Spatiotemporal Dynamics of Forest Carbon Sinks in China’s Qinba Mountains: Insights from Sun-Induced Chlorophyll Fluorescence Remote Sensing

Abstract

Forest carbon sinks are crucial in mitigating climate change as integral components of the global carbon cycle. Accurately estimating forest carbon sinks using traditional remote sensing indices, such as Normalized Difference Vegetation Index(NDVI), presents significant challenges, particularly in complex terrains and regions with variable climates. These limitations hinder the effective capture of photosynthetic dynamics. To address this gap, this study leverages Sun-Induced Chlorophyll Fluorescence (SIF) remote sensing, highlighting its superiority over traditional indices in capturing photosynthetic processes and offering a more precise approach to estimating carbon sinks in climate-sensitive mountainous areas. Using SIF data from GOSIF, alongside models for light-use efficiency and ecosystem respiration, this study estimates forest carbon sinks in the Qinba Mountains of China during the growing season (June to September) from 2011 to 2018. The results are further validated and analyzed in terms of forest age and type. Key findings include: (1) The average annual forest carbon sinks during the growing season was approximately 24.51 TgC; (2) Spatially, higher carbon sinks values (average 36.79 gC·m⁻²·month⁻¹) were concentrated in the western and central Qinba areas, while southeastern and central-northern regions exhibited lower values (average 7.75 gC·m⁻²·month⁻¹); (3) Temporally, minimal interannual variation was observed in the northwest, whereas the southeast showed fluctuating trends, with an initial decline followed by an increase; (4) Forest carbon sinks was significantly influenced by forest age, type, and altitude. Our findings demonstrate that plantation forests aged 10 to 30 years exhibit superior carbon sequestration capacity compared to natural forests, while natural forests aged 70 to 90 years also show significant carbon sinks potential. These results underscore the crucial influence of forest characteristics on carbon sequestration dynamics. By examining these spatiotemporal patterns in the Qinba Mountains, our study offers valuable insights for advancing China’s ‘dual carbon’ goals, emphasizing the importance of strategic forest management in mitigating climate change.

1. Introduction

Human activities have contributed to a 1.1 °C global temperature rise, exacerbating climate change [1] and prompting international initiatives such as the Paris Agreement, which aims to limit global warming to 1.5–2 °C [2]. China’s “dual carbon” goals, targeting carbon peaking by 2030 and carbon neutrality by 2060, underscore the vital importance of carbon sinks. These processes, which remove CO₂ from the atmosphere, are integral to the terrestrial carbon cycle [3] and play a pivotal role in advancing global sustainability and mitigating the impacts of climate change.

Forest carbon sinks are a critical component of terrestrial carbon sinks, significantly contributing to global carbon sequestration and playing an irreplaceable role in mitigating climate change. Among the various strategies to enhance carbon sinks, increasing forest carbon sequestration stands out as the most economical, sustainable, and ecologically friendly approach [4]. Consequently, conducting a scientific assessment of the carbon cycle dynamics within China’s forest ecosystems is essential. Such assessments are pivotal not only for understanding the impact of global environmental changes on forest carbon sequestration functions but also for informing the development of national “dual carbon” policies. Accurate estimation of current forest carbon sinks capacities and reliable predictions of future trends are particularly crucial for achieving these goals.

Several methods are currently employed for monitoring forest carbon sinks, including dendromorphological techniques, eddy covariance systems, and remote sensing technologies [5]. Each approach has unique strengths and limitations. Dendromorphological Methods: This method involves using plot inventories to establish relationships between the size and shape of individual trees or groups of trees. It is relatively simple to implement and allows for scalable results. However, it is highly time-consuming, labor-intensive, and often causes significant disturbances to forest ecosystems [5]. For example, Chen et al. [6] used this method to study Betula alnoides in Guangdong, revealing increased carbon storage when intercropped with nitrogen-fixing species. Similarly, Brown and Lugo [7] estimated global aboveground forest biomass at 205 × 10⁹ tons using this approach. Eddy Covariance Methods: This technique measures CO₂ flux by observing atmospheric fluxes, offering high accuracy and enabling continuous spatiotemporal monitoring. However, it requires expensive equipment and suffers from low data reliability under certain environmental conditions [5]. For instance, You et al. [8] employed eddy covariance to calculate the Net Ecosystem Productivity (NEP) of Inner Mongolia grasslands, estimating an average of 2.43 ± 6.71 TgC annually from 1982 to 2018. Similarly, Hussian et al. [9] used this method to determine that after the 2021 Lymantria dispar outbreak in Ontario, Canada, carbon losses in deciduous and mixed forests of the Great Lakes region reached 21.1 MtC and 21.4 MtC, respectively, compared to 2020. Remote Sensing Technologies: Remote sensing employs satellite data to obtain large-scale observations with high precision. It enables biomass and carbon reserve assessments at regional and global scales [5]. For example, Kong et al. [10] used biomass maps derived from remote sensing data to calculate the forest carbon sinks in China’s Dongting Lake basin, observing an increase from 64.64 TgC in 2000 to 382.56 TgC in 2020. Similarly, Pandey et al. [11] integrated Normalized Difference Vegetation Index(NDVI) and Leaf Area Index(LAI) remote sensing data with ground inventory records to estimate biomass distribution in Tripura, India, at 32–94 Mg ha⁻¹.

Current remote sensing satellites estimate forest biophysical and chemical parameters, such as canopy closure, tree species, structural attributes, and associated indices [12]. While these approaches generally satisfy the requirements for detecting forest carbon sinks, they are hindered by challenges including limited model generalizability, sensitivity to weather conditions, and restricted applicability [13]. Additionally, these methods are insensitive to small or negligible changes in reflectance linked to photosynthetic activity, making them inadequate as indicators of vegetation’s real-time photosynthetic status. They also exhibit limited sensitivity in capturing the complexity of vegetation’s photosynthetic processes [14]. Therefore, it is imperative to develop advanced remote sensing methods that are both robust and efficient, enabling easier observation while accurately reflecting the intensity of vegetation photosynthesis to enhance the precision of forest carbon sinks assessments.

Sun-Induced Chlorophyll Fluorescence (SIF) has emerged as a promising method for forest carbon sinks estimation. SIF is a phenomenon in which plants emit fluorescence when exposed to light, with emissions primarily concentrated in the red light range (approximately 690–750 nm) and the near-infrared range (approximately 725–800 nm). As a byproduct of photosynthesis, SIF provides a more direct representation of vegetation photosynthetic dynamics compared to traditional vegetation indices. This makes it highly effective for applications such as estimating Gross Primary Productivity (GPP) [15] and investigating vegetation phenology [16], crop health and diseases [17], and other related fields.

SIF is intrinsically linked to vegetation photosynthesis, a critical component of the ecological carbon sequestration process in forest ecosystems. This connection makes SIF data a valuable tool for estimating forest carbon sinks. However, due to the influence of physical models and environmental factors, SIF data can exhibit instability, and directly using SIF for forest carbon sinks estimation may lead to significant errors. Previous research has shown a strong linear correlation between SIF and GPP [18,19,20], although this relationship may be influenced by climatic conditions and forest types [21,22]. Despite these influencing factors, the correlation provides a solid foundation for model development. Accordingly, this study adopts GPP as an intermediate variable, using SIF data to estimate GPP. As a key component in calculating forest carbon sinks, GPP, when combined with ecosystem respiration, forms the basis for estimating forest carbon sequestration. For instance, Guanter et al. [23] utilized satellite remote sensing to monitor plant SIF and directly measure GPP in agricultural and grassland ecosystems, achieving results consistent with ground-based observations. Zhao and Liang [24] applied SIF data to estimate forest carbon sinks in northeastern China and validated the method using flux tower data. Similarly, Cheng [25] reviewed the application of SIF in monitoring vegetation dynamics and carbon uptake in Arctic-Boreal regions, finding the method promising. While previous studies have explored methods for estimating carbon fluxes using SIF and highlighted the linear relationship between SIF and GPP, their findings are largely confined to semi-arid regions globally, northeastern China, and the Arctic-Boreal zones. However, it remains uncertain whether this linear relationship holds in the Qinling-Daba Mountain region and whether SIF can reliably estimate forest carbon sinks there, warranting further validation.

This study addresses unresolved challenges in applying SIF to estimate forest carbon sinks in the Qinba Mountain region, where current research faces several limitations: (1) the relationship between SIF and forest carbon sinks remains unclear, (2) the linear correlation between SIF and GPP varies under different environmental conditions and requires further validation in this region, (3) the spatiotemporal characteristics of forest carbon sinks dynamics in the Qinba Mountains are poorly understood, and (4) the factors influencing these changes are not well identified. To address these issues, this study employs SIF to estimate GPP and validates the resulting forest carbon sinks estimates using a remote sensing-based ecosystem respiration model. This approach aims to evaluate the feasibility of the SIF-GPP linear model in the Qinba region and generate remote sensing inversion results for forest carbon sinks during the growing seasons (June to September) from 2011 to 2018. Additionally, potential factors influencing carbon sinks variability will be assessed to provide scientific insights for understanding the role of Qinba Mountain forest ecosystems in China’s ‘dual carbon’ goals.

2. Study Area and Data Sources

2.1. Overview of the Study Area

The study area is situated in the forested region of the Qinling-Daba Mountain range in China (referred to as the Qinba Mountain area), encompassing six provinces: Shaanxi, Henan, Hubei, Chongqing, Sichuan, and Gansu (as shown in Figure 1). The geographic coordinates range from 102°28′E to 113°39′E and 30°43′N to 35°24′N, with an east–west extent of approximately 1000 km, a north–south width of about 300 km, and a total area of 225,000 km². According to the results of China’s ninth national forest resource inventory [26], the country’s total forest area is approximately 220 million hm², with a total vegetation carbon storage of about 918.6 million t. The forest area within the Qinba Mountain region covers 20.8856 million hm². Based on the area proportion, the total vegetation carbon storage in the Qinba Mountain area is estimated at approximately 87.2 million t. This substantial carbon storage plays a critical role in supporting China’s “dual carbon” goals and positions the region as a key player in the global carbon cycle.

The dominant vegetation types in the Qinba Mountain area include coniferous forests (902,100 hm²), broadleaf forests (2,021,100 hm²), and conifer-broadleaf mixed forests (3,015,600 hm²) [27]. The region’s terrain is characterized by a west-to-east elevation gradient, with altitudes decreasing from the Qinling-Daba Mountains toward the surrounding areas and featuring numerous hilly basins and significant topographic variation. Climatically, the eastern part of the region is humid, while the western part is arid, resulting in a pronounced east–west climatic differentiation [28].

This study focuses on the forest area surrounding the Huoditang flux station as a typical case for detailed analysis. Located on the southern slope of the Qinling Mountains in China, the Huoditang forest area spans between 33°18′–33°28′N and 108°21′–108°39′E, with an elevation range from 800 to 2500 m. The region experiences a subtropical mountain climate, with an average annual temperature between 8 °C and 10 °C. Annual precipitation ranges from 900 to 1200 mm, with the majority falling during the months of July, August, and September, accounting for approximately half of the total annual rainfall. These conditions create an ideal environment for vegetation growth during the summer [29] and enhance the area’s representativeness [30].

2.2. Data

2.2.1. SIF Data

Several satellites are currently capable of directly observing SIF, but the SIF products they generate often lack spatial continuity (e.g., GOSAT, OCO-2, TanSat) and typically have coarse spatial resolution (e.g., SCIAMACHY, GOSAT, GOME-2). For this study, which focuses on estimating carbon sinks from 2011 to 2018, we selected the Global SIF dataset (GOSIF) constructed by Li and Xiao [31]. This dataset, based on discrete OCO-2 data, MODIS data, and meteorological data, offers an 8-day time resolution and a spatial resolution of 0.05°. Compared to other SIF data with coarser resolution, GOSIF provides finer spatial resolution, global continuous coverage, and a longer time series. The stability and applicability of GOSIF data have been validated in previous studies [24,32].

2.2.2. GPP Data

The GPP data used in this study is the global GPP dataset published by Zheng et al. [33], which is derived from the improved EC-LUE model. This model integrates several key environmental variables, including atmospheric CO₂ concentration, radiation components, and atmospheric vapor pressure difference. The dataset offers high spatial and temporal resolution (0.05°, 8 days) and is widely regarded as highly reliable [34].

2.2.3. NEP Data

The NEP data used in this study is the global daily NEP simulation dataset for the period from 1981 to 2019, published by He et al. [35]. This dataset is generated by driving the mechanistic ecosystem model BEPS using remote sensing data of vegetation parameters, meteorological data, and atmospheric CO₂ concentration. It provides high spatial resolution (0.072727°) and daily temporal resolution. The dataset is widely regarded as highly reliable [36].

2.2.4. MODIS Data

In this study, the MOD09A1.061 product is used to calculate vegetation indices. This data product has a spatial resolution of 1 km and a daily temporal resolution. Reflectance band data from the corresponding grid points are extracted, and data from the blue, red, near-infrared, and shortwave infrared bands are selected. The Enhanced Vegetation Index (EVI) and the Land Surface Water Index (LSWI) are then calculated. The calculation formulas for these indices are as follows [37]:

$$EVI=2.5{\displaystyle \frac{{\rho }_{NIR}-{\rho }_{red}}{{\rho }_{NIR}+1+6{\rho }_{red}+7.5{\rho }_{blue}}},$$

(1)

$$LSWI={\displaystyle \frac{{\rho }_{NIR}-{\rho }_{SWIR}}{{\rho }_{NIR}+{\rho }_{SWIR}}},$$

(2)

where ${\rho }_{NIR}$ represents the surface reflectance in the near-infrared band, ${\rho }_{red}$ represents the surface reflectance in the red band ${\rho }_{blue}$ represents the surface reflectance in the blue band, and ${\rho }_{SWIR}$ represents the surface reflectance in the shortwave infrared band.

This study utilizes the MOD11A1.061 product to obtain land surface temperature (LST), which has a spatial resolution of 500 m and a daily temporal resolution.

2.2.5. Forest Area Data

The forest area data for the study region is derived from the 2013 land cover dataset of China, sourced from the China 30 m annual land cover dataset and its dynamic changes from 1985 to 2022 [38]. This dataset is provided by the National Glacier, Snow, and Desert Data Center (http://www.ncdc.ac.cn (accessed on 16 July 2024)) and has a spatial resolution of 30 m.

2.2.6. Forest Age Data

This study employs the forest age dataset published by Cheng et al. [39], which estimates the age of forests in China by integrating Landsat images from 1985 to 2020 with forest inventory data. Time series analysis and machine learning algorithms were used to create a 30 m resolution map of forest age distribution for the year 2020. Given the availability of forest age data, this study assumes that the forest age in 2018 is similar to that in 2020. This assumption is supported by the land use transition matrix for the Qinba Mountain area between 2018 and 2020 (

Table 1. Land use transition matrix for the Qinba Mountain area in 2018 and 2020 (Unit: hm2).

	2018	Cropland	Forest	Shrub	Grassland	Water	Snow/Ice	Barren	Impervious	Wetland	Sum
2020
Cropland		4,875,111.99	91,703.88	2297.79	142,520.04	1981.71		1.89			5,113,617.3
Forest		194,373.54	21,035,592	15,239.7	69,887.43	34.38		1.26			21,315,128.31
Shrub		615.96	6767.46	113,637.24	5083.65						126,104.31
Grassland		50,057.82		4970.97	3,382,610.85	1114.47	205.47	4344.66		0.63	3,443,304.87
Water		2489.76			667.53	183,350.16	27.36	78.39	1756.71	0.09	188,370
Snow/Ice							8542.62	520.38			9063
Barren					6166.08	338.76	2779.38	33,993.27			43,277.49
Impervious		13,215.15	721.35		984.24	545.94		10.44	341,382.51		356,859.63
Wetland					0.54					1.89	2.43
Sum		5,135,864.22	21,134,784.69	136,145.7	3,607,920.36	187,365.42	11,554.83	38,950.29	343,139.22	2.61	30,595,727.34

), which indicates that forest cover changed by only about 0.85% during this period. Due to this minimal change, the 2020 forest age distribution map is used as an approximation for the 2018 distribution.

2.2.7. Forest Type Data

This study utilizes the forest type dataset published by Xiao et al. [40], which is based on Landsat images from 1985 to 2021. Time series change detection methods were used to generate training samples, and locally adaptive random forest classifiers were trained on these samples to produce a map of global plantation and natural forests at a 30 m spatial resolution for 2021. The dataset provides high spatial resolution (30 m) and annual temporal resolution. Similarly, based on the available forest type data, this study assumes that the forest types in 2018 are consistent with those in 2021. This assumption is supported by the land use transition matrix for the Qinba Mountain area between 2018 and 2021 (

Table 2. Land use transition matrix for the Qinba Mountain area in 2018 and 2021 (Unit: hm2).

	2018	Cropland	Forest	Shrub	Grassland	Water	Snow/Ice	Barren	Impervious	Wetland	Sum
2021
Cropland		4,794,970.59	119,195.01	3036.96	161,260.29	2303.01		2.61			5,080,768.47
Forest		259,456.32	21,006,311.22	20,258.64	89,368.83	42.75		1.89			21,375,439.65
Shrub		686.43	8437.05	106,182.81	6209.01						121,515.3
Grassland		61,830.09		6667.29	3,342,471.57	1264.32	278.01	5635.44		0.81	3,418,147.53
Water		3247.2			753.21	182,773.89	36.54	89.73	2308.32	0.09	189,208.98
Snow/Ice							8323.56	655.47			8979.03
Barren					6658.38	353.79	2916.72	32,554.08			42,482.97
Impervious		15,673.59	841.41		1198.53	627.66		11.07	340,830.9		359,183.16
Wetland					0.54					1.71	2.25
Sum		5,135,864.22	21,134,784.69	136,145.7	3,607,920.36	187,365.42	11,554.83	38,950.29	343,139.22	2.61	30,595,727.34

), which shows a forest cover change of approximately 1.14% during this period. Given this minimal change, the 2021 forest type distribution map is used as an approximation for the 2018 distribution.

2.2.8. Elevation Data

The DEM data used in this study is the GEBCO 2024 Grid [41], a globally recognized topographic model dataset that is widely used for various geospatial analyses [42].

3. Methods

3.1. Data Preprocessing

The datasets utilized in this study—comprising GOSIF data (0.05°, monthly), GPP data (0.05°, 8-day), NEP data (0.072727°, daily), MODIS data (1 km/500 m, daily), forest age data (30 m), forest type data (30 m), elevation data (30 m), and land cover data (30 m)—exhibit variations in spatial and temporal resolutions. To ensure consistency, all datasets were spatially resampled to a resolution of 0.072727° and temporally aggregated to a monthly scale.

Although high spatial resolution data can more accurately capture dynamic changes in carbon sinks in the Qinba Mountain area, resampling them to lower resolutions may lead to a loss of precision due to inherent estimation biases in the original low-resolution data. In contrast, downscaling high-resolution data through multi-pixel aggregation helps preserve data accuracy effectively. Therefore, in this study, all data were resampled to a unified resolution of 0.072727° to ensure consistency and reliability.

Since the Qinba Mountain area encompasses two MODIS tiles (h26v05 and h27v05), mosaicking of the two tiles was conducted prior to transforming their spatial and temporal resolutions to 0.072727° and a monthly scale, respectively.

In this study, we selected the period from 2011 to 2018 for carbon sinks estimation, as the 9th National Forest Resources Inventory of China, completed in 2018, provides the most recent, systematic, and authoritative forest resource data available. Using this dataset ensures the comparability of our results and underscores the contribution of the Qinba Mountain region to China’s dual carbon goals. Additionally, using data from 2003 to 2010 to project carbon sinks for 2011 to 2018 inherently involves predictions, and the uncertainty in such projections increases with longer time spans. By limiting the study to 2018, we ensure that only the preceding eight years of data (2003–2010) are used to estimate the following eight years (2011–2018), effectively minimizing prediction errors and uncertainties. This approach enhances the reliability of the results, making 2018 the logical endpoint of the study.

3.2. Estimation of Gross Primary Productivity (GPP) Using Solar-Induced Chlorophyll Fluorescence (SIF)

GPP is calculated using the Light Use Efficiency (LUE) model, which expresses GPP as the product of the absorbed photosynthetically active radiation (APAR) by vegetation and the light use efficiency (ε_p). The relationship can be represented as follows:

$$\mathrm{G}\mathrm{P}\mathrm{P}=APAR{\epsilon }_P$$

(3)

SIF is influenced by both APAR and the light use efficiency (ε_p) of vegetation [43]. This relationship can be expressed as:

$$\mathrm{S}\mathrm{I}\mathrm{F}=APAR{\epsilon }_F\left(\lambda \right)f_{esc}\left(\lambda \right),$$

(4)

where ε_F(λ) represents the fraction of photosynthetically active radiation absorbed by the plant that is re-emitted as fluorescence at wavelength λ, and f_esc (λ) denotes the proportion of chlorophyll fluorescence at wavelength λ that escapes the canopy. By combining Equations (3) and (4), the following expression can be derived:

$$\mathrm{G}\mathrm{P}\mathrm{P}=\mathrm{S}\mathrm{I}\mathrm{F}{\displaystyle \frac{{\epsilon }_P}{{\epsilon }_F\left(\lambda \right)}}{\displaystyle \frac{1}{f_{esc}\left(\lambda \right)}},$$

(5)

Furthermore, given that the canopy structure of the vegetation within the satellite’s coverage area remains relatively stable over a specific time period, it can be assumed that f_esc (λ) is constant [43]. Consequently, the relationship between GPP and SIF primarily hinges on the ratio of ε_P to ε_F, expressed as:

$$\mathrm{S}={\displaystyle \frac{{\epsilon }_P}{{\epsilon }_F\left(\lambda \right)f_{esc}\left(\lambda \right)}}{\displaystyle \frac{\mathrm{G}\mathrm{P}\mathrm{P}}{\mathrm{S}\mathrm{I}\mathrm{F}}},$$

(6)

In this equation, S represents the fitting coefficient that quantifies the relationship between GPP and SIF. Once S is determined, SIF can be utilized to estimate GPP, as expressed by the following equation:

$$\mathrm{G}\mathrm{P}\mathrm{P}=\mathrm{S}\mathrm{I}\mathrm{F}\times \mathrm{S},$$

(7)

3.3. Remote Sensing-Based Ecosystem Respiration Model

Ecosystem respiration (R_eco) comprises two main components: plant respiration (R_GPP) and the respiration of organic matter and soil microorganisms within the ecosystem (R_EOM), which includes contributions from litter and soil organic matter. This relationship can be expressed as:

$${\mathrm{R}}_{\mathrm{e}\mathrm{c}\mathrm{o}}={\mathrm{R}}_{\mathrm{G}\mathrm{P}\mathrm{P}}+{\mathrm{R}}_{\mathrm{E}\mathrm{O}\mathrm{M}},$$

(8)

According to Gao et al. [44], R_GPP is influenced by plant growth conditions and water availability, which are primarily represented by EVI and LSWI. The relationship between R_GPP, the EVI influence function (EVI_s), and the water influence function on the maximum light use efficiency (W_s) can be expressed as:

$${\mathrm{R}}_{\mathrm{G}\mathrm{P}\mathrm{P}}=\alpha {\times PC}_{max}\times {EVI}_s\times W_s,$$

(9)

In the equation, α represents a model parameter to be calibrated through fitting. PC_max denotes the maximum photosynthetic capacity specific to each vegetation type, with its value varying across vegetation categories. EVI_s captures the variability in photosynthetic capacity among different vegetation types, and the product of PC_max and EVI_s reflects the effective photosynthetic capacity of each vegetation type. Finally, W_s quantifies the influence of water availability on the photosynthetic process.

Where EVI_s is a function of EVI, expressed as:

$${EVI}_s=EVI-0.1,$$

(10)

when the EVI approaches 0.1, R_GPP converges to 0. In the model proposed by Gao et al. [45], W_s represents a modified form of W_scalar, which is determined using the LSWI, and can be expressed as:

$$W_{scalar}={\displaystyle \frac{1+LSWI}{1+{LSWI}_{max}}},$$

(11)

Since the value of W_scalar depends on LSWI_max, different vegetation types may exhibit similar or identical W_scalar values despite significant variations in actual canopy water content. Thus, W_scalar effectively captures the temporal dynamics and spatial patterns of moisture conditions. By setting LSWI_max to 1, spatial comparisons are simplified. Consequently, the equation for W_scalar is adjusted to W_s, expressed as:

$$W_s={\displaystyle \frac{1+LSWI}{2}},$$

(12)

The relationship between R_EOM and LST can be described using the Lloyd and Taylor [46] equation, expressed as:

$${\mathrm{R}}_{\mathrm{E}\mathrm{O}\mathrm{M}}=R_{ref}\times exp\left[E_0\left({\displaystyle \frac{1}{T_{ref}-T_0}}-{\displaystyle \frac{1}{LST-T_0}}\right)\right],$$

(13)

In the equation, R_ref represents the respiration rate of ecosystem organic matter and soil microorganisms at the reference temperature T_ref; E₀ is a parameter associated with the activation energy; T₀ is the temperature at which R_EOM becomes 0; and LST refers to the land surface temperature. Typically, the reference temperature T_ref is set to 10 °C, and T₀ is fixed at −46.02 °C. Accordingly, the remote sensing-based ecosystem respiration model can be expressed as:

$${\mathrm{R}}_{\mathrm{e}\mathrm{c}\mathrm{o}}=\alpha {PC}_{max}{EVI}_s{W}_s+R_{ref}\times exp\left[E_0\left({\displaystyle \frac{1}{T_{ref}-T_0}}-{\displaystyle \frac{1}{LST-T_0}}\right)\right],$$

(14)

3.4. Calculation of Ecological Carbon Sequestration

Net Primary Productivity (NPP) represents the net amount of carbon fixed by vegetation through photosynthesis after deducting the carbon lost to autotrophic respiration. NEP quantifies the remaining carbon balance within the ecosystem after further subtracting heterotrophic respiration, which accounts for the decomposition of organic matter by soil organisms. The relationship can be expressed as:

$$\mathrm{N}\mathrm{P}\mathrm{P}=\mathrm{G}\mathrm{P}\mathrm{P}-{\mathrm{R}}_{\mathrm{G}\mathrm{P}\mathrm{P}},$$

(15)

$$\mathrm{N}\mathrm{E}\mathrm{P}=\mathrm{N}\mathrm{P}\mathrm{P}-{\mathrm{R}}_{\mathrm{E}\mathrm{O}\mathrm{M}}=\mathrm{G}\mathrm{P}\mathrm{P}-{\mathrm{R}}_{\mathrm{G}\mathrm{P}\mathrm{P}}-{\mathrm{R}}_{\mathrm{E}\mathrm{O}\mathrm{M}}=\mathrm{G}\mathrm{P}\mathrm{P}-{\mathrm{R}}_{\mathrm{e}\mathrm{c}\mathrm{o}},$$

(16)

when NEP > 0, the ecosystem functions as a carbon sink, indicating that it absorbs more carbon than it releases. Conversely, when NEP < 0, the ecosystem acts as a carbon source, emitting more carbon than it sequesters.

For details of the workflow, please refer to Figure A1.

3.5. Comparison of Carbon Sequestration Functions Between Plantations and Natural Forests Across Different Forest Age Groups

To compare the carbon source or sink capacity of plantations and natural forests across different forest age groups, this study incorporates the typical proportion P, defined as the ratio of plantation grid cells to natural forest grid cells within the study area for each age group. The analysis focuses on evaluating the relative performance of plantations and natural forests as carbon sources or sinks under equivalent functional roles (i.e., both acting as carbon sources or both as carbon sinks). For each forest age group, the number of plantation grid cells is divided by the number of natural forest grid cells to calculate a ratio. If this ratio exceeds P, it indicates that plantations exhibit a stronger carbon source or sink capacity at that age group. Conversely, if the ratio falls below P, it suggests that natural forests have a greater capacity as a carbon source or sink for that specific age segment. This approach enables a quantitative assessment of the relative carbon dynamics between the two forest types across varying forest age stages.

4. Results

4.1. GPP Estimation Using the Light Use Efficiency Model

4.1.1. Analysis of the GPP-SIF Correlation and Stability of Fitting Coefficients

Based on the scatter plot of GPP data from the Huoditang forest area (2003–2010) and the corresponding GOSIF data for the respective latitude and longitude (as shown in Figure 2), a strong linear correlation is observed between GPP and SIF, with a correlation coefficient (R²) of 0.63.

The fitting coefficient S was calculated using the GPP data from the Huoditang forest area (2003–2010) and the corresponding GOSIF data. The coefficient S was then plotted as a time series for each year (as shown in Figure 3). The analysis revealed that the fitting coefficient S increased from January to March, reaching its peak value around March. It then declined, remained relatively stable from June to September, increased steadily from September to November, and decreased again from November to December.

Given the significant monthly variations in fitting coefficients across different years, the monthly average fitting coefficients from 2003 to 2010, along with the fitting coefficient values for the same months in different years, were plotted together (as shown in Figure 4). The analysis revealed that, from June to September between 2003 and 2010, the interannual variation in the fitting coefficient S between GPP and SIF was relatively small. This suggests that the photosynthetic efficiency of the forest remains more stable during the growing season (June to September). Consequently, it is recommended to focus research efforts specifically during the vegetation growing season (June to September).

4.1.2. Acquisition of Fitting Coefficients and Estimation of GPP

Table 3. Fitting coefficient analysis of GPP and SIF during the growing season in Huoditang.

Time	2003	2004	2005	2006	2007	2008	2009	2010	Average
Jun.	174.13	211.01	166.01	137.14	201.24	149.57	133.88	211.58	173.07 ± 26.71
Jul.	249.02	187.00	410.01	171.48	284.82	185.82	185.82	286.33	254.55 ± 68.53
Aug.	366.90	354.62	424.29	344.05	288.43	270.15	270.15	326.92	337.60 ± 45.06
Sept.	471.14	527.97	530.80	478.05	471.06	430.94	430.94	416.74	473.43 ± 36.20

Note: The horizontal title reflects the situation of the fitting coefficient S for each year from 2003 to 2010 as well as the multi-year average, while the vertical title reflects the situation of the fitting coefficient S for different months.

presents the calculated fitting coefficients specifically for the Huoditang forest area, which serves as a representative example of the broader forest area, from June to September for the years 2003 to 2010. The data shows an increasing trend in the fitting coefficients at this site during this period.

Similarly, after obtaining the monthly average fitting coefficients for GPP and SIF for each grid in the Qinba Mountain forest area from June to September during 2003–2010, the corresponding monthly average fitting coefficients were applied to the SIF data for each grid from June to September 2011–2018 to estimate GPP. The GPP data product from the GPP dataset for June to September 2011–2018 was then used to validate the GPP estimates derived from SIF, resulting in the scatter plot shown in Figure 5. The estimated GPP demonstrated relatively high accuracy, with an R² value of 0.771.

Since the fitting coefficient in the SIF-GPP relationship is the sole parameter in Equation (7) and is directly multiplied with the SIF data, a 1% perturbation of the fitting coefficient results in only a 1% change in the fitted GPP values.

4.2. Estimation of Ecosystem Respiration Using the Remote Sensing Ecosystem Respiration Model

The remote sensing ecosystem respiration model involves four parameters that need to be fitted: α, PC_max, R_ref and E₀. In the Huoditang forest area, MODIS and NEP data from June to September of 2003–2004 were used to fit these four parameters in the ecosystem respiration model. The NEP and MODIS data from June to September of 2005 were then used for accuracy validation.

Table 4. Fitted parameters of the ecosystem respiration model for the Huoditang forest area.

α	PCmax/(mgCO2 m−2·s−1)	Rref/(gC·m−2)	E0/K
−60.67	4.01	160.94	207.81

presents the fitted respiration model parameters for the Huoditang forest area.

Similarly, for each grid within the Qinba Mountain forest area, the remote sensing ecosystem respiration model parameters were independently fitted using the NEP and MODIS data from June to September of 2003–2004. This resulted in the remote sensing ecosystem respiration model for each grid in the Qinba Mountain forest area. The NEP and MODIS data from June to September of 2005 were then used for accuracy validation. By subtracting the NEP data from the GPP dataset for June to September 2005, the ecosystem respiration values estimated from the literature were obtained. A scatter plot comparing these literature-estimated ecosystem respiration values with the estimated values was created (as shown in Figure 6), showing a good fit with an R² of 0.7682.

Additionally, a 1% perturbation was applied to the fitting parameters of the ecosystem respiration model for the Huoditang forest area (

Table 5. Sensitivity analysis of fitted parameters in the ecosystem respiration model for the Huoditang forest area.

α	PCmax	Rref	E0
7.497065851	113.3945853	8.74436803	0.73069437

). The results showed that the parameter PC_max exhibited high sensitivity, while α and R_ref displayed moderately high sensitivity, and E₀ demonstrated low sensitivity. This discrepancy arises because PC_max, the most sensitive parameter, is incorporated into Equation (14) in a multiplicative form, leading to a compounding effect. Small perturbations in such parameters can accumulate, resulting in significant changes in the model outputs. In contrast, the relatively lower sensitivity of the other parameters suggests that the overall model sensitivity remains within an acceptable range, ensuring a certain level of reliability.

4.3. Remote Sensing-Derived Forest Carbon Sinks Estimates in the Qinba Mountain Region

4.3.1. Spatiotemporal Variation Characteristics of Remote Sensing Inversion Results of Forest Carbon Sinks in the Qinba Mountain Area

Using remote sensing inversion, the forest carbon sinks estimates for the Qinba Mountain region from June to September during the years 2011–2018 were derived, as illustrated in Figure 7. Over the study period, the forests in this region were found to sequester approximately 24.51 TgC annually during the growing season.

From a regional perspective, the high-value areas of forest carbon sinks in the Qinba Mountain area from 2011 to 2018 were primarily concentrated in the western and some central forest regions, with an average value of approximately 36.79 g C m⁻² month⁻¹. In contrast, the low-value areas were mainly located in the southeastern and northern parts of the region, with an average value of about 7.75 g C m⁻² month⁻¹.

As observed from the remote sensing inversion results of forest carbon sinks in the Qinba Mountain area from June to September for each year between 2011 and 2018 (as shown in Figure 8), the western part of the region consistently exhibited high carbon sinks values. In contrast, the southeastern part generally showed low carbon sinks values, except for the years 2012 and 2015. The northern region has consistently maintained low carbon sinks values, while the eastern part displayed notable interannual variability.

4.3.2. Relationship Between Forest Carbon Sinks Remote Sensing Inversion Results and Forest Age and Type in the Qinba Mountain Area

The distribution of forest ages in the Qinba Mountain area (as shown in Figure 9) reveals the following patterns: In the western region, forests aged between 100 and 130 years are concentrated; in the central region, forests predominantly range from 60 to 90 years old; while in the northeastern, southeastern, and northern regions, forests aged between 10 and 50 years are most prevalent.

The forest ages in the Qinba Mountain area are categorized into ten-year intervals, resulting in a total of 14 groups (as shown in

Table 6. Classification of forest ages in the Qinba Mountain area and grid count distribution.

Forest Age Grouping	Number of Grids	Percentage of Total Grids
1–10	4	0.12%
10–20	86	2.51%
20–30	298	8.70%
30–40	364	10.63%
40–50	414	12.09%
50–60	499	14.57%
60–70	454	13.26%
70–80	411	12.00%
80–90	348	10.16%
90–100	180	5.26%
100–110	145	4.23%
110–120	197	5.75%
120–130	22	0.64%
130–140	1	0.03%

Note: Number of Grids represents the quantity of grids in that forest age group, while Percentage of Total Grids represents the proportion of grids in that forest age group relative to the total number of grids.

). The majority of forests in the Qinba Mountain area are concentrated in the 30–90 year age range, accounting for 72.71% of the total forest area. Within this range, forests aged 50–60 years represent the largest proportion, making up 14.58% of the total forest area in the region.

The distribution of forest types in the Qinba Mountain area (as shown in Figure 10 and

Table 7. Proportion of different forest types in the Qinba Mountain area.

	Number of Grids	Percentage of Grids
Natural Forest	2770	85.07%
Plantation Forest	486	14.93%

Note: The horizontal title reflects the number of grids for different forest types and the proportion of these grids relative to the total number of grids, while the vertical title reflects the different forest types.

) reveals that natural forests dominate the region, comprising 85.07% of the total forest area. In contrast, plantations account for 14.93% of the total forest area.

In the Qinba Mountain area, there are 2664 natural forest grids and 437 plantation forest grids. The ratio of plantation forests to natural forests is 16.4%, which serves as the typical proportion (P) for plantation forests relative to natural forests in the region. By combining the forest age and type data, we can calculate the carbon source/sink ratio for each forest age group.

Upon analyzing the data presented in

Table 8. Carbon source/sink ratio across different forest age groups.

Forest Age Grouping	Carbon Source Ratio	Carbon Sink Ratio
1–10	0.00%	0.00%
10–20	33.33%	31.91%
20–30	51.52%	23.03%
30–40	18.37%	19.31%
40–50	28.99%	19.03%
50–60	10.84%	20.93%
60–70	20.00%	19.13%
70–80	3.85%	10.56%
80–90	10.53%	7.55%
90–100	4.17%	10.77%
100–110	0.00%	15.91%
110–120	5.88%	11.45%
120–130	/	25.00%
130–140	/	0.00%

Note: The horizontal title reflects the ratio of artificial forests to natural forests within the same functional groups across different forest age groups, while the vertical title represents the different forest age groups.

, the following conclusions can be drawn:

For forests aged between 10 and 70 years, 91.67% of the carbon source/sink ratios exceed 16.4%, indicating that plantation forests in this age group generally have stronger carbon source and sink capabilities than natural forests. These plantation forests account for 10.64% of the total grid area.

Specifically, for forests aged between 10 and 30 years, the carbon source and sink ratios reach their highest values, indicating that plantation forests in this age range possess the strongest carbon source and sink capabilities. These forests represent 2.29% of the total grid area.

In contrast, for forests aged between 70 and 130 years, 90.91% of the carbon source/sink ratios are below 16.4%, suggesting that natural forests in this age group exhibit stronger carbon source and sink capabilities than plantation forests. These natural forests account for 35.05% of the total grid area.

Within the 70 to 90-year age group, both the carbon source and sink ratios are at their lowest, highlighting that natural forests in this range have the most significant carbon source and sink capabilities. These forests make up 21.64% of the total grid area.

In summary, plantation forests show stronger carbon source and sink capabilities when aged between 10 and 70 years, while natural forests exhibit superior carbon source and sink capabilities when aged between 70 and 130 years. Plantation forests have the strongest carbon source and sink capabilities at ages between 10 and 30 years, while natural forests excel in this regard between the ages of 70 and 90 years.

4.3.3. Relationship Between Remote Sensing Inversion Results of Forest Carbon Sinks and Elevation in the Qinba Mountain Area

Based on the vertical zonal distribution characteristics of vegetation on the southern slope of Mount Taibai in the Qinling Mountains [47], the DEM of the study area was reclassified into nine elevation groups: <750 m, 750–1000 m, 1000–1500 m, 1500–2350 m, 2350–2650 m, 2650–3000 m, 3000–3300 m, 3300–3700 m, and >3700 m. The grid data for these nine groups were then intersected with the grid data from the remote sensing inversion results of forest carbon sinks in the Qinba Mountain area to obtain the forest carbon source/sink results for each elevation group.

Figure 11 shows the average remote sensing inversion results of forest carbon sinks in the Qinba Mountain area for the different elevation groups. It can be observed that as the elevation increases, the forest carbon sinks values also tend to increase.

By summarizing the forest carbon sinks grids within each elevation group and dividing the number of forest carbon sinks grids by the total number of grids in that group, we can calculate the ratio of forest carbon sinks grids to the total number of grids within the same elevation group, as shown in Figure 12.

Figure 12 shows that as elevation increases, the proportion of forest carbon sinks also rises. This indicates that at higher elevations, forests predominantly act as carbon sinks, while the proportion of forests functioning as carbon sources decreases with increasing elevation.

Additionally, this study has analyzed the changes in source/sink status for the same forest grids at different elevations from 2011 to 2018. It calculated the number of forest grids whose source/sink status either remained stable or changed only once during this period for each elevation group. By dividing these numbers by the total number of grids in each group, the proportion of stable source/sink grids was derived, as shown in Figure 13.

According to the results shown in Figure 13, it can be observed that as elevation increases, the stability of forest grids in the Qinba Mountain area also improves, with a decrease in the frequency of transitions between carbon sinks and carbon sources. In contrast, at lower elevations, forest grids are more prone to switching between carbon sinks and carbon sources, indicating lower stability.

5. Discussion

5.1. Comparison with Previous Studies

Forest carbon sinks exhibit distinct seasonal variations, with notable differences in carbon sequestration between the growing and non-growing seasons [48]. Zhao [29] estimated that respiration values for the dominant forest types (Pinus tabuliformis and Quercus aliena) in the Huoditang forest area of the Qinba Mountain region account for approximately 75% of the annual total, while the GPP during the growing season in the Huoditang forest area accounts for about 59% of the annual total (as shown in

Table 9. Proportion of growing season GPP to total annual GPP.

Growing Season GPP (gCm−2 mon−1)	Annual GPP (gCm−2 mon−1)	Growing Season GPP Proportion
431.59	725.55	59.5%

). From this, it can be inferred that the ecological carbon sinks during the growing season contributes roughly 67% of the annual ecological carbon sinks. Based on the results of this study, the estimated annual average carbon sinks value for the Qinba Mountain forest area from 2011 to 2018 is approximately 36.581 TgC/year, derived from the growing season carbon sinks values. By scaling the annual average carbon sinks values from other studies to the area proportion of the Qinba Mountain forest region and comparing these with the remote sensing carbon sinks inversion results obtained in this study (as shown in

Table 10. Comparison of carbon sinks estimation results for Qinba Mountain forests with those from key literature sources.

Typical Literature	Annual Carbon Sinks Estimate (TgC·a−1)
Spatiotemporal dynamics of forest carbon sinks in China’s Qinba mountains: Insights from sun-induced chlorophyll fluorescence remote sensing (this research)	36.58
The current and potential carbon sink in forest ecosystems in China [49]	19.53
The enduring world forest carbon sink [50]	14.88
Forest carbon storage and sink estimates under different management scenarios in China from 2020 to 2100 [51]	21.92
Projections of China’s forest carbon storage and sequestration and ways of their potential capacity enhancement [52]	21.80

), it is evident that the results of this study are comparatively higher than those from previous studies.

The variability in forest carbon sinks estimations can be attributed to the diverse methods employed across different studies. These methods include integrated analysis approaches [49], stock change methods, flux methods [50], and tree growth modeling techniques [51], each relying on distinct data sources. The inherent differences in these methods can introduce various errors and limitations. For example, the stock change method requires a comprehensive national forest inventory system and is primarily applicable to boreal and temperate forests; the flux method may not fully capture carbon stock changes due to land use changes; and tree growth modeling is influenced by factors such as model selection, sample size, and natural disturbances in forests. The choice of method and data source can result in significant differences in forest carbon sinks predictions across studies. These discrepancies stem from several factors, including the mechanisms emphasized by different models, the resolution and accuracy of data sources, the complexity and applicability of the models, and the simplification methods applied in each study.

The higher results of this study compared to others may also stem from the fact that the linear correlation between SIF and GPP can vary significantly under different conditions, such as vegetation types [53], incident radiation, temperature, evaporative components [21], and seasons [54]. These variations can introduce errors when a simple linear model is used to relate SIF and GPP.

Bai et al. [21] argued that when air temperature (Ta) and environmental factor (EF) are low—i.e., under unfavorable thermal and moisture conditions for vegetation growth—the consistency and correlation between SIF and GPP weaken. Directly using SIF to estimate GPP under such conditions can introduce significant errors. The Qinba Mountain area, situated along the climatic divide between northern and southern China, is intersected by the 800 mm annual precipitation isohyet and the January 0 °C isotherm, with pronounced elevation variations. As a result, many regions within the area experience suboptimal Ta and EF for vegetation growth, reducing the accuracy of GPP estimation via SIF and introducing uncertainties into carbon sinks calculations. According to the ninth national forest resource inventory in Shaanxi Province [27], the dominant vegetation types in the Qinba Mountain area are coniferous forests, broadleaf forests, and mixed forests. However, a strong linear relationship between SIF and GPP was observed primarily in deciduous broadleaf forests, while this correlation was weak in coniferous and evergreen broadleaf forests. Consequently, applying a universal SIF-GPP linear model across the Qinba Mountain area could lead to substantial estimation errors.

The simulation performance of remote sensing ecosystem respiration models also differs across ecosystems [37], and the accuracy of ecosystem respiration values can be influenced by various biotic and abiotic factors, such as ecosystem age and nutrient availability [55]. Given the significant role of the Qinba Mountain area’s forest carbon sinks both within China and globally, estimates by Qin et al. [51] suggest that Chinese forests sequester approximately 230 TgC annually, with the forests of the Qinba Mountain area—representing only 9.49% of China’s forest area—accounting for over 10.6% of the country’s forest carbon during the growing season (June to September). Therefore, using the proportion of forest area to estimate the carbon sinks values for the Qinba Mountain area’s forests, as performed in other studies, may lead to underestimations and errors.

Comparing the results of this study with those from other major forest regions worldwide (as shown in

Table 11. Comparison of Qinba Mountain area forests with forests in domestic and international contexts.

Forests	Annual Carbon Sequestration per Unit Area (TgC·a−1·km−2)
China Qinba Mountain Forests	1.17 × 10−4
Amazon Rainforest [57]	1.03 × 10−4
North American Boreal Forest [58]	3.48 × 10−5
Northeast China Forests [59]	2.7 × 10−4
Southern China Forests [60]	7.96 × 10−5

), it is clear that the carbon sinks capacity of the Qinba Mountain area forests ranks among the highest globally. Among the forests included in this study, their carbon sinks capacity is second only to that of the forests in Northeast China. In the Amazon rainforest, GPP declines during the rainy season, while the dry season is characterized by frequent wildfires. Combined with the region’s vast biomass and high ecosystem respiration rates, these factors contribute to a reduction in overall carbon sequestration [56]. However, due to differences in methodologies and data types used for estimation, significant discrepancies arise, with different researchers often producing varying carbon sinks results for the same forest. This highlights the need for ongoing refinement and improvement of the calculation methods for forest carbon sinks.

5.2. Impact of Forest Type and Age on Carbon Sinks Capacity in Forest Ecosystems

Numerous studies have shown that NPP in forests tends to decrease with increasing forest age [61,62]. Smith and Long [61] noted that forest productivity begins to decline in young forests, with a more pronounced decrease in productivity during the early stages of forest growth. However, other studies have highlighted that mature forests can exhibit strong carbon sequestration capacity [63]. The results of this study support this dual perspective, showing that while the carbon sequestration capacity of forests in the Qinba Mountain area decreases during the early stages of growth, older forests still maintain a significant carbon sequestration capacity. This finding aligns with existing research on the topic.

Sheikh et al. [64] found that in the Himalayan region, forest biomass and carbon storage tend to increase with elevation. The results of this study, which show that forest carbon sinks values also increase with elevation, are consistent with these previous findings.

This study indicates that the carbon sequestration capacity of plantation forests is particularly strong when the forest age ranges from 10 to 70 years, with an overall average ecological carbon sinks value of 25.37 gC m⁻² mon⁻¹, which is nearly identical to the average ecological carbon sinks value of natural forests, at 24.36 gC m⁻² mon⁻¹. Although the proportion of carbon sources is lower and the proportion of carbon sinks is higher in natural forests compared to corresponding grid values, this suggests that plantation forests in the Qinba Mountain area also exhibit significant carbon sequestration effects. Guo and Ren [65] found that although plantation forests tend to have lower species diversity, their biomass, productivity, and carbon absorption rates are comparable to or even exceed those of natural forests. In fact, plantation forests aged 0–80 years can already have biomass levels similar to those of natural forests. Similarly, Chen et al. [66] observed that the carbon storage at the ecosystem level in Masson pine plantation forests is comparable to that in natural forests. The results of this study generally align with these earlier findings.

This study finds that the carbon sinks capacity of plantation forests in the Qinba Mountain region peaks when forest age ranges between 10 and 30 years. This highlights the critical role of young plantation forests in carbon sequestration, underscoring their importance in achieving China’s dual carbon goals. Therefore, future efforts should focus on optimizing forest management strategies to enhance the carbon sequestration potential of young plantation forests.

5.3. Impact of Elevation on Forest Carbon Sinks Capacity

Figure 12 reveals a sharp increase in the proportion of forest carbon sinks between 1000 and 1500 m, followed by a decline in subsequent altitude groups. This trend may be attributed to the dominance of cork oak forests mixed with evergreen species at this elevation. Cork oak forests at higher altitudes grow rapidly, and the presence of evergreen species enhances their carbon sequestration capacity [67,68]. As the altitude increases further, there is a slight decline in the proportion of carbon sinks, though it remains relatively high. This decline could be due to the more challenging growing conditions at higher elevations [69]. Figure 13 indicates that forests at lower altitudes exhibit poorer stability, likely because they are more vulnerable to fluctuations in temperature and precipitation. These regions experience significant climate variability, which can easily influence the carbon sequestration status of forests, leading to frequent shifts between carbon sources and sinks in low-altitude areas [70].

5.4. Limitations and Future Prospects of the Study

Although the products used in this study are relatively accurate, errors may still arise due to variations in methods and models, which could affect the final results. Despite conducting independent analyses for each grid point within the Qinba Mountain area forests, issues such as coarse spatial resolution, environmental changes affecting the GPP-SIF linear relationship, and imperfect parameter fitting in the remote sensing ecosystem respiration model remain challenges.

This study spans a long temporal period; however, the forest area data used remain static, potentially failing to capture forest dynamics and leading to an underestimation of carbon losses in harvested areas. To improve accuracy, future research should incorporate annual land cover change data for more precise carbon sinks assessments.

The temporal mismatch between forest age and type data (from 2020 and 2021) and carbon sinks estimation data (from 2011 to 2018) may introduce uncertainties into the analysis and potentially affect the accuracy of the conclusions, despite the relative stability of land use patterns during the study period. Due to data availability constraints, the current data combination represents a practical and necessary compromise. However, this limitation underscores the importance of improving data alignment in future research. To enhance the precision and reliability of carbon sinks assessments, it is recommended that future studies prioritize the acquisition of forest structural data—such as forest age and type—that are temporally and spatially consistent with the study period.

Forest carbon sinks in the Qinba Mountain region were estimated using SIF remote sensing technology. While comparisons were made with other remote sensing methods, the absence of ground-measured carbon sinks data—due to data limitations—precluded direct validation of SIF-based estimates against field measurements, introducing uncertainties in carbon sinks quantification. Additionally, the lack of ground-based GPP and NEP data further complicates the estimation of carbon sinks, thereby increasing the uncertainty in carbon sinks calculations. To address these limitations, future research should establish long-term ground observation plots in the Qinba Mountains to supplement the lack of field data. This would not only allow for more accurate validation of remote sensing estimates but also provide essential ground truth data to refine and improve the reliability of carbon sinks assessments.

The ecosystem respiration model employed in this study is based on a temperature-dependent formulation but does not account for the critical influence of soil moisture in regulating respiration processes. Previous studies have demonstrated that soil moisture significantly affects ecosystem respiration [71]. The current model is a simplification of ecosystem respiration. Incorporating this variable in future models would improve simulation accuracy and enhance process representation.

Although this study examined the relationships between forest carbon sinks and forest type, age, and elevation in the Qinba Mountains, carbon sinks dynamics are also influenced by various factors such as climate and soil conditions. Relying solely on forest type, age, and elevation provides an incomplete understanding of carbon sequestration patterns. Future research should integrate multisource data, including climatic and soil information, to improve the explanatory power of carbon sinks models and refine predictive accuracy.

Despite these limitations, the approach used in this study holds significant promise for estimating forest carbon sinks at regional and global scales due to its large-scale spatial representativeness. The models employed are simple, direct, and clear in their mechanisms, and they do not require auxiliary meteorological data or vegetation index data [43]. In the future, with further exploration of the relationship between SIF and GPP, as well as optimization of the remote sensing ecosystem respiration model, this method is expected to become one of the effective means for estimating forest carbon sinks in areas lacking flux observation data.

To address the issues discussed in this paper, future studies could utilize alternative remote sensing products for estimation, selecting the most accurate ones; incorporate new factors such as radiation intensity, temperature, evaporative components, and soil moisture to refine calculations; and explore the applicability of the method with respect to different time periods and tree species. Continuous improvements in research topics and methodologies can enhance the feasibility, applicability, and accuracy of using SIF remote sensing to estimate forest carbon sinks, leading to a deeper understanding of global carbon cycles and climate change.

6. Conclusions

The estimation of forest carbon sinks in the Qinba Mountain area using SIF remote sensing aims to assess large-scale forest carbon sinks by independently fitting model parameters for each grid point within the study area. The following conclusions were drawn from this study:

(1)

Feasibility and effectiveness of SIF for estimating carbon sinks: The use of SIF for estimating carbon sinks in the Qinba Mountain area is both feasible and effective. SIF remote sensing technology accurately estimates GPP of vegetation, enabling reliable calculation of forest carbon sinks. The forests in the Qinba Mountain area have significant carbon sinks potential, with the ability to sequester approximately 24.51 TgC annually during the growing season (June to September). Notably, seasonal carbon uptake in this region contributes 10.6% of the national annual carbon sequestration, playing a significant role in advancing China’s dual carbon goals.

(2)

Spatial and temporal variation in carbon sinks distribution: There are notable spatial and temporal differences in the distribution of forest carbon sinks in the Qinba Mountain area. From 2011 to 2018, carbon sinks values decreased spatially from the northwest to the southeast. Interannual variability revealed that the northwest exhibited minimal fluctuations, while the southeast experienced a fluctuating decline followed by a fluctuating increase in carbon sinks over time.

(3)

Influence of forest characteristics on carbon sinks capacity: The carbon sinks capacity of the forests in the Qinba Mountain area is influenced by factors such as forest age, type, and altitude. Plantation forests aged 10 to 70 years demonstrate strong carbon sink capacities, while natural forests aged 70 to 130 years exhibit even stronger carbon sequestration potential than plantation forests. Furthermore, the proportion of forest carbon sinks varies significantly across different altitude ranges, with higher-altitude areas displaying stronger carbon sinks capabilities. This suggests that the carbon sinks potential of forests in the Qinba Mountain area is closely related to both topography and forest structure.