf birch dominate. Dwarf birch dominant patches (Figure 2C) have a relatively closed canopy, stem density is higher, and vegetation heights are 40–60 cm, reaching >70 cm on ridge tops. Dwarf birch patches remain quite intermixed with other species, particularly in the understory. The hummock-hollow microtopography influences species composition, with hollows increasingly dominated by lichens, sphagnum, and labrador tea. Dwarf birch patches predominate sloped terrain, hillslopes, and the tops of ridges, and densities are higher on western-facing sloped terrain.

Dwarf birch (Betula glandusa) was chosen as the study species since it is cosmopolitan at TVC and forms dense patches of sufficient size and relative uniformity that can be linked to remote sensing imagery. Two field campaigns were undertaken in June of 2023 during solar noon (plus and minus 3 h from 1 pm) during completely cloud-free days in a stretch of hot sunny weather with no precipitation. As such, individuals had been exposed to constant 24 h sun for several days prior. The first field campaign (Grid) conducted on 25 June utilized a ~30 m grid pattern over a 60 × 100 m area (n = 12) that sampled north to south from open tundra to the middle of a dwarf birch patch at the summit of a hilltop (Figure 3). The terrain was relatively flat, and dwarf birch size and height increased in the dwarf birch patch. The second campaign (Circle) conducted on 29 June originated at the base of the TVC camp and circled the camp clockwise across an ~1200 × 900 m area (n =28). Sites were spaced several hundred meters apart and chosen to reflect top to bottom of a hillslope, east and west side of ridges, the tallest dwarf birch individual, patches surrounded by open tundra, and density gradients. All sites were classified into three dwarf birch density classes: (1) Low, where dwarf birch individuals are separated; (2) High, where birch is the dominant species and there is a consistent overstory; (3) High-Tall, where dwarf birch is the dominant species, and the vegetation height is over 70 cm. Dwarf birch density correlates strongly with both topography and landscape position.

At each site, seven random stems were chosen and the top fully mature leaf per stem was measured using an LI-600f (LICOR Biosciences, Lincoln, NE, USA), which recorded the steady-state (Fs), maximum (F’m), operating efficiency of PSII (PhiPS2), electron transport rate (ETR), and stomatal conductance (gsw) for light-adapted leaves plus other ancillary environmental parameters. Care was taken to collect measurements from a standardized position that did not block the sun and oriented the leaf horizontally. In addition, a metal frost probe was inserted into four randomly selected locations (two local high points and two low points) until it could be pushed no further, which was assumed to represent the active-layer thickness. Site photographs were taken with a Nikon D2600 camera mounted to a metal pole and facing the nadir from an approximately two-meter height. Additionally, measurements of above- and below-canopy ambient light conditions were collected as a crude metric of light extinction by the overstory. The coordinates of each site were recorded to link to remote sensing imagery.

The remote sensing imagery of TVC consisted of a tasked SkySat, Sentinel 2 archive, and AVIRIS-ng flight. The SkySat image was collected at ~10 a.m. local time at 18.9° off nadir on 30 June 2023. The Sentinel 2 image was collected on 4 July 2023, and the AVIRIS image on 26 July 2019. All imagery was atmospherically corrected to reflectance for the purpose of deriving the normalized difference index (NDVI) [13] and the photochemical reflectance index (PRI) [14]. Metrics of NDVI and PRI were extracted at each sample location as the average value of all pixels within a 3 × 3 square buffer. As such, the buffer distance varied from 1.5 to 30 m with satellite resolution. Digital elevation and vegetation height were extracted from 2016 airborne Light Detection and Ranging (LIDAR) collected with a Riegl LMS-Q680i on board the Alfred Wegener Institute’s POLAR-5 science aircraft.

To evaluate whether dwarf birch density influenced the LI-600f measurements, an ANOVA was performed and if significant differences were detected, a post hoc Tukey Host Significant Difference test was utilized to facilitate pairwise comparisons. Analysis was performed in R (version 4.3.2) using the stats and multcompView packages. For each physiology activity metric, an explanatory model was crafted from the field measurements (i.e., average active-layer thickness and height of vegetation), remote sensing NDVI and PRI, and ancillary data (i.e., relative humidity, air temperature, ambient light). Random forest variable importance (randomForest and varImp packages) was utilized to determine the top-performing suite of variables for a linear regression model. Furthermore, the regression coefficients were standardized using the lm.beta package to determine the sign of the coefficients. Several variable combinations were evaluated and non-statistically significant variables (package lm, p > 0.05) were not included. All candidate models were subjected to a due diligence suite of tests, which evaluated multicollinearity, the statistical significance of each explanatory variable, and the improvement in r-squared and AIC.

3. Results

The Grid campaign transitioned north to south between open tundra into the interior of a moderately tall dominant dwarf birch patch. Each row of three had essentially similar characteristics resulting in six samples each for the low and high dwarf birch density classes (Figure 3). At this scale, there was no difference in active-layer thickness. Covering a larger area, the Circle field campaign had larger variations in dwarf birch morphology. Larger birch occurred along the eastern hillslope just south of the base camp and the tallest shrubs were present along the western ridges. The western side of the ridges had higher NDVI values on both the western and eastern ridge lines from the Trail Valley Creek base camp (Figure 3 right). Areas of high NDVI values also include the river channels and individual alders. North of the base camp the open-tundra region has low NDVI values, although the dwarf birch and alder patches stand out. In these patches (e.g., Circle samples 14 and 18), NDVI was ~0.4, which is considerably smaller than the peak NDVI of ~0.58 around the western ride (e.g., Circle patches 9 and 10).

Over the course of the field data collection, air temperature increased while relative humidity decreased (Figure 4). Air temperatures increased nearly 10 °C during the Circle field data collection, a rate of ~1.88° per hour. Interestingly, despite the air temperature being nearly the same between the two sampling days, the relative humidity averaged 28.5% during the Grid sampling and 45.1% during the same time interval of the Circle sampling. The inverse relationship between air temperature and humidity is not unexpected, although the difference in humidity is considerable given the lack of precipitation, suggesting that transpiration in the greater region may have occurred. Atmospheric vapor pressure change during the sampling periods was ~0.05 kpa per hour and there was no difference in insolation between sampling days (average solar insolation ~1634 μmol m⁻² s⁻¹, standard deviation 154 μmol m⁻² s⁻¹), and negligible difference in insolation over time. The constant insolation is likely responsible for the increasing regional air temperature.

3.1. Birch Physiological Dynamics

Stomatal conductance as a function of dwarf birch density had consistent trends of higher conductance at low density for both the Grid and Circle campaigns (Figure 5). In the Grid campaign, the low dwarf birch density class was conducted first, which corresponds to the noted trends in relative humidity and temperature. However, both relative humidity and stomatal conductance were recorded by the same LI-600f instrument, which differs from the Inuvik weather station’s hourly relative humidity. During the Grid campaign, hourly relative humidity increased from 34% to 36%, while it decreased from 32% to 28% during the Circle campaign. The difference in stomatal conductance was statistically significant between low and high dwarf birch density in the Grid campaign, but not the Circle campaign data set (Figure 6), suggesting the change in environmental conditions over time was the dominant controlling variable. Dwarf birch over 70 cm only occurred in the Circle field campaign on the western ridges, yet stomatal conductance was statistically significant from the low and high dwarf birch density. Photochemistry metrics were not statistically significant with dwarf birch density (Figure 6). Mean Fs and PhiPS2 were slightly higher in the high dwarf birch density sites, suggesting higher light-induced stress in the open-tundra and tundra adjacent sites. Taller dwarf birch individuals had the lowest average Fs and quantum efficiency similar to high dwarf birch density. Trends in PhiPS2 were not consistent between the low- and high-density classes in the Grid and Circle campaigns, albeit the Circle campaign took measurements over a wider spatial area and a longer period. The maximum fluorescence was greatest in the Circle high dwarf birch density class, albeit the difference was not statistically significant and there appears to be minimal difference in the Grid sampling between density classes. The electron transport rate for both field campaigns ranged from 65 to 333 μmol m⁻² s⁻¹ (mean 188.9, standard deviation 47.4).

3.2. Explanatory Variables

The top explanatory variables for the four physiological metrics were relatively consistent between the Grid, Circle, and Grid + Circle (All) data sets (Figure 7). Overall, the variable importance in the Grid data set had lower importance values denoting the poorer model fit, likely due to the relatively minor degree of variance in the controlling variables within the small survey area. The Grid + Circle data set does include changes between days, which did have distinct physiological activity, whereas the environmental conditions were more stable within a sampling campaign. The main difference between the Grid and Circle data sets is the larger sampling window for the Circle data set as well as its much larger spatial scale and diversity of dwarf birch density, location, and terrain. For stomatal conductance, relative humidity was the most important variable and positively related, while leaf temperature was negatively related as expected. Derived NDVI and PRI were moderately important and negatively related with the exception of NDVI derived from SkySat.

However, further analysis revealed that NDVI and PRI for any satellite when added to models involving the top two explanatory variables (e.g., relative humidity and leaf temperature) were not statistically significant. Vegetation height and terrain height were moderately important in the Grid + Circle data set, but their influence was negligible when Grid and Circle were separated and their relationship to stomatal conductance varied from positive to negative. The only other variable of statistical significance was active-layer thickness. Although active-layer thickness was of negligible importance in the Grid data set, this is likely because of the proximity of sample sites, minor variances at this scale, and challenges associated with measuring the active layer. Only at a regional scale was active-layer thickness significant (i.e., it was not statistically significant for the Grid data set, but was for the Circle and Grid + Circle data sets). For the Grid + Circle data set, a model involving relative humidity, leaf temperature, and frost table depth had an adjusted R² of 0.702 (F-statistic: 31.63 on 3 and 36 DF, p-value < 0.001), albeit the inclusion of active-layer thickness increased the R² only by 0.0463.

Steady-state and maximum fluorescence could not be explained to the same extent as stomatal conductance, and there were considerable differences between variable importance in each data set. For instance, leaf temperature was negatively correlated to Fs and F’m for the Grid data set, but not for the Circle and Grid + Circle data sets. The importance rankings for these variables were close to the top in the Grid + Circle data sets, but not for either Grid or Circle independently, suggesting the change in humidity magnitude between sampling dates may be the cause. Similarly, terrain was the most important variable, albeit its importance to the Grid and Circle data sets was minimal. Comparing the Grid high dwarf birch density sample point (10–12) to the closest Circle sample points (26–28), there is a clear temporal difference in Fs and F’m (Grid: Fs 144.10│F’m 205.5 vs. Circle: Fs 210.8│F’m 274.4). Since all Grid data points have similar elevation (~85 m) and lower Fs plus F’m values, their inclusion weighs down the one end of the elevation range (total elevation range of 60.4–88.9 m), effectively “tuning” the regression and increasing the terrain elevation importance.

The quantum efficiency of photosystem II (i.e., PhiPS2) was primarily related to ambient light. However, the LI-600f instrument was oriented towards the sun on a horizontal position within a few seconds, albeit the orientation was not controlled and thus varied by a few degrees in the roll, pitch, and yaw axis, in addition to tracking the sun. The Grid data set had the strongest relationship between quantum efficiency and ambient light (R² = 0.4173), albeit there was a slight increase from 1553.6 to 1640.6 μmol m⁻² s⁻¹ during the course of the campaign. For the Grid data set, the change in quantum efficiency has an R² of 0.40, and the inclusion of another variable did not improve the model and no other variable was statistically significant. However, for the Circle data set, there were no statistically significant variables (p < 0.05), although ambient light, leaf temperature, and relative humidity had p-values between 0.12 and 0.15. Depending on the data set, active-layer thickness was statistically significant. No remote-sensing-derived NDVI or PRI were statistically significant for any data set. Overall, the variance in quantum efficiency explained was less than R² = 0.20.

Correlations between the physiological metrics varied by field campaign. Quantum efficiency and stomatal conductance had weak correlations in the Circle campaign (Pearson’s: 0.179, Spearman’s 0.178) and moderate correlations in the Grid campaign (Pearsons: 0.498, Spearman’s 0.389). The inclusion of stomatal conductance in regression models of PhiPS2 for the Grid + Circle data sets did improve the explanatory power (R² = 0.31), was statistically significant, and was positively related to PhiPS2. Fs and F’m were highly correlated (>0.9) regardless of the data set, and both were negatively correlated with PhiPS2. In all data sets, Fs had a stronger correlation to PhiPS2. Correlations between PhiPS2 and Fs were weaker for the Grid data set (Pearson’s: −0.376, Spearman’s −0.226) than the Circle data set (Pearson’s: −0.650, Spearman’s −0.645) as there were fewer and more scattered data points in the >200 Fs range and more variation in the 50–150 Fs range.

4. Discussion

Comparison of this study’s results to previous studies is challenging as there is a paucity of microscale data on dwarf birch physiological activity and comparisons are hindered by differences in equipment and the focus on light- or dark-adapted leaves. Dwarf birch physiology in relation to hillslope position was measured in June 2003 using an FMS II, Hansatech Instruments Limited, Norfolk, U.K., after conditioning the leaves to dark conditions [7]. The dark-adapted ratio of variable to maximum fluorescence (Fv/Fm) was much greater than in our study (mean 0.291, standard deviation 0.078), with average values of ~0.8. The photosynthetically active radiation level that solicited the maximum electron transport rate was approximately 680 μmol m⁻² s⁻¹, whereas, during the Grid and Circle campaigns, the ambient light conditions never dropped below 1200 μmol m⁻² s⁻¹. For dark-adapted leaves, Fv/Fm ratios below 0.69 were suggested as stressed for dwarf birch [15]. However, in terms of stomatal conductance, the measured values in our study were within the normal range. Dwarf birch from western Greenland had stomatal conductance ranges between ~0.750 and 0.175 mol m⁻² s⁻¹ [16], which is similar to our study, which had an interquartile range of 0.110–0.193 mol m⁻² s⁻¹. However, controlled growth experiments for dwarf birch have documented stomatal conductance between 0.6 and 1.2 mol m⁻² s⁻¹ under lowered and elevated carbon dioxide [17]. Stomatal conductance in dwarf birch has a seasonal component increasing from early to late summer, with mid-summer stomatal conductance ~0.2 mol m⁻² s⁻¹ in Toolkit Lake, Alaska, during a particularly hot year [18]. Overall, these results suggest that dwarf birch stomata did not show indications of stress, although the photosynthetic system was likely stressed. Leaf quantum efficiency can be considerably reduced by air temperature or insolation [5]. As insolation was invariable, air temperatures could be the main physiological detriment in this study.

Linkages between dwarf birch physiology and remote sensing metrics were poor. Although NDVI was negligibly correlated to any physiological or environmental metric, it was correlated to both vegetation height (Pearson’s: 0.580, Spearman’s 0.536) and active-layer thickness (Pearson’s: 0.504, Spearman’s 0.494) for SkySat NDVI. Correlations for AVIRIS and Sentinel 2 were not as strong as for SkySat, underscoring the importance of spatial resolution. As all satellite products were the average of a 3 × 3 grid, the larger pixel size likely incurs more mixed pixels between dwarf birch and other vegetation. Even within dominant dwarf birch patches, there are often gaps in the dwarf birch canopy that exposes the understory vegetation and underlaying Sphagnum or lichen. Thus, it is not surprising that high-resolution multispectral imagery would have good alignment with field-measured values of plant morphology. Although a crude metric of morphology, vegetation height was found to encompass several biomass-related components such as stem diameter, canopy closure, and leaf biomass. Furthermore, dwarf birch vegetation height was strongly correlated to active-layer thickness (Pearson’s: 0.633, Spearman’s 0.486) and anecdotally corresponded to landscape position and terrain. Assessments of microscale active-layer thickness in relation to shrub traits have suggested that stem protrusion through snow hastens snowmelt. Combined with the low snow depth due to topographic positioning, the advanced snow-free date increases the thermal load and deepens the active layer [19]. These findings suggest that dwarf birch physiology is only weakly coupled to plant morphology and has stronger ties to current and historical environmental conditions. The change in plant physiology between the Grid and Circle campaigns and their differences in baseline humidity does appear to have a strong influence on photochemistry. The higher humidity in the Circle campaign increased both stomatal conductance and increased Fs and F’m, albeit PhiPS2 was similar.

5. Conclusions

Dwarf birch physiological activity was minimally related to remotely sensed vegetation, whereas relative humidity was often the most important predictor variable whose short temporal variability had a major positive influence on stomatal conductance, steady-state, and maximal fluorescence, while quantum efficiency was reduced by ~10% at higher relative humidity. Subtle variability in physiological activity was weakly correlated to active-layer thickness and birch density. As active-layer thickness was correlated to NDVI, there remain some options for snapshot remote sensing to quantify spatial heterogeneity in physiology. However, linkages between frost depth and NDVI were spatial resolution dependent, with SkySat NDVI having the strongest correlation with active-layer thickness and vegetation height. Although thermal imagery has potential strong relationships to leaf temperatures and atmospheric humidity, the only feasible satellite imagery provider is through the Landsat program, which offers 60 m spatial resolution. This resolution is much too coarse to investigate microscale variations, especially as many birch patches on the open tundra are less than 100 × 100 m and only hillslope and ridge patches would be discernable, limiting their utility. Upcoming instruments such as FLEX may offer imaging of photosynthetic activity from space, but the resolution of 300 m would not facilitate monitoring of microscale heterogeneity, underscoring the importance of field measurements to elucidate physiological drivers.