Trends and Applications of Principal Component Analysis in Forestry Research: A Literature and Bibliometric Review

Abstract

Principal component analysis (PCA) is a widely applied multivariate statistical technique across scientific disciplines, with forestry being one of its most dynamic areas of use. Its primary strength lies in reducing data dimensionality and classifying parameters within complex ecological datasets. This study provides the first comprehensive bibliometric and literature review focused exclusively on PCA applications in forestry. A total of 96 articles published between 1993 and 2024 were analyzed using the Web of Science database and visualized using VOSviewer software, version 1.6.20. The bibliometric analysis revealed that the most active scientific fields were environmental sciences, forestry, and engineering, and the most frequently published journals were Forests and Sustainability. Contributions came from 198 authors across 44 countries, with China, Spain, and Brazil identified as leading contributors. PCA has been employed in a wide range of forestry applications, including species classification, biomass modeling, environmental impact assessment, and forest structure analysis. It is increasingly used to support decision-making in forest management, biodiversity conservation, and habitat evaluation. In recent years, emerging research has demonstrated innovative integrations of PCA with advanced technologies such as hyperspectral imaging, LiDAR, unmanned aerial vehicles (UAVs), and remote sensing platforms. These integrations have led to substantial improvements in forest fire detection, disease monitoring, and species discrimination. Furthermore, PCA has been combined with other analytical methods and machine learning models—including Lasso regression, support vector machines, and deep learning algorithms—resulting in enhanced data classification, feature extraction, and ecological modeling accuracy. These hybrid approaches underscore PCA’s adaptability and relevance in addressing contemporary challenges in forestry research. By systematically mapping the evolution, distribution, and methodological innovations associated with PCA, this study fills a critical gap in the literature. It offers a foundational reference for researchers and practitioners, highlighting both current trends and future directions for leveraging PCA in forest science and environmental monitoring.

1. Introduction

Forest management is essential for environmental protection and biodiversity conservation. The vast diversity of tree species complicates the development of dynamic forest models. These models—whether analytical, numerical, or statistical—require the collection of multiple parameters and long-term historical data. Given the significant number of tree species in forests, a reliable estimation of structural, qualitative, or stability parameters necessitates extensive data collection. These parameters must be categorized into specific groups, such as physical terrain properties, climatic factors, and biological characteristics. Thus, the most effective approach to minimizing prediction bias is to classify trees based on similarity in characteristics.

What is principal components analysis (PCA)? It is a way of identifying patterns in data and expressing the data in such a way as to highlight their similarities and differences. Since patterns in data can be hard to find in data of high dimension, where the luxury of graphical representation is not available, PCA is a powerful tool for analyzing data. The other main advantage of PCA is that once you have found these patterns in the data, you compress the data, i.e., by reducing the number of dimensions, without much loss of information [1].

PCA, proposed by Karl Pearson in 1901 [2,3], is used to analyze and obtain the main components of data using the eigenmatrix transformation. PCA is a simple method for analyzing multivariate statistical distribution with characteristic quantities [4]. The results can be interpreted as an explanation of the variance in the original data. In other words, PCA provides an effective way to reduce the data dimension.

Principal component analysis is widely employed in forestry research. It is among the most frequently used statistical analysis methods for handling multi-variable datasets. The effectiveness of PCA has been demonstrated in numerous fields requiring the grouping, classification, and hierarchical structuring of influential parameter groups. In habitat analysis, PCA is a valuable tool for protected area management, enabling the identification of rare and unique habitats within seemingly homogeneous forest landscapes. For instance, PCA was utilized to develop the compound habitat gradient (CHG), integrating biotic and abiotic factors [5].

Scientific investigations have undergone significant transformations in recent years [6]. Advances in data collection [7], recording methods [8], and data organization [9] have led to increasing automation in classification [10] and analysis [7]. There is a growing tendency to utilize large datasets [11] and historical records [12], particularly those spanning extended timeframes [13].

In most documented studies, multi-variable datasets typically include a combination of physical–chemical [11] and biological parameters [14]. For instance, in coniferous species research, 13 study parameters were included in the analysis [15]. A critical first step in scientific investigations is to study the grouping of state parameters [16] to better understand the factors influencing forest dynamics and the evolution of ecosystems [17].

The most frequently applied method for structuring and organizing influential and action-related parameters is principal component analysis. The effectiveness of PCA lies in its ability to establish a reference system within a representative space, minimizing the dataset’s squared error variance. This system of axes—known as principal factors [5,7,18]—groups dataset parameters based on coordinate values along these axes. Each principal factor classifies affiliated parameters into distinct sets [7]. This approach enhances the clarity of data interpretation [6,10,19] and facilitates the construction of conceptual and dynamic models.

A substantial number of research articles on this topic have been published in specialized scientific journals across different disciplines, including forestry [20,21,22,23]. However, no review articles specifically addressing this subject have been identified so far.

The objective of this study is to provide a systematic review of the evolution of PCA applications in forestry from 1993 to 2024 using bibliometric analysis. The evaluation covers publication types, scientific disciplines, annual distribution of articles, author nationalities, affiliated institutions, journals, editors, and frequently used keywords.

2. Materials and Methods

This study was conducted in two main parts. The first involved a bibliometric analysis aimed at evaluating global scientific research on the application of principal component analysis (PCA) in forestry from 1993 to 2024. Relevant publications were identified using the Science Citation Index Expanded (SCI-Expanded) within the Web of Science (WOS) database [24], as well as Scopus [25].

Web of Science is a multidisciplinary platform comprising six citation databases, including SCI-Expanded, which indexes over 7100 leading journals across 150 disciplines, dating back to 1900. Scopus, similarly, is a comprehensive abstract and citation database that includes content from more than 16,000 peer-reviewed journals, 520 conference proceedings, 650 trade publications, and 315 book series. Both databases are widely used for research evaluation and collection development and are known for their user-friendly interfaces and helpful documentation formats [26]. While both systems offer powerful tools for retrieving scholarly information, increasing their export limits would further streamline the data extraction process.

A comparative study by Adriaanse and Rensleigh [27] evaluated WOS, Scopus, and Google Scholar (GS), finding that WOS retrieved the highest number of citations and the broadest journal coverage, followed by GS and Scopus. WOS also produced the most unique items and, along with Scopus, returned no duplicate entries—unlike GS, which often retrieved multiple copies. In terms of data consistency, Scopus showed the least variability, whereas GS exhibited the most, particularly in areas such as author name formatting and volume or issue numbers. WOS showed intermediate consistency in this regard.

After experimenting with various search strategies and keywords, the phrases “principal component analysis (PCA) in forestry” and “utilization of principal component analysis (PCA) in forest research” were selected as the main search terms. This initial query retrieved 361 articles from Web of Science and 683 from Scopus. After removing duplicates (articles appearing in both databases), 803 unique articles remained. We then excluded records that could not be retrieved, those without abstracts, and articles deemed irrelevant to the topic. This refinement resulted in a total of 382 articles, which formed the basis of the analyses presented in Section 1. From a close evaluation of these 382 articles, only 96 were found to contain elements relevant for citation in our bibliometric and conceptual review (Figure 1).

The bibliometric analysis examined 10 core dimensions: (1) publication types, (2) research areas, (3) publication years, (4) contributing countries, (5) authors, (6) affiliated institutions, (7) publication languages, (8) journals, (9) publishers, and (10) keywords.

Data analysis was carried out using built-in tools from the Web of Science Core Collection, Scopus, Microsoft Excel [28], and Geochart [29]. For visual mapping and cluster analysis, VOSviewer (version 1.6.20) was employed [30].

The second component of the study involved a traditional literature review to provide an in-depth synthesis of the 382 selected articles. The findings were organized into three primary research areas: (1) utilization of principal component analysis in forestry; (2) emerging applications and innovations in PCA-based forestry research; and (3) integration of PCA with other methods and programs.

A visual overview of the methodology is presented in Figure 2.

3. Results

3.1. A Bibliometric Review

The bibliometric analysis of this topic revealed that a total of 96 articles have been published (89% of total publications), comprising 10 proceeding papers (9%) and 2 review articles (2%).

Among the 39 scientific fields to which the published articles belong, the most representative are Environmental Sciences Ecology (35 articles), Forestry (29 articles), Engineering (17 articles), and Science & Technology—Other Topics (13 articles). Other well-represented fields include Agriculture, Plant Sciences, Chemistry, Remote Sensing, Geology, Computer Science, and Water Resources (Figure 3), demonstrating the broad range of research areas where this modern analytical method can be applied.

The distribution of the number of published articles per year is presented in Figure 4.

The articles on this topic have been published in 82 different journals, with the highest number of articles appearing in Forests (10 articles), Sustainability (8 articles), and Journal of Analytical and Applied Pyrolysis (3 articles) (

Table 1. Most representative journals for PCA-related forestry research.

Cur. No.	Journal	Documents	Total Link Strength
1	Forests	10	67
2	Sustainability	8	55
3	Journal of Analytical and Applied Pyrolysis	3	169
4	Forest Policy and Economics	2	50
5	Annals of Forest Science	2	27
6	Forest Ecology and Management	2	22
7	International Journal of Sustainable Development and Word Ecology	2	18
8	Journal of Applied Ecology	1	107
9	Landscape and Urban Planning	1	87
10	Water Air and Soil Pollution	1	74
11	Ecological Indicators	1	69
12	Biodiversity and Conservation	1	68
13	Journal of Hydrology	1	45
14	Applied Soil Ecology	1	34
15	Mathematical Problems in Engineering	1	24
16	IEEE Transactions on Geoscience and Remote Sensing	1	23
17	Quaternary Science Review	1	21
18	Journal of Forestry Research	1	17
19	Scientometrics	1	16
20	Remote Sensing	1	13

).

A total of 36 publishers have been identified, with the most prominent ones being Elsevier (21 articles), MDPI (21 articles), and Springer Nature (15 articles).

In total, 198 authors have contributed to publications on this topic. The most prolific authors, each with two articles, include Paulo Hein, Joaquin Solana-Gutierrez, C.W. Yun, Erika Meszaros, J. Lee, and E. Jakab. The 198 authors contributing to this research are affiliated with 198 institutions, with the most representative being Beijing Forestry University (5 authors), CIRAD—Centre de Cooperation Internationale on Recherche Agronomique pour la Developpement, INRAE—Institut National de la Recherche Agronomique, NorthEast Forestry University China, and Universite de Montpellier, each with 3 authors.

Authors from 72 countries have published articles on this topic. The most well-represented countries are China, Spain, Brazil, and Italy (Figure 5).

The most frequently used keywords are principal component analysis, forestry, climate, indicators, biomass, land use, quality, and conservation (

Table 2. The most used keywords in articles published about principal component analysis (PCA) in forestry.

Crt. No.	Keyword	Occurrences	Total Link Strength
1	principal component analysis	17	18
2	forestry	10	15
3	biomass	6	13
4	biodiversity	4	12
5	climate	7	12
6	indicators	7	12
7	land use	6	12
8	quality	6	12
9	carbon	5	11
10	conservation	6	11
11	dynamics	5	11
12	restoration	4	11
13	growth	6	9
14	impacts	5	9
15	region	4	8
16	vegetation	6	7
17	wood	4	7
18	decomposition	4	6
19	selection	4	6

).

These keywords are grouped into five clusters: the first contains the keywords principal component analysis, forestry, deforestation, biodiversity conservation, GIS, cluster analysis, classification, agriculture, wood; the second: biomass, quality, restoration, carbon, association, biodiversity, soil, ecosystem; the third: land use, impact, river, model, management; the fourth: climate, diversity, growth, region, impacts, sustainability; the fifth: indicators, soil quality, degradation, enzyme activity (Figure 6).

If during the years 2012–2014, the most frequently used keywords were principal component analysis, biodiversity, decomposition, and wood, in the period 2015–2018, terms such as forestry, dynamics, climate, conservation, quality, and growth became more common. In 2019–2020, the dominant keywords shifted to indicators, land use, restoration, impacts, and near-infrared spectroscopy (Figure 7).

3.2. Literature Review

3.2.1. Utilization of Principal Component Analysis in Forestry

To synthesize the scope of PCA applications in forestry, we provide a visual overview that categorizes its uses into thematic areas (e.g., remote sensing, forest health, biodiversity, etc.). This overview supports reader comprehension and provides a rapid reference for the diverse roles PCA plays in forest science (Figure 8).

Additionally, the following table summarizes the benefits and limitations of PCA when applied to forestry research (

Table 3. Benefits and limitations of PCA in forestry research.

Application Area	Benefits of PCA	Limitations of PCA
Forest structure and inventory	Enhanced model performance	Context-dependent performance
Soil and environmental studies	Grouping of multi-parametric indicators	Loss of interpretability beyond PC1–3
Species classification	High accuracy, dimensionality reduction	May miss non-linear patterns
Fire/disease monitoring	Real-time processing	Needs careful calibration
Remote sensing analysis	Feature extraction, fusion, change detection	Sensitive to preprocessing

):

From the large number of areas where PCA is used in forestry, those that have a direct description in the articles published so far are presented in

Table 4. Some areas in forestry where PCA is used (extract from literature).

Cur. No.	Domain	Subdomain	Country	Citing Article
1	Agro-forestry	Chestnut	Italy	Vella et al., 2019 [31]
		Soil quality	China	Luo et al., 2022 [32]
2	Biodiversity conservation	Impact of soil erosion	Nigeria	Asuoha et al., 2019 [33]
3	Biometry	Measurable traits of trunk and crown	Poland	Zawieja et al., 2021 [34]
		Modeling and spatialization of biomass	Brazil; China; Portugal	Oliveira et al., 2021 [35]; He et al., 2024 [36]; Teixeira et al., 2024 [37]
		Estimating tropical forest aboveground biomass stock	Brazil	Rex et al., 2020 [38]
4	Chemical composition	Volatile compounds of sea buckthorn (Hippophae rhamnoides L.)	Romania	Socaci et al., 2013 [39]
		Walnut trees	France	Heim et al., 2022 [40]
		Bioaccumulation of nickel in five edible saprotrophic mushroom species	Croatia	Siric et al., 2023 [41]
5	Education, publications	Forest journals	general	Dobbertin and Nobis, 2010 [42]
		Student creativity in a Forestry engineering degree	Spain	Solana-Gutiérrez et al., 2014 [43]
6	Entomology and Fitopatology	Phenolic profiles of two Melampyrum species	Finland	Kaitera, J. and Witzell, J., 2016 [44]
		Ground-dwelling beetles in a conifer plantation	Japan	Ueda et al., 2024 [45]
		the conifer pathogen Heterobasidion annosum	Finland	Mgbeahuruike et al., 2013 [46]
		Poplar Anthracnose	China	Jia et al., 2024 [47]
		Collembola	Canada	Addison et al., 2003 [48]
		Defoliation during the Dendrolimus tabulaeformis Tsai et Liu disaster outbreak	China	Zhang et al., 2018 [49]
		Leaf blight disease of teak (Tectona Grandis L.)	Nigeria	Dania et al., 2020 [50]
7	Environmental factors	Soil	Poland; Indonesia; Canada; China; Korea; Mexico; India; China	Klamerus-Iwan et al., 2015 [51]; Lee and Yun, 2024 [52]; Staddon et al., 1997 [53]; Zhang et al., 2021 [54]; Siswo et al., 2023 [55]; Rodríguez-Rivera et al., 2023 [56]; Kurien et al., 2021 [57] ; Sharma et al., 2024 [58]; Xiang et al., 2023 [59]
		Topoclimate	Brazil	Tourne et al., 2016 [60]
		Alpine forest drought monitoring	Austria	Lewinska et al., 2016 [61]
		Site factors as predictors for Pinus halepensis Mill. productivity	Spain	Bueis et al., 2017 [62]
		Dynamics of Pinus halepensis Mill. and Pinus sylvestris L. plantations	Spain	Bueis Mellado, 2017 [63]
8	Forest damages	Forest fires	General; Brazil	Wu et al., 2020 [64]; Wang et al., 2021 [65]
		Invasive species	General; Canada	Peerbhay et al., 2015 [66]; Baron and Rubin, 2021 [67]
		Bird response to forest disturbance	Czech Republic	Kebrle et al., 2022 [68]
9	Forest ecosystem services	Socio-environmental determinants	Ethiopia	Mengist et al., 2022 [69]
		Determining forest areas with recreational potential	Turkey	Diktaş Bulut, 2018 [70]
10	Forest management	Sustainable forestry	Hong Kong	Zhang and Jim, 2013 [71]
		Forest indicators	Italy	Salvati et al., 2017 [72]
		State forest organizations	Europe	Liubachyna et al., 2017 [73]
11	Forest structure	Landscape ecology metrics	general	Venema et al., 2005 [74]
		Mixed stands	Brazil; Romania	de Souza et al., 2020 [75]; Murariu et al., 2021 [76]
12	Forest types	Natural forests	China	Qu et al., 2024 [77]; Chen et al., 2025 [78]
		Species association in a broadleaf forest	Romania	Palaghianu and Coșofreț, 2023 [79]
		Comparison between paired plantations versus natural forests	China	Yang et al., 2022 [80]
13	Genetics	Cinnamomum camphora chemotypes	China	Guo et al., 2017 [81]
		Willow, Pinus clones	Slovenia; China	Sandak and Sandak, 2011 [82]; Mu et al., 2024 [83]
		Intraspecific leaf morphological variation	China	Yang et al., 2022 [80]
		Norway spruce (Picea abies L.) provenances	Slovakia	Jamnická et al., 2019 [84]
14	Satellite images, remote sensing	Monitoring forest harvest	USA	Pangaribuan et al., 1997 [85]
		Plantation monitoring	Lao People’s Democratic Republic	Phompila et al., 2014 [86]
		Monitoring tree mortality	Ukraine	Skydan et al., 2022 [87]
		Leaf reflectance spectra	general	Song et al., 2020 [88]
		Vegetation classification	Russia; Iran; Italy	Yancovich et al., 2019 [89]; Oladi et al., 2010 [90]; Pesaresi et al., 2024 [91]; Sa et al., 2023 [92]
15	Urban forests	Peri-urban forests in Beijing	China	Cao et al., 2022 [93]
		Site conservation	Singapore	Hwang and Roscoe, 2017 [94]
		Monitoring of urban forest	Croatia	Krtalic et al., 2021 [95]
		Identifying urban tree species	USA	Pu and Liu, 2011 [96]
16	Vegetation restauration	Carbon pools	China	Feng et al., 2024 [97]; Ju et al., 2024 [98]
		Blanket mire	Ireland	Cooper et al., 2001 [99]
17	Wood anatomy	Genus Maytenus	Argentina	Gimenez et al., 2014 [100]
		Thermal behavior of wood from energy plantation	Hungary; Senegal	Meszaros et al., 2004 [101]; Nganko et al., 2024 [102]
		Wood density and fiber dimensions of the Persian oak wood	Iran	Nazari et al., 2020 [103]
		Basic density of Eucalyptus grandis wood chips	Brazil	de Medeiros et al., 2024 [104]
		Near-infrared spectroscopy	Brazil; India	Ramalho et al., 2017 [105]; Sharma et al., 2024 [106]

.

3.2.2. Emerging Applications and Innovations in PCA-Based Forestry Research

Recent advancements have demonstrated the versatility and growing importance of principal component analysis across various forestry research domains. Several innovative methods have been developed that integrate PCA to enhance data analysis, classification, and interpretation, especially in remote sensing and ecological studies.

An enhanced PCA-based classification scheme (PCABC) was proposed for unsupervised, image-based identification of Mediterranean forest species [6]. This approach applies a binary threshold to the first few principal component bands to iteratively detect and cluster spectral endmembers. By sequentially masking identified classes and reapplying PCA on the residual dataset, the method improves sensitivity to subtle spectral differences, particularly for small or less distinct classes, without relying on supervised classification or inverse PCA transformations.

In addressing the complex analysis of Brazil’s rainforest fires, a novel hybrid model, the principal component analysis-boosted dynamic Gaussian mixture clustering model (PCA-DGM), was developed [4]. PCA-DGM enhances clustering performance by first reducing the dimensionality of fire-related data and emphasizing key ignition features. A dynamic distance loss function capable of adjusting density and distance parameters was integrated to optimize cluster shape determination. This approach allowed for a more accurate temporal and geographical analysis of fire causes, outperforming traditional clustering algorithms.

Another significant application involves the automation of tree crown counting in high-resolution satellite imagery [7]. A MATLAB-based PCA method was employed to transform correlated pixel values into orthogonal principal components, followed by thresholding and edge detection using the Laplacian of Gaussian (LoG) operator. Morphological operations were then applied to delineate and count individual tree crowns. Validation against manual counts across 20 test areas achieved an accuracy of 86.6%, demonstrating the method’s efficiency in large-scale forestry assessments.

Hyperspectral data restoration has also benefited from PCA innovation. A data-driven PCA technique was introduced to restore vegetation reflectance spectra affected by noise in specific bands [88]. Using a training dataset simulated with the PROSPECT-5 model, the first 10 principal components—capturing 99.998% of the variance—were employed to reconstruct noisy spectral bands. This method shows promising potential for improving the quality of hyperspectral remote sensing data critical to vegetation analysis.

In dendrochronology, PCA has been explored as a novel detrending method for tree-ring series [107]. By removing the first principal component—assumed to capture natural growth trends—the method produces tree-ring indices while preserving low-frequency climatic signals. Compared to traditional methods such as cubic splines and regional curve standardization, PCA detrending provides an automated and robust alternative, particularly when using the alternating least squares (ALS) algorithm to handle series of varying lengths.

Finally, advancements in forest structure monitoring have combined PCA with machine learning techniques for improved stem segmentation and measurement. Neuville [108] integrated an improved Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) algorithm for stem segmentation, followed by PCA to extract stem orientations and estimate diameters at breast height (DBH) from LiDAR point clouds. This method achieved stem detection rates of up to 82% (precision of 98%) during leaf-off seasons without requiring site-specific parameter tuning, showcasing the robustness of PCA-driven workflows across different scanning conditions.

Together, these examples highlight the dynamic role PCA continues to play in expanding the boundaries of forestry research, from spectral data enhancement to structural ecosystem analysis.

3.2.3. Integration of PCA with Other Methods and Programs

Principal component analysis (PCA) has started to be integrated into forestry studies, alongside other analytical methods. The method was used to increase model performance, reduce dimensionality, and improve the interpretation of data. This hybrid approach is useful in addressing both the complexity and heterogeneity of environmental data.

Feature selection and dimensionality reduction

He et al. [109] have created an innovative model for predicting soil organic matter by combining Lasso, SCARS, and PCA. As such, they integrated both spectral and profile features and reduced dimensionality and overfitting when the number of features exceeded the number of samples. However, even though PCA has preserved most of the data content and improved computational efficiency, it also obscured nonlinear relationships, which are essential in recognizing ecological patterns.

Object detection and classification

The study by Wu [60] focused on integrating PCA with object detection algorithms (YOLOv3, SSD) in order to detect forest fires in real time. When PCA was used during the feature extraction phase, the mean average precision (mAP) was improved by 7.3% for YOLOv3 and by 4.6% for SSD. However, this improvement depends on good calibration. In addition, aggressive dimensionality reduction can lead to suppressions in signal variance, especially in areas that are prone to fires.

Ecosystem service modeling

Salata and Grillenzoni [110] combined PCA with MATLAB version R2023b and ArcGIS in order to integrate seven ecosystem service models. In their case, PSA offered a mapping that was better and more composite than traditional GIS overlays. However, they also experimented with reductions in the interpretation of individual service contributions, which can impact its usefulness in policy and decision-making processes.

Remote sensing applications

Jia et al. [47] showed how PCA can improve the detection of poplar anthracnose, especially compared with SPA, VCPA, and SDIs. However, Oladi [90] proved that the PCA fusion of panchromatic and multispectral data led to less accuracy in classifications realized in a heterogeneous area. This proves that PCA is sensitive to landscape complexity and risks oversimplifying information.

Spectroscopy and lab analysis

PCA was also used by Heim [40] to differentiate wood fractions in walnut trees and by Sandak [82] to classify clones of willow. Even though PCA improves signal-to-noise rations, it can also decrease visibility in rare and ecologically important features. This limitation is already known in forensic wood or in analyses of trace elements.

Comparative evaluation

López-Amoedo [111] compared PCA with Random Forest and multiple linear regression (MLR) in order to predict the wood weight of Pinus radiata. His results showed that PCA outperformed the other models, with a result of 0.85 R². However, the fact that PCA relies on linearity and orthogonality assumptions can limit its effectiveness, especially in contexts where nonlinear decision boundaries are needed.

Critical considerations and future directions

Although PCA shows considerable benefits for forestry analytics, there are a series of limitations that must be taken into account:

-

PCA assumes linearity, which can mask complex ecological interactions.

-

Secondary PCA interpretations are often unclear.

-

The results depend on variable scaling and data preprocessing decisions.

Based on these aspects, we propose a series of alternatives and improvements:

-

Kernel PCA for extracting nonlinear features.

-

t-SNE and UMAP for visualizing high-dimensional ecological datasets.

-

Adaptive PCA models that adjust to forest heterogeneity.

We also raise a series of questions for future studies:

How can adaptive PCA models be developed to better handle local forest heterogeneity?

Can hybrid PCA methods improve robustness across different forest types and scales?

What is the trade-off between interpretability and accuracy when using PCA in deep learning-based ecological monitoring?

Only if we critically evaluate the integration and hybridization of PCA can we continue to advance analytical workflows in forestry research.

4. Discussion

4.1. Bibliometric Review

Statistical analysis methods are among the most widely used mathematical tools in scientific research. In today’s era of rapid information growth, while there are numerous methods for obtaining data, efficient tools for analysis and data extraction are still lacking. The advancement of pattern classification and machine learning also reflects the needs of both industry and daily life. In this context, the use of statistical methods is now essential. The importance of employing these techniques ensures the coherence of research presentation, as well as the rigor and accuracy of the results included in published studies.

Current studies typically involve large and complex datasets consisting of numerous physical-chemical, biological, and environmental parameters. These datasets are often difficult to interpret without proper analysis. Multivariate statistical techniques and exploratory data analysis are essential tools for reducing data complexity and interpreting chemical and physical measurements with multiple constituents. Techniques such as cluster analysis (CA), factor analysis (FA), principal component analysis (PCA), and discriminant analysis (DA) have been widely applied in water quality assessment, forest structure analysis, and atmospheric phenomena research, providing meaningful insights.

Among all statistical analysis methods, principal component analysis is the most frequently used in the literature. The widespread application of PCA can be attributed to the fact that contemporary studies often involve multiple parameters and measurable variables, all of which contribute to and influence a particular result. In such cases, grouping and ranking input parameters are necessary for a deeper understanding of their effects on the studied outcome.

This study provides a bibliographic overview of how PCA is applied in forestry investigations. The most representative research fields and journals are directly linked to environmental and forestry sciences. The distribution of articles by field of application highlights environmental sciences as the leading domain where PCA has been utilized [112]. The second most dominant field is forestry, which can be explained by the suitability of PCA for analyzing environmental systems with multiple physical–chemical, structural, biological, and ecological parameters. These parameters must be measured together, necessitating an analysis of their interrelationships and classification.

Regarding the distribution of articles over time, an exponential growth trend is evident. This increasing trend in published articles, which has been observed in other bibliometric studies [23,103], also applies to the topic of principal component analysis in forestry. This pattern is commonly found in natural processes with specific growth rates that are independent of external controllable factors [16,18]. This trend reflects a broader phenomenon—a modern and efficient analytical method that, through its diverse applications, naturally drives the expansion of scientific output [8,17]. The number of published articles and citations increased significantly after 2015.

Among the leading journals publishing articles on PCA in forestry, Forests (10 articles) and Sustainability (8 articles) stand out. It is evident that many of the top journals belong to the Q1 and Q2 categories, which naturally serve as the primary platforms for disseminating results obtained using advanced statistical methods such as PCA [7]. This aspect is crucial, emphasizing that the majority of studies in this field employ high-level mathematical methodologies. Today, the majority of research relies on advanced mathematical models, cutting-edge technologies, and sophisticated numerical software. This trend is even more pronounced in high-impact journals, where most published works involve complex statistical analyses.

In this context, it is noteworthy that the most frequently used journals belong to the MDPI group. This can be explained by researchers’ desire to publish their findings quickly in high-impact journals to increase visibility. The urgency to disseminate results is a common trend not only in forestry but also across scientific disciplines, driven by the need for rapid recognition of significant findings. Only after these leading journals do those with lower impact factors come.

Most authors publishing on PCA in forestry are affiliated with universities and research institutes in China. This is partly because many of the journals in which these articles appear are based in China. Additionally, this trend highlights the strong research focus on forestry development in China, particularly in studies related to biodiversity conservation and timber production.

The main keywords associated with PCA-based studies in forestry generally correspond to several major research directions. The first research area focuses on biomass and production [19,72], which involves analyzing state parameters. The second major area is dedicated to species classification and regional categorization, using spectral investigation methods [7]. This category includes studies based on multispectral image processing [72], satellite data analysis [6], and UAV-based or near-infrared imaging techniques. Finally, another group of studies applies PCA to analyze ecosystem dynamics and identify specific indicators, particularly in soil analysis [72], enzyme activity studies [9], and other environmental assessments.

The evolution of keywords over time reflects the development of PCA applications in forestry research. Initially, studies focused on biodiversity and production analysis [19]. In later years, research expanded to indicator identification [72], enzyme-related studies, and broader ecosystem analyses.

4.2. Recent Innovations and Emerging Uses of PCA in Forestry Sciences

Principal component analysis (PCA) has been integrated more and more into forestry research, proving both its adaptability, as well as the complexity of modern environment datasets. The case studies we have reviewed show that PCA can maintain its strength in reducing dimensionality and mitigating noise, as well as play an essential role in providing new innovative methodologies adjusted to specific forestry applications.

Innovative PCA-based workflow

A complex PCA classification scheme (PCABC) was created for unsupervised image-based identification of Mediterranean forest species [6]. This method has improved the separation of classes by masking detected classes and reapplying PCA, without relying on supervised learning or inverse PCA transformations. In a similar manner, Want et al. [4] used PCA-boosted dynamic Gaussian mixture clustering (PCA-DGM) to analyze rainfall fires from Brazil. This model used a dynamic distance loss function for PCA-reduced data and outperformed traditional clustering methods in identifying spatial and temporal patterns in forest fires.

These methods prove the effectiveness of PCA, especially in classifying cases. However, the reliance on variance-based data structure assumptions is a limitation for their applicability. In addition, model accuracies may suffer if the main components fail to identify important ecological gradients or nonlinear trends.

Automation in forest inventory

PCA has been used in automating forest inventory tasks, especially in counting tree crowns through high-resolution satellite images [7]. PCA was used to transform pixels into orthogonal components, allowing for better edge detection and segmentation. This system obtained an accuracy of 86.6% compared to manual tasks. However, its efficiency depends on site-specific factors (canopy density, shadowing, and species diversity).

Spectral restoration and tree-ring analysis

PCA was used by Song et al. [88] to restore noise spectral bands in hyperspectral data. They obtained this by reconstructing reflectance spectra through the dominant principal parts. PCA was also used in dendrochronology to detrend tree-ring series by removing the first main component (long-term growth trends) and preserving low-frequency climatic signals [106]. This provides an automated, data-driven alternative to traditional methods, but poses risks, especially in removing site-specific biological variance or more subtle ecological signals.

Structural forest monitoring

PCA was also used alongside machine learning to monitor forest structures. Neuville [98] used PCA and an HDBSCAN algorithm to extract orientation and diameter metrics after tree stems were clustered. This method obtained high precision detection rates of up to 82% and eliminated the need for site-specific calibration. However, PCA can also generalize stem features and miss unique characteristics, especially in uneven-aged or structurally diverse stands.

Critical Perspectives and Future Directions

As we have seen, PCA is essential for many forestry innovations. However, it also has its limitations, such as the following:

-

assuming linearity and overlooking the nonlinear relationships that are essential for ecological modeling;

-

being sensitive to data processing (normalization, scaling) and affecting result reliability;

-

ambiguous and data-dependent interpretation of later main components.

In order to address these challenges, future studies should explore the following:

-

hybrid methods that combine PCA with tools for reducing nonlinear dimensionality (t-SNE, UMAP). This allows for a better understanding of complex ecological interactions.

-

validating different forest types and ecological gradients. This allows for better testing of general and robust models.

-

integrating deep learning frameworks where PCA is a feature in extracting and preprocessing, allowing for a better performance and a reduced computational load.

Only if we evaluate critically the innovation posed by PCA, as well as its limitations, can we continue to improve this tool, which is essential in forestry sciences. This is even more important when paired with complementary technologies and models.

Enhancement of current approaches

Improving current approaches: PCA implementations can be improved and enhanced by adopting certain strategies: using pipelines adapted to specific domains (vegetation calibration, stratified sampling), using hybrid pipelines that combine PCA with classification (ex. PCA + Random Forests, PCA + SVM), or combining PCA with AI frameworks. All these methods can offer more predictive analytics and interpretation of results, as well as better transparency, elements that are critical in the domain of forestry conservation.

4.3. Integration of PCA with Other Methods and Programs

The integration of principal component analysis with various analytical techniques and technologies in forestry research highlights its versatility and effectiveness across diverse applications. This review of recent literature demonstrates that PCA is not merely a dimensionality reduction tool, but a fundamental component of modern forestry research workflows, particularly when combined with complementary methods.

The combination of PCA with feature selection algorithms, such as Lasso and SCARS [109], reflects an important trend toward hybrid modeling approaches. In high-dimensional datasets like hyperspectral imagery or full-band soil spectra, traditional models often suffer from overfitting. Here, PCA’s ability to preserve essential variance while compressing redundant information has proven critical. By fusing PCA-derived features with profile data, researchers can achieve more accurate and robust predictions, illustrating PCA’s role in building next-generation predictive models.

Similarly, the application of PCA with object detection algorithms like YOLOv3 and SSD (Wu) underscores its value in real-time operational contexts, such as forest fire detection. The observed improvements in mean average precision (mAP) indicate that even deep learning models, typically resilient to high-dimensional input data, benefit from preliminary PCA-based feature engineering. This suggests that PCA continues to be relevant, even in an era dominated by convolutional neural networks and end-to-end learning systems.

In GIS-based ecosystem service modeling [110], PCA offers a solution to the limitations of traditional overlay analysis, which often oversimplifies complex ecological interactions. The move towards PCA-driven composite mapping represents a broader shift in landscape ecology towards more data-driven, statistically robust methodologies.

Remote sensing applications further exemplify the flexibility of PCA. Whether for beetle infestation monitoring [113,114], species identification [115,116], disease detection [45], or forest change analysis [79], PCA consistently improved the interpretability and accuracy of satellite and aerial data. The coupling of PCA with vegetation indices like NDVI and EVI appears particularly powerful, allowing for enhanced thematic mapping and classification accuracy. Nonetheless, it is notable that in some cases, as in Oladi’s [90] study on forest type classification, PCA-based fusion methods may not always yield superior results, especially in heterogeneous landscapes. This underscores the necessity of contextual evaluation when applying PCA: while it reduces dimensionality and redundancy, it may also oversimplify critical spatial patterns if not carefully tuned.

In laboratory-based analytical techniques, such as laser-induced breakdown spectroscopy (LIBS) and near-infrared (NIR) spectroscopy, PCA has facilitated significant advances. In LIBS (Jin), PCA-enhanced models achieved high soil classification accuracy, validating its strength in managing noisy and complex spectral data. Similarly, multiple studies using NIR [38,77,117] and FTIR-ATR spectroscopy [118,119] demonstrated that PCA effectively distinguished subtle compositional differences in biological materials. These examples highlight PCA’s essential role in enhancing signal-to-noise ratios and maximizing the discriminatory power of spectroscopic datasets.

When compared with other machine learning models, PCA showed competitive or superior performance, as seen in López-Amoedo’s work on wood property prediction. Although PCA is fundamentally a linear technique, its ability to structure high-dimensional datasets for subsequent modeling suggests that it remains a highly relevant pre-processing tool even as more complex non-linear models, like Random Forests and neural networks, gain popularity.

The influence of sample size on statistical analysis outcomes is an important consideration in forestry research, especially when employing multivariate methods like principal component analysis (PCA), ANOVA, or correlation analysis. While PCA is relatively robust to moderate variations in sample size, larger datasets typically enhance the reliability of the results, improve the stability of principal components, and reduce variance distortion—especially in heterogeneous ecological datasets. Similarly, in ANOVA, increasing the sample size improves the statistical power and confidence level (lowering p-values), which is essential for detecting subtle but meaningful differences among group means. For Pearson or Spearman correlation coefficients, large datasets can increase precision and reduce the margin of error. In the case of PCA, larger samples facilitate more accurate identification of major variance directions in the data space and allow for better representation of complex ecological gradients. PCA remains a fundamental tool in forestry research because of its ability to reduce dimensionality, reveal patterns, and enhance model interpretability in studies involving many interdependent variables. Given its strengths—particularly when used alongside other techniques such as regression models, clustering, or machine learning classifiers—PCA should continue to be employed, especially in hybrid workflows where it supports both feature extraction and noise reduction. However, the choice of statistical method should always consider data characteristics, including size, dimensionality, and distribution, to ensure meaningful and reproducible outcomes.

In summary, the reviewed studies collectively suggest that PCA serves multiple roles in forestry research: (i) dimensionality reduction and overfitting mitigation, (ii) enhancement of model performance, (iii) improvement of data fusion and multi-source data integration, and (iv) support for real-time and operational forestry applications. However, the success of PCA is contingent on thoughtful application. Challenges remain, particularly when datasets are highly heterogeneous, or when critical information is contained in subtle, non-linear relationships that PCA may not fully capture.

Thus, future research should focus on combining PCA with advanced non-linear dimensionality reduction techniques, such as t-SNE, UMAP, or kernel-PCA, to better handle the complexity inherent in forestry datasets. Furthermore, the development of adaptive PCA-based methodologies, where the dimensionality reduction process is tailored dynamically to the dataset’s characteristics, could further enhance its utility.

5. Conclusions

This study highlights several key aspects: (1) Advanced statistical analysis methods, particularly PCA, are essential tools in contemporary research. The growing emphasis on mathematical and statistical approaches underscores their importance in forestry and environmental science. (2) Current studies involve complex datasets with multiple measured parameters, requiring highly efficient statistical methodologies. Among all the available methods, PCA stands out as one of the most frequently used and widely accepted by the academic community. (3) Various adaptations of statistical analysis methods are currently employed in different configurations and sequences, expanding the range of available analytical tools for forestry researchers.

Future directions and research gaps

Principal component analysis (PCA) has been used in numerous forestry domains. However, some important areas have not been explored. As such, future studies should address the following topics:

Integration with non-linear techniques: While PCA remains a powerful linear dimensionality reduction method, forest ecosystems are inherently complex and often governed by non-linear interactions. Combining PCA with algorithms like t-SNE, UMAP, or kernel-PCA could yield improved results, especially for pattern recognition and anomaly detection in heterogeneous datasets.

Real-time applications and edge computing: The rapid rise in UAVs and in situ sensors generates massive real-time datasets. Applying PCA in edge-computing environments for on-the-fly data reduction and classification could revolutionize forest monitoring, particularly for fire detection, illegal logging, and pest outbreaks.

Climate change scenarios: PCA should be used in the ecological forecast intended for different climate scenarios, such as changes in forest structures, the composition of species, and carbon storage.

The socio-ecological context: PCA should be integrated into the study of socio-environmental variables that impact forestry (stakeholder views, decisions and policies about land usage, the dissemination of forest management practices).

Standard protocols and tools: PCA and standard workflows will help allow better and faster access across studies, especially for areas with low resources.