How does thinning affect the carbon budget of a subtropical coniferous plantation? Authors: Mingjie Xu, Shengtong Li, Fengting Yang, Tao Zhang, Xiao Liu, Hui Zhang, Yuting Wang, Jiaxin Song, Ziyi Wang, Xianjin Zhu, Chuanpeng Cheng, Jianlei Wang, Huimin Wang Published: 2025 ABSTRACT AbstractBackgroundTo mitigate climate change, China has established extensive plantations since the 1970s, making a substantial contribution to the terrestrial carbon sink. However, after decades of growth, plantations require effective management due to potential ecological risks. While thinning can provide long-term economic and ecological benefits for planted forests, it also imposes significant short-term disturbances that may result in temporary carbon sink losses. This raises a critical question: How does thinning affect forest carbon budgets on Earth? This issue is very controversial, largely due to variations in climate conditions, methodological approaches, and study scales. Consequently, in this study, three methods were used to quantify the effects of thinning on carbon fluxes and provide a theoretical and practical basis for carbon budget estimation and forest management.ResultsThe 1-tower, look-up table (LUT), and 2-tower methods were used to investigate the effects of thinning on carbon fluxes. Among them, the 2-tower method is considered the most dependable which tracked the seasonal variation pattern of the carbon fluxes best. The 1-tower method is direct and easy to conduct but may involve estimation biases introduced by climatic interannual variations. While the LUT method could theoretically diminish the effects of the varying climate, it was weak in tracking relatively extreme values. The three methods yielded consistent results, indicating that 25% thinning enhanced gross primary productivity (GPP), net ecosystem productivity (NEP), and ecosystem respiration (Re). According to the 2-tower results, in the first year after thinning, the GPP increased by 9.8%, and it increased more in the second year after thinning, reaching 14.6%. However, a much greater increase in Re was found in the second year after thinning than in the first year, with values of 23.2% and 12.6%, respectively. Consequently, the increases in NEP induced by GPP were offset by increasing Re, which was 5.4% and 0.5% in the first and second years after thinning, respectively.ConclusionsThinning enhanced the GPP, Re, and NEP in the subtropical forest, indicating the positive effects of thinning in a near-mature coniferous plantation, even shortly after thinning.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13021-025-00374-5.Keywords: Thinning, Forest management, Photosynthesis, Respiration, Eddy covariance Background Global warming caused by greenhouse gas emissions greatly threatens the stability of terrestrial ecosystems. Reducing carbon emissions and enhancing carbon sinks is a global consensus for mitigating climate change. Forests are the most important terrestrial carbon sinks. Expanding planted forest areas and enhancing their carbon sequestration capacity are the most promising choices proposed by the IPCC for addressing climate change [1]. Therefore, forests are established through afforestation and reforestation efforts worldwide [2]. According to the Global Forest Resources Assessment Report released by the Food and Agriculture Organization (FAO) of the United Nations, as of 2020, the global area of planted forests has reached 294 million ha, with China accounting for 80 million ha—approximately 27% of the global total plantation area [3]. However, most artificial forests are monoculture forests. After decades of growth, they may enter a slow-growing phase, during which their carbon sequestration capacity gradually declines. Moreover, due to limited tree species diversity and lower biodiversity, these forests typically exhibit weak resistance to natural disturbances, pests, and diseases, thereby with higher ecological risks [4]. Especially, Potter et al. [5] reported that Pinaceae are highly vulnerable to insect infestations and diseases, which can result in widespread mortality in pure coniferous forests. Therefore, thinning and subsequent reconstruction of artificial forests are necessary for their sustainable development [6, 7]. Although thinning is a favorable choice for forest management, it also strongly disturbs forests, especially in the short term, which may lead to carbon sink loss [8, 9]. Consequently, thinning is regarded as a strategy to achieve long-term ecological and economic benefits at the expense of short-term carbon sequestration capacity. However, some studies have reported contradictory findings [10–12], and the effects of thinning on the carbon balance are still controversial. That is, the understanding of the effects of thinning on the carbon sink function of forests remains limited and insufficient. The short-term and long-term impacts of thinning on the carbon sequestration capacity of forests are still unclear. The uncertainties could be attributed to the following two aspects. First, thinning effects vary significantly across different forest types [13], climate backgrounds, thinning intensities [9, 14], and time since thinning; thus, it is difficult to draw a consistent conclusion on the effects of thinning on forests. In addition, changes in forest carbon sink function, which could be represented by net ecosystem productivity (NEP), depend on the balance between ecosystem gross primary productivity (GPP) and ecosystem respiration (Re). Variations in both GPP and Re consequently contribute to substantial variability and uncertainty in NEP. For example, Aslan et al. [15] reported a decrease in GPP in a boreal coniferous forest in southern Finland in the early stage after thinning but reported that GPP increased steadily for two years after thinning. This finding is consistent with the results reported by Dore et al. [16], which indicated that the GPP even exceeded the values before thinning when the forests recovered after thinning. However, another study found no significant effects of thinning on carbon fluxes [12], which might be the result of a tradeoff between the loss of the photosynthetic canopy area and enhanced regrowth of understory vegetation along with compensatory growth of remaining trees. With respect to Re, most studies reported an expected increase after thinning, which is primarily attributed to elevated heterotrophic respiration [17, 18]. However, some studies have reported a decrease in Re after thinning. This phenomenon could be attributed to the removal of trees by thinning, which weakens the autotrophic respiration of plant roots and stems, reduces litter input, and thus reduces the substrate for heterotrophic respiration [19]. Therefore, the compensatory dynamics among different components of carbon flux make it challenging to accurately assess the overall effects of thinning on forest carbon sink function. Moreover, it should be noted that the effect of thinning on forest carbon budgets is a dynamic process, evolving over time following intervention [10]. Therefore, further research on the impact of thinning is needed. Another reason contributing to uncertainty in assessing the effects of thinning lies in the research methods employed. Previous studies have often been limited to leaf and plot scales [20–22]. However, scaling these results to the ecosystem scale and accurately quantifying the impact of thinning on forest NEP remain challenging. Although some studies have indicated that the maximum photosynthetic rate (Pmax) and apparent quantum efficiency (α) of the sampled leaves of the remaining trees increased [23], it is difficult to assess whether the GPP increased or decreased at the ecosystem scale due to the removal of the photosynthetic canopy area. With the widespread application of the eddy covariance (EC) technique [10–12, 24], ecosystem-scale studies have become increasingly feasible. However, if only a single flux tower is set up in the study area, the results will be confounded by the interannual climatic variability and the associated fluctuations in carbon fluxes [12, 25]. Saunders et al. [10] reported an increase in carbon fluxes 1–2 years after thinning. However, due to elevated temperatures in the post-thinning years, they suggested that the observed increase was likely driven by climatic factors rather than the thinning treatment itself. Therefore, an accurate assessment of the effects of thinning on forest carbon sink function requires parallel EC observations or appropriate methodological approaches to minimize the influence of interannual climatic variability. This study was conducted in a typical coniferous plantation in southern China, an area that has experienced severe soil erosion. The plantation was established in the 1980s and has since been recognized by the United Nations as a model for ecological restoration. In addition, such plantations play a significant role in carbon sequestration. The Qianyanzhou station, established in this region, is the pioneer standard station in China conducting EC observations. To avoid possible future ecological risks, a thinning experiment was carried out at this site. Two flux towers have been set up: one in the permanent standard plot without thinning treatment and the other in the thinning region. To accurately assess the impact of thinning on the carbon budget, three methods were applied based on observations from two flux towers: (ⅰ) Based on 1-tower observations, the differences between pre-thinning and post-thinning carbon fluxes (including GPP, Re and NEP; carbon fluxes referred to the same contents in this article) were directly quantified; thus, the differences represented the effects of thinning on carbon fluxes (Fig. 1a). (ii) The paired 2-tower method was employed, with one tower located in the thinned plot (Tower-th) and the other in the unthinned plot (Tower-un). Before thinning, the relationships between the carbon fluxes (GPP, Re, and NEP) observed by Tower-th and Tower-un were established, respectively. After thinning, the expected carbon fluxes in the thinned plot without thinning treatment were estimated using Tower-un data; the differences between the simulated values and the observed carbon fluxes from Tower-th were used to quantify the thinning effect (Fig. 1b). (ⅲ) The look-up table (LUT) method was applied to remove confounding influences from climatic variability (Fig. 1c). In this study, the three methods were cross-validated to evaluate their consistency and assess methodological uncertainties. The impacts of thinning on subtropical plantations and their temporal dynamics following thinning were systematically analyzed. These findings help fill a research gap and provide valuable insights into the effects and underlying mechanisms of thinning on forest carbon budgets. Sketch map of the three methods used in this study: the 1-tower method (a), the 2-tower method (b), and the look-up table (LUT) method (c). Methods Site description This study was conducted in a typical coniferous plantation in subtropical China, which is located in a typical red soil hilly region in Jiangxi Province. The site is the National Qianyanzhou Critical Zone Observatory of the Red Soil Hilly Region (QYZ for short) (26°44’29"N, 115°03’29"E, 102 m above sea level). QYZ is among the first batch of stations carrying out flux observations based on the eddy covariance technique and is a standard forest observation site in the China Flux Observation and Research Network (ChinaFLUX). The first flux tower (Tower-un) was established in 2002. In 2008, to investigate the effects of thinning, a second flux tower (Tower-th) was installed at the QYZ site (Fig. 2). True-color remote sensing imagery map of Qianyanzhou (QYZ) station; the station region is outlined with an orange line, and the shaded area in orange indicates the thinned area. The blue triangle marks the location of the flux tower in the unthinned region (Tower-un). The orange triangle marks the location of the flux tower in the thinned region (Tower-th). The QYZ station is in the region under the control of a typical subtropical monsoon climate. According to meteorological observations in QYZ from 1985 to 2014, the mean annual temperature is 18.0 ± 0.4 °C, and the mean annual precipitation is 1464.5 ± 308.8 mm. Traditionally, the year could be divided into four seasons: spring (March‒May), summer (June‒August), autumn (September‒November) and winter (December‒February). The red soil is weathered from red sandy rock. The soil textures are 2.0–0.05 mm (17%), 0.05–0.002 mm (68%) and < 0.002 mm (15%) [26]. The vegetation at this site is a typical subtropical coniferous plantation. Before 1983, this region experienced severe soil and water erosion. To address this situation, trees were planted in approximately 1985, with pioneer and fast-growing species of Masson pine (Pinus massoniana Lamb.), Slash pine (Pinus elliottii Englem.) and Chinese fir (Cunninghamia lanceolata Hook.). The ecological environment has recovered and improved since then, becoming a model of ecological restoration. However, after years of development, monoculture forests face increasing ecological risks and proper forest management measures are needed to address this issue. Thinning treatment When the trees were planted in the QYZ station in 1985, the density was as high as 1460 stems ha–1. Therefore, many trees were suppressed and relatively small before thinning was conducted. To reconstruct and further improve the forest, the QYZ station conducted relatively large-scale thinning (~ 40 ha) in the winter of 2012 around the second flux tower (Tower-th) established in 2008 (Fig. 2). For study purposes, two parallel tower observations before thinning were conducted during 2009–2012. When thinning was conducted, approximately 25% of the basal area was removed. The commercially valuable parts (timbers) of the tree trunks were removed from the site, and most of the slash was left on the site. The tree stumps were also retained on site, with heights ranging from 25 to 35 cm. Some plots were also set around the tower to further study the effects of thinning. According to the plot investigation conducted in summer, before thinning, the mean stand density was 1329 stems ha−1, and it was 849 stems ha−1 0.5-year after thinning (Table 1). The stand structure characteristics, such as diameter at breast height (DBH) and leaf area index (LAI), were investigated and measured before and after thinning and are listed in Table 1. The LAI was calculated from the TRAC measurements by TRAC-WIN [18]. Table 1Stand structure characteristics before and after thinningStand density (stems ha−1)Leaf area index (LAI)DBH(cm)Basal area(m2 ha− 1)Before thinning13292.915.926.40.5-year after thinning8492.317.319.92.5-year after thinning8252.417.620.1 Stand structure characteristics before and after thinning Observation and instrumentation The EC measurement system was installed on the tower at heights of 23 m (Tower-un) and 18 m (Tower-th), respectively. The EC system included a 3-D sonic anemometer (Model CSAT3, Campbell Scientific Inc., Logan, UT, USA) and an open-path CO2/H2O analyzer (Model LI-7500, LI-COR Inc., Lincoln, NE, USA). All signals were collected at a frequency of 10 Hz, and the CO2 and H2O fluxes were calculated and recorded at 30 min intervals by a CR3000 datalogger (Campbell Scientific Inc.). The environmental factors were observed synchronously. Air temperature (Ta), water vapor pressure, and relative humidity were measured via a sensor (Model HMP45C, Vaisala Inc., Helsinki, Finland) in a ventilated solar radiation shield. The soil temperature and soil water content (SWC) were measured at a depth of 5 cm with a temperature sensor (Model 105 T, Campbell Scientific Inc., Logan, USA) and TDR probes (Model CS615-L, Campbell Scientific Inc.), respectively. Radiation was measured via a four-component net radiometer (Model CNR-1, Kipp & Zonen, Delft, ZuidHolland, Netherlands) and a pyranometer (Model CM11, Kipp & Zonen). Photosynthetically active radiation (PAR) above the canopy was measured with a quantum sensor (Model LI190SB, LI-COR, Inc.). Precipitation (PPT) was monitored with a rain gauge (Model 52203, RM Young Inc., Traverse, MI, USA). The environmental variables were sampled at 1 Hz, and 30 min average data were recorded with a CR1000 datalogger (Campbell Scientific, Inc.). Flux data correction To ensure the reliability and consistency of flux data across China, ChinaFLUX has developed a series of standard methods for processing data [27]. Accordingly, in this study, the flux data were processed by ChinaFLUX standard data processing methods [28]. The raw data were corrected by axis rotation to eliminate the possible influences of local terrain or instrument tilt. Changes in temperature affect water vapor density, thereby affecting latent heat flux, and change CO₂ density, thus inducing “fake” fluxes. Consequently, the Webb Pearman and Leuning correction (WPL correction) [29] has been widely used to eliminate disturbances introduced by water vapor and temperature. The equations are as follows. First, the water vapor flux should be corrected: where µ represents the molar mass ratio of air to water vapor; ρv and ρd represent the densities of water vapor and dry air, respectively; E0 represents the uncorrected latent heat flux; S represents the sensible heat flux; and Cp represents the specific heat of air (J kg⁻¹ K⁻¹). The correction equation for CO₂ flux is as follows: In this equation, E is the corrected latent heat flux, F0 is the observed CO₂ flux, ρc is the concentration of CO₂, and the other parameters are the same as above. After the correction, the spurious data caused by system failures or rainfall were removed, and then the storage fluxes were assessed and added to the flux values. To avoid possible underestimation of the fluxes under poor turbulent conditions at night, nighttime (solar elevation angle < 0) data were excluded when the friction velocity (u*) was lower than the threshold calculated according to the method described by Reichstein et al. [30]. The u* thresholds for tower-un and tower-th for each year are presented in Table 2. Table 2The u* (m s−1) threshold for the tower in the unthinned region (Tower-un) and for the tower in the thinned region (Tower-th) during 2009–2014Year200920102011201220132014Tower-un0.18310.19920.19530.15730.21850.1560Tower-th0.18310.18040.17240.18080.21180.1823 The u* (m s−1) threshold for the tower in the unthinned region (Tower-un) and for the tower in the thinned region (Tower-th) during 2009–2014 After the anomalous data were excluded, data gaps were generated. The data gaps should be filled when estimating the carbon budgets or analyzing seasonal and interannual variations in carbon fluxes. Data gaps were filled using the mean diurnal variation method and linear and nonlinear methods [31]. Further detailed information about the flux data correction and calculation can be found in previous relevant studies [32, 33]. Calculation of the footprint The footprint model proposed by Kormann and Meixner [34] was used in this study for flux source area analysis. The model is constructed based on the Monin-Obukhov similarity theory and mass conservation principles and calculates the spatial distribution of flux contribution functions through analytic methods. The key parameters for the model included the following: measurement height, 23 m for Tower-un and 18 m for Tower-th; canopy height (h); stability parameter (z/L); surface roughness length estimated from canopy structural characteristics; friction velocity (u∗); wind direction (WD); and standard deviation of the lateral wind speed. To present the footprint and the underlying surface situation visually, a spatial overlay analysis between high-resolution land use data and footprint contribution distributions (90% cumulative area) was conducted. The look-up table method To diminish the effects of the climate background, the look-up table (LUT) method was used in this study [35–37]. To create the tables, the 30-min carbon flux data and corresponding meteorological data observed by Tower-th were aggregated to the daily scale for each year. An average table (LUTavg) that included meteorological data and carbon fluxes for the entire time series from DOY 1–366 (including leap years) was established on the basis of the observations before thinning (2009–2012). In LUTavg, the carbon fluxes represent the unthinned carbon fluxes. Then, an index table based on LUTavg was built according to Ta (class width of 0.2 °C) and SWC (class width of 0.02 m3 m− 3); thus, the carbon flux data could be found according to the observed Ta and SWC. Finally, according to the observed meteorological data in 2013 and 2014, the index table based on LUTavg was looked up for, and thus the carbon flux series for 2013 and 2014 representing the unthinned situation could be obtained and marked as flux-lut. The thinning effects could be quantified as the difference between the carbon flux observed by Tower-th and flux-lut. The data corrections, calculations, statistical tests, and visualizations were conducted with MATLAB R2023a software (MathWorks Inc., Natick, MA, USA), PyCharm Professional 2023.1 (JetBrains s.r.o., Prague, Czech Republic) and Origin2021 software (Origin Lab Corporation, Northampton, MA, USA). Results and discussion Variations in carbon fluxes and direct evaluation of the effects of thinning Thinning is an important silvicultural practice widely applied in plantation forests. In the long run, it provides many benefits, such as improving forest structure; accelerating understory vegetation regeneration; enhancing the quality of the remaining trees; increasing resilience to natural disturbances such as pests, diseases, and extreme weather events; and alleviating the contradiction between wood supply and demand [38–40]. Consequently, it could help achieve greater ecological and environmental benefits as well as social and economic benefits [8]. However, in the short term, tree removal during thinning is generally considered to temporarily impair the carbon sink capacity of forests. Specifically, GPP is expected to decline immediately after treatment due to the reduction in the photosynthetic canopy area [41]. Nevertheless, some studies have reported increases in GPP and NEP, which could be attributed to favorable climatic conditions and compensatory growth of the remaining trees [42]. However, there is still a knowledge gap in subtropical plantations in China, and few studies have employed different methods to evaluate the effects of thinning, which is the focus of this study. The variations in the carbon fluxes in the unthinned plot (Tower-un) and thinned plot (Tower-th) are presented in Fig. 3. The variations in carbon fluxes observed by Tower-un and Tower-th followed the same pattern. Both GPP and Re exhibited unimodal trends, peaking during summer. Compared with these two larger components (GPP and Re), NEP displayed greater variability with numerous small fluctuations, as it is the difference between GPP and Re [11, 43]. Variation in either the GPP or the Re would lead to changes in the NEP and even alter its sign. However, the GPP, Re and NEP were all relatively high in summer and relatively low in winter. The seasonal variations were evident, although this forest is an evergreen coniferous plantation. To directly evaluate the effects of thinning, pre-thinning averages from 2009 to 2012 observed by the Tower-th were calculated, and the differences between post-thinning carbon fluxes (in 2013 and 2014) and these baselines were interpreted as thinning effects in 2013 and 2014, respectively. Thinning increased Re, which was expected, as a previous study in QYZ reported that thinning strongly enhanced soil respiration and stump respiration [18]. However, contrary to expectations, thinning also led to increases in GPP and NEP. Before thinning, the annual NEP was 722.6 g C m–2 yr–1, increasing to 856.3 and 791.0 g C m–2 yr–1 in 2013 and 2014, respectively. The GPP was similar in 2013 and 2014, with values of 1918.2 g C m–2 yr–1 and 1954.6 g C m–2 yr–1, respectively, exceeding the pre-thinning average of 1650.2 g C m–2 yr–1. Re increased more substantially in 2014 (236.0 g C m–2 yr–1) than in 2013 (134.3 g C m–2 yr–1). These increases were nonnegligible; thus, we wondered whether they were caused by changing climatic factors or by thinning. Consequently, the parallel two-tower method and LUT method were used to eliminate or minimize confounding influences from climatic variability. Dynamics of the carbon fluxes observed by the unthinned tower (NEP-un, Re-un, GPP-un) (a) and thinned tower (NEP-th, Re-th, GPP-th) (b–d). Flux-avg represents the daily average values of the carbon fluxes of the thinned tower during 2009–2012, with the light blue shaded area indicating the mean ± SD. Estimation of the effects of thinning on the basis of parallel 2-tower observations To obtain credible results, the observations from Tower-th and Tower-un were conducted in parallel for 4 years before the thinning treatment to obtain completely comparable data. Theoretically, the carbon fluxes observed by these two towers are supposed to be closely related because they were established on the same underlying surface. To obtain more accurate results, the footprints of both towers were analyzed (see Additional file, Fig. A1). The interannual variations in the footprint were small, indicating that the observations remained relatively stable due to consistent conditions on the underlying surface. Consequently, the multiyear average footprint would be representative (Fig. 4a and b). Because the observation height was higher at Tower-un, the footprint covered a larger area, including some paddy fields, with some seasonal variations (Fig. A1 and Fig. A2). However, the paddy fields were consistently located in low-contribution areas. Relationships between the carbon fluxes measured at the unthinned tower (NEP-un, Re-un, GPP-un) and the thinned tower (NEP-th, Re-th, GPP-th) (c–e), along with comparisons of their footprints (footprint-un and footprint-th) (a–b). The dynamics of flux-sim and flux-un are presented in Figure (f–h). Flux-sim values were calculated based on the flux-un data from 2013 and 2014 using the fitting equations in Figure (c–e), representing the estimated carbon fluxes at the thinned tower during 2013–2014 under a no-thinning scenario; the light blue shaded area indicates the confidence interval. We initially assumed that the paddy fields exerted negligible influence, given their relatively limited area within the footprint and their location in the low-contribution section of the footprint. Based on the whole flux dataset observed by Tower-un and Tower-th from 2009 to 2012, the relationships between NEP-un and NEP-th, Re-un and Re-th, and GPP-un and GPP-th were modeled using linear equations (Fig. 4c-e). Fluxes measured by Tower-un were used as the independent variable (x), and those from Tower-th were used as the dependent variable (y). Subsequently, we substituted the data observed by Tower-un in 2013 and 2014 into the corresponding equation; thus, the unthinned fluxes around Tower-th under the climatic background of 2013 and 2014 could be estimated, and they were marked as flux-sim. According to the results (Fig. 4f-h), the in situ observed NEP-th and GPP-th values from 2013 to 2014 were still greater than the NEP-sim and GPP-sim values, indicating that thinning enhanced the carbon sink function of this subtropical plantation. Moreover, the Re increased with a greater magnitude. The results were consistent with the results obtained using the 1-tower method, indicating that even when accounting for climatic influences in 2013 and 2014, the carbon fluxes for post-thinning still increased. To obtain more reliable results, the effects of the paddy fields were taken into consideration, although they were located at the edge of the footprint area with relatively small contributions to the observation results. We first compared the carbon fluxes observed by Tower-un and in paddy fields (Fig. A3). What is interesting is that, the paddy was found with higher GPP and NEP values from the tillering stage to the booting stage for both early and late rice (Tab. A1). We marked these periods with gray shading and named them peak periods (Fig. A3). The fluxes observed by Tower-un may involve disturbances induced by the paddy during peak periods. Consequently, the data were binned into two groups: peak periods and non-peak periods. Subsequently, linear fittings were performed separately for each group (Fig. 5a–f). Combined with the records of the rice growth stages in 2013 and 2014 (Tab. A1), the 2-tower method was improved, and the carbon fluxes were re-simulated. The modeled carbon fluxes under non thinning scenario and were labeled as flux-simP (GPP-simP, NEP-simP and Re-simP, respectively). The data for Apr 25–Jun 13 and Aug 10–Sep 17 in 2013 and Apr 20–Jun 7 and Aug 3–Sep 16 in 2014 observed by Tower-un were substituted into the equations in Fig. 5a–c. The data from the other periods were substituted into the equation shown in Fig. 5d–f to calculate the unthinned flux data around the Tower-th. The values were consistent with those of the 2-tower method without separating peak periods (labeled as flux-sim), as shown in Fig. 5g–i. Comparisons of the results of the parallel 2-tower fitting method under two scenarios: considering cropland (paddy field) flux contributions (2-towerP) and not considering them (2-tower) (g–o). Flux-sim values were calculated based on the flux-un data from 2013 and 2014 using the fitting equations in Fig. 4 (c–e). Flux-simP values are calculated based on flux-un data from 2013 and 2014 using the fitting equations in Figure (a–c) during the rice tillering to booting stages (peak periods); for non-peak periods, the fitting equations in Figure (d–f) were applied. Figure (a–f) show the relationships between the carbon fluxes of the unthinned tower (NEP-un, Re-un, GPP-un) and the thinned tower (NEP-th, Re-th, GPP-th) during the peak and non-peak periods of cropland fluxes, respectively. The peak periods of the cropland fluxes are defined in Figure A3. By subtracting flux-sim and flux-simP from the fluxes observed by Tower-th in 2013 and 2014 (i.e., the carbon fluxes after thinning, flux-th), the daily changes in carbon fluxes caused by thinning estimated by the two methods of whole-period fitting (2-tower) and piecewise fitting (2-towerP) were obtained. The results were quite close, with their relationships closely approximating the 1:1 lines (Fig. 5j–l). When daily values were accumulated, the thinning effects on carbon fluxes at the annual scale could be quantified. As expected, the results obtained by the 2-tower and 2-towerP methods were highly consistent (Fig. 5m–o). When the peak period disturbances were considered, the estimated thinning effects on the NEP and GPP were slightly greater than those obtained by the direct 2-tower method. Owing to the slight effects of the paddy fields as quantified, the contribution of paddy fields could be neglected in most cases. Evaluating the effects of thinning by eliminating Climatic effects Previous studies have indicated that the large interannual variabilities in carbon fluxes driven by climatic factors would diminish or even overwhelm the effects of thinning [10]. To further verify the effects of thinning and minimize the influence of meteorological factors, the LUT method was employed. By looking up the average tables established based on the flux and climatic observations of Tower-th during 2009–2012, the unthinned carbon flux values under the climatic background of 2013 and 2014 could be obtained. This method has been widely used in previous studies to disentangle the effects of climatic factors and ecosystem responses, thereby identifying the dominant driver of interannual variations in carbon and water fluxes [35, 44]. Consequently, this method could be reliable for disentangling the effects of climatic factors and thinning combined with ecosystem responses. This would be reasonable, as part of the thinning effect is inherently reflected in ecosystem responses [8, 45]. When the climatic factors in 2013 and 2014 were used to look up the index table established based on LUTavg, the reconstructed carbon fluxes of 2013 and 2014 with whole time series (DOY 1–365) could be obtained, which represented the carbon fluxes of this ecosystem under unthinned conditions (flux-lut). By comparing the flux-lut with the flux-th, the effects of thinning could be estimated and quantified (Fig. 6). The results again verified that thinning enhanced the NEP, Re, and GPP. Dynamics of NEP (a), Re (b), and GPP (c) observed by the thinned tower (NEP-th, Re-th, and GPP-th) and obtained using the look-up table (LUT) method (NEP-lut, Re-lut, and GPP-lut) in 2013 and 2014. Comparison of the three methods for evaluating the effects of thinning Accurate quantification of the effects of thinning is difficult [46, 47]. The EC enables direct measurements of CO2 exchanges between ecosystems and the atmosphere, offering a valuable approach for assessing thinning effects at the ecosystem scale [48]. However, few studies have reported the effects of thinning based on EC observations [19], possibly because of difficulties in conducting large-scale thinning for experimental purposes. Thus, few studies have reported thinning effects on the basis of parallel EC observations [35]. In this study, based on relatively long-term parallel observations, three methods, namely, the 1-tower observation method (1-tower, Fig. 3), parallel 2-tower fitting method (2-tower, Fig. 4), and look-up table method (LUT, Fig. 6), were employed to clarify the effects of thinning on carbon fluxes and verified that thinning enhanced GPP, Re, and NEP. The values obtained from the three methods were similar (Fig. 7a–c), although minor discrepancies existed. Since “true values” are inherently unavailable, identifying which method is best is difficult. Therefore, in this section, the three methods were compared, and the effects of thinning in the first year (2013) and the second year after thinning (2014) were further discussed. The impacts of thinning on carbon fluxes (NEP, Re, and GPP) were evaluated using three approaches: the 1-tower observation method (1-tower), parallel 2-tower fitting method (2-tower) and look-up table method (LUT). Comparisons of the daily scale thinning effects obtained through the three methods are shown in Figure (a–c), and the yellow solid lines are the 1:1 lines. The values were subsequently accumulated at the annual scale to visualize the effects of thinning obtained through the three methods in 2013 and 2014 (d–f). A three-dimensional figure was used to compare each pair of methods. The data points were all near the 1:1 line, indicating that the results can be mutually verified (Fig. 7a–c). Among them, the productivities (including NEP and GPP) obtained by the 1-tower and LUT methods showed the highest consistency, with data points closest to the 1:1 lines (Fig. 7a and c). This consistency could be attributed to the use of the same data source for these two methods. The 1-tower method utilized the observation data from 2013 to 2014, subtracting the baseline mean values derived from 2009 to 2012. The LUT method also used the climatic data from 2013 to 2014 to look up the table established on the basis of the average values from 2009 to 2012. Consequently, although the values obtained by the 1-tower and LUT methods were the most closely aligned, we could not arbitrarily conclude that they are the best methods. According to the variation dynamics (Fig. 4f–h) and the consistency test (Fig. A4), the 2-tower method better tracked the sharp changes in the carbon fluxes, whereas both the LUT method and the 1-tower method tended to smooth out the sudden changes, potentially obscuring ecosystem responses driven by climatic factors. The averages were used in the 1-tower and LUT methods; thus, they underestimated the variabilities in the carbon fluxes (Figs. 3b–d and 6) and underestimated the high values (Fig. 6). However, at the beginning and end of each year, the LUT method tracked the fluctuation in Re well, with similar variability with the observation values (Fig. 6b). This may explain why the results of the 2-tower and LUT methods were more consistent when the effects of thinning on Re were evaluated (Fig. 7b). However, during the period from May–September, the LUT method could not fully track the fluctuations in carbon fluxes, and the variabilities were underestimated. This may be due to the deficiency in the LUT method. This may be because the dataset for establishing the tables is not large enough. When an extreme climatic factor appeared, the corresponding carbon fluxes could not be found in the index table. If the observation years were extended, the LUT method would yield more accurate results. The daily values of carbon flux were summed to quantify the thinning effect at the annual scale (Fig. 7d–f). All three methods yielded consistent results: thinning increased carbon fluxes, with smaller increases in GPP and Re in 2013 than in 2014, while the increase in NEP was greater in 2013 than in 2014. This could be attributed to the improvement of light conditions and reduced competition for water and nutrients [18]. Although the number of trees decreased, the overall photosynthetic efficiency improved [49, 50]. This led to the fact that thinning not only did not reduce the carbon sequestration capacity of the forest ecosystem but also increased GPP by 9.8% (data according to the 2-tower results, same applies hereinafter). After a year of recovery and adaptation, the photosynthetic capacity of the subtropical forest increased in 2014, with the GPP increasing by 14.6% (Fig. 7f). The increase in photosynthesis led to an increase in autotrophic respiration, as they were found with a relatively steady ratio [51, 52]. Furthermore, heterotrophic respiration increased substantially due to thinning-induced debris decomposition, greater substrate availability for soil respiration, and enhanced stump respiration [18]. Consequently, the increase in Re was large, increasing by 12.6% in 2013 and by 23.2% in 2014 (Fig. 7e). Notably, the increase in Re in 2014 was greater than that in GPP, which could be attributed to the further increase in stump respiration and soil respiration. After one year of decomposition, the stumps became more susceptible to decay, and thinning provided additional substrate for heterotrophic respiration [18]. Moreover, respiration rates could be substantially increased under hot and humid subtropical climatic conditions. As the NEP equals GPP minus Re, the sharp increase in Re in 2014 resulted in a marked decline in the increment of NEP in 2014. NEP increased by 5.4% in 2013 but only 0.5% in 2014 (Fig. 7d). These results were similar to the findings of Amiro et al. [24], who observed thinned plots in Arizona and California and reported that the GPP and NEP continued to increase after thinning. Although the Re slightly decreased in the year of thinning, it significantly increased in the second year after thinning [24]. The effects of thinning on the carbon sink function of forests are determined by the dynamic balance between the increased carbon sequestration capacity of the remaining stands and understory vegetation and the carbon emissions from the decomposition of stumps, litter, and woody debris [10, 12, 18]. In the long run, thinning benefits forest health and productivity. However, it also represents a significant disturbance to forest ecosystems, particularly in the years immediately following the intervention [16]. It is generally accepted that thinning entails a short-term loss of carbon sink capacity due to the reduction in the photosynthetic leaf area, yet it supports sustainable forest development [8]. Surprisingly, our study and several previous studies indicated an increase in the NEP even shortly after thinning [53]. Launiainen et al. [54] attributed the increase in NEP primarily to climatic effects. However, our study indicated that even when the effects of climatic factors were diminished, the NEP improved after thinning. The changes in NEP depended on GPP and Re, which represent opposing carbon fluxes; thus, it was difficult to predict the eventual changes in NEP. The GPP was assumed to decrease due to the removal of trees [15]. However, if the removed oppressed trees exhibited low photosynthetic activity, the enhanced photosynthesis of the remaining trees and the vigorous growth of understory vegetation could compensate for the reduction in photosynthetic capacity caused by tree removal [55–57]. Consequently, an increase in GPP is possible and is expected to become more pronounced over time [58]. Studies at this site verified the increase in photosynthesis at the leaf scale [20]. The continuous increase in Re in the first and second years after thinning is consistent with ecological expectations. In addition, this increase offset the gains in GPP, resulting in only a slight increase in NEP, particularly in 2014. However, the Re would probably increase for years, possibly at a faster rate in the short term. Subsequently, its increase will cease and may even decline over time as the supply of thinning-derived debris becomes depleted. Consequently, the carbon sink function changes from year to year after thinning. Thinning can increase forest vitality and maintain sustainable development in the long run, but its impact on carbon sink function, especially after years of thinning, is highly uncertain. Long-term studies across different regions, combined with leaf-scale observations of both canopy and understory photosynthesis as well as land surface respiration, could help quantify the effects of thinning and elucidate the underlying mechanisms in future research. Conclusions Thinning is an important way to improve forest structure and gain both economic and long-term ecological benefits. However, its effects on the carbon balance are controversial, especially in the short term after thinning. A deep and wide understanding of the effects of thinning is important for developing forest policies. In this study, 25% thinning in an approximately 30-year-old coniferous plantation with an initial stand density of 1329 stems ha−1 indicated that thinning did not impair the forest carbon sink function as common sense anticipated. According to previous studies, increases in GPP and NEP might be caused by the changing climate background [10]. To minimize potential confounding effects from climate variability and methodological estimation biases, three complementary analytical approaches were employed to validate the results. The consistent results obtained through the three methods confirmed that thinning in these near-mature forests increased the GPP and NEP. In addition, the Re increased, as expected. Furthermore, the increments in GPP and Re were greater in the second year than in the first year post-thinning. However, a much greater increase was found in Re than in GPP. Consequently, the increase in NEP was greater in the first year after thinning than in the second year after thinning. Climate change is projected to cause widespread forest loss and structural alterations across China in the future [59]. According to our in situ observations in the QYZ station, these plantations are suffering from natural and pest threats. The thinning and further reconstruction of forests may be a positive way to address this problem. In addition, according to our results, harvesting some of the trees would not lead to carbon sink loss, which would provide confidence for conducting thinning in southern China. This would address the dilemma of harvesting and maintaining a carbon sink. In combination with the findings of previous studies, thinning, forest reconstruction and afforestation might be strategies for enhancing future forest carbon sinks [59, 60]. This study provides a scientific foundation for modeling and predicting the effects of forest thinning and can serve as a reference for developing forest management policies in China. Supplementary Information Supplementary Material 1 Acknowledgements We thank the Data Center of Chinese Terrestrial Ecosystem Flux Observation and Research Network (ChinaFLUX) and the Data Center of Chinese Ecosystem Research Network (CERN) for providing data for this study. We thank Yuanfen Huang, Shanyuan Yin, and Jianzhong Zhang for conducting thinning in the field. Abbreviations Net ecosystem productivity Gross primary productivity Ecosystem respiration Eddy covariance technique Tower in the thinned plot Tower in the unthinned plot Carbon fluxes observed by Tower-th Carbon fluxes observed by Tower-un Daily average values of the carbon fluxes Single tower observation method Parallel 2-tower fitting method Parallel 2-tower fitting method considering the cropland contribution Look-up table method Average table established when conducting LUT method Carbon fluxes obtained based on the LUT method Simulated carbon fluxes when conducting 2-tower method: flux-sim represents simulated unthinned flux around Tower-th Re-simulated fluxes considering the cropland contribution in peak periods when conducting 2-towerP method: flux-simP represents simulated unthinned flux around Tower-th China Flux Observation and Research Network National Qianyanzhou Critical Zone Observatory of the Red Soil Hilly Region Air temperature Soil water content Photosynthetically active radiation Precipitation Friction velocity Canopy height Stability parameter Wind direction Author contributions M.J.X. and T.Z. conceived of the study idea. M.J.X. and S.T.L. designed the research. S.T.L., F.T.Y., X.L., X.J.Z., C.P.C. and J.L.W. collected and organized the data. T.Z., S.T.L., H.Z., Y.T.W., J.X.S. and Z.Y.W. designed and drew the graphs. M.J.X. and T.Z. performed the analysis. M.J.X. wrote the original manuscript. T.Z. and H.M.W. edited the manuscript. M.J.X., T.Z., S.T.L. and H.M.W. discussed and interpreted the results and contributed to the manuscript. All the authors contributed to revising and approving the final manuscript. Funding This work was funded by the National Natural Science Foundation of China [grant numbers 42175141, W2523030 and 32471642] and the Liaoning Province Science and Technology Plan Joint Program (Natural Science Foundation-General Project) [grant numbers 2024-MSLH-241, 2024-MSLH-235 and 2024-MSLH-504]. Data availability Data are owned by the ChinaFLUX and will be made available on request. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Mingjie Xu and Shengtong Li are co-first authors. They contributed equally to this work. Contributor Information Tao Zhang, Email: zhangt@syau.edu.cn. Huimin Wang, Email: wanghm@igsnrr.ac.cn. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement Data are owned by the ChinaFLUX and will be made available on request.