Uncertainty quantification for structural response field with ultra-high dimensions

International Journal of Mechanical Sciences, Journal Year: 2024, Volume and Issue: 271, P. 109110 - 109110

Published: Feb. 20, 2024

Language: Английский

---

Incorporating stand parameters in nonlinear height-diameter mixed-effects model for uneven-aged Larix gmelinii forests

Frontiers in Forests and Global Change, Journal Year: 2025, Volume and Issue: 7

Published: Jan. 7, 2025

Tree attributes, such as height (H) and diameter at breast (D), are essential for predicting forest growth, evaluating stand characteristics developing yield models sustainable management. Measuring tree H is particularly challenging in uneven-aged forests compared to D. To overcome these difficulties, the development of updated reliable H-D crucial. This study aimed develop robust Larix gmelinii by incorporating variables. The dataset consisted 7,069 trees sampled from 96 plots Northeast China, encompassing a wide range densities, age classes, site conditions. Fifteen widely recognized nonlinear functions were assessed model relationship effectively. Model performance was using root mean square error (RMSE), absolute (MAE), coefficient determination (R 2 ). Results identified Ratkowsky (M8) best performer, achieving highest R (0.74), lowest RMSE (16.47%) MAE (12.50%), statistically significant regression coefficients (p < 0.05). Furthermore, M8 modified into 5 generalized (GMs) adding stand-variables (i.e., height, volume their combination), results indicate that GM2 0.82% 13.7%. We employed mixed-effects modeling approach with both fixed random effects account variations individual plot level, enhancing predictive accuracy. explained 71% variability trends residuals. calibrated response calibration method, through EBLUP theory. Our findings suggest stand-level variables representing plot-specific can further improve fit mixed- models. These advancements provide authorities enhanced tools supporting

Language: Английский

---

Individual tree detection and estimation of stem attributes with mobile laser scanning along boreal forest roads

ISPRS Journal of Photogrammetry and Remote Sensing, Journal Year: 2022, Volume and Issue: 187, P. 211 - 224

Published: March 18, 2022

The collection of field-reference data is a key task in remote sensing-based forest inventories. However, traditional methods demand extensive personnel resources. Thus, would benefit from more automated methods. In this study, we proposed method for individual tree detection (ITD) and stem attribute estimation based on car-mounted mobile laser scanner (MLS) operating along roads. We assessed its performance six ranges with increasing mean distance the roadside. used Riegl VUX-1LR sensor high repetition rate, thus providing detailed cross sections stems. algorithm propose was designed configuration, identifying (or arcs) point cloud aggregating those into single trees. Furthermore, estimated diameter at breast height (DBH), profiles, volume each detected tree. accuracy ITD, DBH, estimates varied trees' road. general, proximity to branches 0–10 m road caused commission errors ITD over attributes zone. At 50–60 roadside, stems were often occluded by branches, causing omissions underestimation area. ITD's precision sensitivity 82.8% 100% 62.7% 96.7%, respectively. RMSE DBH ranged 1.81 cm (6.38%) 4.84 (16.9%). Stem had RMSEs ranging 0.0800 m3 (10.1%) 0.190 (25.7%), depending sensor. average proportion reference highly affected different zones. This highest 0 10 (113%), zone that concentrated most errors, lowest 50 60 (66.6%), mostly due omission other zones, 87.5% 98.5%. These accuracies are line obtained state-of-the-art MLS terrestrial (TLS) system has potential collect efficiently large-scale inventories, being able scan approximately 80 ha forests per day survey setup. could be increase amount available improve models area-based estimations, support forestry development.

Language: Английский

---

Prediction of tree crown width in natural mixed forests using deep learning algorithm

Forest Ecosystems, Journal Year: 2023, Volume and Issue: 10, P. 100109 - 100109

Published: Jan. 1, 2023

Crown width (CW) is one of the most important tree metrics, but obtaining CW data laborious and time-consuming, particularly in natural forests. The Deep Learning (DL) algorithm has been proposed as an alternative to traditional regression, its performance predicting mixed forests unclear. aims this study were develop DL models for spruce-fir-broadleaf north-eastern China, analyse contribution size, species, site quality, stand structure, competition prediction, compare with nonlinear effects (NLME) their reliability. An amount total 10,086 individual trees 192 subplots employed study. results indicated that all deep neural network (DNN) free overfitting statistically stable within 10-fold cross-validation, best DNN model could explain 69% variation no significant heteroskedasticity. In addition diameter at breast height, showed on CW. NLME (R2 ​= ​0.63) outperformed ​0.54) when six input variables consistent, opposite ​0.69) included 22 variables. These demonstrated great potential prediction.

Language: Английский

---

Nonlinear multilevel seemingly unrelated height-diameter and crown length mixed-effects models for the southern Transylvanian forests, Romania

Forest Ecosystems, Journal Year: 2025, Volume and Issue: unknown, P. 100322 - 100322

Published: March 1, 2025

Language: Английский

---

Comparing Traditional Methods and Modern Statistical Techniques for Tree Height Prediction

Forests, Journal Year: 2025, Volume and Issue: 16(2), P. 271 - 271

Published: Feb. 5, 2025

Forest mensuration is important to gain knowledge and information about forest stands. Because tree height often proves more difficult measure than diameter, different statistical models are used for their estimation instead. In this paper, the data of 986 spruce trees (Picea abies KARST. (L.)), measured in federal states Salzburg Tyrol (Austria), were train compare random with traditional approaches such as linear non-linear mixed a classical uniform curve. For model comparison, RMSE, percent bias, bias used. further visualization differences, residual plots, partial dependence conditional plots shown. The results show that (RMSE 2.23 m) can compete methods, 2.14 2.24 or curves 2.92 m), but not able outperform those especially when it comes extrapolation prediction areas where training sparse available. Furthermore, incorporation additional covariates improve certain models.

Language: Английский

---

Machine learning methods for basal area prediction of Fagus orientalis Lipsky stands based on national forest inventory

Trees, Journal Year: 2025, Volume and Issue: 39(2)

Published: March 19, 2025

Language: Английский

---

Yield of forests in Ankara Regional Directory of Forestry in Turkey: comparison of regression and artificial neural network models based on statistical and biological behaviors

iForest - Biogeosciences and Forestry, Journal Year: 2023, Volume and Issue: 16(1), P. 30 - 37

Published: Jan. 22, 2023

Models of forest growth and yield provide important information on stand tree developments the interactions these with silvicultural treatments. These models have been developed based assumptions such as independence observations, uncorrelated error terms, terms constant variance; if factors are absent, there may be problems multicollinearity, autocorrelation, or heteroscedasticity, respectively. problems, which several adverse effects parameter estimates, statistical phenomena must avoided. In recent years, artificial neural network (ANN) model, thanks to its superior features ability make successful predictions absence requirement for assumptions, has commonly used in forestry modeling. However, while goodness-of-fit measures were taken into consideration assessment ANN models, control biological characteristics model was ignored. this study, variable-density using nonlinear regression techniques. modeling techniques compared some principles yield. The results showed that more meeting expected patterns than models.

Language: Английский

---

Analyzing regression models and multi-layer artificial neural network models for estimating taper and tree volume in Crimean pine forests

iForest - Biogeosciences and Forestry, Journal Year: 2024, Volume and Issue: 17(1), P. 36 - 44

Published: Feb. 29, 2024

The taper and merchantable tree volume equations are the most used models in forestry because of their accuracy estimating both total volume. However, numerous studies reported that artificial neural network show fewer errors a greater success rate as compared to regression models. This study data from 200 Crimean pine trees Turkey’s Central Anatolia Mediterranean Region assess performance (ANN) Max-Burkhart’s equation for accurate results were obtained using 3 hidden layers 10 neurons model 1 layer 100 model. hyperbolic tangent sigmoid function was ANN analysis hyper-parameter customization. Using with customization, AAE Max-Burkhart decreased 9.315 6.939 (-25.5%), RMSE 3.072 2.656 (-13.5%), FI increased 0.964 0.966 (+1.23%). Similarly, 0.056 0.013 (-76.6%), 0.247 0.12 (-51.6%), 0.909 0.979 (+7.69%). Our showed models’ predictions more reliable equations. We resolved overfitting via modification, which also allowed monitoring impact error prediction outputs at various learning rates. It possible develop lower rates training validation data, consistent growth trends sets.

Language: Английский

---

Machine Learning Forest Simulator (MLFS): R package for data-driven assessment of the future state of forests

Ecological Informatics, Journal Year: 2023, Volume and Issue: 75, P. 102115 - 102115

Published: May 8, 2023

Language: Английский

---