Estimating Small-Scale Forest Carbon Sequestration and Storage: i-Tree Eco Model Improved Application

Abstract

Carbon sinks are of great significance for mitigating the greenhouse effect and climate change. However, only a few carbon sink measurement methods are suitable for small-scale research, such as at the city-region scale. Methods that can accurately distinguish the high–low gradients of forest carbon sinks within small-scale areas have not yet been established. To fill this gap, we used a tree allometric growth model—the i-Tree Eco model—and applied it to Tai’an, which is a National Forest City in China. By using indicator conversion methods, we innovatively combined the China Forest Resources Inventory Geographic Information Database with i-Tree Eco. The results showed that i-Tree Eco successfully estimated the carbon sinks provided by urban–rural forests (in 2019)—the total carbon storage in Tai’an forest was 5,828,165.90 t; the average carbon storage per hectare was 37.19 tC·ha⁻¹; the total carbon sequestration was 936,789.03 tC·yr⁻¹; and the annual carbon sequestration was, on average, 5.97 tC·ha⁻¹·yr⁻¹. Our method improved the spatial resolution of carbon sequestration and storage compared to the commonly used InVEST model, from about 350 m × 350 m to 195 m × 195 m. Compared to the traditional IPCC method, the i-Tree Eco model provided greater accuracy and timeliness in small-scale carbon sequestration measurements, eliminating the need to wait for the next forest inventory to be published. Our method yielded results that covered the entire city region and better reflected the spatial heterogeneity of carbon sinks. We conclude that the innovative application of the i-Tree Eco model to urban–rural-scale carbon sink measurements provides stronger technical support for urban green space planning, as well as data guidance, in relation to local carbon mitigation strategies.

1. Introduction

Forest and grassland carbon sinks are the primary sources of regional carbon sinks; however, among the widely used methods, the precision of their statistical measurements remains relatively low. Carbon sinks refer to biological or artificial procedures, actions, and mechanisms that take away CO₂ from the atmosphere [1]; they are essential components in achieving carbon neutrality and mitigating global climate change [2]. Due to their importance, the quantity and distribution of carbon sinks have become key research areas across various disciplines [1]. Among the two major types of carbon sinks, ecosystem carbon sinks are safer, more stable, and more efficient than artificial ones [3]. Terrestrial forest [4], grassland [5], wetland [4], and cultivated land [6] are the four major carbon sinks in ecosystems. Therefore, forest and grassland carbon sinks are important components that researchers must focus on when studying regional carbon sink distributions. For example, larger forests in many Finnish municipalities compensate for carbon emissions from cropland, mires, and waterbodies (which can be calculated using the LUONNIKAS calculation tool) [7]. It can be seen that the importance of forest carbon sinks is very prominent and there is a need to strengthen the body of research in relation to them.

On the other hand, small-scale, detailed carbon sink inventories are more conducive to implementing regional carbon strategies. To support urban planning, land use, and forest management policies, small-scale data on various carbon sinks are required [7,8]. The development of local carbon mitigation strategies requires detailed carbon sink data. Therefore, the need to address the methodological gap in small-scale carbon sink inventories has become apparent [9].

In current research, carbon sink measurements are primarily based on regional- or national-scale methods. Only a few models have been developed for small-scale carbon sink estimation, including sample plot inventory–sample marking tree analysis, tree allometric equations, CITYgreen, InVEST, CENTURY, IBIS, and LPJ-GUESS. The sample plot inventory method is suitable for fine-scale studies but is labor-intensive and error-prone [10,11]. CITYgreen and i-Tree Eco are the only empirical models that support plant or forest sub-compartment-level estimations, effectively capturing spatial heterogeneity [10,12]. However, the results of CITYgreen lead to some uncertainty and the model involves high costs [12,13,14,15].

Process-based models like InVEST lack sufficient spatial resolution and ignore natural tree growth [10], while LPJ-GUESS requires complex parameters and high labor inputs, limiting its application [16]. CENTURY and IBIS are generally not suited for fine-scale urban forest studies [17].

The i-Tree Eco model is a tree allometric growth model that is more comprehensive and accurate than other models [10,18], estimating carbon sinks using diverse forest characteristics and incorporating meteorological conditions. Unlike other commonly used methods, it avoids forest inventory comparisons, reducing labor and cost while improving accuracy.

Despite its advantages, i-Tree Eco’s reliance on local forest surveys restricts its broader use due to high data acquisition costs [19,20]. Airborne radar remote sensing has been introduced to address this; however, it still remains expensive and technically demanding [8,21,22]. Therefore, there is an urgent need to develop high-resolution, easily applicable, and low-cost carbon sink measurement methods that are suitable for small-scale studies.

To tackle these challenges, this study integrates China’s Forest Resources Inventory database into i-Tree Eco through indicator conversion, enabling citywide (a type of small-scale) carbon sink mapping in Tai’an (a National Forest City in China). The i-Tree Eco model was accessed at https://www.itreetools.org/i-tree-tools-download (accessed on 14 August 2025)). Additionally, our improved estimation process is easily applicable and low-cost, leading to results with a high spatial resolution. This way, to a certain extent, the drawback of high data acquisition costs in the i-Tree Eco model has been overcome.

The i-Tree Eco model, which is relatively new in relation to the application of citywide carbon sink estimation and can effectively differentiate between high and low gradients of forest carbon sinks, aims to complement other commonly used carbon sink models in terms of improving accuracy, spatial resolution, and applicability. We applied this model in Tai’an City, with the help of China’s Forest Resources Inventory database, to assess its suitability for fine-scale forest carbon sink measurement at the citywide level. Our core hypothesis is as follows: the i-Tree Eco model is suitable for this application. In summary, this study’s innovation is reflected in three key aspects. First, theoretically, the proposed i-Tree Eco model addresses the gap in applying fine-scale, application-prone, and high-resolution forest carbon sink measurement methods in urban contexts. Second, in terms of model advantages, we have developed and validated the model’s accuracy. Compared to the commonly used models, this method offers better resolution in relation to carbon density and spatial heterogeneity. Lastly, from a technical perspective, the improved data import process enhances the method’s applicability for small-scale carbon sink studies, reducing research labor intensity.

2. Literature Review

2.1. Carbon Sink Estimation Methods

Seven methods are currently used for small-scale carbon sink estimation. We focused on carbon sink estimation methods that are applicable at the small scale in order to provide support for green space planning in cities, as well as for local carbon mitigation strategies. The current carbon sink estimation methods that may be suitable for small-scale studies are listed in

Table 1. Advantages and disadvantages of current carbon sink estimation models.

Models	Applicable Scale	Advantages	Shortcomings of Carbon Sink Estimation Models at the Small Scale and for Green Space Planning and Management
➀ Sample plot inventory–sample marking tree analysis	Individual plant scale [10]	The method is clear and technology is simple [11], directly measures biomass, and is accurate [10].	The inventory investigating workload is heavy; does not take all types of terrestrial ecosystem into account; carbon sinks are calculated based on the annual change in carbon storage, making the carbon sink estimation error larger [11].
Empirical model simulation method
➁ Tree allometric growth equation (i-Tree Eco model)	Individual plant scale and regional scale	Calculations with the tree allometric growth equation method are quick [10]. i-Tree Eco estimates carbon sequestration and carbon storage of forests based on the different growth rates of different tree species, DBH (diameter at breast height), and total heights.	Direct calculation using the allometric growth equation requires a large amount of measured data and the growth stages and regional differences in tree species must be considered [10]. i-Tree Eco also requires a large amount of inventory investigation work. Therefore, it can generally only be used for small-scale research. It does not consider other ecosystem types.
➂ CITYgreen	City and regional scale/park and community scale [12]	Regional-scale analysis is quick. Park- and community-scale analysis can be used to estimate individual trees without requiring remote sensing images [12].	Regional-scale analysis requires image pre-classification; there is uncertainty in the estimation results [13,14]. Park- and community-scale analyses are small in scope, require digitization, and are time-consuming [12]; the input attributes require sample plot inventory [15]; it is mainly used in artificial forests [15].
➃ InVEST	Regional scale	The operation is simple, input data are easy to obtain, and it can cover various ecosystem types.	The internal heterogeneity of forest carbon density is not considered; attribute modifications are needed based on local conditions [10]; does not consider the carbon sequestration effect of the natural growth of trees.
Mechanism model simulation methods
➄ CENTURY model (belongs to the static model class)	Regional scale [17]	It is a carbon element simulation model based on the structure and function of soil [17].	Lack of refined simulation of the photosynthesis process [17].
➅ IBIS model (belongs to the dynamic model class)	Regional scale [17]	It comprehensively considers geochemical, geophysical, and vegetation dynamic processes. It is often used to simulate complex carbon cycle processes [17].	The large number of different types of attributes required limits its wider application [17].
➆ LPJ-GUESS model (belongs to the dynamic model class)	From species to global scales [17]	It can simulate the structure and function of forest ecosystems at multiple scales including individual species [17,23,24].	The model code is complex; vegetation and soil type data with high spatial resolution must be input; to obtain regional data, extensive field plot surveys or interpolations based on flux site data are needed; attribute adjustment takes a long time [16].

Note: The bold text along with dark gray background indicate the type of the following methods.

. Among these, sample plot inventory–sample marking tree analysis is suitable for small-scale studies; however, the required forest inventory investigation consumes considerable manpower and material resources, which limits its usability [10,11].

Notably, the i-Tree Eco and CITYgreen models stand out for their advantages. Among the various empirical models, only the i-Tree Eco and CITYgreen models can complete calculations at the plant or forest sub-compartment scale and are easy to use [10,12]. Carbon sink calculations at such small scales better demonstrate the spatial heterogeneity of carbon density. Therefore, they provide basic data for the spatial analysis of carbon sinks within cities [10,12]. These two methods have a significant advantage in terms of result information content [10,12,13,14,15].

However, the differences between static and dynamic models—such as in their attributes and empirical parameters—lead to uncertainty in the carbon sink results. Among the mechanism model classes, various static models differ with respect to principles, structures, and attributes, resulting in great uncertainties in the assessment of carbon sinks; the errors involved in the models eventually accumulate in the carbon sink calculation results [11]. In various dynamic models, complex biochemical processes involved in carbon sink formation have been simplified into several key driving factors. Key attributes rely on empirical settings that reduce the estimation accuracy [11]. The applicable scales of mechanism model classes are at least regional (except for the LPJ-GUESS model) and are generally not suitable for small-scale studies.

In practice, different models offer varying advantages. However, the plot inventory–sample tree analysis method leads to large errors in carbon sequestration estimation and is costly [11]; the InVEST method lacks sufficient spatial resolution [10]; the CITYgreen model presents uncertain results and is also costly [12,13,14,15]; the LPJ-GUESS model is difficult to widely apply due to parameter acquisition challenges and high labor costs [16]; and other methods are not suitable for small-scale studies [17]. Based on these factors, we chose the i-Tree Eco model for our study. It is ideal for small-scale (forest) carbon sink estimation and offers detailed estimates of forest carbon sinks, making efficient use of official forest inventory data.

2.2. New Applications of the i-Tree Eco Model for Carbon Sink Estimation

The i-Tree Eco model was developed in the early 21st century and has become widely referenced in urban carbon sink research. The i-Tree Eco [25] model (we used the i-Tree Eco v6.0.32 in this study) is a flexible software application that was developed by the Forest Service of the United States Department of Agriculture in 2006 [26]. It uses data collected from individual trees, complete forest inventories, and sample plots to quantify the forest structure, environmental effects, and ecosystem service values [27]. Information on tree species, tree morphological characteristics, air pollution, and meteorological data is required. In this study, the ecosystem service value included carbon storage and sequestration as two carbon sink indicators. The i-Tree series has grown to cover trees in diverse settings such as urban areas and rural regions [28] and has been validated by many case studies worldwide [29]. In urban research, scholars use i-Tree Eco to determine the models of urban green growth [30]; evaluate one particular ecosystem service, like carbon storage [29] or air quality improvement [31]; estimate the ecosystem services provided by urban parks [32]; or compare different urban green designs concerning ecosystem service provision [33]. The timeline of the development of i-Tree Eco for carbon sink estimation is as shown in Figure 1.

In recent years, various innovative applications and method optimizations have been established for i-Tree Eco in urban studies (

Table 2. Innovative applications and method optimizations of i-Tree Eco in urban studies.

Number	Research Area	Research Topic	Data Prepared in Advance	Analysis Method
1 [21]	Santa Barbara, California, United States	Urban forest ecosystem services	Fusion of hyperspectral imagery; waveform radar data	Discriminant analysis method for the tree crown scale—fusion airborne visible light and infrared imaging spectroscopy; laser radar structure measurement; estimating leaf area index using laser penetration and allometric measurement methods
2 [8]	New York, United States	Urbanization and the relationship between canopy coverage morphology and carbon density	High-resolution land cover map; laser radar-derived tree index; fieldwork data	Radar measurement; aggregate carbon density and landscape information at the census tract level; R statistical analysis
3 [19]	A park in Munich, Germany	i-Tree Eco model’s simulation of light sensitivity and tree species classification; development of a remote sensing observation method for crown light exposure (CLE)	Detailed inventory data of the forest	An automated method combined with remote sensing to calculate the CLE based on tree and crown morphology; assesses how ecosystem services and disservices are altered when the degree of species and CLE detail diminishes
4 [38]	Chicopee and Fall River, Massachusetts, United States	Value and distribution of energy savings provided by saplings; value changes in the future	The inventory of 2271 saplings on the streets or in residential areas planted from 2014 to 2015	Spatial correlation analysis of ecosystem services of saplings and socioeconomic factors in census tracts; spatial correlation analysis of forest ecological characteristics and the ecosystem services they provide
5 [40]	Adama City, Ethiopia	Urban forest ecosystem services—effect on climate change mitigation	Investigation data from 214 sample plots	Leaf area/biomass of trees and shrubs, pollution removal, volatile organic compound emissions, hydrological function, forest ecosystem value, and carbon storage and sequestration were estimated; the structure and makeup of urban forests were investigated and analyzed
6 [20]	Royal Forest of Capodimonte, Naples, Italy	Urban forest ecosystem services (ES) and pollen-related disservice assessment (ED)	Plant community survey and biodiversity value assessment	Analysis of ecosystem services and pollen disservices provided by urban forests through data spatialization; evaluation of ES and ED using i-Tree Eco and Index of Urban Green Zones Allergenicity (IUGZA)
7 [22]	Oslo’s built-up areas, Norway	Filling in the missing and incomplete attributes for i-Tree Eco in current urban forest inventories, thereby enabling quick and low-cost estimation of urban ecosystems	Manual measurement, airborne laser scanning crown images, land use map, and land resource map	Using geospatial analytics and machine learning methods; manual measurement was correlated with airborne laser scanning crown images to complete the tree measurements, employing an ancillary spatial dataset to obtain the missing attributes of trees’ spatial background; machine learning with Bayesian networks was used to deduce the output of i-Tree Eco and infer the ecological value of citywide forest inventories
8 [26]	Perugia, Italy	Citizens participated in fieldwork to obtain a tree inventory, assessing the ecosystem services offered by urban trees	Italian Urban Forest Inventory Dataset; public datasets on weather data, pollutant concentrations and urban morphology; urban greening data collected by citizens	A tool was designed to aid decision-making and management for tree planting; four ecosystem services offered by urban trees were assessed by i-Tree Eco

Note: the bold text indicates emphasis and the light grey background is used to enhance the clarity and organization of the table.

).

Additionally, the i-Tree Eco model is highly suitable for precise forest carbon sink research. The i-Tree Eco model is based on tree species and tree morphological characteristics [25,27]. The model calculation program includes the allometric growth formulae of the trees [18]. Factors, such as crown characteristics and light competition, are also considered by the model algorithm to calculate the annual carbon sequestration and storage of each tree [18]. The calculation results have a high spatial resolution. Annual carbon sequestration can be directly estimated based on the forest resource inventory [26]. Carbon storage in interval years does not need to be compared to obtain annual carbon sequestration [18], which reduces the data requirements. The model avoids large errors in measuring the increment in the tree height and DBH by measuring trees in intervals of years [11]. Therefore, there is no need to wait for the next forest resource inventory to calculate forest carbon sequestration for the current year, which helps improve the timeliness of research.

Moreover, compared to the commonly used methods, the i-Tree Eco model is more comprehensive and efficient [10,18]. Forest carbon storage and sequestration are not averaged; instead, they depend on factors like forest type, tree species, forest age, and volume [18,27]. The heterogeneity of forest characteristics becomes more pronounced at smaller spatial scales. Unlike the commonly used models such as InVEST, i-Tree Eco estimates carbon sinks based on a variety of forest factors. Furthermore, i-Tree Eco emphasizes carbon sequestration resulting from natural tree growth, which is an aspect that other models, such as InVEST, do not take into account. In comparison with plot inventory–sample tree analysis and biomass conversion factor methods, i-Tree Eco does not require comparisons of forest resource inventories between survey years and does not rely on tables. This enables quicker carbon storage/sequestration calculations while saving both labor and resources. Additionally, i-Tree Eco accounts for meteorological conditions and canopy features, which significantly improves estimation accuracy.

Building on this, we sought to introduce new data sources to expand the i-Tree Eco model’s applicability and accessibility. Among the new applications of i-Tree Eco listed in the table above, many projects use local forest inventories as basic data. However, because of the huge cost of human resources in forest inventories, many researchers can only apply their local forest surveys to several communities or parks [19,20]. As a result, their applications are limited. Continuous carbon sink distribution patterns were not observed and the extensive research requirements for the spatial analysis of small-scale carbon sinks cannot be met. To solve this issue, remote sensing observation data from airborne radars have been introduced in some studies [8,21,22]. However, the interpretation of radar images may result in certain errors. Furthermore, the airborne remote sensing method consumes considerable human and material resources and requires high technical expertise, which limits its accessibility. Therefore, we introduced a new data source for i-Tree Eco to reduce the cost of the method and extend its application and accessibility. Based on the China Forest Resources Inventory database and several indicator conversions, we used i-Tree Eco to estimate the forest carbon sink distribution at a small scale (taking Tai’an in Shandong Province, China, as an example).

3. Methodology

3.1. Estimation Models

3.1.1. i-Tree Eco Model

i-Tree Eco is based on a database of species-specific allometric equations [35]. We used the i-Tree Eco v6.0.32 (USDA Forest Service, Washington, DC, USA) in this study. The carbon sink estimation equations are as follows (different formulas are used depending on the data and tree species):

Biomass = exp(β₁ + β₂ × LN(DBH) + σ²/2)

(1)

Biomass = exp(β₁ + β₂ × LN(DBH² × HEIGHT) + σ²/2)

(2)

Biomass = β₁ × (DBH^β2)

(3)

Biomass = β₁ × ((DBH² × HEIGHT)^β2)

(4)

where β₁ and β₂ are coefficients specific to species, DBH is measured 1.37 m above the ground, HEIGHT denotes the total tree height, and σ² represents the variance of model errors, which is utilized to correct the transformation bias when predictions are back-transformed from the logarithmic scale to the original scale [41]. The estimated biomass of the whole tree (comprising both above-ground and below-ground biomass) is multiplied by 0.5 for conversion to carbon storage [42].

Based on the Tai’an Forest Resources Inventory database from 2019, the data were converted and input into an attribute table. First, the data year was entered into the i-Tree Eco model software (2019). Second, the data type was entered into a complete inventory. Third, the location was entered, i.e., Jinan, Shandong Province (Jinan is the closest optional address to Tai’an); we chose to use 2019 meteorological data from the Mount Tai weather station, which is in Tai’an. Fourth, the tree species, tree DBH, crown size (including height to live top, height to crown, crown width, and percent crown missing), total tree height, and land use were input. With respect to the attribute “crown health”, we selected to input “dieback”. The above-mentioned input indicators include all attributes the i-Tree Eco software requires or strongly recommends.

Because the input attributes required by i-Tree Eco could not all be found in the Tai’an Forest Resources Inventory database, similar indicators included in the Tai’an Forest Resources Inventory database were converted so that they could be as close as possible to the input attributes required by i-Tree Eco. In previous studies, to supplement incomplete attributes within existing municipal inventories, Cimburova, Z. and Barton, D.N. developed geospatial and machine learning methods that follow attribute definitions in accordance with the i-Tree Eco Field Guide [22]. A similar input attribute transformation was used by Scholz, T. et al. [36] in German urban forest research. Therefore, we can also perform some indicator conversions, provided that the conversion processes are reasonable.

In addition, i-Tree Eco (complete inventory mode) can be used as a carbon sink calculation method for a single tree. However, each item in the Tai’an Forest Resources Inventory database represents a woodland patch in which each tree species is located, providing indicators, such as tree species, average tree height, average DBH, and the overall crown health condition of the woodland patch. Therefore, the “number of trees per hectare” index BCZS and the woodland patch area in the Tai’an Forest Resources Inventory database were used to convert the carbon sink of a single tree evaluated using i-Tree Eco into the total carbon sink of all trees in that woodland patch. The total carbon sink was derived from the average carbon sink per tree. The formula is as follows:

The total carbon sink of all trees in a woodland patch = carbon sink of a representative tree (the species of this tree is the dominant species in this woodland patch; its height is the average height of all trees in this woodland patch; its DBH is the average DBH of all trees in this woodland patch; and its crown health condition is consistent with the overall crown health condition in this woodland patch) × BZCS × area (in hectares)

(5)

BCZS is the number of trees per hectare within this woodland patch; this is a survey indicator included in the Tai’an Forest Resources Inventory database.

Our indicator conversion methods include the following (Figure 2 illustrates where the relevant indicators are located on the tree):

1.

The ZHDJ indicator (“disaster level”) in the Tai’an Forest Resources Inventory database corresponds to the dieback indicator (required by i-Tree Eco). Their correspondence is shown in

Table 3. Correspondence between the “disaster level” ZHDJ indicator and the dieback indicator, as derived from [43]. NAFT refers to the number of affected forest trees. The dieback indicator is set to be equal to the average tree damage severity.

“Disaster Level” ZHDJ Indicator [43]	Disaster Types: Forest Pests and Diseases		Disaster Types: Forest Fire		Disaster Types: Climate Damage and Others
	Evaluation Standard [43]	Average Tree Damage Severity	Evaluation Standard [43]	Average Tree Damage Severity	Evaluation Standard [43]	Average Tree Damage Severity
0 (approximate to zero)	NAFT is less than 10%	10%	No notable damage	5%	No notable damage	5%
1. Light	NAFT is 10–29%	20%	NAFT is less than 20%; still able to resume growth	10%	NAFT is less than 20%	10%
2. Medium	NAFT is 30–59%	44%	NAFT is 20–49%; growth is significantly inhibited	40%	NAFT is 20–59%	44%
3. Severe	NAFT is more than 60%	80%	NAFT is more than 50%, comprising mainly dying and dead wood	80%	NAFT is more than 60%	80%

Note: the light grey background is used to enhance the clarity and organization of the table.

.

2.

The formula for height to live top is as follows:

Height to live top = Total tree height − Total tree height × CR × (Dieback/100)
CR: Crown ratio = (Crown length/Total tree height)

(6)

3.

The formula for height to crown base is as follows:

Height to crown base = Total tree height − Total tree height × CR × (1 − Dieback/100)

(7)

4.

The formula for crown width is as follows:

$$\mathrm{Crown} \mathrm{width}=2\times \sqrt{(1 \mathrm{h}\mathrm{e}\mathrm{c}\mathrm{t}\mathrm{a}\mathrm{r}\mathrm{e}\times \mathrm{c}\mathrm{a}\mathrm{n}\mathrm{o}\mathrm{p}\mathrm{y}{\displaystyle \frac{\mathrm{d}\mathrm{e}\mathrm{n}\mathrm{s}\mathrm{i}\mathrm{t}\mathrm{y}}{\mathrm{B}\mathrm{C}\mathrm{Z}\mathrm{S}}})/\mathsf{\pi }}$$

(8)

BCZS is the number of trees per hectare. BCZS and “canopy density” are survey indicators included in the Tai’an Forest Resources Inventory database.

5.

The formula for percent crown missing is as follows:

Percent crown missing = 39.6 × (1 + Dieback/100)

(9)

The coefficient “39.6” is from Ref. [44].

6.

The “canopy density” indicator in the Tai’an Forest Resources Inventory database corresponds to the crown light exposure indicator. Their correspondence is shown in

Table 4. Correspondence between the “canopy density” indicator and the crown light exposure indicator, as derived from [45].

Canopy Density	Crown light Exposure
≥0.80	1
0.6–0.8 (including 0.6)	2
0.34–0.6 (including 0.3)	3
0.1–0.3 (including 0.1)	4
0–0.1 (not including 0.1)	5

.

• Regarding the correlation between canopy density and canopy lighting conditions, reference was made to previous research on forests in the areas of low mountains and hills of the Yangtze River Basin [45].

The derivation of the indicator conversion (Equations (A1)–(A6)) is shown in the Appendix A. The key variables input into i-Tree Eco for carbon sink estimation are listed in

Table 5. Key variables of the i-Tree Eco model.

Indicator Name		Corresponding to Tai’an Forest Resources Inventory Attribute Table Code or Indicator Conversion Equation	Quantity Unit	Interpretation (i-Tree Software Interface Notes) [25]
Species		YOU_SHI_SZ (means dominant tree species)		The names of the species and genus of each tree were identified and documented
DBH		PING_JUN_XJ (means the average of DBH)	cm	Precise measurement or categories of the DBH of the stem for each individual tree
Land Use		DI_LEI (means land type)		The land use type where the tree is situated
Total tree height		PING_JUN_GAO (means the average of tree height)	m	The height from the ground to the uppermost part (whether alive or dead) of the tree
Crown size	Height to live top	Height to live top = Total tree height-Total tree height × CR ×(Dieback/100)	m	The height from the ground to the living top of the tree
	Height to crown base	Height to crown base = Total tree height − Total tree height × CR × (1 –Dieback/100)	m	The height from the ground to the bottom of the living crown of the tree
	Crown width	2 × SQRT ((1 hectare × canopy density/BCZS)/π)	m	The crown width in two directions, namely north–south and east–west
	Percent crown missing	Percent crown missing = 39.6 × (1 + Dieback/100)		Percentage of the crown volume that is unoccupied by branches and leaves
Crown health	Dieback	ZHDJ corresponds to the percentage of disease, as shown in Table 3 [43]	%	An estimation of the percentage of the crown that consists of dead branches
Crown light exposure		Enter crown light exposure according to canopy density; correspondence is shown in Table 4 [45]		The number of tree sides that receive sunlight from above (maximum of 5)

Note: the bold text indicates emphasis and the light grey background is used to enhance the clarity and organization of the table.

.

The flowchart of this study is shown in Figure 3.

3.1.2. InVEST Model

This method serves as a comparison and supplement; it uses land use maps to estimate carbon sinks. We used InVEST 3.9.2 x64 (Stanford Woods Institute for the Environment, Palo Alto, CA, USA). The estimated results include carbon storage and carbon sequestration.

The “Carbon Storage and Sequestration” module in the InVEST model and the current Tai’an land use map were utilized as the current LULC (land use and land cover). The Tai’an land use map of the next phase was employed as the future LULC.

Carbon pool values are input parameters that are required by the InVEST model. The carbon pool values represent the carbon content of each hectare of various types of land, including the carbon in the above-ground part, the carbon in the below-ground part, soil carbon, and carbon in dead organic matter. The carbon pool values in this study were adopted from the average of the carbon pool values in Shandong Province and the neighboring provinces of Shandong (the data are from previous studies, as shown in Appendix B). Shandong Province is where our study area Tai’an is located. We input

Table 6. Input carbon pool values (tC·ha⁻¹).

Lucode	LULC_Name	C_above	C_below	C_soil	C_dead
112	Cropland	7.05	6.64	65.76	0.43
222	Woodland	31.38	17.28	101.89	3.81
333	Grassland	7.44	11.56	63.93	2.57
444	Water area	0.74	0.37	26.81	0.32
555	Built-up land	2.88	1.58	45.96	0.21
666	Unused land	1.56	2.66	16.34	0.18

Note: The light grey background is used to enhance the clarity and organization of the table.

into the InVEST model to derive the carbon storage and carbon sequestration distributions for Tai’an (Table 6; see Appendix B for the derivation process).

All land use types in our data sources corresponded to the six land use types in the InVEST model—cropland, woodland, grassland, water area, built-up land, and unused land. These categories were used to obtain the three-phase LULC data of Tai’an in 2009, 2017, and 2019. The LULC data were input into the InVEST 3.13.0-workbench software to estimate carbon sinks. The resolution of the LC data used is approximately 350 m × 350 m.

3.1.3. IPCC Forest Carbon Storage Calculation Method

Forest carbon storage was calculated according to the 2006 IPCC (Intergovernmental Panel on Climate Change) on Climate Change Guidelines for National Greenhouse Gas Inventories.

$$C={\sum }_{i,j}\left\{A_{i,j}·V_{i,j}·{BCEF}_{S_{i,j}}·(1+R_{i,j})·{CF}_{i,j}\right\}$$

(10)

$${BCEF}_I={BEF}_I·D$$

(11)

where

C: the total carbon in biomass;

A: area of the land;

V: timber volume;

BCEF: the biomass conversion coefficient that converts the timber volume to above-ground biomass;

R: the ratio of subsurface biomass to above-ground biomass;

CF: the carbon content of dry matter;

D: wood basic density;

BEF: biomass expansion factor;

i: Ecological zone i;

j: Climate domain j.

We primarily performed this calculated based on the “living tree stock per hectare” and “tree species” attributes in the Tai’an Forest Resources Inventory database. We set the “standing stock volume of living trees per hectare” to V.

Our study area belonged to the TeDC temperate continental forest.

• CF: temperate broadleaf forest—0.48; temperate coniferous forest—0.51 [46].
• BCEF_s: The BCEF_s values for hard broadleaf, pine, and other coniferous species were the average BCEF_s recommended by the 2006 IPCC Guidelines for National Greenhouse Gas Inventories. The BCEF_s value of soft broadleaf trees refers to the biomass expansion factor (BEF = 1.586) and basic wood density (D = 0.443) of soft broadleaf species in the temperate broadleaf forests of Heilongjiang Province, China [47]. Based on multiplying BEF and D, the BCEF of soft broadleaf species was 0.7026.
• R: R was set according to the value recommended by the 2006 IPCC Guidelines for National Greenhouse Gas Inventories and based on the tree species and size of the above-ground biomass.

3.2. Study Area and Data Acquisition

3.2.1. Study Area

Tai’an City is located in Shandong Province, China, covering an area of 7762 km². Its location is shown in Figure 4. Tai’an had a population of 5,635,000 in 2019; it is rich in forest resources and is known as China’s “National Forest City”. By the end of 2020, Tai’an City’s forest area was 200,316 ha, the forest coverage was 25.81%, and the total standing stock volume of living trees was 10.81 million m³. Tai’an City had 1189.80 ha of new afforestation between 2021 and 2022 [48]. The forest resource inventory and land use data of Tai’an meet the needs of carbon sink research. Therefore, Tai’an was selected as the study area. The carbon sink function of the forest should be significant, the total carbon sink and carbon density in the municipality should be relatively high, and the spatial heterogeneity of carbon sinks should be evident. The capability of different models to demonstrate the spatial heterogeneity of carbon sinks was tested using this area as a case study.

3.2.2. Data Sources

1.

Land use data

We used Tai’an’s Second National Land Survey data (2009), Tai’an’s Second National Land Survey—Change Survey data (2017), and Tai’an’s Third National Land Survey data (end of 2019) in this study. All the land use data are sourced from the Tai’an Municipal Bureau of Natural Resources and Planning.

2.

Forest resources inventory data

Data were obtained from the Tai’an Forest Resources Inventory Geographic Information Database (2019). The database is based on the technical method of China’s forest resource inventory, which is a combination of forest subcompartment mapping and subcompartment surveys promulgated by the China National Forestry Administration and operated by the Tai’an Forestry Department. China Forest Resources Inventory data comply with unified national survey standards and cover the whole territory. Therefore, these high-quality data were suitable for our research. The mean tree height was 10.00 m and the average DBH was 11.46 cm. The data included 52 tree species, dominated by Populus tomentosa, Robinia pseudoacacia, and Platycladus orientalis.

The Tai’an Forest Resources Inventory database includes 42,682 woodland patches. After excluding invalid data that lacked tree height or DBH information, 41,031 items of valid data were obtained (accounting for 96.13% of all data), representing 41,031 woodland patches. These items represent the patches of inventory plots. Each sample plot was divided into several patches of different shapes according to the dominant tree species. We matched all 52 tree species to predefined species in the i-Tree database. The meteorological data we used were obtained from the 2019 Mount Tai Weather Station in the i-Tree Eco software.

3.3. Methods of Statistical Analysis

We used correlation analysis, linear regression models, and one-way ANOVA (F-test) for statistical analysis. The statistical analysis software used was IBM SPSS Statistics 23 (IBM, Armonk, NY, USA).

4. Results

4.1. Carbon Storage

First, we utilized the InVEST model to estimate carbon storage in Tai’an City based on a land use map. The carbon storage density in Tai’an in 2009 was 0.86 tC·ha⁻¹, while the total carbon storage was 664,407.37 t. In 2017, the carbon storage density in Tai’an was 0.86 tC·ha⁻¹ and the total carbon storage was 663,631.19 t. In 2019, the carbon storage density in Tai’an City was 0.95 tC·ha⁻¹ and the total carbon storage was 735,039.46 t. The model estimation results covered the entire city. The minimum carbon storage density was always 0.21 tC·ha⁻¹, while the maximum was always 1.54 tC·ha⁻¹. The resolution effect of the obtained carbon storage map is shown in Figure 5d; the resolution is approximately 350 m × 350 m. Except for a few locations, the carbon storage density was uniform in the Mount Tai Nature Reserve (1.54 tC·ha⁻¹).

Second, we used the IPCC method to calculate the carbon storage in Tai’an City in 2019. We obtained a total of 6,897,473.91 t of carbon in Tai’an’s forest carbon storage; the average carbon storage density was 44.89 tC·ha⁻¹; and the standard deviation was 31.60 tC·ha⁻¹. The calculation results covered all forestlands in Tai’an. The resolution effect of the carbon storage map is shown in Figure 6; the resolution is approximately 195 m × 195 m. In the Mount Tai Nature Reserve, the carbon storage density can be roughly divided into four levels.

Third, we employed the i-Tree Eco model to estimate the carbon storage of Tai’an City in 2019 based on the Tai’an Forest Resources Inventory database. The total carbon storage in Tai’an was determined to be 5,828,165.90 t; the average carbon storage per hectare was 37.19 tC·ha⁻¹; and the standard deviation was 45.64 tC·ha⁻¹. The resolution effect of the carbon storage map is shown in Figure 7; the resolution is approximately 195 m × 195 m. In the Mount Tai Nature Reserve, the carbon storage density can be roughly divided into five levels. The carbon density at the edge is lower than that in the core area. Patches with high carbon density are scattered in the core area of the reserve and are concentrated in the northwest–southeast direction. As seen in Figure 8, the district with the highest total carbon storage was Daiyue District (according to the results of both methods), while the district with the lowest total carbon storage was Taishan District (according to the IPCC method) or Ningyang County (according to the i-Tree Eco method). The i-Tree Eco model has a higher resolution with respect to measuring carbon storage than the InVEST model (the resolution is improved from 350 m × 350 m to 195 m × 195 m) and can visually display gradual changes in relation to forest carbon storage density. The InVEST model calculates forest carbon storage based on the average carbon storage density of forest land in Shandong Province. However, in the National Forest City Tai’an, Mount Tai and four other nature reserves have long-standing forest protection policies. As a result, the trees in these areas have remained unharvested for years, allowing them to sequester a large amount of carbon in their biomass. Consequently, the forest carbon storage density in Tai’an is higher than the provincial average in Shandong Province; the latter is represented using the InVEST model.

4.2. Carbon Sequestration

First, we used the InVEST model to calculate carbon sequestration in Tai’an City based on Tai’an land use maps for 2009, 2017, and 2019. The average carbon sequestration from 2009 to 2017 was −0.001 tC·ha⁻¹ and the carbon sequestration totaled −776.18 t. This means that the total carbon storage decreased. Tai’an’s average carbon sequestration from 2017 to 2019 was 0.09 tC·ha⁻¹ and the carbon sequestration totaled 71,408.27 t (substantial carbon sequestration may be caused by improvements in land survey technology). The model calculation results covered the entire city.

The resolution effect of the carbon sequestration map from 2017 to 2019 is shown in Figure 9c; the resolution is approximately 350 m × 350 m. Except for a few locations (red and green spots on the map), carbon sequestration was uniform (yellow in the map) in the Mount Tai Nature Reserve.

Second, we applied the i-Tree Eco model to estimate annual forest carbon sequestration in 2019 based on the Tai’an Forest Resources Inventory database.

The estimated total carbon sequestration in Tai’an (2019) was 936,789.03 tC·yr⁻¹; the annual carbon sequestration in Tai’an was, on average, 5.97 tC·ha⁻¹·yr⁻¹; and the standard deviation was 4.11 tC·ha⁻¹·yr⁻¹. The estimation results of the i-Tree Eco model covered all forested land. The forest carbon sequestrations in Tai ‘an’s different districts and counties are shown in

Table 7. Forest carbon sequestration in Tai ‘an’s districts and counties in 2019 using the i-Tree Eco method.

District and County	i-Tree Eco Method: Forest Carbon Sequestration (tC·yr−1)
Taishan District	68,556.02
Daiyue District	290,100.30
Ningyang County	91,263.49
Dongping County	124,533.00
Xintai City	214,052.74
Feicheng City	148,283.49
Sum	936,789.03

. The district with the highest total carbon sequestration was Daiyue District, at 290,100.30 tC·yr⁻¹. The district with the lowest total carbon sequestration was Taishan District, at 68,556.02 tC·yr⁻¹. The resolution effect of the carbon sequestration map is shown in Figure 10; the resolution is approximately 195 m × 195 m. In the Mount Tai Nature Reserve, carbon sequestration per hectare can be roughly divided into six levels. Carbon sequestration at the edge of the reserve is lower than that in the core. Patches with a high carbon sequestration density were scattered throughout the core area and were particularly concentrated in the northwest–southeast direction. The i-Tree Eco carbon sequestration measurement had a higher resolution than the InVEST model. The i-Tree Eco model can visually display gradual changes in the forest carbon sequestration density and is a more refined carbon sink estimation model. The two approaches focus on different components of carbon sinks. The InVEST model estimates the carbon sequestration due to land use change, while the i-Tree Eco model calculates the carbon sequestered by tree growth. That is why the carbon sequestration values obtained by the two models are different. Because the two models calculate different components of carbon sinks, their calculation results can complement each other. The results of the two models can be cross-referenced and provide a multifaceted reference for the measurement of local carbon sinks.

4.3. Spatial Resolution Advantages of the i-Tree Eco Model

We compared the spatial resolution of the carbon sink maps obtained using i-Tree Eco and the commonly used InVEST model. A sampling band was constructed to observe the differences between the two maps. As shown in Figure 11, the sampling band ran from the northeast to the southwest of Tai’an City, passing through the Daiyue District, Taishan District, Feicheng City, and Ningyang County, especially through the Mount Tai Nature Reserve. The sampling band traverses the area where the results of our new methodology are distributed. The sampling band contained one sampling point every 50 m in order to comprehensively test the resolution of the Tai’an carbon sink map.

i-Tree Eco demonstrated a gradient difference in forest carbon storage. Based on the carbon storage distribution maps calculated using the InVEST, IPCC, and i-Tree Eco methods, the carbon storage density of the sampling band was extracted. InVEST calculation of carbon storage density sampling band values is shown in Figure 12. IPCC method calculation of carbon storage density sampling band values is shown in Figure 13. i-Tree Eco calculation of carbon storage density sampling band values is shown in Figure 14. The comparison showed that the carbon storage density obtained with the InVEST model can only be divided into four main levels; this is because the InVEST model is based on the average carbon storage values of six land use types and can only present six values at most. The carbon storage density values acquired using the IPCC method and i-Tree Eco revealed differences between peak-to-valley values and better demonstrated the gradient change in the carbon storage density.

Based on the comparison of i-Tree Eco and InVEST, the i-Tree Eco model demonstrated a higher temporal and spatial resolution for estimating carbon sequestration and was more appropriate for small-scale carbon sink research. Values of the sampling band calculated using InVEST for the carbon sequestration density are shown in Figure 15. Values of the sampling band calculated using i-Tree Eco for the carbon sequestration density are shown in Figure 16. Based on the carbon sequestration per hectare calculated using InVEST and i-Tree Eco, the carbon sequestration density obtained using InVEST fluctuated within ±2 tC·ha⁻¹, and a gradient change could not be observed. The distribution of the peak-to-valley values of carbon sequestration obtained using i-Tree Eco was pronounced and better reflected the gradient changes in carbon sequestration density across different locations.

4.4. Verification of the Accuracy of i-Tree Eco Model Results

As we innovatively combined China Forest Resources Inventory data with the i-Tree Eco model to calculate forest carbon sinks on a citywide scale, the results may have been over- or underestimated. Therefore, we applied the classic IPCC method to the carbon storage results to verify the accuracy of i-Tree Eco. Here, we used SPSS software to conduct multiple statistical analyses.

4.4.1. Correlation Analysis of Carbon Storage Obtained Using the Two Methods

We first conducted a correlation analysis between the carbon storage calculated using the i-Tree Eco model and that calculated using the IPCC method. The results were utilized to preliminarily verify whether the carbon storage of the two methods was similar and served as the basis for further statistical analysis. Because the carbon storage estimated using the IPCC method and i-Tree Eco did not follow a normal distribution, the Pearson correlation coefficient was not appropriate; therefore, we used the Spearman correlation coefficient for the analysis. The analysis results showed that the Spearman correlation coefficient reached 0.968 and the correlation was significant at the 0.01 level (two-tailed). Hence, the correlation between the carbon storage based on the two methods was determined to be significant. This serves as the first step in validating the similarity between the i-Tree Eco model’s carbon storage and that calculated using the IPCC method.

4.4.2. Attempt to Use Carbon Storage Based on the IPCC Method to Correct i-Tree Eco

We constructed a linear regression model to approximately convert the carbon storage obtained using i-Tree Eco into the IPCC-based carbon storage in order to verify the newly applied carbon sink model using the classic method. We used SPSS software to establish a unary linear regression model based on the carbon storage data obtained using the two methods, as well as performing a validity test. It should be noted that we first excluded data from either method where the carbon storage value was 0 (a result of missing data for relevant indicators). In addition, 20 extreme values from the i-Tree Eco carbon storage dataset were removed. After that, we built the regression model to prevent the effect of outliers on the model’s accuracy. We used an over-origin regression (no-intercept model), as carbon storage from the i-Tree Eco model should also be 0 when the IPCC-estimated carbon storage is 0. The validity of the model was tested using one-way ANOVA (F-test).

Table 8. Regression model coefficients ^a,b.

Model	Non-Standardized Coefficient		Standardized Coefficient	t	Significance
	B	Standard Error	Beta
i-Tree Eco method Carbon Storage	1.002	0.003	0.905	400.254	0.000

^a Dependent variable: carbon storage using the IPCC method. Independent variable: carbon storage using the i-Tree Eco method. ^b This model is a univariate linear regression through the origin.

presents the regression model coefficients.

We obtained the following regression equation:

y = 1.002x

(12)

where y is the carbon storage obtained using the IPCC method; x is the carbon storage obtained using the i-Tree Eco method.

As shown in

Table 9. Model summary ^c,d.

R	R2 b	Adjusted R2	Error of Standard Estimation
0.905 a	0.820	0.820	147.099

^a Prognosis variate: carbon storage using the i-Tree Eco method. ^b For regression through the origin (a model without an intercept), R² measures the proportion of variability in the dependent variable—relative to the origin—that is explained by the regression. This R² value is not comparable to the R² value from a model that includes an intercept. ^c Dependent variable: carbon storage using the IPCC method. ^d This model is a univariate linear regression through the origin.

, according to the validity test, the correlation coefficient of the regression model was 0.905, the coefficient of determination (R²) reached 0.820, and the adjusted R² was also 0.820. All three coefficients are relatively high, indicating the good fit of the regression model. In the one-way ANOVA (F-test), at the 0.05 confidence level, the p-value was <0.05, indicating that carbon storage estimated using i-Tree Eco has a statistically significant impact on carbon storage based on the IPCC method. Moreover, the test for the regression coefficient (B) of the independent variable is shown in Table 8, with a p-value < 0.05 at the 0.05 confidence level. This indicates that the regression coefficient (B) of the independent variable was also statistically significant. We also generated a standardized residual histogram and a P–P plot to verify that the residual roughly followed a normal distribution, thereby satisfying the assumptions for establishing a linear regression model. Our regression model passed statistical validation, indicating that our forest carbon storage estimation method, based on the China Forest Resources Inventory database and the i-Tree Eco model, is effective. Additionally, since B = 1.002, which is very close to 1, if the carbon storage obtained using the i-Tree Eco method is converted to the carbon storage obtained using the IPCC method using this linear regression equation (Equation (12)), the pre- and post-conversion values would be nearly identical. Therefore, there is essentially no need for conversion. This further suggests that the carbon storage estimated using the i-Tree Eco method is approximately equivalent to that estimated using the IPCC method (i.e., the conventional approach).

5. Discussion

First, compared with other studies, our estimated carbon sequestration rate in the Tai’an Forest was within a reasonable range. By using the i-Tree Eco model, we determined the total forest carbon sequestration in Tai’an City in 2019 and the average forest carbon sequestration per hectare to be 936,789.03 tC·yr⁻¹ and 5.97 tC·ha⁻¹·yr⁻¹, respectively. The carbon sequestration rate of forest ecosystems in Guizhou Province in 2019 was 3.922 tC·ha⁻¹·yr⁻¹ [49], while the carbon sequestration rate of forest stands in the Xiaoxing’an Mountains ranged from 0.4 to 2.8 tC·ha^–1·yr⁻¹ [50]. According to 22 previous forest carbon sink studies, the carbon sequestration rate of newly planted forests ranges from 0.04 to 7.52 tC·ha⁻¹·yr⁻¹ [51]. Hou, L. et al. (2009) and Zhao, Y. (2023) studied the annual carbon sequestration of Pinus tabuliformis in a forest area of the Qinling Mountains [52,53]. Their measurements of organic carbon sequestration in the arbor layer were 7.81 tC·ha⁻¹·yr⁻¹ and 10.45 tC·ha⁻¹·yr⁻¹, respectively [52,53]. Zhao, Y. (2023) also determined that the annual carbon sequestration of the arbor layer of the Quercus aliena var. acutiserrata forest in this area was 8.57 tC·ha⁻¹·yr⁻¹ [53]. Ding, J. et al. (2017) calculated the annual carbon sequestration rate of each tree species in common economic forests (apple, peach, apricot, cherry, pear, jujube, and walnut) in Beijing, which ranged from 6.73 to 13.80 tC·ha⁻¹·yr⁻¹ [54]. Sun, M. et al. (2020) estimated the carbon sequestration rate in the arbor layer of Robinia pseudoacacia and Quercus wutaishanica forests in the loess hilly area of Shaanxi Province to be between 4.83 and 5.85 tC·ha⁻¹·yr⁻¹, respectively [55]. The forest carbon sequestration rate in Tai’an (our research) was higher than that in Guizhou Province, the Xiaoxing’an Mountains (in Heilongjiang Province), and the loess hilly area (in Shaanxi Province) and lower than that in the Qinling Mountains and the common economic forests of Beijing. It also falls within the empirical range summarized by Liu, B.J. et al. [51]. Our estimation of Tai’an forest’s carbon sequestration is within a reasonable range and is not significantly different from results reported in other provinces in China. This indicates that i-Tree Eco is reasonable and effective for estimating forest carbon sequestration.

Second, not all indicators required by i-Tree Eco exactly matched the tree attributes in the Tai’an Forest Resources Inventory database. Therefore, the relevant attributes in the database were converted into the indicators required for i-Tree Eco. These indicator conversions cannot guarantee absolute accuracy; thus, they affect the estimation accuracy of carbon storage and carbon sequestration. Furthermore, there is a possibility of over- or underestimation to a certain extent. Possible errors introduced by indicator conversions include the following:

• In Table 3, “dieback” may not be completely consistent with “average tree damage severity”. This discrepancy may affect the accuracy of dieback parameter settings.
• No exact figure exists for the crown ratio of coniferous and broad-leaved forests in Tai’an; therefore, we can only refer to corresponding data in North China and Northeast China. In Equations (6) and (7), the crown ratio influences the setting of both “Height to Live Top” and “Height to Crown Base”.
• In Appendix A.2, regarding “Height to Live Top”, we assumed that the ratio of the thickness of the dead-branch layer at the crown top to the crown length ≈ the ratio of the number of dead branches in the crown to the number of all branches (i.e., dieback). However, this assumption may not be accurate in practice. Although “centripetal dieback” is common, there are also trees where dieback mainly occurs in the inner crown area, and the dead-branch layer at the crown top or bottom may not be prominent. This introduces some error, which is also present in the derivation of “Height to Crown Base” (Equation (7)).
• In Appendix A.4, to calculate “Crown Width”, it is necessary to make four assumptions, However, in practice, these ideal assumptions may not hold. For Assumption ➀ (please refer to Appendix A.4), it is unlikely that crown overlap is entirely absent. Some canopy areas are shaded twice by the crowns of two neighboring trees. Such overlapping areas should be included in the crown coverage area of both trees, rather than assigned to only one, leading to an underestimation of total crown coverage. Since Assumption ➁ (please refer to Appendix A.4) cannot be fully realized, the projected crown coverage of each tree is unlikely to be completely closed. Therefore, the estimated crown size and crown width are also likely underestimated.
• Regarding the constant “39.6” in Equation (9), due to the lack of data on the “percent crown missing” indicator in China, we relied on the reference value obtained from Massachusetts, USA—39.6. This substitution may introduce a certain degree of error.
• The correspondence table of crown light exposure converted from canopy density (Table 4) cannot be completely accurate due to the influences of multiple factors. In

  Table A2. Correspondence between canopy density and relative light intensity (%). Data for this table were obtained from [45].

  Canopy Density [45]	Relative Light Intensity
  	Sampled in April (%) [45]	Sampled in May (%) [45]	Sampled in July (%) [45]	Sampled in August (%) [45]	Sampled in October (%) [45]	April, May, July, August, and October Averages (%)	Average Relative Light Intensity in April, May, July, August, and October ≈ Fraction
  0.9	20	22	21	23	19, 19	21.25	1/5
  0.8	33	34	38	38	15, 15	31.25	1/5
  0.7	40	40	52	53	20, 35	43.13	2/5
  0.6	57	47	61	62	27, 35	50.25	2/5
  0.5	60	59	87	87	49, 42	69.63	3/5

  Note: the bold text indicates emphasis.

  , the correspondence between canopy density and relative light intensity (%) is also based on limited references. Although we considered two primary forest types—coniferous (represented by Pinus massoniana) and broadleaf (represented by Cinnamomum camphora)—to reduce error, the gradient mapping between canopy density and relative light intensity may still lack precision. However, Table 4 and Table A2 both reflect the general pattern, showing positive correlations between canopy density, relative light intensity, and crown light exposure.

Future research is needed to address the above-mentioned limitations and improve the new “Forest Resources Inventory Database–i-Tree Eco Carbon Sink Estimation Method”.

6. Application Outlook

By using the i-Tree Eco model with a high spatial resolution to calculate carbon storage and carbon sequestration across Tai’an City, we are able to provide more detailed data on the distribution of carbon sinks and reveal variations in carbon storage and sequestration density within woodlands. Such high-precision carbon sink data can offer robust support for urban planning and policymaking—especially for initiatives related to green space and ecological planning—ensuring that policies and plans are more closely aligned with actual local conditions, thereby improving their effectiveness in implementation.

For instance, in urban green space planning, high-resolution carbon sink data can support the optimization of green space layout and landscape structure, the selection of tree species for planting, the development of management strategies, and the calculation of the ecological benefits of forests, as well as their conversion into economic value.

In ecological and environmental protection planning, high-precision carbon sink data can be used to accurately identify ecologically degraded areas, guide the scientific selection of restoration species, and serve as a foundational tool for building a carbon sink unit network. It can help in unlocking the carbon sink potential in a hierarchical manner, dynamically adjusting forestry management strategies and translating the ecological value of carbon sinks into economic benefits. Additionally, such data can be used for the breakdown and evaluation of progress toward carbon peaking and carbon neutrality goals, the assessment of ecological equity, and as quantitative evidence for compliance with international carbon reduction agreements, such as the Paris Agreement.

For the delineation of the ecological protection red line—a core element of China’s National Land Spatial Planning—high-resolution carbon sink data can assist in accurately identifying high-value carbon sink areas, assessing carbon stock baselines, predicting carbon sink potential, and optimizing the pattern of ecological space. It can further support the construction of carbon sink corridors, the control of development intensity gradients, the prioritization of degraded ecosystem restoration, the dynamic monitoring of carbon sink losses, and the quantitative assessment of ecological restoration effectiveness. Such data also provide scientific support for ecological compensation mechanisms and enhance coordination between carbon peaking, neutrality targets, and related policies.

7. Conclusions

We introduced a new carbon sink estimation method that combined the i-Tree Eco model and the China Forest Resources Inventory. Our approach enables the calculation of both forest carbon sequestration and storage in Tai’an. Thus, it can address the limitations of the commonly used InVEST model, which overlooks annual forest carbon sequestration. It also enhances the spatial resolution of small-scale carbon sink estimations (improving it from 350 m × 350 m to 195 m × 195 m).

Statistical tests showed that the carbon storage calculation results were valid. Similarly, the forest carbon sequestration data produced using this method provided a foundation for the spatial analysis of urban carbon sinks. Compared with the commonly used carbon sink estimation methods (e.g., the IPCC method and InVEST model), our method has the advantage of capturing spatial gradient variations in both carbon storage and sequestration. The i-Tree Eco model estimates forest carbon sequestration with relatively higher temporal and spatial resolutions, which is especially valuable for small-scale (e.g., city-level) carbon sink research. Therefore, this carbon sink estimation method can be used to complement and improve the commonly used approaches. It helps fill the gap in citywide applications of small-scale forest carbon sink estimation methods that require high spatial resolution and precision.