Site quality is a measure of the ability of a site to produce plant biomass, or stemwood volume in forestry, over a period (Skovsgaard and Vanclay, 2008; Weiskittel et al., 2011). In the context of global climate change, accurate assessments of site productivity are no longer just for timber yield prediction but are fundamental for quantifying forest carbon sequestration potential, modeling ecosystem responses to environmental change, and guiding climate-adaptive forest management, such as afforestation planning and optimizing silvicultural practices. Site quality depends on both climatic and edaphic factors prevailing at the site, and due to wide small-scale variability of soil properties, it can vary considerably within relatively confined areas. Classifying sites based on their site quality would therefore be of great value for regional forest management planning, with implications on forest productivity, carbon sequestration, and the ecosystem’s ability to adapt to climate change (Cheveau et al., 2008; Pretzsch, 2020; Sheil and Bongers, 2020). However, determining site quality from environmental factors directly, remains challenging owing to the fact that the dominant factors influencing stand growth vary among tree species, and these factors interact with each other. Historically, methods for evaluating site quality have predominantly relied on two major approaches: biotic-factor-based methods and geographic-factor-based methods. In these methods, site index and dominant height-DBH relationships have been the most commonly used biotic indicators of site quality, while geographic indicators are easily categorized but often lack biological connections (Skovsgaard and Vanclay, 2008; Weiskittel et al., 2011; Ferrero et al., 2013). The basis of the classical site index theory is the finding that cumulative volume correlates with dominant height regardless of site quality, making dominant-height growth rate an easily measurable proxy for volume or biomass accumulation rate. However, the height-based methods are increasingly being challenged, as height growth is subject to external influences such as management practices and climate fluctuations, which affect the reliability and stability of site quality assessments (Molina-Valero et al., 2019). This inherent variability reduces the method’s accuracy, particularly in forests characterized by mixed species, uneven-aged structures, or complex canopy layers (Huang and Titus, 1993; Arias-Rodil et al., 2015; Duan et al., 2018; Hilmers et al., 2019).

An alternative to height-based site quality index is provided by the use of forest biomass as a direct indicator of site productivity. Biomass reflects the cumulative growth of forest stands and provides a more comprehensive measure of forest productivity than tree height alone (Teobaldelli et al., 2009; Hennigar et al., 2017). Using biomass as the primary indicator of site quality can enhance the accuracy and robustness of site evaluations, particularly in regions where forest conditions are highly variable or where management practices have significantly altered forest structure to select the dominant trees (Fu et al., 2017; Yu et al., 2024). This shift from height-based to biomass-based metrics is particularly essential in forestation projects for enhancing forest carbon sequestration in response to global climate change (Gao et al., 2024).

The lack of accurate, nationwide-scale evaluations of site quality has significantly constrained advancements in forest management and monitoring technologies, particularly in China (Wang et al., 2011). This gap highlights the necessity of developing large-scale, standardized models for assessing site productivity across the country’s diverse forest types (Farooq et al., 2020). The need for accurate evaluations of site quality has become increasingly urgent under the strategy of “carbon peak and carbon neutrality,” proposed by China at the 75th UNGA (United Nations General Assembly) General Debate. As carbon sinks play a key role in mitigating climate change, enhancing forest productivity through targeted management practices requires reliable and precise assessments of site potential (Tong et al., 2020). China’s vast and diverse forest ecosystems, both natural and planted forests, present unique challenges for site quality assessment (Liu and Yin, 2012). Differences in silvicultural systems-such as pure vs. mixed stands, even-aged vs. uneven-aged stands, and natural vs. plantation forests-further complicate the evaluation process (Wang et al., 2021). The diversity in forest composition and structure necessitates the development of flexible, large-scale models that can accommodate these differences at national level (Moreno-Fernández et al., 2018; Aguirre et al., 2022; Pretzsch et al., 2022).

The present research aims to bridge the gap by establishing biomass-based site productivity index models for major broadleaf forest types in China, providing a foundational tool for assessing forest productivity on the national scale. Developing a method to normalize species-specific productivity indices onto a generalized scale, that account for the complex dynamics of China’s diverse forest ecosystems, ensures that site quality potential evaluations are conducted on the same baseline level, which aims to provide forest managers and policymakers with the tools necessary to make informed decisions regarding forest productivity improvement.

Site productivity depends on nutrient availability, water use efficiency, etc., influencing the health and stability of broadleaf forest ecosystem. This study has developed a generalized site productivity index model for broadleaf stands that enables the assessment of spatial variation and pattern characteristics of site quality for different broadleaf species across large regional scales. The period of data used for assessing site quality in China covered 15 years, which is relatively short for many sites in relation to environmental variability that may cause fluctuations forest biomass, and probably too short for detecting any significant trend-like dynamic changes expected over longer cycles (Cao et al., 2023). For instance, the long-term accumulation and decomposition of litter can increase available nutrients like phosphorus and nitrogen in stand soil, improving soil texture and structure (Högberg et al., 2017; Kooch et al., 2017). Conversely, unsound management practices may lead to sustained declines in soil organic matter, resulting in deteriorating soil porosity and nutrient levels (Spies, 1997). These factors collectively suggest that in the long term, site quality should exhibit a trend of either improvement or degradation, and evaluation results should capture these dynamics.

Key factors affecting site quality, such as soil type, thickness, texture, and nutrient status, will generally remain stable without human interference and are less affected by climate variations, suggesting that the evaluation should also have stability, which is particularly important for constructing site productivity index models and determining reference ages. Consequently, it is recommended that evaluations focus primarily on middle-aged and near-mature stands to reduce the occurrence of site quality class fluctuation and minimize the uncertainty in the evaluation results when young or over mature stands are used. In practice, fast-growing broadleaf stands in in tropical and subtropical regions (e.g., of Populus spp. and Eucalyptus spp.), are often harvested earlier based on silvicultural objectives. For example, 82.41% of the Populus spp. modeling sample stands had an age of less than 30 years and were primarily distributed in Northwestern forest areas or natural primary forest areas, and thus we lacked sample data for naturally mature stands. It is essential to minimize shifts in the site quality index caused by comprehensive factors such as human interference and climate change (Bontemps and Bouriaud, 2014; Bontemps et al., 2009; Skovsgaard and Vanclay, 2008; Janssen et al., 2018; Aertsen et al., 2010). The evaluation of site quality for fast-growing and high-yield plantations should differ from the general methods used for assessing forest site quality, acknowledging their unique growth dynamics and management practices. This adjustment ensures that the evaluation better aligns with the realities of forest management, particularly for fast-growing species where traditional site quality assessments may not apply effectively. Meanwhile, fitting results suggest that while the Richards model is generally suitable for most broadleaf species stands, exceptions must be made for Robinia spp. stands, where the Logistic model better aligns with biological growth patterns and provides a more accurate representation of site productivity.

This study’s findings align with this principle, showing that broadleaf stand productivity improved continuously from 2003 to 2018, radiating outward from core areas. Although there is a slight fluctuation within 5-year intervals, the general trend is positive. A critical consideration is the observed fluctuation. True site quality, which is governed by stable edaphic and topographic factors, is not expected to change rapidly. Our biomass-based index, however, reflected the stand’s realized productivity, which is sensitive to short-term influences such as inter-annual climate. Variability partly caused by several sources of methodological uncertainties, including propagation of errors from allometric biomass models, potential measurement inaccuracies in plot data, and limited model robustness at the extremes of the age distribution (i.e., in very young or old stands), too. Therefore, some fluctuation between classes is expected. However, the consistent increase in higher-quality sites (Classes I and II) over the entire 15-year period suggests an improvement trend in underlying productive capacity, likely attributable to factors such as improved silvicultural practices. It is important to clarify that our site productivity index, being derived from stand biomass and age, is an indicator of realized stand productivity rather than a direct measure of intrinsic site quality defined by soil properties, topography, and geology. While the trends observed in Figures 4, 5 likely changes in site quality, we cannot disentangle these from the effects of normal stand development or inherent site characteristics without incorporating these environmental variables. Future research should aim to integrate soil physicochemical properties and climate data to establish more mechanistic models of site quality, thereby separating long-term site potential from short-term productivity fluctuations to support the positive improvement.

In forestry practice, the choice of evaluation methods depends on data availability, economic and technical feasibility, and the required scale of analysis (Aguirre et al., 2022; Ding et al., 2021). Our scatter testing curves suggest that the site productivity curves represented a well-defined split of stands into site quality classes and the Spearman correlation coefficient between productivity index classes of major broadleaf stands and the converted generalized index classes exceeded 0.99. These findings corroborate that forest biomass is a direct indicator of stand productivity. As a result, volume or biomass-based site quality assessment can be recommended for use in forest management practice for its effectiveness and potential to minimize the impact of methodological bias on the accuracy of site productivity assessments. Our approach ensures a more realistic and precise reflection of the land’s productive capacity, even though it may result in slightly lower test values compared to traditional height-based evaluations. Indirect methods, such as using stand mean height or dominant height to assess site quality, assume that tree height reflects site quality differences without being affected by stand density or management practices (Adeyemi, 2016). However, this assumption may be challenged, especially in mixed, uneven-aged, or multi-layered stands, where height may not accurately reflect site quality (Padilla-Martínez et al., 2022). Additionally, there is some subjectivity in determining dominant and mean heights in practice, leading to uncertainty when using height as an indicator.

Other indirect methods, like using indicator vegetation types to classify site types, have been found effective in boreal forests where the number of species is limited (Cajander, 1949; Heiskanen et al., 2018). However, the correlation between indicator vegetation types and forest productivity may be weak for more complex ecosystems. Direct evaluation methods, such as using yield volume, do not represent the net primary production due to differences in wood density among species, unless volume is converted to biomass (Gerwin et al., 2018). Biomass-based evaluations have their own limitations, as they assume that stands can achieve full stocking density, a canopy cover of 1.0, and full utilization of site productivity (Ji et al., 2012), which are difficult to achieve in real-world forests. In this study, a requirement of canopy cover greater than 0.70 and no harvesting was set as a prerequisite. Meanwhile, biomass data is more challenging to obtain than height data. All site quality measures share the difficulty of estimating potential productivity from measurements of actual stands that are always more or less constrained by management interference and other non-site factors such as disturbances.

This study was focused on evaluating the range of productivity and hence site quality classes of broadleaved species in China, without attempting to explain these classifications by any independent variables. However, such variables (e.g., soil health and climate variability) have been suggested to be incorporated into site quality assessments (Li et al., 2022). Also, it is important to understand species-specific growth responding to varying environmental conditions (Szatniewska et al., 2022), which further underscores the complexity of accurately assessing site productivity (Babst et al., 2013). Wu et al. (2015, 2022) have pointed out that site quality evaluation should encompass both soil site (reflecting land productivity) and stand site (reflecting the forest’s growth status). Process-based modeling studies have suggested to split the impacts into climate that influences carbon assimilation (Peltoniemi et al., 2015; Tian et al., 2020), and edaphic factors that influence carbon allocation belowground, and that site index and productivity should be derived from a combination of both (Mäkelä et al., 2016, Härkönen et al., 2019). However, it is challenging to analyze site quality changes and construct a unified site quality model on the large spatial scale, as climatic differences among regions affect species growth rates and impact the age-biomass relationship (Repo et al., 2021). Nonetheless, the generalized productivity index developed here has significant applications in modern ecological and environmental management beyond traditional forestry. For example, it can be integrated into large-scale carbon budget models to provide more accurate estimates of regional carbon sinks. It can also serve as a key variable in assessing forest vulnerability to climate change, as sites with higher productivity may exhibit different sensitivities to drought or temperature stress. For policymakers, such a tool enables the targeted allocation of resources for forest restoration projects, ensuring that investments are directed toward areas with the highest potential for biomass accumulation and carbon capture. Thus, developing reliable and scientifically sound models for evaluating site quality is essential for improvement of forest management at national and global levels.

In this study, we established site productivity index models for major broadleaf stands based on the age-biomass relationship to evaluate site quality. On this basis, a generalized broadleaf stand productivity index model was constructed using conversion coefficients. Statistical tests validated the feasibility of the generalized model, making them suitable for use in forestry management and prediction. The generalized productivity index model facilitates the evaluation of spatial changes and pattern characteristics of site quality across different broadleaf species at large regional scales. Analysis of NFI sample plot data from 2003 to 2018, including proportional statistics and spatial analysis, indicated a continuous improvement trend in the site productivity of broadleaf forest, despite fluctuations during the 5-year intervals caused by methodological uncertainties. The generalized productivity index model demonstrates applicability but requires caution regarding potential biases in evaluation results due to uncertainty, especially when based on small sample data.