Solar radiation differences drive karst sun and shade leaf carbon sink contribution shifts

Abstract

Global warming has led to pronounced differences in photosynthesis and respiration between sun and shade leaves. However, assessments of the resulting disparities in carbon sink potential and contributions remain limited, and the underlying mechanisms have yet to be systematically elucidated. This study used three carbon sink indicators—gross primary productivity (GPP), sun leaf GPP (GPPsun), and shade leaf GPP (GPPshade)—to reveal the spatiotemporal dynamics and driving mechanisms of carbon sink in the terrestrial ecosystems of southern China. The Lindeman–Merenda–Gold (LMG) model was applied to quantify the relative contributions of climate change to carbon sink variations. The results showed the following: (1) GPP, GPPsun, and GPPshade exhibited increasing but fluctuating trends during the period 2001–2020, with growth rates reaching 10.88, 5.69, and 5.19 g C m⁻² yr⁻¹, respectively. However, GPPshade increased faster than GPPsun in 44.79% of the study area. (2) GPPshade/GPP showed an increasing trend (0.0003 yr^-1), with a mean value of 0.54. The average contribution of shade leaf to the carbon sink was 1.79 times higher than that of sun leaf. (3) Declining solar radiation (SR) dominated this shift. The contribution rates of SR to GPP, GPPsun, and GPPshade were 28.01%, 24.55%, and 34.52%, respectively. SR was the primary driver in 37.46%, 31.02%, and 50.19% of the entire study area. (4) In areas with decreased SR, GPPsun exhibited slow growth, and GPPshade decreased. In areas with increased SR, GPPshade surged, while GPPsun growth decelerated significantly. Shade leaf carbon sink emerged as the dominant contributor to the overall enhancement of vegetation carbon sink. These findings demonstrate a key mechanism—increased GPPshade potential driven by SR decline, suppression of GPPsun, and a resulting restructuring of carbon sink dynamics. This study provides a theoretical support for enhancing terrestrial ecosystem carbon sink and offers valuable insights for advancing global carbon neutrality objectives.

Introduction

Gross primary productivity (GPP) is a fundamental component of the terrestrial carbon cycle, representing the rate at which vegetation assimilates atmospheric CO₂ through photosynthesis [1–5]. Under global warming, sun and shade leaves exhibit distinct responses in both photosynthetic carbon uptake and respiratory carbon release [6, 7]. Consequently, a detailed investigation of the growth potential, relative contributions, and driving mechanisms of sun and shade leaves carbon sink under warming scenarios is essential for accurately assessing current vegetation carbon sequestration and forecasting its future capacity.

Global warming has always been an important reason for the change of vegetation structure and function in terrestrial ecosystems [8–10]. Under this background, sun and shade leaves exhibit notable similarities and differences in photosynthetic and respiratory responses [11], potentially shifting canopy carbon sink dominance from the traditional sun leaf-dominated state to one increasingly governed by shade leaf—thereby altering overall carbon fixation and storage patterns [12–16]. Conventional theory holds that sun leaf, benefiting from ample irradiance, maintains higher photosynthetic rates than respiratory losses, making them the primary contributors to vegetation carbon sink over long timescales [17–21]. In contrast, shade leaf, situated in low-light microenvironments, exhibit reduced photosynthetic activity and lower net carbon uptake [22, 23]. Research in recent years has shown that this pattern may be changing under the influence of climate warming. Temperature rise may inhibit the photosynthetic efficiency of sun leaf [24, 25], and significantly enhance its respiration [26], thereby reducing its net carbon sink capacity [15]. Conversely, moderate temperature increases can stimulate the photosynthesis of shade leaf [27], increasing both respiration and net carbon uptake [28], which enables them to surpass sun leaf and emerge as the new dominant sink component, thus realizing the transformation from sun leaf leading to shade leaf leading. [16, 21]. However, although the two-leaf light use efficiency (TL-LUE) model can distinguish the differences in photosynthesis and respiration between sun leaf and shade leaf [29, 30], it still generally uses average parameters for the entire canopy in carbon sink assessments at most regional scales and over long-term periods. This approach ignores the spatial heterogeneity of photosynthetic and respiratory characteristics due to leaf position, which can bias carbon sink estimates and obscure the transitions in sink dominance driven by warming.

To address these shortcomings, we implemented the TL-LUE model, incorporating the complex topography of southern China and differential light use efficiencies of sun and shade leaves, to develop an enhanced dynamic simulation framework for vegetation carbon sink. We separately simulated the carbon sink of sun and shade leaves and then applied the Lindeman–Merenda–Gold (LMG) approach to quantitatively assess the relative contributions of climate change to the variability in total GPP, sun leaf GPP (GPPsun), and shade leaf GPP (GPPshade). Furthermore, we used structural equation modeling (SEM) to elucidate the driving mechanisms of vegetation carbon sink in southern China and to test the hypothesis that warming induces a shift in dominance from sun to shade leaf. The three primary objectives of this study are as follows: (1) to elucidate the mechanistic responses of the carbon sink of sun and shade leaves; (2) to characterize the spatiotemporal evolution of ecosystem carbon sink in southern China; (3) to quantify and spatially map the relative contributions of key climate drivers to variations in GPP, GPPsun, and GPPshade.This study clarifies the transition mechanism from sun leaf carbon sink dominance to shade leaf carbon sink dominance. It holds significant scientific importance for advancing the understanding and consolidation of carbon sink in vegetation ecosystems.

Study area

The study area covers eight provinces in southern China (21°N–34°N, 97°E–117°E): Yunnan (YN), Guizhou (GZ), Sichuan (SC), Chongqing (CQ), Hunan (HUN), Hubei (HUB), Guangdong (GD), and Guangxi (GX). This region crosses tropical and subtropical zones, with a total area of approximately 2 million km², and plays a crucial role in the national ecological framework and carbon cycle. The region is predominantly characterized by a subtropical monsoon climate [31], with some areas influenced by tropical monsoon conditions [32]. The region features a synchronized pattern of rainfall and heat, an annual average temperature ranging from 14 °C to 24 °C, and annual precipitation between 800 and 2000 mm, all of which are conducive to the accumulation of vegetation carbon sink. However, extreme events such as typhoons and frost damage can affect the stability of these carbon sink [33]. The topography of the study area is complex. The western part includes the Yunnan-Guizhou Plateau and Hengduan Mountains, with widespread karst landscapes, the eastern part is dominated by plains and hills, where the middle and lower Yangtze River Plain and the Pearl River Basin foster rich wetland carbon sink. The region has a dense river network, including the Yangtze and Pearl River systems, providing favorable conditions for wetland carbon sink (Fig. 1). However, human disturbances also affect the carbon sink functions of aquatic ecosystems [34, 35].

Fig. 1.

Overview of the study area

The region’s soils are diverse, mainly consisting of red and yellow soils, along with lateritic red soils and purple soils. These varying fertility conditions influence vegetation growth and carbon storage. Vegetation is primarily composed of subtropical evergreen broadleaf forests, with tropical seasonal rainforests distributed in some areas. The high forest coverage makes the region a major carbon sink [35]. With a population exceeding 400 million, the area has a diversified economy. GD and HUB are industrially developed, while SC and YN are notable for resource-based industries. The service sector is also expanding rapidly. However, population growth and industrial development have led to high energy consumption and carbon emissions, and the expansion of construction land has encroached upon ecological space [36, 37]. Under the carbon peak and carbon neutrality goals in China, various provincial policies have been introduced to promote carbon sink development. Existing studies indicate that the region has substantial total carbon sink potential, but limitations such as poor forest quality, soil erosion, and environmental pollution constrain further enhancements to carbon sink functions [38–40].

Data

Climate data

Precipitation

The annual precipitation (P) datasets for China from 2001 to 2020 were obtained from the National Earth System Science Data Center (http://www.geodata.cn). The data are stored as annual cumulative values in TIFF format with a spatial resolution of 1 km × 1 km (unit: mm), covering the geographic extent of 56°S–65°N and 169°W–180°E.

Temperature

The average temperature (T) datasets for 2001–2020 were generated by the Earth Resources Data Cloud Platform (http://gis5g.com) through interpolation based on observations from 2472 meteorological stations across China. The data are provided in TIFF format with a spatial resolution of approximately 1 km × 1 km (unit: ^oC), covering the main land areas of China, including Hong Kong, Macao, and Taiwan. The WGS_1984 coordinate system is recommended.

Solar radiation

The annual solar radiation (SR) datasets were derived from the GLDAS-2.1 dataset developed by the National Aeronautics and Space Administration (NASA) (https://www.nasa.gov/). The spatial resolution is 0.25°, covering the region of 60°S–180°N and 90°W–180°E. The dataset is based on GLDAS atmospheric analysis fields, GPCP precipitation data, and AGRMET radiation fields [41]. Its reliability has been validated in multiple studies [42].

CO₂

The CO₂ concentration datasets were sourced from the Global Greenhouse Gas Reanalysis 162 datasets (EGG4) from 2001 to 2019, developed by the European Atmospheric Monitoring and Services Program (https://ads.atmosphere.copernicus.eu/), with a spatial resolution of 0.05°. The 2020 data were derived by merging the Orbiting Carbon Observatory-2 (OCO-2) Version-7 XCO₂ product (https://oco2.gesdisc.eosdis.nasa.gov) with the annual average growth rate data from the U.S. National Oceanic and Atmospheric Administration (NOAA). The data were resampled to a 1 km resolution and provided as 141-band TIFF files representing monthly variation. Missing values were set to NaN, covering 56°S–65°N and 169°W–180°E.

Nitrogen deposition

The nitrogen dioxide (NO₂) column concentration datasets were derived from satellite sensors GOME, SCIAMACHY, and GOME-2, using the Differential Optical Absorption Spectroscopy (DOAS) technique and assimilation algorithms developed by the Royal Netherlands Meteorological Institute. The data are stored as 1 km × 1 km raster grids (http://www.temis.nl/airpollution/no2.html). Specific humidity, eastward wind speed, and northward wind speed data were provided by the China National Meteorological Center (https://data.cma.cn/) at T63 resolution. Based on these datasets, a pixel-by-pixel spatial linear regression model was constructed the relationship between nitrogen deposition (N_Dep) and environmental variables [43]. The model is expressed as follows:

1

where N_Dep was nitrogen deposition, a₀, a₁, a₂, a₃, a₄, and b were regression coefficients of linear equation; T was the surface temperature (℃); P was precipitation (mm); H was relative humidity (%); W is wind speed (m/s); N was the concentration of NO₂ column (mg/L).

Soil moisture

The soil moisture (SM) datasets were obtained from the GLDAS-2.0 and GLDAS-2.1 datasets provided by NASA (https://www.nasa.gov/) from 2001 to 2020, with a spatial resolution of 0.25° × 0.25°. The SM datasets consists of soil moisture values for four depth layers (0–10 cm, 10–40 cm, 40–100 cm, and 100–200 cm) during the growing season (April to October) of each year. These were summed to create an integrated 0–200 cm soil moisture datasets, offering detailed spatiotemporal information for subsequent related studies.

Leaf area index

The annual global leaf area index (LAI) datasets with a spatial resolution of 500 m were derived from the reprocessed MODIS Global Leaf Area Index Product (2001–2021) [44] [J/DB/OL]. Journal of Global Change Data Repository, 2023 (10.3974/geodb.2023.10.03.V1). During the data processing, the datasets were preprocessed to generate LAI data specifically for southern China, with a spatial resolution of 500 m × 500 m.

Solar elevation angle

The solar elevation angle datasets were derived from the MOD13A1 product datasets provided by the National Aeronautics and Space Administration (NASA) (https://www.nasa.gov/). Monthly average solar elevation angle data were obtained using the raster calculator function in ArcGIS to aggregate and average the raster data. The MOD13A1 product used has a spatial resolution of 500 m × 500 m and a temporal resolution of 16 days, providing 23 scenes per year of normalized difference vegetation index (NDVI) data for the period from 2001 to 2020.

Accuracy validation data

This study selected three GPP products, ChinaFLUX, GLASS, and MOD17A3HGF V6.1 to validate the accuracy of simulated GPP data. ChinaFLUX data derive from measurements by ChinaFLUX Observation Network and public datasets, providing annual GPP for terrestrial ecosystems in China (10.57760/sciencedb.o00119.00077) [45]. Spatial resolution is 0.083°. Validation used data from 2020. MOD17A3HGF product is based on MODIS sensor and estimates annual GPP via the LUE model (https://lpdaac.usgs.gov). Spatial resolution is 500 m × 500 m. Validation used data from 2020. GLASS GPP product is derived from NOAA AVHRR data (http://www.geodata.cn) [46]. Spatial resolution is 0.05°. Validation used data from 2018. The product provides a long-term historical series, suitable for monitoring vegetation dynamics at extended temporal scales. All three datasets were resampled to 500 m × 500 m to facilitate further analysis.

In order to ensure data consistency, all data were resampled to 500 m × 500 m resolution, and Albers projection was adopted (Krasovsky-1940-Albers).

Methods

The present study is based on multiple climate and ecological factor datasets, including VPD, SR, P, SM, T, N_Dep, CO₂, vegetation types, and LAI. The TL-LUE model was used to simulate vegetation carbon sequestration in the southern region, and total GPP, GPPsun, and GPPshade were calculated, revealing carbon sink differences between different leaf types under climate change. To ensure data accuracy, simulation results were validated against three GPP product datasets (ChinaFLUX GPP, GLASS GPP, and MOD17A3HGF GPP). The validation results showed that the R², PB, and RMSE indices of the simulated data were within a reasonable range, confirming the reliability of the simulation outcomes. Furthermore, based on the simulated data, the direct, indirect, and total effects of climate factors on carbon sequestration were evaluated through trend analysis, linear regression, and contribution decomposition methods. SEM was used to analyze the mechanisms of climate factor influence on vegetation carbon sequestration, dominant regions were identified, and the driving forces behind carbon sequestration changes were revealed (Fig. 2).

Fig. 2.

Flowchart for analyzing the mechanism of vegetation carbon sink process

Model description

This study integrates remote sensing techniques with the Google Earth Engine (GEE) cloud computing platform, optimizing surface albedo parameters and refining the TL-LUE model. GPP is calculated by separately computing GPPsun and GPPshade, which are then summed. These components are derived based on the maximum light-use efficiency of GPPsun and GPPshade, incident direct and diffuse photosynthetically active radiation above the canopy, water stress, temperature scalar, and atmospheric CO₂ concentration.

2

3

4

where GPP denotes gross primary productivity (g C m⁻² yr⁻¹); GPPshade and GPPsun represent the GPP contributions from shade leaf and sun leaf, respectively (g C m⁻² yr⁻¹); ε_msu (g C MJ⁻¹) and ε_msh (g C MJ⁻¹) refer to the maximum light-use efficiency of sun leaf and shade leaf (Table 1); APAR_su (W m⁻²) and APAR_sh (W m⁻²) represent the absorbed photosynthetically active radiation (PAR) by sun leaf and shade leaf, respectively. Specifically, PAR refers to the solar radiation energy within the wavelength range of 400–700 nm that can be utilized by plant photosynthesis. Sun leaves are directly exposed to sunlight, whereas shade leaves aresheltered by other leaves and receive only scattered or transmitted light. APAR_su and APAR_sh are key parameters for evaluating the distribution and utilization efficiency of light resources within plant canopies, and their values directly affect leaf photosynthetic rates and overall vegetation productivity; W_s is the water stress scalar; T_s is the temperature regulation scalar; C_s represents atmospheric CO₂ concentration during photosynthesis (ppm).

Table 1.

Parameters used in the TL-LUE model

Vegetation types	DBF	EBF	ENF	MXF	CRO	GRS	OSH	SVN	WET
εmsh(g C MJ−1)	3.75 ± 0.52	3.26 ± 0.93	3.40 ± 1.19	3.00 ± 0.66	4.80 ± 1.94	4.57 ± 1.67	3.10 ± 0.42	4.65 ± 0.64	2.53 ± 1.02
εmsu(g C MJ−1)	0.92 ± 0.29	1.44 ± 0.64	0.89 ± 0.49	0.80 ± 0.41	1.43 ± 0.75	1.16 ± 0.45	0.65 ± 0.07	3.45 ± 0.64	1.23 ± 0.92
VPDmax(kPa)	4.1	4.1	4.1	4.1	4.1	4.1	4.1	4.1	4.1
VPDmin(kPa)	0.93	0.93	0.93	0.93	0.93	0.93	0.93	0.93	0.93
Topt(°C)	23.1	25.8	19.7	24.5	23.5	20.9	22.3	25.8	24.2
Albedo(α)	0.18	0.18	0.15	0.17	0.23	0.23	0.16	0.18	0.23
Clumping index (Ω)	0.8	0.8	0.6	0.7	0.9	0.9	0.8	0.8	0.9

Calculation of APARsu and APARsh

Incident direct and diffuse PAR above the canopy derived from surface albedo, solar zenith angle, leaf inclination angle, clumping index, LAI_su, LAI_sh, sky clearness index, solar radiation.

5

6

7

8

9

10

11

12

13

where α is the albedo; θ is the solar zenith angle; β is the leaf angle, fixed at 60°; Ω is the clumping index (Table 1); C is the multiple scattering radiation (W m⁻²); PAR_dir and PAR_dif (W m⁻²) are the direct and diffuse PAR incident above the canopy; PAR_{dif, u} (W m⁻²) represents the diffuse PAR below the canopy; LAI_su and LAI_sh are the LAIs of sun and shade leaf, respectively; R is the sky clarity index, equal to S/(S₀ cosθ), where S is the solar radiation (W m⁻²) and S₀ is the solar constant (1367 W m⁻²); is the zenith angle of diffuse radiation.

Temperature regulation scalar

The temperature regulation scalar, a fundamental parameter for calculating GPP at the spatial pixel scale, is derived based on the maximum, minimum, and optimal temperatures for vegetation photosynthesis.

14

where T_s is the temperature regulation scalar; the maximum (T_max) and minimum (T_min) temperatures for vegetation photosynthesis are set to 313.15 K and 273.15 K, respectively. The optimal temperature (T_opt) for plant photosynthesis is the average value for different vegetation types.

Water stress scalar

Water stress scalar is a core parameter for estimating pixel-scale GPP. Derived from vapor pressure deficit (VPD), specific humidity, saturation vapor pressure, air temperature.

15

16

17

where W_s is the water stress; VPD_max and VPD_min are the VPD values when GPP reaches its maximum and minimum, respectively. W_s equals 0 when VPD ≥ VPD_max; W_s equals 1 when VPD ≤ VPD_min [47]. t_air is the air temperature in °C; VP_sat is the saturated vapor pressure (kPa); spfh is the specific humidity (kg kg⁻¹); and pres is the pressure (Pa).

Atmospheric CO₂ concentration during photosynthesis

Atmospheric CO₂ concentration during photosynthesis is a key parameter in estimating GPP at the pixel scale. It is derived based on the CO₂ compensation point without dark respiration, intercellular CO₂ concentration, Rubisco Michaelis–Menten constants, partial pressure of O₂, and the universal gas constant.

18

19

20

21

22

23

24

25

where C_s represents atmospheric CO₂ concentration during photosynthesis (ppm); Γ* is CO₂ compensation point without dark respiration; C_i is intercellular CO₂ concentration (ppm); C_a is atmospheric CO₂ concentration from NOAA global monthly means (ppm); χ is ratio of intercellular to ambient CO₂; is just a process parameter. K is Rubisco Michaelis–Menten constant; η* is viscosity of water at 25 °C (0.8903); K_c and K_o are the Michaelis–Menten constants of Rubisco for CO₂ and O₂, respectively; P_o is partial pressure of O₂ (21 kPa); R is the molar gas constant (8.314 J mol⁻¹ K⁻¹).

Structural equation modeling

SEM enables path analysis, multiple linear regression, and confirmatory factor analysis. SEM identifies system hierarchy, path structure, and causal relationships. Compared with multiple regression, SEM captures direct, indirect, and total effects among variables. This study constructed SEM incorporating climatic factors (CO₂, N_Dep, SR, T, P, and SM) to disentangle direct, indirect, and total effects on GPP, GPPsun, and GPPshade.

Lindeman-Merenda-Gold model for contribution analysis

To evaluate the relationship between climate factors and terrestrial carbon sink, Pearson correlation coefficients can be used to quantify the association between the GPP time series and its drivers. The LMG method quantifies the relative contribution of each driver to GPP variability.

26

27

Here, x_k denotes the explanatory variable; S represents a set of variables; r indicates the order of regressors (r1, …, rp); and S_k(r) denotes the set of regressors entered into the model before x_k. This study used the interannual time series of T, P, SR, SM, CO₂, and N_Dep as independent variables, and the interannual time series of GPP, GPPsun, and GPPshade as dependent variables. The LMG method quantifies the relative contribution of each climatic factor to interannual variability in GPP, GPPsun, and GPPshade.

Accuracy evaluation

Model performance was evaluated by comparing simulated and observed GPP values using three statistical metrics: coefficient of determination (R²), root mean square error (RMSE), and percent bias (PB). These metrics comprehensively assess correlation, goodness-of-fit, and error characteristics between predicted and observed values. The equations are as follows:

28

29

30

where n is the sample size; Y_i,rec and Y_i,obs are simulated and observed values, respectively; and Ῡ_i,rec and Ῡ_obs are their respective means.

Trend analysis

To assess spatiotemporal trends in GPP, GPPsun, and GPPshade across southern China, we applied the least squares method to estimate annual change rates:

31

where K is the slope of change; n is the number of years; j represents the year; and M_j is the variable value in year j.

F-tests were applied to assess the statistical significance of interannual trends in GPP, GPPsun, and GPPshade under future climate conditions.

32

where y_i and ŷ_i are the observed and fitted values in year i, respectively; ӯ_i is the mean value over the study period. Based on F-test results, trends were classified into four levels: significantly increasing (K > 0 & p < 0.05), non-significantly increasing (K > 0 & p > 0.05), significantly decreasing (K < 0 & p < 0.05), and non-significantly decreasing (K < 0 & p > 0.05).

Result analysis

Accuracy verification

In this study, the reliability of the simulated GPP data was evaluated by comparison with three established products, which are MOD17A3HGF V6.1, ChinaFLUX, and GLASS. The ChinaFLUX dataset was generated by integrating observational data from the Chinese Flux Observation and Research Network with multiple publicly available datasets to produce annual GPP data for Chinese terrestrial ecosystems. Both MOD17A3HGF V6.1 and GLASS have been widely applied in regional and global GPP studies and are commonly used for reliability assessment. The evaluation procedure employed in this study involved randomly sampling a fixed number of 5000 pixels per year from each product dataset, performing the sampling for each year, and then aggregating the sampled data across years for regression fitting. The comparison with ChinaFLUX yielded an R² of 0.86, a PB of 0.95%, and a RMSE of 200.06 g C m⁻² yr⁻¹. For GLASS-GPP, the R² was 0.84, PB was 1.98%, and RMSE was 224.63 g C m⁻² yr⁻¹. When compared with MOD17A3HGF, the R² was 0.90, PB was 2.43%, and RMSE was 211.81 g C m⁻² yr⁻¹ (Fig. 3). These results indicated that the simulated GPP data possess high accuracy and provided a reliable foundation for regional-scale research.

Fig. 3.

Accuracy verification of the simulated GPP in this study. a The consistency check between thesimulated GPP and the ChinaFLUX GPP product. b The consistency check between the simulated GPP andthe GLASS GPP product. c The consistency check between the simulated GPP and the MOD17A3HGFGPP product. d The statistical relationship between the simulated GPP and the GPP product

Evolution trend of the vegetation carbon sink in southern China

During the period 2001–2020, the vegetation GPP in southern China exhibited a significant increasing trend, with an average growth rate of 10.88 g C m⁻² yr⁻¹ (p < 0.05; Fig. 4). An increasing trend occurred in 74.84% of the region, with 28.31% showing a statistically significant increase. Especially in YN, reaching 24.96 g C m⁻² yr⁻¹. In contrast, only 25.16% of the area decreased, and the significantly decreased area accounted for 2.87%. GPPsun increased significantly at a rate of 5.69 g C m⁻² yr⁻¹ (p < 0.05), with an upward trend in 81.63% of the area. In YN, the increasing trend reached 7.28 g C m⁻² yr⁻¹. GPPshade exhibited a significant increase of 5.19 g C m⁻² yr⁻¹ (p < 0.05), with 67.08% of the region showing an increasing trend. The areas of decline were concentrated in the central-eastern region. The areas of increase were mainly distributed along the periphery.

Fig. 4.

GPP change trends in southern China from 2001 to 2020. a The spatial variation trend of GPP. b The spatial variation trend of GPPsun. c The spatial variation trend of GPPshade. d The temporal variationtrend of GPP, GPPsun, and GPPshade

Further analysis shows that vegetation GPP in southern China exhibited significant spatiotemporal heterogeneity (Fig. 5). The annual mean GPP in YN reached 3400 g C m⁻² yr⁻¹, maintaining the highest level. High vegetation density and SR resulted in elevated GPPsun and GPPshade values. The annual mean GPP in GD and GX reached 3200 and 2600 g C m⁻² yr⁻¹, respectively. Shade leaf made a substantial contribution to the carbon sink. The annual mean GPP in CQ and GZ ranged from 1500 to 2000 g C m⁻² yr⁻¹. It is worth noting that although the high vegetation density and light drive the simultaneous growth of GPPsun and GPPshade in YN, the carbon sink in shade leaf has become dominant in complex terrain areas such as CQ and GZ. This emphasized the critical role of shade leaf in carbon sink.

Fig. 5.

Carbon sink fluxes of each province in southern China from 2001 to 2020

Contributions of sun and shade leaf carbon sink to the total carbon sink in southern China

To investigate GPP variation characteristics in southern China, this study analyzed the trend of GPPsun/GPP and GPPshade/GPP ratios (Fig. 6). The results indicated that GPPshade was the dominant contributor to the regional increase in GPP. The mean GPPshade/GPP ratio reached 0.54, showing a slight upward trend over the past two decades with an annual growth rate of 0.0003. In comparison to GPPshade, GPPsun exhibited a less contribution, with the mean GPPsun/GPP ratio being 0.46 and showing an annual decline rate of 0.0003 yr^-1.

Fig. 6.

Contributions of GPPsun and GPPshade to GPP variation. a The spatial variation pattern of GPPsun/GPP. b The spatial variation pattern of GPPshade/GPP. c The temporal trend of GPPsun/GPP and GPPshade/GPP

Spatially, GPPsun drives the region of carbon sink were concentrated along the CQ, HUB, HUN, GZ, and GX belts, and most areas are driven by GPPshade, further confirming the dominant role of GPPshade in regional GPP variation.

Contribution of climate change to the carbon sink of karst vegetation in southern China.

From 2001 to 2020, CO₂, N_Dep, T, and P in terrestrial ecosystems of southern China showed increasing trends, with annual growth rates of 1.92 mg L⁻¹ yr⁻¹, 0.02 mg L⁻¹ yr⁻¹, 0.02 °C yr⁻¹, and 4.25 mm yr⁻¹, respectively. SR and SM exhibited decreasing trends, with annual change rates of − 12.47 MJ m⁻² yr⁻¹ and − 0.18 mm yr⁻¹, respectively (Figs. 7 and 8).

Fig. 7.

Spatial patterns of climate factor trends in terrestrial ecosystems of southern China from 2001 to 2020. a The trend of CO₂. b The trend of N_Dep. c The trend of SR. d The trend of T. e The trend of P. f The trend of SM

Fig. 8.

Trends of climate factors in terrestrial ecosystems of southern China from 2001 to 2020. a The trend of CO₂. b The trend of N_Dep. c The trend of SR. d The trend of T. e The trend of P. f The trend of SM

This study quantified the relative contributions of climatic factors (N_Dep, SR, T, P, SM, and CO₂) to GPP, GPPsun, and GPPshade using the LMG model. According to Table 2, SR was the dominant driver of vegetation GPP variation in southern China, its relative contributions to GPP, GPPsun and GPPshade were 28.01%, 24.55% and 34.52% respectively. Next were N_Dep (15–17%), T (13–19%), CO₂ (12–14%), SM (12–13%) and P (10–12%). Spatial heterogeneity was observed in the relative contributions of different climatic factors to GPP (Fig. 9). For example, the relative contribution of SR to carbon sink was most prominent in YN, SC and GZ, while the role of N_Dep was concentrated in eastern SC, western GZ and the border areas of HUN, GZ and GX. The high contribution area of T was located in the northwest of SC and the coastal areas of GD and GX (Figs. 9, 10, and 11). Generally speaking, the limiting effect of light conditions was the core factor that determines the spatial difference of vegetation carbon sink in southern China, and the increase of N_Dep and T compensates the adverse effects of radiation decline to some extent. In particular, the sensitivity of GPPshade to SR was significantly higher than that of GPPsun, which revealed that the carbon sink of shade leaf was highly dependent on radiation changes in low light environment (Fig. 11).

Table 2.

Mean relative contributions (%) of climate factors to vegetation GPP variation in southern China

Region	Trend	P	T	SM	SR	CO2	NDep
GPP		11.56	16.21	13.05	28.01	14.14	17.03
GPP increasing		11.92	15.55	14.21	26.01	15.06	17.25
GPP decreasing		10.49	18.14	9.64	33.93	11.42	16.38
GPPsun		11.9	19.39	12.34	24.55	14.52	17.3
GPPsun increasing		11.99	18.24	12.77	24.02	15.45	17.53
GPPsun decreasing		11.49	24.54	10.43	26.86	10.36	16.32
GPPshade		10.43	13.82	13.17	34.52	12.52	15.54
GPPshade increasing		11.29	14.19	15.08	28.97	13.88	16.59
GPPshade decreasing		8.69	13.11	9.27	45.83	9.74	13.36

Fig. 9.

Relative contributions of climate factors to vegetation GPP variation in southern China. a The contribution rate of CO₂ to GPP. b The contribution rate of N_Dep to GPP. c The contribution rate of SR to GPP. d The contribution rate of T to GPP. e The contribution rate of P to GPP. f The contribution rate of SM to GPP

Fig. 10.

Relative contributions of climate factors to vegetation GPPsun variation in southern China. a The contribution rate of CO₂ to GPPsun. b The contribution rate of N_Dep to GPPsun. c The contribution rate of SR to GPPsun. d The contribution rate of T to GPPsun. e The contribution rate of P to GPPsun. f The contribution rate of SM to GPPsun

Fig. 11.

Relative contributions of climate factors to vegetation GPPshade variation in southern China. a The contribution rate of CO₂ to GPPshade. b The contribution rate of N_Dep to GPPshade. c The contribution rate of SR to GPPshade. d The contribution rate of T to GPPshade. e The contribution rate of P to GPPshade. f The contribution rate of SM to GPPshade

In addition, based on partial correlation coefficients between climate factors and vegetation GPP, GPPsun, and GPPshade, this study further identified the characteristics of its main control area (Fig. 12). Among all climate factors, SR and N_Dep showed the strongest dominant influence on vegetation GPP in southern China. The proportion of main control area was 37.46% and 17.30%, respectively. Regarding the GPPsun variation, SR-dominated regions covered 31.02% of the whole area. T-dominated regions accounted for 22.53%. Regarding the GPPshade variation, 50.19% of the study area was dominated by SR. N_Dep-dominated regions reached 13.07%. There were significant differences in the main control functions of climate factors in space. GZ and YN were mainly controlled by SR, while GD and GX were more prominent in T and P, and HUN was relatively enhanced by N_Dep. Generally speaking, GPPsun was more easily controlled by T, while GPPshade was dominated by SR in almost all provinces, and the proportion of master control was generally higher than GPPsun. This showed that the carbon sink of sun leaf was more sensitive to heat conditions, while the carbon sink of shade leaf was constrained by light environment for a long time.

Fig. 12.

Dominant control characteristics of climate factors on GPP variation in southern China. a The dominant climate factor regions for GPP variation. b The dominant climate factor regions for GPPsun variation. c The dominant climate factor regions for GPPshade variation. d The proportional area (%) of dominant regions for each climate factor related to GPP. e The proportional area (%) of dominant regions for each climate factor related to GPPsun. f The proportional area (%) of dominant regions for each climate factor related to GPPshade

Driving mechanisms of climate change impacts on the vegetation carbon sink in southern China

In the GPP model, SR, CO₂, SM, P, and T showed highly significant path correlations with GPP (p < 0.001). SM, CO₂, and T drove GPP increases through direct effects. P influenced GPP primarily through indirect effects. CO₂ had the strongest driving effect (total effect value of 0.67). The effect values of SM and T reached 0.57 and 0.45, respectively. Significant path correlations from P to SM, and from CO₂ to N_Dep were also observed (Fig. 13)_.

Fig. 13.

Path effects of climate factors on vegetation carbon sink in southern China. a The driving path effect of climate factors on GPP. b The driving path effect of climate factors on GPPsun. c The driving path effect of climate factors on GPPshade. ***, ** indicate highly significant (p < 0.01) and significant (0.01 < p < 0.05) levels, respectively

Mechanistically, precipitation promoted GPP by increasing soil moisture and affecting microbial activity. Elevated CO₂ concentration enhanced microbial activity and accelerated nitrogen transformation, thereby driving GPP increases. In terms of GPPsun variation, CO₂, SR, T, and SM exhibited significant effects. CO₂ acted through both direct (0.52) and indirect (0.21) effects (Fig. 14). Other factors primarily exerted direct effects. For GPPshade, SR and SM showed significant impacts. CO₂ presented the strongest driving effect (total effect value of 0.63). SM, SR, and T followed with values of 0.62, 0.43, and 0.42, respectively. Except for P, which was primarily influenced through indirect effects, the other factors were dominated by direct effects. Regarding photosynthesis, SR exerted direct effects of 0.77, 0.65, and 0.83 on GPP, GPPsun, and GPPshade variation, respectively. SR promoted photosynthetic processes, benefiting vegetation growth in most regions [48].

Fig. 14.

Effect of climate factors on carbon sink index. a The direct effect of climate factors on GPP variation. b The indirect effect of climate factors on GPP variation. c The total effect of climate factors on GPP variation

Discussion

Comparison with previous studies

Based on the data of remote sensing monitoring and model simulation from 2001 to 2020, this study systematically revealed the temporal and spatial evolution characteristics of vegetation carbon sink function in southern China. It provides new evidence for understanding the dynamic distribution of carbon sink in shade and sun leaf, and clearly points out that the relative contribution of GPPshade is significantly enhanced, while the share of GPPsun is gradually decreasing, which expands the understanding of the structural transformation of vegetation carbon sink. This result is consistent with the existing research direction on the enhancement of the importance of undergrowth system, but it shows regional differences in order of magnitude and mechanism explanation, which further highlights the theoretical value and innovative significance of this study. First of all, in the conclusion of the trend, this study is consistent with the existing observation and simulation results. Previous studies have shown that climate warming promotes the expansion and photosynthetic enhancement of shade-tolerant communities (shrubs and herbs) under forests, thus enhancing the contribution of carbon sink of shade leaf [49]. In tropical forests, the photosynthetic rate of undergrowth increased significantly under the condition of warming, while the sunlight of sun leaf was inhibited due to high temperature stress [50]. The global vegetation model (CLM5) predicts that the proportion of carbon sink in shade leaf of temperate forests will increase from 30% to 45% by 2100 [51]. The results of this study show that GPPshade/GPP increases at a rate of 0.0003 yr⁻¹, which is consistent with the above research and emphasizes the rising role of understory system in the future carbon sink pattern. Secondly, there are differences between this study and the global simulation study in terms of order of magnitude performance and regional response. Previous studies based on TL-CRF model pointed out that the annual scale of GPPsun was about 500 g C m⁻² yr⁻¹ lower than that of GPPshade [34], but the results of this study showed that the absolute growth rate of GPPsun in southern China was slightly higher than that of GPP shade (5.69 vs 5.19 g C m⁻² yr⁻¹). This difference may come from three aspects: first, the scale difference is studied-the global simulation integrates multi-latitude vegetation types, while this study focuses on the forest system in southern China. Second, there are differences in methods-there are inconsistencies between remote sensing inversion and process model in parameterization of shade and sun leaf and distribution of LAI. Third, the ecological background is different-there are many types of forests in the south, and the species composition and disturbance history are different from those in tropical rain forests or temperate forests. Therefore, this study not only verified the universality of the contribution of shade leaves, but also revealed the regional specific mechanism. This regional variation also suggests that the carbon sink growth potential of sun leaf in southern China’s forest systems may be less constrained by macro-climatic factors compared to the global average, which holds practical significance for localized carbon sink management strategies. More importantly, this study emphasizes that although the carbon sink of sun leaf grows faster, its proportion in the overall GPP decreases, while the relative share of shade leaf continues to increase. This result suggests that the growth of sun leaf may have been limited by light inhibition and heat stress, while the undergrowth community gained relative advantages due to the extension of growing season and shade tolerance. This comparative analysis deepens the understanding of the differential response mechanism of the sun-shade system under climate change, and provides a new empirical support for improving the representation of the shady process in the land surface model.

Different process mechanisms between GPPsun and GPPshade

Photosynthesis is the core physiological process of plant growth and carbon fixation, and its characteristics largely reflect the capacity of plants to respond to environmental change [52]. Sun leaf refers to leaves that grow in a sufficient light environment for a long time and usually have the characteristics of thick palisade tissue and small leaf area, while shade leaf refers to leaves that grow in a weak light environment, usually with a large leaf area and a thin cuticle. Leaf responses to SR exhibit a clear spatial hierarchy, particularly in the differences in photosynthetic efficiency and carbon sink capacity between sun and shade leaves [20]. The empirical and model studies show that increasing the proportion of scattered radiation can significantly improve the photosynthetic carbon absorption at canopy scale, and the gain mainly comes from the increase of light utilization efficiency in the shade leaf. Further comprehensive review and multi-source observation also pointed out that in the multi-layer canopy or high LAI ecosystem, the marginal light response of shade leaf is stronger under the condition of scattered light, so when the radiation component changes from direct light to higher scattering ratio, it contributes more to GPP promotion [53, 54]. At the ecosystem scale, sun leaf is exposed to high irradiance and often operates at or near the light saturation point. Further increases in radiation confer limited additional carbon uptake and may even induce photoinhibition or stomatal closure that reduces photosynthetic efficiency [24, 25]. In contrast, shade leaf positioned in the middle and lower canopy remain below light saturation and respond more sensitively to moderate increases in diffuse radiation. Studies have reported that in regions with enhanced SR, increased diffuse light penetrates the canopy and is used by shade leaf, significantly improving their photosynthetic efficiency and carbon sink capacity [28, 32, 55]. Conversely, under reduced radiation, sun leaf may recover from high-light stress with slight gains in photosynthesis, while shade leaf may fall below the light compensation point and exhibit a net carbon release [56, 57].

At the physiological level (Fig. 15), sun leaf in the upper canopy absorb large amounts of light energy, rapidly activating the thylakoid membrane system during the light reaction. Water photolysis generates ATP and reducing equivalents that drive the Calvin cycle, enabling effective CO₂ fixation via C₃ or C₄ pathways. However, excessive irradiance and high temperature can inhibit key enzymes, limiting dark-reaction rates, unutilized ATP and reducing equivalents accumulate and ultimately suppress organic synthesis, reducing the net photosynthetic rate. Shade leaf in the lower canopy receive less light energy and produce limited ATP and reducing equivalents, constraining dark-reaction initiation. The lower leaf temperature further reduces enzyme activity, slowing CO₂ fixation and organic synthesis. Under scenarios of increased total or diffuse radiation, shade leaf receives more usable light, which accelerates the light reaction and increases the production of ATP and reducing equivalents. Higher leaf temperature under enhanced irradiance also elevates enzyme kinetics, accelerating the dark reaction and improving CO₂ fixation efficiency. As a result, organic synthesis is promoted. Because shade leaf operates well below their light saturation threshold, their photosynthetic potential is far from fully exploited, and moderate increases in irradiance often yield greater relative gains in carbon sink capacity than in sun leaf.

Fig. 15.

Driving mechanisms of photosynthesis in sun and shade leaves

In summary, sun and shade leaves display distinct photosynthetic response mechanisms under varying radiation conditions, manifested in enzyme activity, light reaction product generation and utilization efficiency, and CO₂ fixation rates. These physiological responses combined with canopy light distribution patterns determine the overall ecosystem carbon sink. Under global warming and changing radiation regimes, the carbon sink generated by shade leaf may be markedly enhanced through improved light-use efficiency, shifting the traditional sun leaf-dominated sink pattern and revealing more complex ecological regulation and adaptation strategies.

Limitations and future directions

This study systematically revealed the spatiotemporal patterns of carbon sink contributions from sun and shade leaves and clarified the differentiated regulatory mechanisms of climate factors on GPP and its components, GPPsun and GPPshade. However, there are still certain limitations. The current models do not account for the limiting effects of key nutrients such as nitrogen, phosphorus [58], and potassium on photosynthesis. Previous studies have shown that the allocation patterns of nitrogen and phosphorus within plants are of great significance for regulating productivity. The effect of CO₂ fertilization is mainly constrained by the combined limitation of nitrogen and phosphorus, [58, 59], and declines in the nitrogen and phosphorus contents of leaves may also weaken the CO₂ fertilization effect [60, 61]. In general, species with higher levels of nitrogen and phosphorus tend to exhibit greater photosynthetic capacity [62, 63]. Although this study found that the contribution of the carbon sink generated by shade leaf has shown an increasing trend under climate change, it did not analyze the competitive nutrient allocation mechanisms between sun leaf and shade leaf. Therefore, future research will adopt a three-pronged approach involving multi-nutrient coupled model construction, cross-scale observational data integration, and physiological-ecological mechanism analysis in order to overcome the limitations of the traditional carbon-climate binary framework and to clarify the carbon sink trade-off dynamics between sun and shade leaf under nutrient competition. In the process of vegetation ecological restoration and carbon sink consolidation, it is necessary to maintain the carbon sink potential of sun leaf while placing greater emphasis on the management and optimization of the carbon sink generated by shade leaf. Shifting vegetation carbon sink modeling and management from a focus on phenomenon-based explanation to a mechanism-driven approach will establish a scientific framework and provide technical support for achieving global carbon neutrality targets.

Conclusions

This study revealed the spatiotemporal evolution of vegetation carbon sink in southern China during the period 2001–2020 and evaluated the distinct roles of sun and shade leaf in photosynthesis, along with their underlying driving mechanisms. The characteristics of terrestrial vegetation carbon sink in southern China's ecosystems were further examined. The main conclusions are as follows:

GPP, GPPsun, and GPPshade exhibited increasing trends. From 2001 to 2020, GPP increased at a rate of 10.88 g C m⁻² yr⁻¹ (p < 0.05). The growth rates of GPPsun and GPP shade reached 5.69 g C m⁻² yr⁻¹ and 5.19 g C m⁻² yr⁻¹, respectively. The carbon sink contribution from shade leaf was 1.79 times greater than that of sun leaf. From 2001 to 2020, GPP growth in southern China was predominantly driven by GPPshade. The mean GPPshade/GPP ratio reached 0.54, showing an increasing trend at a rate of 0.0003 yr⁻¹. In contrast, GPPsun/GPP showed a decreasing trend with a rate of −0.0003 yr⁻¹.

Changes in SR significantly influenced the carbon sink responses of shade and sun leaf. The decline in SR was the main factor leading to the higher contribution of shade leaf. Contribution rates of SR to variations in GPP, GPPsun, and GPPshade reached 28.01%, 24.55%, and 34.52%, respectively. The areas where SR had a dominant influence on GPP, GPPsun, and GPPshade accounted for 37.46%, 31.02%, and 50.19% of the study area, respectively.

In regions with decreasing SR, carbon sink from sun leaf showed slow growth, while those from shade leaf declined. In regions with increasing SR, shade leaf carbon sink grew rapidly, whereas the growth of sun leaf carbon sink slowed significantly. As a result, the enhancement of vegetation carbon sink was primarily driven by shade leaf.

Collectively, the results have revealed a systematic shift in vegetation carbon sink patterns, characterized by an increased contribution from shade leaf under declining solar radiation. This shift has underscored the pivotal role of within-canopy light partitioning in regulating ecosystem carbon uptake, and has highlighted the necessity of integrating shade canopy processes into terrestrial carbon models in the context of global climate change.

Author contributions

Conceptualization: Luhua Wu. Methodology, and writing—original draft: Jinjun Du and Luhua Wu. Data curation, formal analysis, and visualization: Heng Wei, Dan Chen and Dongni Yang. Investigation and resources: Lusha Xiong and Dongni Yang. Writing—review and editing, project administration, and supervision: Luhua Wu and Yuanyuan Xia. All authors have read and agreed to the published version of the manuscript.

Funding

Funding support for this research was provided by the National Natural Science Foundation of China (Grant Nos. 42261052, 42461047, U22A20619 and U24A20579), the Guizhou Science and Technology Association Young Scientific and Technological Talents Supporting Project (Grant No. GASTYESS202401), the Guizhou Provincial Science and Technology Project (Grant Nos. ZK[2023]-464, [2022]-010, CXTD[2025]057 and [2024]-014), the Science and Technology Project of Tongren City (Grant Nos. 2023-5, 2023-38 and [2025]-30), the High-level Innovative Talents in Guizhou Province (Grant Nos. 2024-(2022)-051 and GCC[2022]015-1), the Program of Engineering Research Center of Intelligent Monitoring and Policy Simulation of Mountainous Land Space, Higher Education Institutions of Guizhou province (Grant No. 2023045).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Competing interests

The authors declare no competing interests.