Recent climate warming has significantly affected the sensitivity of Gross Primary Productivity (GPP) to precipitation within China\'s agricultural ecosystems. Nonetheless, the spatial and temporal nonlinear evolution patterns of GPP-precipitation sensitivity under climate change, as well as the underlying drivers and long-term trends of this sensitivity, are not well understood. This study employs correlation analysis to quantify the sensitivity between GPP and precipitation in China\'s agricultural ecosystems, and utilizes nonlinear detection algorithms to examine the long-term changes in this sensitivity. Advanced machine learning techniques and frameworks are subsequently applied to analyze the driving factors of GPP-precipitation sensitivity in China\'s agricultural ecosystems. The findings reveal that approximately 49.00 % of the analyzed pixels exhibit a significant positive correlation between GPP and precipitation. Nonlinear change analysis indicates spatial heterogeneity in GPP-precipitation sensitivity across China\'s agricultural ecosystems, with patterns showing initial increases followed by decreases accounting for 25.12 %, and patterns of initial decreases followed by increases at 13.27 %. Machine learning analysis identifies temperature, soil moisture, and crop water footprint as the primary factors influencing GPP-precipitation sensitivity in agricultural ecosystems. This study is the first to introduce crop water footprint as a significant factor in the analysis of GPP-precipitation sensitivity. It not only offers new insights into the temporal nonlinear changes and driving factors of GPP-precipitation sensitivity but also underscores the importance of enhancing agricultural water efficiency to maintain agricultural ecosystem health and ensure food security under climate change.

The dune ecosystem plays a significant role in the global carbon cycle. The Horqin Sandy Land is a typical semi-arid fragile ecosystem in northern China. Understanding the magnitudes and dynamics of carbon dioxide fluxes within this region is essential for understanding the carbon balance. Used 6 years (2013-2018) measurements from an eddy-covariance system, we analyzed the dynamic patterns of net ecosystem carbon exchange (NEE), gross primary production (GPP), and ecosystem respiration (Reco) of the dune ecosystem in Horqin Sandy Land and examined their responses to climate factors with a focus on the precipitation. The results showed that the NEE of the dune ecosystem fluctuated from -166 to 100 gCO2·m-2·year-1 across the 6 growing seasons, with an average of -56 gCO2·m-2·year-1. The precipitation was not a key factor influencing the carbon flux variability. During the mid-growth stage, GPP was primarily affected by the effective precipitation frequency (R2 ranging from 0.65 to 0.85, P < 0.05), followed by fractional vegetation cover (R2 ranging from 0.65 to 0.68, P < 0.05). However, in the early and late growth stages, temperature predominantly drove the carbon flux (R2 = 0.75, P < 0.01). The interannual variability of carbon flux can be predominantly elucidated by phenological indicators such as CO2 uptake (CUstart), end of CO2 uptake (CUend), CO2 uptake period (CUP), and Spring lag. The results demonstrated the dune ecosystem is a weak carbon sink in semi-arid ecosystems. Furthermore, we emphasized the significance of effective precipitation frequency in regulating carbon fluxes. Our results provide a foundational understanding of the carbon balance in semi-arid ecosystems.

Carbon dioxide sequestration from the atmosphere is commonly assessed using the eddy covariance method. Its net flux signal can be decomposed into gross primary production and ecosystem respiration components, but these have seldom been tested against independent methods. In addition, eddy covariance lacks the ability to partition carbon sequestration among individual trees or species within mixed forests. Therefore, we compared gross primary production from eddy covariance versus an independent method based on sap flow and water-use efficiency, as measured by the tissue heat balance method and δ13C of phloem contents, respectively. The latter measurements were conducted on individual trees throughout a growing season in a mixed broadleaf forest dominated by three tree species, namely English oak, narrow-leaved ash and common hornbeam (Quercus robur L., Fraxinus angustifolia Vahl, and Carpinus betulus L., respectively). In this context, we applied an alternative ecophysiological method aimed at verifying the accuracy of a state-of-the-art eddy covariance system while also offering a solution to the partitioning problem. We observed strong agreement in the ecosystem gross primary production estimates (R2 = 0.56; P < 0.0001), with correlation being especially high and nearly on the 1:1 line in the period before the end of July (R2 = 0.85; P < 0.0001). After this period, the estimates of gross primary production began to diverge. Possible reasons for the divergence are discussed, focusing especially on phenology and the limitation of the isotopic data. English oak showed the highest per-tree daily photosynthetic rates among tree species, but the smaller, more abundant common hornbeam contributed most to the stand-level summation, especially early in the spring. These findings provide a rigorous test of the methods and the species-level photosynthesis offers avenues for enhancing forest management aimed at carbon sequestration.

Chlorophyll fluorescence is the long-wave light released by the residual energy absorbed by vegetation after photosynthesis and dissipation, which can directly and non-destructively reflect the photosynthetic state of plants from the perspective of the mechanism of photosynthetic process. Moso bamboo has a substantial carbon sequestration ability, and leaf-expansion stage is an important phenological period for carbon sequestration. Gross primary production (GPP) is a key parameter reflecting vegetation carbon sequestration process. However, the ability of chlorophyll fluorescence in moso bamboo to explain GPP changes is unclear. The research area of this study is located in the bamboo forest near the flux station of Anji County, Zhejiang Province, where an observation tower is built to monitor the carbon flux and meteorological change of bamboo forest. The chlorophyll fluorescence physiological parameters (Fp) and fluorescence yield (Fy) indices were measured and calculated for the leaves of newborn moso bamboo (I Du bamboo) and the old leaves of 4- to 5-year-old moso bamboo (Ⅲ Du bamboo) during the leaf-expansion stage. The chlorophyll fluorescence in response to the environment and its effect on carbon flux were analyzed. The results showed that: Fv/Fm, Y(II) and α of Ⅰ Du bamboo gradually increased, while Ⅲ Du bamboo gradually decreased, and FYint and FY687/FY738 of Ⅰ Du bamboo were higher than those of Ⅲ Du bamboo; moso bamboo was sensitive to changes in air temperature(Ta), relative humidity(RH), water vapor pressure(E), soil temperature(ST) and soil water content (SWC), the Fy indices of the upper, middle and lower layers were significantly correlated with Ta, E and ST; single or multiple vegetation indices were able to estimate the fluorescence yield indices well (all with R2 greater than 0.77); chlorophyll fluorescence (Fp and Fy indices) of Ⅰ Du bamboo and Ⅲ Du bamboo could explain 74.4% and 72.7% of the GPP variation, respectively; chlorophyll fluorescence and normalized differential vegetation index of the canopy (NDVIc) could estimate GPP well using random forest (Ⅰ Du bamboo: r = 0.929, RMSE = 0.069 g C·m-2; Ⅲ Du bamboo: r = 0.899, RMSE = 0.134 g C·m-2). The results of this study show that chlorophyll fluorescence can provide a basis for judging the response of moso bamboo to environmental changes and can well explain GPP. This study has important scientific significance for evaluating the potential mechanisms of growth, stress feedback and photosynthetic carbon sequestration of bamboo.

The high uncertainty regarding global gross primary production (GPP) remains unresolved. This study explored the relationships between phenology, physiology, and annual GPP to provide viable alternatives for accurate estimation. A statistical model of integrated phenology and physiology (SMIPP) was developed using GPP data from 145 FLUXNET sites to estimate the annual GPP for various vegetation types. By employing the SMIPP model driven by satellite-derived datasets of the global carbon uptake period (CUP) and maximal carbon uptake capacity (GPPmax), the global annual GPP was estimated for the period from 2001 to 2018. The results demonstrated that the SMIPP model accurately predicted annual GPP, with relative root mean square error values ranging from 11.20 to 19.29% for forest types and 20.49-35.71% for non-forest types. However, wetlands, shrublands, and evergreen forests exhibited relatively low accuracies. The average, trend, and interannual variation of global GPP during 2001-2018 were 132.6 Pg C yr-1, 0.25 Pg C yr-2, and 1.57 Pg C yr-1, respectively. They were within the ranges estimated in other global GPP products. Sensitivity analysis revealed that GPPmax had comparable effects to CUP in high-latitude regions but significantly greater impacts at the global scale, with sensitivity coefficients of 0.85 ± 0.23 for GPPmax and 0.46 ± 0.28 for CUP. This study provides a simple and practical method for estimating global annual GPP and highlights the influence of GPPmax and CUP on global-scale annual GPP.

Stomatal conductance (gs) and compensatory water uptake (CWU) are crucial processes in land surface models, as they directly influence the exchange of carbon and water fluxes between terrestrial ecosystems and the atmosphere. In this study, we integrated a new stomatal scheme derived from optimal stomatal theory (Medlyn\'s gs model), and an empirical CWU scheme into the Common Land Model (CoLM). Assessing the impacts on modeling gross primary productivity (GPP) and latent flux (LE) through observations obtained from eddy covariance (EC) measurements at three forest sites in China. Our results show that replacing the Ball-Berry\'s gs model (termed BB) with Medlyn\'s gs model (termed MED) did not bring about significant changes (had neutral impacts) in the performance of CoLM simulations at three forest sites. Considering the climate factors of annual mean precipitation to optimize key fitting parameters in gs exhibited improvement in model simulations. The average coefficient of determination (R2) achieved to 0.65 for GPP and LE at three sites, and the normalized root mean squared error (NRMSE) decreased from 0.83 to 0.77 at those sites. Besides, incorporating CWU into the model improved its performance. The R2 increased to 0.84 and RMSE decreased to 4.84 μmol m-2 s-1 for GPP, and the R2 increased to 0.62 and RMSE decreased to 55.64 W m-2 for LE. Therefore, modifying the model process of both contributed more to enhancing the model simulations than relying solely on one of these functions. Our study highlights that the response of plant functional types (PFTs) to water stress can be effectively represented in gs models when coupled with biochemical capacity to quantify carbon and water fluxes in forest ecosystems or other ecosystems.

Semi-arid ecosystems have been shown to dominate over tropical forests in determining the trend and interannual variability of land carbon (C) sink. However, the magnitude and variability of ecosystem C balance remain largely uncertain for temperate semi-arid shrublands at the decadal scale. Using eddy-covariance and micro-meteorological measurements, we quantified the interannual variation in net ecosystem production (NEP) and its components, gross primary production (GPP) and ecosystem respiration (Reco, i.e., the sum of autotrophic and heterotrophic respiration), in a semi-arid shrubland of the Mu Us Desert, northern China during 2012-2022. This shrubland was an overall weak C sink over the 11 years (NEP = 12 ± 46 g C m-2 yr-1, mean ± SD). Annual NEP ranged from -66 to 77 g C m-2 yr-1, with the ecosystem frequently switching between being an annual C sink and a C source. GPP was twice as sensitive as Reco to prolonged dry seasons, leading to a close negative relationship between annual NEP and dry-season length (R2 = 0.80, P < 0.01). Annual GPP (R2 = 0.51, P = 0.01) and NEP (R2 = 0.58, P < 0.01) were positively correlated with annual rainfall. Negative annual NEP (the ecosystem being a C source) tended to occur when the dry season exceeded 50 d yr-1 or rainfall dropped below 280 mm yr-1. Increases in dry-season length strengthened the effects of low soil moisture relative to high vapor pressure deficit in constraining NEP. Both GPP and NEP were more closely correlated with C uptake amplitude (annual maximum daily values) than with C uptake period. These findings indicate that dry-season extension under climate change may reduce the long-term C sequestration in semi-arid shrublands. Plant species adapted to prolonged dry seasons should be used in ecosystem restoration in the studied area to enhance ecosystem functions.

Gross primary production (GPP) is a critical component of the global carbon cycle and plays a significant role in the terrestrial carbon budget. The impact of environmental factors on GPP can occur through both direct (by influencing photosynthetic efficiency) and indirect (through the modulation of vegetation structure) pathways, but the extent to which these mechanisms contribute has been seldom quantified. In this study, we used structural equation modeling and observations from the FLUXNET network to investigate the direct and indirect effects of environmental factors on terrestrial ecosystem GPP at multiple temporal scales. We found that canopy structure, represented by leaf area index (LAI), is a crucial intermediate factor in the GPP response to environmental drivers. Environmental factors affect GPP indirectly by altering canopy structure, and the relative proportion of indirect effects decreased with increasing LAI. The study also identified different effects of environmental factors on GPP across time scales. At the half-hourly time scale, radiation was the primary driver of GPP. In contrast, the influences of temperature and vapor pressure deficit took on greater prominence at longer time scales. About half of the total effect of temperature on GPP was indirect through the regulation of canopy structure, and the indirect effect increased with increasing time scale (GPPNT-based models: 0.135 (half-hourly) vs. 0.171 (daily) vs. 0.189 (weekly) vs. 0.217 (monthly); GPPDT-based models: 0.139 vs. 0.170 vs. 0.187 vs. 0.215; all values were reported in gC m-2 d-1 °C-1, P < 0.001); while the indirect effect of radiation on GPP was comparatively lower, accounting for less than a quarter of the total effect. Furthermore, we observed a direct, negative-to-positive impact of precipitation on GPP across timescales. These findings provide crucial information on the interplay between environmental factors and LAI on GPP and enable a deeper understanding of the driving mechanisms of GPP.

Mangrove ecosystems, as an important component of \"Blue Carbon\", play a curial role on global carbon cycling; however, the lack of the global estimates of mangrove ecosystem gross primary production (GPP) and the underlying environmental controls on its estimation remain a gap in knowledge. In this study, we utilized global mangrove eddy covariance data and applied Gaussian Process Regression (GPR) to estimate GPP for global mangrove ecosystems, aiming to elucidate the factors influencing these estimates. The optimal GPR achieved favorable estimation performance through cross-validation (R2 = 0.90, RMSE = 0.92 gC/m2/day, WI = 0.86). Over the study period, the globally annual averaged GPP was 2054.53 ± 38.51 gC/m2/yr (comparable to that of evergreen broadleaf forests and exceeds the GPP of most other plant function types), amounting to a total of 304.82 ± 7.71TgC/yr, hotspots exceeding 3000 gC/m2/yr observed near the equator. The analysis revealed a decline in global mangrove GPP during 1996-2020 of -0.89 TgC/yr. Human activities (changes in mangrove cover area) played a relatively consistent role in contributing to this decrease. Conversely, variations in external environmental conditions showed distinct inter-annual differences in their impact. The spatio-temporal distribution patterns of mangrove ecosystems GPP (e.g., the bimodal annual pattern, latitudinal gradients, etc.) demonstrated the regulatory influence of external environmental conditions on GPP estimates. The model ensemble attribution analysis indicated that the fraction of absorbed photosynthetically active radiation exerted the dominant control on GPP estimations, while temperature, salinity, and humidity acted as secondary constraints. The findings of this study provide valuable insights for monitoring, modeling, and managing mangrove ecosystems GPP; and underscore the critical role of mangroves in global carbon sequestration. By quantifying the influences of environmental factors, we enhance our understanding of mangrove carbon cycling estimates, thereby helping sustain of these disproportionately productive ecosystems.

Understanding the controlling mechanisms of soil properties on ecosystem productivity is essential for sustaining productivity and increasing resilience under a changing climate. Here we investigate the control of topsoil depth (e.g., A horizons) on long-term ecosystem productivity. We used nationwide observations (n = 2401) of topsoil depth and multiple scaled datasets of gross primary productivity (GPP) for five ecosystems (cropland, forest, grassland, pasture, shrubland) over 36 years (1986-2021) across the conterminous USA. The relationship between topsoil depth and GPP is primarily associated with water availability, which is particularly significant in arid regions under grassland, shrubland, and cropland (r = .37, .32, .15, respectively, p < .0001). For every 10 cm increase in topsoil depth, the GPP increased by 114 to 128 g C m-2  year-1 in arid regions (r = .33 and .45, p < .0001). Paired comparison of relatively shallow and deep topsoils while holding other variables (climate, vegetation, parent material, soil type) constant showed that the positive control of topsoil depth on GPP occurred primarily in cropland (0.73, confidence interval of 0.57-0.84) and shrubland (0.75, confidence interval of 0.40-0.94). The GPP difference between deep and shallow topsoils was small and not statistically significant. Despite the positive control of topsoil depth on productivity in arid regions, its contribution (coefficients: .09-.33) was similar to that of heat (coefficients: .06-.39) but less than that of water (coefficients: .07-.87). The resilience of ecosystem productivity to climate extremes varied in different ecosystems and climatic regions. Deeper topsoils increased stability and decreased the variability of GPP under climate extremes in most ecosystems, especially in shrubland and grassland. The conservation of topsoil in arid regions and improvements of soil depth representation and moisture-retention mechanisms are critical for carbon-sequestration ecosystem services under a changing climate. These findings and relationships should also be included in Earth system models.