Sensitivity of soil structure to changes in organic carbon content: Predictions using pedotransfer functions

Transcription

1 Sensitivity of soil structure to changes in organic carbon content: Predictions using pedotransfer functions B. D. Kay 1, A. P. da Silva 2, and J. A. Baldock 3 1 Department of Land Resource Science, University of Guelph, Guelph, Ontario, Canada N1G 2W1; 2 Dept. de Ciência do Sol, Universidade de São Paulo, Piracicaba, Brazil; 3 CSIRO, Division of Soils, Adelaide, SA, Australia. Received 22 November 1996, accepted 23 May Kay, B. D., da Silva, A. P. and Baldock, J. A Sensitivity of soil structure to changes in organic carbon content: Predictions using pedotransfer functions. Can. J. Soil Sci. 77: Pedotransfer functions (PTFs) were used to assess the sensitivity of the structural characteristics of coarse- and medium-textured calcareous illitic soils at different levels of relative compaction (RC) to changes in the organic carbon (OC) content. The analyses predicted that an increase in the OC content of 0.01 kg kg 1 would: increase the available water content from 0.02 to 0.04 m 3 m 3 with the largest increases occurring in coarser-textured soils and not being strongly influenced by RC; decrease the air-filled porosity at field capacity from 0.01 to 0.04 m 3 m 3 with the largest decreases occurring in the finer-textured soils and not being strongly influenced by RC; decrease the soil resistance to penetration with the decreases most pronounced at lower water potentials and higher RC; at the permanent wilting point and a RC of 0.95 the decrease would range from 1.2 to 3.8 MPa; increase the least limiting water range from 0.01 to 0.05 m 3 m 3 with the increase varying with clay content. A comparison with predictions based on PTFs derived from data sets from other parts of the world indicated caution should be exercised in applying PTFs to soil and climatic conditions that are different from those from which the PTFs were derived until the impacts of these conditions are better understood. Key words: Bulk density, field capacity, permanent wilting point, available water; aeration, soil resistance, least limiting water range Kay, B. D., da Silva, A. P. et Baldock, J. A Sensibilité de la structure du sol aux modifications de la teneur en carbone organique : prédiction au moyen de fonctions de pédotransfert. Can. J. Soil Sci. 77: Nous avons utilisé des fonctions de pédotransfert (FPT) pour évaluer la sensibilité des caractères structuraux de sols illitiques calcaires, de texture grossière et moyenne, à différents niveaux de compaction relative (CR), aux modifications de la teneur en C organique (Co). D après ces analyses, on peut prédire qu un accroissement en Co de 0,01 kg 1 : augmenterait de 0,02 à 0,04 m 3 m 3 la teneur en eau disponible du sol, les plus forts accroissements s observant dans les sols à texture plus grossière, peu influencées par CR; abaisserait de 0,01 à 0,04 m 3 m 3 la porosité en air à la capacité au champ, la diminution la plus importante survenant dans les sols de texture plus fine et fortement influencés par CR; réduirait la résistance du sol à la pénétration, les réductions les plus prononcées s observant en présence de potentiel hydrique bas et de forte CR au point de flétrissement permanent. Pour un CR de 0,95, la diminution se situerait entre 1,2 et 3,8 MPa; accroîtrait de 0,01 à 0,05 m 3 m 3 la zone de teneur en eau la moins limitante, l importance de l accroissement variant selon le contenu en argile du sol. Une comparaison faite avec les prédictions basées sur des FPT dérivées de données provenant d autres régions du monde invite à la prudence dans l application des FPT à des conditions de sol et de climat différentes de celles dans lesquelles elles ont été élaborées, au moins jusqu à ce qu on ait une meilleure compréhension des incidences de ces conditions. Mots clés: Densité apparente, capacité au champ, point de flétrissement permanent, eau disponible, aération, résistance du sol à la pénétration, zone de teneur en eau la moins limitante Agricultural practices which increase the sequestration of atmospheric carbon in soils may help to offset the inputs of CO 2 from a range of anthropogenic sources and are, therefore, desirable from an environmental perspective. Increasing the OC content of soils is also generally considered to have a positive impact on soil properties, particularly soil structure. Total porosity [or bulk density (BD)], pore size distribution and resistance to root penetration are structural characteristics that exert great influence on root proliferation and the supply of water and oxygen to the plant at any 655 given time. However, there have been few quantitative assessments of the impact of sequestration of carbon on the structural characteristics that are most important to plant growth. In particular, information is needed on: (a) the sensitivity of the different structural characteristics to changes in OC content, and, (b) the variation in this sensitivity with texture, other soil properties, and/or climatic conditions. The influence of OC soil structure is often assessed on specific sites with different carbon contents resulting from differences in management. Such studies provide valuable

2 656 CANADIAN JOURNAL OF SOIL SCIENCE information in relation to the specific soil and climatic conditions at the site but extrapolation to other conditions is often difficult. A complementary approach which makes it possible to quantitatively assess the sensitivity of the structural characteristics of different soils to changes in OC content involves the use of PTFs. The term pedotransfer function was used by Bouma (1989) to describe the quantitative relations between characteristics of soils or land quality and other soil characteristics that are more readily available (e.g. from resource inventories). Pedotransfer functions have been used extensively in estimating the hydraulic properties of unsaturated soils (c.f. van Genuchten et al. 1992). A recent preliminary analysis (Kay 1998) showed that PTFs could also be used to assess the sensitivity of the structural characteristics of different soils to changes in OC content. However, Kay noted that limitations in the use of PTFs for this purpose may arise if the PTFs are derived from data bases that include too many undefined variables. He suggested that the functions would be of greatest value if they were based on samples from the A horizon of soils that are derived from similar parent materials under the same climatic conditions. Variations in OC content would then be due to differences in management, drainage and position in the landscape. Implicit in the use of PTFs to assess the sensitivity of the structural characteristics of different soils to changes in OC content is the assumption that changes in the proportions, characteristics and impacts on the structural properties of different forms of carbon are either (a) well represented in the data set used to derive the PTFs or (b) that any changes vary linearly with OC content irrespective of the cause of change in OC. This study extends the preliminary analysis of Kay (1998) to provide a quantitative model of the sensitivity of the structural characteristics of soils to changes in OC content. PTFs that are derived for the A horizon of soils developed from similar parent materials under similar climatic conditions and under two tillage treatments (da Silva et al. 1997; da Silva and Kay 1997a) are used in the analyses. Structural characteristics having the greatest influence on plant growth, i.e. water supply, aeration and soil resistance to penetration were examined under different levels of compaction. Comparisons were made, where possible, with predictions based on PTFs derived from data sets from different parts of the world in order to determine the sensitivity of the relations to differences in climate and soil conditions. METHODS The data used by da Silva and Kay (1997a) and da Silva et al. (1997) to develop the PTFs were obtained from a site near Goderich in southwestern Ontario. A side-by-side comparison of zero and conventional tillage traversed a landscape made up of four soil series: Brady sandy loam (Gleyed Brunisolic Gray Brown Luvisol), Fox sandy loam (Brunisolic Gray Brown Luvisol), Perth loam (Gleyed Gray brown Luvisol) and Huron clay loam (Orthic Gray Brown Luvisol). Conventional tillage involved moldboard plowing in the autumn followed by secondary tillage in spring. The tillage comparison and a corn (Zea mays L.), soybean (Glycine max L.), wheat (Triticum aestivum L.) rotation had been maintained on the site for 11 yr prior to the initiation of these studies. The data used to derive the PTFs were collected while the site was planted to corn. The soils on the site were derived from calcareous parent materials, were neutral to slightly alkaline in ph, and their clay mineralogy was dominated by illite and poorly weathered micas. Their textures were broadly representative of coarse- and mediumtextured soils. Clay contents varied from 6 to 42% and OC from 0.8 to 6.7 % in samples taken from 5.0 to 7.5 cm depth (da Silva et al. 1997; da Silva and Kay 1997a). The climate at the site is characteristic of modified continental (Phillips 1990) with mean daily temperatures in July and January of 19.4 and 6.0 C, respectively; and mean annual precipitation of 943 mm (Anonymous 1982). The soil structural conditions at the site were characterized by measurements of BD, the maximum BD under compaction with a static pressure of 200 kpa (BDref), the water release curve, and the soil resistance to penetration-water content curve. Measurements were made at locations along a transect 0.5 km in length in each tillage treatment. A range in soil textures was selected with similar textures at a location for each tillage treatment. Intact cores taken from the cm depth were used to measure the water release curve, the soil resistance to penetration-water content curve and the BD [details are given in da Silva and Kay (1997a)]. Cores were taken three times during the growing season according to the stage of crop growth, i.e., six-leaf stage (row), silking (row and interrow) and harvest (interrow). Cores were taken from 36 pairs of locations along the transect in 1991 but cores from only 32 pairs were used for measurements of the water release curve and the resistance to penetration curve. Sampling for BD measurements was repeated twice in 1992 in the row and interrow area on the 36 locations and the average of the eight measurements (taken over 2 yr) calculated for each location, tillage and position combination (giving 144 average values). Disturbed samples were taken to a depth of 5 cm at each of the 36 pairs of locations in the interrow position for measurements of BDref [details regarding measurement of BD and BDref are given in da Silva et al. (1997)]. Values of penetration resistance varied from 0.09 to 8.64 MPa, BD varied from 0.96 to 1.71 Mg m 3, BDref varied from 1.11 to 1.74 Mg m 3 and the ratio BD/BDref, i.e. the RC, varied from 0.72 to The PTFs that were derived in these studies are summarized in Table 1. The sensitivity of the different structural characteristics to changes in OC content were assessed in one of two ways. The sensitivity of characteristics that were related to soil properties by a single equation (e.g. BD) could be defined by taking the derivative of the equation with respect to OC (e.g. BD/ OC). The sensitivity of characteristics requiring two or more equations to define their functional dependence on soil properties (e.g. soil resistance to penetration) are more difficult to differentiate. In these cases, five samples with different textures and carbon contents were selected from the data set to give a set of test soils providing a uniform variation in texture across the range in the data set, while having similar OC contents

3 KAY ET AL. PREDICTION OF CHANGES IN SOIL STRUCTURE 657 Table 1. Pedotransfer functions describing the different structural characteristics Structural characteristic Pedotransfer function z n R 2 P Reference Bulk density, BD, (Mg m 3 ) BD = T P o 144 y da Silva et al. (1997) T P o OC Cl OC Cl Reference bulk density, BD ref, (Mg m 3 ) BD ref = T OC da Silva et al. (1997) Cl Cl OC Relative compaction, RC RC = T P o da Silva et al. (1997) T P o Water Content, θ v (m 3 m 3 ), as a function θ v = a ψ b da Silva and Kay (1997a) of potential, ψ, (MPa) ln a = ln OC ln Cl ln BD b = ln OC ln Cl ln BD Soil resistance to penetration, SR, (MPa) SR = c θ d BD e da Silva and Kay (997a) ln c = OC Cl d = OC Cl e = Cl z OC is organic carbon content (%), Cl is clay content (%), T is tillage (a class variable assigned a value of 0 for zero tillage and 1 for conventional tillage) and P o is position relative to the row (a class variable assigned a value of 0 for in-row position and 1 for mid-row position). y Each value a mean of eight measurements. Table 2. Properties of test soils used with pedotransfer functions to predict the impact on structural characteristics of increasing the organic carbon content by 0.01 kg kg 1 Loamy Fine sandy Silty clay Silty clay Soil texture sand loam Loam loam loam Clay (%) Silt (%) Sand (%) Organic carbon (%) (Table 2). The value of a structural characteristic was calculated for each of the five test soils using the appropriate PTF (Table 1). The OC content of each test soil was increased by 1% (0.01 kg OC kg 1 soil), the value of the structural characteristic recalculated, and the magnitude of the change in the structural characteristic was determined. The resulting magnitude of change in the structural characteristic was used to approximate the derivative of the characteristic with respect to OC content. RESULTS AND DISCUSSION Bulk Density Bulk density was influenced by soil properties, tillage and position relative to the row (Table 1). Differentiating BD with respect to OC gave: BD/ OC = Cl (1) where Cl is the clay ( <2 mm) content (%). Differentiating a PTF for BD that was derived for 112 horizons of 39 Chernozemic profiles in Manitoba, Canada, in which the texture varied from sands to heavy clay (Shaykewich and Zwarich 1968) gave a relation of similar form, i.e., BD/ OC = Cl (2) Bulk density decreased with increasing OC content and the magnitude of the decrease diminished with increasing clay content. Similar trends were illustrated in the functional dependence of BD on texture and OC contents for 2721 soil horizons from the United States described by Rawls (1983). The magnitude of the changes in BD that were predicted by these three PTFs varied with clay content but, at a given clay content (e.g. 21%), were similar in magnitude (Table 3). Other researchers have either not found the decrease in BD with OC to vary with texture (e.g. Saini 1966; Hudson 1994) or have found that the decrease in BD is least in mediumtextured soils (e.g. Bauer and Black 1992). These apparent inconsistencies may be due to a narrower range of textures or smaller data sets being used by Hudson, Saini, and Bauer and Black (Table 3), but may also reflect fundamental differences in the behavior of different soils. Manrique and Jones (1991) found that the dependence of BD on OC varied with taxonomic classification (although an interaction with clay content does not appear to have been considered in their analyses). Reference Bulk Density The reference BD varied with clay and OC contents (Table 1). Differentiating BDref with respect to OC gave:

4 658 CANADIAN JOURNAL OF SOIL SCIENCE Table 3. Changes in the bulk density and the reference bulk density when the organic carbon content of soils from different regions of the world is increased by 1 g 100 g 1 Bulk Reference Characteristics of data base Details of Geographic density bulk density Range in Cultivation/ analyses location (Mg m 3 ) Reference n soil texture crop R 2 Canada Southwestern 0.081* z da Silva et al. 144 for BD Loamy sand to Arable/corn 0.83 for BD Ontario (at 200 kpa) (1997) 72 for BD ref silty clay loam 0.85 for BD ref Manitoba 0.093* Shaykewich and 112 Sand to clay Not given Zwarich (1968) United States National data base 0.088* Rawls (1983) 2721 Sand to clay Not given Florida, Iowa, Hudson (1994) 20, 18, and 21 Arable 0.73 for sand Wisconsin, in sand, silt loam 0.74 for silt Minnesota, and silty clay loam Kansas loam classes 0.67 for silty respect. clay loam Ohio Saini (1966) 30, 40 and 40 in Not given Not given 0.74, 0.65, and Humic-gley, 0.40 for Humic, imperf. drained gley, imperfectly well drained drained and well soils respect. drained respect. Dakota Bauer and Black 12 in each of sand, Arable and 0.64, 0.66 and (1992) sand, medium and grasslands 0.83 for sand, fine texture classes medium and fine (fine texture mont.) texture classes respect. Europe Scotland Soane (1990) 58 Not given Arable 0.52 z Calculated at a clay content of 21%. Fig. 1. Variation in BD of the test soils at RC of 0.75 (RC l), 0.85 (RC m) and 0.95 (RC h) when the OC content is increased from the initial level (OC l) to initial plus 0.01 kg kg 1 (OC h).

5 KAY ET AL. PREDICTION OF CHANGES IN SOIL STRUCTURE 659 Fig. 2. Variation in total porosity (TP) and the water contents at FC and the PWP of the test soils at a RC of 0.85 when the OC content is increased from the initial level (OC l) to initial plus 0.01 kg kg 1 (OC h). BDref/ OC = Cl (3) Increasing OC content decreased the maximum density under compaction, and once again the magnitude of the effect diminished as the clay content increased. At a clay content of 21%, the change in the reference BD associated with an increase in the OC content of 1% (0.01 kg kg 1 ) was Mg m 3. Soane (1990) related the maximum bulk densities of 58 Scottish soils to OC content using simple linear regression and found that BDref/ OC was Soane determined BDref using the Proctor test, which Häkansson (1990) found to give higher BDref values than those obtained using an oedometer (the same procedure used by da Silva et al. 1997). This could result in Soane s value of BDref/ OC being slightly larger than the value obtained from Eq. 3. Relative Compaction The ratio of the observed BD to the BDref was originally proposed as a measure of management-induced soil compaction that would be largely independent of texture and OC (Häkansson 1990). The ratio has been referred to as the degree of compactness or relative compaction (RC) (Häkansson 1990; Carter 1990). Stepwise multiple regression analyses showed (da Silva et al. 1997) that RC was influenced by tillage and position relative to the row but was not significantly influenced by texture and OC contents. The values of RC in the row position for conventional and zero tillage were 0.79 and 0.87, respectively, while the corresponding values in the interrow position were 0.86 and The mean value of RC across the site was These values are comparable to values reported in the literature. For instance, Carter (1990) found the RC of a fine sandy loam at a depth of 0 8 cm was as low as 0.72 approximately 20 d subsequent to tillage and planting to cereals, and that it gradually increased to 0.84 over a 150-d growing season. The RC of the same soil under zero tillage was Bulk density was included in the PTF describing both the water release curve and the soil resistance to penetration curve (Table 1). The sensitivity of BD to tillage, traffic, weather and biological factors dictate that the RC be considered in assessing the sensitivity of different structural characteristics to changes in OC. This was done by calculating the reference BD for each of the 5 test soils using the PTF, and then calculating the bulk densities for values of RC = 0.85 and for values of RC corresponding to two extremes in compaction, i.e., 0.75 and The sensitivity of the BD of the five test soils at the three values of RC to an increase in OC content of 1% is illustrated in Fig. 1. The decrease in BD with increasing OC content was most pronounced in the coarser-textured soils and was enhanced slightly with increasing compaction. Field Capacity, Permanent Wilting Point and Available Water The water contents at field capacity (FC), permanent wilting point (PWP) and the potentially available water (AWC = FC PWP) have been predicted by different researchers

6 660 CANADIAN JOURNAL OF SOIL SCIENCE Fig. 3. Variation in the potentially available water content of the test soils at a RC of 0.75 (RC l), 0.85 (RC m) and 0.95 (RC h) when the OC content is increased from the initial level (OC l) to initial plus 0.01 kg kg 1 (OC h). Table 4. Changes in the water content at field capacity ( 0.01 MPa), permanent wilting point ( 1.5 MPa), and the available water capacity of the loam test soil (21% clay) when the organic carbon content was increased by 0.01 kg kg 1 as predicted from PTF derived for soils from different regions of the world θ0.01 θ1.5 θ Characteristics of data set Details of Geographic Horizon number Range in analyses location (m 3 m 3 ) Reference and depth soil texture Cultivation/crops R 2 Canada Southwestern Ontario da Silva and Kay (1997a) z 256 Loamy sand Arable/corn 0.94 A horizon to silty clay loam U S A national data base Rawls and Brakensiek 2541 Sand to clay Not given 0.81 for FC (1982) y All Horizons 0.80 for PWP Europe Belgium Vereecken et al. (1989) z 182 Sand to clay Not given 0.85 for θ s All horizons 0.70 for θ r 0.68 for log α 0.56 for log n United Kingdom Thomasson and Carter Not given Not given 0.72 for FC (1992) z A horizon 0.71 for PWP Australia Williams et al. (1992) z, 78 Sand to clay Not given 0.90 function No. 1 All horizons z Soil properties related to coefficients of a model of a water release curve. y Soil properties related to water content at specific potential. using PTFs to describe either the entire water release curve or the water contents at the relevant potentials (Ψ). Where the entire water release curve is described, different models may be used to relate water content to potential. The PTF developed by da Silva and Kay (1997a) was based on the entire ( 1.5 MPa Ψ MPa) water release curve, using a model similar to those used by Campbell (1974), Hutson and Cass (1987) and Williams et al. (1992). The

7 KAY ET AL. PREDICTION OF CHANGES IN SOIL STRUCTURE 661 Fig. 4. Variation in air-filled porosity at FC of the test soils at a RC of 0.75 (RC l), 0.85 (RC m) and 0.95 (RC h) when the OC content is increased from the initial level (OC l) to initial plus 0.01 kg kg 1 (OC h). coefficients in the model were described as functions of BD, clay content and OC content (Table 1). The sensitivity of the total porosity, TP, (TP = [2.65 BD]/2.65) and the water contents at FC ( 0.01 MPa) and PWP ( 1.5 MPa) to increases in OC content were predicted for the five test soils using the PTFs for the water release curve and the bulk densities calculated for the three degrees of compaction. Values of TP and the water contents at FC and PWP for RC = 0.85 are given in Fig. 2. The variation in available water content with texture at the three levels of compaction and different OC contents were also calculated (Fig. 3). Increases in OC increased the water content at FC and the PWP across the range of clay contents. The increase at FC was greater than that at the PWP, resulting in an increase in AWC at all clay contents. However, the magnitude of the increase in AWC varied with clay content. The increase in water contents at FC and PWP, due to increased OC content, became greater as the clay content increased with the largest changes exhibited at the PWP. Consequently, the increase in AWC due to increasing OC content diminished as the clay content increased. Similar trends were predicted at the different values of RC. The changes in the volumetric water contents at FC and PWP and the change in AWC arising from an increase in the OC content of 1 % predicted from the PTF in Table 1 were calculated for the loam test soil (clay content of 21%) at RC = 0.85 and compared to corresponding predictions using PTFs from the literature (Table 4). PTFs were selected to give a combination of the two approaches (PTFs that predict the entire water release curve and the water content at a specific potential) and to give a broad geographic distribution of the data bases used to derive the functions. In addition, only functions were selected that: (a) were based on measurements made on intact cores, (b) predicted water contents on a volumetric basis and (c) had coefficients of determination that were The PTF derived by Vereecken et al. (1989) and Williams et al. (1992) described the entire water release curve and, as was the case for the function derived by da Silva and Kay (1997a), the magnitude of the increase in water contents with OC at the different potentials varied with texture. Changes in water contents were, therefore, calculated for the loam test soil (21% clay) for comparative purposes. The PTFs derived by Vereecken et al. (1989) and Thomasson and Carter (1992) included BD as a variable. In order to standardize the comparison, the PTF for BD derived by da Silva et al. (1997) (Table 1), or its derivative (Eq. 1), was used to calculate the BD (for the row position, zero till) and the changes in water content calculated at 21% clay. It should be noted, however, that the changes in water contents predicted by the functions derived by Thomasson and Carter (1992) were strongly influenced by the BD/ OC term, and therefore will include errors if Eq. 1 does not apply to their data base. The PTF derived by Williams et al. (1992) included terms for both coarse and fine sand content; values for the total sand content of the test soils (Table 2) were substituted for the fine sand content. Functions derived by Rawls and Brakensiek (1982) indicated that the increase in water content with OC content at the different potentials was independent of texture. The changes in water contents with OC that are predicted from the PTF derived by da Silva and Kay (1997a) at 21% clay are similar to those predicted from the function derived by Rawls and Brakensiek (1982) and are much larger than predictions based on soils from Belgium (Vereecken et al.

8 662 CANADIAN JOURNAL OF SOIL SCIENCE (a) (b) 1989) and the UK (Thomasson and Carter 1992). It should be noted, however, that both of the European data sets included a wider range of soil textures than was considered by da Silva and Kay (1997a) and that half of the samples in the data set used by Vereecken et al. (1989) came from the B horizon where the organic matter content would be expected to be low. The changes in water content with OC predicted from the PTF derived by Williams et al. (1992) for some Australian soils are small in comparison with predictions derived from other data bases. The model used by Williams et al. to describe the water release curve and the statistical procedures used to develop the PTF were identical to those used by da Silva and Kay suggesting that differences in the coefficients in the two PTFs reflect differences in the characteristics of soils in the data sets. Soils in the data set used by da Silva and Kay were topsoils, dominated by illite and poorly weathered mica. Williams et al. (1983) found that montmo-

9 KAY ET AL. PREDICTION OF CHANGES IN SOIL STRUCTURE 663 (c) Fig. 5. Variation in soil resistance to penetration of the test soils when the OC content is increased from the initial level (OC l) to initial plus 0.01 kg kg 1 (OC h) at FC and the PWP and RC of (a) 0.75 (RC l), (b) 0.85 (RC m), and (c) 0.95 (RC h). rillonite, iron oxide, vermiculite and quartz had a stronger influence on the water release curve than illite and that iron oxide and vermiculite were present in approximately half of the soils in the data set used to develop their PTF (their data set was also made up of samples from depths of up to 2.4 m). The extent to which mineralogy or climate influences different forms of OC and subsequently pore characteristics is not known, but a comparison of predictions from the two PTFs does suggest that extreme caution should be used in applying a PTF to soils that are developed under conditions that are much different to those used to derive the PTF. A similar conclusion can be drawn from analyses carried out by Kern (1995) and Batjes (1996). A comparison of six PTFs by Kern (1995) indicated that the best prediction of the volumetric water content at FC and the PWP for in excess of soil samples from the United States of America were obtained using the PTFs that were derived from data collected from a wide range of soils from the United States of America. Batjes (1996) found that the predictive capability of a PTF for the calculation of available water, that was based on a world data set, was relatively low and consideration of several factors including FAO-Unesco soil unit type was required to improve predictions. Aeration Soil aeration is directly related to soil water retention characteristics. The flux of oxygen in soils decreases as an increasing proportion of the pore space becomes waterfilled and approaches zero when there are no continuous airfilled pores remaining. This limiting condition has generally been found to occur when the air-filled porosity (the fraction of the soil volume filled with air) approaches 10% (e.g. Xu et al. 1992). The air-filled porosity that exists when rapid drainage has ceased is, therefore, an important structural characteristic of soils (Thomasson 1978). The air-filled porosity at FC, i.e. TP FC, was calculated for each of the test soils at the three values of RC. The OC content was increased by 1% and the air-filled porosity was recalculated (Fig. 4). Air-filled porosity decreased with increasing clay content and increasing compaction and fell below the limit of 10% in a number of cases, indicating that an inadequate supply of oxygen could limit plant growth, even if the soil water potential was at FC. An increase in the OC content of 1% reduced the air-filled porosity at FC for all clay contents and values of RC (Fig. 4). This unexpected trend was a consequence of the increase in OC giving rise to a larger increase in FC than in TP (Fig. 3). The trend was, however, compatible with unpublished measurements of saturated hydraulic conductivity from the site. Stepwise multiple regression analyses indicated that saturated hydraulic conductivity decreased with increasing OC content. There are few data in the literature with which the prediction of decreasing air-filled porosity with increasing OC content can be compared. Thomasson and Carter (1992) found no correlation between the air-filled porosity at a potential of 0.05 MPa of topsoils in the United Kingdom and other soil properties. However, a trend similar to the trend observed in this study was predicted using the PTFs derived for BD by Rawls (1983) and for FC by Rawls and Brakensiek (1982) for soils from the United States of America. The predicted change in BD of Mg m 3 (Table 3) with an increase in OC of 1% is equivalent to a

10 664 CANADIAN JOURNAL OF SOIL SCIENCE Fig. 6. Variation in the LLWR of the test soils at a RC 0.75 (RC l), 0.85 (RC m) and 0.95 (RC h) when the OC content is increased from the initial level (OC l) to initial plus 0.01 kg kg 1 (OC h). change in TP of m 3 m 3 and is smaller than the change in FC predicted for the same data set ( m 3 m 3, Table 4). Soil Resistance to Penetration Root growth decreases with increasing soil resistance to penetration. Although the functional dependence of root growth on soil resistance depends on a number of factors, root growth is often found to approach minimum values when the soil resistance, as measured with a penetrometer, is about 2 MPa (e.g. Taylor et al. 1966). Soil resistance to penetration was described by da Silva and Kay (1997a) using a model proposed by Busscher (1990) with water content and BD as the primary variables. The PTF linked the coefficients in the model to clay and OC contents (Table 1). The sensitivity of soil resistance to increasing OC content was predicted by calculating the soil resistance to penetration at FC and PWP of the five test soils at each of the three states of compaction (Fig. 5a, b and c). Soil resistance decreased with increasing OC content when the soil was held at a constant water potential and the decrease was greatest under the most compacted conditions and the lowest (most negative) water potential. The implications of increasing the OC content on soil resistance were most obvious at a RC = 0.85 (Fig. 5b). In all but the coarsest textured soil the soil resistance exceeded 2 MPa at a potential of 1.5 MPa at the original OC contents (OC l). As these soils dry, root growth would be increasingly curtailed by soil resistance at potentials above PWP in all but the coarsest textured soil. An increase in OC content of 1% resulted in the soil resistance falling below 2 MPa on all soils implying that water potential, rather than soil resistance, would impose progressively greater limitations to plant growth as the water content decreased to that at PWP. At higher levels of compaction, the effect of OC on soil strength was larger but, at the PWP, the soil resistance of the test soils was so large that the decrease in soil resistance arising from the increased OC content was not enough to lower the soil resistance below 2 MPa on three of the five test soils. There have been few attempts to describe soil resistance to penetration using PTFs with which these predictions can be compared. The most comprehensive model has been developed by Canarache (1990) but his model does not include OC as a variable. Least Limiting Water Range Plant growth can be influenced by the collective changes in aeration, available water and soil resistance to penetration that arise from changes in soil OC and/or soil compaction. The influence of soil structure on aeration, available water and soil resistance to penetration are incorporated into a parameter referred to as the non- or least limiting water range (LLWR) (Letey 1985; da Silva et al. 1994). The LLWR represents the range in water contents defined at the upper end by the water content at FC or the water content at which aeration becomes limiting (whichever is smaller) and at the lower end by the water content at PWP or the water content at which soil resistance to penetration by roots becomes limiting (whichever is larger). Da Silva and Kay

11 KAY ET AL. PREDICTION OF CHANGES IN SOIL STRUCTURE 665 Fig. 7. Change in the SGR of corn as a fraction of the maximum shoot growth rate (SGR/SGR max ) corn in the test soils at a RC of 0.75 (RC l), 0.85 (RC m) and 0.95 (RC h) when the OC content is increased from the initial level (OC l) to initial plus 0.01 kg kg 1 (OC h). (1997a) used the water content at which air-filled porosity was 10% as the aeration limit (Xu et al. 1992) and a soil resistance to penetration of 2 MPa as the soil strength limit (Taylor et al. 1966). Using these limits and values of FC and PWP of 0.01 and 1.5 MPa, respectively, the LLWR was calculated for each of the five test soils at each of the three states of compaction using the PTF for BDref, the water release curve and the soil resistance curve. The OC content was then increased by 1% and the LLWR recalculated for the test soils. The LLWR became smaller as clay content and RC increased but the magnitude of these decreases was diminished as the OC content was increased (Fig. 6). The values of LLWR at different clay contents formed a band bounded on the upper side by the conditions of least compaction and highest OC contents, and on the lower side by the highest compaction and the lowest OC contents. The change in LLWR with OC varied from 0.05 m 3 m 3 (7% clay, RC = 0.95) to 0.01 m 3 m 3 (35% clay, RC = 0.95). Changes in the LLWR with an increase in OC of 0.01 kg kg 1 at RC = 0.85 varied from 0.01 at 35% clay to 0.05 m 3 m 3 at 14 and 21% clay. The sensitivity of plant growth to an increase in the LLWR has been found to vary with the magnitude of the LLWR of the surface horizon. Da Silva and Kay (1997b) found that the shoot growth rate (SGR) of corn at a given LLWR divided by the maximum growth rate under given climatic and tillage conditions (SGR max ) increased with the LLWR of the surface horizon according to a logistic function, i.e. SGR/SGR max = {1 + exp[ g (LLWR h)]} 1 (4) where g and h are constants equal to and 0.023, respectively. Substituting arbitrary values for the LLWR of 0.20, 0.15 and 0.10 into Eq. 4 gave values of the SGR/SGR max of 0.97, 0.93 and A given increase in the LLWR would, therefore, have the largest impact on plant growth at lower values of the LLWR, i.e. soils on which management practices are leading to increased compaction. The sensitivity of plant growth on different soils to an increase in the OC content of 0.01 kg kg 1 will depend on the sensitivity of the LLWR to the change in OC and the sensitivity of plant growth to the change in LLWR. Values of the LLWR given in Fig. 6 were substituted into Eq. 4 and the SGR/SGR max calculated. The change in the SGR/SGR max arising from increasing the OC content by 1% is given in Fig. 7 for each of the test soils at the three values of RC. An increase in OC content was predicted to have the greatest effect on SGR/SGR max in soils that were most compacted (where the LLWR was lowest) and in medium-textured soils. The LLWR of soils will vary throughout the growing season as RC changes in response to tillage, traffic, wetting/drying and freezing/thawing events and the activity of soil fauna. The rate of change of the LLWR will be a function of the stability and resiliency characteristics of soils and some of these characteristics are influenced by OC contents (Kay 1998). However, no functional relationships have been established to date that permit quantitative predictions of rates of change of the LLWR (or other characteristics of structural form) in response to changes in stability or resiliency. Consequently, the impact of changes in OC contents on the change in the LLWR over time cannot be assessed.

12 666 CANADIAN JOURNAL OF SOIL SCIENCE CONCLUSIONS The possibility of increasing the sequestration of carbon in soils, thereby offsetting anthropogenic inputs of CO 2 to the atmosphere while at the same time improving soil physical conditions, has highlighted the need for quantitative information on the sensitivity of different soil structural characteristics to changes in OC contents. Information on the variation in this sensitivity with texture, other soil properties and climatic conditions is required if generalizations on a broad geographic basis are to be made. This study has illustrated that PTFs can be used to model the sensitivity of the structural characteristics of soils to changes in OC contents. However, the changes in structural characteristics that are predicted by PTFs, which have been derived from data bases from different geographic regions of the world, exhibit considerable variation. The impacts of soil and climatic conditions on these PTFs have received little attention and merit more research. Until these impacts are better understood, considerable caution should be exercised when attempting to apply functions to soil and climatic conditions that are different from the conditions associated with the data bases used to derive the functions. The PTFs derived by da Silva and Kay (1997a) and da Silva et al. (1997) would be expected to be most applicable to soils from surface horizons that are formed under a modified continental climate from calcareous parent materials and are: neutral to slightly alkaline in ph, broadly representative of coarse- and medium-textured soils, OC contents ranging from 0.8 to 6.7% and clay mineralogy dominated by illite and poorly weathered micas. The magnitude of the changes in the structural characteristics of soils that were predicted to arise from an increase in OC content of 0.01 kg kg 1 varied with the characteristic, texture and RC. Changes in soil resistance to penetration with changes in OC were found to be more sensitive to RC than were the water contents between FC and the PWP. The LLWR exhibited the greatest variation with OC content at RC of 0.85 and The increase in the LLWR (which can include limits related to aeration, available water and soil resistance to penetration) varied with clay content and was found to be greatest (0.05 m 3 m 3 ) at clay contents between 7 and 21%. The sensitivity of plant growth rate to an increase in the OC content of 0.01 kg kg 1 was predicted to vary with clay content and was greatest when the variation in LLWR with OC and the variation in plant growth rates with LLWR were optimized. The analyses suggest that sequestration of OC will not confer benefits of similar magnitude on the structural characteristics of soils of different texture (within the range of textures considered in this study). Speculation on mechanisms that could account for the relations between structural characteristics and OC content and the interaction with clay content is beyond the scope of this paper. In a recent review of literature on the influence of OC on soil structure, Kay (1998) noted the lack of information on mechanisms through which OC influences pore characteristics. Greater use of PTFs could lead to the development of a number of hypotheses that would provide opportunities for collaboration between soil physicists, soil biologists, and biochemists that, heretofore, appear to have remained unexplored. Greater use of PTFs will complement ongoing long-term studies that are being carried out under specific soil and climatic conditions on the influence of management practices on soil properties and crop productivity. Site-specific studies provide data that can be used to assess the predictions based on the PTFs while the PTFs themselves might be used, with caution, to extrapolate results to a broader range of conditions. Anonymous Canadian climate normals. Temperature and precipitation Ontario. Canadian Climate Program, Environment Canada, Ottawa, ON. Batjes, N. H Development of a world data set of soil water retention properties using pedotransfer rules. Geoderma 71: Bauer, A. and Black. A. L Organic carbon effects on available water capacity of three textural groups. Soil Sci. Soc. Am. J. 56: Bouma, J Using soil survey data for quantitative land evaluation. Adv. Soil Sci. 9: Busscher, W. J Adjustment of flat-tipped penetrometer resistance data to a common water content. Trans. ASAE 33: Campbell, G. S A simple method for determining unsaturated conductivity from moisture retention data. Soil Sci. 117: Canarache, A PENETR a generalized semi-empirical model estimating soil resistance to penetration. Soil Tillage Res. 16: Carter, M. R Relative measures of soil bulk density to characterize compaction in tillage studies on fine sandy loams. Can. J. Soil Sci. 70: da Silva, A. P. and Kay, B. D. 1997a. Estimating the least limiting water range of soils from properties and management. Soil Sci. Soc. Am. J. 61: da Silva, A. P. and Kay, B. D. 1997b. The sensitivity of shoot growth of corn to the least limiting water range of soils. Plant Soil. 184: da Silva, A. P., Kay, B. D. and Perfect, E Characterization of the least limiting water range of soils. Soil Sci. Soc. Am. J. 58: da Silva, A. P., Kay, B. D. and Perfect E Management versus inherent properties effects on bulk density and relative compaction. Soil Tillage Res. (in press). Häkansson, I A method for characterizing the state of compactness of the plough layer. Soil Tillage Res. 16: Hudson, B. D Soil organic matter and available water capacity. J. Soil Water Conserv. 49: Hutson, J. L. and Cass, A A retentivity function for use in soil-water simulation models. J. Soil Sci. 38: Kay, B. D Soil structure and organic carbon: a review. Adv. Soil Sci. (in press). Kern, J. S Evaluation of soil water retention models based on basic soil physical properties. Soil Sci. Soc. Am. J. 59: Letey, J Relationship between soil physical properties and crop production. Adv. Soil Sci. 1: Manrique, L. A. and Jones, C. A Bulk density of soils in relation to soil physical and chemical properties. Soil Sci. Soc. Am. J. 55: Phillips, D The climates of Canada. Environment Canada, Ottawa, ON. p. 97. Rawls, W. J Estimating soil bulk density from particle size analysis and organic matter content. Soil Sci. 135: Rawls, W. J. and Brakensiek, D. L Estimating soil water

13 KAY ET AL. PREDICTION OF CHANGES IN SOIL STRUCTURE 667 retention from soil properties. J. Irrig. Drain. Div., Proc. ASCE 198 (IR2): Saini, G. R Organic matter as a measure of bulk density of soil. Nature (Lond.) 210: Shaykewich, C. F. and Zwarich, M. A Relationships between soil physical constants and soil physical components of some Manitoba soils. Can. J. Soil Sci. 48: Soane, B. D The role of organic matter in soil compactibility: A review of some practical aspects. Soil Tillage Res. 16: Taylor, H. M., Roberson, G. M. and Parker, Jr., J. J Soil strength-root penetration relations for medium to coarse-textured soil materials. Soil Sci. 102: Thomasson, A. J Towards an objective classification of soil structure. J. Soil Sci. 29: Thomasson, A. J. and Carter, A. D Current and future uses of the UK soil water retention data set. Pages in M. Th. van Genuchten, F. J. Leij, and L. J. Lund, eds. Proc. International Workshop on Indirect Methods for Estimating the Hydraulic Properties of Unsaturated Soils. USDA/ARS/University of California, Riverside, CA. van Genuchten, M. Th., Leij, F. J. and Lund, L. J. (eds.) Indirect methods for estimating the hydraulic properties of unsaturated soils. Proc. of an International Workshop. USDA- ARS/University of California, Riverside, CA. Vereecken, H., Maes, J., Feyen, J. and Darius, P Estimating the soil moisture characteristic from texture, bulk density and carbon content. Soil Sci. 148: Williams, J., Prebble, R. E., Williams, W. T. and Hignett, C. T The influence of texture, structure and clay mineralogy on the soil moisture characteristic. Aust. J. Soil Res. 21: Williams, J., Ross, P. and Bristow, K Prediction of the Campbell water retention function from texture, structure, and organic matter. Pages in M. Th. van Genuchten, F. J. Leij, and L. J. Lund, eds. Proc. International Workshop on Indirect Methods for Estimating the Hydraulic Properties of Unsaturated Soils. USDA/ARS/University of California, Riverside, CA. Xu, X., Nieber, J. L. and Gupta, S. C Compaction effect on the gas diffusion coefficient in soils. Soil Sci. Soc. Am. J. 56: