Flower Mapping in Grasslands With Drones and Deep Learning

Phenological responses of alpine snowbed communities to advancing snowmelt

Climate change is leading to advanced snowmelt date in alpine regions. Consequently, alpine plant species and ecosystems experience substantial changes due to prolonged phenological seasons, while the responses, mechanisms and implications remain widely unclear. In this 3-year study, we investigated the effects of advancing snowmelt on the phenology of alpine snowbed species. We related microclimatic drivers to species and ecosystem phenology using in situ monitoring and phenocams. We further used predictive modelling to determine whether early snowmelt sites could be used as sentinels for future conditions. Temperature during the snow-free period primarily influenced flowering phenology, followed by snowmelt timing. Salix herbacea and Gnaphalium supinum showed the most opportunistic phenology, while annual Euphrasia minima struggled to complete its phenology in short growing seasons. Phenological responses varied more between years than sites, indicating potential local long-term adaptations and suggesting these species' potential to track future earlier melting dates. Phenocams captured ecosystem-level phenology (start, peak and end of phenological season) but failed to explain species-level variance. Our findings highlight species-specific responses to advancing snowmelt, with snowbed species responding highly opportunistically to changes in snowmelt timings while following species-specific developmental programs. While species from surrounding grasslands may benefit from extended growing seasons, snowbed species may become outcompeted due to internal-clock-driven, non-opportunistic senescence, despite displaying a high level of phenological plasticity.

Grassland vertical height heterogeneity predicts flower and bee diversity: an UAV photogrammetric approach

The ecosystem services offered by pollinators are vital for supporting agriculture and ecosystem functioning, with bees standing out as especially valuable contributors among these insects. Threats such as habitat fragmentation, intensive agriculture, and climate change are contributing to the decline of natural bee populations. Remote sensing could be a useful tool to identify sites of high diversity before investing into more expensive field survey. In this study, the ability of Unoccupied Aerial Vehicles (UAV) images to estimate biodiversity at a local scale has been assessed while testing the concept of the Height Variation Hypothesis (HVH). This hypothesis states that the higher the vegetation height heterogeneity (HH) measured by remote sensing information, the higher the vegetation vertical complexity and the associated species diversity. In this study, the concept has been further developed to understand if vegetation HH can also be considered a proxy for bee diversity and abundance. We tested this approach in 30 grasslands in the South of the Netherlands, where an intensive field data campaign (collection of flower and bee diversity and abundance) was carried out in 2021, along with a UAV campaign (collection of true color-RGB-images at high spatial resolution). Canopy Height Models (CHM) of the grasslands were derived using the photogrammetry technique "Structure from Motion" (SfM) with horizontal resolution (spatial) of 10 cm, 25 cm, and 50 cm. The accuracy of the CHM derived from UAV photogrammetry was assessed by comparing them through linear regression against local CHM LiDAR (Light Detection and Ranging) data derived from an Airborne Laser Scanner campaign completed in 2020/2021, yielding an [Formula: see text] of 0.71. Subsequently, the HH assessed on the CHMs at the three spatial resolutions, using four different heterogeneity indices (Rao's Q, Coefficient of Variation, Berger-Parker index, and Simpson's D index), was correlated with the ground-based flower and bee diversity and bee abundance data. The Rao's Q index was the most effective heterogeneity index, reaching high correlations with the ground-based data (0.44 for flower diversity, 0.47 for bee diversity, and 0.34 for bee abundance). Interestingly, the correlations were not significantly influenced by the spatial resolution of the CHM derived from UAV photogrammetry. Our results suggest that vegetation height heterogeneity can be used as a proxy for large-scale, standardized, and cost-effective inference of flower diversity and habitat quality for bees.

Animal board invited review: Grassland-based livestock farming and biodiversity

Grasslands dominate land cover nationally and globally, and their composition, structure and habitat value are strongly influenced by the actions of domestic and wild grazing animals that feed on them. Different pastures are characterised by varying opportunities for selective feeding by livestock; agronomically improved, sown swards generally consist of a limited range of plant species whereas longer-term leys and semi-natural grasslands are characterised by a more diverse mixture of plants. In the case of botanically diverse permanent pastures/grazing lands, the dietary preferences of different grazers have a more pronounced effect on the botanical composition of the sward in the longer term. Selection of a dominant species within the sward can give less abundant components a chance to compete, increasing community evenness and species richness. Conversely, the selection of minor components reduces sward compositional heterogeneity and hence plant species richness and evenness. Body size, gut type (foregut vs hindgut fermentation), physiological status (growing, pregnant, lactating), metabolic status (extent of body reserves) and environmental conditions all influence the nutrient requirements of a given animal and related foraging priorities. The diet selected is also strongly influenced by the availability of preferred food items, and their vertical and horizontal distribution within the sward. In general, larger animals, such as cattle and horses, are less selective grazers than smaller animals, such as sheep and goats. They are quicker to switch to consuming less-preferred sward components as the availability of preferred resources declines due to their greater forage demands, and as a result can be very effective in controlling competitive plant species consistently avoided by more selective grazers. As a result, low-intensity mixed grazing of cattle and sheep has been shown to improve the diversity and abundance of a range of taxa within grazed ecosystems. Mixed/co-species grazing with different animals exploiting different grassland resources is also associated with increased pasture use efficiency in terms of the use of different sward components and related improvements in nutritional value. In situations where cattle are not available, for example if they are not considered commercially viable, alternative species such as goats, ponies or South American camelids may offer an opportunity to diversify income streams and maintain productive and biodiverse pastures/grazing lands. Stocking rate and timing of grazing also have a considerable role in determining the impact of grazing. Regardless of the species grazing or the pasture grazed, grazing systems are dynamic since selective grazing impacts the future availability of sward components and subsequently dietary choices. New technologies under development provide opportunities to monitor plant/animal interactions more closely and in real time, which will in future support active management to deliver targeted biodiversity gains from specific sites.