The use of R in forestry research

Abstract

The field of forestry research has greatly benefited from the integration of computational tools and statistical methods in recent years. Among these tools, the programming language R has emerged as a powerful and versatile platform for forestry research, ranging from data analysis, modeling to visualization. However, the key trends in general reported R use and patterns in forestry research remain unknown. We analyzed R and R package usage frequencies for 14 800 research articles published in eight top forestry journals across a span of 10 years, from 2013 to 2022. Among these articles, a notable number of 6790 (accounting for 45.7%) explicitly utilized R as their primary tool for data analysis. The adoption of R exhibited a linear growth trend, rising from 28.3% in 2013 to 60.9% in 2022. The top five used packages reported were vegan, lme4, nlme, MuMIn, and ggplot2. Diverse journals have their unique areas of emphasis, resulting in disparities in the frequency of R package application among journals. The average number of R packages used per article also showed an increasing trend over time. The study underscores the recognition that R, with its powerful data statistical and visualization capabilities, plays a pivotal role in enabling researchers to conduct thorough analyses and acquire comprehensive insights into various aspects of forestry science.

摘要

R在林业研究中的应用

R是用于数据分析、建模和可视化的强大编程语言之一，广泛运用于各个科学学科，但其林学研究领域受欢迎程度还尚待探索。为了回答这个问题，我们对8种主要林学期刊上2013至2022年间发表的14 800多篇研究文章R和R包的使用频率进行了全面分析。分析结果显示，有6700篇文章(占总数的45.7%)明确使用R进行数据分析。R的使用率呈现出持续增长的趋势，从2013年的28.3%上升到2022年的60.9%。使用最多的5个R软件包包括vegan、lme4、nlme、MuMIn和ggplot2。各期刊的关注点有所不同导致R软件包使用频率格局不同。该分析表明，R语言凭借其强大的统计和数据可视化功能，在林学各个领域具有广阔的使用前景。

INTRODUCTION

Forestry research is an interdisciplinary science, drawing upon knowledge and techniques from ecology, biology, genetics, economics, sociology, and other fields (Helms 2002). It aims to generate knowledge that supports sustainable forest management, conservation, and the sustainable use of forest resources for current and future generations (Kimmins 2002). In recent years, the field of forestry has undergone a remarkable transformation, largely driven by the advent of data-driven approaches and advanced statistical analysis techniques (Zou et al. 2019). Statistical analysis plays a pivotal role in forestry research, enabling researchers to derive meaningful insights from data and make well-informed decisions (Polinko and Coupland 2021). It encompasses various techniques such as the experimental design, descriptive statistics, inferential statistics, multivariate analysis, spatial analysis, and time series analysis. The relevant software packages, such as SPSS, SAS, R, Python, and MATLAB, provide powerful capabilities for data analysis, facilitating in-depth research and enhancing our understanding of forest ecosystems (Atkins et al. 2022).

R is an incredibly versatile software environment designed for statistical computation and graphics development, and the best advantage is that it is freely available to the public. It is compatible with various operating systems, including Windows, MacOS, and many UNIX platforms. In the domain of data analysis in scientific fields, the R program has emerged as a highly favored choice for data analysts (Bollmann et al. 2017; Lawlor et al. 2022). Its popularity can largely be attributed to the extraordinary diversity of packages it offers. An R package is an extension of R containing specific functions and datasets to solve specific questions. These packages are mostly crafted by R users typically stored and distributed through the Comprehensive R Archive Network (CRAN) (https://cran.r-project.org/). The true strength of R lies in the availability of these packages, which can provide an extensive repertoire of functions tailored to various analytical needs.

Scientists commonly employ academic journals as the primary platform to disseminate their research findings to the academic community and the public. The frequency of R use in papers published in forestry journals could serve as an indicative gauge of its overall acceptance in the field. However, despite the widespread use of R in various scientific domains, the frequency of R and associated packages in forestry research remains relatively unexplored. While a previous study focused on the utilization of R in the field of ecology, revealing an increasing trend in popularity and reported usage of R between 2008 and 2017 (Lai et al. 2019), a substantial knowledge deficit exists regarding its utilization in forestry research. Obtaining a comprehensive grasp of the extent of R’s usage within the discipline of forestry offers immense potential benefits. Relevant research provides valuable insights for both novice R users who may consider the integration of R into their research methodologies and researchers who are actively involved in developing R packages for future use by their peers. By shedding light on R’s role in forestry research, this exploration can potentially pave the way for more comprehensive and efficient data analysis in the discipline, ultimately leading to further advancements in sustainable forest management and conservation efforts.

This study examines a comprehensive dataset consisting of more than 14 800 research articles published in the eight top forestry journals over a period of 10 years (2013–2022). The primary objective is to evaluate the reported usage of R and R packages within these articles, with the aim of identifying key trends and patterns that emerge in terms of their general adoption and the prevalence of specific packages. This examination may lead to a deeper appreciation of the benefits and challenges associated with employing R in this context, and ultimately aid researchers and practitioners in making more informed decisions when applying R and its packages into their work.

METHODS

In order to comprehensively assess the popularity of R and R packages in forestry research, we selected representative high-impact journals that are widely recognized in the field. Our selection criterion involved the journals with an impact factor exceeding 3.0 in the year 2021, quantified by the Web of Science (WoS) Journal Citation Reports (JCR) in the “Forestry” category (available at www.webofknowledge.com). Furthermore, we excluded journals which have published fewer than 50 paper annually since 2013, to guarantee the sample size required by statistics. This rigorous process screened out the following reputable journals: Agricultural and Forest Meteorology, Annals of Forest Science, Applied Vegetation Science, European Journal of Forest Research, Forest Ecology and Management, Forestry, Journal of Vegetation Science, and Tree Physiology.

We firstly recorded all papers with R or R packages cited in the “References.” However, we acknowledged that some articles might mention R or R packages in the “Methods” but not cited explicitly in the “References.” To ensure the accuracy and comprehensiveness of our analysis, we conducted a thorough manual review of the “Methods” of each article in our dataset. This laborious process allowed us to identify and include all articles that utilized R or R packages, even if they did not explicitly cite them in the “References.” During the review, when a specific R package was employed, we made a note of its name. This approach guarantees that our findings provide a precise reflection of the role of R in the field of forestry research and its associated packages.

All calculations in this paper were conducted using R statistical language (R Core Team 2022). The data (R file) and code of this paper are in Supplementary Appendix 1, and readers can easily reproduce the figures.

RESULTS

Overall trends in R utilization

A total of 14 856 research articles from the selected eight forestry journals have been accessed (Supplementary Appendix 1). This comprehensive dataset covered a decade-long period, ranging from the year 2013 to 2022. Within this extensive dataset, a noteworthy 6790 articles, constituting approximately 45.7% of the total, explicitly cited the use of R as their chosen statistical software for data analysis. This finding not only highlights the substantial adoption of R in forestry research but also indicates a clear and growing trend. Over the last decade, the proportion of articles reporting the utilization of R displayed a steady growth, starting at 28.3% in 2013 and progressively climbing to an impressive 60.9% in 2022 (as illustrated in Fig. 1). The correlation between time and the percent of R used was significant, with a coefficient of r = 0.98 (P < 0.001).

Notably, in the most recent 3 years (2020, 2021, and 2022), over half of the research papers preferred R for statistical analysis, with percentage varying from 52.2% in 2020 to 60.9% in 2022. This finding reinforces the affirmation that the R language has firmly established itself as the primary tool for data analysis in forestry research.

Throughout our comprehensive investigation covering the period from 2013 to 2022, a clear and notable upward trajectory emerged in the proportion of articles utilizing R for data analysis across the selected journals (Fig. 2). However, it is important to highlight that certain journals displayed significant fluctuations, and the rate of R usage increase varied among the different publications (Fig. 2). Journal of Vegetation Science (JVS) notably excelled in its dedication to utilizing R, achieving an outstanding usage percentage of 73.1% (793 out of 1085) over the course of a decade. Impressively, as early as 2013, JVS demonstrated a substantial R usage percentage of 57.3%, and this rate exhibited an average annual growth of 3.4%. By 2022, the R usage percentage within JVS had soared to an astonishing 85.7%, highlighting a strong and consistent preference to R within this journal.

Applied Vegetation Science (AVS) ranked the second in terms of the average R usage ratio, with a percentage of 66.1% (444 out of 672). Similar to JVS, AVS exhibited a steady growth trend, with an average annual increase of 3.2%. It progressed from a 47.1% R usage percentage in 2013 to an impressive 80.3% in 2022, as illustrated in Fig. 2. The similarity in their patterns can be attributed to the common focus of both journals on vegetation science.

Forest Ecology and Management (FEM) started with a modest R usage rate of 18.1% in 2013, but it showcased the most rapid average annual growth among all the journals at a rate of 4.5%. By 2022, the R usage percentage in FEM had surged to a notable 67.4%. Consequently, FEM ranked the third position in the overall 10-year average R usage percentage of 51.2% (3283 out of 6412).

In comparison, Forestry, Tree Physiology, and Annals of Forest Science (AFS) demonstrated moderately 10-year average R usage percentages at 42.8%, 42.6%, and 40.8%, respectively. European Journal of Forest Research (EJFR) and Agricultural and Forest Meteorology (AFM) displayed relatively lower proportions of R language use with a 10-year average of 30.9% and 27.3%, respectively. It is important to note that, except for AFS, which had a minimal growth rate of 2.8%, the average annual growth rates for the remaining journals all surpassed 3%, indicating a general rapid upward trajectory in R adoption within these journals.

R package usage patterns

In our extensive analysis of research articles, we observed that researchers harnessed a wide array of over 1020 R packages for data statistics and visualization purposes. Notably, 17 packages emerged as stalwarts, being utilized in more than 100 articles (Fig. 3). At the forefront of the list was the “vegan” package, a versatile and widely recognized tool primarily used for multivariate analysis in community ecology (Oksanen et al. 2022). The second most common R package was “lme4,” a versatile package renowned for its broad functionality in fitting and dissecting linear mixed models (Bates et al. 2015). The third one was “nlme,” a versatile package for modeling both linear and nonlinear mixed models (Pinheiro et al. 2020). In the fourth position was “MuMIn,” a package that significantly streamlines information-theoretic model selection and averaging using information criteria (Bartoń 2022). And the fifth one was the “ggplot2” package, renowned for its pivotal role in enhancing data visualization (Wickham 2016). For a comprehensive breakdown of the 17 most frequently used packages, please see Supplementary Table S1.

The rich tapestry of research fields covered by the various journals naturally gave rise to distinct sets of frequently employed R packages, as illustrated in Fig. 4. Among these, the “vegan” package clearly emerged as the dominant choice in JVS, AVS, and FEM. The significance of this finding is underscored by the fact that articles published in these three journals accounted for a substantial 54.9% of the total articles (8169 out of 14 864), firmly establishing “vegan” as the package of choice within these journals, as illustrated in Fig. 3.

Conversely, the “nlme” package claimed the title of being the most commonly employed package in four journals: AFM, AFS, Forestry, and Tree Physiology. However, due to the relatively lower overall proportion of R usage in these four journals, the “nlme” package found itself in third place in the rankings of overall usage frequency. Interestingly, the “lme4” package only stood out by being the most frequently employed package in the JFR. Nevertheless, it managed to secure the second position in the usage rankings across the remaining seven journals, ultimately earning the second spot in the overall package rankings.

These observations underscore the diverse and strategic choices made by researchers in different fields and across various journals when it comes to selecting R packages. This diversity highlights the adaptability of R to cater to a broad spectrum of research needs and underscores the vital role played by specific packages in supporting various facets of data analysis within the realm of forestry research.

A noteworthy occurrence is the observable trend of a rising average number of packages utilized per article, particularly accentuated after 2019 (refer to Fig. 5). This observation signifies a gradual deepening of sophistication in data analysis approaches within the field of forestry studies, characterized by an increased reliance on a greater number of R packages per paper. Additionally, the phenomenon can be elucidated by the expanding quantity of R packages. The escalating abundance of these tools enhances researchers’ capabilities, offering a diverse range of resources for conducting more intricate and sophisticated analyses in the realm of forestry studies. This growing availability of R packages equips researchers with a more extensive set of tools, empowering them to explore and interpret data in increasingly nuanced and advanced ways within the context of forestry research.

DISCUSSION

In recent years, the landscape of modern forestry science has undergone a profound transformation, primarily driven by the seamless integration of computational techniques and programming into the discipline (Polinko and Coupland 2021). Within the field of forestry research, a variety of statistical software programs are commonly employed, each offering its own unique strengths and limitations (Atkins et al. 2022). Nevertheless, what’s particularly noteworthy is the notable surge in the adoption of R as the main statistical tool in research articles published in the eight top forestry journals. This adoption percentage has surged from 28.3% in 2013 to an impressive 60.9% by 2022. This substantial growth underscores the ascent of the R language as the foremost data analysis tool in contemporary forestry research. These trends align with findings from other bibliometric studies in related fields, such as ecology (Lai et al. 2019) and photosynthesis (Liu et al. 2022). R’s robust statistical capabilities, advanced data visualization tools, strong community support, open-source nature, and accessibility collectively, establish it as an attractive and preferred choice for data analysis and research across various disciplines, including forestry.

As our literature survey process focused solely on assessing the prevalence of R usage, we could not quantify trends related to other computer programs in forestry journals. Nevertheless, a more comprehensive analysis across diverse academic journals indicates a significant increase in the use of R, coupled with a decrease in the adoption of proprietary tools like SAS, SPSS and MATLAB from 2008 to 2016 (source: http://r4stats.com/articles/popularity/). Additionally, there is a possibility that researchers gradually reduced their dependence on expensive commercial software, shifting toward free open-source alternatives like R or Python, as proposed by Tippmann (2015). Although we did not evaluate Python’s popularity in these articles, our data reveal that R holds dominance in the top eight forestry journals. The primary factor contributing to R’s prominence in this context is its provision of more specialized libraries tailored for forest field analysis (Atkins et al. 2022).

The dominance of the R language in the Journal of Vegetation Science and Applied Vegetation Science, among these eight forestry journals, can be credited to the robust alignment of these journals with the field of vegetation ecological science. The specific focus on vegetation ecology in these journals necessitates robust data analysis tools like R to handle the intricate forest community data often encountered in this domain. As a result, these journals have naturally gravitated toward R as their main statistical software. Conversely, the relatively lower utilization of R in journals like Forestry, Annals of Forest Science, and European Journal of Forest Research can be linked to their primary emphasis on pure forestry science. Forestry, as a discipline, traditionally places a stronger emphasis on traditional forestry practices and may not require the extensive data-intensive analysis that R excels in. This difference in research focus has led to a less frequent adoption of R in these journals. In contrast to pure forestry, the realm of ecology constitutes a data-intensive domain of study, often requiring advanced computational skills (Carey et al. 2019; Lortie et al. 2020; Davis and Kay 2023).

The scope of Forest Ecology and Management encompasses both ecological and forest manage fields. This dual focus results in an intermediate level of R language usage. While this journal handles ecological aspects, it also covers the forest management and forestry aspects, leading to a balanced adoption of data analysis tools like R. Lastly, in Agricultural and Forest Meteorology, articles predominantly center on the complex dynamics of ground-atmosphere fluxes and the intricate interplay between ecosystems and climate. This specialized field demands highly complex models, and there is a relative scarcity of extensive packages available in R. Consequently, this journal exhibits the lowest proportion of R language usage among all the journals analyzed, underscoring the unique demands and emphasis of this specific research area that are currently not fully addressed in R. Moving forward, it becomes imperative to consider developing additional R packages with the capacity to handle large-scale ecosystem and climate models.

The impressive statistical capabilities of R owe much of their strength to the extensive library of packages available, playing a crucial role in enhancing R’s analytical prowess and adaptability. These packages provide researchers with a wide array of specialized tools to perform intricate data analysis. In parallels with the patterns of package utilization observed in ecology journals (Lai et al. 2019), the three most frequently employed packages are “lme4,” “vegan,” and “nlme.” However, minor variations in their rankings indicate differences between the fields of ecology and forestry. In ecology journals, “lme4” claims the top position as the most frequently used package. This is not surprising as “lme4” is specifically designed to address the common issue of non-independence often encountered in ecological data (Harrison 2014; Harrison et al. 2018; Schielzeth et al. 2020, Lai et al. 2022). The nature of ecological data frequently involves complex relationships, hierarchies, and repeated measurements (Bolker et al. 2009), and “lme4” is well suited to handle these scenarios. As a result, it has gained significant attention in ecological research, helping researchers model and analyze data with the level of sophistication necessary for ecological studies. On the other hand, in the realm of forest ecology, where the central focus is on multispecies forest communities, it is understandable that the analytical emphasis is on diverse communities composed of multiple species. This unique focus has led to a shift in preference to the “vegan” package, which excels in multivariate analysis, a common requirement in forest ecology. “vegan” provides the tools needed to explore and understand the intricate relationships within these diverse forest communities, making it the preferred package for this field. This observation underscores the existence of both shared and distinctive characteristics in terms of statistical analysis between ecology and forestry, reflecting the specialized needs and nuances of each discipline.

Certainly, different journals have their unique focuses, but there is significant overlap among the papers within these journals. Due to the extensive volume of literature, we refrained from classifying the articles by research field during the survey process, resulting in the inability to categorize our data into more detailed areas. The limitation of this study pertains to the focus on R and R packages at the journal level, and we recognize this as a drawback. Future literature surveys will incorporate more detailed categorizations to obtain a more comprehensive and nuanced dataset.

Reproducibility is a crucial characteristic across various research domains in contemporary natural sciences (Powers and Hampton 2019). Essentially, the popularity of the R language greatly promotes research reproducibility by providing a transparent, standardized, and well-documented platform for data analysis and statistical modeling. This widespread adoption of a common tool not only encourages consistency but also boosts transparency, creating an environment where research findings can be independently validated and verified by other scientists. The widespread acceptance of the R language within top forestry journals has substantial implications for advancing the field of forestry as an open science. By adopting R as a common tool, researchers in forestry science embrace a commitment to transparency, consistency, and collaboration. This collective effort not only strengthens the scientific rigor of forestry research but also invites a broader audience to evaluate, validate, and build upon existing findings. Ultimately, the open science ethos cultivated by R’s popularity in forestry journals is a driving force for scientific progress and innovation in the field.

Despite the manifold advantages that R brings to the realm of forestry research, it is essential to acknowledge that certain challenges persist on the path to its full utilization. These hurdles encompass steep learning curves, demanding computational requirements, and the intricacies involved in integrating diverse data sources. The intricate nature of R, while empowering for seasoned users, can indeed present a steep learning curve for newcomers. Overcoming this challenge necessitates the development of user-friendly interfaces and comprehensive educational resources that can facilitate the adoption of R within the forestry community. Simplifying the user experience and providing accessible training materials can significantly reduce the barriers to entry and promote wider acceptance of this powerful tool.

Another challenge lies in the computational demands of working with substantial forestry datasets. As research in forestry often involves complex, large-scale data, optimizing R to handle big data efficiently is of paramount importance. This entails the development of more robust algorithms, improved memory management, and the capacity to leverage parallel processing to expedite computations. By enhancing R’s capabilities in this regard, it becomes better equipped to address the data-intensive requirements of forestry research. Additionally, the integration of diverse data sources, which is a common necessity in forestry studies, can be intricate. R should evolve to facilitate seamless data integration from a variety of formats and sources, streamlining the process for researchers who must work with heterogeneous datasets.

Moving forward, the future of R in forestry research should center on enhancing its usability through intuitive interfaces, reinforcing its capacity to handle big data efficiently, and fostering interdisciplinary collaborations. By doing so, R can become not only more accessible and powerful but also better positioned to address the intricate and multifaceted challenges that characterize modern forestry research. Such collaborative efforts will not only enrich the utility of R but also promote innovation and progress in the field of forestry science.

Supplementary Material

Supplementary material is available at Journal of Plant Ecology online.

Appendix 1: The data (R file) and code.

Table S1: The top 17 most popular (highest use frequency) packages in the eight main forestry journals.

Acknowledgements

We express our sincere gratitude to the numerous graduate students who enthusiastically enrolled in the “R course” instructed by Dr Jiangshan Lai at the University of Chinese Academy of Sciences and Nanjing Forestry University. Their committed endeavors in conducting comprehensive literature surveys have significantly contributed to establishing the empirical foundation for this research.

Funding

This work was supported by the National Natural Science Foundation of China (32271551), Jiangsu Social Development Project (BE2022792) and the Metasequoia funding of Nanjing Forestry University.

Conflict of interest statement

The authors declare that they have no conflict of interest.

REFERENCES