Remote sensing can transform the speed, scale, and cost of biodiversity and forestry surveys. Data acquisition currently outpaces the ability to identify individual organisms in high resolution imagery. We outline an approach for identifying tree-crowns in RGB imagery while using a semi-supervised deep learning detection network. Individual crown delineation has been a long-standing challenge in remote sensing and available algorithms produce mixed results. We show that deep learning models can leverage existing Light Detection and Ranging (LIDAR)-based unsupervised delineation to generate trees that are used for training an initial RGB crown detection model. Despite limitations in the original unsupervised detection approach, this noisy training data may contain information from which the neural network can learn initial tree features. We then refine the initial model using a small number of higher-quality hand-annotated RGB images. We validate our proposed approach while using an open-canopy site in the National Ecological Observation Network. Our results show that a model using 434,551 self-generated trees with the addition of 2848 hand-annotated trees yields accurate predictions in natural landscapes. Using an intersection-over-union threshold of 0.5, the full model had an average tree crown recall of 0.69, with a precision of 0.61 for the visually-annotated data. The model had an average tree detection rate of 0.82 for the field collected stems. The addition of a small number of hand-annotated trees improved the performance over the initial self-supervised model. This semi-supervised deep learning approach demonstrates that remote sensing can overcome a lack of labeled training data by generating noisy data for initial training using unsupervised methods and retraining the resulting models with high quality labeled data.

Broad scale remote sensing promises to build forest inventories at unprecedented scales. A crucial step in this process is to associate sensor data into individual crowns. While dozens of crown detection algorithms have been proposed, their performance is typically not compared based on standard data or evaluation metrics. There is a need for a benchmark dataset to minimize differences in reported results as well as support evaluation of algorithms across a broad range of forest types. Combining RGB, LiDAR and hyperspectral sensor data from the USA National Ecological Observatory Network’s Airborne Observation Platform with multiple types of evaluation data, we created a benchmark dataset to assess crown detection and delineation methods for canopy trees covering dominant forest types in the United States. This benchmark dataset includes an R package to standardize evaluation metrics and simplify comparisons between methods. The benchmark dataset contains over 6,000 image-annotated crowns, 400 field-annotated crowns, and 3,000 canopy stem points from a wide range of forest types. In addition, we include over 10,000 training crowns for optional use. We discuss the different evaluation data sources and assess the accuracy of the image-annotated crowns by comparing annotations among multiple annotators as well as overlapping field-annotated crowns. We provide an example submission and score for an open-source algorithm that can serve as a baseline for future methods.

Mapping species through classification of imaging spectroscopy data is facilitating research to understand tree species distributions at increasingly greater spatial scales. Classification requires a dataset of field observations matched to the image, which will often reflect natural species distributions, resulting in an imbalanced dataset with many samples for common species and few samples for less common species. Despite the high prevalence of imbalanced datasets in multiclass species predictions, the effect on species prediction accuracy and landscape species abundance has not yet been quantified. First, we trained and assessed the accuracy of a support vector machine (SVM) model with a highly imbalanced dataset of 20 tropical species and one mixed-species class of 24 species identified in a hyperspectral image mosaic (350–2500 nm) of Panamanian farmland and secondary forest fragments. The model, with an overall accuracy of 62% ± 2.3% and F-score of 59% ± 2.7%, was applied to the full image mosaic (23,000 ha at a 2-m resolution) to produce a species prediction map, which suggested that this tropical agricultural landscape is more diverse than what has been presented in field-based studies. Second, we quantified the effect of class imbalance on model accuracy. Model assessment showed a trend where species with more samples were consistently over predicted while species with fewer samples were under predicted. Standardizing sample size reduced model accuracy, but also reduced the level of species over- and under-prediction. This study advances operational species mapping of diverse tropical landscapes by detailing the effect of imbalanced data on classification accuracy and providing estimates of tree species abundance in an agricultural landscape. Species maps using data and methods presented here can be used in landscape analyses of species distributions to understand human or environmental effects, in addition to focusing conservation efforts in areas with high tree cover and diversity.

Species identification and characterization in tropical environments is an emerging field in tropical remote sensing. Significant efforts are currently aimed at the detection of tree species, of levels of forest successional stages, and the extent of liana occurrence at the top of canopies. In this paper we describe our use of high resolution imagery from the Quickbird Satellite to estimate the flowering population of Tabebuia guayacan trees at Barro Colorado Island (BCI), in Panama. The imagery was acquired on 29 April 2002 and 21 March 2004. Spectral Angle Mapping via a One-Class Support Vector machine was used to detect the presence of 422 and 557 flowering tress in the April 2002 and March 2004 imagery. Of these, 273 flowering trees are common to both dates. This study presents a new perspective on the effectiveness of high resolution remote sensing for monitoring a phenological response and its use as a tool for potential conservation and management of natural resources in tropical environments.

Aim: Tree crowns determine light interception, carbon and water exchange. Thus, understanding the factors causing tree crown allometry to vary at the tree and stand level matters greatly for the development of future vegetation modelling and for the calibration of remote sensing products. Nevertheless, we know little about large‐scale variation and determinants in tropical tree crown allometry. In this study, we explored the continental variation in scaling exponents of site‐specific crown allometry and assessed their relationships with environmental and stand‐level variables in the tropics.Location: Global tropics.Time period: Early 21st century.Major taxa studied: Woody plants.Methods: Using a dataset of 87,737 trees distributed among 245 forest and savanna sites across the tropics, we fitted site‐specific allometric relationships between crown dimensions (crown depth, diameter and volume) and stem diameter using power‐law models. Stand‐level and environmental drivers of crown allometric relationships were assessed at pantropical and continental scales.Results: The scaling exponents of allometric relationships between stem diameter and crown dimensions were higher in savannas than in forests. We identified that continental crown models were better than pantropical crown models and that continental differences in crown allometric relationships were driven by both stand‐level (wood density) and environmental (precipitation, cation exchange capacity and soil texture) variables for both tropical biomes. For a given diameter, forest trees from Asia and savanna trees from Australia had smaller crown dimensions than trees in Africa and America, with crown volumes for some Asian forest trees being smaller than those of trees in African forests.Main conclusions: Our results provide new insight into geographical variability, with large continental differences in tropical tree crown allometry that were driven by stand‐level and environmental variables. They have implications for the assessment of ecosystem function and for the monitoring of woody biomass by remote sensing techniques in the global tropics.

Summary: 1. We present a model to quantify tropical forest structure and explain variance in dynamic rates (growth and mortality) that is computationally simple and can be applied to landscape‐scale forest inventory and, potentially, remote sensing‐derived data.2. The model is a modification of the perfect plasticity approximation (PPA) based on tree allometry, tree locations and sizes. The model quantifies crown area index (CAI) (number of crowns per unit ground area) and assigns trees to crown layers, which determines the expected number of crowns above each tree and thus its light environment.3. The structural model, parameterized and tested for the Barro Colorado Island, Panama 50‐ha forest dynamics plot using data from forest inventories and stereo aerial photographs, reproduces most canopy and understorey structural and dynamic properties. The PPA model worked as well or better than a computationally intensive, spatially explicit model. A single allometry for all trees worked equally well as functional group or species allometries. Models of growth and mortality were always improved by adding crown layers as defined by the PPA model.4. The mean CAI of the 50‐ha plot was 3.1 with low variance. The observed variance was lower than when tree locations were randomized, which drastically lowered the variance in tree density per plot, indicating that there are regulating forces towards a small range of crown area indices.5. Synthesis. A number of simplifying characteristics in structure were uncovered with the PPA structural model applied to a tropical forest: species allometries were not needed despite the high species diversity in the forest; the model worked on a range of plot sizes; and the variance in CAI was surprisingly low, suggesting regulatory mechanisms. The PPA structural model can be used to develop a fully dynamic simulation model for tropical forests. The ability of the simulation model to predict temporal changes in landscape patterns of biomass, dynamic rates, and species and/or functional group composition will provide validation for the partitioning of dynamic rates by crown layers in the PPA structural model.