JEQ Grow Your Career With ASA
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Related articles in JEQ
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (9)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Heckrath, G.
Right arrow Articles by Zhang, Y.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Heckrath, G.
Right arrow Articles by Zhang, Y.
Agricola
Right arrow Articles by Heckrath, G.
Right arrow Articles by Zhang, Y.
Related Collections
Right arrow Soil Fertility and Productivity
Right arrow Tillage
Right arrow Field-Scale Studies
Right arrow Spatial Variability
Right arrow Soil Erosion
Published in J. Environ. Qual. 34:312-324 (2005).
© ASA, CSSA, SSSA
677 S. Segoe Rd., Madison, WI 53711 USA

TECHNICAL REPORTS

Landscape and Watershed Processes

Tillage Erosion and Its Effect on Soil Properties and Crop Yield in Denmark

G. Heckratha,*, J. Djurhuusa, T. A. Quineb, K. Van Oostc, G. Goversc and Y. Zhangb

a Department of Agroecology, Danish Institute of Agricultural Sciences, Research Centre Foulum, 8830 Tjele, Denmark
b Department of Geography, University of Exeter, Amory Building, Exeter, EX4 4RJ, UK
c Laboratory for Experimental Geomorphology, Catholic University of Leuven, Redingenstraat 16, 3000 Leuven, Belgium

* Corresponding author (goswin.heckrath{at}agrsci.dk)

Received for publication April 12, 2004.

   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Tillage erosion had been identified as a major process of soil redistribution on sloping arable land. The objectives of our study were to investigate the extent of tillage erosion and its effect on soil quality and productivity under Danish conditions. Soil samples were collected to a 0.45-m depth on a regular grid from a 1.9-ha site and analyzed for 137Cs inventories, as a measure of soil redistribution, soil texture, soil organic carbon (SOC) contents, and phosphorus (P) contents. Grain yield was determined at the same sampling points. Substantial soil redistribution had occurred during the past decades, mainly due to tillage. Average tillage erosion rates of 2.7 kg m–2 yr–1 occurred on the shoulderslopes, while deposition amounted to 1.2 kg m–2 yr–1 on foot- and toeslopes. The pattern of soil redistribution could not be explained by water erosion. Soil organic carbon and P contents in soil profiles increased from the shoulder- toward the toeslopes. Tillage translocation rates were strongly correlated with SOC contents, A-horizon depth, and P contents. Thus, tillage erosion had led to truncated soils on shoulderslopes and deep, colluvial soils on the foot- and toeslopes, substantially affecting within-field variability of soil properties. We concluded that tillage erosion has important implications for SOC dynamics on hummocky land and increases the risk for nutrient losses by overland flow and leaching. Despite the occurrence of deep soils across the study area, evidence suggested that crop productivity was affected by tillage-induced soil redistribution. However, tillage erosion effects on crop yield were confounded by topography–yield relationships.

Abbreviations: AWC, available water capacity • LS, length–slope • SOC, soil organic carbon


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 
DURING THE PAST DECADE tillage erosion has been identified as a major process of soil redistribution within sloping arable fields in many different agroenvironments (e.g., Quine et al., 1997; Lobb et al., 1999; Van Muysen et al., 2002). Tillage erosion is caused by a variation in the magnitude of soil movement during tillage along a hillslope. Characteristically, tillage erosion removes soil at convexities such as crests and shoulderslopes and deposits it again at the concavities of footslopes and hollows. Hence, tillage-induced soil redistribution primarily depends on the change in slope gradient (i.e., profile curvature), which is why the linear slope sections remain stable (Govers et al., 1994).

The magnitude of soil redistribution due to tillage on hummocky land in temperate climates may often exceed that of water erosion (e.g., Govers et al., 1996). Recent estimates of tillage erosion and deposition rates based on radionuclide (137Cs) tracer studies, which integrated soil redistribution over a period of approximately 40 yr, frequently exceeded 10 Mg ha–1 annually at eroding and aggrading sites of intensively cultivated land (e.g., Govers et al., 1996; Van Oost et al., 2003). Controlled tillage experiments employing modern tillage practices with a combination of several tillage operations showed even higher tillage erosivities (e.g., Lindstrom et al., 1990; Lobb et al., 1995; Gerontidis et al., 2001). Typically, the moldboard plow was the most erosive implement (Van Muysen et al., 2002). Therefore, tillage has the capacity to alter soil profile height by several millimeters annually at both eroding and aggrading sites and is a major contributor to the distributions of colluvial soils on arable land.

Such extent of soil redistribution affects the spatial pattern of soil properties and eventually soil productivity of arable land (Miller et al., 1988; Moulin et al., 1994). Subsoil is incorporated into the plow layer at eroding sites while former plow layer soil is buried below the reach of the plow with the continuous buildup of soil at aggrading sites. This has important implications for dynamic processes such as SOC and nitrogen turnover and storage in soils (Liu et al., 2003). Additionally, when subsoil with different textural characteristics is mixed into the plow layer, soil structure and water regime may be affected (Schumacher et al., 1999). It is well documented that convexities often are the least productive parts of the landscape and erosion was seen as one of the most probable causes (e.g., Battiston et al., 1987; Pennock et al., 1994; Verity and Anderson, 1990).

However, only a few studies have explicitly investigated the relationship between tillage erosion and soil quality (Kosmas et al., 2001; Quine and Zhang, 2002). In Northern Europe large, intensively cultivated areas extend over the rolling topography of morainal landscapes and are hence vulnerable to tillage erosion. We need to improve our understanding of the effect of tillage erosion on soil property evolution for maintaining a high level of soil quality and especially productivity. To this end a sound basis is required for supporting decisions on long-term land management strategies regarding tillage practices. Additionally, the indirect effects of tillage-induced soil redistribution on nutrient losses from agricultural land need to be assessed. We undertook a detailed investigation of the extent of tillage erosion and its effect on soil quality and productivity on a hummocky field site in typical arable production in Northern Jutland, Denmark. The objectives of our study were to (i) determine the extent of past soil redistribution by tillage erosion on an arable hillslope in Denmark; (ii) map the spatial distribution of selected soil properties; and (iii) relate soil redistribution to the spatial variation of soil properties, yield, and topographic attributes.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Site Description
The field site was situated in Northern Jutland, Denmark (UTM N 6357607, E 589761), near the town of Sæby. In this part of the country, the 137Cs depositions from Chernobyl reactor accident were comparatively low, thus limiting the interference with the 137Cs inventories resulting from atmospheric bomb tests. The field had been in continuous arable cultivation for more than 100 yr and was bordered by permanent grassland suitable for obtaining the 137Cs reference inventories. In the 1960s and 1970s, the crop rotation typically included spring barley (Hordeum vulgare L.), turnip (Brassica rapa L.), oat (Avena sativa L.), and fodder beet (Beta vulgaris L.), while later mainly cereals and small grains (oil seed rape, Brassica napus L.) were grown. Up to approximately 1970 the field received farmyard manure and thereafter only mineral NPK fertilizers. Before 1970 about 30 kg P ha–1 were applied annually; thereafter, the rate dropped to 20 to 30 kg P ha–1. Straw was always removed. The field had been moldboard-plowed once annually since the mid 1950s nominally at a depth of 0.2 to 0.25 m before 1985 and a depth of 0.25 to 0.27 m thereafter. Additionally, the field was harrowed and tined in connection with stubble cultivation and seedbed preparation. At a nearby weather station (Sæby) the mean annual rainfall (1961–1990) and the mean daily temperature were 666 mm and 7.3°C, respectively. Rainfall is distributed fairly evenly throughout the year with on average 318 mm falling between April and September and 348 mm between October and March.

Following a detailed topographic survey of the Sæby field with an electronic theodolite in July 1997, a topographic analysis was performed in Surfer (Golden Software, 1999). The radial basis function was used as interpolation method and a digital elevation model (DEM) was constructed at a grid resolution of 2.5 m, corresponding to the surveying resolution (Fig. 1a) . The terrain attributes slope gradient, profile curvature, and plan curvature (Mitasova and Hofierka, 1993) were used to subdivide the study area into the landform elements shoulder (SH), backslope (BS), footslope (FS), and toeslope (TS), which additionally were classified as divergent (D) or convergent (C) (Pennock and Corre, 2001) (Fig. 1b). The field occupied a typical hillslope of terminal moraine rising to 58 m above sea level and dating from the latest (Weichsel) glaciation about 15000 yr ago. The central feature of the site was a major knoll that occupied about one-third of the whole field. A relatively narrow zone of shoulderslopes turned into a steep backslope (12°) stretched toward the east while the northwestern and southern sides had gentler backslopes of 8°. A shallow valley (thalweg) ran in the west–east direction at the northern field border. The relief thus turned concave to the northwest and the east and ran out in <4° toeslopes. Of the actual study area, 18% was covered by shoulderslopes, 35% by backslopes, 18% by footslopes, and 29% by toeslopes. The soils developed on glacial till that was interspersed with a large clay lens of dislodged Older Yoldia marine deposits. The soils were classified in accordance with WRB (World Reference Base for Soil Resources) as mainly associations of Mollic Stagnic Cambisols and Mollic Stagnic Luvisols.




View larger version (93K):
[in this window]
[in a new window]
  		Fig. 1. (a) Terrain model of the Sæby site indicating the soil and crop sampling points. The open circles represent additional sampling points for 137Cs inventories only. The boundary of the digital elevation model corresponds to the field border. (b) Distribution of different landform elements within the sampling area. The lines denote elevation contours and the cell size shown is 10 by 10 m. D, divergent; C, convergent; SH, shoulder; BS, backslope; FS, footslope; TS, toeslope.

 
Soil and Crop Sampling and Analyses
The study was conducted on a representative 1.9 ha in the eastern half of the field covering the relevant landscape positions for investigating tillage erosion. For measuring 137Cs inventories, 135 sampling points were established on the basis of a regular 15- by 15-m grid, though the grid spacing was reduced to 7.5 by 15 m at the most convex and concave parts of the sampling area (Fig. 1a). Additionally, 10 sampling points were selected along a 75-m transect at the reference site. In spring 1998 undisturbed soil cores of 93 mm in diameter and 900 mm in length were collected with a cylindrical auger. The depth of the A horizon was recorded before the cores were divided into sections corresponding to soil depths of 0 to 0.25, 0.25 to 0.35, 0.35 to 0.45, and 0.45 to 0.90 m. After air-drying and separation of stones > 6 mm, the soils were ground to <2 mm. For 137Cs analysis, the subsamples of each core were proportionally recombined to represent the whole core, and 137Cs activity (Bq kg–1) was measured by {gamma} spectrometry, using a coaxial germanium detector at 662 keV and multichannel system in the Department of Geography, University of Exeter. Counting times of approximately 30000 s gave an analytical precision of ±6%. The 137Cs activities were converted into 137Cs inventories (Bq m–2) based on sample weight and cross-section. The 137Cs reference inventory was 2430 Bq m–2 (SD = 32.9, n = 10).

Further soil analyses were performed on 114 soil cores omitting the marginal transects along the northern and southern field border (Fig. 1a). From each core the 0- to 0.25-, 0.25- to 0.35-, and 0.35- to 0.45-m layer was analyzed for pH (CaCl2), soil texture including CaCO3 (Gee and Bauder, 1986), SOC by dry combustion (Nelson and Sommers, 1982), and total P and plant-available P (i.e., Olsen P; Olsen and Sommers, 1982). As tillage-induced soil translocation on arable land affects the soil properties of the upper part of the solum, the 0.45- to 0.9-m layer was not further analyzed. Soil pH varied only little within the study area, partly due to regular liming around every fifth year, and will not be further discussed. Average pH values of 6.7, 6.4, and 6.3 were determined in the three respective soil layers. The available water capacity (AWC) to a depth of 0.45 m was derived from pedotransfer functions (Madsen, 1986). To assess the spatial pattern of crop yield, winter barley was harvested at the beginning of August 1997 by taking 0.5-m2 plot samples at the 114 grid points. Grain yield is reported at 15% water content.

The Cesium-137 Technique
A detailed discussion of the 137Cs technique is given elsewhere (e.g., Walling and Quine, 1991; Quine, 1999). The radionuclide 137Cs is present in the environment mainly as the result of fallout from nuclear weapon tests, which occurred mainly between the mid-1950s and the mid-1960s. As deposition took place over such a long period, it is generally assumed that fallout was essentially uniform at the local scale. Due to its strong sorption in mineral soils, subsequent 137Cs redistribution has almost exclusively occurred with sediment transport. Therefore, the spatial distribution of 137Cs at a given time represents the net effect of all soil redistribution processes that had occurred since fallout began. The application of the 137Cs technique to erosion assessment is based on the comparison of the contemporary spatial distribution of 137Cs inventories in the cultivated study area and the 137Cs inventory of an adjacent undisturbed and hence uneroded area (e.g., Walling and Quine, 1991). The differences between 137Cs inventories of the cultivated and reference sites (i.e., the relative loss or gain in 137Cs) may be used directly to examine qualitative patterns of soil redistribution. To derive quantitative estimates of erosion, however, it is necessary to establish a site-specific calibration relationship relating the change of the 137Cs inventories to erosion rates (Quine, 1999). The 137Cs calibration model of Quine et al. (1997) was used to convert 137Cs inventories into soil redistribution rates, which combines a 137Cs mass-balance model with a two-dimensional model of radionuclide redistribution by tillage. This approach assumes that water erosion plays an insignificant role in the total soil and 137Cs redistribution, and the calibration does not yield quantitative estimates of water erosion. Soil fluxes due to tillage for successive passes in opposing directions are calculated using a simple topographic relationship:

[1]
where Qt is the net unit soil transport rate (kg m–1) in the downslope direction, {alpha} is the slope gradient, and kt (kg m–1) is the tillage transport coefficient that is indicative of the intensity of tillage practices (Govers et al., 1994). Soil movement by tillage was assumed to occur in the direction of the steepest slope. The model was used to establish the intensity of tillage erosion in the period from 1954 to 1998 that would have been required to derive the measured redistribution of 137Cs. The intensity of tillage erosion in the simulation was varied by alteration of the tillage transport coefficient. The estimated soil redistribution rates by tillage were those required to obtain optimum agreement between the observed and the simulated pattern of 137Cs loss and gain (Quine et al., 1997). Soil redistribution rates are annual averages for the simulation period.

Topographic Indices
To characterize the water erosion risk at the site, a distributed, topography-dependent index for soil erosion by water was calculated based on the specific catchment area and the slope gradient of a given grid point (Desmet and Govers, 1996), corresponding to a two-dimensional length–slope (LS) factor of the Universal Soil Loss Equation (USLE) (Wischmeier and Smith, 1978). The two-dimensional LS factors were estimated by the erosion modeling tool WATEM (Van Oost et al., 2000) with exponents taken from Govers (1991) and the transport capacity coefficient set to 150 m (Van Oost et al., 2000). A topographic wetness index was calculated according to Moore et al. (1991) based on the ratio of specific catchment area and slope gradient. The index indicates the distribution of drier and wetter zones in landscapes, which may be an important control of the spatial pattern of soil properties (Moore et al., 1993). The indices were used to discuss the potential effects of water erosion and water redistribution on a hillslope on soil properties at the study site.

Statistical Methods
Correlation analyses were done for selected soil properties, grain yield, erosion variables, and topographic attributes. To map the spatial variability of soil properties a geostatistical analysis was done. Empirical semivariograms were estimated for all variables assuming intrinsic stationarity (Eq. [2]) (i.e., trends within the field were not allowed). This assumption was judged graphically. The experimental semivariances were estimated for 5-m intervals up to a distance of 105 m. This distance was about 53 to 60% of the maximum distance between any pair of observations. The reason for not using the entire experimental semivariogram is that it becomes unreliable at larger distances:

[2]
where (h) is the semivariance for distance h, Nh is the number of pairs of observations within distance class h, and Z(xk) is the measurement at point xk. The data did not support the assumption of anisotropy when the semivariances were evaluated for four different direction classes of 45° each. Exponential (Eq. [3]), spherical (Eq. [4]), or linear (Eq. [5]) semivariogram models were fitted to the experimental semivariances:

[3]

[4]

[5]
where C0 is the nugget effect, C1 is the partial sill, a is the range for the spherical model, and r is a distance parameter controlling the range of influence for the exponential model. The effective range, defined as the distance at which the spatial-dependent part of the model is approximately 0.95C1, equals 3r1 for the exponential model.

The parameters were fitted by the least square method in nonlinear regression analyses using the number of pairs of observations within each distance class as weights. When the nugget effect (C0) was not significantly different from zero it was left out and the remaining parameters were reestimated. When more than one model could be chosen, the choice was based on the Akaike Information Criterion for choosing models (Akaike, 1973). All calculations were performed in SAS (SAS Institute, 1999).

The fit of semivariogram models was assessed with the Jackknifing procedure (Vauclin et al., 1983) implemented in GSTAT (Pebesma and Wesseling, 1998). Based on the estimated semivariograms, spatial averages of the measured soil and crop variables were interpolated in GSTAT for 10- by 10-m pedocells by ordinary blockkriging (Cressie, 1993). The observations used in the kriging were restricted to 25 within a radius of 100 m from the center of the predicted pedocell.


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Soil Redistribution
Substantial soil redistribution had occurred at the Sæby site since the middle of the 1950s indicated by the distribution of the observed 137Cs losses or gains (Fig. 2) , which followed a clear spatial pattern (Table 1). Maximum 137Cs losses were found on the divergent shoulder and backslope positions averaging –31 and –24%, respectively. Also, the convergent backslopes had on average incurred substantial 137Cs losses, while losses from the convergent shoulderslopes were only one-third that from the divergent shoulderslopes (Table 2). The highest 137Cs gains occurred in concave areas of the thalweg and a section of the eastern footslopes. Further 137Cs gains were found on the convergent footslopes and toeslopes; however, on average, 137Cs gains here were much lower compared with losses from the eroding areas. Due to analytical and sampling variability, only 137Cs losses or gains differing by ±10% from the 137Cs inventory of the reference site were assumed significant (Walling and Quine, 1991). Hence, the range between –10 and +10% represented stable conditions regarding soil redistribution and comprised 36% of the study area at Sæby (Fig. 2). Accounting for 49%, areas of 137Cs losses clearly exceeded those with 137Cs gains (15%).



View larger version (34K):
[in this window]
[in a new window]
  		Fig. 2. Spatial distribution of relative 137Cs losses or gains calculated as percentages of the 137Cs reference inventory for the sampling area. Negative values indicate soil loss and positive values indicate soil gain. Values between –10 and +10% represent stable areas.

 

View this table:
[in this window]
[in a new window]
  		Table 1. Descriptive statistics of the measured variables and parameters of semivariogram models.

 

View this table:
[in this window]
[in a new window]
  		Table 2. Mean values of 137Cs losses or gains, estimated tillage erosion rates, and soil properties in different soil layers for different landform elements.

 
The predicted net soil redistribution rates due to tillage (i.e., tillage translocation rates) are shown in Fig. 3 for the 45-yr period. The highest tillage erosion rates occurred on the divergent shoulderslopes around the central knoll averaging 2.7 kg m–2 yr–1 for this landform element and exceeding 4.5 kg m–2 yr–1 in some places. In comparison, the predicted tillage deposition rates were generally lower. Considerable deposition was predicted mainly on the convergent foot- and toeslopes averaging between 1.1 and 1.3 kg m–2 yr–1 (Table 2). Deposition exceeded 2.5 kg m–2 yr–1 only in a small area of the thalweg (Fig. 3). Hence, areas of maximum erosion had lost about 0.15 m of soil while depositional areas had gained up to 0.10 m over a period of 45 yr. Tillage erosion rates in access of 0.5 kg m–2 yr–1 were found on 48% of the study area, with more than 30% of the area exhibiting high to very high erosion rates (>1.5 kg m–2 yr–1). In contrast, significant deposition rates were restricted to 32% of the study area. In summary, eroding sites averaged soil losses of 2 kg m–2 yr–1 whereas depositional sites received 1.1 kg m–2 yr–1 by tillage translocation. The net erosion rate was 0.6 kg m–2 yr–1, indicating net soil loss from the analyzed portion of the field. It is expected that a proportion of this exported soil was deposited at the eastern field margins not covered by the 137Cs sampling scheme. During the study period there was no evidence for export of soil from the field by water erosion.



View larger version (28K):
[in this window]
[in a new window]
  		Fig. 3. Spatial pattern of modeled tillage erosion and deposition rates within the sampling area.

 
Soil Property and Yield Measurements
The measured soil properties, like 137Cs losses and gains, exhibited a systematic spatial pattern in the study area (Table 1). Soil texture varied considerably. In the 0- to 0.25-m layer, silt content remained between 20 and 40%, clay content ranged between 10 and 60%, and sand content showed a trend opposite to clay (Fig. 4) . Hence, soil types ranged from sandy loam over loam, clay loam to clay. The spatial variability of clay content increased below the plow layer ranging from 7 to 74 and 4 to 67% at 0.25- to 0.35- and 0.35- to 0.45-m depths, respectively (Fig. 5) . The highest clay contents throughout the soil profile were found on the shoulderslopes, the lowest on the footslopes, while backslopes and toeslopes had intermediate contents. Clay contents in this order of magnitude are likely to have a negative effect on the available water capacity (AWC) (Madsen, 1986). Thus the lowest values of AWC at the 0- to 0.45-m depth coincided with the highest clay contents at Sæby (r = –0.64, p < 0.001; data not shown). Because AWC was only determined to a 0.45-m depth, we estimated the AWC of the rootzone to a 0.9-m depth by assuming the AWC of the 0.35- to 0.45-m layer was representative for the lower part of the rootzone, too. This way we obtained AWC values averaging 165 mm, with 5th and 95th percentiles of 143 and 200 mm, respectively.



View larger version (39K):
[in this window]
[in a new window]
  		Fig. 4. Textural composition of soil samples (n = 114) from the 0- to 0.25-m layer.

 


View larger version (24K):
[in this window]
[in a new window]
  		Fig. 5. Distribution of blockkriged clay contents in three soil layers within the sampling area.

 
Inevitably, soil redistribution in agrolandscapes affects the spatial distribution of the A-horizon depths. At Sæby the thickness of the A horizon at the sampling points was highly variable, ranging between 0.15 and 0.72 m (Table 1). Shallow A horizons of <0.25 m were predominantly observed on the upper shoulder and the northeastern backslope, covering about 35% of the study area (Fig. 6) . Deep A horizons generally occurred in the thalweg and at the southeastern footslopes. Average A-horizon depths for landform elements measured 0.24, 0.28, 0.33, and 0.34 m on shoulder-, back-, foot-, and toeslopes, respectively (Table 2). At a few isolated sampling points in the lower part of the field unusually thick A horizons were found exceeding the 0.5-m depth. Around 30% of the sampling area had deep A horizons of between 0.3 and 0.45 m thickness.



View larger version (35K):
[in this window]
[in a new window]
  		Fig. 6. Variation in depth of the A horizon within the sampling area.

 
Soil organic carbon (SOC) showed a high degree of variability and spatial dependence in all three soil layers (Table 1). The 5th and 95th percentiles of the measured SOC concentrations in the three soil layers were 7.6 and 15.9, 2.4 and 12.9, and 1.8 and 12.9 g kg–1 in order of increasing depth (Fig. 7) . The averages of the spatially interpolated SOC concentrations were 12.4, 6.3, and 5.3 g kg–1 soil in the respective layers. These SOC concentrations are low compared with those reported for clayey agricultural soils in Denmark (Heidmann et al., 2001). Overall SOC contents showed a moderate increase from the shoulder- toward the toeslopes in the 0- to 0.25-m layer. Between the divergent shoulderslopes with the lowest and the divergent toeslopes with the highest SOC contents the increase amounted to 25%. A much larger trend was observed at the 0.25- to 0.45-m depth, where the SOC contents of the same landform elements differed by a factor of more than two (Table 2).



View larger version (23K):
[in this window]
[in a new window]
  		Fig. 7. Distribution of soil organic carbon (SOC) concentrations in the three soil layers within the sampling area.

 
The measured total P also varied substantially in each soil layer (Table 1) with 5th and 95th percentile concentrations of 314 and 748, 175 and 623, and 143 and 565 mg kg–1 soil in the 0- to 0.25-, 0.25- to 0.35-, and 0.35- to 0.45-m layers, respectively (Fig. 8) . The averages of the spatially interpolated total P concentrations were 538, 366, and 316 mg kg–1 soil in the respective layers. In a survey of agricultural soils in Denmark average total P concentrations of 560 and 460 mg kg–1 soil were determined in the 0- to 0.25- and 0.25- to 0.5-m layers of cultivated clay loams (Rubæk et al., 2000), a soil type representative of one-third of our study area. Heavier soils, as found on large parts of the study area, tended to contain higher total P concentrations according to the survey by Rubæk et al. (2000). Therefore, the total P concentrations observed at Sæby were comparatively low. This was especially the case for the 0.25- to 0.35-m subsoil layer. Due to a characteristically exponential decline of total P in soil profiles (e.g., Johnston and Poulton, 1993), higher total P concentrations would be expected in the 0.25- to 0.35- than in the 0.25- to 0.5-m layer of a given soil. Averaged over the two lower depths total P concentrations of 342 mg kg–1 soil at Sæby were very close to those found at the same subsoil depths under ancient beech forests (Rubæk et al., 2000).



View larger version (23K):
[in this window]
[in a new window]
  		Fig. 8. Distribution of total phosphorus (TP) concentrations in three soil layers within the sampling area.

 
The spatial distribution of total P followed a similar trend compared with SOC with gradually increasing total P contents toward the lower lying areas (Fig. 8). At both the 0- to 0.25- and 0.25-to 0.45-m depths the divergent shoulder- and backslopes had the lowest total P contents, while the convergent toeslopes had the highest (Table 2). However, other than for SOC the relative increases in total P contents downslope were of the same order at both depths and total P contents increased by at most 1.5 times. Notable are the relatively low total P contents at both depths on the diverging footslopes. This leads to a different order of landform elements after increasing total P contents compared with SOC contents (Table 2).

The measured grain yield of winter barley varied between 3.3 and 9.7 Mg ha–1 within the experimental area (Table 1; Fig. 9) . About one-third of the area had yields below the average of 6.6 Mg ha–1 on the basis of the spatially interpolated data. The spatial pattern of yield distribution was similar to that of total P concentrations in the 0- to 0.25-m layer (Fig. 8). The yields were lowest around the summit region but tended to increase toward the low-lying parts of the field.



View larger version (32K):
[in this window]
[in a new window]
  		Fig. 9. Spatial pattern of winter barley grain yield in 1997.

 
Correlation Analyses
Correlation analyses were done to assess the potential interactions between tillage translocation, soil properties, and yield (Table 3). Clay content was significantly negatively correlated with the observed 137Cs losses or gains, as high clay contents coincided with soil loss at the central knoll. The highest positive correlations were obtained between 137Cs losses or gains and SOC concentrations, the SOC contents at 0- to 0.45-m depth, and A-horizon depth. Slightly weaker were the positive correlations between the 137Cs losses and gains and soil P. Similar correlations were found between tillage translocation rates and soil properties. However, while SOC variables were less strongly related to tillage translocation rates than to 137Cs losses and gains, P variables showed the opposite trend (Table 3). Expectedly, A-horizon depth and SOC contents at the 0- to 0.45-m depth were strongly correlated (r = 0.79, P < 0.001). The SOC contents at the 0- to 0.45-m depth were in turn particularly influenced by SOC at the 0.25-to 0.35-m depth (r = 0.91, P < 0.001). Clay content was not correlated with total P in the upper two soil layers, but was negatively correlated with SOC concentrations (r = –0.45 and r = –0.48, P < 0.001, 0–0.25 and 0.25–0.35 m, respectively). In the 0- to 0.25-m layer Olsen P and total P were highly correlated (r = 0.88, P < 0.001). Grain yield was not correlated with clay content. At the 0- to 0.25-m depth SOC had a moderate effect on grain yield, though not at greater depths. In contrast, P was more strongly correlated with grain yield also in the upper subsoil layer (Table 3).


View this table:
[in this window]
[in a new window]
  		Table 3. Pearson correlations between soil variables, 137Cs losses or gains, modeled tillage translocation, and yield (based on n = 114 data points).

 

   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
CONCLUSIONS
 REFERENCES
 
Soil Redistribution
The pattern of 137Cs loss at convexities and gain at concavities at the Sæby site was consistent with the characteristic spatial signature of soil redistribution by tillage (e.g., Govers et al., 1994; Lobb et al., 1999). A comparison between observed and predicted 137Cs losses and gains is a basis for evaluating the modeled tillage translocation rates. Reasonable agreement is found across the study area (Table 3). Around the shoulderslopes of the central knoll, the predicted 137Cs losses were, however, consistently overestimated corresponding to an overestimation of soil loss. Such deviation may be due to simplifications embedded in the calibration procedure that assumes tillage translocation to occur in the direction of the steepest slope and does not take into account all details of the plowing history including plowing direction (Quine and Zhang, 2004).

In a small area on the eastern footslope, predicted 137Cs gain was underestimated. This coincided with occasionally very deep A horizons and large 137Cs gains in the same area (see Fig. 2 and 6). In this respect, it should be noted that simulated rates of soil redistribution are based on the current topography, not the unknown initial (1950) topography. If the latter evidenced a pronounced small-scale concavity in this area, then higher rates of deposition would have occurred through normal tillage erosion and would have been simulated with appropriate data. The deep A horizons and excess 137Cs are clearly evidence of locally high soil deposition rates; however, without additional data it is not possible to establish whether these were due to rapid surface evolution or deliberate soil movement. Predicted 137Cs loss was underestimated on convergent backslopes and partly also on convergent footslopes, while 137Cs gain was overestimated on the toeslopes. This suggested that water erosion might have contributed to soil redistribution.

No record exists of the extent of water erosion at the field site. Neither the farmer nor extension officers had considered it significant from an agronomical point of view. Typically the risk of soil erosion by water is considered low in Denmark, mainly due to relatively low relief and erosivity (Veihe et al., 2003). A comprehensive, recent study found that the 75% quantile of annual soil loss by rill erosion during autumn and spring amounted to only about 0.2 kg m–2 on eroding arable sites in Denmark (Jørgen Djurhuus, unpublished data). Based on the USLE, Olsen and Kristensen (1998) calculated a low erosion risk for the Sæby parish. We used the two-dimensional LS factor of Desmet and Govers (1996) as a qualitative index of rill erosion at the study area. Clearly, the pattern of maximum 137Cs loss and gain could not be explained by the water erosion potential. Small LS factors (0 < LS < 5) indicating low water erosion potential coincided with the shoulderslopes (Fig. 10) , where the observed 137Cs losses were largest (Fig. 2). The northeastern toeslopes exhibited a low water erosion or even deposition potential, while soil deposition due to tillage was already overestimated in this area. Moderate LS factors failed to explain the substantial underestimation of soil loss on the northeastern backslope. Large LS factors indicated high water erosion potentials in the northern thalweg (Fig. 10), yet the observed 137Cs data indicated soil gain. The cluster of 137Cs gains on the eastern footslope (Fig. 2) could not be explained by LS factors indicating erosion, nor could the large 137Cs gains in the northern thalweg, which coincided with some of the largest LS factors (Fig. 2 and 10).



View larger version (29K):
[in this window]
[in a new window]
  		Fig. 10. Spatial distribution of the two-dimensional length–slope (LS) factor within the sampling area. Increasing LS values correspond to a higher water erosion potential. Values below zero (white) indicate soil deposition.

 
Our results strongly suggest that tillage translocation was by far the dominant process of soil redistribution at Sæby during the past decades, exceeding the typical magnitude of water erosion in Denmark by a factor of 10 (Veihe et al., 2003). The spatial pattern and the magnitude of tillage translocation at Sæby (Fig. 3) corresponded directly to findings from recent studies of erosion under comparable tillage systems in other Northern European countries employing the 137Cs technique (e.g., Walling and Quine, 1991; Quine et al., 1997). In the Mediterranean up to five times larger tillage erosion rates compared with Sæby were determined on intensively tilled, steep cropland (Tsara et al., 2001). Pennock (2003) presented landform element–soil redistribution associations for nine hummocky fields in Saskatchewan based on 137Cs studies. The tillage-attributed soil loss on the divergent shoulder- and backslopes were almost exactly of the same magnitude as those found at Sæby, and the soil deposition rates were similar (Table 2). Furthermore, the trend of soil redistribution within the suite of landform elements corresponded closely to the Sæby findings. This exemplifies the importance of tillage translocation across a broad range of agroenvironments as a major process of soil redistribution.

The tillage transport coefficient (Eq. [1]) is a direct measure of tillage erosivity (Govers et al., 1994). The tillage transport coefficient obtained by modeling the 137Cs redistribution was 456 kg m–1 at Sæby. Transport coefficients ranging from 230 to 350 kg m–1 for a single pass with a moldboard plow were reported from controlled tillage experiments (Lindstrom et al., 1990; Govers et al., 1994; Lobb et al., 1995). Considering that the transport coefficient at Sæby integrated various annual tillage operations, a value higher than found in tillage experiments would be expected (Van Oost et al., 2003).

Soil Property Measurements
Though tillage erosion does not have an immediate "off-site" effect, the large extent of soil redistribution within the field had important implications for soil property distribution at Sæby. Tillage translocation led to truncated A horizons on shoulder- and upper backslopes and conversely buried plow layers on foot- and toeslopes (Table 2). Hence, the positive spatial correlations between SOC variables at the different depths and tillage translocation (Table 3) suggest that plow layer soil became depleted in SOC on eroding sites as the original plowing depth was maintained and SOC-poor subsoil was progressively incorporated into the plow layer. Repeated tillage moved soil downslope where it accumulated to form deep, SOC-enriched soils. Similar trends of SOC distribution on sloping arable fields have been reported from various agroecosystems and have been linked to erosional processes, although rarely to tillage erosion (e.g., Kreznor et al., 1989; Gregorich et al., 1998; Manies et al., 2001). In a study on erosion–soil productivity relationships from southwestern England (Quine and Zhang, 2002), the observed high spatial correlations between 137Cs losses or gains, tillage translocation, and SOC contents were practically identical to those found at Sæby (Table 3). Clay content did not seem to affect the spatial pattern of SOC at Sæby as indicated by the rather uniform SOC contents in the plow layer across the study area (Table 2). Neither could the trend of increasing subsoil SOC contents toward the low-lying areas be related to clay. This is in accordance with another Danish study, which found no correlation between clay and SOC contents on a sloping arable field with a similarly wide range of clay contents (Schjønning et al., 1999). In a related study Thomsen et al. (1999) concluded that in the temperate climate in Denmark land use and soil management were the determining variables for SOC contents.

The significance of tillage translocation for soil profile anisotropy at Sæby is illustrated by a comparison of averaged soil property values in different erosion classes (Table 4). Stable areas are represented by tillage translocation rates of –0.5 to +0.5 kg m–2 yr–1. While SOC contents in the 0- to 0.25-m layer were 13% higher on aggrading compared with eroding areas, the difference was 38% in the 0.25- to 0.45-m layer. Further the ratio between SOC contents in the 0- to 0.25- and 0.25- to 0.45-m layer was 3.1 on eroding and 2.4 on aggrading areas. Ignoring dynamic processes of SOC turnover, a first approximation of SOC redistribution due to tillage translocation between erosion classes can be calculated based on the plow layer SOC concentrations and the soil redistribution rates. We obtained SOC changes of –22 and 15 g SOC m–2 yr–1 on eroding and aggrading areas, respectively (Table 4). There have been few comparable studies reporting SOC redistribution rates within arable fields. At hummocky arable field sites in the Canadian prairie, shoulderslopes had lost 45 g SOC m–2 yr–1, while footslopes had gained 23 g SOC m–2 yr–1 due to soil redistribution over a 56-yr period (Pennock et al., 1994). The authors did not quantify the contribution of tillage erosion, but considered it important. Interestingly, despite the differences in climate, soil type, and cropping history, and despite the fact that rates referred to distinct landform elements, the results of the Canadian and the Sæby study are broadly similar. Our findings suggest that tillage erosion may have important implications for SOC storage at the field scale as eroded SOC is deposited in a subsoil environment with typically much longer turnover times (Gill and Burke, 2002). Additionally, the denuded shoulderslope positions may have the capacity to bind extra atmospheric C (Liu et al., 2003). Therefore, the effect of tillage-induced SOC redistribution on C dynamics is an important area of future research.


View this table:
[in this window]
[in a new window]
  		Table 4. Arithmetic means of spatially interpolated (blockkriged) soil properties for eroding, aggrading, or stable areas at Sæby.

 
Like SOC, total P was another soil property that evidenced a spatial distribution and appears to be strongly affected by tillage translocation (Table 3). Quine and Zhang (2002) reported similarly strong spatial correlations between SOC and total P in the plow layer of a loamy soil, which had similar median SOC but higher total P concentrations. At Sæby, total P concentrations at aggrading sites exceeded those at eroding sites by 45, 62, and 30% in the 0-to 0.25-, 0.25- to 0.35-, and 0.35- to 0.45-m layers, respectively (Table 4). This indicated a lower rate of decline of total P in soil profiles on aggrading compared with eroding areas and therefore a larger P enrichment of the subsoil. As P is rather immobile in soil profiles exhibiting exponential depth distributions (e.g., Johnston and Poulton, 1993), subsoil enrichment is consistent with topsoil burial. However, the ratio between total P content in the plow layer and in the subsoil was the same at eroding and aggrading areas, especially due to high total P contents in the plow layer on aggrading areas. This behavior distinguished total P from SOC (Table 4) and together with the different pattern of spatial dependence for all P variables (Table 1) pointed to additional controls of total P distribution at the field. In general, the heterogeneity of the glacial till at Sæby can be expected to affect P soil–landscape relationships (e.g., Brubaker et al., 1993). Other than SOC, total P in soil profiles depends highly on parent material and fertilizer inputs and the effect of these variables is confounded with the effect of tillage translocation. From a soil fertility perspective it is noteworthy that approximately 35% of the equivalent of a typical annual P fertilizer application of 25 kg ha–1 is lost from eroding sites due to tillage (Table 4). Conversely, on aggrading sites substantial P accumulation occurs.

Implications for Soil Productivity
Due to its effect on the spatial distribution of soil properties, soil redistribution may eventually have an effect on crop yields (e.g., Lal, 1998). In a modeling exercise Schumacher et al. (1999) could directly show that tillage erosion reduced the crop production potential on hummocky cropland in South Dakota. We determined the crop yield at Sæby in one season only. As cereal yields are sensitive to the year-to-year changes in weather and disease pressure our results must be interpreted cautiously and can only provide a preliminary indication of the potential effect of tillage erosion on yields in a temperate climate. The average winter barley yield in the region (Northern Jutland) was 5.4 Mg ha–1 in the study year and hence about 20% lower than the yield at Sæby. This is largely explained by better soils in the study area compared with those prevalent in the region (A. Christensen, Agricultural Advisory Service, Northern Jutland, personal communication). Grain yield was only moderately correlated with most soil properties. Neither A-horizon depth nor AWC had a direct effect on yield (Table 3). In contrast, in many southern European agrolandscapes, A-horizon depth is perhaps the most important crop productivity parameter on land heavily affected by tillage erosion (Kosmas et al., 2001; Tsara et al., 2001). Other than for these studies, rooting depth should not have been limiting in our study area. The predicted AWC in the root zone at Sæby should, therefore, have been sufficient for cereals crops with a potential soil moisture deficit of 130 to 140 mm in Northern Jutland. Rainfall during the growing season 1996–1997 was close to the 30-yr average with a slightly wetter spring than usual. In a comprehensive study of soil–crop relationships on sloping loess soils, Auerswald et al. (1997) concluded that AWC had a minor influence on yield, and instead interactions of rooting depth, upslope drainage area, and fungal infection controlled crop performance. Clay content was not related to yield. However, the combination of very high clay and low SOC contents in the soils on the shoulderslopes resulted in an unfavorable soil structure and tilth and this may be expected to have a negative effect on crop production (Kay and Munkholm, 2004). It is, therefore, interesting to note that the farmer reported poor crop emergence and establishment on the shoulderslopes both in wet and dry years, and he attributed this to the difficulty of adequate seedbed preparation under these soil conditions.

Of the variables tested, total P at the 0- to 0.25-m depth was most strongly correlated with grain yield at Sæby. Interestingly, the effect of plant-available P (Olsen P) was lower pointing to an interrelation between total P and another yield-determining variable. Olsen P values at the field fell mainly within the low range of agronomic soil test P values with a 25th percentile of 14 and a median of 18 mg P kg–1 soil. This corresponded roughly to one-half the recommended range for cereal crops, and P deficiency may therefore have affected crop performance. As plant-available P contents in soils can readily be corrected by fertilization, P deficiency can only be considered a minor problem associated with tillage erosion. However, the example illustrates that when maintaining an environmentally desirable narrow nutrient balance, tillage erosion-induced soil nutrient variability may affect crop productivity.

A definitive effect of soil fertility parameters on crop productivity in the context of eroding landscapes is difficult to determine by simple statistical analyses. Terrain modifies the distribution of hydrological processes and energy inputs in landscapes and ultimately affects soil formation (Huggett, 1995). In a system of soil–landscape–climate interactions, topographic variables are separately correlated with soil attributes and crop productivity. Additionally, as yield and erosion are both correlated with the same topographic variables, the effect of erosion on yield cannot be separated (Moore et al., 1993; Moulin et al., 1994). Therefore, topography–yield relationships confound erosion effects as a result of water redistribution on a hillslope (Stone et al., 1985). Table 5 shows this for the Sæby site. The LS factor was poorly correlated with soil properties, indicating that water erosion played a minor role for soil property distribution at Sæby. Soil organic C and to a lesser degree also P parameters were highly positively correlated with profile curvature. Together with high correlations between profile curvature and 137Cs losses or gains as well as tillage translocation, this confirmed tillage erosion as the dominant process of soil redistribution at Sæby in accordance with earlier studies (Govers et al., 1996; Quine et al., 1997). In contrast to SOC, P contents were also strongly negatively related to terrain slope (i.e., soils on steep slopes contained less P). This had also been observed by Quine and Zhang (2002) and suggests other processes than mass movement of soil. Possible explanations may be nonuniformity of parent material, redistribution of dissolved P by overland flow, or likely uneven application of P fertilizers on steep slopes. Profile and plan curvature were both positively correlated with grain yield. The steady-state topographic wetness index (Barling et al., 1994) exhibited the overall strongest correlations with SOC and P contents and yield in accordance with other studies (Moore et al., 1993). This index accounts for flow convergence and divergence and was previously used to indicate "wet" and "dry" areas within landscapes. Hence, redistribution of water may also be linked to crop productivity at the Sæby site (Auerswald et al., 1997). The wetness index further suggests that organic matter formation was affected by the soil moisture regime at Sæby as a result of higher plant production and perhaps reduced organic matter decomposition in the wetter, and periodically stagnic, toeslope areas of the field (Gregorich et al., 1998).


View this table:
[in this window]
[in a new window]
  		Table 5. Pearson correlations between soil variables, yield, tillage translocation, and terrain attributes at Sæby.

 

   CONCLUSIONS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSIONS
REFERENCES
 
Substantial soil redistribution had occurred on the undulating field at Sæby during the past decades. Our results strongly suggest that tillage erosion was by far the dominating process of soil redistribution. With its distinct spatial pattern of erosion at convexities and deposition at concavities, tillage erosion led to truncated soil profiles on the shoulderslopes and was a major control on the distributions of colluvial soil in the field. Therefore, tillage erosion was a major contributor to within-field variability of soil properties with important implications for the field's carbon budget and nutrient losses. The burial of SOC below plow depth at the footslopes will preserve some SOC from decomposition and ought to be taken into account in assessments of carbon storage in agrolandscapes. With the progressive accumulation of nutrient-rich soil material in low-lying areas of fields exposed to concentrated overland flow and leaching, the risk of nutrient loss is prone to increase. Despite the occurrence of deep soils across the whole Sæby field, preliminary evidence suggests that crop productivity was affected by the spatial distribution of soil properties and in turn tillage erosion. However, the redistribution of water across the hillslope at Sæby could not be ruled out as a contributing factor. An integration of topographic models of soil redistribution and soil property evolution with crop growth models will be a way forward to isolate the soil–crop–erosion interactions. The effect of tillage erosion on soil quality and productivity will vary with the agroenvironment. It is therefore likely that shallower soils on hummocky terrain in drier climates will suffer more adverse effects than observed at Sæby. Additionally, the formation of deep, nutrient-rich soils in limited areas within fields presents a soil resource inefficiently used.


   ACKNOWLEDGMENTS
 
The authors gratefully acknowledge the funding of this work by the European Commission under Contract FAIR3-CT96-1478 as part of the TERON project. We also wish to thank Birthe Johannsen and the late Anker Giversen for their help with the soil sampling. We would in particular like to acknowledge the late Erik Sibbesen's role in initiating this work.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSIONS
 REFERENCES