Continental-scale hyperspectral tree species classification in the United States National Ecological Observatory Network

Marconi, Sergio; Zou, Sheng; Bohlman, Stephanie A.; Zare, Alina; Singh, Aditya; Stewart, Dylan; Harmon, Ira; Steinkraus, Ashley; White, Ethan P.

doi:10.1016/j.rse.2022.113264

Cited by 32 publications

(42 citation statements)

References 93 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Pinaceae) and angiosperms. As these relationships become better understood, remote sensing (Nzimande et al ., 2021; Marconi et al ., 2022) is a promising avenue for correlating tree composition and abundance with bedrock types and scaling tree‐level soil observations and processes to regional CO 2 consumption. A substantial increase in plot‐based tree and soil sampling is also needed in upland areas, particularly in New Guinea.…”

Section: Discussionmentioning

confidence: 99%

Do Southeast Asia's paleo‐Antarctic trees cool the planet?

Wilf

Kooyman

2023

New Phytologist

View full text Add to dashboard Cite

Summary Many tree genera in the Malesian uplands have Southern Hemisphere origins, often supported by austral fossil records. Weathering the vast bedrock exposures in the everwet Malesian tropics may have consumed sufficient atmospheric CO2 to contribute significantly to global cooling over the past 15 Myr. However, there has been no discussion of how the distinctive regional tree assemblages may have enhanced weathering and contributed to this process. We postulate that Gondwanan‐sourced tree lineages that can dominate higher‐elevation forests played an overlooked role in the Neogene CO2 drawdown that led to the Ice Ages and the current, now‐precarious climate state. Moreover, several historically abundant conifers in Araucariaceae and Podocarpaceae are likely to have made an outsized contribution through soil acidification that increases weathering. If the widespread destruction of Malesian lowland forests continues to spread into the uplands, the losses will threaten unique austral plant assemblages and, if our hypothesis is correct, a carbon sequestration engine that could contribute to cooler planetary conditions far into the future. Immediate effects include the spread of heat islands, significant losses of biomass carbon and forest‐dependent biodiversity, erosion of watershed values, and the destruction of tens of millions of years of evolutionary history.

show abstract

Section: Discussionmentioning

confidence: 99%

Do Southeast Asia's paleo‐Antarctic trees cool the planet?

Wilf

Kooyman

2023

New Phytologist

View full text Add to dashboard Cite

show abstract

“…The mean (min, max) of each metric is shown. vegetation structure data alone (Marconi et al, 2022, evaluation accuracy in Table S2), we doubled the species number predicted by incorporating auxiliary, non-NEON data. As a result, 25% of crowns predicted at OSBS in this study were of species not included in previous efforts.…”

Section: Discussionmentioning

confidence: 99%

“…For example, as part of the multisite data competition (Graves et al, 2021), Scholl et al (2021) modeled 27 selected species classes, but only seven of these classes had non-zero evaluation accuracy. Marconi et al (2022) attempted the first NEON-wide model for 77 species across 27 sites using a pixel-based ensemble machine learning classifier. In all of these studies, only the most common species were classified, largely due to insufficient field data on rare species for model development and evaluation.…”

Section: Introductionmentioning

confidence: 99%

Capturing long‐tailed individual tree diversity using an airborne imaging and a multi‐temporal hierarchical model

Marconi

Graves

Zare

et al. 2023

Remote Sens Ecol Conserv

Self Cite

View full text Add to dashboard Cite

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, many classification models only include the most abundant species, leading to biased predictions at broad scales. For example, if only common species are used to train the model, this assumes that these samples are representative across the entire landscape. Extending classification models to include rare species requires targeted data collection and algorithmic improvements to overcome large data imbalances between dominant and rare taxa. We use a targeted sampling workflow to the Ordway Swisher Biological Station within the US National Ecological Observatory Network (NEON), where traditional forestry plots had identified six canopy tree species with more than 10 individuals at the site. Combining iterative model development with rare species sampling, we extend a training dataset to include 14 species. Using a multi‐temporal hierarchical model, we demonstrate the ability to include species predicted at <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 61% accuracy of a baseline deep learning model. After filtering out dead trees, we generate landscape species maps of individual crowns for over 670 000 individual trees. We find distinct patches of forest composed of rarer species at the full‐site scale, highlighting the importance of capturing species diversity in training data. We estimate the relative abundance of 14 species within the landscape and provide three measures of uncertainty to generate a range of counts for each species. For example, we estimate that the dominant species, Pinus palustris accounts for c. 28% of predicted stems, with models predicting a range of counts between 160 000 and 210 000 individuals. These maps provide the first estimates of canopy tree diversity within a NEON site to include rare species and provide a blueprint for capturing tree diversity using airborne computer vision at broad scales.

show abstract

“…Technological and theoretical advances are now linking remote observations to estimates of biodiversity change (e.g. for plant species [110][111][112]). Extensive effort is required to scale-up observations of biodiversity change at specific sites to broader regions [112] and reconstruct historical observations.…”

Section: Next Stepsmentioning

confidence: 99%

“…for plant species [110][111][112]). Extensive effort is required to scale-up observations of biodiversity change at specific sites to broader regions [112] and reconstruct historical observations. New biodiversity indices are being developed to quantify the link between the changing state of biodiversity and the resulting impacts of these changes on ecosystem benefits to people.…”

Section: Next Stepsmentioning

confidence: 99%

A framework for the detection and attribution of biodiversity change

Gonzalez

Chase

O’Connor

2023

Phil. Trans. R. Soc. B

View full text Add to dashboard Cite

The causes of biodiversity change are of great scientific interest and central to policy efforts aimed at meeting biodiversity targets. Changes in species diversity and high rates of compositional turnover have been reported worldwide. In many cases, trends in biodiversity are detected, but these trends are rarely causally attributed to possible drivers. A formal framework and guidelines for the detection and attribution of biodiversity change is needed. We propose an inferential framework to guide detection and attribution analyses, which identifies five steps—causal modelling, observation, estimation, detection and attribution—for robust attribution. This workflow provides evidence of biodiversity change in relation to hypothesized impacts of multiple potential drivers and can eliminate putative drivers from contention. The framework encourages a formal and reproducible statement of confidence about the role of drivers after robust methods for trend detection and attribution have been deployed. Confidence in trend attribution requires that data and analyses used in all steps of the framework follow best practices reducing uncertainty at each step. We illustrate these steps with examples. This framework could strengthen the bridge between biodiversity science and policy and support effective actions to halt biodiversity loss and the impacts this has on ecosystems. This article is part of the theme issue ‘Detecting and attributing the causes of biodiversity change: needs, gaps and solutions’.

show abstract

Continental-scale hyperspectral tree species classification in the United States National Ecological Observatory Network

Cited by 32 publications

References 93 publications

Do Southeast Asia's paleo‐Antarctic trees cool the planet?

Do Southeast Asia's paleo‐Antarctic trees cool the planet?

Capturing long‐tailed individual tree diversity using an airborne imaging and a multi‐temporal hierarchical model

A framework for the detection and attribution of biodiversity change

Contact Info

Product

Resources

About