Greenhouse experiment

A pot experiment was carried out inside a plastic shed (with no temperature controlling) at the Northwest Agriculture and Forestry University, (N 34°16′, E 108°4′), China to evaluate the growth performance of T. chinensis treated with different combination rates of W, N and P. Mean annual temperature, precipitation, relative humidity and photosynthetically active radiation of the experimental site (outside the plastic shed) were 13.7°C, 650 mm, 59% and 2450 MJ m⁻², respectively. Coal mine spoil samples were collected from Yangchangwan coal mining area in Lingwu, Ningxia of China (106° 35′–106° 38′E, 37° 59′–38° 03´N). Average annual temperature and evaporation rates in Lingwu, Ningxia were 8.6°C and 2682 mm, respectively. Mean annual precipitation, relative humidity and photosynthetically active radiation of Lingwu, Ningxia were 245 mm, 50% and 2870 MJ m⁻², respectively (https://www.timeanddate.com/weather/china/yinchuan/climate). Annual average and maximum wind speeds were 3.1 and 23.3 m s⁻¹, respectively. The frost-free period lasted for 145 days. Mine spoils were sampled according to the method followed by Sikdar et al. (2020). Collected materials were homogenized followed by bulking, air drying, manual crushing and sieving through 2 mm mesh. The chemical properties of spoils are listed in Table S1.

One-year-old T. chinensis seedlings with similar height (105.55 ± 14.58 cm) and stem diameter (5.91 ± 0.98 mm) were collected from a forestry nursery, in Inner Mongolia, China. Each plastic pot (32 cm top diameter, 27 cm bottom diameter and 30 cm height) was filled with 14 kg of coal mine spoils and one seedling was planted in each pot in early March 2018. Sufficient water was supplied daily to the seedlings to ensure their normal growth in the first month and after that the water-stress treatment started according to the experimental design and continued up to October 2018. To maintain the water levels at field capacity for each water stress treatment, pots were watered daily by a weighing method following the process reported by Roy et al. (2020). The pots with seedling were randomly rearranged fortnightly to minimize the possible positional effects. Fertilizers (N and P) were added in holes close to the seedlings' root zone. The N fertilizer (urea-46% N) was added as granular form in four times (¼ May 6th, ¼ June 6th, ¼ July 6th and ¼ August 6th). P fertilizer (triple super phosphate-46% P₂O₅) was added also as granular form in two equals halves (¹/₂ May 6th and ¹/₂ July 6th) during the experimental period. We did not add K or other elements such as Mg, Ca or Fe in our experiment because our soils have relatively high levels of K (Table S1).

Experimental design

The RSM was used to optimize soil water and fertilizers application rates to get maximum growth performance of T. chinensis. A central composite design (CCD) was selected for the optimization process. The independent variables chosen for this study were W, N and P, while different morphological, physiological and biochemical growth parameters were selected as the dependent or response variables. Application rates of W, N and P in our study were chosen based on some previous findings (Singh and Singh 2006, Zhou et al. 2011, Roy et al. 2020) and five different levels (+α, +1, 0, −1, −α) of W, N and P was selected (Table 1) according to the equation:

(())

Here, X_i denotes the coded value of the independent or process variables; Z_i denotes the real value of the process variable; Z_i0 indicates the actual value of Z_i at the center point; and ∆Z_i denotes the step change value. Total experimental treatments were 21 as per confirmed by RSM (including control) and each treatment was replicated three times resulting in a total of 63 pots that were randomly arranged in the greenhouse. The detailed description of the experimental treatments, along with their coded and actual levels are presented in Table 2. Experimental data obtained were fitted to the second-order polynomial model, as presented:

(())

Here, Y represents response variables; β₀ determine constant-coefficient; β₁, β₂ and β₃ represents interpret linear coefficients; β₁₂, β₁₃ and β₂₃ represents interaction coefficients; β₁₁, β₂₂ and β₃₃ determine quadratic coefficients; A, B and C represents coded value of W, N and P.

To get interaction effects of W, N and P on integrated growth performance (IGP) of T. chinensis, the plant morphological, physiological and biochemicals data were transformed according to the experimental design using the following equation and presented in Table S2.

(())

In Eqn 3, X is the coded value for each treatment, x is the measured value of each treatment, x_min is the minimum, and x_max is the maximum value recorded for each parameter for different treatments.

Field experiment

A field experiment in the Yangchangwan coal mining area was conducted to assess the effects of W, N and P application rates on growth and development of T. chinensis. W, N and P application rates were selected based on evidence from our pot experiment as well as from several previous studies. Schachtsiek et al. (2014) recommended that the application of 155 mm year⁻¹ is an effective strategy for early survival and growth of six afforestation desert shrub species including T. chinensis in abandoned degraded areas. Xu et al. (1998) concluded that effective water supply for artificial revegetation and desertification control with T. chinensis on the northwestern region of China was 115–150 mm year⁻¹. Moreover, water levels from 80 to 160 mm year⁻¹ were previously applied in several studies focusing on the establishment of desert shrub species in degraded areas of northwestern region of China (Khamzina et al. 2008, Li et al. 2016, Zhang et al. 2017). Specifically a L-4 orthogonal array design was assigned with two levels of W addition (W₁₅₀, 150; and W₁₀₀, 100 mm year⁻¹), two levels of N addition (N₁₆₀, 160; and N₈₀, 80 kg ha⁻¹) and two levels of P addition (P₈₀, 80 kg ha⁻¹; and P₄₀, 40 kg ha⁻¹). In May 2019, 20 plots measuring 4 × 4 m were established with five experimental treatments: (1) W at 150 mm year⁻¹ + N at 160 kg ha⁻¹ + P at 80 kg ha⁻¹ (W₁₅₀N₁₆₀P₈₀); (2) W at 150 mm year⁻¹ + N at 80 kg ha⁻¹ + P at 40 kg ha⁻¹ (W₁₅₀N₈₀P₄₀); (3) W at 100 mm year⁻¹ + N at 160 kg ha⁻¹ + P at 40 kg ha⁻¹ (W₁₀₀N₁₆₀P₄₀); (4) W at 100 mm year⁻¹ + N at 80 kg ha⁻¹ + P at 80 kg ha⁻¹ (W₁₀₀N₈₀P₈₀); and (5) W at 100 mm year⁻¹ without N and P (control). In our study, we added low W level and no N and P fertilizers to the control treatment seedlings in both greenhouse and field experiments. We followed this approach to guarantee seedling survival based on indications from previous studies under similar environmental conditions (Li et al. 2016, Wang et al. 2016, Zhang et al. 2017). Under field conditions, each treatment was replicated four times (12 seedlings per each replicate i.e., 48 seedlings per treatment), separated by 2-m buffers and randomly assigned to plots within each block. N and P were applied in two halves (¹/₂ June 10th and ¹/₂ August 10th) during the experimental period. Beside natural precipitation, additional 100 and 150 mm water was added to experimental plots between early-May and end of October. Specifically, seedlings were watered every week in doses of 5–7.5 mm of water per meter square if an effective precipitation event (> 5 mm) did not occur within 7 days. Watering was suspended if significant rainfall occurred. During each of five consecutive monthly irrigation events we added 15, 20, 20, 20, 15 and 10 mm of water, respectively, to give a total of 100 mm for specific plots. Similarly, we added 22.5, 37.5, 37.5, 37.5, 22.5 and 15 mm of water each month to give a total of 150 mm to other plots between May and October 2019. Watering was carried out using a pump-line injector system for a total of 20 times during the experimental period. After 6 months of transplantation (October, 2019), eight representative seedlings from each treatment were randomly selected and harvested for determination of tap root length and biomass production (fresh and dry weight). The rest of the seedlings were allowed to grow further for revegetation and restoration of coal mine degraded spoils in Yangchangwan coal mining area of Ningxia, China.

Measurement of growth parameter

Plant morphological, physiological and biochemicals measurements for T. chinensis were taken at the end of the experimental period in both pot and field experiments.

Assessment of morphological parameters

Growth analysis of T. chinensis was performed by measuring plant height, stem diameter, maximum root length of the main tap root, plant fresh weight (FW) and plant dry weight (DW). The plant DW was measured by drying plant samples in an oven at 80°C for 72 h. The root-shoot (R/S) biomass ratio (dry weight basis) was calculated by dividing root and shoot biomass.

Measurements of physio-biochemical constituents and antioxidant enzyme activities

Leaf gas exchange parameters such as net photosynthesis rate (P_n), transpiration rate (T_r), stomatal conductance (G_s), the intercellular concentration of CO₂ (C_i) and water use efficiency (WUE, as P_n/T_r) were measured on sunny days between 08:30 and 11:30 h by a portable photosynthesis system CIRAS-3 (PP Systems). To avoid systematic bias in gas exchange parameter changes due to temporal differences, we measured leaf gas exchange parameters using an alternate model of different leaves in different plants, which ensured the comparability and accuracy of the measured data between different treatments. For three seedlings within each treatment, we randomly selected 2–3 healthy, mature, sun-exposed leaves in the mid-upper canopy of each plant as our sample for fixed observations. Each blade was measured three times. Given the irregular leaf shape of T. chinensis, we distributed the selected leaves throughout the leaf chamber as much as possible to obtain representative data. The plant leaves were removed and scanned with a scanner. The actual leaf surface area within the gas exchange cuvette was estimated using a LiCor 3100 leaf area meter (LiCor), and the photosynthetic parameters were recalculated based on the actual photosynthetically active area.

Chlorophyll (Chl) was extracted from fully expanded 0.1 g fresh leaves using a 10 ml mixture of ethanol, acetone and distilled water (4.5:4.5:1) according to the technique described by Roy et al. (2020). The absorbance of the extracts was recorded at 645, 663 and 470 nm, and the concentration of Chl a, Chl b, total Chls and catotenoids (Cars) were calculated as mg g⁻¹ FW according to the formula of Arnon (1949).

Leaf water potential (LWP) was measured with the help of plant water potential meter (PMS-Model 1000, PMS Instrument Company) before morning 06:00. Relative water content (RWC) and membrane stability index (MSI) of T. chinensis leaves were measured according to the method described by Bandeppa et al. (2019). Free proline (Pro) from fresh leaf samples (0.1 g) was extracted using a sulphosalicylic acid solution and measured with a ninhydrin solution, according to Bates et al. (1973). The content of total soluble sugar (SS) and soluble protein (SP) were colorimetry-measured using anthrone-H₂SO₄ (Joseph 1955) and Bradford (1976) methods, respectively. Glucose and bovine serum albumin were used to develop standard curves for SS and SP, respectively.

The method from Roy et al. (2020) was adopted to extract, and measure the content of malondialdehyde (MDA) in leaf samples of T. chinensis. The superoxide anion (O₂^●–) was estimated by testing the formation of nitrite from hydroxylamine, as described by Ke et al. (2002). Hydrogen peroxide (H₂O₂) content was measured according to Velikova et al. (2000).

Fresh T. chinensis leaf samples (0.3 g) were homogenized in a mortar and pestle using 8 ml of 50 mM sodium phosphate buffer (pH 7.8), followed by centrifugation at 10 000g for 20 min at 4°C. The enzyme extracts obtained were used for the estimation of enzyme activities (Roy et al. 2020). The superoxide dismutase (SOD) activity was assayed following the method from Roy et al. (2020). The catalase (CAT) activity was measured based on the method of Beers and Sizer (1952). The peroxidase (POD) activity was carried out using the guaiacol oxidation method described by Ekmekci and Terzioglu (2005). Ascorbate peroxidase (APX) activity was determined following Nakano and Asada (1980).

Data analysis

Optimum water and fertilizer rates were obtained from an optimization process combined with the derringer's desired function approach using Design Expert statistical software version 11 (Stat-Ease). Origin 2018 (OriginLab Inc.) was used to perform principal component analysis of the traits. A heatmap was generated using the https://biit.cs.ut.ee/clustvis/online program package with Euclidean distance as the similarity measure, and hierarchical clustering with complete linkage heatmap summarized all the plant responses to the different combination of W, N and P treatments.

Greenhouse experiment

Analysis of variance (anova) results are presented in Table S3. The models for different growth parameters had P-values of less than 0.0001, except for the root-shoot ratio (R/S). In contrast, lack of fit P-values was non-significant (greater than 0.05) for all models. Among different models (linear, interactive [2FI], quadratic, and cubic polynomial), the quadratic model was found to be the most suitable model for growth parameters and integrated growth performance (IGP) of T. chinensis (except plant dry weight (DW), R/S, leaf water potential (LWP) and catalase (CAT) activity) (Table S3). The differences between predicted R-squared (R²_p) and adjusted R-squared (R²_a) values were less than 0.2, indicating a good agreement between them (Roy et al. 2020; Table S3). A low CV value (less than 10%) for most of the growth parameters as well as IGP of T. chinensis (except H₂O₂ content) designated that the deviations between predicted and experimental values were very low, thus suggesting a reproducibility of the experimental model (Table S3). Finally, an adequate precision value above 4 suggests that the models predicted by the software could also be favorably used (Table S3).

Morphological parameters

The addition of W, N and P caused a significant increase in plant height and stem diameter in all experimental treatments compared to control. The maximum increase in plant height and stem diameter were observed under W₆₀N_1.68P_1.26 treatment (W at 60% field capacity [FC], N at 1.68 g plant⁻¹ and P at 1.26 g plant⁻¹). Here plant height and stem diameter were 141.7 and 125.4% greater, respectively, than those of control seedlings (Table 3). DW dramatically increased from 23.75 to 34.18 g plant⁻¹ when W level increased from 40% FC (W₄₀N_0.84P_1.26) to 80% FC (W₈₀N_0.84P_1.26; Table 3). The root: shoot ratio (R/S) significantly increased when seedlings were treated with W_71.1N_1.34P₂ (+8.8%) and W_71.1N_1.34P_0.51 (+10.5%) and decreased with W_48.1N_0.34P₂ (−10.5%) compared to control seedlings (Table 3).

Gas-exchange attributes and photosynthetic pigment contents

Net photosynthesis rate (P_n) and transpiration rate (T_r) both increased rapidly with W, N and P application rates until reaching peak values of 14.01 and 6.46 mmol m⁻² s⁻¹, respectively at moderate W, N and P doses (W₆₀N_0.84P_1.26). Then P_n and T_r decreased under greater W, N and P applications (Table 3). Leaf water use efficiency (WUE) significantly upregulated in all treatments, when compared to control. However, WUE reached maximum increase (+76.9% over control) under the W_48.1N_1.34P₂ treatment (Table 3). Moderate-to-high W and N doses, regardless of P additions, caused significant increases in photosynthetic pigments compared to control. Therefore, plants treated with W_71.9N_1.34P₂, W_71.9N_1.34P_0.51, W₆₀N_0.84P_2.52 and W₆₀N_0.84P₀ showed a remarkable improvement in the contents of Chlorophyll (Chl) a by 257.1, 228.6, 219.0 and 247.6%; Chl b by 244.4, 222.2, 200.0 and 222.2%; total Chls by 241.9, 216.1, 203.2 and 229.0%; and Carotenoids (Cars) by 154.5, 109.1, 127.3 and 163.6%, respectively, over that of control treatment (Table 3).

Water-status, osmolytes and soluble proteins

The addition of W, N and P across the various treatments caused a remarkable increase in leaf water potential (LWP) values, which were 4.2–74.2% higher than those in control seedlings (Fig. 1A). Seedlings exposed to W_71.9N_1.34P₂, W_71.9N_1.34P_0.51 and W₈₀N_0.84P_1.26 presented a significant enhancement in relative water content (RWC), which was 27.4, 29.2 and 29.6% higher, respectively, than the control. Also, membrane stability index (MSI) under these treatments increased by 82.8, 86.9 and 80.0%, respectively, over control seedlings (Fig. 1B,C). Addition of W, N and P remarkably reduced the contents of proline (Pro) and soluble sugar (SS) and increased soluble protein (SP) content over that of control (Fig. 1D–F). Seedlings exposed to W additions ≤48% FC, regardless of N and P additions, had greater Pro and SS and lower SP content. Pro and SS contents decreased by 17.7–54.9 and 6.7–57.9%, respectively, and SP content increased by 3.4–110.6%, across the various treatments compared to control (Fig. 1D–F).

Lipid peroxidation, ROS production and activities of antioxidant enzymes

Leaf malondialdehyde (MDA), H₂O₂ and O₂^●– contents significantly decreased across treatments by 5.3–33.5, 9.7–81.8 and 16.7–77.8%, respectively, when compared to control (Table 4). Under the same N, P additions (N_0.84P_1.26), production rate of H₂O₂ and O₂^●– contents dramatically decreased from 0.65 to 0.13 (μmol g⁻¹ FW) and 0.15 to 0.04 (nmol min⁻¹ g⁻¹ FW), respectively, when W levels increased from 40% FC (W₄₀N_0.84P_1.26) to 80% FC (W₈₀N_0.84P_1.26) (Table 4). Seedlings treated with very low- to low-W doses (≤48 %FC) such as W_48.1N_1.34P₂, W_48.1N_1.34P_0.51, W_48.1N_0.34P₂, W_48.1N_0.34P_0.51 and W₄₀N_0.84P_1.26 showed less reduction in MDA contents compare to seedlings treated with moderate-to-very high W doses, regardless of N and P fertilization (Table 4). Treatments containing very low-to-low W and N doses such as W_48.1N_0.34P₂, W_48.1N_0.34P_0.51, W₄₀N_0.84P_1.26 and W₆₀N₀P_1.26 resulted in a notable increase in the activity of superoxide dismutase (SOD) (+11.6, +12.9, +2.2 and +8.1%, respectively, compared to control). Similarly, catalase (CAT) activity increased by 2.7, 6.2, 3.6 and 0.2% and ascorbate peroxidase (APX) activity by 15.9, 8.1, 4.7 and 5.9% respectively, compared to the control treatment (Table 4). In comparison with control, the peroxidase (POD) activity significantly decreased with the addition of W, N and P doses but the treatment W₄₀N_0.84P_1.26 showed a slight improvement in POD activity (by 2.6%) compared to control (Table 4).

The absolute value of the regression coefficient in the equation can be used to judge the significant degree of each factor on the different growth responses of T. chinensis (Table 5). The positive and negative values of the coefficient can indicate the synergistic or antagonistic effect of each factor on the different growth parameters. Coefficients of β₁, β₂ and β₃ are positive values suggesting that W, N and P additions have synergistic effects on various growth traits of T. chinensis. For instance, W additions have synergistic effects on plant height, stem diameter, DW, R/S, P_n, T_r, Chl a, Chl b, total Chls, Cars, LWP, RWC, MSI and SP but antagonistic effects on other parameters. N additions have antagonistic effects on DW, T_r, Pro, SS, MDA, H₂O₂, O₂^●–, SOD, CAT and APX, but synergistic effects on other growth parameters (Table 5). P additions have synergistic effects on stem diameter, DW, WUE, LWP, RWC, MSI, Pro, SP, MDA, SOD and POD, but antagonism effects on other parameters (Table 5). Higher values of negative coefficients of β₁², β₂² and β₃² imply that too much W, N and P additions can inhibit increases in stem diameter, P_n, T_r, and most growth traits (see Table 5). Absolute values of coefficients of β₁₂, β₁₃ and β₂₃ show the coupling effects of W × N, W × P and N × P, respectively on different growth traits of T. chinensis (Table 5).

Interactive effects of W, N and P on IGP of T. chinensis

To investigate the interactive effects of W, N and P on integrated growth performance (IGP) of T. chinensis, 3D surface plots were generated according to CCD. T. chinensis seedlings showed highest IGP of 0.60 when treated with high W, N and P doses (W_71.9N_1.34P₂) (Table S2). The interaction between the two factors (x and y) with IGP of T. chinensis at the z-axis is shown in Fig. 2A–C. Each figure shows the effect of two factors when the third factor is kept at a central level (moderate dose).

Fig. 2A shows that IGP of T. chinensis dramatically increased with the mutual increase of W and N additions. ANOVA results also show that the interactive effects of W × N had a significant effect (P < 0.05) on the IGP of T. chinensis (Table 5). We found that IGP rapidly increased as the W regime (Fig. 2B) and N doses (Fig. 2C) also increased either at low- or high-P doses. At very low-W and N doses, the growth performance of T. chinensis remarkably decreased and the addition of P did not reveal any positive effect. Therefore, no significant interactive effects (P > 0.05) of W × P and N × P were detected on the IGP of T. chinensis (Table 5). The magnitude of the effects of W, N and P on IGP of T. chinensis was W > N > P (Table 5).

Optimization of process variables and validation of the optimized condition

According to CCD analysis the optimal W, N and P application rate was 58.90% FC, 1.14 g plant⁻¹, 0.61 g plant⁻¹, respectively (Fig. 2D), which is equivalent to 160 and 85 kg P ha⁻¹ and appropriate for a planting density of 1.4 × 10⁵ seedlings ha⁻¹ (1-year-old T. chinensis individuals) under field conditions (Su et al. 2014). The accuracy of the optimized outcome was tested by comparing mean experimental values with predicted values resulted by the RSM model. A desirability ramp was developed from optimal points presented in Fig. 2D.

Field experiment

We measured the effects of optimum W, N and P additions on multiple morpho-physiological features of field-grown T. chinensis including plant height, stem diameter, maximum root length, FW, DW, R/S ratio, P_n, T_r, C_i, WUE, RWC and MSI (Fig. 3A–L).

In comparison with control, W₁₅₀N₁₆₀P₈₀ (W at 150 mm year⁻¹, N at 160 and P at 80 kg ha⁻¹), W₁₅₀N₈₀P₄₀, W₁₀₀N₁₆₀P₄₀ and W₁₀₀N₈₀P₈₀ treatments markedly increased plant height (by 31.3, 57.3, 24.4 and 12.2%, respectively), stem diameter (by 39.2, 49.3, 18.2 and 5.4%, respectively), maximum root length (by 67.8, 74.7, 40.2 and 24.1%, respectively), FW (by 39.7, 57.6, 26.5 and 9.3%, respectively) and DW (by 33.8, 51.4, 26.2 and 7.3%, respectively) (Fig. 3A–E). However, the R/S ratio was considerably lower in the seedlings treated with W₁₅₀N₁₆₀P₈₀ and W₁₅₀N₈₀P₄₀ over control (Fig. 3F). T. chinensis treated with W₁₅₀N₁₆₀P₈₀, W₁₅₀N₈₀P₄₀, W₁₀₀N₁₆₀P₄₀ and W₁₀₀N₈₀P₈₀ showed increases in P_n by 68.9, 102.4, 49.5 and 19.5% and C_i by 31.2, 19.6, 23.0 and 4.2%, respectively, compared to control (Fig. 3G and J). The T_r was significantly higher in W₁₅₀N₁₆₀P₈₀ (by 87.73%), W₁₅₀N₈₀P₄₀ (by 90.92%) and W₁₀₀N₈₀P₈₀ (by 43.67%) over control (Fig. 3H). Compared to control, W₁₅₀N₈₀P₄₀, W₁₀₀N₁₆₀P₄₀ and W₁₀₀N₈₀P₈₀ enhanced WUE by 3.0, 0.5 and 37.0%, respectively, but W₁₅₀N₁₆₀P₈₀ considerably reduced WUE by 12.6% (Fig. 3I). RWC in the leaf of T. chinensis increased by 15.13, 18.88, 11.75 and 7.13% and MSI increased by 31.2, 46.22, 25.04 and 19.52% when seedlings were treated with W₁₅₀N₁₆₀P₈₀, W₁₅₀N₈₀P₄₀, W₁₀₀N₁₆₀P₄₀ and W₁₀₀N₈₀P₈₀, respectively, as compared to control seedlings (Fig. 3K–L).

Growth performance of T. chinensis in response to combined applications of W, N and P assessed by heatmap and principal component analysis methods

Results from heatmap analyses show morphological, physiological and biochemical responses of T. chinensis to different W, N and P combinations under pot- (Fig. 4A) and field-grown (Fig. 4B) conditions. For both pot- and field-grown T. chinensis seedlings, measured traits that cluster together are strongly correlated to each other.

The aggregated data heatmap analysis (Fig. 4A) shows that treatments containing W doses ≤48% FC, regardless of N and P fertilization, clustered on the left while the rest of the treatments clustered on the right suggesting that W was the main clustering factor. Treatments containing moderate-to-very high W doses (≥ 48% FC; i.e. W_71.9N_1.34P_0.51, W_71.9N_1.34P₂, W₈₀N_0.84P_1.26, W_71.9N_0.34P₂ and W₆₀N_1.68P_1.26, W₆₀N_0.84P₀, W₆₀N_0.84P_2.52 and W₆₀N_0.84P_1.26) clustered together because of reduced antioxidant enzyme activities (i.e. SOD, APX, CAT and POD), ROS production (H₂O₂ and O₂^●–), osmolytes (SS and Pro) and MDA. The same treatments showed high levels of leaf water-status (RWC, LWP and MSI), morphological growth traits (DW, plant height [PH], R/S) and photosynthetic pigments contents (Chl a, Chl b, total Chls, and Cars) and gas-exchanges parameters (P_n and T_r). On the contrary, treatments containing very low- and low-W doses (≤48% FC) such as W_48.1N_1.34P₂, W_48.1N_1.34P_0.51, control, W_48.1N_0.34P₂, W_48.1N_0.34P_0.51 and W₄₀N_0.84P_1.26, clustered on the left because of their high accumulation of ROS, MDA, antioxidative enzymes, osmolytes, but low levels of morphological growth traits, gas-exchanges parameters, photosynthetic pigments and leaf water-status (Fig. 4A).

For field grown T. chinensis, the heatmap identified two main clusters (Fig. 4B), where high-W containing treatments like W₁₅₀N₁₆₀P₈₀ and W₁₅₀N₈₀P₄₀ were separated from the rest of the treatments because of their high contents of C_i, T_r, SD (stem diameter), PH, DW, FW, P_n, MSI, RL (root length) and RWC. On the other hand, W₁₀₀N₁₆₀P₄₀, W₁₀₀N₈₀P₈₀ and control treatments clustered separately on the right because of their high contents of R/S ratio and WUE (Fig. 4B).

Principal component analysis (PCA) shows that for both pot and field experiments, the first five and first two principal components (PCs), respectively, had Eigen values higher than one. For pot and field experiments, the first two PCs explained 70.53 and 95.33% of the total variance, respectively, with PC1 and PC2 accounting for 55.86 and 14.67% for pot experiment (Fig. 4C) and 82.93 and 12.4% for field experiment (Fig. 4D).

Fig. 4C shows how T. chinensis treated with very low- to low-W doses such as W_48.1N_1.34P₂ (T5), W_48.1N_1.34P_0.51 (T6), W_48.1N_0.34P₂ (T7), W_48.1N_0.34P_0.51 (T8), W₄₀N_0.84P_1.26 (T10) and control (T21), positioned on the negative side of PC1 of the PCA score plot, were strongly and positively correlated with MDA content, ROS production, osmolytes accumulation and antioxidant enzyme activities. On the contrary, moderate- to very high-W treated seedlings, regardless of N and P fertilization positioned on the positive side of PC1 of the PCA score plot showing positive correlations with photosynthetic pigments, leaf water-status, morphological parameters and gas-exchange parameters. Moreover, PC2 was positively correlated to WUE and negatively correlated to T_r (Fig. 4C).

Increases in W levels contributed to the sharp separation of PC1 in seedlings growing under field conditions (Fig. 4D). High W-treated seedlings under W₁₅₀N₁₆₀P₈₀ and W₁₅₀N₈₀P₄₀ treatments are distributed on the positive side of PC1 of the PCA score plot. These seedlings showed a strong positive correlation with MSI, PH, RWC, P_n, DW, RL, SD, T_r and C_i. Instead, low W-treated treatments like W₁₀₀N₈₀P₈₀ and control treatments were distributed under the negative direction of PC1 showing a positive relationship with R/S ratio and WUE (Fig. 4D).