The study site was in Zhong County of Chongqing, China, which is located in the Three Gorges Reservoir area of the upper reaches of Yangtze River valley (107°32′~108°18′E, 30°03′~30°53′N) (Figure 1). This site has a subtropical monsoon climate, with a mean annual precipitation of 1,200 mm and a mean annual temperature of 18.2°C, which are particularly suitable for citrus cultivation. The region is the main citrus-producing area of China, and Zhong County is the core citrus production demonstration area. In 2020, the citrus plantation area in Zhong County was 23,600 ha, accounting for approximately 31% of the total arable land and 82.1% citrus planted on hilly areas in this county. Citrus annual production averaged 338 million tons with the output value of ~30 billion yuan RMB, thus contributing 6.15% to the total GDP of this country.

Various citrus varieties are cultivated in the study region, which include the early-maturing Daya mandarin (Citrus reticulata ‘Daya’) (DM), Harumi mandarin (Citrus reticulata ‘Harumi’) (HM), Lane Late Navel Orange (Citrus sinensis (L.) Osb.) (LLN), and Orah mandarin (Citrus reticulate ‘Orah’) (OM) and the late-maturing Ehime mandarin (Citrus reticulata ‘Ehime’) (EM) and Shatangju mandarin (Citrus reticulata ‘Shatangju’) (StjM) varieties. They were all used in this study. From 2017 to 2021, the Zhong county government encouraged farmers to implement different forms OSCF by offering free organic fertilizers and cover crop seeds, and thus various nutrient management strategies in citrus production can be found in this area.

In this study, the CF and WF of different citrus production systems were evaluated with a cradle-to-farm gate life cycle assessment methodology. The system boundary included agricultural material production and citrus production sections (Figure 2), and both GHG emissions (GE) and C sequestration were also considered. The GHG emissions included CO₂ (from input production and energy use), direct and indirect N₂O emissions (ammonia (NH₃), and the N losses in runoff and leaching), and CH₄ emissions from organic amendments. Carbon sequestration in soil and perennial biomass (from organic inputs, cover crops, no-tillage, and tree growth) was included as a negative emission. Excluded processes were post-farmgate activities such as fruit transport, processing, and distribution, as the study focuses on comparing on-farm management practices. Capital goods and infrastructure (e.g., machinery, irrigation systems, buildings) were also excluded, as their contribution to the annual footprint of perennial cropping systems is typically negligible compared with operational inputs.

In the system boundary of citrus production, the WF was estimated based on the consumption of rainfall (green WF) and irrigation water (blue WF), and the freshwater required to dilute contaminated water by total N and P losses from fertilizer (grey WF). The specific impact results of each management measure on each component of water footprint are shown in Figure 3.

The data sources included a 2-year county-scale survey and a 3-year citrus nutrient management experiment. The survey was conducted systematically through face-to-face meeting supported by a comprehensive questionnaire encompassing all aspects of citrus production in 2020 and 2021, covering 273 smallholder or large-scale citrus agribusinesses and 48% townships in Zhong County. All of these in-house surveys were conducted by professional research staff to ensure that the survey samples are representative. The questionnaire collected information on citrus management, including fertilizer application, pesticide application, diesel fuel and electricity consumption, fruit bagging, cover crop and tillage management, irrigation, citrus variety, yields, citrus tree annual growth biomass, farmer’s cost, and economic outcomes.

Following the different OSCF Policy strategies carried out in the orchard farms, the survey data was divided into four nutrient management strategies, namely, chemical fertilizer only (Che), chemical fertilizer plus organic fertilizer (Che+Org), chemical fertilizer plus cover cropping management (Che+CC), and chemical fertilizer plus combined organic material inputs (Che+Org+CC). Optimal fertilization plus cover crop (OPT), for which data were obtained from a 3-year orchard field experiment, was also included for analysis. The OPT mode aims to integrate the core principles of OSCF: reduction of chemical fertilizer use, precise application of organic fertilizers, and management of cover crop systems. In the OPT treatment, the total nitrogen input is 300 kg N per hectare, of which 220 kg N per hectare is chemical nitrogen fertilizer (accounting for 73.3%) and 80 kg N per hectare is organic nitrogen fertilizer (accounting for 26.7%), with the ratio of inorganic nitrogen to organic nitrogen being approximately 2.75:1. Compared with the standard OSCF policy recommended dosage, the physical input of organic fertilizer in the OPT treatment is numerically lower. This design is based on optimization principles rather than simply following the policy. Our field experiments aim to determine, while maintaining yield and enhancing carbon sequestration capacity, the minimum effective organic input quantity required to maximize fertilizer reduction. Therefore, OPT represents a refinement and local calibration of the OSCF policy, shifting the focus from the general recommended dosage of organic substitution to a precise nutrient-integrated management strategy. This experiment was carried out in a typical citrus-growing hilly orchard in Zhong country and aimed to define and recommend an optimal nutrient management strategy to farmers, which include the criteria such as chemical fertilizer reduction, organic fertilizer application, and cover crop management. The agricultural inputs and outputs data from survey and experiment are shown in Table 1.

This study adopted a volumetric approach for quantification of WF of citrus production (PWF, m³ t⁻¹), including those of green WF (WF_green), blue WF (WF_blue), and grey WF (WF_grey). WF_green and WF_blue represent the consumption of water use (CWU, m³ ha⁻¹) for citrus production and can be calculated as the following:

where ET_c, ET_green, and ET_blue represent, respectively, total evapotranspiration, green water evapotranspiration, and blue water evapotranspiration (m³ ha⁻¹) in citrus production annually. ET_c was calculated by the CROPWAT model, which is listed in Supplementary Figure S1; P_eff is the total effective rainfall used for citrus growth and was estimated and is shown in Supplementary Figure S1. The digital 10 is the conversion factor of water depth into water volume per unit area.

WF_grey represents water required diluting N and P pollution during citrus production and is mainly caused by the loss of fertilizers and pesticides. In this study, only loss of total N and P from fertilizer was accounted for, as the process of pesticides utility is difficult to detect. WF_grey can be calculated as the following:

where αN and α_P are the proportion of total N and P losses (by runoff and leaching) from chemical fertilizer. According to our previous research carried out in the same research region (Liu et al., 2021b), 46.22% N and 2.22% P were lost by runoff and leaching in clean tillage management, whereas 26.11% N and 1.37% P were lost in cropping management, respectively. AR is the chemical fertilizer application rate in kg ha⁻¹. C_N-max, C_N-min and C_P-max, C_P-min (kg m³) refer to the maximum holding capacity of N and P in the water body and the concentration in the natural state, respectively. According to China’s environmental quality standard of surface level V water body (GB3838-2002), the highest permissible concentrations for total N (C_N-max) and P (C_P-max) were 2 and 0.4 mg L⁻¹, and the minimum permissible concentrations for total N (C_N-min) and P (C_P-min) were 0 and 0 mg L⁻¹, respectively.

Finally, the total WF was calculated by the sum of WF_green, WF_blue, and WF_grey.

Data processing was performed using Microsoft Office Excel 2016 and ArcGIS 10.0. The survey and experiment data were divided into four groups according to average CF (FCF or PCF) and citrus yield, including the low yield and low CF level (LL), low yield and high CF level (LH), high yield and low CF level (HL), and high yield and high CF level (HH) groups. All statistical analyses were performed using SPSS 17.0, and figures were performed using SigmaPlot 12.5. Significant differences of the FCF, PCF, carbon efficiency, and WF among different nutrient management strategies were analyzed using one-way ANOVA and least significant difference (LSD). The levels of significance were defined at P < 0.05 (*) and P < 0.01 (**). Spearman correlations between each WF and CF fractions were calculated using SigmaPlot 12.5.

The resource inputs of different nutrient management strategies are shown in Table 1. The OPT group is a predefined integrated strategy aimed at utilizing localized fertilizer application amounts to fulfill the principles of the OSCF policy. Therefore, as its design objective indicates, the amount of chemical nitrogen and phosphorus fertilizers used for citrus production in the OPT group was less than that in the Che+CC group (P < 0.05) Due to the large variation in organic fertilizer inputs among farmers, there was no significant difference in organic nutrient inputs among the Che+Org, Che+Org+CC, and OPT groups (P < 0.05). The total N inputs of the Che+CC and Che+Org+CC groups were significantly greater than that of the OPT group (P < 0.05). In contrast, the OPT group consumed more fuel than the Che, Che+CC, and Che+Org+CC groups (P < 0.05). Although the chemical input was intentionally reduced, there was no significant difference in the citrus yield among the group. Moreover, the net economic benefit of farmers in the Che+CC, Che+Org+CC, and OPT groups were significantly greater than those in the Che and Che+Org groups (P < 0.05), demonstrating the feasibility of this design strategy.

The survey data of 273 farms and 3 years of optimal management experiment data were divided into LL, LH, HL, and HH levels based on CF (FCF or PCF) and yield (Figure 4). The average FCFs in the LL, HL, LH, and HH levels were 0.51, 0.47, 4.12, and 4.81 t CO₂ eq ha⁻¹, respectively (Figure 4a). The average PCFs in the LL, HL, LH, and HH levels were 0.01, 0.004, 0.34, and 0.33 t CO₂ eq t⁻¹, respectively (Figure 4b). Approximately 47.5%, 42.1%, and 30.0% farms treated with Che, Che+Org, and Che+CC were all attributed to LH level (Figure 4a). However, approximately 51.2% farms treated with Che+CC+Org were attributed to HL level (Figure 4a). Two out of three OPT experiment fields were also in the HL level (Figure 4a). Although there are differences in data sources and sample sizes, this comparison remains valid because both sets of data used the same local emission factors and system boundaries and conducted completely identical LCA calculations. Thus, it is possible to fairly assess the performance gap between current practices and the optimized schemes based on scientific design.

In the citrus field experiment of this study, the FCF ranged between 0.31 and 3.83 t CO₂ eq ha⁻¹, and the PCF between 0.02 and 0.34 t CO₂ eq t⁻¹ depending on the nutrient management treatment (Figure 5). The OPT showed significantly reduced FCF compared with the Che, Che+Org, and Che+CC groups (Figure 5a). This declining trend was more pronounced for PCF with OTP having the lowest value (Figure 5b). Also, the Che+CC and Che+Org+CC groups showed significant reductions in PCF compared with Che and Che+Org (Figure 5b).

Chemical nitrogen fertilizer use (GE_N) accounted for 70.71% of greenhouse gas (GHG) emissions on average, making it the primary CF source in citrus production (Figure 5). The use of organic fertilizer (GE_{Organic fertilizer}) and the use of phosphate fertilizer (GE_P2O5) were the second and third CF contributors, accounting for 14.34% and 6.18% of GHG emissions, respectively (Figure 4). GEK₂O, GE_Pesticides, GE_Fuel, GE_Electricity, and GE_Fruit bagging accounted for 3.21%, 2.33%, 2.10%, 0.86%, and 0.28% of GHG emissions, respectively (Figure 5a). The main CF mitigation factors in the citrus system were soil organic carbon sequestration by cover crops (CSC), organic fertilizer use (CS_Org), citrus tree pruning litter (CS_Litter), and no-tillage management (CS_No-tillage), and they accounted for 23.81%, 15.89%, 15.58%, and 7.21% of GHG emissions, respectively (Figure 5a). For GHG emissions per ton (citrus yield), nitrogen fertilizer, organic fertilizer, phosphate fertilizer, potassium fertilizer, electricity, pesticides, fuel oil, and paper bags contributed 70.22%, 14.58%, 6.10%, 3.18%, 2.48%, 2.24%, 0.91%, and 0.29%, respectively. Carbon sequestration by tree growth (CS_Tree) and no-till practice offset 23.67%, 16.74%, 15.98%, and 7.80% (Figure 5b).

There are significant differences between OPT management and farmers’ crop management practices, and it impacted both FCF and PCF (Figure 5). FCF was drastically diminished under OPT, with the reduction ranging from 76.34% to 91.91%, compared with the Che, Che+Org, Che+CC, and Che+Org+CC methods. In the OPT management, carbon emission was decreased by 3.18%, 15.23%, 24.86%, and 22.39% compared with Che, Che+Org, Che+CC, and Che+Org+CC, respectively, whereas OPT carbon sequestration was increased by 230.97% compared with Che. The PCF associated with cover crop management varied significantly between Che+CC, Che+Org+CC, and OPT compared with those without cover crop (Che, Che+Org). The PCF of OPT was 94.12%, 93.10%, 90.47%, and 60.00% lower than those of Che, Che+Org, Che+CC, and Che+Org+CC, respectively. Compared with PCF, the reduction in carbon emission of OPT was smaller, and it was 34.84%, 47.36%, 32.14%, and 27.67% lower than those of Che, Che+Org, Che+CC, and Che+Org+CC, respectively. However, carbon sequestration in the OPT group was increased by 121.71% and 18.21% compared with Che and Che+CC and reduced by 1.51% and 23.46% in Che+Org and Che+Org+CC, respectively.

From the perspective for FCF, there is no significant difference between different citrus varieties that were cultivated under similar management methods (Figure 6). Compared with the FCF of varieties OM, LLN, DM, HM, and EM, FCF in StjM was decreased by 27.69%, −5.79%, 38.34%, 1.95%, and 16.60%, respectively. Also, StjM’s carbon emission, compared with OM, LLN, DM, HM, and EM, was decreased by 5.08%, −12.80%, 7.14%, −1.00%, and 5.79%, and carbon sequestration was increased by 11.17% and 16.60%, respectively. The PCF of StjM was significantly different from that of other varieties. (Figure 6). Compared with OM, LLN, DM, HM, and EM, the PCF of StjM was reduced by 80.00%, 63.64%, 85.71%, 75.00%, and 77.77%, respectively. Carbon emission of StjM was 35.48%, 28.33%, 36.66%, 26.54%, and 25.48% lower than those of OM, LLN, DM, HM, and EM, respectively. In contrast, StjM’s carbon sequestration was increased by 9.48% and 19.73% compared with OM, LLN, DM, HM, and EM.

Orchards managed by OPT and Che+Org+CC achieved significantly higher carbon efficiency compared with other management methods (Figure 7a). The carbon efficiency of orchards managed by OPT was 8.00-, 8.86-, 6.98-, and 4.00-fold greater than those of Che, Che+Org, Che+CC, and Che+Org+CC, respectively. Among the varieties studied, StjM had the highest carbon efficiency (22.23 × 104 Yuan t⁻¹ CO₂eq) and it was 1.93, 3.05, 3.50, 11.00, and 19.16 times higher than those of EM, DM, HM, LLN, and OM, respectively (Figure 7b).

Greenhouse gas emissions from agricultural operations in citrus cultivation include CO₂ emissions from agricultural material production, direct N₂O emissions from N₂O escape caused by nitrogen fertilizer use, indirect N₂O produced from NH₃ emission and NH₄ leaching, and methane emissions from organic fertilizer application. In our experiments, CO₂ caused the most significant greenhouse effect, followed by the direct and indirect emissions of N₂O and CH₄ (Figure 8). Che, Che+Org, Che+CC, Che+Org+CC, and OPT account for 66.89%, 66.27%, 66.66%, 67.21%, and 67.21%, respectively, of the greenhouse effect caused by CO₂ emissions. Direct N₂O was estimated to contribute 24.97%, 21.86%, 25.14%, 23.17%, and 22.03% greenhouse effect in Che, Che+Org, Che+CC, Che+Org+CC, and OPT, respectively. The greenhouse effect caused by indirect N₂O was 8.14%, 10.44%, 8.20%, 8.20%, and 7.83% for Che, Che+Org, Che+CC, Che+Org+CC, and OPT, respectively. Methane emitted by Che+Org, Che+Org+CC, and OPT account for 1.43%, 1.42%, and 1.32% of greenhouse effect in Che+Org, Che+Org+CC, and OPT, respectively.

Orchards under OPT management reduced CO₂ and N₂O emissions and increased carbon sequestration, resulting in much lower FCF and PCF compared with other management methods evaluated in this study (Figures 9, 10). Fertilizer is still the primary source of greenhouse gas emissions in FCF. The greenhouse effect of CO₂ emitted from fertilizer production and application ranged from 3.08 to 4.16 t CO₂eq ha⁻¹, accounting for 89.06% to 92.11%. Additionally, N₂O and methane emissions are also associated with fertilizer application (Figure 9). Similarly, fertilizer remains the primary source of greenhouse gas emissions in PCF. The greenhouse effect of CO₂ emitted from fertilizer production and application ranged from 0.19 to 0.35 t CO₂eq ha⁻¹. N₂O and methane emissions were also attributed to fertilizer application (Figure 10).

As shown in Figure 3, there was no significant difference for blue and green water footprints among the management models tested here. Compared with Che and Che+Org, OPT’s blue water footprint was reduced by 4.63% and 13.10%, respectively. The green water footprint of OPT was reduced by 29.39% and 35.66%, respectively, when compared with Che and Che+Org.

The grey water footprint crop cover management models (Che+CC, Che+Org+CC, and OPT) is significantly different from those without cover crop (Che and Che+Org). Crops in OPT had their grey water footprint decreased by 71.58%, 67.98%, 50.33%, and 28.60%, respectively, compared with Che, Che+Org, Che+CC, and Che+Org+CC. The overall water footprint of OPT was reduced by 67.98%, 64.62%, 45.97%, and 26.94%, respectively, when compared with Che, Che+Org, Che+CC, and Che+Org+CC.

As shown in Figure 11, PCF is negatively correlated with the blue and green water footprints; that is, the higher the PCF, the smaller the blue and green water footprints. An opposite trend was evident for PCF and the grey water footprint and overall water footprint.

In 2020, China’s agricultural carbon emissions amounted to 856 million tons of CO₂eq (FAO, 2022). Among them, the fruit industry, especially citrus production, has become a significant contributor to GHG emissions due to its extensive planting area and high resource input. Its carbon emission intensity is notably higher than the global average (Lin et al., 2015; Yan et al., 2016).

Fertilizers, especially nitrogen fertilizers, are the primary source of CF in agriculture. This study is consistent with the conclusions of most studies, confirming that the production and application of chemical fertilizers are the main contributors to CF, with a contribution rate ranging from 68.16% to 94.49%. Among them, the emission of nitrogen fertilizer accounts for as high as 88.28% (Chen et al., 2020; Yan et al., 2016). Previous studies have also identified a positive correlation between nitrogen fertilizer input and carbon emissions (Chen et al., 2020). Nitrogen use efficiency in Chinese agriculture is relatively low, with nitrogen fertilizer application rates practiced by farmers, including for citrus, being excessively high. Additionally, nitrogen fertilizer partial productivity is only approximately 40% of that of developed countries (Chen et al., 2020; Yang et al., 2020). Therefore, the primary focus for reducing carbon emissions in orchards should be on decreasing nitrogen use and enhancing nitrogen use efficiency. From our study, it is evident that even if the amount of nitrogen applied is greatly reduced, the fruit yield can be maintained at a high level. In general, a reduction of nitrogen input by 30.0% can result in a reduction of greenhouse gas emissions by 15.0% to 24.0% (Yan et al., 2016). The OPT model in this study has been verified in practice to achieve significant emission reduction while maintaining high yields through optimized management.

The use of organic fertilizers significantly reduced the farm carbon footprint (FCF) and product carbon footprint (PCF), which is consistent with the findings of Ribal et al. (2016) and Aguilera et al. (2014). However, our FCF and PCF values were notably higher than those reported by Martin-Gorriz et al. (2020). Through the research conducted by Martin-Gorriz et al. (2020), it provides a suitable reference for analyzing the mechanism of the impact of fertilization management on carbon footprint. Because this discrepancy primarily stems from differences in nitrogen input level, the lemon, citrus, and orange trees in Martin-Gorriz et al. (2020) received substantially lower nitrogen fertilization rates than those used in our study. Our results further demonstrate that chemical fertilizers constitute the primary source of agricultural carbon emissions, with a positive correlation observed between nitrogen application rates and carbon emissions. Consequently, the variance in fertilization rates became the key determinant explaining the higher PCF and FCF values in our study relative to that of Martin-Gorriz et al. (2020).

Apart from chemical fertilizers, organic fertilizers are a significant source of GHG in orchards. However, organic fertilizers can also increase soil organic carbon in orchards through carbon sequestration, which helps alleviate GHG. In this study, the carbon sink benefit of organic fertilizer outweighed its carbon emission attribute. The addition of 10 t ha⁻¹ (dry weight) of organic fertilizer can significantly enhance the carbon sequestration rate (Aguilera et al., 2013). Zhou et al. (2022) reported that increasing the application ratio of organic fertilizer to chemical fertilizer enhances the utilization efficiency of the latter, leading to improved yield and quality of citrus.

Previous studies on CFs have primarily focused on GHG emissions caused by agricultural production activities. These include direct GHG emissions from the production of agricultural products such as fertilizers (Chen et al., 2015; Li et al., 2016; Clark et al., 2016; Pishgar-Komleh et al., 2013), direct greenhouse gas emissions from fertilizer application (Gu et al., 2019; Zhang et al., 2017), and indirect greenhouse gas emissions from soil denitrification (Zhang et al., 2017; Liu et al., 2021a; Kang et al., 2022). However, there are limited studies on carbon sequestration, which includes the CO₂ absorbed through photosynthesis during crop growth and development (Wu, 2008), as well as the increase in SOC resulting from practices like cover crops, organic fertilizers (Aguilera et al., 2013; Yang et al., 2020), and no tillage to reduce SOC decomposition (Lu et al., 2009). Cover crops are an important component of soil carbon sequestration and the most significant component that mitigates CF. Carbon input from cover crops is transferred from the atmosphere to the soil through photosynthesis, increasing SOC (Blanco-Canqui, 2017). Additionally, they reduce runoff losses and soil erosion in steep-slope orchards significantly (Ledo et al., 2020; Liu et al., 2021a). This practice can decrease the amount of both soil carbon and nitrogen leaving the orchard system through runoff, ultimately reducing orchard CF (Repullo-Ruibérriz de Torres et al., 2018).

The study area is a typical hilly landform experiencing serious soil erosion and low fertilizer utilization. Among the issues, a low nitrogen utilization efficiency leads to a high carbon footprint and a high grey water footprint. The greenhouse effect, as well as the eutrophication of rivers and lakes, is largely caused by the excessive use of nitrogen fertilizers (Liu et al., 2013). The loss of chemical nitrogen and phosphate fertilizers causes grey water footprint, and thus reducing the grey water footprint indirectly promotes the utilization of fertilizers. Moreover, because the research results (Figure 11) show that the carbon footprint (CF) is significantly positively correlated with the grey water footprint and the overall water footprint, reducing the grey water footprint indirectly reduces the carbon footprint.

Secondly, cover crops are a significant contributor to the reduction in the grey water footprint, as they substantially reduce the loss of nitrogen and phosphorus, as much as 53.4% and 56.9%, respectively, as reported in a previous study (Liu et al., 2021a). The grey water footprint varies significantly based on management practices, and it was significantly lower in cover-cropped management models (Che+CC, Che+Org+CC, and OPT) than those without cover crop (Che and Che+Org). Additionally, OPT-managed orchards have a significantly lower grey water footprint compared with the models using farmer practice. The overall water footprint (WF) under OPT management ranged between 32.02% and 73.06%. In addition, cover crops increased the green water footprint, but the increase was smaller than the decrease in grey water footprint. The overall water footprint of OPT was significantly lower than that of other models tested.

In different nutrient management models studied here, both CFs and the overall water footprint were influenced by environmental factors and farming practices. As shown in Figure 3, there was little difference in the blue water and green water footprints across different management models. Blue water and green water are mainly affected by local climatic conditions and crop varieties. The CF is significantly positively correlated with the grey water footprint and the overall water footprint. The grey water foot was however the most important factor affecting the CFs of citrus production. By reducing application of chemical fertilizers and cover crop management between tree rows, grey water footprint can be greatly reduced, thereby reducing the CFs of citrus production.

Organic fertilization and cover crop management practices have emerged as promising strategies to reduce GHG emissions and mitigate global warming. In this context, we studied five distinct nutrient management methods (Che, Che+Org, Che+CC, Che+Org+CC, and OPT) and found that the OPT method performed the best in terms of environmental and economic benefits. The fundamental reason for this is that OPT does not simply add organic inputs or cover crops; instead, it achieves a synergistic enhancement of nutrient cycling and carbon sequestration functions through systematic optimization. (Figure 5).

The results of our study indicate that OPT can reduce FCF to 0.31 t CO₂ eq ha⁻¹ and PCF to 0.02 t CO₂ eq ha⁻¹. In comparison with other studies, our study demonstrates relatively lower CFs of citrus production though with notable disparities (Table 4). These differences can be attributed to several factors: Firstly, our approach involved not only reducing the application of chemical fertilizers but also increasing investment in organic fertilizers and implementing cover crop management. Secondly, while other studies typically employ IPCC-recommended values for carbon emission parameters, our study optimized these parameters specifically for citrus production, incorporating information obtained from local experiments. This refinement enhances the accuracy of carbon emissions estimates for Chinese citrus production. For example, the estimated NH₃ volatilization (EFChe-NH₃) and N leaching (EFChe-RL) in dryland apple orchards in northern China were 10.8% (Ge et al., 2011) and 10.0% of total chemical N fertilizer inputs, whereas the emission parameter EFChe-N2O obtained by field tests in the study area is 13.38% (Kang et al., 2022). EFChe-RL was 46.22% without cover crops and 26.11% with cover crops (Liu et al., 2021a). Moreover, while our study fully considered the role of carbon sequestration, some studies (Ribal et al., 2016; Maestre-Valero et al., 2018; Mazis et al., 2021) neglect to account for greenhouse gas offset by carbon sequestration, resulting in higher CF values than what was observed in our study. Additionally, other studies’ boundary systems (Maestre-Valero et al., 2018) encompass CF resulting from irrigation, a factor not included in our study as there was no irrigation in our production system, thereby contributing to difference in CFs. The citrus boundary system studied by Yan et al. (2016) includes CF influenced by irrigation, leading to a slightly larger CF than ours. Furthermore, variations in fertilizer quantities, as observed in other studies (Chen et al., 2020; Maestre-Valero et al., 2018; Ribal et al., 2019), lead to differing CF levels. Additionally, the citrus boundary system examined by Li et al. (2021) encompasses five links, four more than in our study: post-harvest processing, storage, transportation, and consumption, resulting in significantly higher CF values than our findings. Lastly, the difference in climate can also influence CF levels (Aguilera et al., 2014). For instance, in a study in the Mediterranean climate region (Martin-Gorriz et al., 2020), the emphasis on reducing CO₂ emissions resulted in the system functioning as an ideal carbon sink, with protective measures implemented over 7 years leading to an optimal CF state.

Among the 273 smallholder or large-scale citrus agribusinesses examined in this study (Che, Che+CC, Che+Org, and Che+Org+CC), both FCF and PCF were markedly higher than those observed in the 2-year OPT trial. This difference primarily arises from the intensive management practices involving high fertilizer and pesticide inputs prevalent in the study region. This stands in stark contrast to the findings of Zhang et al. (2022) on grain crop production in plain areas of China, where large-scale agriculture reduces the trade-offs between crop productivity and resource input, enhancing energy efficiency and achieving the co-benefits of high crop productivity and low energy consumption. The citrus cash crops in our study area predominantly grow in relatively steep hilly and mountainous terrains, facing challenges such as limited mechanization and higher labor cost. Farmers or large landowners typically rely on high resource inputs to attain high yields and economic returns. Knudsen’s (2011) and Beccali’s (2009) studies further corroborate that GHG emissions from citrus orchards used for fruit juice production with low intensification are lower compared with those from orchards employing intensive agricultural practices to increase fruit yield. Hence, to foster green and low-carbon development of citrus production in our study area, it is imperative to adopt low-intensity OPT management practices.

In our study, there was no significant difference in FCF among different citrus varieties, yet substantial disparities in PCF were evident. StjM and LLN varieties exhibited lower PCF compared with other varieties, primarily due to their higher yields. Therefore, cultivating varieties with relatively high yields such as StjM and LLN in this region while phasing out varieties with high carbon emissions proves to be an effective strategy for promoting the green development of regional citrus.

Although our research provides strong evidence for the benefits of integrated management in mountain citrus orchards, some limitations must also be acknowledged. Firstly, the CF and WF evaluations rely on the combination of farmer survey data and controlled experimental data from OPT treatments. Although we applied unified emission factors and system boundaries, the differences in management precision, tree age, and micro-topography among the surveyed farms, when directly compared with the optimization experiments, may introduce some uncertainties. Secondly, our carbon sink calculation uses conversion coefficients based on meta-analysis of organic inputs and cover crops. Although these values are derived from regional or global studies, long-term measurements of SOC dynamics specific to the location will further refine the accuracy of carbon sink quantification. Thirdly, this study focuses on the farm gate. Incorporating downstream processes (such as transportation and processing) into the full life cycle assessment can provide a more comprehensive view of the environmental impact of citrus production, although this exceeds the scope of this farm management comparison. Finally, the economic analysis considers direct costs and benefits. A more detailed cost–benefit analysis, covering labor input for cover crop management and potential long-term soil health benefits, will strengthen its practicality assessment.