Integrating National Ecological Observatory Network (NEON) Airborne Remote Sensing and In-Situ Data for Optimal Tree Species Classification

Scholl, Victoria; Cattau, Megan E.; Joseph, Maxwell B.; Balch, Jennifer K.

doi:10.3390/rs12091414

Cited by 33 publications

(35 citation statements)

References 65 publications

(97 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The t-SNE algorithm can project data into two-dimensional or three-dimensional space and uses good visualization to verify the effectiveness of the dataset or algorithm. The t-SNE method was used in various fields as a The proposed two-stage cyclical learning rate method is calculated as shown in Equation (2).…”

Section: T-distributed Stochastic Neighbor Embedding (T-sne) Analysis Methodsmentioning

confidence: 99%

“…At the second stage, using the traingular2 method, the learning-rate cycle was gradually reduced to confine the model results until, finally, the solution stayed at a fixed position with no large swings (see Figure 6). The proposed two-stage cyclical learning rate method is calculated as shown in Equation (2).…”

Section: Cyclical Learning Ratementioning

confidence: 99%

“…Remote sensing images can be used in many technical aspects of scene classification, such as land-cover detection, urban planning, disaster relief, traffic control, etc. Therefore, how to classify large amounts of remote sensing image data covering land areas is an important research topic [1][2][3][4][5][6][7][8].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Remote Sensing Scene Classification and Explanation Using RSSCNet and LIME

Hung

Tseng

2020

Applied Sciences

View full text Add to dashboard Cite

Classification is needed in disaster investigation, traffic control, and land-use resource management. How to quickly and accurately classify such remote sensing imagery has become a popular research topic. However, the application of large, deep neural network models for the training of classifiers in the hope of obtaining good classification results is often very time-consuming. In this study, a new CNN (convolutional neutral networks) architecture, i.e., RSSCNet (remote sensing scene classification network), with high generalization capability was designed. Moreover, a two-stage cyclical learning rate policy and the no-freezing transfer learning method were developed to speed up model training and enhance accuracy. In addition, the manifold learning t-SNE (t-distributed stochastic neighbor embedding) algorithm was used to verify the effectiveness of the proposed model, and the LIME (local interpretable model, agnostic explanation) algorithm was applied to improve the results in cases where the model made wrong predictions. Comparing the results of three publicly available datasets in this study with those obtained in previous studies, the experimental results show that the model and method proposed in this paper can achieve better scene classification more quickly and more efficiently.

show abstract

Section: T-distributed Stochastic Neighbor Embedding (T-sne) Analysis Methodsmentioning

confidence: 99%

Section: Cyclical Learning Ratementioning

confidence: 99%

See 1 more Smart Citation

Remote Sensing Scene Classification and Explanation Using RSSCNet and LIME

Hung

Tseng

2020

Applied Sciences

View full text Add to dashboard Cite

show abstract

“…All rights reserved. Initial efforts have used field data in combination with NEON airborne mapping products to improve remote-sensing based vegetation classifications (e.g., Scholl et al, 2020), infer structures that may influence ecosystem function (LaRue et al, 2019), or to map biodiversity patterns that are difficult to assess from field data (Hakkenberg et al, 2018;Musavi et al, 2017).…”

Section: Accepted Articlementioning

confidence: 99%

Synergies Among Environmental Science Research and Monitoring Networks: A Research Agenda

et al. 2021

View full text Add to dashboard Cite

Many research and monitoring networks in recent decades have provided publicly available data documenting environmental and ecological change, but little is known about the status of efforts to synthesize this information across networks. We convened a working group to assess ongoing and potential cross-network synthesis research and outline opportunities and challenges for the future, focusing on the US-based research network (the US Long-Term Ecological Research network, LTER) and monitoring network (the National Ecological Observatory Network, NEON). LTER-NEON crossnetwork research synergies arise from the potentials for LTER measurements, experiments, models, and observational studies to provide context and mechanisms for interpreting NEON data, and for NEON measurements to provide standardization and broad scale coverage that complement LTER studies. Initial cross-network syntheses at co-located sites in the LTER and NEON networks are addressing six broad topics: how long-term vegetation change influences C fluxes; how detailed remotely sensed data reveal vegetation structure and function; aquatic-terrestrial connections of nutrient cycling; ecosystem response to soil biogeochemistry and microbial processes; population and species responses to environmental change; and disturbance, stability and resilience. This initial study offers exciting potentials for expanded cross-network syntheses involving multiple long-term ecosystem processes at regional or continental scales. These potential syntheses could provide a pathway for the broader scientific community, beyond LTER and NEON, to engage in cross-network science. These examples also apply to many other research and monitoring networks in the US and globally, and can guide scientists and research administrators in promoting broad-scale research that supports resource management and environmental policy. Plain Language Summary Today many research networks and monitoring networks exist in ecology and environmental science. Their complementary designs and publicly available results and data can create powerful synergies. Long-term, hypothesis-based mechanistic research can provide context and JONES ET AL.

show abstract

“…And, their measurement accuracy and data stability are difficult to guarantee compared with point cloud data. Measuring tree parameters through lidar data can realize forest monitoring (including tree species classification [23], forest structure [24], canopy height [25], parameter extraction, etc.) and mapping [26].…”

Section: A Backgroundmentioning

confidence: 99%

Trunk Model Establishment and Parameter Estimation for a Single Tree Using Multistation Terrestrial Laser Scanning

Pan

Zhang

et al. 2020

IEEE Access

View full text Add to dashboard Cite

Parameter estimation and effective tree trunk model establishment are significant to tree biomass measurements and three-dimensional (3D) reconstruction. In this study, we proposed a trunk model establishment and parameter estimation method for a single tree through multistation terrestrial laser scanning (TLS). This approach obtains a more accurate and dense point cloud and yields several improvements. First, to correct the tree position, the angle between the trunk centreline and the vertical line was extracted to normalize the tree point clouds. Subsequently, some tree model parameters were accurately calculated. The tree height was extracted based on individual differences rather than directly adopting the Digital Elevation Model (DEM) model; The B-Spline curve fitting was added to the diameters at breast height (DBH) measure, which better determined the shape of the DBH compared with circle fitting; The CBH was measured by multi-view voxel projection instead of Single-view fixed threshold segmentation, which confirmed the position of the CBH more accurately. To highlight the effectiveness and pertinence of the proposed method, we analysed three plots that the proposed method can achieve more accurate results, where the mean relative error (MRE) of the tree height, DBH and CBH were 1.89%, 1.74% and 2.91% satisfying the accuracy standard specified by the technical indicators (± 5%). The experimental results validate the trunk model accuracy and the effectiveness of the parameter measurements (close to the field measurements). The proposed method is significantly different from the accuracy of the previous method. The proposed method has the best accuracy in Multistation Terrestrial Laser Scanning. Hence the proposed method has outstanding performance in the high-precision extraction of trunk parameters and the construction of trunk models which will perform forest inventory and construction of virtual forest more accurately.

show abstract

Integrating National Ecological Observatory Network (NEON) Airborne Remote Sensing and In-Situ Data for Optimal Tree Species Classification

Cited by 33 publications

References 65 publications

Remote Sensing Scene Classification and Explanation Using RSSCNet and LIME

Remote Sensing Scene Classification and Explanation Using RSSCNet and LIME

Synergies Among Environmental Science Research and Monitoring Networks: A Research Agenda

Trunk Model Establishment and Parameter Estimation for a Single Tree Using Multistation Terrestrial Laser Scanning

Contact Info

Product

Resources

About