Modelling seasonal dynamics of secondary growth in R

Ecography, Journal Year: 2022, Volume and Issue: 2022(9)

Published: July 20, 2022

The monitoring of seasonal radial growth woody plants addresses the ultimate question when, how and why trees grow. Assessing dynamics is important to quantify effect environmental drivers understand species will deal with ongoing climatic changes. One crucial steps in analyses model xylem phloem formation based on increment measurements samples taken at relatively short intervals during growing season. most common approach use Gompertz equation, while other approaches, such as general additive models (GAMs) generalised linear (GLMs), have also been tested recent years. For first time, we explored artificial neural networks Bayesian regularisation algorithm (BRNNs) show that this method easy use, resistant overfitting, tends yield s‐shaped curves therefore suitable for deriving temporal secondary tree growth. We propose two data processing algorithms allow more flexible fits. main result our work XPSgrowth() function implemented Tree Growth (rTG) R package, can be used evaluate compare three modelling approaches: BRNN, GAM function. newly developed function, intra‐seasonal data, has potential applications many ecological disciplines where expressed a time. Different approaches were evaluated terms prediction error, fitted visually compared derive their characteristics. Our results suggest there no single best fitting method, recommend testing different methods selection optimal one.

Language: Английский

---

Enhancing Height Predictions of Brazilian Pine for Mixed, Uneven-Aged Forests Using Artificial Neural Networks

Forests, Journal Year: 2022, Volume and Issue: 13(8), P. 1284 - 1284

Published: Aug. 13, 2022

Artificial intelligence (AI) seeks to simulate the human ability reason, make decisions, and solve problems. Several AI methodologies have been introduced in forestry reduce costs increase accuracy estimates. We evaluate performance of Neural Networks (ANN) estimating heights Araucaria angustifolia (Bertol.) Kuntze (Brazilian pine) trees. The trees are growing Uneven-aged Mixed Forests (UMF) southern Brazil under different levels competition. dataset was divided into training validation sets. Multi-layer Perceptron (MLP) networks were trained Data Normalization (DN) procedures, Neurons Hidden Layer (NHL), Activation Functions (AF). continuous input variables diameter at breast height (DBH) base crown (HCB). As a categorical variable, we consider sociological position (dominant–SP1 = 1; codominant–SP2 2; dominated–SP3 3), output variable (h). In hidden layer, number neurons varied from 3 9. Results show that there is no influence DN ANN accuracy. However, NHL above certain level caused model’s over-fitting. this regard, around 6 stood out, combined with logistic sigmoid AF intermediate layer identity layer. Considering best selected network, following values statistical criteria obtained for (R2 0.84; RMSE 1.36 m, MAPE 6.29) 0.80; 1.49 6.53). possibility using numerical same modeling has motivating use techniques applications. presented generalization consistency regarding biological realism. Therefore, recommend caution when determining DN, amount NHL, during modeling. argue such great potential forest management procedures suggested other similar environments.

Language: Английский

---

Comparative analysis of machine learning algorithms and statistical models for predicting crown width of Larix olgensis

Earth Science Informatics, Journal Year: 2022, Volume and Issue: 15(4), P. 2415 - 2429

Published: Aug. 5, 2022

Language: Английский

---

Predicting stem taper using artificial neural network and regression models for Scots pine (Pinus sylvestris L.) in northwestern Türkiye

Scandinavian Journal of Forest Research, Journal Year: 2023, Volume and Issue: 38(1-2), P. 97 - 104

Published: Feb. 17, 2023

ABSTRACTStem taper models are helpful tools for predicting diameter of a tree at any height or volume stem section. In this study, traditional and artificial neural network (ANN) approaches were used to predict tapers Scots pine individuals. The data in study correspond destructively sampled trees even-aged forest stands located the three important locations where grows naturally northwestern Türkiye. total, regression type from different categories an ANN model developed evaluated both statistically graphically. best results obtained by Kozak's accounting 99% total variance predictions.KEYWORDS: Artificial intelligenceBayesianmachine learningsegmented modelstem diametertaper modelvariable-form AcknowledgementsI very much appreciate comments associate editor four anonymous reviewers. assistance field collection staff Küre, Taşköprü Yenice Forest Enterprises, Dr. Ferhat Bolat about modeling greatly appreciated.Disclosure statementNo potential conflict interest was reported author(s).Funding statementThe author declares no specific funding work.

Language: Английский

---

Modeling Height—Diameter Relationship Using Artificial Neural Networks for Durango Pine Species in Mexico

Published: July 12, 2023

The total tree height (h) and diameter at breast (dbh) relationship is an essential tool in forest management planning. height—diameter (h-dbh) had been studied with several approaches for species worldwide. nonlinear mixed effect modeling (NLMEM) has extensively used lately the resilient backpropagation artificial neural network (RBPANN) approach a trend topic this relationship. (ANN) computing system based intelligence inspired biological supervised learning. In study NLMEN RBPANN were h—dbh Durango pine (Pinus durangensis Martínez) mixed-species from Mexico. dataset considered 1,000 (11,472 measured trees) randomly selected 14,390 temporary inventory plots was divided into two parts; 50% training testing. An unsupervised clustering analysis to grouped 10 clusters on k-means method plot-variables like density, basal area, mean dbh, h, quadratic diameter, altitude aspect. performed tangent hyperbolicus (RBPANN-tanh), softplus (RBPANN-softplus), logistic (RBPANN-logistic) activation functions cross product of covariate or neurons weights ANN analysis. For both testing, classical statistics (e.g., RMSE, R2, AIC, BIC, logLik) computed residual values assess ANNs outperformed NLMEM approach, RBPANN-tanh best performance testing phases.

Language: Английский

---

Incorporated neighborhood and environmental effects to model individual-tree height using random forest regression

Scandinavian Journal of Forest Research, Journal Year: 2023, Volume and Issue: 38(4), P. 221 - 231

Published: May 19, 2023

In forest resource inventory, tree height is often estimated by easily measurable diameter from height-diameter model. this study, we tried to use random (RF), an important machine learning method, model individual-tree height. Results showed that the optimized RF had better fitting and prediction accuracy (R2 = 0.8146 RMSE 2.2527 m). terms of relative importance, at breast (DBH) was most factor, followed neighborhood-related variables other related environmental conditions. Further, generally positively affected DBH, mean neighbors, DBH dominance, number annual precipitation, but negatively elevation. The results indicated RF-based statistically reliable highly accurate, it strong interpretability with ecological significance. Our study will provide a new perspective for application algorithms dynamic modeling.

Language: Английский

---

Plant functional traits and tree size inequality improved individual tree height prediction of mid-montane humid evergreen broad-leaved forests in southwest China

Forest Ecology and Management, Journal Year: 2023, Volume and Issue: 551, P. 121526 - 121526

Published: Nov. 6, 2023

Language: Английский

---

Shifting potential for high-resolution climate reconstructions under global warming

Quaternary Science Reviews, Journal Year: 2023, Volume and Issue: 325, P. 108486 - 108486

Published: Dec. 30, 2023

Language: Английский

---

Evaluation of Machine and Deep Learning Algorithms for Subtropical Forest Above-Ground Biomass Estimation Based on Multi-Source Data

Published: Jan. 1, 2024

Forest above-ground biomass (AGB) is an important indicator to quantify forest carbon sequestration and also a key parameter in the study of cycle terrestrial ecosystem. AGB great significance ecosystem management. However, there are few studies on systematical evaluation machine deep learning algorithms for estimation subtropical China at large scales. Moreover, often incorporates spectral variables but ignores influence environmental factors. Based multi-source data including Landsat 8 images, MODIS DEM data, climate inventory 680 sample plots, we used convolutional neural networks (CNN), random (RF), artificial (ANN), k-nearest neighbors (KNN) estimate AGB, selected model with best performance retrieve across area. The findings revealed that CNN-based had highest accuracy evergreen broad-leaved (EBF) (R2cv = 0.415 RMSEcv 51.324 Mg/ha) needle-leaved (ENF) 0.421 51.331 Mg/ha), while KNN performed deciduous (DBF) 0.364 33.673 Mg/ha). Spectral contributed most estimation, inclusion further improved models. EBFs ENFs might be positively affected by topography, negatively climate. DBFs were less Therefore, our provides new idea explore dynamics, value remote sensing

Language: Английский

---

UAV and satellite-based prediction of aboveground biomass in scots pine stands: a comparative analysis of regression and neural network approaches

Earth Science Informatics, Journal Year: 2024, Volume and Issue: 18(1)

Published: Dec. 18, 2024

Language: Английский

---