Introduction

The spatial distribution of soil nutrients under agricultural systems is affected by natural conditions as well as management practices (Lal 1998; Huang 2000; Barton et al. 2004; Atreya et al. 2008), while soil nutrient distribution in natural systems is mainly affected by topography, climate and biological activities (DeBusk et al. 1994; Tang and Yang 2006). In recent years, geo-statistics, linear models, neural networks, regression trees, fuzzy systems and other analytical procedures have been used to analyze soil nutrient distributions and are considered good tools for use in understanding nutrient dynamics in the field (Zhang et al. 2007; Srividya et al. 2002; Liu et al. 2006; DeBusk et al. 1994; Park and Vlek 2002). Despite the sophisticated analytical techniques available and the recognition of the importance of understanding nutrient variability, however, the nature and degree of soil nutrient variability with respect to landscape position is still poorly understood (Silveira et al. 2009; Gallardo 2003). One of the primary factors affecting nutrient distribution is physical movement of the soil, as runoff from upslope areas carries topsoil to lower slope positions, thus altering the spatial distribution of soil and water and affecting soil nutrient content in both affected areas (Balasundram et al. 2006; Noorbakhsh et al. 2008; Verity and Anderson 1990; Moulin et al. 1994). As a result, soil organic matter, nitrogen and phosphorus contents are generally higher in the foot-slope area than at higher slope positions (Soon and Malhi 2005; Verity and Anderson 1990; Moulin et al. 1994; Jowkin and Schoenau 1998; Malo and Worcester 1975; Voroney et al. 1981). It is apparent that the steepness of a slope influences the intensity of soil erosion and thus affects soil nutrient distribution in the field, but soil erosion as affected by slope also varies according to the type of soil (Agassi et al. 1990; Walson and Laflen 1986; Liu et al. 2006; Morgan 2005).

Slope aspect can also be important in affecting nutrient distribution, as equatorial slopes tend to be dryer than polar-facing ones because have greater evapo-transpiration rates (Rundel 1981), while the amount of rainfall, and thus runoff, tends to be greater on a windward slope than on the leeward side (Agassi et al. 1990). This affects nutrient levels because higher soil moisture tends to retard the decomposition of soil organic matter and thus sustains higher nutrient contents (Huang 2000; Wang and Cai 1988). Much research reported in the literature has been focused on north vs south-facing slopes because of the difference in evapotranspiration, but wind and storms can affect all aspects and should not be neglected when considering nutrient distribution (Rundel 1981; Agassi et al. 1990; Wang and Cai 1988; Gong et al. 2008). Previous research in nutrient distribution has been primarily conducted in Europe, Australia and North American and has focused on the effects of topography, soil and climate, with little documentation of the combined effect of land management and landscape processes (McBratney et al. 2003; Gessler et al. 1995; Grunwald 2006; Burrough 1993).

Studies of the effect of land management on nutrients has shown that cultivation generally increases the potential for soil erosion due to the breakdown of soil aggregates and reduction of soil cohesion, and thus decreases soil nutrient content in the profile (Horn et al. 1995; Walson and Laflen 1986). By extension, since cross-slope tillage reduces soil loss compared to down-slope tillage (Voroney et al. 1981; Van Doren et al. 1950), it can be expected that nutrient levels would also be less affected. It is also reported that soil nutrient content varies considerably under different crops (Huang, 2000; Gardner and Gerrard 2003), and that generally continuous corn (Zea mays L.) and corn-soybean (Glycine max L.) rotations do not result in a significant accumulation of SOC (West and Post 2002). Fertilizer application rates can also affect soil nutrient content dynamics in the field (Lal 2004; Liu et al. 2006).

Most research studies mentioned above were carried out on plots or in a small study area, but regional analysis of nutrient variability is becoming of more interest in the agricultural community. Most of the “black soil” region of North-eastern China has been a major production base for corn and soybeans for more than 100 years. The long, narrow fields generally have slopes of <6° and slope lengths from 500 to 2,000 m (Yan et al. 2008) and can suffer from extreme erosion under intense rainfall events. Little research has been conducted regarding the effects of slope aspect, slope steepness and land management on soil nutrient distribution, and slope aspect particularly has been neglected (Liu et al. 2006; Zhang et al. 2007; Wei et al. 2006; Yan et al. 2008). In this study, the spatial distribution of soil nutrients in 2007 was investigated in a 6.55 km2 region of typical Mollisol soils using Kriging procedures (Yost et al. 1982; Oliver 1987; Debusk et al. 1994) and traditional statistical methods to describe the spatial variability of soil characteristics. Soil nutrient variability was analysed in relation to slope position, slope steepness, slope aspect and land management practices.

Materials and methods

Study area description

Our 6.55 sq. km (1.57 km by 4.17 km) study area is located in Guangrong village (47.21–47.23°N, 126.50–126.51°E) in Hailun city, Heilongjiang province, Northeast China (Fig. 1). The area falls in the North Temperate Zone and has a continental monsoon climate of cold and arid weather in winter and hot and rainy conditions in summer. Average annual precipitation is 530 mm, with 65% falling in June, July and August. Average precipitation from March to October in the 2002–2008 period was 472.3 mm. The average annual temperature is 1.5°C and annual sunshine averages between 2,600 and 2,800 h. Total annual solar radiation is 113 MJ cm−2 and annual average available accumulated temperature (≥10°C) is 2,450°C. The prevailing wind is from the north-west in winter and spring and from the south-west in summer (Soil survey service of Hailun 1985). Formation of soil in the study area began during the Quaternary period on loess deposits under natural grasses and now have a rich, dark organic horizon and are classified as Mollisols (Zhang et al. 2007). These soils have a silty clay loam texture (Table 1), and most slopes are inclined at less than 5°, but are over 200 m in length. Slope aspects can be defined as north-east, south-east, north-west or south-west.

Fig. 1
figure 1

Location of Guangrong region, Hailun city, Heilongjiang province, Northeast China

Full size image
Table 1 Soil physical and chemical properties before planting on the experimental field in 2007
Full size table

Land management

Our study area is farmed by local villagers, but all cropping activities are prescribed and monitored by our team of researchers working with a local manager. Fields are generally long and narrow and the orientation between up-and-down or cross-slope is essentially random. The crop rotation has consisted of alternating one year of soybean with one year of corn for at least the last 3 decades, so approximately 50% of fields were in each crop during the study year of 2007. Fields were ridged-tilled on 65 cm intervals using a small tractor operating a roto-tiller after harvest in the autumn of 2006, and both crops were planted in the first 10 days of May and harvested in October of 2007. Average crop yield for soybean, calculated by our research group, was 3,000 kg ha−1 and for corn 4,500 kg ha−1. Prescribed chemical fertilization on soybean consisted of 20.25 kg N ha−1, 51.75 kg P ha−1 and 15 kg K ha−1 applied at planting, and on corn 69 kg N ha−1 at planting and an additional 69 kg N ha−1 as side dressing at the three-leaf stage. Weeds were controlled with herbicides after emergence early in the growing season, and two or three between-ridge machine tillage operations were carried out before July.

Soil sample collection and measurement

Four hundred and thirty-five soil samples were collected on a grid pattern from a depth of 0–20 cm in the autumn of 2007 after harvest (Fig. 2). Each soil sample was comprised of a mixture of six cores taken randomly from within a 20 m2 plot. Samples were air-dried and sieved at 0.25 mm for analyzing total nitrogen (TN) and soil organic matter (SOM) and at 2 mm for total phosphorus (TP). SOM and TN were measured using a Vario ELIII and TP was determined using the molybdenum-blue method after digestion with concentrated HClO4–H2SO4 (Lu 1999). Farmers in the study area plant their fields on ridges running the length of the field, and since field direction varies randomly with respect to slope, the effect of ridge direction on soil nutrient content was analysed by selecting 20 sample points to represent each of cross-slope and down-slope tillage. These selections were chosen from sites where slope steepness and aspect were identical over an area of 0.5 km2 around the sample point. The site selection process was performed using ArcGIS 9.3 software (ESRI 2008).

Fig. 2
figure 2

Distribution of sample sites and elevation in the study area

Full size image

Calculation of slope and aspect

A Digital Elevation Model (DEM) was built according to a triangulation network calculated from contour lines with 1 m intervals using ArcGIS 9.3 (ESRI 2008) and the DEM was then transformed into slope percent and aspect. Aspect was designated as north-east, north-west, south-east or south-west, and slope was initially classified as: 0–1, 1–2, 2–3, 3–4, 4–5, 5–6, 6–7%, and >7%. Four hundred and thirty-five observations were then statistically clustered using slope, SOM, TN and TP, and slope was determined to group naturally into 5 classes (0–2, 2–3, 3–4, 4–5, and >5%).

Statistics and Kriging analysis

Pearson correlations and multiple comparisons using the Least Significant Difference (LSD) method were carried out using SPSS 13.0 statistical software. The spatial distribution of SOM, TN and TP was determined by geostatistical analysis using GS+5.0 (Robertson 2000). Semivariograms were used in an autocorrelation analysis in order to evaluate the spatial dependence of the values, and best-fit models were optimized to predict SOM, TN and TP in space (Oliver et al. 2000). Semivariograms are calculated according to the formula: \( \gamma \left( h \right) = 1/2 N\left( h \right)\sum \left[ {z\left( {x_{i} + h} \right) - z\left( {x_{i} } \right)} \right]^{2} \) (Isaaks and Srivastava 1989), where γ(h) is the experimental semivariogram value at distance interval h, N(h) is the number of sample pairs within the distance interval h and z(x i ), z(x i  + h) is the sample value at two points separated by the distance interval h. The kriging algorithm was used to create an interpolated grid for development of isarithmic maps of the parameters SOM, TN and TP. The distribution of SOM, TN and TP values based on the 435 sample points were not normally distributed, Kurtosis and skewness values were then used to determine the goodness of fit to a normal distribution. So, all SOM, TN and TP data were transformed by the logarithm method in order to satisfy the required normal distribution. Kurtosis values of SOM, TN and TP were all close to 3, and skewness values were close to 0. After anisotropic analysis, the average semivariance of SOM and TN between 0° (North–South) and 90° (East–West), or between 45° (Northeast–Southwest) and 135° (Southeast-Northwest) were closer in the same separation distances, and K (h) (the ratio of average semivariance to 0°and 90°, or to 45°and 135°) values were close to 1 (Fig. 3). By this measure, SOM and TN can be considered as isotropic in the study area. The average semivariance of TP between 0°and 90°, or between 45° and 135°were closer at separation distances from 0 to 850 m and K(h) values were close to 1, while differences were larger at separation distances greater than 850 m. Separation distance was thus considered isotropic at distances under 850 m, and anisotropic at distances greater than 850 m. Thus, SOM and TN were predicted by ordinary Kriging methods for all separation distances in the study area, while TP was predicted by ordinary Kriging methods for separation distances of less than 850 m. The regression coefficient (R2) provides an indication of how well a model fits variogram data, but this value does not serve as well as the residual sum of squares (RSS) value for best-fit calculations involving changes in a model parameter. The lower the RSS, the better the model fits (Table 2). A Gaussian model was selected as the best for predicting SOM and TN distribution, while an Exponential model was selected for predicting TP (Table 2) .

Fig. 3
figure 3

Anisotropic experimental semivariogram of SOM (a), TN (c) and TP (e); the ratio of anisotropic semivarigram of SOM (b), TN (d) and TP (f)

Full size image
Table 2 Parameters from analysis of ordinary Kriging for soil nutrients
Full size table

The Gaussian isotropic model can be depicted as follows:

$$ \gamma (h) = C_{0} + C\left[ {1 - \exp ( - h^{2} /A_{0}^{2} )} \right] $$

Where h = lag interval, C 0  = nugget variance ≥ 0, C = structure variance ≥ C 0 , and A 0  = range parameter (not range).

The Exponential isotropic model can be depicted as follows:

$$ \gamma (h) = C_{0} + C\left[ {1 - \exp ( - h/A_{0} )} \right] $$

where h = lag interval, C 0  = nugget variance ≥ 0, C = structure variance ≥ C 0 , and A 0  = range parameter (not range) (Robertson 2000).

Results

Spatial distribution of soil nutrients

Values of SOM, TN and TP from the 435 samples ranged from 22.51 to 86.61, 0.98 to 4.26 and 0.26 to 1.80, respectively. Kriging analysis indicated that N/S (C 0/C 0 + C) of SOM, TN and TP ranged from 24 to 50%. N/C was lower in SOM than in TN, and TP was the highest. TP had a relatively longer A distance than TN, while SOM had the longest A distance (Table 2).

There was a strong positive correlation between SOM and TN (Table 3), and their spatial distributions were similar in the study area (Fig. 4a, b). Spatial variability of both SOM and TN were highly significant. SOM and TN typically increased from southeast to northwest.

Table 3 Pearson correlation coefficients among soil nutrients and slopes
Full size table
Fig. 4
figure 4

Distribution of SOM (a), TN (b) and TP (c) in the study area, as estimated by kriging

Full size image

In contrast to SOM and TN, TP exhibited only minor spatial variability (Fig. 4c), with only one area in the northwest showing a concentration of over 1.5 g kg−1, which is more than 7 times higher than the lowest values in the southeast. TP was positively correlated with SOM and TN (Table 3) in that it typically increased from the southeast to the northwest (Fig. 4).

Slope position and soil nutrients

It is obvious that SOM, TN and TP were all higher at higher elevations than at lower elevations (Figs. 2 and 4). Concentration gradients were steep from the back slope to the summit,but gentle from the back-slope to the toe-slope. Mean SOM values at the summit were 9 and 10% greater than on the back-slope and toe-slope (Table 4), respectively, and mean TN values were 9 and 9% greater at the summit than on the backslope and toe-slope respectively. There was no significant difference in TP at different slope positions (Fig. 5).

Table 4 Soil nutrient content and slope class
Full size table
Fig. 5
figure 5

Soil nutrient content at different slope positions. Histograms denoted by the same letter within a group are not significantly different by the F test (sig. ≤ 0.05)

Full size image

Slope steepness and soil nutrients

Slope steepness was negatively correlated with SOM (P < 0.01), TN (P < 0.05) and TP (Table 3), and can be considered as a key factor affecting the spatial distribution of soil nutrients in the study area. TN was significantly higher on slopes of 0–2 and 2–3% than on slopes of 4–5 and >5% (P < 0.05) and was 15.7% greater on slopes of 0–2% than on slopes of >5% (Table 4). TN was also lower on slopes of 4–5% than on slopes of >5%, but the relationship was not significant. SOM was significantly higher on slopes of 0–2, 2–3 and 3–4% than on slopes of >4% (P < 0.05) and it was 26.3% higher on slopes of 0–2% than on slopes of >5%. TP decreased with increasing slope degree at all 435 sample sites, but no significant difference was found among slopes.

Slope aspect and soil nutrients

In this section, SOM, TN and TP were analysed with respect to slope aspect only (Fig. 6). Both SOM and TN were typically highest on slopes with a north-eastern aspect, with decreasing values on south-east facing slopes, flat land, north-west facing slopes and slopes with a south-west aspect. TP was also highest on slopes facing north-east. The lowest values of SOM and TN were found on slopes with a southern aspect, while TP was lowest on non-sloping land. SOM, TN and TP were significantly greater (21.7, 22.6 and 9.7%, respectively) on north-eastern facing slopes than on those with a south-western aspect (P < 0.05), while SOM and TN were significantly greater (14.5 and 13.3%, respectively) on south-eastern slopes than on south-western slopes (P < 0.05) (Fig. 6).

Fig. 6
figure 6

Soil nutrient content under aspect (NE, Northeast; SE, Southeast; NW, Northwest; SW, Southwest). Histograms denoted by the same letter within the same group are not significantly different by LSD’s multiple range test (P ≤ 0.05)

Full size image

Slope steepness and aspect and soil nutrients

On slopes with a north-eastern aspect, both SOM and TN values were higher on slopes of 0–2% than on any other slope class and were lowest on slopes of 4–5% (Table 5), but the only significant difference (P < 0.05) was between slopes of 0–2% and those of 3–4% and steeper. SOM and TN were 18.3 and 18.7% greater, respectively, on slopes of 0–2% than on slopes of 4–5%. TP decreased with all slope increases, but not significantly.

On slopes with a north-western aspect, SOM was significantly higher on slopes of 0–2% than on any of the steeper slope classes, and was 32.4% greater than on slopes of 2–3%. TN showed the same trend (higher on slopes of 0–2% than on any of the steeper slope classes), except that the difference between 0 and 2% and >5% was not significant. TN was 31.3% greater on slopes of 0–2% than on slopes of 2–3%. TP showed the lowest values on slopes of 2–3 and 3–4%, but there were no significant differences.

On slopes with a south-eastern aspect, SOM was higher on slopes of 3–4% than on slopes of 0–2 and 2–3%, but then decreased with further increases in slope steepness. SOM was highest on slopes of 3–4%, but was only significantly higher (22.0%) than on slopes of >5%. Both TN and TP were higher on slopes of 3–4% than on any other slope class, but no significant difference was found.

On slopes with a south-western aspect, SOM was significant higher on slopes of 0–2, 2–3 and 3–4% than on slopes >5%, and was 23.7% lower on slopes >5% than on slopes of 3–4%. Both TN and TP showed a declining trend as slope steepness increased up to 4–5%, but then showed a slight increase again on slopes >5%. However, none of the differences were significant.

Table 5 Distribution of soil nutrient content under slope and aspect
Full size table

Land management and soil nutrients

Based on the F test analysis (P < 0.05), SOM, TN and TP were different under down-slope tillage than under cross-slope tillage. Under down slope tillage, mean SOM ranged from 26.82 to 59.49 g kg−1, TN ranged from 1.32 to 2.95 g kg−1, and TP ranged from 0.54 to 1.13 g kg−1. Under cross slope tillage, mean SOM ranged from 28.40 to 67.46 g kg−1, TN ranged from 1.24 to 2.95 g kg−1, and TP ranged from 0.86 to 1.15 g kg−1. SOM, TN and TP were statistically greater under cross-slope tillage than under down-slope tillage by 33.8, 23.3 and 22.4%, respectively.

When slope position, slope steepness and tillage practice under identical aspects were considered together to evaluate soil nutrient content in the study area, regression equations between the landscape and management factors and SOM and TN were as presented in Table 6.

Table 6 Regression between soil nutrient content and factors of slope steepness, slope position and tillage practice under identical aspects
Full size table

Discussion

The proportion of the spatial structure (Nugget/sill, N/C) of <0.25, 0.25–0.75, and >0.75 can be used to describe strong, moderate, and weak spatial autocorrelation, respectively (Zhang et al. 2007). This proportion of spatial structure determines the ratio by which random factors induce spatial variability (Cambardella et al. 1994). Geo-statistical results indicate that spatial autocorrelations as influenced by human activities were moderate for TP (0.50) and TN (0.43) and strong for SOM (0.24), due primarily to the long history of intensive management on these fields (Soil survey service of Hailun 1985; Zhang et al. 2007). The relatively shorter autocorrelation distance (R) for TP and TN compared to SOM can be attributed to long-term nitrogen and phosphorus fertilizer application in the study area (Yan et al. 2008; Liu et al. 2006).

Surface soil in the middle of slopes has suffered serious erosion due to slope steepness and length. In many cases parent material has been exposed in the middle of the slopes and has been eroded down the slopes such that it covers topsoil at the bottom. Under these conditions, SOM, TN and TP showed lower values at lower slope positions. In this study, TP in one part of the study area was seven times greater at the summit than in at the toe-slope of the slope. Phosphorus compounds in surface soils are mostly insoluble and unavailable for plant uptake; even when soluble compounds in fertilizer and manure are added to the soil, they become fixed and not easily moved (Brady and Weil 2000). Applied phosphorus is thus concentrated in the surface layer of the soil and is easy removed from slopes through erosion of the soil.

When aspect was not considered, SOM, TN and TP were all shown to generally decrease with increases in slope steepness. SOM showed significantly lower levels on slopes of >4% and steeper, and TN showed decreased values on slopes of >3% and steeper. These results could be attributed to soil erosion on the steep portions of slopes, since soil erosion occurs readily in areas with steep slopes, heavy textured soils and poor drainage such as in the black soil area of northeast China (Agassi et al. 1990; Shen 1998; Yan et al. 2008). When slope steepness was neglected, SOM and TN were typically higher on slopes with a north-eastern aspect, followed by those facing south-east, flat, north-west and south-west. This can be explained by the fact that wet soil conditions reduce decomposition of soil organic matter (Anderson 1987) and since evaporation is higher (Soil survey service of Hailun 1985) and crop growth is more vigorous on south-facing slopes (Rundel 1981), more nutrients would be taken up by the crops and SOM, TN and TP levels would be reduced. Another factor reducing SOM, TN and TP in slopes with a south-western aspect could be that the intense summer rain with the prevailing winds from the south-west cause more runoff of the intensively cultivated soils (Agassi et al. 1990).

However, when both aspect and slope steepness were considered together, SOM and TN were significantly higher on slopes of 0–2% than on slopes of 3–4 and 4–5%, and were higher in areas with a northern aspect (including northeast and northwest) than southern aspect (including southeast and southwest). SOM and TN were also significantly higher on the slopes of 0–2, 2–3 and 3–4% than on the slope of >5% in locations with a south-eastern aspect, and SOM was significantly higher on slope classes of 0–2, 2–3 and 3–4% than on slopes of >5% in areas facing south-west. The results obtained when both slope conditions were considered together differed from those when only slope steepness or aspect were considered independently because climate affects SOM, TN and TP distribution differently in both situations. This indicates that in order to analyze the spatial distribution of SOM, TN and TP realistically, and thus to manage nutrients effectively in the field, both aspect and slope should be considered with respect to local climate.

When compared across identical slope aspect or slope degree, SOM, TN and TP levels under cross-slope farming were significantly different than under down-slope farming, a result that is due to the fact that cross-slope tillage reduces soil loss effectively (Van Doren et al. 1950) and thus retains higher nutrient levels in the field. These results support the use of cross-slope tillage as an effective management option, especially on steep slopes and those with a south-western aspect.

Conclusions

It is clear from the results that both landscape and land management practices affect the dynamics of SOM, TN and TP in the study area. SOM, TN and TP distributions are spatially heterogenous in the study area, and in-depth analysis lead to the conclusion that nutrient spatial variability is mainly due to soil loss, which is in turn influenced by local climate, growing conditions and tillage practices. Generally, SOM and TN levels were much lower on slopes with a northern aspect and steeper than 3% and on all slopes with a southern aspect. TP was not significantly different on slopes of different steepness, but was significantly higher on slopes with a north-eastern aspect than on slopes facing other directions. Corn and soybean crops showed no significant effect on the distribution of SOM, TN and TP, but tillage direction is a very important management factor affecting soil loss and soil nutrient distribution. It was concluded that cross-slope tillage could be adopted as a good soil and nutrient management practice in the study area. It can also be recommended that farmers increase (e.g. by applying more fertilizer) and maintain residue cover in fields with a southern aspect, especially when slope degree is >3% in south-western facing slopes.

From the results presented in this study, it can be concluded that reducing soil erosion would be a highly effective means of improving and maintaining soil fertility in the black soils of north-eastern China. The highest priority for controlling erosion would be to increase land cover, especially on south-western facing slopes during the rainy season. Farmers could apply more fertilizer (i.e. chemical fertilizer and manure) on slopes with a southern aspect, especially steep slopes facing south-west, and on slopes with a north-western aspect, because SOM, TN and TP tend to be lower in those areas. Cross-slope tillage is also beneficial for reducing soil loss and holding nutrients in place. However, in the black soil region in north-eastern China, in order to improve water drainage in spring for early sowing most farmers cultivate their fields down–slope, leading to soil loss during the rainy season and hence decreased soil productivity (Yan et al. 2008). Tile drainage systems have been successfully applied in fields to control flooding and soil water movement in North America, and this has been helpful in lowering soil–water content and reducing soil and nutrient loss (Tang et al. 1998; Bottcher et al. 1981; Gentry et al. 1998; Buhler et al. 1993). A complete, integrated management system consisting of increased land cover on particularly sensitive slopes, cross-slope tillage and installation of tile drainage systems could be highly beneficial in maintaining soil and nutrient levels in the Black soil region of north-eastern China. Cross-slope tillage could be particularly difficult to implement, as the fields tend to be long and narrow, and where the narrow dimension coincides with the contour, the ability to use tillage machinery cross-slope is very limited. However, the results of this study could be used to identify particularly vulnerable locations where field realignment would enable cross-slope tillage.