• Tidak ada hasil yang ditemukan

Directory UMM :Data Elmu:jurnal:S:Soil & Tillage Research:Vol57.Issue1-2.Sept2000:

N/A
N/A
Protected

Academic year: 2017

Membagikan "Directory UMM :Data Elmu:jurnal:S:Soil & Tillage Research:Vol57.Issue1-2.Sept2000:"

Copied!
9
0
0

Teks penuh

(1)

Soil compactibility in relation to physical and organic

properties at 156 sites in UK

B.C. Ball

a,*

, D.J. Campbell

a,1

, E.A. Hunter

b

aEnvironment Division, Scottish Agricultural College, West Mains Road, Edinburgh EH9 3JG, UK

bBiomathematics and Statistics, Scotland, James Clerk Maxwell Building, The King's Buildings, Edinburgh EH9 3JZ, UK

Received 9 February 2000; received in revised form 19 June 2000; accepted 28 June 2000

Abstract

The variation of soil compactibility and its relationship to plasticity, texture, organic matter and particle density is considered for 156 sites and for one intensively sampled site. These sites were concentrated in east Scotland and were the locations of Scottish Agricultural College (SAC) ®eld experiments and surveys related to tillage and compaction. Compactibility was determined by a rammer method on sieved soil. The coef®cient of variability of compactibility (as maximum dry bulk density) was relatively low between sites (9.5%) and within site (4%). The soils covered a wide range of textures; sand contents ranged from 1.7 to 93.5 g 100 gÿ1

and clay contents ranged from 2.5 to 49.1 g 100 gÿ1

. However, liquid limit was more important than particle size fractions in the prediction of compactibility. Loss-on-pretreatment prior to measurement of particle soil distribution was taken as a measure of readily oxidisable soil organic matter. This fraction was more variable and more relevant than total organic matter in determining mechanical behaviour. Compactibility was predicted adequately by a combination of loss-on-pretreatment and liquid limit. Maximum dry bulk density and liquid limit were identi®ed as important characteristics of the dataset and would be suitable parameters for measurement of soil physical/ behavioural quality. Although particle density was not particularly important in predicting compactibility, it ranged from 2.36 to 2.87 Mg mÿ3

. Awareness of this variability is important for properties estimated by a calculation involving particle density.

#2000 Elsevier Science B.V. All rights reserved.

Keywords:Compaction; Organic matter; Plasticity; Survey

1. Introduction

Soil physical and mechanical properties were mea-sured in order to characterise soils used in laboratory and ®eld experiments on soil tillage and compaction in south-east Scotland. Soane et al. (1972) tested 58

of these agricultural soils and found that compact-ibility and plasticity were better related to intrinsic soil properties such as particle size distribution and clay mineralogy than to classi®cations of soil type, such as soil series, based on parent material and drainage.

Many researchers have attempted to predict proper-ties that are dif®cult or expensive to measure from simpler, cheaper or more readily obtained measure-ments such as particle size distribution or organic matter. Compactibility is time-consuming and expen-sive to measure because it requires measurement of *Corresponding author. Tel.:‡44-131-535-4392;

fax:‡44-131-667-2601.

E-mail address: b.ball@ed.sac.ac.uk (B.C. Ball).

1Present address: 11 Broomhill Road, Penicuik, Midlothian,

EH26 9EE, UK.

(2)

the response of soils to stress repeated at a range of water contents (Horn and Lebert, 1994). Compact-ibility is usually speci®ed as a maximum bulk density which can be used as a reference in describing the relative compaction state of ®eld soils (Carter, 1990). Compactibility is in¯uenced not only by the content of organic matter but also the type of organic material (Soane, 1990). Partly decomposed material, highly humi®ed material and material of fungal origin are particularly important in increasing resistance to com-paction. In our tests, a simple method of destroying organic material with hydrogen peroxide was used prior to measurement of particle size distribution and was used as an estimate of oxidisable organic matter (Day, 1965).

Compactibility has also been related to clay content (Gupta and Allmaras, 1987). Particle density also varies with clay content and soil mineralogy, particu-larly with the content of iron-containing minerals (Culley, 1993) and may in¯uence compactibility.

This paper considers a database of soil compact-ibility and other physical, mechanical and organic matter properties created from analysis of soils col-lected from 156 sites over a period of 30 years at the Scottish Agricultural College (SAC). Our objective is to identify readily measured soil properties which relate to compactibility. These include organic matter determined as loss-on-pretreatment, plasticity indices and particle density. We also show how compactibility indices relate to other soil variables and vary between sites, and within a single site.

2. Materials and methods

2.1. Soils

Soil samples were obtained from 156 sites, mainly of experiments or areas monitored for advisory pur-poses. The regional distribution of sites is given in Table 1. Most of the sites are in eastern Scotland near the location of SAC in Edinburgh. All samples came from within the cultivated layer and usually from a depth of about 100 mm; six to eight sub-samples from an area of about 0.3 ha typically being combined to form the sample tested. In addition, a detailed analysis of variation within one site was made in 1975±1976 by taking 48 samples from an experimental ®eld site at

South Road, Bush estate, near Edinburgh. Pidgeon (1980) described this experiment in detail. The sam-ples were taken from the top 36 cm of soil (sampled at 6 cm depth intervals) from eight replicate plots of four tillage treatments located on two soil types. Plots were 48 m12 m. Samples were combined over replicates prior to measurement.

2.2. Soil measurements

Soil properties were determined essentially by the methods described in British Standards Institution (1990) but with some minor modi®cations as described below. The dry bulk density/water content relationship was determined using the 2.5 kg `Proc-tor' rammer method (British Standards Institution, 1990). This provided both the maximum dry bulk density and the optimum water content for compac-tion (the water content at maximum density). We also include the total porosity at maximum density. Par-ticle size distribution was determined by pretreatment with 20 volume hydrogen peroxide followed by dry sieving and then sedimentation using the pipette method. The estimate of ``readily oxidisable organic matter'' is the loss-on-pretreatment expressed as a gravimetric fraction of the original mass of soil. Particle density was measured on the treated residue. The small pyknometer method was used with kero-sene as the displacement ¯uid (British Standards Institution, 1990). The cone penetrometer method was used for the liquid limit. Plastic limit was deter-mined by the standard method. For a limited number of soils, total organic matter was obtained by the acid± dichromate oxidation method described by Tinsley (1950), a factor of 1.724 being used to convert from organic carbon.

B Angus, Fife and central Scotland 37

C Lothians and Borders 93

D South-west Scotland 3

E North England and south-west Wales 5

(3)

2.3. Statistical analysis

Summary statistics (Table 2) were derived for the ``156 sites'' data set and histograms of each variable were produced. The histograms for compactibility and some of the more important and interesting related variables are given in Fig. 1. Experimental data from the ``South Road'' site were used to determine the within-site variability as measured by the coef®cient of variation from an analysis of variance (Table 2).

The method of Orchard and Woodbury (1972) as implemented by the GenstatTM procedure MULT-MISS was used to ®ll gaps in the data matrix (mainly for plastic limit and plasticity index, but also for liquid limit) prior to the computation of the correlation matrix, given in Table 3. This technique allows units with only partial data to be used.

The correlation matrix (Table 3) led naturally to a principal components analysis (PCA) Ð see, e.g. Krzanowski (1993). This technique is used to sum-marise the data whilst still preserving the major features. Instead of trying to interpret 10 variables which are correlated (Table 3), the data were reduced to a smaller number (two) of summary variables (scores) which were not correlated with each other.

The formulae for transforming the original variables to the summary variables are called the loadings. Using this technique the data were summarised by the two principal scores which separated the sites best and are used to plot them in Fig. 2. The scores are de®ned by the equations (loadings) given in Table 4. The correlations in columns 4 and 5 of Table 4 are between the scores and the original variables. The higher the correlation the more important the variable is for forming the score. Correlations of contrasting sign indicate contrasting effects of variables on the scores. Multiple correlation (column 6 of Table 4) integrates the correlations of the two scores.

Partial least squares (PLS), which was originally developed as a calibration method for chemical data (Hoskuldsson, 1988), was used to determine predic-tive equations for compactibility properties from other properties (Table 5). Principal components regression (PCR) summarises the independent vari-ates by deriving principal component scores for a smaller number of signi®cant components. Multiple linear regression (MLR) is then used to estimate a prediction equation. These equations are then con-verted into equations in the original independent variables using the matrix of PCA loadings. A

Table 2

Variation of soil properties for the 156 sites and for the South Road ®eld experiment (48 samples)

Variable 156 sites South Road experiment

Mean Minimum Maximum C.V. (%) No. of sites not tested

Mean C.V. (%)

Maximum dry bulk density (Mg mÿ3) 1.60 0.76 1.82 9.5 1.60 4.0

Water content at maximum density (g 100 gÿ1) 19.1 7.4 41.2 21.7 18.2 5.6

Total porosity at maximum density (cm3cmÿ3) 37.9 27.9 66.0 15.0 38.0 5.5

Liquid limit (g 100 gÿ1) 37.8 21.3 67.4 22.6 10 33.4 2.5

Plastic limit (g 100 gÿ1) 28.5 18.4 44.9 20.6 27 16.6 34.2

Plasticity index 9.4 0 28.5 59.1 27 3.64 99.0

Treated particle density (Mg mÿ3) 2.64 2.36 2.87 2.3 1 2.66 2.0

Loss-on-pretreatment (g 100 gÿ1) 4.22 0.10 30.10 65.1 1 3.28 16.7

Total organic matter (g 100 gÿ1) 4.25 0.92 6.71 35.8 121 5.18 10.5

Total sand (2±0.06 mm) (g 100 gÿ1) 48.6 1.7 93.5 31.4 48.4 2.8

Coarse sand (2±0.6 mm) (g 100 gÿ1) 6.2 0 16.9 58.3 4.7 8.4

Medium sand (0.6±0.2 mm) (g 100 gÿ1) 16.2 0.2 52.8 53.2 13.6 4.5

Fine sand (0.2±0.06 mm) (g 100 gÿ1) 26.3 0.7 65.3 36.5 30.0 3.1

Total silt (0.06±0.002 mm) (g 100 gÿ1) 35.2 1.6 84.4 33.4 35.0 2.2

Coarse silt (0.06±0.02 mm) (g 100 gÿ1) 18.9 1.6 77.6 43.2 18.4 6.2

Medium silt (0.02±0.006 mm) (g 100 gÿ1) 9.5 0.4 35.5 48.9 9.6 10.5

Fine silt (0.006±0.002 mm) (g 100 gÿ1) 6.85 0.8 15.6 39.9 7.1 13.5

(4)

measure of the importance of each independent vari-able can be found by correlating it with each score in turn and summing the values ofR(squared). In PLS, in contrast to PCR, the scores are determined in part by correlation with the response variable. The number of scores (components) used for prediction are

deter-mined using cross-validation and the PRESS statistic. The process is known to give equations with good predictive ability (Osten, 1988).

Linear regression was also used to derive an equa-tion of predicequa-tion of total organic matter from loss-on-pretreatment (Fig. 3).

Fig. 1. Frequency distribution of selected variables for the 156 survey sites.

Table 3

Correlation coef®cients for comparisons among 10 properties

Property 1 2 3 4 5 6 7 8 9 10

1 Maximum dry bulk density 1.000

2 Water content at maximum density ÿ0.775*** 1.000

3 Total porosity at maximum density ÿ0.973*** 0.711*** 1.000

4 Plastic limit ÿ0.536*** 0.603*** 0.484*** 1.000 5 Liquid limit ÿ0.655*** 0.693*** 0.64*** 0.791*** 1.000

6 Plasticity index ÿ0.41*** 0.389*** 0.449*** 0.065 0.662*** 1.000

7 Particle density 0.086 ÿ0.267** 0.094 ÿ0.112 0.024 0.177 1.000

8 Loss-on-pretreatment ÿ0.497*** 0.524*** 0.413*** 0.408*** 0.414*** 0.176 ÿ0.378*** 1.000

9 Coarse sand (2±0.6 mm) 0.111 ÿ0.014 ÿ0.146 0.213* 0.038 ÿ0.2* ÿ0.119 0.047 1.000

10 Clay (<0.002 mm) ÿ0.405*** 0.387*** 0.445*** 0.193 0.577*** 0.706*** 0.184 0.157 ÿ0.186 1.000

(5)

3. Results

The variation in soil properties for the 156 survey sites is summarised in Table 2. It shows the mean, minimum and maximum value for each property together with the coef®cient of variation for the mean. The overall mean and coef®cient of variation for the soil properties in the South Road site as determined within an analysis of variance are also given in Table 2.

Some values for total organic matter are missing because measurements of these properties began at a relatively late stage in the assembly of the database. Most properties, including particle density, covered a wide range. Coef®cients of variation ranged from low for particle density (2.3%) to very high for loss-on-pretreatment (65.1%) and plasticity index (59.1%). The high coef®cients of variation for plasticity index and plastic limit occurs despite the assigning of `miss-ing values' to non-plastic soils. Coef®cients of varia-tion within the South Road site covered a smaller range, typically less than 10% for directly measured properties.

Frequency distributions for compactibility as indi-cated by maximum dry bulk density and water content and total porosity at this maximum and properties useful in its prediction viz. loss-on-pretreatment, liquid limit and particle density are given in Fig. 1. Total organic matter is not shown as there were insuf®cient data. Compactibility and particle density show a near normal distribution whereas liquid limit and loss-on-pretreatment are skewed.

The characteristic properties shown in Fig. 1 along with the plastic limit, plasticity index, coarse sand and clay fractions formed a reduced set of variables for further analysis. The correlation coef®cient and its signi®cance for comparisons between each pair of these 10 properties are shown in Table 3. Correlation coef®cients and their signi®cances are generally lower than those reported by Soane et al. (1972), presumably because of the wider distribution of sites represented Fig. 2. PCA Scores (1) vs. Scores (2) for survey sites. The letters

are de®ned in Table 1.

Table 4

PCA loadings for scores, correlations between individual variables and scores and multiple correlations

Loadings Correlations Multiple

correlation Component 1

(accounts for 46.6% variation)

Component 2 (accounts for 18.0% variation)

Scores 1 Scores 2

Maximum dry bulk density ÿ0.413 ÿ0.048 ÿ0.892 ÿ0.064 0.894 Water content at maximum density 0.397 0.169 0.858 0.226 0.888 Total porosity at maximum density 0.402 ÿ0.062 0.867 ÿ0.083 0.871

Plastic limit 0.318 0.30 0.686 0.402 0.795

Liquid limit 0.414 ÿ0.037 0.894 ÿ0.049 0.896

Plasticity index 0.286 ÿ0.429 0.618 ÿ0.575 0.844

Particle density ÿ0.034 ÿ0.518 ÿ0.073 ÿ0.695 0.699

Loss-on-pretreatment 0.269 0.337 0.582 0.452 0.736

Coarse sand (2±0.6 mm) ÿ0.033 0.380 ÿ0.071 0.509 0.514

(6)

in Table 3. Soil compactibility, as indicated by max-imum dry bulk density and water content at this maximum, is essentially unrelated to particle size distribution apart from a weak relation with clay content. Compactibility properties, however, are cor-related with both the liquid and plastic limits which in turn are both correlated with loss-on-pretreatment. In

summary, compactibility (as indicated by both increased maximum dry bulk density and decreased water content at the maximum) decreased with loss-on-pretreatment, plasticity limits and clay content.

Results of PCA are given as the ®rst two nents and their loadings in Table 4. The ®rst compo-nent accounted for almost half of the variation Table 5

Equations of prediction for compactibility properties using (A) all other properties and (B) the two properties giving the best prediction. Each equation is the sum of the product of the coef®cients and the value of the variables plus a constant within a columna

Maximum dry bulk density

Water content at maximum density

Total porosity at maximum density

(A)All properties

% variation accounted for 58.4 58.1 47.0

No. of significant components 2 2 1

Constant 1.338 44.23 22.30

Coefficient

Liquid limit ÿ0.004 0.127 0.123

89.4 89.1 87.5

Plastic limit ÿ0.005 0.151 0.122

53.5 61.9 45.9

Particle density 0.244 ÿ13.8 1.80

28.5 56.6 1.0

Loss-on-pre-treatment ÿ0.014 0.350 0.223

67.9 68.8 17.7

Coarse sand (0.6±2 mm)b 0.608 ÿ10.14 ÿ9.44

9.2 0.7 1.4

Medium sand (0.6±0.2 mm)b 0.190 ÿ3.02 ÿ10.6

53.7 55.0 51.0

Fine sand (0.2±0.06 mm)b ÿ0.062 ÿ0.691 ÿ5.12

75.6 68.9 39.7

Coarse silt (0.06±0.02 mm)b 0.126 ÿ1.514 ÿ0.694

14.7 9.7 0.8

Medium silt (0.02±0.006 mm)b ÿ0.009 ÿ1.271 12.76

62.0 63.4 44.1

Fine silt (0.006±0.002 mm)b ÿ0.114 14.89 21.31

65.1 63.0 53.9

Clay (<0.002 mm)b ÿ0.409 7.00 12.60

43.0 38.7 49.2

(B)Two properties giving the best prediction

% variation accounted for 48.8 55.2 42.8

No. of significant components 1 1 1

Constant 1.998 7.371 23.33

Coefficient

Liquid limit ÿ0.009 ÿ0.251 0.318

79.2 79.9 82.7

Loss-on-pretreatment ÿ0.0194 0.552 0.621

62.5 61.6 58.1

aThe percentage of variation of each ``explanatory'' variable which is related to the ``signi®cant PLS scores'' appears in italics below the

value of the coef®cient.

(7)

between samples (46.6%) and together the ®rst two components account for almost two-thirds (64.6%) of the variation. The remaining variation is spread over eight further components. The loadings given in Table 4 show how the summary scores are constructed from the original variables and the correlations allow the scores to be interpreted. Score 1 is strongly positively related to water content and total porosity at maximum density and also to liquid limit. Score 2 is related positively to coarse sand and negatively to particle density, plasticity index and clay content. The multiple correlation shows that maximum dry bulk density and liquid limit were the most effective whilst coarse sand was the least effective in describing the major features of the data.

A plot of PCA scores (1) against scores (2) is shown in Fig. 2. This pairing of scores showed the maximum separation of the data, splitting the data set according to location. The data appear to lie in four groups. Most samples lie in a group (Fig. 2) corresponding to values of score (1) betweenÿ4 and 0.5. Forty three samples lie in a group where values of score (1) are between 0.5 and 4.5. This group contains 10 of the 12 sites from north-east Scotland (symbol A). The soils in this group are mostly loams and silt loams of lower than average compactibility (i.e. maximum dry bulk densities below average, and water contents at maximum bulk density above average) and with higher than average liquid limits. The next two groups, each containing

three sites, correspond to score (1) greater than 4.5. The groups are split according to whether score (2) is positive or negative. In the group where score (2) is positive, particle density is very low but in the group where it is negative, particle density is very high.

The equations produced for prediction of compact-ibility properties by PLS regression on intrinsic soil properties are given in Table 5. Use of all the chosen variables accounted for 58.4% of the variability of maximum dry bulk density, 58.1% of the variability of water content at maximum density and 47% of the variability of total porosity at maximum density. Information on the relationship of the compactibility variables to the ``explanatory'' variables is given from the percentage variation (given in italics below each coef®cient in Table 5) of each coef®cient accounted for by the signi®cant scores. From these, liquid limit appears to be the most important variable and coarse sand and coarse silt the least important. Progressive removal of variables beyond ®ve had little effect on the percentage variation accounted for. Prediction from liquid limit and loss-on-treatment, the two `best predictors', accounted for 48.8% of the variability of maximum dry bulk density, 55.2% of the variability of water content at maximum density and 42.8% of the variability of total porosity at maximum density.

4. Discussion

Dry bulk densities measured in recent SAC ®eld experimentation are about 82±95% of the maxima reported here, see, e.g. (Ball et al., 1994; Ball and Ritchie, 1999). Although compactibility was mea-sured on remoulded soil with the in¯uence of macro aggregation excluded, the test is severe enough for aggregation effects to be relatively minor. Thus, the maximum bulk density provides a useful reference level for assessing relative compaction status from ®eld-measured bulk density.

Loss-on-pretreatment and plasticity indices (parti-cularly liquid limit) were shown to be important characteristic variables useful in predicting soil com-pactibility. Indeed, when the dataset was reduced to include only sites where both organic matter and loss-on-pretreatment were measured, correlations between compactibility properties and loss-on-pretreatment exceeded those between compactibility properties Fig. 3. Organic matter by wet oxidation vs. loss-on-pretreatment

for samples for the 35 survey sites where both were measured. The regression line is organic matterˆ0:74loss-on-pretreatment‡

(8)

and organic matter. Loss-on-pretreatment attacks the colloidal, humi®ed organic matter, but not the ®brous residues (Day, 1965). The humi®ed portion of organic matter was considered by Soane (1990) to have a complex role in stabilising natural structure; in general the greater the content of highly humi®ed material the greater the stability and strength of aggregates and hence the lower the compactibility. However, the effectiveness of hydrogen peroxide for pretreatment may be reduced in the presence of manganese dioxide (Day, 1965). Also the oxidation of the organic matter by hydrogen peroxide is further restricted in soils containing iron which tends to stabilise organic mate-rials during their oxidative degradation (Oades and Townsend, 1963). Nevertheless, a highly signi®cant relationship was found between loss-on-pretreatment and organic matter for those sites where both were measured (Fig. 3).

Liquid limit related better to compactibility than plastic limit (Table 5). Liquid limit has been shown to relate to soil surface area better than plastic limit (Hammel et al., 1983). Thus, soil compactibility may be related to surface area. Liquid limit may integrate the effects of organic matter, and clay con-tent and type to provide a relevant estimate of surface area. This may relate to how soil resists compaction which is a combination of the total number of contact points between single particles and the shear resis-tance per contact point (Horn and Lebert, 1994).

Soil particle density is often assumed to be a constant at 2.6 or 2.65 Mg mÿ3

, however, for the soils considered here, particle density ranged from 2.36 to 2.87 Mg mÿ3

with six soils having values <2.5 Mg mÿ3

and 10 with values >2.8 Mg mÿ3

. We found a wide range (2.49±2.83 Mg mÿ3

) within one soil series (Macmerry) with the low particle densities mostly on sloping sites and high particle densities on level sites at the bottom of slopes. Such variation may arise from variation in the types of heavy minerals in soils derived from glacial till (Ragg and Futty, 1967), or displacement by erosion of heavy minerals asso-ciating with iron oxides and silicate clays (Padma-nabhan and Mermut, 1995) or even to tillage effects on the halloysitic (kaolinite) clay mineralogy (Heredia et al., 1996). Such variation has implications for the use of particle density in the calculation of porosities. Although particle size fractions were important in determining soil behaviour and physical properties,

there was little variation between fractions suggesting that a single, easily-determined fraction would be suf®cient to characterise particle size.

The sites in this database were the locations of ®eld experiments and surveys usually related to tillage or compaction. Thus, the sites are concentrated in arable cropping areas and do not represent an even spread of soil types. This may help explain why maximum dry bulk density and liquid limit are most effective in distinguishing sites.

5. Conclusions

Compactibility and liquid limit were important characteristics distinguishing soil types. Coef®cients of variation of compactibility were relatively low, particularly within a single site, where they were <5%. Compactibility was predicted well by a combi-nation of liquid limit and loss-on-pretreatment (for particle size analysis).

Acknowledgements

The work was funded by the Scottish Executive Rural Affairs Department. The assistance of many colleagues in the Land Management Department, SAC in sampling and testing the soils is gratefully acknowl-edged. Copies of the complete dataset are available from the ®rst author in electronic form.

References

Ball, B.C., Ritchie, R.M., 1999. Soil and residue management effects on arable cropping conditions and nitrous oxide ¯uxes under controlled traf®c in Scotland. 1. Soil and crop responses. Soil Till. Res. 52, 117±189.

Ball, B.C., Lang, R.W., Robertson, E.A.G., Franklin, M.F., 1994. Crop performance and soil conditions on imperfectly drained loams after 20±25 years of conventional tillage or direct drilling. Soil Till. Res. 31, 97±118.

British Standards Institution, 1990. British Standard Methods of test for soils for civil engineering purposes, BS 1377. British Standards Institution, London.

Carter, M.R., 1990. Relative measures of soil bulk density to characterise compaction in tillage systems on ®ne sandy loams. Can. J. Soil Sci. 70, 425±433.

(9)

Society of Soil Science, Lewis Publishers, CRC Press, Boca Raton, FL, pp. 529±539.

Day, P.R., 1965. Particle fractionation and particle size analysis. In: Black, C.A. (Ed.), Methods of Soil Analysis, Part 1, Physical and Mineralogical Properties, including Statistics of Measure-ment and Sampling.Agronomy Monograph No. 9. American Society of Agronomy, Madison, WI, pp. 545±577.

Gupta, S.C., Allmaras, R.R., 1987. Models to assess the susceptibility of soils to excessive compaction. Adv. Soil Sci. 6, 65±100.

Hammel, J.E., Sumner, M.E., Burema, J., 1983. Atterberg limits as indices of external surface areas of soils. Soil Sci. Soc. Am. J. 47, 1054±1056.

Heredia, Y., Kass, D., Heredia-Volquez, Y., 1996. Cambios en las propiedades ®sicas del suelo despues de seis anos de cultivos en callejones con dos sistemas de labranza. Agrofor. en las Americas 3, 11±12, 16±19.

Horn, R., Lebert, M., 1994. Soil compactibility and compressi-bility. In: Soane, B.D., van Ouwerkerk, C. (Eds.), Soil Compaction in Crop Production. Elsevier, Amsterdam, pp. 45± 69.

Hoskuldsson, A., 1988. PLS regression methods. J. Chemom. 2, 211±228.

Krzanowski, W.J., 1993. Principles of Multivariate Analysis: A User's Perspective. Oxford University Press, Oxford.

Oades, J.M., Townsend, W.N., 1963. The in¯uence of iron on the stability of soil organic matter during peroxidation. J. Soil Sci. 14, 134±143.

Orchard, T., Woodbury, M.A., 1972. A missing information principle: theory and applications. In: Proceedings of the Sixth Berkeley Symposium of Mathematical Statistics and Prob-ability, Vol. 1, Berkeley, CA, pp. 697±713.

Osten, D.W., 1988. Selection of optimal regression models via cross-validation. J. Chemom. 2, 39±48.

Padmanabhan, E., Mermut, A.E., 1995. The problem of expressing the speci®c surface areas of clay fractions. Clays Clay Mins. 43, 237±245.

Pidgeon, J.D., 1980. A comparison of the suitability of two soils for direct drilling of spring barley. J. Soil Sci. 31, 581±594. Ragg, J.M., Futty, D.W., 1967. The Soils of the Country Round

Haddington and Eyemouth. Her Majesty's Stationery Of®ce, Edinburgh.

Soane, B.D., 1990. The role of organic matter in soil compactibility: a review of some practical aspects. Soil Till. Res. 16, 179±202. Soane, B.D., Campbell, D.J., Herkes, S.M., 1972. The character-ization of some Scottish arable topsoils by agricultural and engineering methods. J. Soil Sci. 23, 93±104.

Referensi

Dokumen terkait

Pengumuman juga diumumkan pada aplikasi SPSE kota Banjarmasin. Pokja XV ULP

Panitia Pengadaan Barang/Jasa akan melaksanakan Pemilihan Langsung Ulang dengan Pascakualifikasi untuk paket Pekerjaan Pengadaan kontruksi pada Dinas Pertanian dan

konsepsi mengenai nihongo dan kokugo, nihongogaku dan kokugogaku, penutur pahasa Jepang,1. dan karakteristik bahasa Jepang; (2) fonologi bahasa Jepang; (3) sistem ortografis

IMPLEMENTASI NILAI MORAL TATA TERTIB SEKOLAH SEBAGAI BENTUK KEKERASAN SIMBOLIK DALAM.. MENCEGAH

Diakses pada tanggal 19 Februari 2015 dari

PENGARUH PERSEPSI KEMUDAHAN (PERCEIVED EASE OF USE) DAN PERSEPSI KEGUNAAN (PERCEIVED USEFULNESS) TERHADAP PENGGUNAAN AKTUAL (ACTUAL USAGE) E -..

Kurikulum 2013 KOMPETENSI DASAR Sekolah Dasar (SD)/Madrasah Ibtidaiyah (MI).Kementerian Pendidikan dan Kebudayaan.. Argumentasi

Peraturan Pemerintah Nomor 27 Tahun 1994 tentang Pengelolaan Perkembangan Kependudukan sebagai pelaksanaan dari Undang-Undang Nomor 10 Tahun 1992 tentang Perkembangan Kependudukan