SW—Soil and Water

J. agric. Engng Res. (2001) 79 (2), 231}237 doi:10.1006/jaer.2000.0693, available online at http://www.idealibrary.com on SW*Soil and Water

Crop Residue Management E!ects on Soil Mechanical Impedance M. Maiorana; A. Castrignano`; F. Fornaro Istituto Sperimentale Agronomico, Via C. Ulpiani 5, 70125 Bari, Italy; e-mail of corresponding author: a. castrignano`: [email protected] (Received 29 May 2000; accepted in revised form 15 December 2000; published online 5 April 2001)

Soil penetration resistance was evaluated in di!erent crop residue management practices. The experimental design was a randomized complete block with "ve replications (block). The soil strength data, measured to a depth of 52)5 cm using a recording penetrometer, were corrected to a common water content and then submitted to a multivariate approach consisting of a combination of principal component analysis (PCA) and multivariate variance analysis, considering crop season and sampling date within season as repeated factors. The "rst two principal components (PC) explained 95% of the variance. The "rst one, related to the &25}52 cm deep layer, accounted for 84% of total variance, while the second PC, which seems to measure the penetration resistance at 3}25 cm depth, explained more than 11% of total variance. The results of variance analysis proved that crop residue management had signi"cant e!ects on soil penetration resistance only for the "rst PC. Crop season, date of measurement within each season, the two-way interactions: season with date, season with management, date with management and the three-way interactions: season with date and block, and season with date and management were highly signi"cant in both PCs. This study has also shown the location e!ect of penetrometer measurements is fundamental when soil characteristics are altered by crop residue management.  2001 Silsoe Research Institute

1. Introduction In southern Italy, the scanty use of animal manure and the cultivation of winter cereals in continuous cropping are causing, as years go by, a decrease in soil organic matter and main nutrient components, not always counterbalanced by the application of increasing levels of mineral fertilizers. Soil management procedures, that address soil chemical and organic fertility conservation on sustainable crop yield need to be developed. Among these, ploughing in of crop residue can compensate for the loss of soil organic matter content (Convertini et al., 1998) and slow down the progressive deterioration of some chemical and physical soil properties. Nevertheless, stubble and straw burning is still widely practised in cropping systems in Mediterranean areas and is often utilized as a means of reducing crop residue loads on soil surface. In the light of these considerations, the Agronomic Research Institute has been carrying out a long-term study, started in 1978 and still in progress, on di!erent straw and stubble management practices in 0021-8634/01/060231#07 $35.00/0

a continuous cropping of durum wheat. The results concerning the quanti}qualitative aspects of production and some chemical characteristics of soil have been reported (Convertini et al., 1985; Di Bari et al., 1987; Maiorana et al., 1992). This paper reports the e!ects over time of those treatments on one of the most important physical soil parameters, the soil strength. Its measurement is important in agriculture as it characterizes mechanical impedance to root penetration. Nevertheless, cone penetration resistance measurements have not yet been developed to such a point that a routine measurement is unambiguous enough for precise recommendations for modi"cation of soil. An important aspect of why it is di$cult to interpret resistance data unambiguously is that soil strength is a dynamic characteristic, depending on many soil physical and chemical properties, water content, position, depth and soil management procedures. Therefore, we developed a statistical analysis able to show statistically signi"cant di!erences in mechanical impedance for various management practices.

231

 2001 Silsoe Research Institute

232

M . MA IO RA N A E ¹ A ¸.

2. Materials and methods

Table 1 List of the crop residue management treatments

2.1. ¹he site and the climate The experiment was carried out at Foggia (41327 latitude N, 15336 longitude E, 90 m above sea level), in a typical coastal area of southern Italy, the Apulian Tavoliere, in the experimental farm of the Institute. The main crop in the area is durum wheat in continuous cropping or in rotation with industrial crops (sugar beet, sun#ower, tomato), pulses and vegetables. The soil is a silty}clay Vertisol of alluvial origin, classi"ed as "ne, mesic, Typic Chromoxerert by the USDA Soil Taxonomy, with a satisfactory content of total nitrogen (0)122%), available phosphorus (measured as P O ,   41 p.p.m.) and exchangeable potassium (as K O,  1561 p.p.m.) and a good supply of organic matter (2)07%). In summer the soil often shows several cracks, both at the surface (4}5 cm wide) and throughout the 50 cm deep layer (1}2 cm wide). The climate is classi"ed as &accentuated thermo-mediterranean', according to the FAO-UNESCO classi"cation, with summer temperatures which can rise above 403C, winter temperatures which can fall below 03C and rains concentrated mainly in the winter months. The weather during the period of soil strength measurements (1994}1996) was characterized by great variability. Referring to the interannual cycle August}April (ranging from the main soil ploughing to the end of penetrometer use), in the 3 trial years the rainfall was lower (416)0, 319)8 and 447)4 mm, respectively) than the long-term average 1952}1992 (461)5 mm).

2.2. Experimental design and treatments The experimental "eld design is a randomized complete block with "ve replications (block) and sub plots of 80 m each. The straw and stubble treatments, namely, burning of crop residues (B) and ploughing in of crop residues (P), the latter with or without the application on the residues of three nitrogen levels (as urea) (N , N and N ) and of    500 m ha\ of water (W), are shown in Table 1. The whole trial "eld received 100 kg P O ha\ in   summer, at the time of main soil ploughing and 100 kg N ha\ (NH NO ) as a top dressing on the wheat,   half in February (at the 5th}6th leaf stage) and half in April (at the start of booting stage). After harvesting of wheat, straw and stubble were chopped to 10}15 cm lengths and spread back on the plot. The various doses of urea were then applied, and everything were buried with a ploughing depth of 40 cm. In the treatments requiring water, it was applied before

Crop residue treatments Burning Ploughing in Ploughing in Ploughing in

Nitrogen on residue, kg ha\1

Water on residue, m3 ha\1

* * 50 100 150 50 100 150

* * * * * 500 500 500

Treatments identixcation B P PN  PN  PN  PN W  PN W  PN W 

ploughing. Finally, the residues of the B treatment were burnt. The sowing of wheat was carried out in the last ten days of November, with a rate of 450 seeds per m and a 15 cm row spacing.

2.3. Resistance to penetration The measurements of soil penetration resistance (cone penetration resistance) were taken on two of the "ve replications four times per year, in January}April 1994}1996, the most important months of the durum wheat cropping cycle, that starts with the sowing in November and ends with the harvesting in June. In May, because of the soil dryness, most readings fell at the bottom of measurement scale (exceeding 50 MPa), then resulting not very accurate. Five random measurements were made in each experimental plot between the wheat rows. The soil strength was measured using a Bush recording soil penetrometer (Findlay Irvine) with a 303 angle and 12)83 mm diameter cone, corresponding to the American Society of Agricultural Engineering standard. The penetrometer resistance measurements were done at 3)5 cm depth increments up to the total depth of 52)5 cm (labelled from R to R ) and recorded on a data   storage unit. The "rst depth of measurement (R , 0}3)5 cm depth) was excluded from the analyses, as it  usually showed values near 0 MPa, owing to an incomplete contact of the penetrometer base plate on the uneven soil surface.

2.4. Correction of resistance to penetration values for soil water content Since the penetrometer resistance is mainly a!ected by soil moisture (Perumpral, 1987; Vyn & Raimbault, 1993;

233

S O IL M E CH A N IC AL I MP ED AN C E

Table 2 Parameter estimate, its con5dence internal and residual error of the relationship among penetration resistance, soil water content and bulk density at the two soil depths Parameter

Estimate

Conxdence interval Lower

Depth 0}25 cm A B C

0)14949 !1)94560 0)47469

Residual error Depth 25}50 cm A B C Residual error

Upper

0)09035 !2)23013 !0)01516

0)20864 !1)66106 0)96454

0)15675 !2.01433 0)01101

0)37879 !1)45180 0)75803

0)025012 0)26777 !1)73307 0)38452 0)019128

Busschler et al., 1997), gravimetric soil water contents h, from regular depth intervals (0}20, 21}40 and 41}60 cm) and for all treatments and recording dates, were also measured. Correcting strengths for signi"cant di!erences in water contents among the various treatments permits examination of soil strength aspects other than those caused by water (Busscher, 1987). Perumpral (1987) examined various soil}water correction methods; Campbell and O'Sullivan (1991) showed that cone penetration resistance could be predicted as a function of moisture content and bulk density by R "Ah (o )! N @

(1)

where R is cone penetration resistance in MPa, o N @ is bulk density in g cm\, h is gravimetric water content in g g\ and A, B and C are positive coe$cients depending on soil properties. To correct cone penetration resistance measurements for di!erences in water content and then make them comparable, the model (1) takes the form R /R "(h /h ) A S A S

(2)

where the subscripts c and u are for corrected and uncorrected values. The empirical coe$cients, A, B and C were determined "tting the Eqn (1) to the experimental data collected on the same site and at two depths (0}25 and 25}50 cm) in a previous experiment. Two di!erent equations were estimated by least-squares method separately at the two above depths, as reported in Table 2. All penetrometer values were standardized to a common water content equal to the maximum soil water

content recorded over the whole three-season period, to obtain values varying within the interval [0, 1].

2.5. Statistical analysis As measurements made at di!erent depths in each penetration are not independent, principal component analysis * PCA * (Rao, 1964) was applied to penetration data, using correlation matrix (Stelluti et al., 1998). That allowed the derivation of a smaller number of linear combinations (principal components, PC) of the original variables (R }R ), which retained most of the informa  tion. To make easier the interpretation of PCs, a Varimax rotation was applied to PCs. The penetration data distributions were tested to be very skewed; in this case the arithmetic mean is an ine$cient estimator, because the estimation error is very large (Webster & Oliver, 1990). As the distributions were approximately lognormal, penetration data were transformed to neperian logarithms and all means and variances were computed in the logarithms. Finally, to obtain unbiased estimates of the means, the logarithms of penetration data (y) were transformed back using the standard formula (Aitchison & Brown, 1957): R "exp(1y2#1/2p) N W

(3)

where 1y2 is the mean of the transformed values of R and N p is the estimated variance of the logarithms. Since the W values of penetration resistance represent repeated measurements over time on the same experimental unit, each PC was submitted to multivariate variance analysis (Winer, 1971; Castrignano`, 1990). This technique allowed

234

M . MA IO RA N A E ¹ A ¸.

testing hypotheses about the factors of the statistical model. Di!erent types of sources of variation were considered: (1) measurement factor, also called within-subject factor; in this case the crop season (SEASON) and the measurement date within the season (DATE); (2) the main factor, also called between-subject factor; in this case the crop residue management (MANAGEMENT); (3) all the two- and three-way interactions among the two above within-subject factors and the betweensubject factor. To study temporal trend of penetration resistance and the likely factors a!ecting it, orthogonal polynomial contrasts were generated for each within-subject factor and a variance analysis was produced for each contrast. Finally, a variety of single degree-of-freedom contrasts were performed to test some of the hypotheses deemed the most interesting: (1) burning against ploughing in without any fertilization and irrigation; (2) ploughing in plus nitrogen against ploughing in plus nitrogen and irrigation; (3) PN against PN W; (4) PN against    PN W; (5) PN against PN W and (6) burning against    all the other treatments using ploughing in, pooled together. All the statistical procedures used were applied using of the statistical software package of SAS/STAT (SAS Institute, 1998).

3. Results and discussion The results of the PCA indicated that two of the PCs provided a good summary of the data. The "rst PC accounted for 84% of the total variance and the second PC more than 11%, so both totalling more than 95%. Therefore, only the "rst two PCs are retained in the analysis (Table 3). The "rst principal component (PC ) is  a measure of the overall penetration resistance representative of the layer &25}52 cm, having high positive loads on all variables R }R . On the contrary, the second   principal component (PC ) is more related to the  penetration resistance measurement in the topsoil. Principal component analysis has then disclosed the existence of a discontinuity along the soil pro"le, approximately at 25 cm depth, which is thought to be caused by annual tillage at the same depth. From the multivariate variance analysis, it resulted the "rst principal component was signi"cantly a!ected by crop residue management at the probability level of P(0)0001. As regards the within-subject factors, SEASON and DATE, were signi"cant at P(0)0001, whereas their two-way

Table 3 Rotated principal component pattern Variables

R  R  R  R  R  R  R  R  R  R  R  R  R  R 

Principal component value PC1

PC2*

0)92R 0)92R 0)92R 0)90R 0)89R 0)87R 0)84R 0)81R 0)32 0)20 0)47 0)58 0)25 0)64

0)30 0)31 0)30 0)31 0)38 0)42 0)45 0)49 0)86R 0)84R 0)79R 0)72R 0)70R 0)67R

Note: R }R , penetrometer readings.   *PC , PC , "rst and second principal component.   RSigni"cant values at P(0)05.

interactions with the between-subject factor (MANAGEwere signi"cant at P(0)05 and P(0)0001, respectively. Moreover, the two-way interaction between the two within-subject factors (SEASON with DATE) as well as the three-way interactions SEASON WITH DATE AND BLOCK and SEASON WITH DATE AND MANAGEMENT were signi"cant at P(0)0001. The signi"cance of the former three-way interaction points out the in#uence of surface location (BLOCK) on soil strength, revealing the probable existence also of a horizontal gradient in soil structure. The latter signi"cant three-way interaction stresses the importance of the speci"c meteorological pattern within a crop season on the di!erentiation of the crop residue management treatments. To disclose any temporal trend in soil strength more clearly, polynomial contrasts were also performed for both within-subject factors. As regards the factor SEASON, only the quadratic component was signi"cant at P(0)0001 and was signi"cantly in#uenced by crop residue management at P(0)01. On the contrary, as regards the factor DATE, both linear and quadratic components were highly signi"cant at P(0)0001, but only the linear component di!ered signi"cantly among the various crop residue management treatments at P(0)0001. These results then revealed a general increase of soil resistance over a crop season (means not reported for brevity), owing to the decrease of soil water content; this trend was di!erent among the management treatments, but it was mainly a!ected by the meteorological #uctuations characterizing each crop season. Since the above statistical analysis revealed that the meteorological pattern of each season caused the main MENT)

235

S O IL M E CH A N IC AL I MP ED AN C E

Table 4 Results of linear contrast between treatments for the 5rst principal component in terms of probability level Contrasts for component 1

B versus P PN versus PNW PN versus PN W   PN versus PN W   PN versus PN W   B versus P 2-2

Probability level s.d. 1

1994 s.d. 2 s.d. 3

s.d. 4

s.d. 1

1995 s.d. 2 s.d. 3

s.d. 4

s.d. 1

1996 s.d. 2 s.d. 3

s.d. 4

0)0216 0)0061 0)1643 0)0419 0)1543 0)0366

0)0215 0)7828 0)2689 0)0245 0)0854 0)0016

0)2641 0)0001 0)0005 0)4206 0)0099 0)9112

0)0045 0)0487 0)5352 0)0077 0)2265 0)0043

0)1737 0)0165 0)0001 0)2714 0)4184 0)0314

0)7779 0)0288 0)8857 0)3595 0)0082 0)5249

0)1404 0)0014 0)0128 0)2950 0)0278 0)0002

0)5461 0)0019 0)0069 0)0188 0)6305 0)5551

0)1904 0)0018 0)0996 0)1634 0)0119 0)6163

0)3053 0.0278 0)4996 0)0217 0)4102 0)0689

0)1823 0)0104 0)0888 0)0024 0)7542 0)0025

0)1697 0)0835 0)0612 0)8860 0)1825 0)3179

Note: s.d., sampling date: 1 in January, 2 in February, 3 in March, 4 in April; B, burning; P, ploughing in without N and water; N, all nitrogen treatment; N , N and N , nitrogen levels; W, water application; P , all treatments using ploughing in.    2-2

variation in subsoil strength, a variety of univariate contrasts was performed at the measurement dates relative to each individual crop season (Table 4). The objective was to disclose some relevant di!erences among the treatments. From Table 4, there is no well-de"ned behaviour of a treatment in comparison with another treatment. In particular, burning di!ered from all treatments using crop residue incorporation in late season and also in February 1995. The application of water on residues in summer produced signi"cant e!ects on soil impedance, but not in all years and in di!erent periods of the crop season. The results of Student}Newman}Keuls multiple test among the means of the "rst principal component (not reported) did not show any clear trend among the treatments, therefore it is not possible to prefer one treatment against another (burning against ploughing in; fertilization against fertilization with irrigation of the straws) on the ground of the only subsoil impedance criterion. The interpretation of results becomes still more confused for the second principal component. In fact, the management factor did not result in signi"cant di!erences, whereas the factor SEASON was signi"cant at P(0)01 as well as the two-way interaction SEASON WITH MANAGEMENT at P(0)0001. In the same way the factor DATE was signi"cant at P(0)0001 as well as the two-way interactions DATE WITH BLOCK at P(0)0002 and DATE WITH MANAGEMENT at P(0)05. Also, the two-way interaction SEASON WITH DATE resulted in signi"cant di!erences at P(0)0001 as well as the three-way interactions SEASON WITH DATE AND BLOCK and SEASON WITH DATE AND MANAGEMENT at P(0)002 and P(0)001, respectively. All these statistically signi"cant interactions prove that any probable e!ect of management on soil impedance is greatly a!ected by spatial heterogeneity and meteorological pattern.

The results of the temporal trend analysis showed that for the factor SEASON only the linear component was signi"cant at P(0)005, whereas for the factor DATE, the linear, quadratic and cubic components were signi"cant at P(0)0001, 0)004 and 0)0001, respectively. The interpretation of the results of the univariate contrasts (Table 5) is still more di$cult than the one for the "rst PC, owing to the greater in#uence of spatial and temporal variability on the soil impedance at the surface. Even though the treatments di!erentiated, such di!erences appeared randomly distributed during the crop season and no clear trend could then be disclosed. To have a clearer idea of the behaviour of soil strength over time, the temporal pattern of the two PCs averaged over the crop residue treatments is reported in Fig. 1. The soil impedance was generally greater in depth than at the top, with the exception of the season 1995. However, it underwent many #uctuations over the three crop seasons, mostly a!ected by rainfall pattern. The mean values of soil strength in January and in February 1994 appeared very high, quite probably because the soil conditions were not well "tted for mouldboard ploughing. The exceptionally plentiful autumn rains of 1993, in fact, did not allow working the mellow soil, with the production, as a consequence, of large and dense clods after the main tillage. Even if the generally accepted criteria (Taylor & Gardner, 1963; Taylor & Brar, 1991) assume that cone penetration resistance values greater than 2 MPa frequently reduce crop yields, nevertheless the "nal grain yields were not a!ected by soil compaction observed in January and February 1994 (Maiorana, 1998). The muddle aspect of these results derives quite probably from the fact that this type of study was aimed at characterizing soil strength in "eld using the means of several measurements without speci"c regard to

236

M . MA IO RA N A E ¹ A ¸.

Table 5 Results of linear contrast between treatments for the second principal component in terms of probability level Contrasts for component 2

B versus P PN versus PNW PN versus PN W   PN versus PN W   PN versus PN W   B vs P 2-2

Probability level s.d. 1

1994 s.d. 2 s.d. 3

s.d. 4

s.d. 1

1995 s.d. 2 s.d. 3

s.d. 4

s.d. 1

1996 s.d. 2 s.d. 3

s.d. 4

0)0027 0)0007 0)1050 0)0278 0)0256 0)0124

0)6250 0)6948 0)9367 0)0538 0)2213 0)1922

0)1313 0)0269 0)2743 0)0205 0)6679 0)8563

0)1989 0)6355 0)4858 0)5138 0)6144 0)0738

0)2014 0)0223 0)2629 0)0139 0)7024 0)1126

0)1450 0)3277 0)8477 0)6490 0)0561 0)0009

0)8907 0)9344 0)8540 0)2061 0)2159 0)6498

0)0740 0)2413 0)0230 0)9152 0)7756 0)2673

0)2483 0)2236 0)4373 0)0486 0)5056 0)1646

0)8476 0)0935 0)7727 0)1068 0)3316 0)7210

0)1383 0)9523 0)0417 0)0003 0)0468 0)0300

0)5107 0)8651 0)0925 0)0279 0)7039 0)0067

Note: s.d., sampling date: 1 in January, 2 in February, 3 in March, 4 in April; B, burning; P, ploughing in without N and water; N, all nitrogen treatment; N , N and N , nitrogen levels; W, water application; P , all treatments using ploughing in.    2-2

Fig. 1. Temporal pattern of the mean values for the xrst and the second principal components over the three crop seasons: , component 1, , component 2

horizontal position. Such a procedure might be su$cient for measurements of cone penetration resistance in very homogenous soil conditions. This study has then shown that the inclusion of position is fundamental when soil properties may be altered by crop residue management as a function of location. Not recognizing this important position e!ect means that the penetration data were not used e$ciently, i.e. without drawing the full information from them. In the past the practise usually was to take several random samples, without speci"c regard to location, in an attempt to overcome the variability of measured values. This study has proved that sometimes splitting the "eld in large blocks cannot be su$cient to assess the e!ect of horizontal position on soil strength. A more e$cient way to record soil impedance would then require that each measure is to be georeferred: in this case, other techniques of spatial data processing, such as geostatistics, which take into account spatial correlation, should be preferred.

Even if this study has not showed a clear, direct e!ect of straw burning on soil impedance, however, we feel to promote alternative methods to that technique, largely spread in Italy, as it is the cheapest way of clearing the "eld after wheat harvesting and before the main soil tillage. But crop residue burning and repeated deep-plough tillage have led to a progressive deterioration of soil fertility, above all in south Italy. Therefore, we suggest the application of more conservative management systems, such as no or minimum tillage, as they did not cause signi"cant yield loss in previous trials at the same site (De Giorgio et al., 1994). Moreover, leaving crop residues on soil surface, it is possible to reduce the impact of rain drops and then preserve the continuity of water conducting pores (Pagliai et al., 2000), improving water availability to the crops.

4. Conclusions Statistically signi"cant di!erences in mechanical impedance, as measured by a cone penetrometer, were found as functions of crop residue management, crop season and date of measurement within each season. Nevertheless, trend analysis and multiple comparisons among the experimental treatments did not disclose the superiority of someone of the examined crop residue management practices and then it is di$cult to give clear, practical advice to farmers on the ground of only the improvement of soil physical conditions. In this case other considerations, more properly of environmental type, could induce one to prefer ploughing crop residues into the soil; "rst of all as it avoids the use of "re and minimizes the loss of organic matter. Moreover, crop residue is a major renewable resource, which has an important impact also on the global carbon cycle: crop

S O IL M E CH A N IC AL I MP ED AN C E

residue conservation, in fact, is a fundamental tool for enhancing carbon sequestration in soil. Among the possible reasons of the confounding e!ect in the data interpretation, there could be the few years considered and the intrinsic soil variability. In fact, soil conditions vary locally as well as by depth and crop residue management. This made it di$cult to unambiguously interpret cone penetration resistance data. Therefore, measurement and description of soil impedance in soil management studies, importing heterogeneity to soil, must account for localization in addition to depth and management e!ects. To allow a more complete and clear interpretation of the results of penetrometer measurements, future studies should be then designed so as to include location e!ects and data should be processed by geostatistical techniques.

References Aitchinson J; Brown J A C (1957). The Lognormal Distribution. Cambridge University Press, Cambridge Busscher W J (1987). E!ect of water content changes on strength for south-eastern coastal plain soils. Agronomy Abstracts, pp. 237, ASA, Madison, Wisconsin Busscher W J; Bauer P J; Camp C R; Sojka R E (1997). Correction of cone penetration resistance for soil water content di!erences in a coastal plain soil. Soil and Tillage Research, 43, 205}217 Campbell D J; O:Sullivan M F (1991). The cone penetrometer in relation to tra$cability, compaction and tillage. In: Soil Analysis, Physical Methods (Smith K A; Mullins C E, eds), pp 399}429, Marcel Dekker Inc. New York Castrignano` A (1990). Applicazione dell'analisi multivariata della varianza ad esperimenti con misure ripetute nel tempo [Multivariate analysis of variance applied to experiments with measurements repeated over time]. Annali Istituto Sperimentale Agronomico Bari, 21, 183}195 Convertini G; Ferri D; Perniola M (1985). Confronto tra bruciatura e interramento dei residui vegetali del frumento (¹riticum durum Desf.). II. Aspetti agrochimici [Comparison between burning and ploughing in of durum wheat. II. Agro-chemicals aspects]. Monogra"e Genetica Agraria, 7, 299}316 Convertini G; Ferri D; Maiorana M; Giglio L; La Cava P (1998). In#uenza dell'interramento dei residui colturali sulla sostanza organica e su alcune proprieta` biologiche del

237

terreno in una prova a lungo termine in ambiente mediterraneo [E!ects of crop residue ploughing in on organic matter and some biological soil characteristics of a long-term research in Mediterranean environment]. Bollettino della Societa` Italiana della Scienza del Suolo, 2, 169}181 De Giorgio D; Castrignano` A; Rizzo V (1994). A multivariate approach to assess the e!ects of di!erent tillage systems on biometric parameters and weeds in durum wheat continuous cropping. Proc. of 13th International Conference of International Soil Tillage Research Organization. Aalborg, Denmark, 24}29 July, pp. 983}989 Di Bari V; Maiorana M; Rizzo V (1987). L'interramento dei residui vegetali del frumento (¹riticum durum Desf.) con la somministrazione di dosi crescenti di azoto [Ploughing in of durum wheat crop residues with increasing nitrogen level]. Rivista di Agronomia, 21, 85}89 Maiorana M (1998). L'interramento dei residui colturali di frumento duro. Risultati di studi pluriennali [Ploughing in of durum wheat crop residue. Results of long-term researches]. L'Informatore Agrario, 18, 41}45 Maiorana M; Convertini G; Di Bari V; Rizzo V (1992). Yield and quality of durum wheat (¹riticum durum Desf.) under continuous cropping after nine years of straw incorporation. European Journal of Agronomy, 1, 11}19 Pagliai M; Vignozzi N; Pellegrini S; Ceccon P; Giovanardi R; Coiutti C (2000). Impact of di!erent cropping systems on soil porosity and structure. Italian Journal of Agronomy, 1, 43}51 Perumpral J V (1987). Cone penetrometer application: a review. Transactions of the American Society of Agricultural Engineers, 4, 939}944 Rao C R (1964). The use and interpretation of Principal Component Analysis in applied research. Sankhya A, 26, 329}358 SAS Institute Incorporation (1998). User's Guide, Version 6.12, SAS/STAT, Cary Stelluti M; Maiorana M; De Giorgio D (1998). Multivariate approach to evaluate the penetrometer resistance in di!erent tillage systems. Soil and Tillage Research, 46, 145}151 Taylor H M; Brar G S (1991). E!ect of soil compaction on root development. Soil and Tillage Research, 19, 111}119 Taylor H M; Gardner H R (1963). Penetration of cotton seedling taproots as in#uenced by bulk density, moisture content and strength of soil. Soil Science, 96, 153}156 Vyn T J; Raimbault B A (1993). Long-term e!ect of "ve tillage systems on corn response and soil structure. Agronomy Journal, 5, 1074}1079 Webster R; Oliver M A (1990). Statistical Methods in Soil and Land Resource Survey. Oxford University Press, Oxford. Winer B J (1971). Statistical Principles in Experimental Design (2nd Edn.). McGraw-Hill Book Co., New York