The Spectral Species Concept in Living Color.

Biodiversity monitoring is an almost inconceivable challenge at the scale of the entire Earth. The current (and soon to be flown) generation of spaceborne and airborne optical sensors (i.e., imaging spectrometers) can collect detailed information at unprecedented spatial, temporal, and spectral resolutions. These new data streams are preceded by a revolution in modeling and analytics that can utilize the richness of these datasets to measure a wide range of plant traits, community composition, and ecosystem functions. At the heart of this framework for monitoring plant biodiversity is the idea of remotely identifying species by making use of the 'spectral species' concept. In theory, the spectral species concept can be defined as a species characterized by a unique spectral signature and thus remotely detectable within pixel units of a spectral image. In reality, depending on spatial resolution, pixels may contain several species which renders species-specific assignment of spectral information more challenging. The aim of this paper is to review the spectral species concept and relate it to underlying ecological principles, while also discussing the complexities, challenges and opportunities to apply this concept given current and future scientific advances in remote sensing.

Tallo: A global tree allometry and crown architecture database.

Data capturing multiple axes of tree size and shape, such as a tree's stem diameter, height and crown size, underpin a wide range of ecological research-from developing and testing theory on forest structure and dynamics, to estimating forest carbon stocks and their uncertainties, and integrating remote sensing imagery into forest monitoring programmes. However, these data can be surprisingly hard to come by, particularly for certain regions of the world and for specific taxonomic groups, posing a real barrier to progress in these fields. To overcome this challenge, we developed the Tallo database, a collection of 498,838 georeferenced and taxonomically standardized records of individual trees for which stem diameter, height and/or crown radius have been measured. These data were collected at 61,856 globally distributed sites, spanning all major forested and non-forested biomes. The majority of trees in the database are identified to species (88%), and collectively Tallo includes data for 5163 species distributed across 1453 genera and 187 plant families. The database is publicly archived under a CC-BY 4.0 licence and can be access from: https://doi.org/10.5281/zenodo.6637599. To demonstrate its value, here we present three case studies that highlight how the Tallo database can be used to address a range of theoretical and applied questions in ecology-from testing the predictions of metabolic scaling theory, to exploring the limits of tree allometric plasticity along environmental gradients and modelling global variation in maximum attainable tree height. In doing so, we provide a key resource for field ecologists, remote sensing researchers and the modelling community working together to better understand the role that trees play in regulating the terrestrial carbon cycle.

Dynamics of a human-modified tropical peat swamp forest revealed by repeat lidar surveys.

Tropical peat swamp forests (PSFs) are globally important carbon stores under threat. In Southeast Asia, 35% of peatlands had been drained and converted to plantations by 2010, and much of the remaining forest had been logged, contributing significantly to global carbon emissions. Yet, tropical forests have the capacity to regain biomass quickly and forests on drained peatlands may grow faster in response to soil aeration, so the net effect of humans on forest biomass remains poorly understood. In this study, two lidar surveys (made in 2011 and 2014) are compared to map forest biomass dynamics across 96 km2 of PSF in Kalimantan, Indonesia. The peatland is now legally protected for conservation, but large expanses were logged under concessions until 1998 and illegal logging continues in accessible portions. It was hypothesized that historically logged areas would be recovering biomass while recently logged areas would be losing biomass. We found that historically logged forests were recovering biomass near old canals and railways used by the concessions. Lidar detected substantial illegal logging activity-579 km of logging canals were located beneath the canopy. Some patches close to these canals have been logged in the 2011-2104 period (i.e. substantial biomass loss) but, on aggregate, these illegally logged regions were also recovering. Unexpectedly, rapid growth was also observed in intact forest that had not been logged and was over a kilometre from the nearest known canal, perhaps in response to greater aeration of surface peat. Comparing these results with flux measurements taken at other nearby sites, we find that carbon sequestration in above-ground biomass may have offset roughly half the carbon efflux from peat oxidation. This study demonstrates the power of repeat lidar survey to map fine-scale forest dynamics in remote areas, revealing previously unrecognized impacts of anthropogenic global change.

Individual tree crown delineation and tree species classification with hyperspectral and LiDAR data.

An international data science challenge, called National Ecological Observatory Network-National Institute of Standards and Technology data science evaluation, was set up in autumn 2017 with the goal to improve the use of remote sensing data in ecological applications. The competition was divided into three tasks: (1) individual tree crown (ITC) delineation, for identifying the location and size of individual trees; (2) alignment between field surveyed trees and ITCs delineated on remote sensing data; and (3) tree species classification. In this paper, the methods and results of team Fondazione Edmund Mach (FEM) are presented. The ITC delineation (Task 1 of the challenge) was done using a region growing method applied to a near-infrared band of the hyperspectral images. The optimization of the parameters of the delineation algorithm was done in a supervised way on the basis of the Jaccard score using the training set provided by the organizers. The alignment (Task 2) between the delineated ITCs and the field surveyed trees was done using the Euclidean distance among the position, the height, and the crown radius of the ITCs and the field surveyed trees. The classification (Task 3) was performed using a support vector machine classifier applied to a selection of the hyperspectral bands and the canopy height model. The selection of the bands was done using the sequential forward floating selection method and the Jeffries Matusita distance. The results of the three tasks were very promising: team FEM ranked first in the data science competition in Task 1 and 2, and second in Task 3. The Jaccard score of the delineated crowns was 0.3402, and the results showed that the proposed approach delineated both small and large crowns. The alignment was correctly done for all the test samples. The classification results were good (overall accuracy of 88.1%, kappa accuracy of 75.7%, and mean class accuracy of 61.5%), although the accuracy was biased toward the most represented species.

How to map forest structure from aircraft, one tree at a time.

Forest structure is strongly related to forest ecology, and it is a key parameter to understand ecosystem processes and services. Airborne laser scanning (ALS) is becoming an important tool in environmental mapping. It is increasingly common to collect ALS data at high enough point density to recognize individual tree crowns (ITCs) allowing analyses to move beyond classical stand-level approaches. In this study, an effective and simple method to map ITCs, and their stem diameter and aboveground biomass (AGB) is presented. ALS data were used to delineate ITCs and to extract ITCs' height and crown diameter; then, using newly developed allometries, the ITCs' diameter at breast height (DBH) and AGB were predicted. Gini coefficient of DBHs was also predicted and mapped aggregating ITCs predictions. Two datasets from spruce dominated temperate forests were considered: one was used to develop the allometric models, while the second was used to validate the methodology. The proposed approach provides accurate predictions of individual DBH and AGB (R2 = .85 and .78, respectively) and of tree size distributions. The proposed method had a higher generalization ability compared to a standard area-based method, in particular for the prediction of the Gini coefficient of DBHs. The delineation method used detected more than 50% of the trees with DBH >10 cm. The detection rate was particularly low for trees with DBH below 10 cm, but they represent a small amount of the total biomass. The Gini coefficient of the DBH distribution was predicted at plot level with R2 = .46. The approach described in this work, easy applicable in different forested areas, is an important development of the traditional area-based remote sensing tools and can be applied for more detailed analysis of forest ecology and dynamics.

Topography shapes the structure, composition and function of tropical forest landscapes.

Topography is a key driver of tropical forest structure and composition, as it constrains local nutrient and hydraulic conditions within which trees grow. Yet, we do not fully understand how changes in forest physiognomy driven by topography impact other emergent properties of forests, such as their aboveground carbon density (ACD). Working in Borneo - at a site where 70-m-tall forests in alluvial valleys rapidly transition to stunted heath forests on nutrient-depleted dip slopes - we combined field data with airborne laser scanning and hyperspectral imaging to characterise how topography shapes the vertical structure, wood density, diversity and ACD of nearly 15 km2 of old-growth forest. We found that subtle differences in elevation - which control soil chemistry and hydrology - profoundly influenced the structure, composition and diversity of the canopy. Capturing these processes was critical to explaining landscape-scale heterogeneity in ACD, highlighting how emerging remote sensing technologies can provide new insights into long-standing ecological questions.

Allometric equations for integrating remote sensing imagery into forest monitoring programmes.

Remote sensing is revolutionizing the way we study forests, and recent technological advances mean we are now able - for the first time - to identify and measure the crown dimensions of individual trees from airborne imagery. Yet to make full use of these data for quantifying forest carbon stocks and dynamics, a new generation of allometric tools which have tree height and crown size at their centre are needed. Here, we compile a global database of 108753 trees for which stem diameter, height and crown diameter have all been measured, including 2395 trees harvested to measure aboveground biomass. Using this database, we develop general allometric models for estimating both the diameter and aboveground biomass of trees from attributes which can be remotely sensed - specifically height and crown diameter. We show that tree height and crown diameter jointly quantify the aboveground biomass of individual trees and find that a single equation predicts stem diameter from these two variables across the world's forests. These new allometric models provide an intuitive way of integrating remote sensing imagery into large-scale forest monitoring programmes and will be of key importance for parameterizing the next generation of dynamic vegetation models.

Tree-centric mapping of forest carbon density from airborne laser scanning and hyperspectral data.

Forests are a major component of the global carbon cycle, and accurate estimation of forest carbon stocks and fluxes is important in the context of anthropogenic global change. Airborne laser scanning (ALS) data sets are increasingly recognized as outstanding data sources for high-fidelity mapping of carbon stocks at regional scales.We develop a tree-centric approach to carbon mapping, based on identifying individual tree crowns (ITCs) and species from airborne remote sensing data, from which individual tree carbon stocks are calculated. We identify ITCs from the laser scanning point cloud using a region-growing algorithm and identifying species from airborne hyperspectral data by machine learning. For each detected tree, we predict stem diameter from its height and crown-width estimate. From that point on, we use well-established approaches developed for field-based inventories: above-ground biomasses of trees are estimated using published allometries and summed within plots to estimate carbon density.We show this approach is highly reliable: tests in the Italian Alps demonstrated a close relationship between field- and ALS-based estimates of carbon stocks (r2 = 0·98). Small trees are invisible from the air, and a correction factor is required to accommodate this effect.An advantage of the tree-centric approach over existing area-based methods is that it can produce maps at any scale and is fundamentally based on field-based inventory methods, making it intuitive and transparent. Airborne laser scanning, hyperspectral sensing and computational power are all advancing rapidly, making it increasingly feasible to use ITC approaches for effective mapping of forest carbon density also inside wider carbon mapping programs like REDD++.

EEG decoding of semantic category reveals distributed representations for single concepts.

Achieving a clearer picture of categorial distinctions in the brain is essential for our understanding of the conceptual lexicon, but much more fine-grained investigations are required in order for this evidence to contribute to lexical research. Here we present a collection of advanced data-mining techniques that allows the category of individual concepts to be decoded from single trials of EEG data. Neural activity was recorded while participants silently named images of mammals and tools, and category could be detected in single trials with an accuracy well above chance, both when considering data from single participants, and when group-training across participants. By aggregating across all trials, single concepts could be correctly assigned to their category with an accuracy of 98%. The pattern of classifications made by the algorithm confirmed that the neural patterns identified are due to conceptual category, and not any of a series of processing-related confounds. The time intervals, frequency bands and scalp locations that proved most informative for prediction permit physiological interpretation: the widespread activation shortly after appearance of the stimulus (from 100 ms) is consistent both with accounts of multi-pass processing, and distributed representations of categories. These methods provide an alternative to fMRI for fine-grained, large-scale investigations of the conceptual lexicon.