http://orcid.org/0000-0002-3820-9585
Figures
Abstract
Although the spatial mapping and fertility assessment of soil chemical properties (SCPs) are well studied in the Loess Plateau region of China at farmland scale, little is known about spatial mapping the SCPs and their fertility and their influence factors at urban forest scale. The objectives of this study were to (1) compare the performance of two spatial interpolation methods, Ordinary kriging (OK) and regression kriging (RK), and (2) explain the relationships of the vegetation, terrain, and soil layer depth between the eight SCPs and their fertility, and (3) find the limiting factors of soil comprehensive fertility in this study area? The Yan’an urban forest was taken as study case, used hybrid spatial interpolation methods based on OK and RK to mapping eight SCPs and the soil fertility in each soil layer (0–20 cm, 20–40 cm, and 40–60 cm) for 285 soil samples. The results indicated that RK outperformed OK for total nitrogen (TN), available potassium (AK), organic matter (OM) in 0–60 cm profile and available phosphorus (AP) in the 0–20 cm and 40–60 cm soil layers because RK considered the impact of terrain. The terrain factors, comprising the relative terrain position, slope, aspect, and relative elevation significantly affected the SCPs and spatial heterogeneity of fertility, where the vegetation cover types determined the average SCPs to some extent. On average, the six SCPs (except total potassium and AP) and the fertility decreased as the soil layer depth increased. Ten vegetation cover types comprising broadleaved mixed natural forest (BM), cultivated land (CL), economic forest (EF), grassland (GL), Platycladus orientalis natural forest (PON), Platycladus orientalis plantation (POP), Pinus tabuliformis plantation (PT), Quercus wutaishanica natural forest (QW), Robinia pseudoacacia plantation (RP), and Shrubwood (SW) were associated with significant differences in TN, OM, AN, AP, and AK, across the three soil layers. QW, PON, and BM also had higher content of TN, OM, AN, and AK contents than the other vegetation cover types. There were small differences in TK, AK, and pH among the 10 vegetation cover types. We concluded that AN, TN, and OM are the limiting factors of soil comprehensive fertility in this region. These results improve understanding of the spatial mapping, influence and limiting factors of SCPs and their fertility at urban forest scales.
Citation: Zhang X, Zhang F, Wang D, Fan J, Hu Y, Kang H, et al. (2018) Effects of vegetation, terrain and soil layer depth on eight soil chemical properties and soil fertility based on hybrid methods at urban forest scale in a typical loess hilly region of China. PLoS ONE 13(10): e0205661. https://doi.org/10.1371/journal.pone.0205661
Editor: Suzannah Rutherford, Fred Hutchinson Cancer Research Center, UNITED STATES
Received: February 13, 2018; Accepted: September 28, 2018; Published: October 18, 2018
Copyright: © 2018 Zhang et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the paper and its Supporting Information files.
Funding: This study was financially supported by the National 12th Five-Year Scientific and Technological Support Plan (grant no. 2015BAD07B06), and the Nonprofit Industry Special Research Fund of the State Forestry Administration of China (grant no. 201104045).
Competing interests: The authors have declared that no competing interests exist.
Introduction
Soil plays an essential role in the biosphere by governing plant productivity, organic matter (OM) degradation, and nutrient cycles [1]. Soil fertility is one of the major drivers of ecological processes, and thus it is frequently investigated in ecological research [2]. The eight soil chemical properties (SCPs), comprising total nitrogen (TN), total potassium (TK), total potassium (TP), available nitrogen (AN), available phosphorus (AP), available potassium (AK) and organic matter (OM) are important chemical components of soil fertility. Hence, spatially continuous mapping of these soil chemical properties and soil fertility is required to facilitate the sustainable management of land resources in precision agriculture and forestry [3–5], while it is also helpful to understand the belowground food webs [6], plant species distributions [7], and other factors. However, abundant observations of soil properties cannot always be obtained easily across a large landscape because of cost and time constraints on soil sampling and analysis [8]. Therefore, spatial interpolation is commonly used to generate soil property maps from discrete point-based data [9]. Previous studies have shown that auxiliary variables are important for predicting soil properties [10]. In recent years, the availability of high-resolution topographical data has provided ancillary variables for accurately mapping soil chemical properties [11–14]. The interpolation method employed critically affects the accuracy of interpolation. Among the existing interpolation techniques, the two commonly used methods are ordinary kriging (OK) [3,12,15,16] and regression kriging (RK) [8,16–19]. RK performs better than OK because its uses auxiliary variables, as well as reducing the number of observations needed for target variables [20]. Previous studies have investigated the effects of topography and the dominant trees species on the spatial distribution of soil physicochemical properties [3,12,21–23]. Terrain is one of the main factors that affect the soil C, N, and P contents at the landscape scale [3,21,24]. In recent decades, many studies have shown that a geostatistical approach based on the integration of terrain factors is an effective tool for accurately predicting the spatial distribution of soil chemical properties [3,15,21,25–29]. Fraterrigo et al. [30] showed that vegetation cover types have persistent, long-term effects on the spatial heterogeneity of soil resources, which may not be detectable when the values are equalized across sites. Differences in the distribution and supply of soil chemical properties could alter the composition and diversity of forest ecosystems by interacting with the patterns of variability in the plant and heterotrophic organisms. These activities may continue to influence the distributions of soil nutrients by altering their spatial heterogeneity in ecologically sensitive regions, such as, the Loess Plateau in China (LPC).
The LPC is well known due to the presence of severe soil erosion and ecosystem degradation, which have resulted in great soil nutrient losses and extreme terrain conditions [31–33]. Spatial information about soil nutrients and its fertility is required to understand and manage loess hilly ecosystems [24,34]. Previous studies in this area mainly focused on agricultural systems [35–37], whereas the continuous spatial distributions of these soil chemical properties and fertility, as well as the effects of the vegetation, terrain, and soil layer depth remain unclear in urban forest ecosystems. Thus, in order to determine the spatial distributions of the main soil chemical properties and their fertility, we investigated Yan’an urban forest as a real-world case study.
Urban forests are integral components of urban ecosystems and they can generate significant ecosystem services. Yan’an urban forest is coupled of artificial and natural forest ecosystem, which comprises suburban forest (including grassland), urban green spaces, and street trees. This region has been selected as one of the pioneer demonstration areas for the large-scale ecological restoration project known as “Grain for Green” in China in 1999 [38–41]. After ecological restoration for 15 years, the forest coverage of Yan’an urban forest increased significantly from 36.6% to 65.8%, and which effectively improved the ecological functions and services of Yan'an urban forest, such as offsetting carbon emissions, removing air pollutants, regulating the microclimate, allowing recreation [42], and mitigating urban heat islands [38]. The annual total value of these services is approximately 10.02 billion Chinese yuan [43].
In this study, in order to compare the performance of OK and RK in SCPs spatial predication, and to reveal the influence of vegetation, terrain and soil layer depth on eight soil chemical properties and soil fertility at urban forest scale, Ordinary Kriging (OK), Regression Kriging (RK), and an improved Nemerow index were used to mapping and assess soil chemical properties and their fertility, based on field vegetation investigations and the determination of soil chemical properties from 855 soil samples in 95 sample plots distributed evenly in the study area.
Materials and methods
Study area
According to the concept of urban forest and the local forestry practice [38,43], nine towns near (within 30 km) the center of Yan’an city, in the north of Shaanxi province, China, namely, Zaoyuan, Qiaogou, Chuankou, Liqu, Liulin, Yaodian, Wanhuashan, Hezhuangping, Nanniwan were identified as the spatial range of Yan'an urban forest (109°11′ to 109°47′ N, 36°11′ to 36°46′ E, elevation = 815–1466 m, total area = 1545.05 km2), which located in the hilly and gully region of the Loess Plateau (Fig 1). The main characteristics of Yan'an urban forest are as follows: (a) By the end of 2015, the total area and coverage rate of forests and trees in the study area was 1016.64 km2 (suburb forests = 998.76 km2, urban green spaces = 16.73 km2, street trees = 1.15 km2) and 65.8%, respectively. (b) The soil parent material is relatively homogeneous, the soil type is mainly loessial soil and the thickness ranges from approximately 50 m to 200 m [44], the soil has been deposited by the wind since the Quaternary period [45]. (c) However, the water retention properties of loessial soil are very poor because of its low anti-scourability [43]. The landforms of this area exhibit significant topographic variability, which corresponds to the different developmental stages and patterns of the Loess Plateau, including river valley, loess tableland, loess ridge, loess hill and loess residual tableland [44,46]. The river valley areas are relatively flat and solid, hence, the overwhelming majority of residents and human activities are mainly concentrated in this region. Because the other landforms have strong ecological fragility, state and local forestry sectors had proposed a series of protective measures, such as “Close Hillsides to Facilitate Afforestation”, “Returning Sloping Cultivated Land to Forest and Grasslands”, “Tending of Existing Woodlands”, and “Grazing Prohibition for Existing Grasslands”, since 1999. Which have effectively controlled the disturbance to Yan’an urban forest, especially the soils. (d) This area has a typical semiarid continental climate with average annual rainfall of approximately 470 mm, over 65% occurs during June and September. (e) The northern region mainly consists of Robinia pseudoacacia plantation and Platycladus orientalis plantation, and a few of economic forests, i.e., apple (Malus pumila), walnut (Juglans regia), etc.. While the southern region mainly covered by Quercus wutaishanica natural forest. The dominant tree species in the study area were listed in the Supporting Information Section (S1 Table and S2 Table).
Maps of the study area showing: (a) its location in China and; (b) the spatial distribution of 95 sample plots (black filled squares: modeling dataset, n = 70; red filled squares: validation dataset, n = 25).
Technical details
The workflow and technical details of this study are shown in Fig 2.
Soil sampling and soil chemical analysis
In July and August during 2014 and 2015, 855 soil samples (95 sample plots × 3 layers × 3 positions) were collected from depths of 0 cm to 60 cm and divided into three layers, i.e., surface soils (0–20 cm depth), near-surface soils (20–40 cm depth), and subsurface soils (40–60 cm depth). Details of the 95 sample plots are shown in Table 1. We used a split core sampler with a diameter of 5 cm at each location after removing any litter. In each sample plot, we tested three soil sampling locations according to the slope position (up, middle, and down) along a diagonal path, where the soil samples from the three position were mixed by hand to obtain one homogeneous soil sample (Fig 3) [47]. Environmental factors (slope, aspect, slope position, altitude, and GPS geographical coordinates) were also recorded for each sample plot with using a hand-held geological compass and a Garmin GPS receiver. About 1.0 kg of soil was collected from the corresponding soil layer at each location and returned to the laboratory to dry them under indoor natural ventilation condition. Samples were then ground and sieved them before chemical analyses. According to relative conferences [48–51], we could get the following knowledge: comprehensive evaluation of urban soil fertility is an effective means to judge urban soil quality. Soil fertility is the essential characteristics of soil, and a comprehensive reflection of soil physical, chemical and biological characteristics, among which soil pH, soil organic matter, nitrogen (N), phosphorus (P) and potassium (K) and their available state are very important soil properties for urban trees. Hence, we selected these eight soil chemical properties (SCPs), i.e., soil pH, soil organic matter (OM), total nitrogen (TN), total phosphorus (TP), total potassium (K), available nitrogen (AN), available phosphorus (AP), and available potassium (AK) as the soil fertility parameters. Details of the determination of the eight soil chemical properties are given in Table 2.
Sample plot vegetation survey
The canopy closure, stem height (height of the first major branch), tree height, diameter at breast height (DBH), crown width, plant community structure, and health status were surveyed for the trees in each plot in 95 sample plots (20 m × 20 m) when the soil samples were collected (S1 Table and S2 Table). These parameters were determined according to the Chinese Forestry Standards “Observation Methodology for Long-term Forest Ecosystem Research” (LY/T 1952–2011). Our field surveys and soil sampling procedures were authorized by local forestry administration (Forestry Bureau of Baota District in Yan’an City). Our study only involved soils and plants, and no humans or animals, and did not involve endangered or protected species.
Semivariance model optimization and spatial interpolation
In this study, empirical semivariogram values were obtained for the eight soil chemical properties using Eq 1:
(1)
where γ(h) is the sample semivariance between all observations Z(Xi), Z is the measured value at a particular location, N(h) represents the number of paired data at distance h, and (h) is the lag distance that separates the total numbers of data pairs. The semivariogram may be fitted with spherical, exponential, Gaussian, or linear models (Eqs 2–5, respectively):
(2)
(3)
(4)
(5)
where C0 is the nugget (N), C + C0 is the sill (S), and a is the correlation length. The detailed meanings of semivariogram parameters, including a, C0, C and h were explained in relevant literatures [59–61].
2.5.1 Ordinary kriging (OK).
Predictions are usually obtained by calculating some weighted average of the observations and the interpolation procedure is as follows, Eq 6:
(6)
where
is the predicted value of the target variable at an unvisited location S0 given its map coordinates, the sample data Z(S1), Z(S2),…, Z(Sn), and their coordinates. The weights λi are selected in order to minimize the prediction error variance, thereby yielding weights that depend on the spatial autocorrelation structure of the variable.
Regression kriging (RK).
Predictions were obtained by modeling the relationships between the target and auxiliary environmental variables at the sample locations, and by applying these relationships to unvisited locations by using the known values of the auxiliary variables at these locations. In order to obtain spatial predictions of the soil chemical properties with regression kriging, the usual auxiliary environmental predictors employed comprised the land surface parameters, remote sensing images [13,18], and geological information [8], soil data [19], and land-use maps [18]. In this study, we used terrain factors as auxiliary environmental predictors, as described in detail in Table 2. The principles of mathematical principles are shown in Eq 7 [62,63]:
(7)
where qk are the estimated regression coefficients, p is the number of predictors or auxiliary variables, qk are the estimated drift model coefficients, qk(S0) are the values of the auxiliary variables at the target location,
is the estimated intercept, λi are kriging weights determined by the spatial dependence structure of the residual, and e(Si) is the residual at location Si. The regression coefficients
were estimated from the sample using a fitting method. In this study, we used the stepwise ordinary least squares method to estimate
with collinearity diagnostics in SAS.
is the fitted drift, and
is the interpolated residual. The detailed computational process employed for making regression kriging predictions was described in previous studies [18,19].
Auxiliary terrain variables
Topography controls the flow of water, solutes, and sediments, thereby affecting soil development and the formation of the typical patterns of spatially distributed soil properties [64]. We calculated the relative elevation (Hr), slope (β), aspect (A), sunny slope (cosA), shady slope (sinA), terrain wetness index (TWI), stream power index (SPI), vertical curvature (Cv), horizontal curvature (Ch), range of change in the elevation (QFD), macroscopic information on the surface form (M), and relative position index (RPI) based on the digital elevation model at a spatial resolution of 5 m according to the appropriate computing methods in ArcGIS 10.0 and SimDTA-V1.0.3 [65–67]. Further details of the computation of the terrain factors are given in Table 3.
Model validation
We performed leave-one-out cross-validation [3,24–26,28]. This process was repeated for all of the observations. Three standard indices comprising the mean error (ME), mean relative error (MRE), and root mean squared error (RMSE) were used to compare the accuracy of interpolation. These indices were calculated as follows with Eqs 8, 9 and 10 [3,24,25,73]:
(8)
(9)
(10)
where m is the number of points, Zk is the observed content of the kth measurement, and
is the predicted value of the kth measurement. ME is a measure of the bias of the interpolation, which should be close to zero for unbiased methods, whereas MRE and RMSE are measures of prediction accuracy, which should be as low as possible.
Soil fertility assessment
The assessment of the enrichment and the paucity for every single soil chemical property.
The time-consuming collection of detailed objective soil content measures is justified when biophysical analysis is warranted. The classification criteria for the enrichment and the paucity of the eight soil chemical properties were assigned according to the second soil survey in Shaanxi province [48], as shown in Table 4.
Comprehensive soil fertility assessment based on the Nemerow index.
In order to eliminate the dimensional differences among the soil properties, the eight soil chemical properties, comprising, AN, AP, AK, TN, TP, TK, OM, and pH, were tested and nondimensionalized according to the following computational formulae, Eq 11:
(11)
where Fi is the nondimensionalized fertility index of i soil properties, Ci are the measured values of i soil properties estimated by ordinary kriging or regression kriging at a spatial resolution of 5 m, and xa, xc, and xp are the indices for the eight soil chemical properties, as shown in Table 5.
The comprehensive soil fertility index (F), was calculated as follows:
(12)
where
is the average value of each fertility index, Fiminis the minimum value of each fertility index, and n is the index number [49,50]. i.e., n = 8 in this study. Soil fertility grading was based on the method of Kan and Wu (1994), as shown in Table 6 [50]. We improved Nemerow index (F) and replaced the maximum value for a single factor in the original Nemerow index (F0), which is mainly used for soil pollution assessment, with the minimum value of single factor based on the Liebig's law of the minimum. In the improved Nemerow index, additional amendments (n –1)/n are used to increase the credibility of the evaluation.
Results
General normality analysis and optimization of semivariograms parameters for soil chemical properties
Table 7 shows the descriptive statistics (maximum, minimum, mean, standard deviation, coefficient of variation, skewness and kurtosis) obtained from the modeling dataset for the eight soil chemical properties at three soil depths. In general, the average value of each of the soil chemical properties decreased as the soil depth increased in the 0–60 cm soil profile with 20 cm intervals (Table 7). The skewness is a statistical index that indicates the degree of the symmetrical distribution for data. The skewness of a normal distribution is zero, and the tail length is symmetrical on both sides. If the skewness is positive, the data are relatively scattered on the right side of the mean. Otherwise, the data on the left side of the mean are more diffuse [74]. The raw data of TN (40–60 cm), TP (40–60 cm), TK (40–60 cm), and pH were less than one, so these data all conformed to normal distributions and they could be used to directly the optimize relevant parameters (nugget, partial sill, sill, range, and nugget/sill) for the semi-variance functions. By contrast, a natural logarithmic transformation should be considered where the data’s skewness is larger than 1.0 [3]. Logarithmic transformations were performed for TN (0–20 cm, 20–40 cm), TP (0–20 cm, 20–40 cm), TK (0–20 cm, 20–40 cm), AN, AP, AK, and OM to make the data suitable for optimizing the semi-variance function model. The majority of measured soil chemical properties exhibited moderate spatial dependency according to the N/S value in semi-variance function models, where specific parameters used in this study are shown in S3 Table, S1 and S2 Figs.
Effects of terrain factors on eight soil chemical properties in three soil layers
The multiple linear stepwise regression (MLSR) models for different soil chemical properties are presented in Table 8. All of the regression models were significant (P < 0.05). The predicted residual sum of squares (PRESS) and the determination coefficients (R2) of MLR models ranged from 0.03 to 37.22 and from 0.32 to 0.90, respectively. The terrain variables in bold in Table 8, such as, RPI, TWI, sinA, cosA, M, QFD, and TWI had great impact on the soil chemical properties in the corresponding MLR models. RPI had significant negative correlations with TN in the 0–40 cm soil layer, with AP in the 0–20 cm soil layer, and with OM in the 0–60 cm soil layer, where these correlations became weaker as the soil depth increased. sinA had significant negative correlations with TN in the 0–20 cm soil layer, and AN in the 0–40 cm soil layer. M had significant positive correlations with AN and AP in the 20–60 cm soil layer, and with AK in the 0–20 cm soil layer, but a significant negative correlation with OM in the 0–20 cm soil layer. QFD had significant positive correlations with AP and AK in the 40–60 cm soil layer, and with OM in the 0–20 cm soli layer. TWI had a significant positive correlation with TN in the 20–40 cm soil layer, but a significant negative correlation with OM in the 0–20 cm soil layer.
Comparison of two spatial interpolation methods for eight soil chemical properties at the urban forest scale
Validation of predictive accuracy using ordinary kriging and ordinary kriging for the eight soil chemical properties.
The predictions accuracy are shown in Table 9. The variogram for the soil chemical properties had the lowest nugget (C0) variance for TN (0–20 cm depth), followed by TN (40–60 cm depth) and AK (0–20 cm depth) using the OK method. The nugget value (1.872) for TK in the 40–60 cm soil layer was the largest compared with the other soil chemical properties, thereby demonstrating the great variations in this value over small distances. The small nugget value for almost all of the soil chemical properties indicated that the sampling density was adequate for determining the spatial structure, e.g., TN in the soil layers of 0–20 cm and 40–60 cm soil layers, and AK in the 0–20 cm soil layer. The range of the soil chemical properties in the three soil layers varied with the distance from 1290 m to 71100 m (S1 and S2 Figs). The predictive accuracies of ordinary kriging and regression kriging performed were good based on the lower ME, MRE, and RMSE. Our results showed that regression kriging performed better than ordinary kriging for TN (0–20 cm and 40–60 cm), OM (0–20 cm and 20–40 cm), and AP (20–40 cm and 40–60 cm). By contrast, ordinary kriging performed better than regression kriging for AN, AK, TP, TK, and pH of in three soil layers, and TN, OM, and AP of in other soil layers (Table 9).
Comparison of spatial distribution mapping for the eight soil chemical properties using ordinary kriging and regression kriging.
The spatial mapping technique was employed to assess the extent and magnitude of spatial heterogeneity of the soil chemical properties and fertility, especially soil nutrients and their deficiencies. Fig 4 shows that the general spatial distributions of the eight soil chemical properties determined by the two kriging interpolation methods were similar in the three soil layers. However, there were some differences between the two predictions. Firstly, the ranges of the predictions obtained by regression kriging were larger than those using ordinary kriging. Second, the vector polygons were more fragmented in the regression kriging maps, which were more sensitive to the variations in terrain and vegetation. The best kriging interpolation method for specific soil chemical properties are indicated in Table 9 (ninth column), mainly based on the topographical features, vegetation cover types, prediction accuracy (Table 7), and the continuity of the spatial distribution maps (Fig 4).
(a–f), available nitrogen; (g–l), available phosphorus; (m–r), available potassium; (s–x), total nitrogen; (y, z, aa–ad), total phosphorus; (ae–aj), total potassiumTK; (ak–ap), organic matter; and (aq–aw), pH.
Fig 4A, Fig 4B and 4E indicates that AN contents were generally low in the three soil layers, and AN was highly deficient at the margin of the study area with steep terrain. Fig 4G, Fig 4J and 4L indicates that the AP distributions were similar to these of AN. However, the AP values in the valley landform, especially, near Yanhe River, were higher than those in the other areas, mainly the river carried sediments containing phosphorus. AK was mainly above average in the surface and near-surface layers (Fig 4M and 4O). The AK contents of the subsurface layer were mainly low, and below average (Fig 4Q). In general, AK was highly deficient in the near-surface and subsurface layer soils in the low vegetation cover areas along the Yanhe River. Fig 4T indicates that the TN contents of the surface layers were generally low (accounting for 26.21% of the study area), below average (accounting for 30.69%), or above average (accounting for 23.55%). The TN contents of the near-surface layers were mainly low (accounting for 66.21%) or very low (accounting for 25.31%) (Fig 4U). The TN contents of the subsurface layers were mainly very low (accounting for 57.60%) or low (accounting for 34.17%). Spatially, the lower TN contents were usually distributed in farmlands before the implementation of the “Grain for Green” ecological restoration project in China. Fig 4Y, Fig 4Z, Fig 4Aa, Fig 4Ab, Fig 4Ac and 4Ad indicated that the TP contents of the surface layer, near-surface layer, and subsurface layer with predication by ordinary kriging were generally in the ranges of 0.39–0.93 g/kg, 0.34–0.87 g/kg, and 0.34–0.89 g/kg, respectively, where there were four grades comprised below average, above average, high, and very high,. In the whole study area, the soil TP contents in the 0–60 cm depth layer were sufficient for the existing agriculture and forestry requirements. Fig 4Ae, Fig 4Af, Fig 4Ag, Fig 4Ah, Fig 4Ai and 4Aj showed that the TK contents of the surface layer, near-surface layer, and subsurface layer were mostly in the ranges of 15.01–25.00 g/kg, 15.50–23.06 g/kg, and 10.05–15.39 g/kg, respectively, where most were above average. In the whole study area, the TK contents in the 0–60 cm depth layer were sufficient for the existing agriculture and forestry requirements. Fig 4Al, Fig 4Am, Fig 4An, Fig 4Ao and 4Ap shows that the OM range and contents decreased as the soil depth increased. In the whole study area, the OM contents in the 0–60 cm depth layer were insufficient for the existing agriculture and forestry requirements. Fig 4Ar, Fig 4As and 4At showed that the pH values in the surface layer and near-surface layer were mostly in the range of 7.00–8.00, where the two grades were above average (accounting for 93.53% and 84.18%, respectively). The pH in the subsurface layer was usually high in the range of 7.00–8.30, where the three grades comprised above average, high, and very high, and these three levels accounting 98.97% in total (Fig 4Aw). In the whole study area, the pH in the 0–60 cm depth soil layer was rather alkaline for existing agriculture and forestry utilization.
Effects of vegetation cover types and soil layer depth on the eight soil chemical properties and the comprehensive fertility
The effects of the vegetation cover types and soil layer depth on the eight soil chemical properties and the comprehensive fertility are shown in Fig 5. First, the TN, OM, AN, AP and AK differed significantly in the 10 vegetation cover types across the three soil layers. In particular, sites with Quercus wutaishanica natural forest (QW), Platycladus orientalis natural forest (PON), and broadleaved mixed natural forest (BM) had higher TN, OM, AN, and AK contents than the other vegetation cover types. By contrast, there were small differences in TK, AK, and pH among the ten vegetation cover types. Second, the soil layer depth had different effects on the eight soil chemical properties. The contents of TN, OM, AK and pH contents with the corresponding vegetation cover types decreased as the soil depth increased. However, TK was larger in near-surface layer than the other two soil layers.
BM, Broadleaved mixed natural forest; CL, Cultivated land; EF, Economic forest; GL, Grassland; PON, Platycladus orientalis natural forest; POP, Platycladus orientalis plantation forest; PT, Pinus tabuliformis plantation forest; QW, Quercus wutaishanica natural forest; RP, Robinia pseudoacacia plantation forest; SW, Shrubwood.
Soil comprehensive fertility
The Fig 6 shows that the soil fertility index was the highest in the surface soil layer (0–20 cm) with the grassland (GL) because the soil chemical properties in most of the grassland areas were at the equilibrium state for natural cycling as artificial harvesting was performed and grazing was prohibited. The average F values for natural forests (PON, QW, BM, Platycladus orientalis plantation (POP), and shrubwood (SW)) were larger than 0.90, thereby indicating that the fertility at general level because of low human intervention in these ecosystems. The average F values for artificial vegetation types (cultivated land (CL), Robinia pseudoacacia plantation (RP), and economic forest (EF)) were less than 0.90, and thus their fertility was at the barren level due to major artificial harvesting from the CL and EF ecosystems. In addition, a dried soil layer was widespread in the RP forests due to their vigorous taproot systems and active transpiration, which reduced the plant diversity and slowed down the rate of litter decomposition on the topsoil. For the soil fertility index in near-surface soil layer (20–40 cm), the average F values for PON and QW were both 0.91 and at the general level because that PON and QW had both been near-natural forest for over 50 years so there was rich litters from the complex plant communities and their soil environmental conditions were suitable for the litter decomposition, while the soil void structure was better in the 0–40 cm soil layer than that under the other vegetation cover types. There were not obvious differences in the average F values among the other eight vegetation cover types because of their similar soil physical properties [75] The average F values for the subsurface soil layer (40–60 cm) were at the barren level, which may be attributed to the soil physical properties gradually homogenizing as the soil layer depth increased [76].
The abbreviations for the vegetation cover types were the same as those used in the Fig 5.
According to the selection results obtained by the kriging interpolation method (Table 9), the F values for the three soil layers were computed with ArcGIS 10.0, as shown in Fig 1B, Fig 7, Table 6, and Table 8. (1) Spatially, west of Nanniwan, east of Chuankou, and parts of Wanhuashan, Hezhuangping, and Yaodian were at the general level (0.90 ≤ F < 1.80) based on the perspective of the comprehensive fertility index (F). (2) In the vertical soil profile, the cover area percentage with the barren fertility level (F ≤ 0.90) tended to increase from the surface layer (0–20 cm depth) to the subsurface layer (40–60 cm depth). However, the cover area percentage with the general fertility level (0.90 ≤ F < 1.80) exhibited a declining trend. (3) The spatial distribution maps of soil comprehensive fertility (F index) clearly matched with the terrain texture, which may facilitate the efficiency of land utilization.
Spatial distribution maps for the soil comprehensive fertility index (F) integrated based on eight soil chemical properties in the three soil layers predicted using the high accuracy spatial interpolation method: (a) 0–20 cm soil layer, (b) 20–40 cm soil layer, and (c) 40–60 cm soil layer.
Discussion
Effects of vegetation cover types, terrain, and soil layer depth on the soil chemical properties
In this study, we found that the mixed forests (Q. wutaishanica + P. orientalis, P. hopeiensis + R. pseudoacacia, R. pseudoacacia + P. davidiana, etc.) had higher levels of AN, AP, AK, TN, TP, TK, and OM, compared with the pure plantations (S1 Table, and Fig 5). The average concentrations level of the single and comprehensive soil chemical properties in the forest stands with the three dominant tree species (R. pseudoacacia, Q. wutaishanica and P. orientalis) in S2 Table were higher than those with other tree species because of differences in the total number of plant species (TNPS) and DBH in S1 Table. Analysis using Pearson’s correlation coefficients based on the relationships between the soil chemical properties with TNPS and DBH (S4 and S5 Tables), showed that the TP contents in the 0–60 cm soil layer and TK in the 20–40 cm were negatively correlated with TNPS, whereas the TK and AN contents in 0–20 cm soil layer, and AN in the 40–60 cm were positively correlated with TNSP. In addition, the DBH was positively correlated with the concentration of TN in the 0–20 cm soil layer, AN and AK in all three soil layers, and OM and pH in the 0–40 cm soil layer. High vegetation cover and diversity (i.e., TNPS) are helpful for soil conservation [77]. Moreover, forest land can improve the availabilities of soil N and organic C, and R. pseudoacacia cou1d enhance the concentration of NO3– as well as the availabilities of soil TN and OM in the rhizosphere soil because of the N2-fixing functions of its roots [78]. The Hippophae rhamnoides mixed forests significantly increased the OM, TN, AP, and AK contents compared with the pure forests [79]. Different types (tree, shrub, and grass) and sources (artificial and natural) of vegetation (R. pseudoacacia, Caragana korshinskii, and grassland) strongly affected the soil AP, TN, NH4+-N, and pH levels [80]. We also found that the soils under natural vegetation (QW, PON, and BM) were accumulated more TN, OM, AN, and AK than the other vegetation cover types.
We found that the TN, TP, AN, AP, AK, and OM contents were highly variable in the 0–60 cm soil layers (CV ≥ 20%) due to complex topographical changes over a short scale in the study area, even within same land use type (S3 Fig). Therefore, this case limited the results to interpolate for a long distance. Terrain factors, including the relative elevation (Hr), slope (β), aspect (A), sin of aspect (sinA), and the relative position index (RPI) significantly affected the concentrations and spatial heterogeneity of the eight soil chemical properties (Table 8) because topography has important effects on the runoff potential and soil-forming in hilly landscapes [3]. Similar studies have also demonstrated that topographical factors, including elevation (H), slope (β), upslope contributing area (CA), terrain wetness index (TWI), and stream power index (SPI) have important effects on the redistribution of soil total carbon (TC), OM, N, and P across the landscape [3,24,35,80,81]. In conclusion, the TN, AN, AP and OM contents in 0–60 cm soil layers were significantly correlated with terrain factors (in bold and italic in Table 8), including RPI, M, QFD, and TWI.
We found that the TN, TP, AN, AP, AK, OM, and pH levels, and the soil fertility in Yan’an urban forest all decreased as the soil depth increased under various vegetation cover types (Figs 5 and 6), possibly due to differences in nutrient infiltration and absorption by soil colloids along the vertical profile [82,83], as well as variations in the microbial community [84] and litter [85]. The specific causal relationships require further investigation.
Single soil fertility and comprehensive fertility at the urban forest scale
To the best of our knowledge, there is a lack of criteria for synthetically assessing the fertility status of forest soils. Previous studies of agricultural soil fertility assessments mainly on relatively stable properties rather than unstable properties, because that the former usually exhibit greater variation. However, in this case study, the CV values for the unstable properties (AN, AP and AK) were slightly larger than (0.39–1.74 fold) that of the corresponding those for the relatively stable properties (TN, TP, and TK), but, they all close to similar range (17%–91%) and trend (relatively stable among different soil layers). Which resulted from relatively homogeneous soil parent material (Quaternary loess) and most soils were free from anthropic intervention for ten years. Furthermore, some previous studies also used kriging interpolation methods to predict and assess the unstable soil properties, such as, AK (including, water-soluble potassium and exchangeable potassium) [26,86], extractable phosphorus (EP) [86]. According to the Liebig's law of the minimum, we referenced the soil fertility assessments for cropping lands in agricultural ecosystems, based on an improved Nemerow index, where we replaced the maximum value for a single-factor in the original Nemerow index, and an additional amendments (n –1)/n were used to increase the credibility of the evaluation. Relatively stable properties and unstable properties were both integrated into our comprehensive soil fertility index (F). The differential grading of each soil chemical property made them compatible. Our conclusions are similar to these obtained in related studies conducted in the Loess Plateau region of China. The soil OM contents decreased as the soil depth increased and they were insufficient in most areas with low vegetation cover because of scarce litter in the surface layer. Moreover, the AN was deficient in the 0–60 cm soil profile, especially at the margins of the study area with steep terrain because of more acute soil erosion due to surface runoff [41]. Furthermore, the soil AP in the 0–60 layer was at a low level in 90% of areas surveyed. The soil TN levels were below average in 66% areas of tested, especially in farmland. Chen et al. (2017) and Liu et al. (2013) had found that the soil C, TN, and TP levels on the Loess Plateau were lower compared with the whole of China [37,87]. We found that the soil AK contents of the 0–60 cm profile were at the average level in most areas (more than 60%). The soil TP and TK were both sufficient in the 0–60 cm layer. The soil pH was alkaline in the 0–60 cm depth, mostly in the range of 7.00–8.30. Hence, the AN, TN, and OM were the limiting factors in our study area, which were consistent with the previous studies [37,87].The soil comprehensive fertility index (F) was higher in the southwest of the study area compared with other areas, possibly due to differences in vegetation (mainly natural broad-leaved) and climate (precipitation and temperature, S4 Fig) [5]. In subsequent studies, the characteristics of soil nutrient requirements of specific target plants should be determined to obtain more accurate outputs.
Conclusions
This case study based on the spatial mapping and the variations in the eight soil chemical properties (TN, TK, TP, AN, AP, AK, OM, and pH) as well as the soil fertility in the hilly gully Loess Plateau at the urban forest scale obtained three main findings. First, the majority of the soil chemical properties exhibited moderate spatial dependencies and they were suitable for ordinary kriging interpolation, whereas others with weak spatial dependencies required regression kriging interpolation with topographic factors (elevation, slope, aspect, the sin of aspect, relative position index, etc.) as auxiliary variables. Second, the concentrations of the eight soil chemical properties were significantly influenced by the vegetation cover types due to differences in TNPS and DBH at the sample-plot scale. Third, the AN, TN, and OM were the limiting factors in our study area, and which could be improved by natural broad-leaved forests (Q.wutaishanica forests, Betula platyphylla forests, etc.).
Supporting information
S1 Table. Overview of the 95 sample plots used in this study.
https://doi.org/10.1371/journal.pone.0205661.s001
(DOCX)
S2 Table. Importance values (%) for tree species in the 95 sample plots.
https://doi.org/10.1371/journal.pone.0205661.s002
(DOCX)
S3 Table. The optimized semivariograms parameters used in the OK and RK based on the modeling dataset.
https://doi.org/10.1371/journal.pone.0205661.s003
(DOCX)
S4 Table. Pearson correlation coefficients for the eight SCPs of each soil layers and the total number of plant species (TNPS).
https://doi.org/10.1371/journal.pone.0205661.s004
(DOCX)
S5 Table. Pearson correlation coefficients for the eight SCPs of each soil layers and the diameter at breast height (DBH).
https://doi.org/10.1371/journal.pone.0205661.s005
(DOCX)
S1 Fig. Semivariograms of the raw or log-transformed data for the eight SCPs using the OK interpolation method.
TN, TP, TK, AN, AP, AK, OM and pH are proxy for total nitrogen, total phosphorus, total potassium, available nitrogen, available phosphorus, available potassium and pH, respectively. The range data: 0–20, 20–40, 40–60 represents the depth of soil layer with 0–20 cm, 20–40 cm and 40–60 cm, respectively. These have the same meanings in the S2 Fig.
https://doi.org/10.1371/journal.pone.0205661.s006
(PDF)
S2 Fig. Semivariograms based on the interpolated residuals obtained by OLS for the eight SCPs using the RK interpolation method.
https://doi.org/10.1371/journal.pone.0205661.s007
(PDF)
S3 Fig. Spatial distribution maps for the auxiliary terrain variables used in the RK interpolation method.
(a) Hr, relative elevation; (b) β, slope; (c) A, aspect; (d) cosA; e) sinA; (f) Cv, slope of slope; (g) Ch, slope of aspect; (h) QFD, range of change in the elevation; (i) M, terrain roughness on the surface; (j) RPI, relative position index; (k) TWI, terrain wetness index; and (l) SPI, stream power index.
https://doi.org/10.1371/journal.pone.0205661.s008
(PDF)
S4 Fig. Spatial distribution maps of total annual precipitation (TAP) and mean annual air temperature (MAT) during 1951–2013.
https://doi.org/10.1371/journal.pone.0205661.s009
(PDF)
Acknowledgments
This study was financially supported by the National 12th Five-Year Scientific and Technological Support Plan (Grant no. 2015BAD07B06), and the Nonprofit Industry Special Research Fund of the State Forestry Administration of China (Grant no. 201104045). The authors wish to thank the Baota District Branch of Yan’an Municipal Bureau of Land and Resources for providing the digital elevation model at a spatial resolution of 5 m. The authors also thank the editor and anonymous reviewers for their insightful comments and suggestions. Hongke Hao provided technical guidance concerning the analysis of spatial interpolation.
References
- 1. Haileslassie A, Priess J, Veldkamp E, Teketay D, Lesschen JP. Assessment of soil nutrient depletion and its spatial variability on smallholders’ mixed farming systems in Ethiopia using partial versus full nutrient balances. Agriculture, Ecosystems & Environment. 2005; 108: 1–16. https://doi.org/10.1016/j.agee.2004.12.010.
- 2. Wigley BJ, Coetsee C, Hartshorn AS, Bond WJ. What do ecologists miss by not digging deep enough? Insights and methodological guidelines for assessing soil fertility status in ecological studies. Acta Oecologica. 2013; 51: 17–27. https://doi.org/10.1016/j.actao.2013.05.007.
- 3. Kumar S, Singh RP. Spatial distribution of soil nutrients in a watershed of Himalayan landscape using terrain attributes and geostatistical methods. Environmental Earth Sciences. 2016; 75: 473. https://doi.org/10.1007/s12665-015-5098-8.
- 4. Berazneva J, McBride L, Sheahan M, Güereña D. Empirical assessment of subjective and objective soil fertility metrics in east Africa: Implications for researchers and policy makers. World Development. 2018; 105: 367–382. https://doi.org/10.1016/j.worlddev.2017.12.009.
- 5. Ma YX, Minasny B, Wu CF. Mapping key soil properties to support agricultural production in Eastern China. Geoderma Regional. 2017; 10: 144–153. https://doi.org/10.1016/j.geodrs.2017.06.002.
- 6. Laliberte E, Kardol P, Didham RK, Teste FP, Turner BL, et al. Soil fertility shapes belowground food webs across a regional climate gradient. Ecol Lett. 2017; 20: 1273–1284. https://doi.org/10.1111/ele.12823. pmid:28853198
- 7. Copeland SM, Harrison SP. Shading and litter mediate the effects of soil fertility on the performance of an understorey herb. Ann Bot. 2016; 118: 1187–1198. https://doi.org/10.1093/aob/mcw172. pmid:27604279
- 8. Zhu Q, Lin HS. Comparing Ordinary Kriging and Regression Kriging for Soil Properties in Contrasting Landscapes. Pedosphere. 2010; 20: 594–606.
- 9. Schloeder CA, Zimmerman NE, Jacobs MJ. Comparison of methods for interpolating soil properties using limited data. Soil Science Society of America Journal. 2001; 65: 470–479.
- 10. Ahmed S, De Marsily G. Comparison of geostatistical methods for estimating transmissivity using data on transmissivity and specific capacity. Water Resources Research. 1987; 23: 1717–1737.
- 11. Xu YM, Smith SE, Grunwald S, Abd-Elrahman A, Wani SP. Incorporation of satellite remote sensing pan-sharpened imagery into digital soil prediction and mapping models to characterize soil property variability in small agricultural fields. ISPRS Journal of Photogrammetry and Remote Sensing. 2017; 123: 1–19. https://doi.org/10.1016/j.isprsjprs.2016.11.001.
- 12. Mulder VL, de Bruin S, Schaepman ME, Mayr TR. The use of remote sensing in soil and terrain mapping—A review. Geoderma. 2011; 162: 1–19. https://doi.org/10.1016/j.geoderma.2010.12.018.
- 13. Barnes EM, Sudduth KA, Hummel JW, Lesch SM, Corwin DL, et al. Remote- and ground-based sensor techniques to map soil properties. Photogrammetric Engineering & Romote Sensing. 2003; 69: 619–630.
- 14. Fitzpatrick BR, Lamb DW, Mengersen K. Ultrahigh Dimensional Variable Selection for Interpolation of Point Referenced Spatial Data: A Digital Soil Mapping Case Study. PLoS One. 2016; 11: e0162489. https://doi.org/10.1371/journal.pone.0162489. pmid:27603135
- 15. Liu R, Xu F, Yu W, Shi J, Zhang P, et al. Analysis of field-scale spatial correlations and variations of soil nutrients using geostatistics. Environ Monit Assess. 2016; 188: 126. https://doi.org/10.1007/s10661-016-5130-9. pmid:26832723
- 16. Vašát R, Pavlů L, Borůvka L, Drábek O, Nikodem A. Mapping the Topsoil pH and Humus Quality of Forest Soils in the North Bohemian Jizerske hory Mts. Region with Ordinary, Universal, and Regression Kriging: Cross-Validation Comparison. Soil & Water Res. 2013; 8: 97–104.
- 17. Odeh IOA, McBratney AB, Chittleborough DJ. Further results on prediction of soil properties from terrain attributes: heterotopic cokriging and regression-kriging. Geoderma. 1995; 67: 215–226.
- 18. Hengl T, Heuvelink GBM, Rossiter DG. About regression-kriging: From equations to case studies. Computers & Geosciences. 2007; 33: 1301–1315. https://doi.org/10.1016/j.cageo.2007.05.001.
- 19. Hengl T, Heuvelink GBM, Stein A. A generic framework for spatial prediction of soil variables based on regression-kriging. Geoderma. 2004; 120: 75–93. https://doi.org/10.1016/j.geoderma.2003.08.018.
- 20. Knotters M, Brus DJ, Oude Voshaar JH. A comparison of kriging, co-kriging and kriging combined with regression for spatial interpolation of horizon depth with censored observations. Geoderma. 1995; 67: 227–246.
- 21. Ceddia MB, Vieira SR, Villela ALO, Mota LdS, Anjos LHCd, et al. Topography and spatial variability of soil physical properties. Scientia Agricola. 2009; 66: 338–352. https://doi.org/10.1590/s0103-90162009000300009.
- 22. Chang RY, Fu BJ, Liu GH, Yao XL, Wang S. Effects of soil physicochemical properties and stand age on fine root biomass and vertical distribution of plantation forests in the Loess Plateau of China. Ecological Research. 2012; 27: 827–836. https://doi.org/10.1007/s11284-012-0958-0.
- 23. Conforti M, Lucà F, Scarciglia F, Matteucci G, Buttafuoco G. Soil carbon stock in relation to soil properties and landscape position in a forest ecosystem of southern Italy (Calabria region). Catena. 2016; 144: 23–33. https://doi.org/10.1016/j.catena.2016.04.023.
- 24. Jeong G, Oeverdieck H, Park SJ, Huwe B, Ließ M. Spatial soil nutrients prediction using three supervised learning methods for assessment of land potentials in complex terrain. Catena. 2017; 154: 73–84. https://doi.org/10.1016/j.catena.2017.02.006.
- 25. Li QQ, Zhang X, Wang CQ, Li B, Gao XS, et al. Spatial prediction of soil nutrient in a hilly area using artificial neural network model combined with kriging. Archives of Agronomy and Soil Science. 2016; 62: 1541–1553. https://doi.org/10.1080/03650340.2016.1154543.
- 26. Chatterjee S, Santra P, Majumdar K, Ghosh D, Das I, et al. Geostatistical approach for management of soil nutrients with special emphasis on different forms of potassium considering their spatial variation in intensive cropping system of West Bengal, India. Environ Monit Assess. 2015; 187: 183. https://doi.org/10.1007/s10661-015-4414-9. pmid:25773896
- 27. Yuan F, Wu JG, Li A, Rowe H, Bai YF, et al. Spatial patterns of soil nutrients, plant diversity, and aboveground biomass in the Inner Mongolia grassland: before and after a biodiversity removal experiment. Landscape Ecology. 2015; 30: 1737–1750. https://doi.org/10.1007/s10980-015-0154-z.
- 28. Mudrak EL, Schafer JL, Fuentes-Ramirez A, Holzapfel C, Moloney K. Predictive modeling of spatial patterns of soil nutrients related to fertility islands. Landscape Ecology. 2014; 29: 491–505. https://doi.org/10.1007/s10980-013-9979-5.
- 29. Zhang ZQ, Yu DS, Shi XZ, Wang N, Zhang GX. Priority selection rating of sampling density and interpolation method for detecting the spatial variability of soil organic carbon in China. Environmental Earth Sciences. 2014; 73: 2287–2297. https://doi.org/10.1007/s12665-014-3580-3.
- 30. Fraterrigo JM, Turner MG, Pearson SM, Dixon P. Effects of past land use on spatial heterogeneity of soil nutrients in southern appalachian forests. Ecological Monographs. 2005; 75: 16.
- 31. Xiong LY, Tang GA, Strobl J, Zhu AX. Paleotopographic controls on loess deposition in the Loess Plateau of China. Earth Surface Processes and Landforms. 2016; 41: 1155–1168. https://doi.org/10.1002/esp.3883.
- 32. Gong J, Chen LD, Fu BJ, Huang Y, Huang Z, et al. Effect of land use on soil nutrients in the loess hilly area of the Loess Plateau, China. Land Degradation & Development. 2006; 17: 453–465. https://doi.org/10.1002/ldr.701.
- 33. Fu BJ, Wang S, Liu Y, Liu JB, Liang W, et al. Hydrogeomorphic Ecosystem Responses to Natural and Anthropogenic Changes in the Loess Plateau of China. Annual Review of Earth and Planetary Sciences. 2017; 45: 223–243 https://doi.org/10.1146/annurev-earth-063016020552.
- 34. Tian Q, Taniguchi T, Shi WY, Li G, Yamanaka N, et al. Land-use types and soil chemical properties influence soil microbial communities in the semiarid Loess Plateau region in China. Scientific Reports. 2017; 7: 45289. https://doi.org/10.1038/srep45289. pmid:28349918
- 35. Lü JL, Yan MJ, Song BL, Guan JH, Shi WY, et al. Ecological stoichiometry characteristics of soil carbon, nitrogen, and phosphorus in an oak forest and a black locust plantation in the Loess hilly region. Chinese Acta Ecologica Sinica. 2017; 37: 3385–3393. https://doi.org/10.5846/stxb201603120437.
- 36. Wang J, Fu BJ, Qiu Y, Chen LD. Soil nutrients in relation to land use and landscape position in the semi-arid small catchment on the loess plateau in China. Journal of Arid Environments. 2001; 48: 537–550. https://doi.org/10.1006/jare.2000.0763.
- 37. Chen XJ, Hou FJ, Matthew C, He XZ. Soil C, N, and P Stocks Evaluation Under Major Land Uses on China’s Loess Plateau. Rangeland Ecology & Management. 2017; 70: 341–347. https://doi.org/10.1016/j.rama.2016.10.005.
- 38. Zhang XP, Wang DX, Hao HK, Zhang FF, Hu YN. Effects of Land Use/Cover Changes and Urban Forest Configuration on Urban Heat Islands in a Loess Hilly Region: Case Study Based on Yan’an City, China. International Journal of Environmental Research and Public Health. 2017; 14: 840. https://doi.org/10.3390/ijerph14080840.
- 39. Lü YH, Fu BJ, Wei W, Yu XB, Sun RH. Major ecosystems in China: dynamics and challenges for sustainable management. Environ Manage. 2011; 48: 13–27. https://doi.org/10.1007/s00267-011-9684-6. pmid:21553106
- 40. Lü YH, Fu BJ, Feng XM, Zeng Y, Liu Y, et al. A policy-driven large scale ecological restoration: quantifying ecosystem services changes in the Loess Plateau of China. Plos One. 2012: e31782. https://doi.org/10.1371/journal.pone.0031782.g001. pmid:22359628
- 41. Wang JL, Lü YH, Zeng Y, Zhao ZJ, Zhang LW, et al. Spatial heterogeneous response of land use and landscape functions to ecological restoration: the case of the Chinese loess hilly region. Environmental Earth Sciences. 2014; 72: 2683–2696. https://doi.org/10.1007/s12665-014-3175-z.
- 42. Jim CY, Chen WY. Ecosystem services and valuation of urban forests in China. Cities. 2009; 26: 187–194. https://doi.org/10.1016/j.cities.2009.03.003.
- 43.
Hu YN Assessment of urban forest ecosystem services value in Yan'an, China [doctoral]. Yangling,Shaanxi,China: Northwest A&F University. 160 p. 2015.
- 44. Zhao WF, Xiong LY, Ding H, Tang GA. Automatic recognition of loess landforms using Random Forest method. Journal of Mountain Science. 2017; 14: 885–897. https://doi.org/10.1007/s11629-016-4320-9.
- 45. Gao L, Shao MA. The interpolation accuracy for seven soil properties at various sampling scales on the Loess Plateau, China. Journal of Soils and Sediments. 2011; 12: 128–142. https://doi.org/10.1007/s11368-011-0438-0.
- 46. Wu B, Guo SH, Li BL, Zhang LY. Draft of soil environmental function regionalization of China. Chinese Journal of Applied Ecology. 2018; 29. https://doi.org/10.13287/j.1001-9332.201803.038.
- 47. Kleinman PJA, Bryant RB, Pimentel D. Assessing ecological sustainability of slash-and-burn agriculture through soil fertility indicators. Agronomy Journal 1996; 88: 122–127. https://doi.org/10.2134/agronj1996.00021962008800020002x.
- 48.
Zhao YT Spatial characteristics and changes of soil nutrients in cultivated land of Guanzhong region in Shaanxi province based on GIS [Doctor]: Northwest A&F University. 2014.
- 49. Zhou W, Wang WJ, Zhang B, Xiao L, Lü HL, et al. Soil fertility evaluation for urban forests and green spaces in Changchun City. Acta Ecologica Sinica. 2017; 37: 1211–1220.
- 50. Kan WJ, Wu QT. A preliminary approach to the quantitative and comprehensive evaluation of soil fertility. Chinese Journal of Soil Science. 1994; 25: 245–2457.
- 51. Emamgholizadeh S, Shahsavani S, Eslami MA. Comparison of artificial neural networks, geographically weighted regression and Cokriging methods for predicting the spatial distribution of soil macronutrients (N, P, and K). Chinese Geographical Science. 2017; 27: 747–759. https://doi.org/10.1007/s11769-017-0906-6.
- 52.
Bremner JM, Mulvaney CS. Nitrogen-total. In: Page AL MR, Keeney DR (eds), editor. Methods of soil analysis,part 2—chemical and microbiological properties. Madison: ASA-SSSA. pp. 576–624; 1982.
- 53. Subbiah BV, Asija GL. A rapid procedure for estimation of available nitrogen in soils. Current Science. 1956; 25: 259–260. https://doi.org/10.2136/sssaj2001.1751.
- 54. Grewal JS, Kanwar JS. Forms of potassium in Punjab soils. Journal of the Indian Society of Soil Science. 1966; 14: 63–67.
- 55.
Gomez KA, Gomez AA. Stastical procedures for agricultural research. New York: John Wiley and Sons. 1976.
- 56.
Jackson ML. Soil Chemical Analysis. New Delhi: Prentice Hall of India Pvt. Ltd. 1973.
- 57. Walkley A. A critical examination of a rapid method for determining organic carbon in soils: effect of variations in digestion conditions and inorganic soil constituents. Soil Science. 1947; 63: 251–263.
- 58.
McLean EO. pH and lime requiremnets. In: Page ALE, editor. Methods of Soil Analysis. Madison, WI,USA: American Society of Agronomists Inc.; 1982.
- 59. Olea RA. A six-step practical approach to semivariogram modeling. Stochastic Environmental Research and Risk Assessment. 2006; 20: 307–318. https://doi.org/10.1007/s00477-005-0026-1.
- 60. Cambardella CA, Moorman TB, Novak JM, Parkin TB, Karlen DL, et al. Field-Scale Variability of Soil Properties in Central Iowa Soils. Soil Science Society of America Journal. 1994; 58: 1501–1511.
- 61. Qiu WW, Curtin D, Johnstone P, Beare M, Hernandez-Ramirez G. Small-Scale Spatial Variability of Plant Nutrients and Soil Organic Matter: An Arable Cropping Case Study. Communications in Soil Science and Plant Analysis. 2016; 47: 2189–2199. https://doi.org/10.1080/00103624.2016.1228945.
- 62. Breiman L, Friedman JH. Predicting Multivariate Responses in Multiple Linear Regression. Journal of the Royal Statistical Society Series B-Methodological. 1997; 59: 3–54.
- 63. Qiu Y, Fu BJ, Wang J, Chen LD. Spatiotemporal prediction of soil moisture content using multiple-linear regression in a small catchment of the Loess Plateau, China. Catena. 2003; 54: 173–195. https://doi.org/10.1016/s0341-8162(03)00064-x.
- 64. Li QQ, Yue T.X.;, Wang CQ, Zhang WJ, Yu Y, et al. Spatially distributed modeling of soil organic matter across China: An application of artificial neural network approach. Catena. 2013; 104: 210–218. https://doi.org/10.1016/j.catena.2012.11.012.
- 65. Fan HY, Qi PC, Huang TY. Study on three surface curvatures in GIS. Journal of Nanyang Normal University. 2011; 10: 66–69.
- 66. Qin CZ, Zhu AX, Shi X, Li BL, Pei T, et al. Quantification of spatial gradation of slope positions. Geomorphology. 2009; 110: 152–161. https://doi.org/10.1016/j.geomorph.2009.04.003.
- 67.
ESRI (2011) ArcGIS Desktop: Release 10. Redlands, CA. 2011. Environmental Systems Research Institute.
- 68. Chang KT, Tsai BW. The effect of DEM resolution on slope and aspect mapping. Cartography and geographic information systems. 1991; 18: 69–77. https://doi.org/10.1559/152304091783805626.
- 69. Qiu Y, Fu BJ, Wang J, Chen LD. Variablity of the soil Physical Properties on the Loess Plateau. Acta Geographica Sinica. 2002; 57: 587–594.
- 70. Zevenbergen LW, Thorne CR. Quantitative analysis of land surface topography. Quantitative analysis of land surface topography. 1987; 12: 47–56.
- 71. Beven KJ, Kirkby MJ. A physically-based, variable contributed area model of basin hydrology. Hydrological Sciences Bulletin des Sciences Hydrologiques. 1979; 24: 43–69.
- 72. Tarboton DG. A new method for the determination of flow directions and upslope areas in grid digital elevation models. Water Resources Research. 1997; 33: 309–319. https://doi.org/10.1029/96wr03137.
- 73. Vasu D, Singh SK, Sahu N, Tiwary P, Chandran P, et al. Assessment of spatial variability of soil properties using geospatial techniques for farm level nutrient management. Soil and Tillage Research. 2017; 169: 25–34. https://doi.org/10.1016/j.still.2017.01.006.
- 74. Joanes DN, Gill CA. Comparing Measures of Sample Skewness and Kurtosis. Journal of the Royal Statistical Society Series D (The Statistician). 1998; 47: 7.
- 75.
Cronan CS. Ecosystem Biogeochemistry Element Cycling in the Forest Landscape; Wagner M, Lambert S, editors. Switzerland, Cham: Springer. 2018.
- 76. Wang R, Liu WZ, Li Z. Physical properties of soils along a 10 m deep soil profile in loess tableland. Acta Pedologica Sinica. 2008; 45: 550–554.
- 77. Hou J, Fu BJ, Wang S, Zhu HX. Comprehensive analysis of relationship between vegetation attributes and soil erosion on hillslopes in the Loess Plateau of China. Environmental Earth Sciences. 2014; 72: 1721–1731. https://doi.org/10.1007/s12665-014-3076-1.
- 78. Hu B, Zhou MH, Dannenmann M, Saiz G, Simon J, et al. Comparison of nitrogen nutrition and soil carbon status of afforested stands established in degraded soil of the Loess Plateau, China. Forest Ecology and Management. 2017; 389: 46–58. https://doi.org/10.1016/j.foreco.2016.12.011.
- 79. Zhang JT, Chen TG. Effects of mixed Hippophae rhamnoides on community and soil in planted forests in the Eastern Loess Plateau, China. Ecological Engineering. 2007; 31: 115–121. https://doi.org/10.1016/j.ecoleng.2007.06.003.
- 80. Xu M, Zhang J, Liu GB, Yamanaka N. Soil properties in natural grassland, Caragana korshinskii planted shrubland, and Robinia pseudoacacia planted forest in gullies on the hilly Loess Plateau, China. Catena. 2014; 119: 116–124. https://doi.org/10.1016/j.catena.2014.03.016.
- 81. Hengl T, Leenaars JGB, Shepherd KD, Walsh MG, Heuvelink GBM, et al. Soil nutrient maps of Sub-Saharan Africa: assessment of soil nutrient content at 250 m spatial resolution using machine learning. Nutrient Cycling in Agroecosystems. 2017; 109: 77–102. https://doi.org/10.1007/s10705-017-9870-x.
- 82. Bell LW, Sparling B, Tenuta M, Entz MH. Soil profile carbon and nutrient stocks under long-term conventional and organic crop and alfalfa-crop rotations and re-established grassland. Agriculture, Ecosystems & Environment. 2012; 158: 156–163. https://doi.org/10.1016/j.agee.2012.06.006.
- 83. Martínez I, Chervet A, Weisskopf P, Sturny WG, Etana A, et al. Two decades of no-till in the Oberacker long-term field experiment: Part I. Crop yield, soil organic carbon and nutrient distribution in the soil profile. Soil and Tillage Research. 2016; 163: 141–151. https://doi.org/10.1016/j.still.2016.05.021.
- 84. Yao XD, Zhang NL, Zeng H, Wei W. Effects of soil depth and plant–soil interaction onmicrobial community in temperate grasslands of northern China. Science of the Total Environment. 2018; 630: 96–102. https://doi.org/10.1016/j.scitotenv.2018.02.155. pmid:29475117
- 85. An SS, Mentler A, Mayer H, Blum WEH. Soil aggregation, aggregate stability, organic carbon and nitrogen in different soil aggregate fractions under forest and shrub vegetation on the Loess Plateau, China. Catena. 2010; 81: 226–233. https://doi.org/10.1016/j.catena.2010.04.002.
- 86. Li XY, Li H, Liu HJ. Mapping the Spatial Variability of Soil Properties: A Comparative Study of Spatial Interpolation Methods in Northeast China. Applied Mechanics and Materials. 2014; 535: 483–488. https://doi.org/10.4028/www.scientific.net/AMM.535.483.
- 87. Liu ZP, Shao MA, Wang YQ. Spatial patterns of soil total nitrogen and soil total phosphorus across the entire Loess Plateau region of China. Geoderma. 2013; 197–198: 67–78. https://doi.org/10.1016/j.geoderma.2012.12.011.