FennoTraits: Dataset of plant functional traits and community composition in northern European flora

28 We thank the personnel at the following research stations for their support during our 29 fieldwork and laboratory work: Kilpisjärvi Biological Station and Värriö Subarctic Research 30 Station, Oulanka Research Station, and Kevo Subarctic Research Institute. We thank the 31 University of Helsinki, University of Oulu, and University of Turku, our work would not be 32 possible without these invaluable research stations across northern Finland. We thank Berit 33 Tønsberg Gaski and Kari Anne Bråthen for their support during fieldwork at Máttavárri and 34 giving us access to the Climate-ecological Observatory for Arctic Tundra cabin. We thank 35 Tuuli Rissanen for sharing species lists of the Rásttigáisa study sites with us. We thank 36 Johanna Lehtinen and Miska Luoto for sharing species lists of the Pallas study sites with us. 37 We thank Ian Brown from Stockholm University for helping to establish the Vindelfjällen 38 study design. We thank the 4th Plant Functional Trait Course held in Svalbard 2018, 39 particularly Vigdis Vandvik, Aud H. Halbritter, Brian Maitner, and Brian J. Enquist who 40 taught us how and why to sample plant functional traits. 41 42 Funding 43 PN acknowledges funding from the Research Council of Finland (grant no. 378397; 347558; 44 PROFI8: 365202), Kone Foundation, and Nessling Foundation. JK acknowledges funding 45 from the Research Council of Finland (grant no. 349606; 353218; 370245) and the GeoDoc 46 programme at the University of Helsinki. 47 48 49 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 2 2 Permits 50 Permission to carry out fieldwork was granted by Metsähallitus. 51 52 Conflict of interest 53 The authors have no conflict of interest. 54 55 56

57 We present here FennoTraits, which is a dataset of plant functional trait and community 58 composition data which we collected from Fennoscandia across northern Finland, Norway, and 59 Sweden in 2016-2025. This dataset has 42 049 abundance estimations and 155 794 functional 60 trait observations from 10 traits representing 373 vascular plant species collected from 1 235 61 study sites within seven study areas. The trait measurements consist of size -structural, leaf 62 economic, leaf spectral, and reproductive traits. The species represent the majority of the native 63 vascular plant species that occur at the seven study areas, and many of the species occur in all 64 seven areas across the two biomes and their ecotone: tundra and boreal forests. Each study area 65 has distinct characteristics and a range of habitats: tundra, meadows, wetlands, shrublands, and 66 boreal forests. These areas are under low anthropogenic influence, and many of the sites are 67 within protected areas that are reserved for nature conservation and scientific research. Finally, 68 we provide with this dataset a general description of the main trait patterns and profiles of the 69 northern European flora. 70 71 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 3 3

72 Plant functional traits can reveal the mechanisms driving ecosystem dynamics and species 73 interactions. Functional traits are thus a powerful tool and have emerged into a critical research 74 topic across ecology, biogeography, and environmental sciences (Violle et al. 2007, Diaz et al. 75 2016, Jennifer L. Funk et al. 2017). Plant functional traits are measurable plant characteristics 76 that influence the growth, survival, and reproduction of plants, and also plant interactions with 77 the environment and other organisms (Pérez -Harguindeguy et al. 2013). Therefore, it is 78 essential to understand how and why functional traits vary across and within species 79 (Kemppinen and Niittynen 2022, Laughlin 2024). Variation in functional traits can explain 80 how plants adapt to changing environmental conditions, such as climate change, and how plants 81 contribute to ecosystem services, such as carbon sequestration (Gerlinde B. De Deyn et al. 82 2008, Christiane Roscher et al. 2012, Georges Kunstler et al. 2016). Ultimately, functional 83 diversity is a key component of biodiversity, contributing to ecosystem resilience and stability 84 (Cadotte et al. 2011, Mammola et al. 2021, Carmona et al. 2021). 85 86 Northern European ecosystems are facing rapid warming due to anthropogenic climate change, 87 which is challenging the resilience of these ecosystems and their provided services (Cohen et 88 al. 2014, Rantanen et al. 2022). Northern ecosystems play significant roles in carbon storage 89 regulation and nature-based livelihoods, and they also harbor unique biodiversity (Hobbie et 90 al. 2000, Ford et al. 2021). Northern Europe encompasses Finland, Sweden, and Norway, 91 representing ecosystem diversity ranging from northern boreal forests to sub -Arctic and oro-92 Arctic tundra. These ecosystems are characterised by strong environmental gradients due to 93 their rich geodiversity, providing mosaics of habitats and vegetation types that are adapted to 94 cold and wet climates (Wielgolaski 1975, Austrheim and Eriksson 2001, Kuuluvainen and 95 Aakala 2011). The northern boreal forests are primarily composed of coniferous species 96 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 4 4 Norway spruce (Picea abies) and Scots pine ( Pinus sylvestris). Towards higher latitudes, the 97 coniferous forest shifts to sub -Arctic mountain birch forest ( Betula pubescens ssp. 98 czerepanovii). Towards higher altitudes, the deciduous forest transitions to dwarf -shrub 99 dominated oro -Arctic heath (chiefly, Empetrum nigrum and Betula nana ), and finally, to 100 mostly barren mountain tops. 101 102 Measuring plant functional traits across wide environmental gradients is important for 103 investigating how plants are shaped by environmental change. Ultimately, functional trait data 104 can be leveraged in establishing more effective conservation strategies and restoration efforts, 105 which requires extensive trait measurements (Carlucci et al. 2020). Some plant functional traits 106 are informative but laborious, and in turn, expensive to measure, such as root traits. Whereas, 107 other traits are more cost -efficient to collect, enabling replication of a high number of plant 108 species, communities, and study sites across large gradients. Such cost -efficient traits include 109 plant height, leaf area, specific leaf area (SLA), and leaf dry matter content (LDMC). These 110 are relatively fast and easy to measure, and require only a ruler, scale, scanner, and oven (Figure 111 1) (Pérez-Harguindeguy et al. 2013). Plant height and leaf area represent size -structural traits, 112 and SLA and LDMC represent leaf economic traits, and together these four traits form the two 113 principal trait variation axes globally, and are thus most often used in ecological research (Diaz 114 et al. 2016). 115 116 Here, we present FennoTraits, which is a taxonomically and spatially comprehensive dataset 117 of plant functional trait measurements in Fennoscandia across Finland, Norway, and Sweden 118 (Figure 1). We collected this dataset primarily from protected areas at seven study areas in 119 2016-2025. With this documentation of northern European flora, we aim to advance the 120 understanding of biodiversity, trait variability, and ecological responses across different 121 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 5 5 environmental conditions and large gradients. The dataset includes 42 049 abundance 122 estimations from 373 vascular plant species and 155 794 functional trait observations from 10 123 traits representing 1 235 study sites, allowing analyses on intraspecific trait variability and its 124 responses to a range of environmental gradients. The dataset consists of plant community 125 composition with a nested plot structure, enabling analyses on community structure, diversity, 126 and mechanisms linked to the locally measured plant traits. Lastly, this dataset is unique in the 127 sense that it was collected and processed only by two researchers, maximising data 128 comparability across the many study designs and study areas. We followed the best practices 129 for open and reproducible science in planning, collecting, documenting, and publishing this 130 dataset (Pérez-Harguindeguy et al. 2013, Hampton et al. 2015, Wilkinson et al. 2016, Jenkins 131 et al. 2023). 132 133 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 6 6 134 Figure 1. Sampling protocol and study designs (a -f). We present a comprehensive dataset of 135 plant functional trait measurements that we collected from Fennoscandia across northern 136 Finland, Norway, and Sweden in 2016-2025. 137 138

139 Study areas 140 We collected the data at seven study areas (Figure 1; Table 1; see Supporting information for 141 maps Figures S1 -S7), four in Finland, two in Norway, and one in Sweden. We chose these 142 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 7 7 study areas due to their floristic diversity, accessibility and on -going research collaboration. 143 Many of the study areas are nearby research stations, facilitating both the field and laboratory 144 work. 145 146 Máttavárri area 147 The Máttavárri area is located in northern Norway. In this low -Arctic area, the dominating 148 vegetation type is dwarf shrub tundra. The summits are bare. 149 150 Rásttigáisá area 151 The Rásttigáisá area is located in northern Norway, close to the Norway-Finland border. In this 152 sub-Arctic area, the dominating vegetation type is dwarf shrub tundra. The summits are bare. 153 154 Kilpisjärvi area 155 The Kilpisjärvi area is located in north -western Finland, partly extending to the Salloaivi 156 mountain close to the border of Finland and Norway. In this sub -Arctic area, the dominating 157 vegetation type is dwarf shrub tundra and mountain birch forest. The summits are bare. The 158 area is diverse and heterogeneous with both acidic and calcareous bedrocks. The area has 159 several conservation areas, including Malla Strict Nature Reserve and Saana Nature Reserve. 160 161 Pallas area 162 The Pallas area is located in north -western Finland. In this hilly northern boreal area, the 163 dominating vegetation type is coniferous forest, which forms the tree line. The summits are 164 tundra heath vegetation. Aapa mires are common in the lowlands. The area is within the Pallas-165 Yllästunturi National Park. 166 167 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 8 8 Värriö area 168 The Värriö area is located in north -eastern Finland, on the Finland -Russia border. In this 169 northern boreal area, the dominating vegetation type is coniferous forest, which nearly reaches 170 the mountain tops. The summits are tundra heath vegetation. Aapa mires are common in the 171 lowlands. The area is within the Värriö Strict Nature Reserve. 172 173 Oulanka area 174 The Oulanka area is located in eastern Finland, on the Finland -Russia border. In this boreal 175 area, the dominating vegetation type is mixed forest, which covers the entire area except the 176 open wetlands and small patches of herb-rich deciduous forests. The area is within the Oulanka 177 National Park. 178 179 Vindelfjällen area 180 The Vindelfjällen area is located in northern Sweden, close to the Sweden -Norway border. In 181 this mountain area, the dominating vegetation type is mountain tundra heath and extensive 182 herb-rich snowfields. The summits are bare. The area is within the Vindelfjäll Nature Reserve. 183 184 Table 1. Summary statistics of the study areas. N trait obs = number of unique trait 185 observations; N species = number of unique species in data; N sites = number of unique study 186 sites with full plant community surveyed. 187 Area, acronym Lat, Lon Size (km2) Elevation (m a.s.l.) Sampling years N trait obs N species N sites Máttavárri, MAT 70.3; 29.1 50 196–468 2023 4 649 126 49 Rásttigáisa, 70.0; 26.3 6 420–761 2023 1 988 47 - .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 9 9 RAS Kilpisjärvi, KIL 69.1; 20.8 160 477–1008 2016-2025 130 815 306 771 Pallas, PAL 68.0; 24.1 35 268–558 2024 1 365 54 - Värriö, VAR 67.7; 29.6 30 262–475 2021 3 785 87 47 Oulanka, OUL 66.4; 29.3 100 143–384 2023-2025 10 849 200 103 Vindelfjäll en, VIN 65.8; 15.2 14 931–1408 2021-2023 2 343 73 43 188 Study designs 189 At the seven study areas, we used different study designs, and in the Kilpisjärvi area we had 190 11 different study designs (Figure 1a-f; Table 2). Across all study areas and designs, the plant 191 functional trait and plant community composition data are fully comparable because we used 192 the same methods and observers. Yet, each study design was established for a specific purpose, 193 and therefore, the spatial, environmental, and temporal structures differ among the areas and 194 designs. Where possible, we have provided references to the original study designs for detailed 195 information on the design and related environmental data. We have provided coordinates to the 196 entire dataset with GPS accuracy, and at most sites, we used a high-accuracy Global Navigation 197 Satellite System (chiefly, GeoExplorer GeoXH 6000 Series; Trimble Inc., Sunnyvale, CA, 198 USA; or a comparable device) that provides up to centimeter-scale positioning accuracy. 199 200 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 10 10 For many of the study designs, we used random stratification to select the study sites. We used 201 the random stratification to pre-select a set of candidate sites, maximising the coverage of the 202 main environmental gradients within each study area. The variables that we used for stratifying 203 the environmental space vary among the study areas and designs. We provide references for 204 each design when possible, and in short, the variables included, e.g., total canopy cover, 205 deciduous canopy cover, distance to forest edge, altitude, potential incoming solar radiation, 206 and topographic wetness. 207 208 At many of the study designs, we used a nested plot structure (Figure 1a), unless we mention 209 otherwise (Figure 1b -f). The nested plot structure means that at each site we had three plot 210 scales nested so that the centre of the plots align. The plot scales and forms are: 0.2 m x 0.2 m 211 squared plot, 1.0 m x 1.0 m squared plot, and 2.0 m radius circular plot (Figure 1a). 212 213 In addition to the study designs (Figure 1a -f), we also collected extra leaf samples 214 opportunistically in the field (coded as “EXT” in the leaf trait data). We targeted these efforts 215 to gain more trait data for species that were underrepresented in the functional trait data 216 collected at the studied plots. This means that the plant functional trait data also contains leaf 217 trait measurements from these extra leaf samples, without accompanying height trait 218 measurements or the plant community composition data. 219 220 The study sites and plots are primarily unmanipulated, except for the leaf trait sampling, unless 221 we mention otherwise, for instance, the Kilpisjärvi: Community experiment design. 222 223 Máttavárri: Gradient design 224 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 11 11 In the Máttavárri area, we sampled plant functional trait and plant community composition data 225 in 2023. This design c onsists of 49 study sites, and here, we used the nested plot structure 226 (Figure 1a). 227 228 Rásttigáisá: Gradient design 229 In the Rásttigáisá area, we sampled plant functional trait data in 2023. This design c onsists of 230 49 study sites with 1 m x 1 m study plots (Figure 1b). These data consist only of leaf traits of 231 the most abundant plant species in the plant communities. This means that the data do not 232 include height measurements or the plant community composition data. The original study 233 design is described in detail in Rissanen et al. (2023). 234 235 Kilpisjärvi: Woody plant design 236 In the Kilpisjärvi area, we used a design focusing on woody plant species in tundra. In this 237 design, we measured woody species cover and height in 2016 -2017. This design consists of 238 223 study sites with five study plots, each 1 m x 1 m. In total, we had 1 053 plots which we 239 placed hierarchically, so that at each site, one plot was at the centre of the site and four plots 240 were placed in the four cardinal compass directions five meters from the centre plot (Figure 241 1c). These data consist of species-specific height measurements (median, maximum) and cover 242 percentages of woody plant species. This means that the data do not include other trait 243 measurements or the plant community composition data. The original study design is described 244 in detail in Kemppinen et al. (2021b) and Kemppinen et al. (2018). 245 246 Kilpisjärvi: Saana-Jehkas gradient design 247 In the Kilpisjärvi area, we sampled plant functional trait and plant community composition data 248 in 2017-2025. This design consists of 228 study sites with 1 m x 1 m study plots. We have 249 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 12 12 studied 50 of these sites more intensively, and used here the nested plot structure (Figure 1a) 250 and annual resampling of leaf traits. This design overlaps with the centre plots of the Kilpisjärvi 251 Woody plant design. The original study design is described in detail in Kemppinen et al. 252 (2021b) and Tyystjärvi et al. (2022). 253 254 Kilpisjärvi: Malla gradient design 255 In the Kilpisjärvi area, we sampled plant functional traits and plant community composition in 256 2020-2025. This design consists of 70 study sites, and here, we used the nested plot structure 257 (Figure 1a). The original study design is described in detail in Aalto et al. (2022) and 258 Kemppinen et al. (2023). 259 260 Kilpisjärvi: Rare Arctic design 261 In the Kilpisjärvi area, we used a design focusing on rare Arctic vascular plant species and thus 262 targeted their habitats, such as calcareous heaths and meadows. In this design, we sampled 263 plant functional trait and plant community composition data in 2020-2025. This design consists 264 of 182 study sites, and here, we used the nested plot structure (Figure 1a). 265 266 Kilpisjärvi: Ailakkavaara gradient design 267 In the Kilpisjärvi area, we sampled plant functional trait and plant community composition data 268 in 2021-2025. This design consists of 41 study sites, and here, we used the nested plot structure 269 (Figure 1a). The original study design is described in detail in Aalto et al. (2022) and 270 Kemppinen et al. (2023). 271 272 Kilpisjärvi: Arthropod sampling design 273 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 13 13 In the Kilpisjärvi area, we used a design focusing on arthropods. In this design, we sampled 274 plant functional trait and plant community composition data in 2021. This gradient design 275 consists of 35 study sites, and here, we used the nested plot structure (Figure 1a). The sites 276 were used for sampling arthropods using malaise traps and ground traps, however, these were 277 not placed in the plots. The original study design is described in detail in Peña -Aguilera et al. 278 (2023). 279 280 Kilpisjärvi: Microclimate grid design 281 In the Kilpisjärvi area, we used a design focusing on within-species microclimate relationships 282 of six common tundra vascular plant species. In this design, we sampled plant functional trait 283 data in 2021. This design consists of six study grids with 25 study plots, each 1 m x 1 m. In 284 total, we had 150 plots which we placed in a grid layout, so that in each grid, the 25 plots were 285 placed at six m intervals (Figure 1d). These data consist only of plant functional traits of the 286 six species. This means that the data do not include the plant community composition data. In 287 two plots, the focal species were not present and thus we measured traits from 148 plots. The 288 original study design is described in detail in Kemppinen & Niittynen (2022). 289 290 Kilpisjärvi: Spring design 291 In the Kilpisjärvi area, we used a design focusing on springs. In this design, we sampled plant 292 functional trait and plant community composition data in 2022 -2023. This design consists of 293 32 study sites at or around springs, and here, we used the nested plot structure (Figure 1a). 294 295 Kilpisjärvi: Geodiversity gradient design 296 In the Kilpisjärvi area, we used a design focusing on geodiversity. In this design, we sampled 297 plant functional trait and plant community composition data in 2023-2025. This design consists 298 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 14 14 of 39 study sites at locations with pronounced geomorphological processes (e.g., cryoturbation 299 and fluvial activity), and here, we used the nested plot structure (Figure 1a). 300 301 Kilpisjärvi: Community experiment design 302 In the Kilpisjärvi area, we used a design with a plant community experiment. In this design, 303 we sampled plant functional trait and plant community composition data in 2023, retrieving a 304 baseline data prior to the community manipulations that followed in 2024 and 2025 when trait 305 sampling and community surveys were also repeated. This design consists of 24 replicated 306 study grids (each 2 m x 3 m) with six 1 m x 1 m study plots (Figure 1e). In total, we had 144 307 plots that represent herb-rich meadow vegetation. Within a grid, we had one control plot and 308 the rest had a different manipulation: 1) species with the highest LDMC removed; 2) species 309 with the lowest LDMC removed; 3) tallest species removed; 4) shortest species removed; 5) 310 species removed randomly. We conducted the manipulations in a given plot community by 311 active removals (i.e., cutting the above -ground parts) so that at least half of the total vascular 312 plant cover was removed. The removals were repeated 2-3 times per growing-season. 313 314 Kilpisjärvi: Seasonal monitoring design 315 In the Kilpisjärvi area, we used a design focusing on seasonality of intra-specific trait variation 316 of 11 common tundra vascular plant species. In this design, we sampled leaf functional trait 317 data in 2024. This design consists of three study sites with 10 m -radius circular study plots 318 (Figure 1f). We conducted a weekly sampling for 15 weeks, the entire growing season. These 319 data consist only of the leaf functional traits of the 11 species. This means that the data do not 320 include height measurements or the plant community composition data. The original study 321 design is described in detail in Niittynen et al. (2026). 322 323 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 15 15 Pallas: Gradient design 324 In the Pallas area, we sampled plant functional trait data in 2024. This design consists of 23 325 study sites with 1 m x 1 m study plots (Figure 1b). These data consist only of leaf traits of the 326 most abundant plant species in the plant communities. This means that the data do not include 327 height measurements or the plant community composition data. The larger original study 328 design is described in detail in Lehtinen et al. (2025). 329 330 Värriö: Gradient design 331 In the Värriö area, we sampled plant functional trait and plant community composition data in 332 2021. This design consists of 47 study sites, and here, we used the nested plot structure (Figure 333 1a). The original study design is described in detail in Aalto et al. (2022) and Kemppinen et al. 334 (2023). 335 336 Oulanka: Gradient design 337 In the Oulanka area, we sampled plant functional trait and plant community composition data 338 in 2023 -2025. This design consists of 103 study sites, and here, we used the nested plot 339 structure (Figure 1a). 340 341 Vindelfjällen: Gradient design 342 In the Vindelfjällen area, we sampled plant functional trait and plant community composition 343 data in 2021-2023. This design consists of 43 study sites with 1 m x 1 m study plots (Figure 344 1b). 345 346 Table 2. The data from the Kilpisjärvi study area originated from 11 different study designs. 347 N trait obs = number of unique trait observations; N species = number of unique species in 348 data; N sites = number of unique study sites with trait or abundance data. 349 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 16 16 Study design Acronym Sampling years N trait obs. N species N sites Woody plant design KIL_WOO 2016-2017 9679 21 223 Saana-Jehkas gradient design KIL_MI 2017-2025 16084 141 228 Malla gradient design KIL_MAL 2020-2025 6966 136 70 Rare Arctic design KIL_RA 2020-2025 23862 247 182 Ailakkavaara gradient design KIL_AIL 2021-2025 4874 129 41 Arthropod sampling design KIL_ROS 2021 4665 114 35 Microclimate grid design KIL_ITV 2021 4107 6 148 Spring design KIL_L 2022-2023 4169 128 32 Geodiversity gradient design KIL_X 2023-2025 4451 141 39 Community experiment design KIL_EXP 2023-2025 48184 139 144 Seasonal monitoring design KIL_SEA 2024 1551 11 3 350 Taxonomy 351 We identified the plants to species level always when it was possible. There were only few 352 exceptions: 1) The genus Taraxacum is known for its complex taxonomy, and therefore, we 353 identified it only to genus level; 2) Alchemilla species can be difficult to identify to species 354 level in field, and therefore, in some of our subdatasets it is only at genus level. An exception 355 to Alchemilla spp. is A. alpina, which we always identified at species level, because it is easy 356 to identify due to its separated leaflets. We used the Leipzig Plant Catalogue as the backbone 357 for our taxon nomenclature. We harmonised the taxon names using the lcvplants R package 358 (Freiberg et al. 2020). Hybrids are common in Salix and Carex genera. When we suspected a 359 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 17 17 hybrid, we named the taxon after the most likely pair of species that formed the hybrid. In the 360 dataset, all hybrids are easy to separate using the provided taxonomic rank of each recorded 361 taxon. 362 363 Plant community composition data 364 We collected the plant community composition data from the plots by identifying all vascular 365 plant species and visually estimating their species-specific cover percentages. The sum of cover 366 values within plots may exceed 100 %, as plants often overlap each other. 367 368 In the community data, there are four species (total of 10 occurrences in the plant community 369 composition data) that are classified as sensitive species in Finland. Therefore, we anonymised 370 these species in the entire dataset (i.e., SpeciesA-D) to protect their precise locations in Finland, 371 following the guidelines of the Finnish Biodiversity Information Facility (FinBIF). 372 373 Plant functional trait data 374 We collected the plant functional trait data by following the protocol outlined in Kemppinen 375 & Niittynen (2022) and Niittynen et al. (2026) which are based on the handbook for 376 standardised measurements of plant functional traits (Cornelissen et al. 2003, Pérez -377 Harguindeguy et al. 2013). We collected data on 10 plant functional traits (Table 3), namely, 378 median height, maximum height, reproductive effort, fresh weight, dry weight, leaf area, SLA, 379 LDMC, leaf brightness index (BITM; Equation 1), and leaf greenness index (Excess Green 380 index; ExG; Equation 2). The height traits are measured at plot -level, which means that a 381 species can have multiple height measures per site in those study designs in which we used the 382 nested plot structure. We measured the rest of the traits at species- and site-level, which means 383 that we measured or sampled several individuals from each focal species and pooled the 384 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 18 18 samples by site. Below, we explain our 9 -step plant functional trait protocol in chronological 385 order (Figure 1.1-9). 386 387 Fieldwork 388 We measured plant height (Figure 1.1) with a ruler to centimeter precision (millimeter precision 389 for the shortest plants), recording the median and maximum vegetative heights of the focal 390 species and excluding inflorescences. The median height is a visual estimation of the typical 391 height of the canopy of a given species within the plot. In the nested plot structure, we measured 392 plant heights from the two smaller plots, but not from the largest circular plot size (Figure 1a). 393 394 We estimated reproductive effort (Figure 1.1) by documenting the sexual reproductive effort 395 of the focal species. We used a scale from 0 to 5, where 0 means that we did not observe any 396 signs of sexual reproduction efforts at the study site (i.e., flowers, berries, fresh seed capsules), 397 and 5 means that all individuals showed signs of sexual reproduction efforts. If the individuals 398 were large and hard to separate (e.g., Empetrum nigrum), 5 indicated exceptionally high density 399 of flowers or berries. Therefore, the reproductive effort should be used as an index that 400 indicates the relative intensity of sexual reproductive effort in a form that is comparable across 401 species and sites. We did not estimate reproductive effort for ferns. The reproductive effort 402 data can also be missing in cases when we were not able to determine if sexual reproduction 403 was present. For instance, if the plants were still developing or if the trees were too tall for us 404 to reach. 405 406 We collected leaf samples (Figure 1.2) from or near the plots using the plant community 407 composition data to determine which species to sample at each site. In general, we sampled the 408 entire community at a given site, except for species with a low coverage ≤ 1%. The exceptions 409 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 19 19 to this rule are: 1) all protected species, because we did not sample them in the countries in 410 which they were protected, and 2) specific study designs, where we did not collect any leaf 411 samples (e.g., Kilpisjärvi: Woody plant design) or focused only on a set of focal species (e.g., 412 Kilpisjärvi: Microclimate grid design). We selected fully opened and developed leaves (i.e., 413 not curled) that showed no signs of damage, such as pathogens or herbivory. In general, we 414 sampled one leaf from 3-4 individuals of the focal species at a given study site at each sampling 415 time point. The exceptions to this rule were species with very small leaves, such as Empetrum 416 nigrum. For those small -leaf species, we sampled 10 -15 leaves per individual and three 417 individuals per given study site at each sampling time point. Regarding all species, we pooled 418 the sampled leaves at the species and site level to reduce the workload. Regarding evergreen 419 shrub species, we selected leaves that were from previous years instead of new leaves that had 420 emerged during the sampling season. We transported the leaf samples (or branches of e.g., 421 Empetrum nigrum) from the field to the laboratory in zip -lock bags with a drop of water to 422 keep the leaves fresh or to rehydrate them. In the laboratory, we stored the sample bags at 4°C 423 and we processed the leaves within 48 hours. 424 425 Laboratory work 426 We determined fresh weight (Figure 1.3) by first preparing the leaves by removing the petioles, 427 and then, gently patting them dry from any excess water on their surfaces. Then, we weighed 428 the leaves using a Mettler AE 100 scale (0.0001 g precision) or a comparable scale. 429 430 We scanned fresh leaves (Figure 1.4) right after weighing using a Canon CanoScan LiDE 400 431 scanner (600 dpi) or a comparable scanner. We imaged the adaxial leaf surface, i.e., the sun -432 facing side of the leaves. Some leaves were too large to fit the scanner, so we chopped them 433 before scanning (e.g., Matteuccia struthiopteris). 434 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 20 20 435 We determined dry weight (Figure 1.5-6) by first preparing the leaves by drying them at 70°C 436 for 48 hours using VWR VENTI -Line ovens. Then, we weighed the leaves right after drying 437 them using a Mettler AE 100 scale (0.0001 g precision) or a comparable scale. 438 439 Leaf segmentation 440 We calculated leaf area (Figure 1.7 -8) using the leaf scans. We established the following 441 procedure for leaf segmentation and shadow removal through an iterative optimisation process. 442 The leaf segmentation method with accompanying computer code was published in Niittynen 443 et al. (2026). This process involved applying the spectral index based rules and thresholds to a 444 large dataset of scans across multiple species, visually inspecting the outcomes, and then fine-445 tuning the used indices and threshold parameters until no significant errors or artifacts were 446 observed. The final procedure goes as follows. First, we conducted an initial leaf segmentation 447 by applying thresholds to the blue and red channels, removing pixels in which the values of the 448 blue channel were >180 or the red channel <30. Next, we applied a Normalized Difference 449 Yellowness Index (NDYI, Equation 3) threshold (< 0.13) at the image borders (20 pixels 450 closest to the margins), excluding shadows that may occasionally appear on the image margins. 451 Then, we converted the filtered pixels into polygons, assuming that each polygon represented 452 an individual leaf. We discarded any polygons with an area <200 pixels to eliminate dirt on the 453 scans. We filled any small holes within the leaf polygons which are likely artifacts using the 454 fill_holes function from the smoothr R package (Strimas-Mackey 2025), with a threshold of 1 455 000 pixels. We also excluded polygons at the image borders (<50 pixels to the margins) to 456 remove possible shadows of the margins. 457 458 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 21 21 Next, we used an iterative refinement process on individual leaf polygons, further reducing 459 shadows on the scans. Shadows often occurred around thick leaves, and the previous 460 procedures were insufficient to exclude these shadows. So, first, we buffered each leaf polygon 461 with 10 pixels, and then, we cropped and masked the original RGB image to this buffered area. 462 Next, we determined the position of the leaf, which we were able to determine based on the 463 scanner sensor geometry that systematically casts the shadows on the same side of the leaf 464 scans (towards the upper right corner). We leveraged this to determine the position of the leaf 465 and to estimate potential shadows in a specific direction based on that position. First, we 466 calculated a centerline of the leaf using the centerline R package (Tsyplenkov 2025), involving 467 the creation of a skeleton of the leaf polygon and tracing a path between two 'end points' defined 468 by the maximum and minimum X or Y coordinates (depending on the aspect ratio of the leaf). 469 Then, we were able to estimate the potential shadow locations relative to the orientation of the 470 leaf. Next, we calculated a distance raster to the centerline skeleton towards the direction where 471 shadows were possible. The shadows gradually shift from dark to light in a known direction, 472 so we calculated a focal Pearson correlation coefficient between the distance raster and the blue 473 channel using a moving window of 15 x 15 pixels. Then, we could exclude a pixel as shadow 474 if the correlation coefficient was above 0.5, the Red Chromatic Coordinate index (RCC; 475 Equation 4) exceeded 0.2, and NDYI was less than 0.15. We considered pixels with a blue 476 channel value greater than 200 as white background and excluded them. Then, we refined the 477 leaf margins with a focal majority filter (5x5 kernel). Finally, we used the processed RGB 478 images from the leaf scans to quantify leaf area. 479 480 We calculated leaf brightness index (BITM, Equation 1) and leaf greenness index (ExG, 481 Equation 2; Figure 1.7 -8) using the leaf scans. We calculated colour indices that are used in 482 satellite-based remote sensing of vegetation. These indices can be derived from red, green, and 483 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 22 22 blue wavelengths (Montero et al. 2023). We conducted all routine raster and vector processing 484 using functions from the terra (Hijmans 2022) and sf (Pebesma 2018) R packages. 485 The formulas for the spectral indices were as follows: 486 Equation 1: BITM = ((B² + G² + R²) / 3)^0.5 487 Equation 2: ExG = 2 * G - R - B 488 Equation 3: NDYI = (G - B) / (G + B) 489 Equation 4: RCC = R / (R + G + B) 490 where R represents the red, G the green, and B the blue channel in the scanned RGB images. 491 492 Finally, we quantified specific leaf area (SLA; Figure 1.7 -8) by calculating the ratio between 493 leaf area and dry weight. We quantified leaf dry matter content (LDMC; Figure 1.7 -8) by 494 calculating the ratio between dry weight and fresh weight. 495 496 Table 3. Summary of the 10 plant functional traits and their prevalence in the data. 497 Trait Unit Explanation N trait obs. N species N sites median height cm Estimated median vegetative height (inflorescence excluded) of a focal species within a plot. 27201 327 1157 maximum height cm Maximum vegetative height (inflorescence excluded) of a focal species within a plot. 27200 327 1157 reproductive effort unitless Intensity of the sexual reproduction effort of a species at a study site. From 0 to 5, 0 indicating no signs of sexual 18653 325 754 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 23 23 reproduction effort (flowers, berries, fruits), 3 indicating that half of the individuals show efforts of sexual reproduction, and 5 indicating that all individuals show efforts on sexual reproduction. fresh weight g Water-saturated fresh mass of the leaf 11814 350 1039 dry weight g Dry mass of the oven-dried leaf 11805 350 1039 leaf area cm2 Area of the leaf derived from scanned leaf images taken of the fresh leaf 11826 350 1040 specific leaf area, SLA mm²/mg⁻¹ Specific leaf area (SLA) is the ratio of dry weight and leaf area. 11809 350 1039 leaf dry matter content, LDMC g/g Leaf dry-matter content (LDMC) is the oven-dry mass of a leaf, divided by its water- saturated fresh mass 11814 350 1039 Brightness index, BITM unitless An spectral index about the brightness of the upper side of the leaf (Equation 1) 11836 350 1040 Excess Green Index, ExG unitless A spectral index indicating the greenness of the upper side of the leaf (Equation 2) 11836 350 1040 498 Data 499 We provide here a general description of the main trait profiles of the northern European flora 500 and most frequent species in the dataset (Figure 2). 501 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 24 24 502 Figure 2. Distributions of the nine continuous traits across the seven study areas (a). The most 503 frequent species in the plant functional trait data colored by functional group (b). 504 505 Data and code availability 506 We provide the dataset in the supporting information of this article. Up -to-date version of the 507 dataset will be maintained and openly available at a GitHub repository that we will link here 508 after acceptance for publishing. Stable versions of the future dataset with Digital Object 509 Identifiers (DOI) will be published in the Zenodo repository annually after major updates. 510 511 Dataset structure and data dictionary 512 We provide a dataset that consists of five files: three data tables as text files, one metadata file 513 as OpenDocument Spreadsheet, and an R script. 514 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 25 25 515 1) The file FennoTraits_community_heights.csv includes the plant community 516 composition data and plant height data (Table 4). These data are at plot-level. 517 2) The file FennoTraits_leaf_traits.csv includes the leaf trait data and reproductive effort 518 data (Table 5). These data are at site-level. 519 3) The file FennoTraits_species.csv is the lookup table for the observed or sampled taxa 520 with higher taxonomy and sampling summary statistics (Table 6). 521 4) The file FennoTraits_metadata.ods includes the common metadata in three 522 spreadsheets. This is the data dictionary of the full dataset, which includes the 523 information in Tables 4-6. 524 5) The file FennoTraits_combine_data.R is an R script to facilitate the use of the dataset. 525 526 Table 4. Data dictionary for the plant community composition data and plant height data. 527 Variable Type Description Units / Values site character Unique study site identifier, matching site in the other files — design character Study design the observation belongs to — area character Abbreviation of the study area — plot_type character Type of vegetation plot a = 20cm x 20cm; b = 1m x 1m; c = circular plot with 2m radius full_commu nity logical Whether the record represents the full plant community (TRUE) or a focal taxon subset (FALSE) TRUE / FALSE .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 26 26 treatment character Experimental treatment applied to the plot high_LDMC = species with the highest LDMC removed; tall = the tallest species removed; low_LDMC = species with the lowest LDMC removed; random = random subset of species removed; short = the shortest species removed date date Date of observation YYYY-MM-DD year numeric Year of observation — Lat numeric Latitude of the plot in decimal degrees (WGS84) — Lon numeric Longitude of the plot in decimal degrees (WGS84) — country character Country where the observation was made FIN = Finland; NOR = Norway; SWE = Sweden taxon character Taxon name as used in the dataset (may be subspecies, species, genus, or family) — cvr numeric Percentage cover of the taxon in the plot % median_heig ht numeric Median vegetation height of the taxon in the plot cm max_height numeric Maximum vegetation height of the taxon in the plot cm 528 Table 5. Data dictionary for the reproductive effort data and leaf trait data. 529 Variable Type Description Units / Values .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 27 27 site character Unique study site identifier, matching site in the other files — design character Study design the observation belongs to — area character Abbreviation of the study area — treatment character Experimental treatment applied to the plot high_LDMC = species with the highest LDMC removed; tall = the tallest species removed; low_LDMC = species with the lowest LDMC removed; random = random subset of species removed; short = the shortest species removed date date Date of leaf sample collection YYYY-MM-DD year numeric Year of sample collection — Lat numeric Latitude of the observation in decimal degrees (WGS84) — Lon numeric Longitude of the observation in decimal degrees (WGS84) — country character Country where the observation was made FIN = Finland; NOR = Norway; SWE = Sweden taxon character Taxon name as used in the dataset (may be subspecies, species, genus, or family) — .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 28 28 reproduction numeric Sexual reproductive effort index of the species at site at year of sampling — n_inds numeric Number of individuals sampled (and pooled) for leaf traits — n_leaf numeric Total (sum) of leaves sampled, pooled and processed for leaf traits — leaf_area numeric One-sided leaf area cm² SLA numeric Specific leaf area (leaf area per unit dry mass) mm² mg⁻¹ LDMC numeric Leaf dry matter content (dry mass per unit fresh mass) g g⁻¹ w_weight numeric Fresh (wet) mass of a leaf g d_weight numeric Dry mass of a leaf after drying g BITM numeric Brightness index of upper side of a plant leaves derived from scanned RGB imagery; formula = ((B² + G² + R²) / 3)^0.5 dimensionless ExG numeric Excess Green Index of upper side of a plant leaves from scanned RGB imagery; formula = (2G − R − B) dimensionless 530 Table 6. Data dictionary for the taxonomic lookup table of the observed or sampled taxa. 531 Variable Type Description Units / Values taxon character Taxon name as used across all dataset files; primary key for joining to the other files — .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 29 29 species character Accepted species name under the LCVP backbone taxonomy — author character Author(s) of the accepted species name — genus character Genus — family character Family — order character Order — rank character Taxonomic rank of the taxon entry sub-species / species / genus / family / hybrid n_areas numeric Number of distinct study areas in which the taxon was recorded — n_locations numeric Number of distinct plot locations at which the taxon was recorded — n_leaf_samples numeric Number of individual leaf trait measurements available for the taxon — 532 Technical validation 533 We conducted a systematic quality control procedure on all plant community composition and 534 plant functional trait observations. In the procedure, we combined visual inspection, 535 biologically motivated hard thresholds, and within -species trait covariance analysis. We 536 designed the procedure to be conservative: instead of removing observations outright based on 537 statistical criteria alone, each step aimed to identify the specific measurement most likely to be 538 erroneous before we set any values as missing. 539 540 First, we produced two sets of diagnostic figures for each species with sufficient data: 1) Per -541 species histograms of all trait distributions, allowing visual identification of extreme isolated 542 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 30 30 values and measurement artefacts, and then 2) Pairwise trait-trait scatterplots with fitted linear 543 trends for all ecologically meaningful trait pairs: log(SLA) against LDMC, log(d_weight) 544 against log(w_weight), log(leaf_area) against log(w_weight), log(leaf_area) against 545 log(d_weight), ExG against BITM, and log(max_height) against log(median_height). We used 546 these plots and species -level summary statistics to identify any suspicious trait observations, 547 which we then manually double -checked using our field notes and raw data, and finally, we 548 corrected those observations if we found a clear source of error, such as a typing error. 549 550 After the visual inspection, we applied a set of hard bounds derived from the known biological 551 and physical constraints of each measured variable. We flagged observations with the 552 following criteria: 1) Cover values outside the range 0.25 –100%; 2) Plant height values 3000 cm; 3) Observations in the plant community composition data, where the maximum 554 height was less than the median height; 4) LDMC values at or outside the bounded interval 555 0−1; 5) SLA values below 200 mm²/mg⁻¹; 6) Leaf area, fresh mass, and dry mass values 556 of ≤0; 7) Observations in the plant functional trait data, where dry mass exceeded fresh mass; 557 8) Image-derived colour indices we checked against their theoretical ranges: −1-1 for ExG and 558 0-1 for BITM. Based on these criteria, we inspected all flagged observations and set the traits 559 to missing when we confirmed that there were errors or if we were not able to locate and correct 560 the source of the error. 561 562 When inspecting the data, we noticed that a simple statistical outlier detection based on 563 univariate methods is poorly suited to our ecological trait data collected along wide 564 environmental gradients. This is because the extreme values often reflect genuine biological 565 variations, and also, because the distributions are typically right -skewed and highly species -566 specific. We therefore based the main automated detection procedure on the within -species 567 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 31 31 covariance structure of functionally related trait pairs. This approach rests on the principle that 568 most of the traits are not independent: leaf wet and dry weight share a near-constant ratio within 569 a species, and leaf area scales predictably with both wet and dry weight. Consequently, their 570 derivatives (.ie., LDMC and SLA) are also typically highly correlated. Severe single -571 observation deviations from these expected relationships within a species are therefore more 572 indicative of measurement error than of true biological variation. 573 574 Next, we focused on four trait pairs: log(d_weight) against log(w_weight), log(leaf_area) 575 against log(d_weight), log(leaf_area) against log(w_weight), and log(SLA) against LDMC. 576 For these pairs, we fitted a robust linear regression per species using Huber M -estimation as 577 implemented in the rlm function of the MASS package (Venables and Ripley 2002). Here, we 578 preferred robust regression over ordinary least squares because the latter is sensitive to the very 579 outliers we seek to detect: a single erroneous observation can pull the fitted line toward itself, 580 reducing its own residual and evading detection. The Huber estimator down -weights 581 observations with large residuals during fitting, so the resulting line reflects the central 582 tendency of the majority of observations and outlying points to receive appropriately large 583 residuals. Because all continuous trait distributions were right -skewed and spanning several 584 orders of magnitude, prior to fitting, we applied log-transformation for all pairs except LDMC 585 against SLA, where LDMC was left on its original scale as a bounded near-symmetric variable. 586 Here, we included only species with ≥10 complete observations for a given pair. We 587 studentised residuals from each robust fit by dividing them with their median absolute deviation 588 (MAD), yielding a dimensionless measure of how many MADs each observation lies from the 589 species-specific regression trend. Then, we retained this MAD -studentised residual as a 590 continuous diagnostic variable for each trait pair. 591 592 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 32 32 Since the three weight and leaf area pairs share raw trait measurements (i.e., leaf area, wet 593 weight, dry weight), we could use the pattern of high residuals across the pairs to identify which 594 specific raw measurement is most likely in error. This is because each of the three raw trait 595 measurements is absent from exactly one of the three pairs: leaf area does not appear in the 596 dry-against-fresh-weight pair, dry weight does not appear in the area-against-fresh-weight pair, 597 and fresh weight does not appear in the area -against-dry-weight pair. Consequently, if a 598 measurement was erroneous, the one pair that did not involve it would show a low residual 599 while the two pairs that did involve it would show high residuals. We used this logic to assign 600 each observation a suspect variable: observations that had a high residual in the leaf area pairs 601 but a low residual in the weight pair indicated a suspicious leaf area; observations that had a 602 high residual in the dry-weight pairs combined with a low residual in the wet-weight-area pair 603 indicated a suspicious dry weight; and the complementary pattern indicated a suspicious fresh 604 weight. If observations did not fit any of these three clean patterns, we classified them as 605 ambiguous and we inspected them manually. We used the SLA–LDMC pair as a confirmatory 606 signal rather than a diagnostic one, since either or both of these derived traits are affected if 607 any of the raw trait measurements is erroneous. We subjected to diagnosis observations with a 608 MAD residual >5 in any of the three raw trait measurement pairs, and subsequently, we set the 609 traits to missing according to the identified suspect variable and its downstream consequences 610 for derived traits: a suspicious leaf area propagated to missing SLA, BITM, and ExG; a 611 suspicious fresh weight propagated to missing LDMC; and a suspicious dry weight propagated 612 to missing SLA and LDMC. Additionally, we observations to missing if they had a MAD 613 residual >8 in the SLA–LDMC pair that had not been resolved by the pattern-based diagnosis, 614 and we did this across all derived trait measurements and raw trait measurements, as deviations 615 of this magnitude indicated a measurement inconsistency too severe to attribute to a single 616 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 33 33 source. In total, we determined 14 leaf samples containing clearly erroneous trait values and 617 we set their particular traits to missing. 618 619 Usage notes 620 Data use and best practice notes 621 We provide this dataset under a CC-BY license. We recommend users of the dataset to cite this 622 data article when referencing and using these data. We encourage contact from users for 623 guidance, advice, and collaboration. We appreciate users contacting us before visiting the study 624 sites due to our long-term monitoring programs. 625 626 Strengths 627 A major strength of this dataset is the high level of operator consistency. This means that all 628 field and laboratory work was conducted by the same two researchers. Therefore, we were able 629 to limit subjectivity in this dataset, reducing errors that may arise from observer bias and 630 maximising comparability across study designs and study areas. 631 632 A second major strength is the nested plot structure that we used in many of the study designs, 633 enabling multi -level analyses. However, this structure also requires careful consideration, 634 because the trait measurements represent different levels of observation. This means that we 635 documented the plant community composition and plant height at the plot -level, whereas, we 636 measured leaf traits and reproductive effort at the site-level. 637 638 Finally, a third key strength is the high spatial accuracy of the recorded study site coordinates. 639 We provide coordinates up to centimetre -scale positioning accuracy, ensuring that the exact 640 same study sites can be reliably relocated in the future, and that the spatial context of the 641 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 34 34 observations incorporated into analyses. High positioning accuracy also enabled us to do true 642 resurveys of the exact same sites, plots, and plant communities in the Kilpisjärvi Saana-Jehkas 643 gradient design, Kilpisjärvi Community experiment design, and Kilpisjärvi Seasonal 644 monitoring design. 645 646 Considerations 647 Temporal context is one the most important factors to consider when using this dataset. We 648 collected the data over more than a decade. This means that different parts of the dataset 649 originate from different years, both within and among study areas and study designs. During 650 our observation and sampling period 2016 –2025, the climatic conditions in northern Europe 651 were extreme at times, including winter warming and snow -on-ice events (Aalto et al. 2026). 652 Moreover, 2024 was likely the warmest growing season in northern Europe in the past 2000 653 years (Rantanen et al. 2025). 654 655 Seasonal dynamics should also be considered. We collected the data during peak growing 656 seasons, except in the Kilpisjärvi Seasonal monitoring design, in which we documented 657 seasonal patterns in leaf traits, see Niittynen et al. (2026). However, growing season dynamics 658 can vary substantially in northern ecosystems. For instance, the onset of the growing season 659 may shift between years and vary across local environmental gradients (Rantanen et al. 2026). 660 661 Land use practices and particularly grazing pressure contribute to variation among study areas 662 and study sites. For example, the study sites in Kilpisjärvi study area are located within three 663 distinct reindeer pastures. Thus, the grazing pressure by the semi -domesticated reindeer 664 (Rangifer tarandus tarandus ) and its spatiotemporal dynamics vary greatly even within the 665 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 35 35 study area. Reindeer grazing affects the plant community composition and likely also plant 666 heights (Olofsson et al. 2009, Maliniemi et al. 2018, Happonen et al. 2019). 667 668 The dataset captures a wide range of environmental variation within study areas, because we 669 selected the study sites within each study area using stratified random sampling. However, this 670 dataset should not be considered an unbiased representation of entire landscapes, and this 671 should be considered especially when calculating species -level average trait values. For 672 example, atypical habitats are overrepresented relative to their true frequency, such as springs 673 in the Kilpisjärvi Spring design and geofeatures in the Kilpisjärvi Geodiversity gradient design. 674 675 Furthermore, we measured vegetative height excluding inflorescences. However, plant height 676 measurements are dependent on the presence of inflorescence in many species. In our study 677 areas, species such as Solidago virgaurea often occurred only with the leaf rosettes close to the 678 ground, but their height measurements were considerably taller, if they produced flowers at the 679 top of a tall shoot with many leaves. 680 681 Regarding leaf traits, we also measured the leaf traits of horsetails (Equisetum spp.). However, 682 using these data calls for consideration because these species do not really have leaves. 683 Therefore, we used the smaller branches as equivalents of leaves for some species, such as E. 684 sylvaticum. Whereas, for the non -brancing species (E. scirpoides, E. variegatum, E. hyemale , 685 and E. fluviatile), we calculated the leaf traits using the whole green shoot. 686 687 Finally, we emphasise that the plant community composition data and the plant height data 688 represents the entire plant community of each site, whereas the leaf trait data does not. This is 689 because we collected leaf samples of the entire community at each site, except for species with 690 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 36 36 a low coverage ≤ 1%. More importantly, we did not sample any protected species, in the 691 countries in which they were protected. Therefore, the leaf trait data do not represent the entire 692 plant community. 693 694

and uncertainties 695 Regarding taxonomy, we consider the main source of uncertainty to be species 696 misidentification. We mitigated this risk through extensive prior experience in species 697 identification of these northern ecosystems (Niittynen et al. 2018, Kemppinen et al. 2021a). In 698 addition, all leaf sample identifications were verified by both of us during laboratory work. 699 Nevertheless, we acknowledge that hybridisation is common in certain groups, such as Salix 700 spp. and Carex spp., and such cases we have explicitly marked as hybrids in the dataset. 701 702 Regarding the plant community composition data, we consider the main limitation to be the 703 visual estimation that we used for estimating the species coverages. We minimised observer 704 bias because all estimates were made by the same two researchers, and we cross-calibrated our 705 estimations. However, visual estimation is inherently more subjective than, for instance, image-706 based analysis or the point-intercept method. As a result, direct comparisons of species cover 707 with future resurveys should be conducted with particular care. 708 709 Regarding plant height, we consider the main source of uncertainty to be the median height 710 which we visually estimated. This means that we first visually inspected the representative 711 median height of the given species in the plot, and then based on this inspection, we measured 712 an individual representative of median height. We chose this approach to reduce workload, but 713 a more robust approach would have involved measuring a larger number of individuals and 714 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 37 37 then calculating the median based on their measured heights. Consequently, additional 715 measurements would be required to obtain more objective estimates of median height. 716 717 Finally, regarding leaf trait data, we consider the main limitation to be that we pooled the leaf 718 samples at site-level. We chose this approach to reduce workload, but a more comprehensive 719 approach would have involved conducting the leaf trait measurements at plot -level and 720 individual-level. Our approach limits the ability to quantify variation within -plot and within-721 population. Consequently, more detailed sampling at finer scales and sampling more local 722 replicates would be required to reduce potential noise arising from inter -individual and intra-723 individual variation which affects the trait averages at plot-level (Maitner et al. 2023). 724 725 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 38 38 Supporting information 726 727 Figure S1. Map of the Máttavárri study area. Colours represent elevation (m a.s.l.) overlaid 728 with hillshade. Black dots represent the study site locations. Water bodies in white. The 729 coordinate system is UTM34N WGS84. 730 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 39 39 731 Figure S2. Map of the Rásttigáisá study area. Colours represent elevation (m a.s.l.) overlaid 732 with hillshade. Black dots represent the study site locations. Water bodies in white. The 733 coordinate system is UTM34N WGS84. 734 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 40 40 735 Figure S3. Map of the Kilpisjärvi study area. Colours represent elevation (m a.s.l.) overlaid 736 with hillshade. Black dots represent the study site locations. Water bodies in white. The 737 coordinate system is UTM34N WGS84. 738 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 41 41 739 Figure S4. Map of the Pallas study area. Colours represent elevation (m a.s.l.) overlaid with 740 hillshade. Black dots represent the study site locations. Water bodies in white. The coordinate 741 system is UTM34N WGS84. 742 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 42 42 743 Figure S5. Map of the Värriö study area. Colours represent elevation (m a.s.l.) overlaid with 744 hillshade. Black dots represent the study site locations. Water bodies in white. The coordinate 745 system is UTM34N WGS84. 746 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 43 43 747 Figure S6. Map of the Oulanka study area. Colours represent elevation (m a.s.l.) overlaid with 748 hillshade. Black dots represent the study site locations. Water bodies in white. The coordinate 749 system is UTM34N WGS84. 750 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 44 44 751 Figure S7. Map of the Vindelfjällen study area. Colours represent elevation (m a.s.l.) overlaid 752 with hillshade. Black dots represent the study site locations. Water bodies in white. The 753 coordinate system is UTM34N WGS84. 754 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 45 45

755 Aalto, J., M. Kämäräinen, M. Rantanen, P. Niittynen, G. K. Phoenix, J. Lenoir, I. 756 Maclean, and M. Luoto. 2026. A new era of bioclimatic extremes in the terrestrial 757 Arctic. Science Advances 12:eadw5698. https://doi.org/10.1126/sciadv.adw5698. 758 Aalto, J., V. Tyystjärvi, P. Niittynen, J. Kemppinen, T. Rissanen, H. Gregow, and M. 759 Luoto. 2022. Microclimate temperature variations from boreal forests to the tundra. 760 Agricultural and Forest Meteorology 323:109037. 761 https://doi.org/10.1016/j.agrformet.2022.109037. 762 Austrheim, G., and O. Eriksson. 2001. Plant Species Diversity and Grazing in the 763 Scandinavian Mountains: Patterns and Processes at Different Spatial Scales. Ecography 764 24:683–695. 765 Cadotte, M. W., K. Carscadden, and N. Mirotchnick. 2011. Beyond species: functional 766 diversity and the maintenance of ecological processes and services. Journal of Applied 767 Ecology 48:1079–1087. https://doi.org/10.1111/j.1365-2664.2011.02048.x. 768 Carlucci, M. B., P. H. S. Brancalion, R. R. Rodrigues, R. Loyola, and M. V. 769 Cianciaruso. 2020. Functional traits and ecosystem services in ecological restoration. 770 Restoration Ecology 28:1372–1383. https://doi.org/10.1111/rec.13279. 771 Carmona, C. P., R. Tamme, M. Pärtel, F. de Bello, S. Brosse, P. Capdevila, R. 772 González-M., M. González-Suárez, R. Salguero-Gómez, M. Vásquez-Valderrama, and 773 A. Toussaint. 2021. Erosion of global functional diversity across the tree of life. Science 774 Advances 7:eabf2675. https://doi.org/10.1126/sciadv.abf2675. 775 Christiane Roscher, Jens Schumacher, Marlén Gubsch, Annett Lipowsky, Alexandra 776 Weigelt, Nina Buchmann, Bernhard Schmid, and Ernst Detlef Schulze. 2012. Using 777 Plant Functional Traits to Explain Diversity–Productivity Relationships. PLOS ONE 7. 778 https://doi.org/10.1371/JOURNAL.PONE.0036760. 779 Cohen, J., J. A. Screen, J. C. Furtado, M. Barlow, D. Whittleston, D. Coumou, J. 780 Francis, K. Dethloff, D. Entekhabi, J. Overland, and J. Jones. 2014. Recent Arctic 781 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 46 46 amplification and extreme mid -latitude weather. Nature Geoscience 7:627 –637. 782 https://doi.org/10.1038/ngeo2234. 783 Cornelissen, J. H. C., S. Lavorel, E. Garnier, S. Díaz, N. Buchmann, D. E. Gurvich, P. 784 B. Reich, H. T. Steege, H. D. Morgan, M. G. A. V. D. Heijden, J. G. Pausas, and H. 785 Poorter. 2003. A handbook of protocols for standardised and easy measurement of plant 786 functional traits worldwide. Australian Journal of Botany 51:335 –380. 787 https://doi.org/10.1071/BT02124. 788 Diaz, S., J. Kattge, J. H. C. Cornelissen, I. J. Wright, S. Lavorel, S. Dray, B. Reu, M. 789 Kleyer, C. Wirth, I. C. Prentice, E. Garnier, G. Bonisch, M. Westoby, H. Poorter, P. B. 790 Reich, A. T. Moles, J. Dickie, A. N. Gillison, A. E. Zanne, J. Chave, S. J. Wright, S. N. 791 Sheremet’ev, H. Jactel, C. Baraloto, B. Cerabolini, S. Pierce, B. Shipley, D. Kirkup, F. 792 Casanoves, J. S. Joswig, A. Gunther, V. Falczuk, N. Ruger, M. D. Mahecha, and L. D. 793 Gorne. 2016. The global spectrum of plant form and function. Nature 529:167-U73. 794 https://doi.org/10.1038/nature16489. 795 Ford, J. D., T. Pearce, I. V. Canosa, and S. Harper. 2021. The rapidly changing Arctic 796 and its societal implications. WIREs Climate Change 12:e735. 797 https://doi.org/10.1002/wcc.735. 798 Freiberg, M., M. Winter, A. Gentile, A. Zizka, A. N. Muellner -Riehl, A. Weigelt, and 799 C. Wirth. 2020. LCVP, The Leipzig catalogue of vascular plants, a new taxonomic 800

list for all known vascular plants. Scientific Data 7:416. 801 https://doi.org/10.1038/s41597-020-00702-z. 802 Georges Kunstler, Georges Kunstler, Daniel S. Falster, David A. Coomes, Francis K. 803 C. Hui, Robert M. Kooyman, Robert M. Kooyman, Daniel C. Laughlin, Lourens 804 Poorter, Mark C. Vanderwel, Ghislain Vieilledent, S. Joseph Wright, Masahiro Aiba, 805 Christopher Baraloto, Christopher Baraloto, John P. Caspersen, J. Hans C. Cornelissen, 806 Sylvie Gourlet -Fleury, Marc Hanewinkel, Bruno Hérault, Jens Kattge, Hiroko 807 Kurokawa, Yusuke Onoda, Josep Peñuelas, Hendrik Poorter, María Uriarte, Sarah J. 808 Richardson, Paloma Ruiz -Benito, Paloma Ruiz -Benito, I -Fang Sun, Göran Ståhl, 809 Nathan G. Swenson, Jill Thompson, Bertil Westerlund, Christian Wirth, Miguel A. 810 Zavala, Hongcheng Zeng, Jess K. Zimmerman, Niklaus E. Zimmermann, and Mark 811 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 47 47 Westoby. 2016. Plant functional traits have globally consistent effects on competition. 812 Nature 529:204–207. https://doi.org/10.1038/NATURE16476. 813 Gerlinde B. De Deyn, Johannes H. C. Cornelissen, and Richard D. Bardgett. 2008. Plant 814 functional traits and soil carbon sequestration in contrasting biomes. Ecology Letters 815 11:516–531. https://doi.org/10.1111/J.1461-0248.2008.01164.X. 816 Hampton, S. E., S. S. Anderson, S. C. Bagby, C. Gries, X. Han, E. M. Hart, M. B. Jones, 817 W. C. Lenhardt, A. MacDonald, W. K. Michener, J. Mudge, A. Pourmokhtarian, M. P. 818 Schildhauer, K. H. Woo, and N. Zimmerman. 2015. The Tao of open science for 819 ecology. Ecosphere 6:1–13. https://doi.org/10.1890/ES14-00402.1. 820 Happonen, K., J. Aalto, J. Kemppinen, P. Niittynen, A. M. Virkkala, and M. Luoto. 821 2019. Snow is an important control of plant community functional composition in 822 oroarctic tundra. Oecologia 191:601–608. https://doi.org/10.1007/s00442-019-04508-823 8. 824 Hijmans, R. J. 2022. terra: Spatial Data Analysis. R package ver. 1.8 -54, 825 https://rspatial.org/. 826 Hobbie, S. E., J. P. Schimel, S. E. Trumbore, and J. R. Randerson. 2000. Controls over 827 carbon storage and turnover in high-latitude soils. Global Change Biology 6:196–210. 828 https://doi.org/10.1046/j.1365-2486.2000.06021.x. 829 Jenkins, G. B., A. P. Beckerman, C. Bellard, A. Benítez‐López, A. M. Ellison, C. G. 830 Foote, A. L. Hufton, M. A. Lashley, C. J. Lortie, Z. Ma, A. J. Moore, S. R. Narum, J. 831 Nilsson, B. O’Boyle, D. B. Provete, O. Razgour, L. Rieseberg, C. Riginos, L. Santini, 832 B. Sibbett, and P. R. Peres‐Neto. 2023. Reproducibility in ecology and evolution: 833 Minimum standards for data and code. Ecology and Evolution 13:e9961. 834 https://doi.org/10.1002/ece3.9961. 835 Jennifer L. Funk, Julie E. Larson, Gregory M. Ames, Bradley J. Butterfield, Jeannine 836 Cavender-Bares, Jennifer Firn, Daniel C. Laughlin, Ariana E. Sutton -Grier, Ariana E. 837 Sutton-Grier, Laura Williams, and Justin P. Wright. 2017. Revisiting the Holy Grail: 838 using plant functional traits to understand ecological processes. Biological Reviews 839 92:1156–1173. https://doi.org/10.1111/BRV.12275. 840 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 48 48 Kemppinen, J., and P. Niittynen. 2022. Microclimate relationships of intraspecific trait 841 variation in sub‐Arctic plants. Oikos 2022:e09507. https://doi.org/10.1111/oik.09507. 842 Kemppinen, J., P. Niittynen, P. C. Le Roux, M. Momberg, K. Happonen, J. Aalto, H. 843 Rautakoski, B. J. Enquist, V. Vandvik, A. H. Halbritter, B. Maitner, and M. Luoto. 844 2021a. Consistent trait –environment relationships within and across tundra plant 845 communities. Nature Ecology & Evolution 5:458–467. https://doi.org/10.1038/s41559-846 021-01396-1. 847 Kemppinen, J., P. Niittynen, H. Riihimaki, and M. Luoto. 2018. Modelling soil 848 moisture in a high -latitude landscape using LiDAR and soil data. Earth Surface 849 Processes and Landforms 43:1019–1031. https://doi.org/10.1002/esp.4301. 850 Kemppinen, J., P. Niittynen, T. Rissanen, V. Tyystjärvi, J. Aalto, and M. Luoto. 2023. 851 Soil Moisture Variations From Boreal Forests to the Tundra. Water Resources Research 852 59:e2022WR032719. https://doi.org/10.1029/2022WR032719. 853 Kemppinen, J., P. Niittynen, A. M. Virkkala, K. Happonen, H. Riihimaki, J. Aalto, and 854 M. Luoto. 2021b. Dwarf Shrubs Impact Tundra Soils: Drier, Colder, and Less Organic 855 Carbon. Ecosystems 24:1378–1392. https://doi.org/10.1007/s10021-020-00589-2. 856 Kuuluvainen, T., and T. Aakala. 2011. Natural forest dynamics in boreal Fennoscandia: 857 a review and classification. Silva Fennica 45. https://doi.org/10.14214/SF.73. 858 Laughlin, D. C. 2024. Unifying functional and population ecology to test the adaptive 859 value of traits. Biological Reviews. https://doi.org/10.1111/brv.13107. 860 Lehtinen, J., J. Aalto, P. Niittynen, and M. Luoto. 2025, July 24. Synoptic atmospheric 861 conditions drive microclimatic variability in a high -latitude landscapes. Research 862 Square. https://doi.org/10.21203/rs.3.rs-6784547/v1. 863 Maitner, B. S., A. H. Halbritter, R. J. Telford, T. Strydom, J. Chacon, C. Lamanna, L. 864 L. Sloat, A. J. Kerkhoff, J. Messier, N. Rasmussen, F. Pomati, E. Merz, V. Vandvik, 865 and B. J. Enquist. 2023. Bootstrapping outperforms community‐weighted approaches 866 for estimating the shapes of phenotypic distributions. Methods in Ecology and 867 Evolution 14:2592–2610. https://doi.org/10.1111/2041-210X.14160. 868 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 49 49 Maliniemi, T., J. Kapfer, P. Saccone, A. Skog, and R. Virtanen. 2018. Long -term 869 vegetation changes of treeless heath communities in northern Fennoscandia: Links to 870 climate change trends and reindeer grazing. Journal of Vegetation Science 29:469–479. 871 https://doi.org/10.1111/jvs.12630. 872 Mammola, S., C. P. Carmona, T. Guillerme, and P. Cardoso. 2021. Concepts and 873 applications in functional diversity. Functional Ecology 35:1869 –1885. 874 https://doi.org/10.1111/1365-2435.13882. 875 Montero, D., C. Aybar, M. D. Mahecha, F. Martinuzzi, M. Söchting, and S. Wieneke. 876 2023. A standardized catalogue of spectral indices to advance the use of remote sensing 877 in Earth system research. Scientific Data 10:197. https://doi.org/10.1038/s41597 -023-878 02096-0. 879 Niittynen, P., R. K. Heikkinen, and M. Luoto. 2018. Snow cover is a neglected driver 880 of Arctic biodiversity loss. Nature Climate Change 8:997 –1001. 881 https://doi.org/10.1038/s41558-018-0311-x. 882 Niittynen, P., M. Merkurjeva, and J. Kemppinen. 2026. Seasonal variation of leaf 883 functional traits in sub -Arctic plants. Oikos (in press):e11849. 884 https://doi.org/10.1002/oik.11849. 885 Olofsson, J., L. Oksanen, T. Callaghan, P. E. Hulme, T. Oksanen, and O. Suominen. 886 2009. Herbivores inhibit climate-driven shrub expansion on the tundra. Global Change 887 Biology 15:2681–2693. https://doi.org/10.1111/j.1365-2486.2009.01935.x. 888 Pebesma, E. 2018. Simple Features for R: Standardized Support for Spatial Vector 889 Data. The R Journal 10:439–446. 890 Peña-Aguilera, P., N. M. Schmidt, L. Stewart, B. Parisy, R. van der Wal, L. Lindman, 891 E. J. Vesterinen, I. M. D. Maclean, T. Kankaanpää, H. Wirta, and T. Roslin. 2023. 892 Consistent imprints of elevation, soil temperature and moisture on plant and arthropod 893 communities across two subarctic landscapes. Insect Conservation and Diversity 894 16:684–700. https://doi.org/10.1111/icad.12667. 895 Pérez-Harguindeguy, N., S. Díaz, E. Garnier, S. Lavorel, H. Poorter, P. Jaureguiberry, 896 M. S. Bret -Harte, W. K. Cornwell, J. M. Craine, D. E. Gurvich, C. Urcelay, E. J. 897 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 50 50 Veneklaas, P. B. Reich, L. Poorter, I. J. Wright, P. Ray, L. Enrico, J. G. Pausas, A. C. 898 de Vos, N. Buchmann, G. Funes, F. Quétier, J. G. Hodgson, K. Thompson, H. D. 899 Morgan, H. ter Steege, M. G. A. van der Heijden, L. Sack, B. Blonder, P. Poschlod, M. 900 V. Vaieretti, G. Conti, A. C. Staver, S. Aquino, and J. H. C. Cornelissen. 2013. New 901 handbook for standardised measurement of plant functional traits worldwide. 902 Australian Journal of Botany 61:167–234. https://doi.org/10.1071/BT12225. 903 Rantanen, M., S. Helama, J. Räisänen, and H. Gregow. 2025. Summer 2024 in northern 904 Fennoscandia was very likely the warmest in 2000 years. npj Climate and Atmospheric 905 Science 8:158. https://doi.org/10.1038/s41612-025-01046-4. 906 Rantanen, M., A. Y. Karpechko, A. Lipponen, K. Nordling, O. Hyvärinen, K. 907 Ruosteenoja, T. Vihma, and A. Laaksonen. 2022. The Arctic has warmed nearly four 908 times faster than the globe since 1979. Communications Earth & Environment 3:1–10. 909 https://doi.org/10.1038/s43247-022-00498-3. 910 Rantanen, M., P. Niittynen, M. Kämäräinen, J. Kemppinen, T. Rissanen, V. Tyystjärvi, 911 M. Luoto, and J. Aalto. 2026, March 30. Substantial spatial heterogeneity in growing 912 season dynamics across northern environments. In Review . 913 https://doi.org/10.21203/rs.3.rs-8955565/v1. 914 Rissanen, T., A. Aalto, H. Kainulainen, O. Kauppi, P. Niittynen, J. Soininen, and M. 915 Luoto. 2023. Local snow and fluvial conditions drive taxonomic, functional and 916 phylogenetic plant diversity in tundra. Oikos 2023:e09998. 917 https://doi.org/10.1111/oik.09998. 918 Strimas-Mackey, M. 2025, June 25. smoothr: Smooth and Tidy Spatial Features. R 919 package ver. 1.2.1, https://CRAN.R-project.org/package=smoothr. 920 Tsyplenkov, A. 2025. centerline: Extract Centerline from Closed Polygons. R package 921 ver. 0.2.5, https://CRAN.R-project.org/package=centerline. 922 Tyystjärvi, V., J. Kemppinen, M. Luoto, T. Aalto, T. Markkanen, S. Launiainen, A. -J. 923 Kieloaho, and J. Aalto. 2022. Modelling spatio -temporal soil moisture dynamics in 924 mountain tundra. Hydrological Processes 36:e14450. 925 https://doi.org/10.1002/hyp.14450. 926 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint 51 51 Venables, W. N., and B. D. Ripley. 2002. Modern Applied Statistics with S. Fourth. 927 Springer, New York. 928 Violle, C., M.-L. Navas, D. Vile, E. Kazakou, C. Fortunel, I. Hummel, and E. Garnier. 929 2007. Let the concept of trait be functional. Oikos 116:882 –892. 930 https://doi.org/10.1111/J.0030-1299.2007.15559.X. 931 Wielgolaski, F. E., editor. 1975. Fennoscandian Tundra Ecosystems. Springer, Berlin, 932 Heidelberg. https://doi.org/10.1007/978-3-642-80937-8. 933 Wilkinson, M. D., M. Dumontier, Ij. J. Aalbersberg, G. Appleton, M. Axton, A. Baak, 934 N. Blomberg, J. -W. Boiten, L. B. Da Silva Santos, P. E. Bourne, J. Bouwman, A. J. 935 Brookes, T. Clark, M. Crosas, I. Dillo, O. Dumon, S. Edmunds, C. T. Evelo, R. Finkers, 936 A. Gonzalez-Beltran, A. J. G. Gray, P. Groth, C. Goble, J. S. Grethe, J. Heringa, P. A. 937 C. ’T Hoen, R. Hooft, T. Kuhn, R. Kok, J. Kok, S. J. Lusher, M. E. Martone, A. Mons, 938 A. L. Packer, B. Persson, P. Rocca -Serra, M. Roos, R. Van Schaik, S. -A. Sansone, E. 939 Schultes, T. Sengstag, T. Slater, G. Strawn, M. A. Swertz, M. Thompson, J. Van Der 940 Lei, E. Van Mulligen, J. Velterop, A. Waagmeester, P. Wittenburg, K. Wolstencroft, J. 941 Zhao, and B. Mons. 2016. The FAIR Guiding Principles for scientific data management 942 and stewardship. Scientific Data 3:160018. https://doi.org/10.1038/sdata.2016.18. 943 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 9, 2026. ; https://doi.org/10.64898/2026.04.07.716889doi: bioRxiv preprint