A field experiment was set up in 2016 in Boming tea plantation, Anji County, Zhejiang Province, China. The soil pH was 4.48 and the contents of organic matter, total N, available P, available K and available Mg were 14.36 mg g^-1, 0.86 mg g^-1, 52 mg kg^-1, 131 mg kg^-1 and 30 mg kg^-1, respectively. The tea cultivar was Baiye-1 and the plants were 20 years old. Before the start of field experiment, tea plants received common compound fertilizer (N-P₂O₅-K₂O = 15-15-15) together with rapeseed cake manure (N-P₂O₅-K₂O = 5.8-2.7-1.5), a practice widespread in tea plantations of China according to a previous survey (Ni et al., 2019). This practice is hereafter referred to as farmers’ practice (FP) and was compared to three optimized schemes (treatments) ( Table 1 ). In the first scheme referred to as RSM, the chemical compound fertilizers were fully substituted by rapeseed cake manure (RSM). In the second scheme referred to as SCF+RSM, the common compound fertilizer of equal nutrient formular was replaced by a specially formulated compound fertilizer (N-P₂O₅-K₂O-MgO = 18-8-12-2) together with organic manure rapeseed cake and urea. In the third scheme referred to as SCF+LSM, the specially formulated compound fertilizer (SCF) was applied together with commercial livestock manure (LSM) and urea. The treatment referred to as HCF received high rates of fertilizers (as common compound fertilizer, rapeseed cake manure and urea) to investigate the effect of over-application of fertilizers (Ni et al., 2019). The rates of fertilizers in treatments are presented in Table 1 . RSM, SCF+RSM and RSM+LSM had the same total amounts of nutrient N, P and K, which were reduced by 25% compared to that of FP ( Table 1 ). Compared to FP, N rate was increased (by 5.7%) in RSM, decreased in SCF+RSM (by 4.2%) and in SCF+LSM (by 19.5%). The ratio of N-P-K fertilizers was optimized to be close to their ratios in young shoots. The ratio was decreased in RSM, SCF+RSM and HCF whereas that of K was slightly increased in SCF+LSM compared to that of FP ( Table 1 ). Organic manures and compound fertilizers were applied in October. Urea in SCF+RSM, SCF+LSM and HCF was separately into two applications with equal amounts, one in early February and the second in the end of April after the harvest of spring tea. Fertilizers were applied to furrows between rows followed by covering with soil. There were totally six treatments including a control (CK) without any fertilizers. The plot area was 20 m² and plots were randomly arranged in the field. Each treatment was replicated for three times. All other field management was the same as in the FP.

Young spring shoots were harvested by hand and were weighed. The weights of each harvest were summed as the annual fresh yield. Yield data of 2018-2020 were presented in the present work whereas those of 2017 was not included to eliminate the turnover effect from farmer’s practice to the fertilization treatments. Samples of young shoots were collected to measure the concentrations of nutrients in harvested teas. The concentration of N in teas and mature leaf samples was determined by an elemental analyzer (Vario Macro Cube, Elementar Analysensysteme GmbH, Langenselbold, Germany) and those of P and K were determined by Inductive Coupled Plasma-Atomic Emission Spectrometer (iCAP™ 7400 ICP-OES, Thermo Fisher Scientific, USA) after digestion at 550°C and re-dissolved in dilute nitric acid. Mature leaves of tea plants were sampled on March 21, 2020, dried in an electric oven at 60 °C, and finely ground to determine the concentrations of nutrients.

The contents of quality-related metabolites in teas were determined in the fourth year of field experiment assuming that the nutrition status of plants reached stable conditions. Tea samples of the first and second harvest were taken on March 22 and 26, 2020, quickly frozen in liquid nitrogen, freeze-dried and finely ground. Powder of tea samples (100 mg) were extracted with 5 mL of H₂O in a boiling water bath for 5 min (Ma et al., 2021). The extract was used for the determination of total free amino acid (TFAA) by spectrophotometry after reaction with ninhydrin reagent and total polyphenol (TP) after reaction with Fe-tartrate reagent (Ruan et al., 2013). The composition of free amino acids in the extract was determined by High Performance Liquid Chromatography (HPLC) using Waters AccQ•Tag™ pre-column derivatization kit following the manufacturer’s instruction (Waters Corporation, Milford, MA, USA). The content of caffeine and the composition of catechins in the extract were determined by HPLC (Waters Corporation, Milford, MA, USA) equipped with a C₁₈ reverse phase column (250 × 4.6 mm) according to the method previously described (Wu et al., 2012). Contents of amino acids, caffeine and catechins were quantified by their areas of chromatographic peaks against those of authentic standards.

Soil samples were taken in the end of field experiment (October 2020) from 5 randomly selected sites to 1 m depth at five separate layers, i.e. 0-20, 20-40, 40-60, 60-80 and 80-100 cm and mixed thoroughly per plot. Stones and debris of roots were removed and samples were separated into two portions after thorough-out mixture. Fresh soil was temporarily stored in a refrigerator at 4°C before the determination of water content and inorganic nitrogen. Ammonium and nitrate in fresh soil were extracted by 2 mol l^-1 potassium chloride and determined by Discrete Chemistry Analyzer (Smartchem 140, AMS Alliance, Frepillon, France). Soil pH was measured in 1:2.5 water paste of air-dried soil samples by a glass electrode (Orion 3 Star, Thermo Ltd., Waltham, MA, USA).

The average price of fresh spring teas was approximately 150 Yuan kg^-1 according to the local market. The net profit of fertilization was calculated as the difference between the values of fresh young shoots of treatments (V_F) and CK (V_CK) and the fertilization cost (C_F) according to the following Equation (1). Cost of harvesting young shoots was not taken into account as this provided important employment and cash income for local labors.

Agronomical efficiency (AE, kg kg^-1) was calculated from the yield (Y_F) and the rate (R_F) of N, P and K ( Table 1 ) in the treatments relative to the yield of CK (Y_CK) according to the following Equation (2):

The greenhouse gas (GHG) emissions derived from fertilization were divided into four parts: production and transportation of fertilizers, and direct and indirect N₂O emissions caused by N fertilizer application according to default method of the Intergovernmental Panel on Climate Change (IPCC) (Forster et al., 2007; Hergoualc’h et al., 2019). The area scaled GHG (CO₂ equivalent) generated from the production and transportation of chemical fertilizers (GHG_A-Pr and GHG _A-Tr) was estimated from their application rates per hectare (F) and respective emission factors (EF) according to the following Equation (3). The emission factors of manufacture and transportation were 8.21 and 0.09 for N fertilizer, 0.73 and 0.06 for P fertilizer, 0.5 and 0.05 for K fertilizer, respectively (Forster et al., 2007; Zhang et al., 2013).

The direct emission of N₂O (dE_N2O) from the application of chemical and organic N fertilizers was estimated from their application rates per hectare (F_N ) and their respective emission factors according to the following Equation (4). The EF for chemical and organic N fertilizers were 0.0175 and 0.0261 according to the most recent meta-analysis (Wang et al., 2020a). These values were slightly lower than 0.0272 (or 2.72%) which used for both chemical and organic N fertilizers in the previous works (Tang et al., 2021).

The indirect N₂O emissions (idE_N2O) caused by the application of N fertilizers were calculated from N rate per hectare, fractions of N loss through leaching (FN_L, %) and runoff (FN_R, %), and default emission factor (0.011, Hergoualc’h et al., 2019) according to the following Equation (5). FN_L was 14.5% based on our three-year lysimeter experiment of similar soil type and texture with the same cultivar Baiye-1 (Zheng, 2022). FN_R was 8.2% according to a field experiment of green tea with similar soil, weather and topographical conditions (Xie et al., 2021). The current value of FN_L was higher while that of FN_R was lower than 9.8% which was adopted for both in the previous work (Tang et al., 2021).

The direct and indirect emissions of N₂O (dE_N2O) were converted to CO₂ equivalent greenhouse gas emission (GHG_A-FN) according to the following Equation (6).

The area scaled total emission (GHG_A) from production, transportation and application of fertilizer was calculated according to the following Equation (7) and was further converted to per yield (GHG_Y) and profit (GHG_P) scales according to the following Equations (8), (9), respectively.

To test the effect of fertilization data were subjected to one-way analysis of variance (ANOVA) combined with the least significant difference (LSD) test by SigmaStat embedded in SigmaPlot (Version 12.5, Systat Software Inc., Palo Alto, CA 94303).

The concentrations of N, P and K in teas of both harvests appeared not directly related to the application rates of fertilizers ( Figure 1 ). However, the concentrations of N in teas of both harvests were significantly decreased in SCF+LSM ( Figures 1A, B ). The concentrations of P and K in teas were inconsistently affected by fertilization treatments between the two harvests. Their concentrations in the first harvest (March 22) were considerably decreased in RSM and SCF+RSM compared to those in CK ( Figures 1C, E ). Their concentrations in the second harvest were unaffected by fertilization treatments ( Figures 1D, F ).

Compared to CK, the concentrations of N in mature leaves of tea plants were significantly increased by fertilization. Their concentrations were most considerably increased in HCF and HCF+RSM but least increased in SCF+LSM ( Figure 2A ). There were significantly positive relation between the N concentration of mature leaves and the N application rate, which could be well described by a quadratic equation (R^{2 =} 0.742, p<0.0001). The concentrations of P in mature leaves of tea plants were significantly decreased in RSM, SCF+RSM and SCF+LSM and unchanged in FP and HCF compared to that of CK ( Figure 2B ). The concentrations of K in mature leaves of tea plants were significantly increased in SCF+LSM, HCF and RSM but were unaffected in SCF+RSM and FP compared to that CK ( Figure 3B ). The responses of P and K concentrations in mature leaves to the rates of P and K fertilizers, respectively could not be described by any defined equations.

The concentrations total free amino acid (TFAA) in teas were significantly increased in HCF but decreased in SCF+LSM in both harvests compared to those of CK ( Table 2 ). Compared to that of CK, the TFAA concentration in the first harvest was significantly increased in SCF+RSM. The TFAA concentrations in both harvests were not significantly affected in RSM and FP. There were significantly (p<0.01) linear positive correlations between TFAA concentrations in both harvests and the N concentrations in mature leaves of tea plants ( Figure 3A ).

The compositions of free amino acids were significantly affected by fertilizers but the effects were inconsistent between the two harvests ( Table 2 ). Compared to CK, FP had little effect on the compositions of free amino acids. RSM affected the composition of free amino acids only in the second harvest, increasing that of theanine (Thea) but decreasing those of glutamate (Glu) and phenylalanine (Phe). SCF+RSM increased Thea concentrations in both harvests whereas had little effect on those of other amino acids. SCF+LSM decreased the concentrations of Thea and glutamine (Gln) in both harvests compared to CK. HCF increased concentrations of most amino acids including Thea compared to CK in the first harvest but had smaller effects in the second harvest. The concentrations of Thea and Gln of both harvests linearly related to the concentrations of N in mature leaves of tea plants ( Figures 3B, C ).

Compared to CK, SCF+LSM decreased the concentrations of caffeine in both harvests ( Table 3 ). SCF+RSM increased while HCF decreased the concentrations of caffeine in teas of the second harvest. The concentrations of caffeine was negatively correlated with K concentrations of mature leaves ( Figure 3D ). Compared to CK, SCF+RSM increased the concentrations of total polyphenol (TP) in both harvests mostly among the treatments. By contrast, HCF decreased the concentrations of TP, catechins and the ratio of TP/TFAA in both harvests compared to CK. The concentrations of TP and catechins in HCF were the lowest among fertilization treatments with only a few exceptions. The concentrations of catechins were also affected by other fertilization treatments but inconsistently between the two harvests ( Table 3 ). SCF+LSM decreased the concentrations of EGC and ECG in the first harvest but increased those in the second harvest. FP decreased the concentrations of ECG in the first harvest and those of EGCG and EC in the second harvest. The concentrations of catechins were only weakly affected in RSM compared to CK.

The yield was relatively stable with small variations (CVs 7.7 − 12.1%) between experimental years ( Table 4 ). Compared to CK without any fertilizers, fertilization treatments significantly increased tea yield by 33.6% − 58.7% ( Table 4 ). Yields were slightly (p>0.05) higher in SCF+RSM, HCF and SCF+LSM and slightly lower in RSM than in FP. The profit of fertilization varied from 19794 to 51284 Yuan ha^-1 ( Table 4 ). The profit and profit/cost ratio were the lowest in RSM and the highest in SCF+RSM. SCF+RSM and SCF+LSM had 80% and 35% higher profits than FP, respectively. The HCF had higher profit than FP but much lower profit and profit/cost ratio than SCF+RSM and SCF+LSM.

The contents of N, P and K in the young shoots, which were calculated from shoot dry tea yield and nutrient concentrations, were significantly increased by fertilization ( Table 4 ). The contents of N and P were significantly increased, in a decreasing order, by HCF and SCF+RSM, followed by SCF+LSM, FP and RSM. Compared to CK, K contents of young shoots were increased significantly by HCF, SCF+RSM, SCF+LSM but unaffected by FP and RSM. The agronomic efficiency (AE) of N, P and K fertilizers were the highest in SCF+RSM and the lowest in HCF ( Table 4 ).

Fertilization significantly decreased soil pH of all layers ( Figure 4A ). The lowest soil pH was found in RSM and HCF. The concentrations of residual NO₃ ^–N and NH₄ ⁺-N varied largely among different soil layers and mainly accumulated at the surface soil (depth 0-20 cm) ( Figures 4B, C ). The amounts of NO₃ ⁻-N and NH₄ ⁺-N in the surface soil (0-20 cm) accounted for 49.9% and 35.1% of the total profile (assuming an unform bulk density) down to 1 m depth, respectively. HCF had extremely high residual NO₃ ⁻-N in the soil down to the depth of 60-80 cm. The concentration ratios of NO₃ ⁻-N/NH₄ ⁺-N were significantly higher in HCF than in other treatments.

Area (GHG_A) and yield (GHG_Y) scaled total greenhouse gas emissions were decreased in SCF+RSM, SCF+LSM and RSM compared to FP ( Table 5 ). SCF+LSM had the lowest GHG_A and GHG_Y. Compared to FP, the profit scaled total greenhouse gas emissions (GHG_P) were decreased in SCF+RSM and SCF+LSM but increased in RSM. GHG_A, GHG_Y and GHG_P were the highest in HCF, which were 3.7 − 4.4, 3.5 − 4.4 and 2.0 − 4.9 times higher than those in other fertilization treatments, respectively.

Compared to CK, tea yield was significantly increased (33.6% − 58.7%) by fertilization, indicating the important contribution of fertilization to improve the productivity of Baiye-1. However, tea yields did not statistically differ among fertilization treatments in spite of different rates of fertilizers. The N concentrations of mature leaves and the contents of amino acids, TP and TP/TFAA significantly responded to the application rate of N fertilizers. In the present experiment, we included a special treatment HCF to evaluate the effect of extremely high rate of fertilizers which had been used in plantations of normal green cultivars (Mishima et al., 2010; Ni et al., 2019). High N rate in HCF significantly increased yield and the concentrations of TFAA and main amino acids (such as Thea and Gln) but decreased the concentrations of total polyphenol and catechins, resulting in the unbalance of TP and TFAA and weakening the intensity of the tea infusion. These results suggest that application N fertilizers at the rate of 250~276 kg ha⁻¹ in SCF+RSM, FP and RSM appeared to supply sufficient N to tea plants. However, HCF produced higher profit regardless of higher fertilization cost than FP because of good price of tea. This might help explain the reason why over-application of fertilizers had been popular in tea plantations (Ni et al., 2019).

The contents of free amino acid and polyphenols were insignificantly affected in FP although the yield was significantly increased (by 38%) compared to CK. Compared to FP and CK, SCF+RSM increased the concentrations of TFAA, Thea, total polyphenol and catechins, thereby improved the overall tea quality. This was associated with the significantly increased concentration of N in mature leaves of tea plants in SCF+RSM ( Figure 2A ). Our previous work showed that N in the mature leaves is removable to support the growth and metabolism of spring tea hence significantly affects the contents of amino acids of spring tea (Liu et al., 2016), which is also supported by the present findings of their close relationships ( Figure 3 ). The photosynthesis rate was likely improved in leaves of increased N content (Okano et al., 1997). SCF+RSM had slightly lower N rate than FP and it was likely that their different effects on tea quality might be partially related to the optimization of P, K rates and their ratio against N to better meet the requirement of tea plants. The amounts of P and K input in SCF+RSM were reduced by 47% and 31%, respectively, compared to FP, by replacing compound fertilizer of equal nutrient formular (N-P₂O₅-K₂O = 15-15-15) with that of specialized formular (N-P₂O₅-K₂O = 18-8-12). The ratio of N, P and K in SCF+RSM was optimized to 1-0.17-0.39, which was more closed to the ratio of their contents in young shoots (1-0.10-0.35, Table 4 ). This result is consistent with recent findings showing that over application of P and K fertilizers reduce green tea quality (Wei et al., 2022; Zhang et al., 2023). The concentrations of P and K in young shoots and mature leaves hardly responded to the rates of fertilizers ( Figures 1 , 2 ). The specially formulated compound fertilizer contained Mg and provided additional Mg (16 kg MgO ha^-1), which might also play a role (Ruan et al., 2012; He et al., 2023). A recent field experiment showed that application of tea-specific fertilizer together with rapeseed cake manure improves the aroma of green tea (Huang et al., 2022).

In the present work, chemical fertilizers were substituted by organic manures fully in RSM and partially in SCF+RSM and SCF+LSM. The yield increase and profit were much lower in the full substitution (RSM) than in the partial substitution (SCF+RSM, SCF+LSM) and even lower than those in FP for the high fertilization cost and relative lower yield of RSM. On the other hand, the quality of tea was improved in RSM and SCF+RSM for the significant increase of Thea content compared to in FP and CK. Nevertheless, it appeared that tea quality was improved more considerably in SCF+RSM relative to RSM for greater increase of the contents of amino acids (Glu, Phe and Ser), caffeine, TP and EGCG. These results confirmed our previous findings that partial substitution of chemical fertilizers by organic manure had better effect than the full substitution (Ji et al., 2018; Ma et al., 2022). These effects might be explained by the increased N concentration of mature leaves indicative of improved N nutrition of tea plants as the result of more readily available N supply in the partial (SCF+RSM) than in the full substitution (RSM) ( Figure 2A ).

The concentrations of TFAA and free amino acids in SCF+LSM were generally decreased compared to other fertilization treatments. Tea plants in SCF+LSM had low N status which was indicated by low N concentration in mature leaves of plants ( Figure 2A ). This is partly explained by its low N application rate, which was 16 − 24% lower than in FP, RSM and SCF+RSM. Furthermore, it was likely that the availability of N in livestock manure with a high C/N ratio was lower than that of rapeseed cake manure with a low C/N ratio. Previous work showed that there was a significant negative correlation between manure C:N ratio and N mineralization in the manure-amended soils (Qian and Schoenau, 2002). On the other hand, SCF+LSM decreased the concentrations of TFAA and main amino acids (Thea and Gln) compared to CK, which likely had been caused by a dilution effect as the result of significantly increased yield. The SCF+LSM scheme may be further optimized by increasing N application rate to improve N supply. The present work suggests that the effect of substitution is dependent on the type of organic manure and the bioavailability of nutrients.

Greenhouse gas emission in tea field has been estimated by IPCC default method in recent works. Production and application of N fertilizer accounted for a large portion of the GHG emissions in tea industry (Liang et al., 2021; Tang et al., 2021). SCF+LSM had the lowest area scaled GHG_A emission for its low N fertilizer rate. RSM had lower GHG_A than SCF+RSM and both of them had lower GHG_A than FP. However, an opposite result was found in greenhouse gas emissions (GHG_Y and GHG_P) per yield and profit, which were much higher in RSM than in SCF+RSM and SCF+LSM. In all these treatments, it was assumed that organic manure had been locally recycled without long-distance transportation. In case of long-distance transportation of organic manures, RSM would have higher gas emissions due to its low nutrient content and high moisture content. Therefore, it appears that the full substitution of chemical fertilizers by rapeseed cake manure had not economic and sustainability advantages over the partial substitution. The application rate of N and the estimated GHGs in the present work were lower than those of the previous work in normal green-leaf cultivars based on Nutrient Expert (Tang et al., 2021).

Nutrient use efficiency is an important indicator evaluating the impact of an applied fertilizer on crop production and economic return (Fixen et al., 2015). A recent work showed that the agronomic efficiency (AE) was positively correlated with the free amino acid content, income, and negatively with global warming potential of tea (Tang et al., 2023). The present work showed that AEs were increased in SCF+RSM together with increases of tea productivity, the contents of free amino acids, total polyphenols and economic profit as well as reduced greenhouse gas emissions. The AEs of the present work were relatively low compared to the reported values in tea (Tang et al., 2023) and other crops (Fixen et al., 2015), which was likely attributed to the low yield of Baiye-1. On the other hand, AEs in HCF were decreased because of high amount of applied fertilizers. However, the low AEs of HCF did not necessarily reflect low income and low contents of free amino acids, a finding inconsistent with the recent finding (Tang et al., 2023). However, HCF remarkably increased greenhouse gas emission as well as the contents of residual NO₃ ⁻-N in the deep soil hence enhanced the risk of N leaching. These results indicated that over-use of fertilizers deteriorated Baiye-1 tea quality and imposed serious environmental risks.