*2.1. Research Area*

The research was conducted in the Loess Plateau at the Wuliwan catchment of Ansai County, Shaanxi Province, China (36◦39–36◦52 N, 109◦20–109◦21 E). The site has a hilly loess landscape and an elevation of 1010–1400 m above sea level. The study area has a typical semiarid climate with a mean annual rainfall of 505 mm, an annual average temperature of 8.8 ◦C, and a frost-free period of 159 days. Approximately 60% of the precipitation occurs between July and September, and the precipitation varies greatly in different years. The soil in this region is mainly loessal soil, developed from wind-accumulated loess, which has extremely poor resistance to erosion. The proportion of the differently sized soil particles in the 0–20 cm layer was 63.6% sand, 29.2% silt, and 7.2% clay. Artificial forests have been planted at the study site since the 1970s to mitigate the serious soil erosion on steeply sloped crop land [29]; therefore, the large area of plantations have different stand ages. The dominant plantation species is black locust and *Caragana korshinskii* Kom.

### *2.2. Experimental Design and Sampling*

Based on the afforestation records of the local forestry department, black locust forests of five age classes (i.e., 10, 20, 30, 36, and 45 years old during field sampling in 2016, hereafter referred to as RP10, RP20, RP30, RP36, and RP45, respectively) were selected to represent a chronosequence. All of the sample lands were converted from cropland subjected to similar farming practices, namely, mainly planting maize (*Zea mays* L.), foxtail millet (*Setaria italic* L.), and broomcorn millet (*Panicum miliaceum* L.) with watering mainly through the rainfall and a small amount of manure was the main fertilizer. After trees were planted, the sample lands were fenced to prevent grazing and other human disturbance including thinning, wood usage, or other practices. Detailed information of the sampled lands is shown in Table 1. The climatic and edaphic conditions of the forests of five age classes were similar. Three replicate fields were measured for each black locust stand age. The three replicate fields for each stand age were within 1.5 km of one another to ensure consistent climatic variables and other conditions. In each replicate field, we randomly selected three replicated plots (20 m × 20 m) for sampling. All sites were topographically similar (i.e., slope aspect and slope degree) with similar elevation. Finally, we established 45 sampling plots (5 stand ages × 3 replicate fields × 3 sample plots).

Green and senesced leaves were collected in mid-August and mid-October 2016, respectively, when the green leaves were at peak biomass, and senesced leaves were ready to fall. At least five individual trees were selected in each sample plot. At least 20 branches were chosen around each tree from the lower, middle, and upper canopy. Approximately 50 g leaves were sampled from each individual (picked using a long reach pruner), all the leaves from the same sample plots were mixed (about 250–300 g), and then a quantity sufficient for chemical analysis (about 100 g) was taken out for analysis. The leaves with obvious diseases and/or insect pests were excluded. To avoid nutrient decomposition and leaching, the fully senesced leaves were collected from trees rather than from litter in mid-October, when they were completely yellow, dry, without any signs of deterioration, and would fall from the branch with a gentle touch [10,34]. The sample leaves were transfer to the laboratory in paper bags, then oven-dried at 105 ◦C for 30 min and maintained at 60 ◦C until a constant mass was reached. Dried sample leaves were ground and sieved with a 0.1 mm sieve.


**Table 1.** Field site information for *Robinia pseudoacacia* L. plantations with different stand ages.

ES and EN means East by South and Ease by North; DBH means diameter at breast height.

The soil samples were collected in mid-August 2016. In each sample plot, we selected five sample points along the diagonal. Because the surface soil (0–10 cm) is the most closely related to plant nutrients [10,29], we collected soil samples at each point from the 0–10 cm layer using a soil auger with a 4 cm diameter. The soil samples from the same plot were adequately mixed and reduced by coning and quartering to achieve appropriate quantities (about 400 g). Then, the soil samples were transported to the laboratory, air-dried at room temperature (25–28 ◦C), and passed through a 0.25 mm sieve in preparation for soil C, N, and P analysis. To avoid the impact of roots and litter, all samples were collected 80 cm from trees after the litter layer was removed and the plant roots, stones and debris were removed before drying through hand picking and sieving. A vegetation survey was also conducted in each plot in August 2016 using five randomly selected 1 m × 1 m small plots.

### *2.3. Laboratory Analysis and Determination*

The TOC analyzer (Total organic carbon analyzer, Shimadzu Corp., Kyoto, Japan) was used to measure the total C from soil and leaf samples (0.1 mol L−<sup>1</sup> HCl were used to destroy the carbonates in soil). The total nitrogen (TN) content was measured using the semimicro-Kjeldahl method with a Kjeldahl auto-analyzer (KDY-9830, Beijing, China) after digestion in sulfuric acid hydrogen peroxide (for leaves) or sulfuric acid perchloric acid (for soil). Total phosphorus (TP) content was measured using the molybdenum antimony colorimetric method by using an ultraviolet spectrometer subsystem (UV-6100, Shanghai, China) after digestion following the methods as in TN measurement [35].

### *2.4. Calculations and Statistical Analysis*

NuRE was used to quantify the nutrient resorption, which was calculated as follows [19]:

$$\text{NuRE} = (1 - \frac{W\_1}{W\_2} \times \text{MLCF}) \times 100\% \tag{1}$$

where NuRE is the nutrient resorption efficiency, and W1 and W2 are the nutrient concentrations (TN or TP, g kg−1) in senesced and green leaves, respectively. MLCF is the mass loss correction factor to correct the unbiased resorption value, which is 0.784 for broad-leaved deciduous trees as introduced by Vergutz et al. [12].

The soil and leaf C:N:P stoichiometry ratios were calculated as the mass ratio. There were no missing values in our data. All data were distributed normally after the Kolmogorov–Smirnov test (*p* > 0.05 for each null hypothesis). Then, the Bartlett test was used to investigate the homogeneity of variance, and *p* > 0.05 for each null hypothesis. There were no significant differences for each soil and leaf properties between different replicate fields of the same stand ages, or between different sample

plots of the same replicate fields (*p* > 0.05). Therefore, one-way analysis of variance (ANOVA) was used to test the differences of NuRE, nutrients and stoichiometric ratios among the plantations of different ages (the significant level was *p* < 0.05). A least significant difference (LSD) test was conducted for post-hoc multiple comparisons. A Pearson correlation analysis was used to estimate the relationship between NuRE and other leaf or soil nutrition properties. The relationships between the stand age and other properties were estimated by linear regression. All of the statistical analyses were conducted using SAS (SAS Institute Inc., Cary, NC, USA), and the figures were plotted in Origin 2016 (OriginLab Corporation, Northampton, MA, USA).
