No abstract

No abstract

Measuring forest biodiversity using terrestrial surveys is expensive and can only capture common species abundance in large heterogeneous landscapes. In contrast, combining airborne imagery with computer vision can generate individual tree data at the scales of hundreds of thousands of trees. To train computer vision models, ground-based species labels are combined with airborne reflectance data. Due to the difficulty of finding rare species in a large landscape, the majority of classification models only include the most abundant species, leading to biased predictions at broad scales. Extending classification models to include rare species requires targeted data collection and algorithmic improvements to overcome large data imbalances between dominant and rare taxa. In addition, large landscapes often require multiple acquisition events, leading to significant within-species variation in reflectance spectra. Using a multi-temporal hierarchical model, we demonstrate the ability to include species predicted at less than 1 frequency in landscape without losing performance on the dominant species. The final model has over 75% accuracy for 14 species with improved rare species classification compared to a baseline deep learning model. After filtering out dead trees, we generate landscape species maps of individual crowns for over 670,000 individual trees at the Ordway Swisher Biological Station within the National Ecological Observatory Network. We estimate the relative abundance of the species within the landscape and provide three measures of uncertainty to generate a range of counts for each species. These maps provide the first estimates of canopy tree diversity within NEON sites to include rare species and provide a blueprint for capturing tree diversity using airborne computer vision at broad scales.

Accurate field data are essential to understanding ecological systems and forecasting their responses to global change. Yet, data collection errors are common, and data analysis often lags far enough behind its collection that many errors can no longer be corrected, nor can anomalous observations be revisited. Needed is a system in which data quality assurance and control (QA/QC), along with the production of basic data summaries, can be automated immediately following data collection. Here, we implement and test a system to satisfy these needs. For two annual tree mortality censuses and a dendrometer band survey at two forest research sites, we used GitHub Actions continuous integration (CI) to automate data QA/QC and run routine data wrangling scripts to produce cleaned datasets ready for analysis. This system automation had numerous benefits, including (1) the production of near real‐time information on data collection status and errors requiring correction, resulting in final datasets free of detectable errors, (2) an apparent learning effect among field technicians, wherein original error rates in field data collection declined significantly following implementation of the system, and (3) an assurance of computational reproducibility—that is, robustness of the system to changes in code, data and software. By implementing CI, researchers can ensure that datasets are free of any errors for which a test can be coded. The result is dramatically improved data quality, increased skill among field technicians, and reduced need for expert oversight. Furthermore, we view CI implementation as a first step towards a data collection and analysis pipeline that is also more responsive to rapidly changing ecological dynamics, making it better suited to study ecological systems in the current era of rapid environmental change.

Determining mechanisms of plant establishment in ecological communities can be particularly difficult in disturbance-dominated ecosystems. Longleaf pine (Pinus palustris Mill.) and its associated plant community exemplify systems that evolved with disturbances, where frequent, widespread fires alter the population dynamics of longleaf pine within distinct life stages. We identified the primary biotic and environmental conditions that influence the survival of longleaf pine in this disturbance-dominated ecosystem. We combined data from recruitment surveys, tree censuses, dense lidar point clouds, and a forest-wide prescribed fire to examine the response of longleaf pine individuals to fire and biotic neighborhoods. We found that fire temperatures increased with increasing longleaf pine neighborhood basal area and decreased with higher oak densities. There was considerable variation in longleaf pine survival across life stages, with lowest survival probabilities occurring during the bolt stage and not in the earlier, more fire-resistant grass stage. Survival of grass-stage, bolt-stage, and sapling longleaf pines was negatively associated with basal area of neighboring longleaf pine and positively related to neighboring heterospecific tree density, primarily oaks (Quercus spp.). Our findings highlight the vulnerability of longleaf pine across life stages, which suggests optimal fire management strategies for controlling longleaf pine density, and—more broadly—emphasize the importance of fire in mediating species interactions.