Gaussian Process Regression for Forest Attribute Estimation From Airborne Laser Scanning Data

Varvia, Petri; Lähivaara, Timo; Maltamo, Matti; Packalén, Petteri; Seppänen, Aku

doi:10.1109/tgrs.2018.2883495

Cited by 15 publications

(8 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Thus, our findings suggested the use of BART rather than KNN for this specific application; this was in agreement with the study by Varvia et al [57], which reported that the Bayesian method exhibits better RMSE value than KNN in the total basal area and volume estimation. Hence, the BART algorithm seems to be a perfect option in forestry, where heterogeneity and spatial autocorrelation structures exist and might involve inconsistencies in the linearity, homogeneity, and independency of the variables [41].…”

Section: Discussionsupporting

confidence: 93%

“…Machine learning methods are regarded as popular and advanced models among the research communities. Their ability and flexibility of data modelling with a large set of variables together with their learning schemes and control over the non-linearity in the datasets have been tested and proved, especially in vegetation applications [38,57]. However, further investigations in terms of the various challenges and limitations of the ML algorithms is required.…”

Section: Machine Learning Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Comparison of Machine Learning Methods for Mapping the Stand Characteristics of Temperate Forests Using Multi-Spectral Sentinel-2 Data

Ahmadi

Kalantar

Saeidi³

et al. 2020

Remote Sensing

View full text Add to dashboard Cite

The estimation and mapping of forest stand characteristics are vital because this information is necessary for sustainable forest management. The present study considers the use of a Bayesian additive regression trees (BART) algorithm as a non-parametric classifier using Sentinel-2A data and topographic variables to estimate the forest stand characteristics, namely the basal area (m2/ha), stem volume (m3/ha), and stem density (number/ha). These results were compared with those of three other popular machine learning (ML) algorithms, such as generalised linear model (GLM), K-nearest neighbours (KNN), and support vector machine (SVM). A feature selection was done on 28 variables including the multi-spectral bands on Sentinel-2 satellite, related vegetation indices, and ancillary data (elevation, slope, and topographic solar-radiation index derived from digital elevation model (DEM)) and then the most insignificant variables were removed from the datasets by recursive feature elimination (RFE). The study area was a mountainous forest with high biodiversity and an elevation gradient from 26 to 1636 m. An inventory dataset of 1200 sample plots was provided for training and testing the algorithms, and the predictors were fed into the ML models to compute and predict the forest stand characteristics. The accuracies and certainties of the ML models were assessed by their root mean square error (RMSE), mean absolute error (MAE), and R-squared (R2) values. The results demonstrated that BART generated the best basal area and stem volume predictions, followed by GLM, SVM, and KNN. The best RMSE values for both basal area (8.12 m2/ha) and stem volume (29.28 m3/ha) estimation were obtained by BART. Thus, the ability of the BART model for forestry application was established. On the other hand, KNN exhibited the highest RMSE values for all stand variable predictions, thereby exhibiting the least accuracy for this specific application. Moreover, the effectiveness of the narrow Sentinel-2 bands around the red edge and elevation was highlighted for predicting the forest stand characteristics. Therefore, we concluded that the combination of the Sentinel-2 products and topographic variables derived from the PALSAR data used in this study improved the estimation of the forest attributes in temperate forests.

show abstract

Section: Discussionsupporting

confidence: 93%

Section: Machine Learning Methodsmentioning

confidence: 99%

Comparison of Machine Learning Methods for Mapping the Stand Characteristics of Temperate Forests Using Multi-Spectral Sentinel-2 Data

Ahmadi

Kalantar

Saeidi³

et al. 2020

Remote Sensing

View full text Add to dashboard Cite

show abstract

“…We predicted the species-specific volumes using the multivariate Gaussian process (GP) regression [39]. The GP regression is a widely used machine learning method that is related to kriging and support vector machines.…”

Section: E Estimation Methodologymentioning

confidence: 99%

“…The predicted total volume was computed as a sum of the predicted species-specific volumes. The postprocessing of negative predictions was implemented, as presented in detail by Varvia et al [39].…”

Section: E Estimation Methodologymentioning

confidence: 99%

See 1 more Smart Citation

A Comparison of Linear-Mode and Single-Photon Airborne LiDAR in Species-Specific Forest Inventories

Räty

Varvia

Korhonen

et al. 2022

IEEE Trans. Geosci. Remote Sensing

Self Cite

View full text Add to dashboard Cite

Single-photon airborne light detection and ranging (LiDAR) systems provide high-density data from high flight altitudes. We compared single-photon and linear-mode airborne LiDAR for the prediction of species-specific volumes in boreal coniferous-dominated forests. The LiDAR data sets were acquired at different flight altitudes using Leica SPL100 (single-photon, 17 points • m −2 ), Riegl VQ-1560i (linear-mode, 11 points • m −2 ), and Leica ALS60 (linear-mode, 0.6 points • m −2 ) LiDAR systems. Volumes were predicted at the plot-level using Gaussian process regression with predictor variables extracted from the LiDAR data sets and aerial images. Our findings showed that the Leica SPL100 produced a greater mean root-meansquared error (RMSE) value (41.7 m 3 • ha −1 ) than the Leica ALS60 (39.3 m 3 • ha −1 ) in the prediction of species-specific volumes. Correspondingly, the Riegl VQ-1560i (mean RMSE = 33.0 m 3 • ha −1 ) outperformed both the Leica ALS60 and the Leica SPL100. We found that the cumulative distributions of the first echo heights > 1.3 m were rather similar among the data sets, whereas the last echo distributions showed larger differences. We conclude that the Leica SPL100 data set is suitable for area-based LiDAR inventory by tree species although the prediction errors are greater than with data obtained using the modern linear-mode LiDAR, such as Riegl VQ-1560i. Index Terms-Airborne laser scanning, Gaussian processes (GPs), light detection and ranging (LiDAR) intensity, photon-counting LiDAR. I. INTRODUCTION A IRBORNE light detection and ranging (LiDAR) has established its role in forest inventories [1]-[3]. LiDAR systems have developed rapidly during the last decade, and the development has mainly focused on linear-mode LiDAR systems. These conventional linear-mode LiDAR systems can operate in discrete return or full-waveform modes. Discrete return LiDARs record distinct echoes, whereas full-waveform systems can store the continuous shape of both the outgoing Manuscript

show abstract

Forest parameters inversion by mean coherence set from single-baseline PolInSAR data

Thieu

Pham

2021

Advances in Space Research

View full text Add to dashboard Cite

Gaussian Process Regression for Forest Attribute Estimation From Airborne Laser Scanning Data

Cited by 15 publications

References 26 publications

Comparison of Machine Learning Methods for Mapping the Stand Characteristics of Temperate Forests Using Multi-Spectral Sentinel-2 Data

Comparison of Machine Learning Methods for Mapping the Stand Characteristics of Temperate Forests Using Multi-Spectral Sentinel-2 Data

A Comparison of Linear-Mode and Single-Photon Airborne LiDAR in Species-Specific Forest Inventories

Forest parameters inversion by mean coherence set from single-baseline PolInSAR data

Contact Info

Product

Resources

About