Intra‐ and interspecific variation in spectral properties of dominant Sphagnum moss species in boreal peatlands

Abstract Boreal peatlands store ~25 % of global soil organic carbon and host many endangered species; however, they face degradation due to climate change and anthropogenic drainage. In boreal peatlands, vegetation indicates ecohydrological conditions of the ecosystem. Applying remote sensing would enable spatially and temporally continuous monitoring of peatland vegetation. New multi‐ and hyperspectral satellite data offer promising approaches for understanding the spectral properties of peatland vegetation at high temporal and spectral resolutions. However, using spectral satellite data to their fullest potential requires detailed spectral analyses of dominant species in peatlands. A dominant feature of peatland vegetation is the genus Sphagnum mosses. We investigated how the reflectance spectra of common boreal Sphagnum mosses, collected from waterlogged natural conditions after snowmelt, change when the mosses are desiccated. We conducted a laboratory experiment where the reflectance spectra (350–2500 nm) and the mass of 90 moss samples (representing nine species) were measured repetitively. Furthermore, we examined (i) their inter‐ and intraspecific spectral differences and (ii) whether the species or their respective habitats could be identified based on their spectral signatures in varying states of drying. Our findings show that the most informative spectral regions to retrieve information about the Sphagnum species and their state of desiccation are in the shortwave infrared region. Furthermore, the visible and near‐infrared spectral regions contain less information on species and moisture content. Our results also indicate that hyperspectral data can, to a limited extent, be used to separate mosses belonging to meso‐ and ombrotrophic habitats. Overall, this study demonstrates the importance of including data especially from the shortwave infrared region (1100–2500 nm) in remote sensing applications of boreal peatlands. The spectral library of Sphagnum mosses collected in this study is available as open data and can be used to develop new methods for remote monitoring of boreal peatlands.


| INTRODUC TI ON
Northern peatlands have a critical role in the global carbon cycle and climate system (FAO, 2020;Harenda et al., 2018). Constituting up to 25% of soil organic carbon (Loisel et al., 2021;Yu et al., 2010), they have functioned as a strong sink and stock of organic carbon throughout the Holocene (Nichols & Peteet, 2019). They are also hotspots for biodiversity (Fraixedas et al., 2017;Saarimaa et al., 2019) and can store large quantities of water, making them important water-retaining and water-regulating ecosystems (Blodau & Moore, 2003).
The ecohydrology of a peatland site, that is, the spatial, temporal, and qualitative variation of the incoming water on the site, determines the structure and species composition of the site's vegetation communities (Laine & Vasander, 1996). An established way to classify peatlands is according to their ecohydrological qualities into ombrotrophic and minerotrophic and further into eu-, meso-, and oligotrophic peatlands (Rydin & Jeglum, 2006). Sites referred to as minerotrophic receive their water from nutritious mineral soil and surface water as well as from precipitation and thus, have a higher nutrient availability than the ombrotrophic sites. Sites with a thick enough peat layer to isolate the surface vegetation from the nutritious mineral soil water, making precipitation their only source of water are called ombrotrophic. Sub-categories of minerotrophic sites, eu-, meso-, and oligotrophic sites form a sequence of nutrient levels with decreasing nutrient availability.
One of the predominant features of the vegetation of boreal peatlands is the genus Sphagnum mosses, also known as peat mosses . Because of their substantial biomass production rate as well as relatively slow decay rate (Limpens & Berendse, 2003), Sphagnum mosses are the leading component of peat formation in boreal peatland ecosystems (Robroek et al., 2007).
They are also commonly used as indicators of a peatland site's ecohydrology because of their narrow habitat preferences (Johnson et al., 2015). Thus, remote monitoring of Sphagnum mosses and changes in their coverage is critical for understanding how the vegetation and growth conditions in peatlands are developing due to climate change and anthropogenic drainage.
Hyperspectral satellite data that will be increasingly available from new and forthcoming satellite missions (e.g., EnMAP, PRISMA, CHIME) will open novel possibilities for monitoring peatland vegetation at the species level due to the higher spectral coverage.
However, to use the novel hyperspectral airborne or satellite data sets to their full spectral extent, detailed measurements of the spectral properties of key plant species are needed. The potential for using spectral data to differentiate Sphagnum species from each other has been speculated in previous literature, as the spectral attributes of different Sphagnum species have been found to be remarkably different from each other in laboratory and field measurements (Harris & Bryant, 2009;Tucker et al., 2022;Vogelmann & Moss, 1993). Nevertheless, intraspecific variation in Sphagnum species' spectra remains poorly understood, both because the sample sizes in previous studies have been very small, often only one sample per species, or the number of studied species has been very limited. Furthermore, nearly all recent research on Sphagnum's spectral characteristics has been conducted in temperate peatlands (Bryant & Baird, 2003;Harris, 2008;Lees et al., 2019Lees et al., , 2020 instead of boreal, where the growing conditions differ considerably from the temperate zone in terms of the length of the snow-free season, and seasonal variations in solar irradiance and temperature. Additionally, much of the peatland surface vegetation studies in situ have been conducted using sensors that only record VIS-and NIR-spectral regions (McPartland et al., 2019;Pang et al., 2022;Tucker et al., 2022).
This means that the information regarding the species and the moisture content derived using data from the SWIR-wavelength regions are currently unexplored for boreal peatland vegetation.
The aim of this study was to collect and analyze a spectral library of nine Sphagnum species common in European boreal peatlands and which form large growths, and that could, thus, be monitored with remote sensing sensors in the future. The moss samples were collected from natural, waterlogged peatlands after snowmelt. Using the newly collected spectral library, we (i) analyzed the intra-and interspecific variations in Sphagnum species' spectra in different moisture conditions and (ii) investigated how the moisture content of the mosses was related to their spectral properties. We also included key species whose spectral qualities have been studied very little, such as Sphagnum centrale, Sphagnum girgensohnii, Sphagnum riparium, and Sphagnum rubellum. This study has, to our knowledge, a larger sample size and a larger number of species than any previous study on Sphagnum spectra.

T A X O N O M Y C L A S S I F I C A T I O N
Botany 2 | MATERIAL S AND ME THODS

| Studied species
The Sphagnum moss species included in this study are common in southern boreal peatlands in northern Europe and represent a gradient of peatland nutrient levels. All studied species form homogenous growths that are large enough to be observed by high spatial resolution remote sensing sensors and, thus, are also large enough to be collected as samples for spectral measurements in a laboratory. All studied species form growths of at least 1 m 2 , whereas some species (e.g. Sphagnum fuscum and Sphagnum cuspidatum) can cover even over 10 m 2 areas.

| Sample collection
The Sphagnum moss samples were collected from four peatlands in Uusimaa and Häme areas in southern Finland in May 2022 ( Table 2). The sites were chosen based on the proximity to Aalto University, where the spectral measurements were conducted, to make the transportation time as short as possible. The sampling took place directly after the snow-melting period, that is, all sampling sites were completely waterlogged during the sample collection. Each species was collected within 1 day (S. angustifolium, S. capillifolium, S. centrale, S. fallax, S. girgensohnii, and S. riparium) or two consecutive days (S. cuspidatum, S. fuscum, S. rubellum). The ~460 cm 2 (21.5 × 21.7 cm) sized samples were collected by cutting the sample carefully using a shovel and scissors, and by trimming the dead Sphagnum from the bottom of the sample so that the depth of each sample varied between 5 and 7 cm. After the trimming, the sample was placed in an 8 cm deep measurement container, which had been painted black to prevent light scattering from the walls of the box during the spectral measurements ( Figure 1).
A total of 10 living samples for each species were collected. To prevent multiple samples coming from one growth, the distance between each sampling location in a peatland site was at least 10 m.
After being collected, sample containers were lidded and transported by car to Aalto University. The containers were unlidded promptly after the transportation and were kept open for the rest of the study. Before the spectral measurements, all litter on top of the moss, such as leaves or tree needles, was carefully removed with tweezers. The samples were measured for the first time within 4-6 h from the time they were collected.
Between the spectral measurements, the samples were stored in open containers in a semi-dark room with no direct sunlight and standard temperature (average 20.5°C) and air humidity (average 29%). The samples were kept in the same container all the time to make sure that their structure would not be damaged during or between the spectral measurements.

| Reflectance measurements and preprocessing of spectra
Reflectance measurements of the samples were conducted in a dark laboratory equipped for spectral measurements. The walls, doors, and ceiling of the laboratory were painted with black paint, and the measurement table was covered with optically black fabric. Each sample was measured four times: fresh (0 h), 1 day later (24 h), 2 days later (48 h), and 1 week later (1 week).
The samples were photographed after the first and last spectral measurements.

TA B L E 1
The Sphagnum moss species selected for the study, as well as their nutrient status and habitat, in the descending order of nutrient level. Before each reflectance measurement, the sample was weighed with a standard scale. The dry mass of the samples was determined by weighing them 2 months after the spectral measurements so that they had been allowed to dry completely in the room where they were stored between the measurements. The dry mass was then used to determine the moisture content (Moist %) of each sample at the time of the measurement by first extracting the mass of the empty container and then applying Equation (1):

Nutrient level Description of habitats and habit of growth
where F w is the mass of the sample at the time of the measurement, and D w is the dry mass of the sample.
Nadir-view reflectance measurements were done with a FieldSpec 4 spectroradiometer ( Figure 2) (Analytical Spectral Devices Inc (ASD), serial number 18641), which has a spectral range of 350-2500 nm, and its resolution at 700 nm is 3 nm and at 1400 and 2100 nm is 10 nm. The field of view was 25°. The measurement height was 30 cm, with the optical fiber of the spectroradiometer The raw spectral data consisted of three repetitions of a measurement of an individual sample, which were then averaged as one.
Similarly, at the beginning, end, and every 1 h from the start of a measurement session, white reference and dark current measurements were taken by repeating the measurement thrice and averaging the three measurements as one. The conical-conical reflectance factor (CCRF) (Schaepman-Strub et al., 2006), hereafter called only reflectance, was calculated from them (Equation 2): where I s is the spectrometer reading of the sample, I dc is the reading of the dark current measurement, I wr is the reading of the white reference measurement, and RF wr is the reflectance factor of the white ref-

| Data analysis
First, a mean spectral signature for each species for each measurement time was calculated by averaging the reflectance values over the 10 samples of each species. As the data were tested for normality and found not-normally distributed, a Wilcoxon rank-sum test was applied with a 5% significance level to compare the spectra of the first  (Miller et al., 1990), was calculated for each spectral signature, and the variance of the different species REP position was compared with a Wilcoxon rank-sum test to examine whether REP has statistically significant differences between species.
Next, we analyzed how much of the spectral variation is explained by species or habitat. To achieve this, we used least squares estimation to fit a linear regression model ( To visualize the overall spectral variation between the species and to avoid multicollinearity and reduce the dimensionality in initial spectral data, principal component analysis (PCA) was conducted for the data collected in the first and last measurement times (i.e., 0 h and 1 week). From the PCA results and based on the first three principal components, Euclidean distance was used to calculate how the spectral data could be further grouped into smaller clusters. Here, we applied the hierarchical clustering analysis and dendrogram method. Both the PCA and the hierarchical clustering analysis were done with the R package FactoMineR (Lê et al., 2008). where R is the biconical reflectance factor at wavelengths λ 1 and λ 2 .
The RI values were then fitted into a linear regression model, where the RI value was the response variable and the moisture % was the explanatory variable.

| Reflectance properties of Sphagnum species
The spectral signatures of the fresh Sphagnum mosses exhibited general features typical to green vegetation in the VIS and NIR regions but often had very low reflectance in the SWIR region compared with vascular vegetation (Figure 3). As the moss became drier, it also became whitish in appearance, which resulted in higher reflectance at especially NIR and SWIR wavelength regions (Figure 3). The only species that started to appear whitish before the 1-week measurement was S. riparium, while S. fallax had the highest reflectance of all during the final measurement. Throughout the experiment, S. cuspidatum had the lowest reflectance of all species, being close to zero in VIS and SWIR regions.
For the first 2 days after being collected (i.e., corresponding to three measurement times), the mean reflectance spectra remained similar for all species (Figure 3). However, after drying for 1 week,

| Relating species and habitats with reflectance data
The results of fitted linear model indicated that species explained the variation in the reflectance more than the habitat ( Figure 6).

| Estimating moisture content with reflectance data
The samples' mass and moisture content varied between the species, the mesotrophic species being the lightest and the visually sparsest and the ombrotrophic species being the heaviest as well as visually densest (Table 3). During the experiment, the mesotrophic S.
girgensohnii and S. riparium lost the greatest amounts of their moisture out of all the species and the ombrotrophic species lost less of their moisture content than the other species (Table 3).
Across all study species, the Ratio Index analysis demonstrated that the most informative bands for detecting moisture content of the samples were located in the 1100-1900 nm wavelength region, consistently for all four measurement times ( Figure 9, Table 4). In the 1week measurements, the spectral regions, mainly in SWIR, that were strongly related to moisture content were considerably wider than for the fresh samples. The relationships between the RI and moisture content were strong (R 2 > 0.7) in the λ 1 = 1200-1400 nm, λ 2 = 1000-1200 nm regions throughout the first three measurements (Figure 10).
The regions where the relationship was mainly weaker (R 2 < 0.2) were, for the 0 and 24 h measurements, in the VIS wavelengths and for the 48 h measurement, in the λ 1 = 500-1000 nm, λ 2 = 1900-2500 nm regions. For the 1-week measurement, the regions with high R 2 were generally wider than in the first three measurements, R 2 ranging from 0.5 to 0.7.

| Reflectance properties of Sphagnum
The nine Sphagnum species we analyzed were, to different extents, distinct from each other in their reflectance spectra. They were also visually distinct from the reflectance properties of vascular plants (e.g., Girard et al., 2020;Hovi et al., 2017;Meireles et al., 2020) and lichens (e.g., Kuusinen et al., 2020). Furthermore, the spectral responses the species had for desiccation were, especially for the mesotrophs, characteristic of their respective habitats. Between-species differences of the spectra were the smallest during the 0 h measurements, at mainly VIS wavelengths and generally increasing after 700 nm toward 1400 nm wavelengths. A similar observation was made for the 24 and 48 h measurements.
During the 1-week measurement, the between-species variation was high throughout the whole spectral range, increasing in the SWIR region from 1000 nm onward. We speculate that many of the between-species spectral differences could be due to the different biochemical compositions of the species, yet the lack of literature on Sphagnum mosses' biochemistry does not allow linking typical biochemical and spectral features, similar to what has been done for vascular species in peatlands (Girard et al., 2020). Overall, the reflectance of the species that have been explored in the previous studies, such as S. capillifolium and S. fallax, were similar in shape but a little lower than what Bubier et al. (1997) and Bryant and Baird (2003) have reported. The reflectance spectrum of S. cuspidatum reported by Bryant and Baird (2003) was both lower and of a different shape, suggesting that the samples we measured were considerably wetter.
As our study had a larger sample size per species than previous studies on Sphagnum spectra, we were also able to analyze intraspecific (within-species) variation in the reflectance spectra of different species. Our results showed that, for the fresh samples, intraspecific spectral variation was small, but after 1 week of drying, the variation increased for all the mesotrophic and intermediate species, especially in the NIR and SWIR regions. For the ombrotrophic species, the effects of drying were not as distinctive, which suggests that they are less susceptible to desiccation during a short-term, such as 1 week, drought.

| Relating species and habitats with reflectance data
Species explained over 90% of the variability in spectra of the fresh samples in 1135-1341 nm wavelengths. This underlines the importance of having access to species-specific spectral libraries of Sphagnum mosses, also extending beyond traditional VIS-NIR data to longer NIR or SWIR wavelengths, when mapping species distributions in peatlands using high spatial resolution remote sensing data.
The habitats, on the other hand, explained clearly less of the spectral variation, exceeding 65% explicability at 352-402, 525-543, 1388-1575, 1856-2154, and 2321-2500 nm wavelengths during the 1-week measurement. For the fresh measurements, the habitats usually explained less than 50% of the spectral variation. This is likely due to the high interspecific variation of the ombrotrophic species: they do not resemble each other at any point of the experiment as much as the mesotrophic species resemble each other. During the measurement of fresh samples, the species that had the most similar spectral properties with each other were from the mesotrophic habitats. Conspicuously, after 1 week of drying, the mesotrophic species became whitish in their appearance and resembled each other also spectrally. The ombrotrophs, and to some extent, the oligotrophic S. angustifolium, did not undergo such dramatic natural bleaching during the week of spectral mea- surements. An important mechanism behind this phenomenon was probably that the ombrotrophic species had a higher mass and their surface was much denser in appearance than the mesotrophs'. With less surface area than what the mesotrophs have, the ombrotrophs can restrict their evapotranspiration (Bengtsson et al., 2016;Hájek et al., 2009;Laing et al., 2014). This is also strengthened because of the structure of their pending branches which form the capillary network (Hayward & Clymo, 1983;Rydin, 1985). This is likely due to the less layered structure of their growth environments: the tree-covered mesotrophic habitats provide more shade for the undergrowth mosses than the treeless ombrotrophic habitats (Laing et al., 2014). In our experiment, the species that are more used to the extreme conditions in their natural habitats could uphold their moisture for a week, and the species that came from more stable environments could not.

| Estimating moisture content from reflectance data
The spectral combinations that were strongly associated with the moisture content were, even more notable than in the other analy- The relationships between the moisture content and reflectance data were the strongest after 2 days of drying (R 2 > 0.8). We also observed that when the drought proceeded, the SWIR region most strongly associated with moisture moved toward longer wavelengths ( as EnMAP, PRISMA, and CHIME, will also include SWIR bands, and thus, interpreting data collected by them will need to be based on a fundamental understanding of the spectral properties of peatland surface vegetation not only in VIS and NIR but also in SWIR. The Sphagnum moss spectra (350-2500 nm) measured in this study are publicly available (Salko et al., 2023). In the future, our spectral library can be applied for different purposes, such as in developing new remote sensing methods for mapping boreal Sphagnum mosses.

ACK N OWLED G M ENTS
We thank Dr. Aarne Hovi and Dr. Daniel Schraik for scientific collaboration. We would also like to thank Metsähallitus and the foundations Pääkaupunkiseudun Partiosäätiö and Kai, Eeva ja Aarne Vähäkallion Muistosäätiö for allowing us to conduct research on their peatlands.

FU N D I N G I N FO R M ATI O N
The study was mainly funded by the Academy of Finland (grant: PEATSPEC 3341963). This study has also received funding from No. 771049). The text reflects only the authors' view, and the Agency is not responsible for any use that may be made of the information it contains.

CO N FLI C T O F I NTE R E S T S TATE M E NT
The authors declare there are no competing interests.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data are available in Mendeley Data (https://doi.org/10.17632/ wm5fc xdmzd.1).