Vegetation impacts on ecosystem functioning are mediated by mycorrhizas, plant–fungal associations formed by most plant species. Ecosystems dominated by distinct mycorrhizal types differ strongly in their biogeochemistry. Quantitative analyses of mycorrhizal impacts on ecosystem functioning are hindered by the scarcity of information on mycorrhizal distributions. Here we present global, high-resolution maps of vegetation biomass distribution by dominant mycorrhizal associations. Arbuscular, ectomycorrhizal, and ericoid mycorrhizal vegetation store, respectively, 241 ± 15, 100 ± 17, and 7 ± 1.8 GT carbon in aboveground biomass, whereas non-mycorrhizal vegetation stores 29 ± 5.5 GT carbon. Soil carbon stocks in both topsoil and subsoil are positively related to the community-level biomass fraction of ectomycorrhizal plants, though the strength of this relationship varies across biomes. We show that human-induced transformations of Earth’s ecosystems have reduced ectomycorrhizal vegetation, with potential ramifications to terrestrial carbon stocks. Our work provides a benchmark for spatially explicit and globally quantitative assessments of mycorrhizal impacts on ecosystem functioning and biogeochemical cycling.

Invasion by alien species is a worldwide phenomenon with negative consequences at both natural and production areas. Acacia longifolia is an invasive shrub/small tree well known for its negative ecological impacts in several places around the world. The recent introduction of a biocontrol agent (Trichilogaster acaciaelongifoliae), an Australian bud-galling wasp which decreases flowering of A. longifolia, in Portugal, demands the development of a cost-efficient method to monitor its establishment. We tested how Unmanned Aerial Vehicles (UAV) can be used to map A. longifolia flowering. Our core assumption is as the population of the biocontrol agent increases, its impacts on the reduction of A. longifolia flowering will be increasingly visible. Additionally, we tested if there is a simple linear correlation between the number of flowers of A. longifolia counted in field and the area covered by flowers in the UAV imagery. UAV imagery was acquired over seven coastal areas including frontal dunes, interior sand dunes and pine forests considering two phenological stages: peak and off-peak flowering season. The number of flowers of A. longifolia was counted, in a minimum of 60 1m2 quadrats per study area. For each study area, flower presence/absence maps were obtained using supervised Random Forest. The correlation between the number of flowers and the area covered by flowering plants could then be tested. The flowering of A. longifolia was mapped using UAV mounted with RGB and CIR Cannon IXUS/ELPH cameras (Overall Accuracy > 0.96; Cohen’s Kappa > 0.85) varying according to habitat type and flowering season. The correlation between the number of flowers counted and the area covered by flowering was weak (r2 between 0.0134 and 0.156). This is probably explained, at least partially, by the high variability of A. longifolia in what regards flowering morphology and distribution. The very high accuracy of our approach to map A. longifolia flowering proved to be cost efficient and replicable, showing great potential for detecting the future decrease in flowering promoted by the biocontrol agent. The attempt to provide a low-cost method to estimate A. longifolia flower productivity using UAV failed, but it provided valuable insights on the future steps.

Remote sensing (RS) of biophysical variables plays a vital role in providing the information necessary for understanding spatio-temporal dynamics in ecosystems. The hybrid approach to retrieve biophysical variables from RS by combining Machine Learning (ML) algorithms with surrogate data generated by Radiative Transfer Models (RTM). The susceptibility of the ill-posed solutions to noise currently constrains further application of hybrid approaches. Here, we explored how noise affects the performance of ML algorithms for biophysical trait retrieval. We focused on synthetic Sentinel-2 (S2) data generated using the PROSAIL RTM and four commonly applied ML algorithms: Gaussian Processes (GPR), Random Forests (RFR), and Artificial Neural Networks (ANN) and Multi-task Neural Networks (MTN). After identifying which biophysical variables can be retrieved from S2 using a Global Sensitivity Analysis, we evaluated the performance loss of each algorithm using the Mean Absolute Percentage Error (MAPE) with increasing noise levels. We found that, for S2 data, Carotenoid concentrations are uniquely dependent on band 2, Chlorophyll is almost exclusively dependent on the visible ranges, and Leaf Area Index, water, and dry matter contents are mostly dependent on infrared bands. Without added noise, GPR was the best algorithm (<0.05%), followed by the MTN (<3%) and ANN (<5%), with the RFR performing very poorly (20%) even at low levels of added noise (≈5%). Overall, both neural networks performed significantly better than GPR and RFR when noise was added with the MTN being slightly better when compared to the ANN. Our results imply that the performance of the commonly used algorithms in hybrid-RTM inversion are pervasively sensitive to noise. The implication is that more advanced models or approaches are necessary to minimize the impact of noise to improve near real-time and accurate RS monitoring of biophysical trait retrieval.

Invasive plant management is both challenging and expensive and as such strategic and well-informed decisions may contribute for a more effective management. Remote sensing time series can inform decision-making, improving management strategies. Acacia longifolia is one of the most widespread invasive plants in Portuguese coastal areas. We used this species to test a new approach to assess how different management practices and disturbances may influence invasive plants' distribution. The Mann-Kendall statistical test was applied to a 15-year time series (2000-2015) of Landsat TM/ETM+ derived normalized difference vegetation index to detect statistically significant vegetation trends. These maps were interpreted together with a land-cover map derived from a combined rule-based and supervised classification of a Landsat OLI image from November 27th 2013; this interpretation was then complemented with information about management practices and disturbances allowing to identify the processes influencing the current distribution of A. longifolia. The overall accuracy of the classification was 0.785 (Kappa 0.753) and A. longifolia was detected in 9% (8691 ha) of the study area. Of the nine processes of land cover change identified, “vegetation cover intensification” and “vegetation recovery after removal” were the main drivers of expansion. Pine forests were vulnerable to invasion but offered resistance to conversion into A. longifolia monospecific stands, while herbaceous and shrub habitats seemed less resistant to invasion. Some interventions aiming to control A. longifolia were shown to aggravate the invasion. This approach facilitates monitoring the invasion and allows managers to allocate resources to areas where management may be most effective.