Spatial variability of soil nitrogen in a hilly valley: Multiscale patterns and affecting factors

Science of the Total Environment 563–564 (2016) 10–18

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

Spatial variability of soil nitrogen in a hilly valley: Multiscale patterns and affecting factors Shirong Zhang a,⁎, Chunlan Xia a, Ting Li b, Chungui Wu c, Ouping Deng b, Qinmei Zhong a, Xiaoxun Xu a, Yun Li b, Yongxia Jia b a b c

College of Environmental Science, Sichuan Agricultural University, Wenjiang 611130, PR China College of Resources, Sichuan Agricultural University, Wenjiang 611130, PR China Agricultural Bureau in Shehong, Shehong 629200, PR China

H I G H L I G H T S

G R A P H I C A L

A B S T R A C T

• Spatial patterns of soil TN and AN among three hill types showed obvious difference. • Hill type with different heights was a key factor for spatial variabilities. • Effects of slope position, parent material, land use, etc. increase with upscaling.

a r t i c l e

i n f o

Article history: Received 13 February 2016 Received in revised form 15 April 2016 Accepted 16 April 2016 Available online xxxx Editor: Jay Gan Keywords: Total nitrogen Available nitrogen Hilly region Multiscale General linear model Affecting factor

a b s t r a c t Estimating the spatial distribution of soil nitrogen at different scales is crucial for improving soil nitrogen use efficiency and controlling nitrogen pollution. We evaluated the spatial variability of soil total nitrogen (TN) and available nitrogen (AN) in the Fujiang River Valley, a typical hilly region composed of low, medium and high hills in the central Sichuan Basin, China. We considered the two N forms at single hill, landscape and valley scales using a combined method of classical statistics, geostatistics and a geographic information system. The spatial patterns and grading areas of soil TN and AN were different among hill types and different scales. The percentages of higher grades of the two nitrogen forms decreased from low, medium to high hills. Hill type was a major factor determining the spatial variability of the two nitrogen forms across multiple scales in the valley. The main effects of general linear models indicated that the key affecting factors of soil TN and AN were hill type and fertilization at the single hill scale, hill type and soil type at the landscape scale, and hill type, slope position, parent material, soil type, land use and fertilization at the valley scale. Thus, the effects of these key factors on the two soil nitrogen forms became more significant with upscaling. © 2016 Elsevier B.V. All rights reserved.

1. Introduction

⁎ Corresponding author. E-mail address: [email protected] (S. Zhang).

http://dx.doi.org/10.1016/j.scitotenv.2016.04.111 0048-9697/© 2016 Elsevier B.V. All rights reserved.

Soil total nitrogen (TN) plays a key role in building soil fertility and enhancing soil productivity (Franzluebbers and Stuedemann, 2009). As an important part of soil TN, available nitrogen (AN) supplies plant

S. Zhang et al. / Science of the Total Environment 563–564 (2016) 10–18

nutrients during the growth period (Mengel et al., 2006). Nitrogen deficiency in soils impedes plant growth and leaf photosynthesis (Boussadia et al., 2010). However, excess nitrogen in soil is the largest contributor of non-point-source pollution (Basso et al., 2016) and may result in severe environmental issues, such as eutrophication, soil acidification and gaseous emissions (Oorts et al., 2007; Rode et al., 2009; Velthof et al., 2014). Soil nitrogen heterogeneity is caused by widely varying soil-forming factors including climate, topography, parent material, land use and human activity (Aubert et al., 2005; Basso et al., 2016). Therefore, knowledge of the spatial variability of soil nitrogen is indispensable in environmental monitoring and management (Foster et al., 2005; Córdova et al., 2012). In recent years, studies of the variation of soil nitrogen have benefited from the combination of geostatistics with global positioning system (GPS) and geographic information system (GIS) (Raines, 2002; Lamsal et al., 2006; Yamashita et al., 2010). Recent many studies have focused on the spatial variability inherent in soil nitrogen at a single scale (Dharmakeerthi et al., 2005; Momtaz et al., 2009; Jackson-Blake et al., 2012). Nevertheless, relatively little attention has been devoted to studying their spatial patterns in the same region at multiple scales (Yemefack et al., 2005; Oenema et al., 2010) although the various environmental conditions, land use and management may lead to highly diverse distribution patterns in different regions or at multiple spatial scales (Rode et al., 2009). The spatial patterns of soil nitrogen are the cumulative result of all acting soil-forming factors (Zeleke and Si, 2006). Among all factors affecting soil nitrogen heterogeneity, only a few are usually studied at one time in most literature. For example, Rodríguez et al. (2011) reported the effect of plant species on the spatial variability of the labile organic nitrogen and inorganic nitrogen in a forest soil with singlefactor analysis of variance (ANOVA). Wang et al. (2009) suggested that land use and topography were the dominant factors affecting TN by single-factor ANOVA and correlation analysis and Tremblay et al. (2011) evaluated the effect of nitrogen fertilizer rates on soil

11

nitrate-N using ANOVA. However, it remains unclear how the dominant factors together influence the variability of soil nitrogen (Aubert et al., 2005; Wang et al., 2009; Oenema et al., 2010). Consequently, it is essential to improve our understanding of the main effects of the multiple factors affecting the variability of soil TN and AN, especially by applying the univariate general linear model (GLM) across different spatial scales (Huang et al., 2007; Mairura et al., 2007; Cobo et al., 2010). Hills are one of the most extensive landforms in East Asia including south China (Wang et al., 2009), Vietnam (Schmitter et al., 2010) and Thailand, and cover up to 180 million ha (Aumtong et al., 2009). Despite the large areas of hill soils, only a limited number of researchers have examined the spatial variability of soil TN and AN in these regions (Yemefack et al., 2005; Oenema et al., 2010; Li et al., 2016). Therefore, it is essential to develop systematic studies into understanding the spatial patterns and crucial affecting factors in hilly regions at multiple scales (Zhang et al., 2007; Li et al., 2016). The Fujiang River Valley, as a typical hilly region of the central Sichuan Basin (Fig. 1), is one of the most vulnerable ecological regions in the upper Yangtze River Valley. This is mainly because of the high population density, erosion-prone soil, high cropping intensity and traditional management (He et al., 2007; Wang et al., 2009). In the valley, three hill topographies (low hill, medium hill and high hill) are present, and their geographical landscapes vary according to the assemblages of landforms, parent material, soil type, land use and management. However, the spatial patterns and dominant factors affecting soil N are unclear. Hence, we hypothesized that soil TN and AN show different spatial patterns and dominant affecting factors within a single valley at multiple scales. The objectives of this study were to: (1) characterize the spatial patterns of soil TN and AN using a combination of classical statistics, geostatistics, GPS and GIS at the single hill, landscape and valley scales; and (2) evaluate the main effects of the dominant affecting factors on soil TN and AN across multiple scales using univariate GLM.

Fig. 1. Sampling sites for the single hill, landscape and valley scales in the study area. Nine plots (A1–A9) were selected for single hill scale. Three areas (B1–B3) were selected for the landscape scale. Region C was the whole river valley.

12

S. Zhang et al. / Science of the Total Environment 563–564 (2016) 10–18

2. Material and methods 2.1. Study area The study area is located along the Fujiang River (105°13′–105°39′ E, 30°40′–31°10′ N) in Sichuan, China. The selected region covers an area of 1002.97 km2 (Fig. 1). The whole river valley (C), a typical hilly region of central Sichuan Basin, is a combination of low, medium and high hilly areas. A low hilly area (B1) lies in the southeast of the region and along the Fujiang River. A medium hilly area (B2) is positioned in the west of the study area, and a high hilly area (B3) is located in the north (Fig. 1). The elevations range from 256 to 619 m a.s.l. throughout the whole study region. The climate in the region is classified as moist subtropical. The mean annual rainfall and temperature from 1980 to 2009 were 934 mm and 17.2 °C, respectively. At the landscape scale, the three landform areas of low hill with relative heights from gully floor to hilly top b50 m (Zhou, 2007), medium hill with relative heights from 50 to 100 m and high hills with relative heights from 100 to 200 m occupy 39.36, 31.41 and 29.23% of the valley area, respectively. In each of the three hilly areas, three single hill plots (each containing a single hill and its near gully) were selected as study units at the single hill scale. Plots A1, A2 and A3 were located in the low hilly area, Plots A4, A5 and A6 in the medium hilly area, and Plots A7, A8 and A9 in the high hilly area (Fig. 1). 2.2. Geology and hydrology The Fujiang River Valley is located in the central part of the Sichuan Basin. The lithology is dominated by shale, sandstone and alluvium. There are three major tributaries including the Zijiang River, the Yangxi River and the Lulian River in the valley (Fig. 1).

According to field investigations, we categorized the inorganic fertilizer nitrogen rates into four classes: b150, 150–270, 271–330 and N331 kg ha−1 yr−1. Phosphatic and potassic fertilizers were applied as P2O5 of 150–185 kg ha−1 yr−1 and K2O of 75–135 kg ha−1 yr−1, respectively. No fertilizer was applied on the grasslands or forests.

2.5. Soil sampling and analysis The sampling sites were determined on the basis of hill types, parent material, dominant soil types, land use classes and management practices. At the landscape and valley scales, we applied a stratified sampling scheme to sampling sites. At the single hill scale, sampling sites were designed with a 25 × 25 m grid. The sampling grid size for the landscape scale was 640 m. The numbers of sampling points for low (B1), medium (B2) and high hill areas (B3) were 965, 788 and 669, respectively. Some sampling sites were offset 2–5 m for surface features such as field paths, ditches and other artificial constructions. Soil samples in the 0–20-cm soil layer were collected in 2007 and a GPS (GPS MAP 76, Garmin Ltd., USA) was used to georeference the sites. Soil samples were air-dried and passed through a 2-mm sieve. The TN concentration in soil was measured using the Kjeldahl method (Bremmer and Mulvaney, 1982). Soil AN concentration was determined using 2 mol L−1 NaOH (Cornfield, 1960). In brief, air-dried soil (2 g) was scattered into the outer ring of a cell with outer and inner chambers, and 2 mL of 2% boric acid was placed in the central chamber. Then, 5 mL of 2 mol L−1 sodium hydroxide was added into the outer cell without mixing with soil. After sealing the lid with gum acacia fixative, the cell was rotated gently to mix soil and sodium hydroxide, and incubated at 40 °C for 24 h. The ammonia absorbed by boric acid was determined by titration with 0.01 mol L−1 sulfuric acid using a methyl red– methylene blue indicator.

2.3. Soil types In the study area, the soil types are complex as they are influenced by the high diversity of landforms and parent materials. The soil parent materials are composed of residium and colluvium from thick sandstone and thin shale (TSA), residium and colluvium from thick shale and thin sandstone (TSH), and alluvium (AL). Major soil types are Cambosols, Primosols and Anthrosols (Gong et al., 2003). Typic Feaccumuli-Stagnic Anthrosols and Typic Fe-leachi-Stagnic Anthrosols, mainly developed from residium, colluvium and alluvium, widely distribute in the toeslopes and gullies. These soils have the finest texture of all the studied soils. Calcaric Purpli-Orthic Primosols and Carbonatic Udi-Orthic Primosols, derived from TSA and TSH, occur on the backslope and footslope of the hills. Calcaric Purpli-Udic Cambosols, mainly developed from TSH, are distributed at the tops of hills.

2.6. Statistical and geostatistical analysis 2.6.1. Conventional statistical analysis Data were subjected to descriptive analysis using Statistical Product and Service Solutions (SPSS) software 17.0 (SPSS Institute Inc., 2007). The mean, minimum and maximum, standard deviation, skewness and kurtosis for soil TN and AN were computed to describe the central and dispersion trend. We also conducted exploratory data analyses for outliers. We applied the one-dimensional Kolmogorov–Smirnov (K–S) test together with skewness and kurtosis values to evaluate the normality of datasets because asymmetry in the distribution of data has an important effect on the geostatistical analysis (Fu et al., 2010). Non-normal data were transformed to stabilize the variance. Then a normality test was conducted to confirm that the transformed data were now normally distributed.

2.4. Land use and farming system A mixed farming system that integrates cropping, grass and forest planting is adopted in the valley (integral agroforestry ecosystem), but cropping remains the main activity. Therefore, the main land use classes include paddy fields, dryland, grassland and forests across the study area. The paddy fields are distributed on the toeslopes of the hill and in gullies, and drylands on the hill slopes. Grassland occurs sporadically on the top and steep hillslopes, and forests are present on the top of hills or around the house. The main crops in cropland areas were wheat (Triticum aestivum), rape (Brassica napus), horsebean (Vicia faba), or pea (Pisum sativum) in winter and spring, and rice (Oryza sativa), maize (Zea mays), sweet potato (Ipomoea batatas) or peanut (Arachis hypogaea) in summer and autumn. On the grassland and forest areas, the dominant species were Plantago asiatica, Amaranthus hypochondriacus, Dactylis glomerata, Alnus cremastogyne, Morus alba and Metasequoia glyptostroboides.

2.6.2. General linear model The general linear model becomes even more useful when the analysis includes both numeric (interval level) and categorical (nominal level) variables, since both can be directly entered into the analysis, and SPSS will do any required dummy coding. The inclusion of both variables is advantageous because the method can then take into account the relationships of the predictor variables with the dependent variables (Park et al., 2005). In this study, the affecting factors of soil TN or AN including slope position, parent material, soil type, land use and N fertilizer rate levels were attributed to categorical variables. Therefore, we used univariate GLM of SPSS software 17 to test the main effect of the multiple affecting factors on soil TN or AN across the multiple scales. The Student–Newman–Keuls or Games–Howell tests were used for separating means at P b 0.05, regardless of whether the error variances of the dependent variable across groups were equal.

S. Zhang et al. / Science of the Total Environment 563–564 (2016) 10–18

2.6.3. Geostatistical analysis Geostatistical analysis uses the semivariogram to quantify the spatial variation of a regionalized variable (Matheron, 1963). If the variables showed global trends, their semivariograms were calculated using the residuals after the trend removal (Johnston et al., 2001). The semivariogram (γ(h)) of regionalized variables is defined by the following equation as: γ ðhÞ ¼

n 1 X ½Z ðxi Þ−Z ðxiþh Þ2 ; 2n i¼1

ð1Þ

where Z(xi) is the measure value for a soil property at the location of xi; n is the number of data pairs separated by lag distance h; and γ(h) represents the semivariogram for a distance h between observations Z(xi) and Z(xi + h). The semivariance analysis and kriging interpolation for soil TN and AN were performed on the ArcGIS 10.0 platform, and different theoretical semivariograms γ(h) were used to fit the calculated values. In this study, semivariogram model validation was conducted using the mean error (ME), root mean square standardized error (RMSSE) and coefficient of determination (R2). The best-fitted models were selected for soil TN and AN based on the minimum ME and RMSSE being close to 0 and 1 (Johnston et al., 2001), respectively, and the maximum R2. Information provided by the best-fitted model was used to analyze spatial structure and provided the input parameters for kriging interpolation. Where a dataset showed a global or local trend, kriging interpolation was conducted on the residuals after the trend was removed. The final prediction value on the ArcGIS10.0 platform includes the trend and ordinary kriging interpolation (Johnston et al., 2001). 3. Results and discussion 3.1. Descriptive statistics The skewness, kurtosis and K–S normal distribution tests showed that soil TN and AN in plots belonged to normal or log-normal distributions at the single hill scale (P N 0.05, Table 1). However, both variables were neither normally nor log-normally distributed (P b 0.05), except for soil TN in Area B3 (high hill area) and AN in Area B1 (low hill area) as the landscape and valley scales for these datasets showed a first-

13

order trend. After trend removal for each dataset, the residuals were normally distributed. The distribution difference of the variables at multiple scales might be because the soil-forming factors such as land use and management practices were more similar to each other at the single hill scale than at landscape and valley scales. The coefficients of variation of soil TN at the three spatial scales differed little, ranging from 0.20 to 0.32, whereas the values of soil AN decreased from 0.41 at the single hill scale to 0.30 at the landscape scale, and 0.33 at the valley scale. This result was consistent with the related study (Zhang et al., 2007) in hilly region of southern Jiangxi, China. Although the mean value of soil TN or AN of the three plots within each hill type at the single hill scale was close or equal to the mean value of the corresponding hill type at the landscape scale (Table 1), it was difficult to depict the spatial variability of soil nitrogen using a conventional statistical approach. 3.2. Spatial variability analysis 3.2.1. Semivariance analysis The best-fitted models of soil TN and AN, based on a combination of the ME, RMSSE and R2, are shown in Table 2. At the single hill scale, the TN and AN can select spherical, exponential or Gaussian models as their best-fitted models; however, only exponential models were able to accurately reflect their spatial variability at the two larger scales. At the single hill scale, the R2 values of the semivariogram models in all plots except for A1 and A4 for soil TN exceeded 0.9 (Table 2). In contrast, their values in the plots (except for A4, A5 and A8) for soil AN were below 0.9. Therefore, soil AN distribution is more variable than soil TN, as the latter was easily influenced by land use and fertilization (Oenema et al., 2010; Basso et al., 2016). At the landscape and valley scales, the R2 values of the semivariogram models were over 0.9, which indicates a high prediction capability for soil TN and AN (Table 2). The semivariograms of soil TN and AN exhibited some spatial autocorrelation at all three scales. Generally, the mean spatial autocorrelation ranges at the single hill, landscape and valley scales were 536, 5436 and 1614 m for soil TN, and 697, 4888 and 3138 m for soil AN (Table 2). Thus, these results showed increasing trends from the single hill to the landscape scale, and are similar to the change trend of soil TN across multiple scales in Pingguo County, Beijing (Hu et al., 2014). Nevertheless, they displayed a decreasing tendency from the landscape to valley scales. It is noteworthy that the autocorrelation ranges of the two variables varied

Table 1 Descriptive statistics for soil total nitrogen (TN) and available nitrogen (AN) across the multiple scales. Data seta

nb

TN (g kg−1)

AN (mg kg−1)

DT

Mean

Min

Max

CV

Skew

Kurt

P

DT

Mean

Min

Max

CV

Skew

Kurt

P

Mono-hill scale A1 27 A2 30 A3 28 A4 27 A5 25 A6 24 A7 26 A8 29 A9 25

N N N LN N N N N N

1.02bc 1.05b 0.90ab 0.92ab 0.90ab 0.97ab 0.88ab 0.88ab 0.79a

0.32 0.68 0.46 0.56 0.62 0.67 0.42 0.66 0.48

1.67 1.49 1.45 1.78 1.33 1.62 1.33 1.34 1.06

0.27 0.28 0.29 0.32 0.23 0.21 0.30 0.20 0.24

−0.30 0.32 0.55 0.56 0.56 1.43 0.10 0.85 −0.07

1.09 −1.67 −0.29 −0.24 −0.62 2.59 2.03 0.16 −1.38

0.82 0.06 0.99 0.12 0.93 0.28 0.38 0.36 0.72

N N N LN N N N N N

88ab 78ab 87ab 102b 94ab 91ab 72a 70a 78ab

19 12 40 53 47 63 21 23 22

175 144 143 199 134 124 132 149 154

0.44 0.53 0.28 0.36 0.23 0.14 0.42 0.40 0.57

−0.24 −0.06 −0.05 1.00 −0.55 −0.26 0.20 0.89 0.38

−0.35 −1.43 0.12 0.96 0.20 1.61 2.10 1.01 −1.37

0.72 0.15 0.97 0.68 0.86 0.32 0.84 0.83 0.22

Landscape scale B1 965 B2 788 B3 669

LN NN N

0.99c 0.90b 0.86a

0.48 0.48 0.49

1.60 1.58 1.29

0.19 0.22 0.16

0.51 0.64 0.22

0.71 −0.29 −0.65

0.00 0.00 0.68

N NN NN

93c 88b 71a

18 27 20

170 175 126

0.28 0.24 0.27

0.51 0.51 0.47

0.71 0.69 −0.07

0.37 0.00 0.00

Valley scale C 2422

NN

0.92

0.48

1.60

0.22

0.53

0.18

0.00

NN

85

18

175

0.29

0.50

0.28

0.00

a

Plots A1–A9 were the nine plots at single hill scale. Landscapes B1–B3 were the three areas at landscape scale. Region C was the whole valley. b n, sample number; DT, distribution type; N, normal distribution; LN, log-normal distribution; NN, non-normal distribution; CV, coefficient of variation; Skew, skewness; Kurt, kurtosis; P, P value of Kolmogorov–Smirnov test. c Means with different letters are significantly different from each other based on Student–Newman–Keuls or Games–Howell test when the error variances of the dependent variable across groups are equal or not (P b 0.05).

14

S. Zhang et al. / Science of the Total Environment 563–564 (2016) 10–18

Table 2 Parameters and cross-validation errors for the semivariogram models for soil total nitrogen (TN) and available nitrogen (AN) at the different scales. Dataset setab

Fittingb model

Range (m) Major

TN Mono-hill scale A1 A2 A3 A4 A5 A6 A7 A8 A9 Landscape scale B1 B2 B3 Valley scale C AN Mono-hill scale A1 A2 A3 A4 A5 A6 A7 A8 A9 Landscape scale B1 B2 B3 Valley scale C

Direction (°)

Nugget

Sill

Nug/Sill (%)

ME

RMSSE

R2

Minor

286 265 197 584 596 566 297 412 503

155 40 0 63 109 0 178 121 16

0.04 0.07 0.04 0.03 0.04 0.02 0.01 0.02 0.02

0.08 0.08 0.10 0.08 0.05 0.08 0.09 0.04 0.03

50.0 87.5 40.0 37.5 80.0 25.0 11.1 50.0 66.7

−0.02 −0.01 −0.01 0.00 −0.01 0.00 0.02 0.01 0.01

0.94 1.03 1.05 0.96 1.04 1.18 1.09 0.87 0.99

0.88 0.96 0.93 0.86 0.96 0.92 0.95 0.96 0.97

12,145 1366 2798

4773 694 2366

31 55 49

0.03 0.02 0.02

0.04 0.05 0.03

75.0 40.0 66.7

0.00 0.00 0.00

1.05 0.97 1.00

0.97 0.98 0.97

E

1614

1166

44

0.04

0.05

80.0

0.00

0.99

0.97

S G S S G S E G E

730 620 721 1380 799 357 579 794 294

603 497 575 1380 635 312 468 381 207

16 68 114 0 85 113 19 128 73

797 1211 493 793 349 371 234 1663 64

1122 1331 625 1620 484 545 1005 1811 644

71.0 91.0 78.9 49.0 72.1 68.1 23.3 91.8 9.9

0.89 0.09 −0.98 −0.54 −0.09 −0.23 0.97 0.87 −0.45

0.99 1.00 1.04 1.08 0.98 1.05 1.00 0.88 0.94

0.89 0.87 0.86 0.93 0.92 0.87 0.85 0.90 0.87

E E E

6491 7507 667

4862 4209 457

167 137 75

173 463 43

577 541 360

30.0 85.6 11.9

−0.38 −0.08 0.01

0.97 0.98 1.01

0.96 0.94 0.97

E

3138

3037

75

470

718

65.5

0.07

1.02

0.96

G S G E E S S E E

468 300 197 820 750 566 475 694 550

E E E

a

A1–A9 were the nine plots at single hill scale; B1–B3 were three areas at landscape scale. C was the whole valley at valley scale. b S, spherical model; E, exponential model; G, Gaussian model; Nug/Sill, the ratio of Nugget to Sill; ME, Mean Error; RMSSE, root mean square standardized error; R2, determination coefficient.

among the different hill types at the smallest scale. At the single hill scale, the range of soil TN was significantly smaller in the low hill than in the medium hill (P b 0.05); however, there was no statistical difference in autocorrelation ranges for soil AN among the three hill types at landscape scale (P N 0.05). The differences in the autocorrelation ranges may result from the similarities of soil TN or AN at the same positions on the adjacent hill, and the variation of other controlling factors. The nugget to sill ratio of the semivariogram model represents the degree of spatial dependence for a variable. As the ratio increases, the strength of spatial dependence decreases. In this study, the ratios of soil TN and AN varied across the different scales (Table 2). According to the classification system of spatial dependence for soil properties proposed by Cambardella et al. (1994), Plots A6 and A7 for soil TN, and Plots A7 and A9, and Landscape B3 for soil AN belonged to strong spatial dependence. Plots A2 and A5, and Region C for soil TN, and Plots A2, A3 and A8, and Landscape B2 for soil AN were considered to have a weak spatial dependence. The remaining plots exhibited moderate spatial dependence. From the change ranges of the nugget to sill ratios across the different scales, the spatial variability of the two soil N forms at the single hill scale showed strong uncertainty. A similar result was reported in the spatial variability of extractable soil ammonium-N at different spatial scales in a deciduous, mixed-species forest on the Oak Ridge Reservation, near Oak Ridge, TN, USA (Garten et al., 2007). 3.2.2. Spatial distribution The spatial distributions for soil TN and AN at the three scales were obtained by ordinary kriging based on the fitted semivariogram models (Fig. 2). Overall, the spatial distributions varied from striped bands at

the single hill scale and irregular patches at the valley scale. The values of soil TN and AN were classified into six classes based on the China National Soil Survey in the 1980s. Soil TN values of classes I–VI were b0.5, 0.5–0.8, 0.8–1.0, 1.0–1.2, 1.2–1.5 and N1.5 g kg−1, respectively, and classes I–VI of soil AN were b30, 30–60, 60–90, 90–120, 120–150 and N150 mg kg−1, respectively. At the single hill scale, the concentrations of soil TN and AN were both distributed as striped bands or patches, and decreased with the increasing elevation (Fig. 2a and b). As every plot was selected as a single hill topography, the slope with dryland and forest land occupied more than half of the whole plot area. For areas with minimal farm management practices (i.e. fertilization) on the middle–upper slope of the hill, the TN and AN concentrations were lower than those in the hilly edge and gully bottom. Nevertheless, the concentration classifications among the three hill types differed (Table 3). In low hilly plots, the areas of Class III and Class IV on average occupied 58.2 and 40.0% of the total plot area for TN, and 70.4 and 29.4% for AN, respectively. In medium hill plots, the class distributions of TN were similar to those in low hill plots; however, the areas of the third and fourth AN classes were on average 73.4 and 15.8% for TN, and 25.6 and 71.4% of total plot areas for AN, respectively (Table 3). In the high hill plots, the areas of the second and third TN classes on average accounted for 50.6 and 45.1% of the plot areas; by contrast, the land with the second, third and fourth AN classes on average occupied 25.5, 44.8 and 25.9% of the total plot areas, respectively. As a whole, the mean values of the second, third, fourth and fifth classes from the three hill types occupied 19.5, 58.9, 19.8 and 1.8%, respectively, of the single hill area for soil TN, and they were 8.6, 46.9, 42.2 and 2.3% of the single hill area for soil AN, respectively.

S. Zhang et al. / Science of the Total Environment 563–564 (2016) 10–18

15

Fig. 2. Spatial distributions for soil total nitrogen (TN, g kg−1) and available nitrogen (AN, mg kg−1) in Plot A4 at single hill scale (a, b), Area B2 at landscape scale (c, d) and Region C at valley scale (e, f).

Compared with the area percentage of the grades at the single hill scale, the percentages of higher-grade areas of soil TN increased in the three hill regions at the landscape scale (Table 3). For example, the percentage of the fourth TN class in the low hill area (B1, 49.5%) was higher than the mean value of Plots A1, A2 and A3 (40.0%). Nevertheless, the distribution changes of soil AN class from the single hill to the landscape scale showed three situations: increase in the low hill area, decrease in the medium hill area and concentration on the second grade in the high hill area (Table 3). As shown in Fig. 2e and f, at the valley scale, soil TN and AN concentrations showed an irregular patchy or striped distributions and

decreased from the bank areas of the main rivers toward the north or the west. Their lowest values were found mostly in the north (high hilly areas) and west (medium hilly area), while the highest values were found in the southeast or south areas along the branched rivers (low hilly area). In the valley, the areas of the second, third and fourth classes occupied 16.4, 63.5 and 19.8% of the whole valley area for soil TN, and 2.4, 64.7 and 32.1% for soil AN (Table 3), respectively. An interesting result presented in Table 3 indicates that the mean value distributions of area percentage in different classes of soil TN and AN at the single hill and landscape scales were similar to the area percentage distribution of their relative grades at the valley scale;

16

S. Zhang et al. / Science of the Total Environment 563–564 (2016) 10–18

Table 3 The percentage of grading areas for soil total nitrogen (TN) and available nitrogen (AN) across the multiple scales (%). Data seta

TN (g kg−1)

Class

I

II

III

IV

V

VI

I

II

III

IV

V

VI

≤0.5

0.5–0.8

0.8–1.0

1.0–1.2

1.2–1.5

N1.5

≤30

30–60

60–90

90–120

120–150

N150

0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

0.0 0.0 1.3 11.3 0.0 11.4 60.3 31.7 59.9

48.0 32.3 94.3 75.7 98.3 46.1 28.8 66.4 40.1

51.3 64.4 4.4 13.0 1.7 32.8 9.1 1.9 0.0

0.7 3.3 0.0 0.0 0.0 9.6 1.8 0.0 0.0

0.0 0.0 0.0 0.0 0.0 0.1 0.0 0.0 0.0

0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

0.0 0.7 0.0 0.0 0.0 0.0 0.0 41.8 34.7

58.6 88.5 64.1 12.7 21.2 42.8 35.6 58.2 40.5

41.4 10.8 35.9 78.2 78.8 57.1 53.1 0.0 24.7

0.0 0.0 0.0 9.1 0.0 0.1 11.3 0.0 0.1

0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

Landscape scale B1 0.0 B2 0.0 B3 0.0

0.4 17.3 19.4

45.4 61.9 79.4

49.6 18.6 1.2

4.6 2.2 0.0

0.0 0.0 0.0

0.0 0.0 0.0

0.3 0.2 7.9

47.2 58.8 88.8

45.3 40.1 3.3

7.2 0.9 0.0

0.0 0.0 0.0

Valley scale C

16.4

63.5

19.7

0.4

0.0

0.0

2.4

64.7

32.1

0.8

0.0

Mono-hill scale A1 A2 A3 A4 A5 A6 A7 A8 A9

a

0.0

AN (mg kg−1)

Plots A1–A9 were the nine plots at single hill scale. Landscapes B1–B3 were the three areas at landscape scale. Region C was the whole valley.

however, the similarity of soil AN was lower than that of soil TN because of the high intensive land management (Wang et al., 2009). This finding that the spatial patterns of soil TN and AN were coincident across multiple scales is different from the diverse spatial distribution characteristics of soil TN identified at small, medium and large scales in Pinggu, Beijing (Hu et al., 2014). 3.3. Factors affecting spatial patterns of soil TN and AN at the multiple scales To achieve a better understanding of the characterization of soil spatial variability, it is essential to study the complex relations between soil nitrogen and environmental factors or land use systems (Yemefack et al., 2005). In the current study, the main effects of hill types, slope positions, soil types, land uses and fertilization on soil TN and AN were evaluated using univariate GLM. 3.3.1. Hill type Landforms are regarded as one of the most important factors controlling pedogenic processes, and result in large and complex spatial variation of soil properties at multiple scales (Yamashita et al., 2010). It is expected that hill types with different relative heights may influence spatial patterns of soil nitrogen by affecting the spatial variability of microclimate, runoff, biotic activity and management practices (Zhang et al., 2007). However, the effect of hill types on soil nitrogen has been rarely reported so far. In the present study, the two soil nitrogen forms did not show the same variation among the three hill types (Tables 1 and 2). At the single hill scale, the soils in the three hill types showed significant differences for TN (P = 0.002) but not for AN (P = 0.134). The TN concentrations were significantly higher in the low hill than in the medium and high hills (P b 0.05). At the landscape scale, soil TN and AN showed significant differences among the three hill types, descending as follows: low hill N medium hill N high hill (Table 1, P b 0.05). Furthermore, the differences in soil AN levels among the three hill types were not significant at the single hill scale (P N 0.05), but were significant at the landscape scale (P b 0.05). These results may be caused by the variation of hilly landforms with different parent materials, relative heights, land uses and management practices (Wang et al., 2009; Wall et al., 2012). According to the soil AN level, the risk possibilities of N non-point pollution varied among the three hill types. At the single hill scale, Plots A7 and A4 had a high risk for N non-point pollution and Plots A8 and A9 had a high risk for N deficiency (Table 3). Similarly, at the landscape scale, the risk of N non-point pollution decreased in the following

order: low hill N medium hill N high hill. In contrast, the risk of N deficiency increased in the following order: low hill b medium hill b high hill. 3.3.2. Slope position Slope position also affects soil properties through surface runoff and redistribution of the material from erosion or accumulation (Schwanghart and Jarmer, 2011). In the current study, only soil TN and AN in low hills and AN in medium hills at the single hill scale indicated significant differences among the different slope positions (Table 4, P b 0.05). At the landscape scale, soil TN and AN in the low hill, and TN in the high hill were significantly different in the different slope positions (P b 0.05). However, the two forms of soil nitrogen in medium and high hills generally increased in the following order: top b backslope b footslope b toeslope b gully bottom (Table 4). However, at the valley scale they increased in the following order: top and backslope b footslope b toeslope b gully bottom (P b 0.05). Therefore, the variation of two soil nitrogen forms became gradually more significant with the scale extending. This result is seemingly contrary to the findings of Wanshnong et al. (2013), who showed that potentially mineralizable nitrogen on a hill slope in India was higher at the summit than on the backslope, because the soil was deeper at the summit than on the backslope (P b 0.05). 3.3.3. Parent material Parent materials determine soil composition and formation by weathering, and affect the supply of soil nutrient elements including nitrogen. In the current study region, soil TN and AN concentrations from the different parent materials in sequence were thick shale and thin sandstone N thick sandstone and thin shale or alluvium at the single hill and landscape scales (Table 4). However, they decreased in the order: thick shale and thin sandstone N alluvium N thick sandstone and thin shale at the valley scale (P b 0.05). Different soil nitrogen concentrations from diverse parent materials have been reported in similar research (Kooijman et al., 2005; Mayes et al., 2014). 3.3.4. Soil type Nitrogen is present in the soil ecological systems in many forms and responses to diverse soil compositions and processes with respect to different soil types. Thus, the different soil types might contain distinct TN and AN concentrations. In the current study, the two forms of soil nitrogen among soil types indicated different concentration levels: they were higher in Typic Fe-accumuli-Stagnic Anthrosols than in Calcaric Purpli-

S. Zhang et al. / Science of the Total Environment 563–564 (2016) 10–18

17

Table 4 Effects of slope positions, parent materials, soil types, land uses and nitrogen fertilizer rates on soil total nitrogen (TN, g kg−1) and available nitrogen (AN, mg kg−1) across the multiple scales. Factora

Mono-hill scale Low hill

Slope position

Parent material

Soil type

Land use

N fertilizer rate

TO BS FS TS GB AL TSA TSH CPOP CPUC CUOP TFASA TFLSA PF AR GL FL N1 N2 N3 N4

Landscape scale Medium hill

High hill

Low hill

Valley scale Medium hill

High hill

n

TN

AN

n

TN

AN

n

TN

AN

n

TN

AN

n

TN

AN

n

TN

AN

n

TN

AN

15 34 20 16 – 10 15 60 14 39 – 16 16 30 41 7 7 8 25 31 21

0.80ab 1.01b 1.07b 1.03b – 0.68a 0.79a 1.09b 0.77a 1.00b – 1.18c 0.97b 1.10b 0.97b 0.63a 1.01b 0.74a 0.78a 1.02b 1.29c

65a 88ab 100b 81ab – 60a 74ab 92b 55a 88b – 114c 75ab 96 81 61 87 63a 68a 92b 105b

12 25 28 11 – – 17 59 11 48 – 17 – 17 52 – 7 7 11 35 23

0.74 0.89 0.86 0.98 – – 0.76a 0.90b 0.65a 0.88b – 0.98b – 0.98 0.83 – 0.84 0.84b 0.65a 0.88b 0.97b

72a 89ab 77a 107b – – 79 86 53a 85b – 103b – 103b 80ab – 73a 73a 55a 80a 109b

4 24 35 17 – – 25 55 6 56 – 18 – 17 54 – 9 8 14 21 37

0.9 0.81 0.9 0.81 – – 0.91b 0.83a 0.9 0.88 – 0.76 – 0.77 0.9 – 0.73 0.72ab 0.63a 0.82b 0.98c

47 66 79 76 – – 70 74 72 75 – 68 – 64ab 79b – 51a 53a 54a 62a 91b

203 284 233 95 150 168 618 179 125 448 133 92 167 244 625 38 58 96 12 280 577

0.93a 0.92a 0.96a 1.07b 1.17c 0.96a 0.94a 1.16b 0.86a 0.94b 0.97b 1.08c 1.17d 1.13b 0.94a 0.87a 0.92a 0.90a 0.92ab 1.07c 0.96b

90a 85a 90a 101b 110b 97 88 107 78a 87b 98cd 106de 109e 107 89 83 88 86a 92ab 102b 90ab

60 408 89 134 97 22 708 58 51 477 19 185 56 230 533 9 16 29 23 262 474

0.86 0.89 0.92 1.08 1.11 0.92 0.88 1.19 0.70a 0.83b 0.88b 1.07c 1.21d 1.1 0.83 0.74 0.88 0.84a 0.94ab 1.03b 0.84a

78 82 83 97 111 85a 85a 121b 74a 82a 83a 100b 116c 103 81 82 81 86a 91ab 98b 83a

103 300 107 98 61 15 589 65 123 384 12 129 21 158 450 26 35 59 31 218 361

0.77a 0.82b 0.87c 0.95d 0.95d 0.79a 0.85a 0.92b 0.78ab 0.84b 0.71a 0.96c 1.06d 0.97b 0.83a 0.76a 0.79a 0.79a 0.85ab 0.88b 0.86ab

65 67 72 76 80 71 71 69 66a 68a 64a 80b 79b 79 68 60 66 64 66 70 74

366 992 429 327 308 205 1915 302 299 1309 164 406 244 632 1608 73 109 184 66 760 1412

0.85a 0.85a 0.92b 1.04c 1.12d 0.94b 0.89a 1.11c 0.80a 0.87b 0.94c 1.04d 1.17e 1.08c 0.87b 0.82a 0.87b 0.86a 0.90b 1.00c 0.89b

81a 79a 84b 93c 104d 94b 89a 102c 72a 79b 94c 95c 108d 98c 81b 74a 80b 79a 79a 93b 82a

a Slope positions: TO, top (summit and shoulder); BS, backslope; FS, footslope; TS, toeslope; GB, gully bottom. Parent materials: AL, alluvium; TSA, thick sandstone and thin shale; TSH, thick shale and thin sandstone. Soil type: CPOP, Calcaric Purpli-Orthic Primosols; CPUC, Calcaric Purpli-Udic Cambosols; CUOP, Carbonatic Udi-Orthic Primosols; TFASA, Typic Fe-accumuliStagnic Anthrosols; TFLSA, Typic Fe-leachi-Stagnic Anthrosols. Land uses: PF, paddy field, AR, dryland; GL, grass land; FL, forest land. N fertilizer rates: N1, b150 (kg ha−1 yr−1); N2, 151– 270 (kg ha−1 yr−1); N3, 271–330 (kg ha−1 yr−1); N4, N331 (kg ha−1 yr−1). b Means with different letters are significantly different from each other based on Student–Newman–Keuls or Games–Howell test when the error variances of the dependent variable across groups are equal or not (P b 0.05).

Orthic Primosols in low and medium hills at the single hill scale. At the landscape and valley scales, concentrations of the two N forms generally decreased as follows: Typic Fe-leachi-Stagnic Anthrosols N Typic Feaccumuli-Stagnic Anthrosols N Carbonatic Udi-Orthic Primosols N Calcaric Purpli-Udic Cambosols N Calcaric Purpli-Orthic Primosols. 3.3.5. Land use Land use is one of the most important factors affecting soil nitrogen concentrations (Oenema et al., 2010; Wiesmeier et al., 2013). Some research has indicated that soil TN and AN showed significantly different concentrations among different land uses (Cobo et al., 2010; Mayes et al., 2014). In the current study, however, soil nitrogen concentrations showed no clear difference in characteristics among the different land uses at the single hill scale, and only TN was higher in paddy fields than in the other land uses in low and high hills at the landscape scale. At the valley scale, soil concentrations of the two N forms in dryland and forests were lower than those in paddy fields (Table 4, P b 0.05), and higher than in grasslands (P b 0.05). This is because higher fertilizer rates were generally applied to paddy fields than to drylands. In addition, these land uses were distributed in the footslope and gullies where the erosion and leaching N from the slope is deposited into the soil. In contrast, the grasslands and forest lands applied no fertilizer or only small amounts. These results were similar to those of Mayes et al. (2014), who showed that TN concentrations in soils from alluvium among the different land uses were higher in farmland than in grazing land (P b 0.05) in the Konya Basin, Turkey. In addition, incorporation of crop residues into soils usually affects the soil N content (Huang et al., 2007). In the current study valley, incorporations of straws and roots in paddy field and drylands were 4800–7000 and 3675– 4725 kg ha−1 yr−1. As a result, soil TN and AN were higher in paddy fields than in drylands (P b 0.05) at the valley scale. 3.3.6. Fertilization Fertilizer application provides an important source of soil N. As shown in Table 4, higher concentrations of soil TN and AN occurred in

the soils with higher N fertilizer rates at the single hill scale; however, the soils with the N rates of 271–330 kg ha−1 yr−1 showed a slightly higher TN and AN at the landscape and valley scales. This is because the N fertilizer rate was one of the key factors affecting soil N, and the other land management practices such as crop harvest and residue, manure application, irrigation and drainage also affected their concentrations (Huang et al., 2007; Wang et al., 2009). In general, the key factors affecting the two N forms at the single hill scale were hill type and N fertilizer rate for soil TN (P b 0.01), and N fertilizer rate, hill type and soil type for soil AN, based on analysis of variance and comparison of means. At the landscape scale, hill type and soil type were the most crucial factors affecting the two N forms (P b 0.01). In addition, the other key factors showed some difference among the three hill types. In the low hill area (B1), slope position and N fertilizer rate were the key factors (P b 0.01); in the medium hill area (B2), N fertilizer rate was the crucial factor (P b 0.01); and in the high hill area (B3), parent material was the key factor (P b 0.05). It is noteworthy that hill type, slope position, parent material, soil type, land uses and N fertilizer rate were the key factors affecting the two N forms at the valley scale (P b 0.01). 4. Conclusions The spatial variability of soil nitrogen at the single hill, landscape and valley scales in the hilly region of Sichuan Basin was analyzed using a combined method of classical statistics and geostatistics. The soil TN and AN in low, medium and high hills showed different spatial distribution patterns at the single hill and landscape scales. The percentages of the higher grades of the two N forms clearly decreased from low hill to medium hill to high hill. The mean area percentage patterns in different grades at the single hill and landscape scales were similar to those at the valley scale; however, the similarities of soil TN were higher than those of soil AN. Their key affecting factors under multiple scales were hill type and fertilization for the single hill scale, hill type and soil type at the landscape scale, and hill type, slope position, parent material,

18

S. Zhang et al. / Science of the Total Environment 563–564 (2016) 10–18

soil type, land uses and fertilization at the valley scale. Therefore, the effects of these key factors became more significant with the scale increase. Nevertheless, future studies are needed to explore the interaction effects of the most crucial factors for soil nitrogen across multiple scales in the hilly valley. Acknowledgments This study was supported by the Projects of Sci-tech Support, Sichuan, China (No. 2014NZ0044) and the Projects of National Sci-tech Support, China (No. 2012BAD14B18-2). We thank Chuer Zhang and Guiyin Wang from Sichuan Agricultural University for supporting the investigation and research work. References Aubert, M., Bureau, F., Vinceslas-Akpa, M., 2005. Sources of spatial and temporal variability of inorganic nitrogen in pure and mixed deciduous temperate forests. Soil Biol. Biochem. 37, 67–79. Aumtong, S., Magid, J., Bruun, S., de Neergaard, A., 2009. Relating soil carbon fractions to land use in sloping uplands in northern Thailand. Agric. Ecosyst. Environ. 131, 229–239. Basso, B., Dumont, B., Cammarano, D., Pezzuolo, A., Marinello, F., Sartori, L., 2016. Environmental and economic benefits of variable rate nitrogen fertilization in a nitrate vulnerable zone. Sci. Total Environ. 545–546, 227–235. Boussadia, O., Steppe, K., Zgallai, H., Hadj, S.B.E., Braham, M., Lemeur, R., Labeke, M.C.V., 2010. Effects of nitrogen deficiency on leaf photosynthesis, carbohydrate status and biomass production in two olive cultivars ‘Meski’ and ‘Koroneiki’. Sci. Hortic. 123, 336–342. Bremmer, J.M., Mulvaney, C.S., 1982. Nitrogen: total. In: Page, A.L., Miller, R.H., Keeney, D.R. (Eds.), Methods of Soil Analysis, Part 2—Chemical and Microbiological Properties. ASA-SSSA, Madison, WI, pp. 595–641. Cambardella, C.A., Moorman, T.B., Novak, J.M., Parkin, T.B., Karlen, D.L., Turco, R.F., Konopka, A.E., 1994. Fieldscale variability of soil properties in Central Iowa soils. Soil Sci. Soc. Am. J. 58, 1501–1511. Cobo, J.G., Dercon, G., Cadisch, G., 2010. Nutrient balances in African land use systems across different spatial scales: a review of approaches, challenges and progress. Agric. Ecosyst. Environ. 136, 1–15. Córdova, C., Sohi, S.P., Lark, R.M., Goulding, K.W.T., Robinson, J.S., 2012. Resolving the spatial variability of soil N using fractions of soil organic matter. Agric. Ecosyst. Environ. 147, 66–72. Cornfield, A.H., 1960. Ammonia released on treating soils with N sodium hydroxide as a possible means of predicting the nitrogen — supplying power of soils. Nature 187, 260–261. Dharmakeerthi, R.S., Kay, B.D., Beauchamp, E.G., 2005. Factors contributing to changes in plant available nitrogen across a variable nitrogen across a variable landscape. Soil Sci. Soc. Am. J. 69, 453–462. Foster, N., Spoelstra, J., Hazlett, P., Schiff, S., Beall, F., Creed, I., David, C., 2005. Heterogeneity in soil nitrogen within first-order forested catchments at the Turkey Lakes Watershed. Can. J. For. Res. 35, 797–805. Franzluebbers, A.J., Stuedemann, J.A., 2009. Soil-profile organic carbon and total nitrogen during 12 years of pasture management in the Southern Piedmont USA. Agric. Ecosyst. Environ. 129, 28–36. Fu, W., Tunney, H., Zhang, C., 2010. Spatial variation of soil nutrients in a dairy farm and its implications for site-specific fertilizer application. Soil Tillage Res. 106, 185–193. Garten, C.T., Kang, S., Brice, D.J., Schadt, C.W., Zhou, J.Z., 2007. Variability in soil properties at different spatial scales (1 m–1 km) in a deciduous forest ecosystem. Soil Biol. Biochem. 39, 2621–2627. Gong, Z.T., Zhang, G.L., Chen, Z.C., 2003. Development of soil classification in China. In: Eswaran, H., Rice, T., Ahrens, R., Stewart, B.A. (Eds.), Soil Classification: A Global Desk Reference. CRC Press, Boca Raton, pp. 101–125. He, X.B., Xu, Y.B., Zhang, X.B., 2007. Traditional farming system for soil conservation on slope farmland in southwestern China. Soil Tillage Res. 94, 193–200. Hu, K.L., Wang, S.Y., Li, H., Huang, F., Li, B.G., 2014. Spatial scaling effects on variability of soil organic matter and total nitrogen in suburban Beijing. Geoderma 226–227, 54–63. Huang, B., Sun, W.X., Zhao, Y.C., Zhu, J., Yang, R.Q., Zou, Z., Ding, F., Su, J.P., 2007. Temporal and spatial variability of soil organic matter and total nitrogen in an agricultural ecosystem as affected by farming practices. Geoderma 139, 336–345. Jackson-Blake, L., Helliwell, R.C., Britton, A.J., Gibbs, S., Coull, M.C., Dawson, L., 2012. Controls on soil solution nitrogen along an altitudinal gradient in the Scottish uplands. Sci. Total Environ. 431, 100–108. Johnston, K., Hoef, J.M.V., Krivoruchko, K., Lucas, N., 2001. Using ArcGIS Geostatistical Analyst. ESRI, Redlands, CA, pp. 59–77.

Kooijman, A.M., Jongejans, J., Sevink, J., 2005. Parent material effects on Mediterranean woodland ecosystems in NE Spain. Catena 59, 55–68. Lamsal, S., Grunwald, S., Bruland, G.L., Bliss, C.M., Comerford, N.B., 2006. Regional hybrid geospatial modeling of soil nitrate–nitrogen in the Santa Fe River Watershed. Geoderma 135, 233–247. Li, Q.Q., Luo, Y.L., Wang, C.Q., Li, B., Zhang, X., Yuan, D.G., Gao, X.S., Zhang, H., 2016. Spatiotemporal variations and factors affecting soil nitrogen in the purple hilly area of Southwest China during the 1980s and the 2010s. Sci. Total Environ. 547, 173–181. Mairura, F.S., Mugendi, D.N., Mwanje, J.I., Ramisch, J.J., Mbugua, P.K., Chianu, J.N., 2007. Integrating scientific and farmers' evaluation of soil quality indicators in Central Kenya. Geoderma 139, 134–143. Matheron, G., 1963. Principles of geostatistics. Econ. Geol. 58, 1246. Mayes, M., Marin-Spiotta, E., Szymanski, L., Erdoğan, M.A., Ozdoğan, M., Clayton, M., 2014. Soil type mediates effects of land use on soil carbon and nitrogen in the Konya Basin, Turkey. Geoderma 232–234, 517–527. Mengel, K., Hütsch, B., Kane, Y., 2006. Nitrogen fertilizer application rates on cereal crops according to available mineral and organic soil nitrogen. Eur. J. Agron. 24, 343–348. Momtaz, H.R., Jafarzadeh, A.A., Torabi, H., Oustan, S., Samadi, A., Davatgar, N., Gilkes, R.J., 2009. An assessment of the variation in soil properties within and between landform in the Amol region, Iran. Geoderma 149, 10–18. Oenema, J., Burgers, S., Verloop, K., Hooijboer, A., Boumans, L., Berge, H.T., 2010. Multiscale effects of management, environmental conditions, and land use on nitrate leaching in dairy farms. J. Environ. Qual. 39, 2016–2028. Oorts, K., Merckx, R., Gréhan, E., Labreuche, J., Nicolardot, B., 2007. Determinants of annual fluxes of CO2 and N2O in long-term no-tillage and conventional tillage systems in northern France. Soil Tillage Res. 95, 133–148. Park, S.J., Hwang, C.S., Vlek, P.L.G., 2005. Comparison of adaptive techniques to predict crop yield response under varying soil and land management conditions. Agric. Syst. 85, 59–81. Raines, G.L., 2002. Description and comparison of geologic maps with FRAGSTATS—a spatial statistics program. Comput. Geosci. 28, 169–177. Rode, M., Thiel, E., Franko, U., Wenk, G., Hesser, F., 2009. Impact of selected agricultural management options on the reduction of nitrogen loads in three representative mesoscale catchments in Central Germany. Sci. Total Environ. 407, 3459–3472. Rodríguez, A., Durán, J., Covelo, F., Fernández-Palacios, J.M., Gallardo, A., 2011. Spatial pattern and variability in soil N and P availability under the influence of two dominant species in a pine forest. Plant Soil 345, 211–221. Schmitter, P., Dercon, G., Hilger, T., Ha, T.T.L., Thanh, N.H., Lam, N., Vien, T.D., Cadisch, G., 2010. Toposequential variation in methane emissions from double-cropping paddy rice in Northwest Vietnam. Geoderma 155, 298–307. Schwanghart, W., Jarmer, T., 2011. Linking spatial patterns of soil organic carbon to topography — a case study from south-eastern Spain. Geomorphology 126, 252–263. Tremblay, N., Bouroubi, M.Y., Vigneault, P., Bélec, C., 2011. Guidelines for in-season nitrogen application for maize (Zea mays L.) based on soil and terrain properties. Field Crop Res. 122, 273–283. Velthof, G.L., Lesschen, J.P., Webb, J., Pietrzak, S., Miatkowski, Z., Pinto, M., Kros, J., Oenema, O., 2014. The impact of the Nitrates Directive on nitrogen emissions from agriculture in the EU-27 during 2000–2008. Sci. Total Environ. 468-469, 1225–1233. Wall, D.P., Murphy, P.N.C., Melland, A.R., Mechan, S., Shine, O., Buckley, C., Mellander, P.-E., Shortle, G., Jordan, P., 2012. Evaluating nutrient source regulations at different scales in five agricultural catchments. Environ. Sci. Pol. 24, 34–43. Wang, H.J., Shi, X.Z., Yu, D.S., Weindorf, D.C., Huang, B., Sun, W.X., Ritsema, C.J., Milne, E., 2009. Factors determining soil nutrient distribution in a small-scaled watershed in the purple soil region of Sichuan Province, China. Soil Tillage Res. 105, 300–306. Wanshnong, R.K., Thakuria, D., Sangma, C.B., Ram, V., Bora, P.K., 2013. Influence of hill slope on biological pools of carbon, nitrogen, and phosphorus in acidic alfisols of citrus orchard. Catena 111, 1–8. Wiesmeier, M., Hübner, R., Barthold, F., Spörlein, P., Geuß, U., Hangen, E., Reischl, A., Schilling, B., Lützow, M.V., Kögel-Knabner, I., 2013. Amount, distribution and driving factors of soil organic carbon and nitrogen in cropland and grassland soils of southeast Germany (Bavaria). Agric. Ecosyst. Environ. 176, 39–52. Yamashita, N., Ohta, S., Sase, H., Luangjame, J., Visaratana, T., Kievuttinon, B., Garivait, H., Kanzaki, M., 2010. Seasonal and spatial variation of nitrogen dynamics in the litter and surface soil layers on a tropical dry evergreen forest slope. For. Ecol. Manag. 259, 1502–1512. Yemefack, M., Rossiter, D.G., Njomgang, R., 2005. Multi-scale characterization of soil variability within an agricultural landscape mosaic system in southern Cameroon. Geoderma 125, 117–143. Zeleke, T.B., Si, B.C., 2006. Characterizing scale-dependent spatial relationships between soil properties using multifractal techniques. Geoderma 134, 440–452. Zhang, S.R., Sun, B., Zhao, Q.G., Li, T., Chen, M.M., Huang, L.Q., 2007. Distribution characteristics of soil nitrogen at multi-scales in hilly region of South China. Acta Pedol. Sin. 44, 885–892. Zhou, X.L., 2007. Fissure water and its rational use in shallow weathering zone of red beds in the Sichuan Basin. Acta Geol. Sichuan 27, 184–191.