Cryptogams (bryophytes and lichens) are ubiquitous non-vascular species that contribute significantly to total biodiversity and play an essential ecological role in ecosystem functioning worldwide. Specifically, cryptogams influence water, carbon and nutrient cycles, as well as physical and chemical weathering, and increase stability of soils, preventing their erosion and regulating their temperature and humidity. Cryptogams facilitate ecosystem recovery following disturbances, and provide microhabitats for micro- and macroorganisms, and a food source for invertebrates and herbivores. These species are also reliable and highly sensitive indicators to environmental disturbances and currently face numerous human-induced threats mainly derived from land use and climate change. Despite this, cryptogams are generally neglected in conservation planning mostly due to current knowledge gaps in their diversity, ecology and distribution, which jeopardizes the maintenance of their species and ecological role. New technologies and data sources such as remote sensing (RS) can significantly help to fill these gaps and ultimately improve the representation of cryptogams in systematic conservation planning. The contribution of RS to cryptogam biodiversity assessments can be particularly valuable in vast and largely unknown regions such as boreal forests, where these species and their habitats face increasing human-induced threats. The general objective of this thesis is to elucidate the role that RS can play in the evaluation and generation of information on cryptogam biodiversity in a boreal context. The study region is located in the Canadian boreal forest, within the Eeyou-Istchee James Bay region in Northern Quebec. As specific objectives, Chapter II aims to predict and map diversity (species richness) patterns of i) total bryophytes, and ii) bryophyte guilds (mosses, liverworts and sphagna) using RS data; Chapter III focusses on producing predictive models of rare bryophyte species using RS-derived predictors in an Ensembles of Small Models (ESMs) framework; and Chapter IV is intended to describe and model the lichen alpha diversity (species richness) and beta diversity (species turnover) components parallelly using two set of RS-derived variables (Red and NIR; EVI2) from two sensors (Wordlview-3, WV3; Sentinel-2, S2) at different high spatial resolutions (1.2m; 10m), and ii) to identify which habitat types represent lichen biodiversity hotspots.
The Random Forest algorithm used in Chapter II allowed us to develop spatially explicit models and to generate predictive cartography at 30m resolution of total bryophyte, moss, liverwort and sphagna richness. These models explained a significant fraction of the variation in total bryophyte and guild level richness, both in the calibration (42 to 52%) and validation sets (38 to 48%), and consistently identified vegetation (mainly NDVI) and climatic variables (temperature, precipitation, and freeze-thaw events) as the most important predictors for all bryophyte groups modeled. Guild-level models identified differences in important factors determining the richness of each of the guilds and thus in their predicted richness patterns, which provide valuable information for management and conservation strategies for bryophytes. The RS-based ESMs developed in Chapter III built from Random Forest and Maxent techniques using predictors related to topography (TPI) and vegetation (EVI2, NDWI1, Vegetation Continuous fields, and PALSAR HVHH) yielded poor to excellent prediction accuracy (AUC > 0.5) for 38 of the 52 modeled species despite their low number of occurrences (< 30), with AUC values > 0.8 for 19 species. The actual presences of the 38 species modeled better than random (AUC ≤ 0.5) were accurately predicted, as supported by the high sensitivity values obtained that ranged from 0.8 to 1 with an average of 0.959 ± 0.063. The distribution of these 38 species and the richness patterns both for total rare bryophytes and rare species at the guild level were mapped at 30m resolution. Chapter III also revealed a spatial concordance between rare (present chapter) and overall bryophyte richness patterns (Chapter II) in different regions of the study area, which has important implications for conservation planning. In Chapter IV, a total of 116 lichen species were identified. While high lichen richness was generally found across our plots (36.5 ± 9 species), those richer in microhabitats often harbored more species (R2 = 0.22) regardless of the habitat type. Differences in species composition were identified among plots (25.6% explained by PCoA) and habitat types (PERMANOVA R2 = 0.35), both being supported by differences in microhabitat composition (Mantel r = 0.22 and PERMANOVA R2 = 0.29, respectively). Rocky outcrops and undisturbed coniferous forests represented the main lichen biodiversity hotspots, while other habitat types were also important for maintaining overall biodiversity. Red and NIR variables were effective for modeling alpha and beta diversity at both resolutions, while EVI2, either from WV3 or S2, was only informative for assessing beta diversity. Poisson models explained up to 32% of the variation in lichen richness. Generalized dissimilarity models described well the relationship between beta diversity and spectral dissimilarity (R2 from 0.25 to 0.30), except for the S2 EVI2 model (R2 = 0.07), confirming that more spectrally and thus environmentally different areas tend to harbor different lichen communities. While WV3 often outperformed the S2 sensor, the latter still provides a powerful tool for the study of lichens and their conservation.
This thesis demonstrated the ability for RS at medium and high spatial resolutions to characterize the habitat of inconspicuous cryptogam species, to capture diverse meaningful ecological features shaping their distribution, and thus to better understand and/or predict their biodiversity patterns. RS-based modeling frameworks proved to be informative even when the available baseline information on cryptogam biodiversity was limited. By identifying environmental drivers of cryptogam biodiversity that can guide specific management actions, and by providing predictive mapping of their spatial patterns at high level of detail across the landscape, this work unequivocally highlighted the high potential of RS technology for conservation purposes of cryptogams. This thesis thus represents a very important step to achieve the inclusion of these inconspicuous and generally overlooked species into systematic conservation planning.