1 Introduction

Decline in soil structure is commonly associated with decreased aggregate stability following loss of soil carbon. Aggregate stability is a measure of the ability of soil aggregates to withstand breakdown to small fragments when quickly moistened. Aggregate stability is commonly related to soil properties, including organic carbon, texture, clay mineralogy and the proportion of monovalent verses polyvalent cations. Factors responsible for soil aggregation are understood to be size dependent (Tisdall and Oades 1982; Oades and Waters 1991). Soil organic carbon or fractions of carbon such as labile carbon and biologically active carbon are strongly associated with the stability of macro-aggregates (Tisdall and Oades 1982). In contrast, particle size, mineralogy, cation ratios and cementing agents are strongly associated with the stability of micro-aggregates (Tisdall and Oades 1982; Chenu et al. 2000; Duchicela et al. 2012; Portella et al. 2012).

Sand content has been reported to have negative influence on aggregate stability (Idowu 2003), while clay content has positive influence (Fernandez-Ugalde et al. 2013). Sands tend to have low aggregate stability due to their large size and low surface area compared with clays, which have high surface area and negatively charged surfaces, which readily bond together via metal cations or organic molecules (Bazzoffi et al. 1995). Different types of clay minerals influence aggregate stability, due in large part to differences in surface area, charge of clay platelets and swelling behaviour (Chenu and Guerif 1991; Wakindiki and Ben-Hur 2002; Lado and Ben-Hur 2004; Fernandez-Ugalde et al. 2013). For example, smectitic clays are more dispersive than kaolinitic clays (Singer 1994), due to the large internal surface area of smectite compared to kaolinite.

Aggregate breakdown by dispersion is influenced by the proportion of monovalent verses polyvalent cations on the exchange complex, which is commonly expressed by ratios such as exchangeable sodium percentage (ESP) and sodium adsorption ratio (SAR) (Shainberg and Letey 1984; Amezketa and Aragues 1995). Traditionally, these measures have focussed on the proportion of Na+ relative to other cations due to sodium’s large hydrated radius and low electronegativity (Laurenson et al. 2011). However, the role of K+ in dispersion is being increasingly recognised. Although K+ has smaller hydration radius than that of Na+, in low CEC soils, presence of moderate levels of K+ may also result in dispersion (Auerswald et al. 1996; Igwe and Okebalama 2006). In recent years, cation ratios have been developed to incorporate the dispersive potential of K+, including the exchangeable potassium percentage (EPP), exchangeable cation ratio (ECR), monovalent cation of adsorption ratio (MCAR) and the cation ratio of soil structural stability (CROSS) (Marchuk et al. 2014, Rengasamy and Marchuk 2011). Presence of polyvalent cations (Fe3+, Al3+, Ca2+, Mg2+) commonly act as flocculating agents promoting soil aggregation (Six et al. 2000; Igwe et al. 2009). Baohua and Doner (1993) reported that soils with a low concentration of polyvalent cations are more susceptible to dispersion because of the repulsion between the negatively charged clay particles. Polyvalent cations may act together with clay or organic matter to further strengthen soil aggregates (Chan and Heenan 1999; Wuddivira and Camps-Roach 2007).

Aggregate stability may be influenced by soil management, often in association with changes in organic carbon. Intensively cultivated soils are particularly prone to carbon loss due to loss of carbon by (i) erosion following tillage and irrigation; (ii) oxidation, especially the use of powered implements that lift and throw soil into the air; (iii) mechanical breakdown due to the impact of tillage implements; and (iv) limited input of crop residues, due to above ground biomass being harvested, bailed for stock feed or directly grazed (Elliott 1986; Ross 1993; Six et al. 2000).

Measurement of aggregate stability has not been standardised. Le Bissonnais (1996) demonstrated that soils differed in their susceptibility to aggregate breakdown depending on the type of aggregate stability test as each type of test applied different forms and levels of disruptive energy. Aggregate stability is usually measured by wet sieving, which involves raising and lowering aggregates above a sieve or sieves while immersed in water. Consequently, wet sieving emphases aggregate breakdown via slaking and dispersion. However, soil crusting and disaggregation of field soils are in large part due to raindrop impact (Roth and Eggert 1994), which is not replicated in wet sieving tests. Consequently, understanding and predicting aggregate breakdown may require use of multiple forms of analysis.

The intensive management practices at trial sites in Southern Tasmania have led to loss of soil organic carbon, soil structure decline and surface crusting, which have reduced infiltration of rainfall and irrigation causing runoff, erosion and poor irrigation performance (Hardie et al. 2013). Observations between different blocks demonstrate considerable differences in the degree of soil crusting despite all blocks having similar management practices. Production of packet salad requires very high levels of cultivation, prolonged fallows and considerable irrigation. On a yearly basis, soils are cropped 2–3 times, resulting in 6–12 cultivations with a rotary hoe, and an average of 2.5 ML/ha of irrigation. Additional vehicle traffic also occurs during sowing, spraying for weeds and pests and harvesting. This study was established to (i) explore the role of specific soil attributes or combination of attributes on aggregate stability, (ii) examine to what extent associations between soil properties and aggregate stability were influenced by methodology for measuring aggregate stability and (iii) identify management options to increase aggregate stability.

2 Materials and methods

2.1 Field sites

This study was conducted on farms located in the lower Coal Valley, southern Tasmania, Australia (42° 44′ 00″, 147° 26′ 00″). The Coal Valley has a Mediterranean climate with 500 mm annual rainfall; yearly temperature and rainfall range are presented in the Fig. 1 (Bureau of Meteorology 2014). Sampling was conducted at 20 sites distributed across five properties for packet salad production. Sample sites were stratified to include topsoils developed on six different Cainozoic geological units, largely Tertiary and Quaternary sediments. Topsoil texture was fairly consistent between sites, with 18 of the 20 sites having sandy clay loams, and two sites having sandy loams.

Fig. 1
figure 1

Monthly rainfall and temperature at Richmond, Coal Valley, Tasmania, Australia

Full size image

2.2 Soil sampling and preparation

At each site, approximately 2000 g of soil aggregates was collected from the crop bed at 0–5 cm depth within 6 h after tillage and bed formation, and prior to rainfall or irrigation, using a small hand shovel. Samples were collected in triplicate, in which replicates were located 10 to 15 m apart. Soil aggregates were placed in plastic containers for transport to the laboratory. Gravimetric moisture content at the time of sampling was determined for each site by drying duplicate samples at 105 °C for 24 h. All soils were air dried at 40 °C for 24 h and then carefully hand sieved to separate out the 2.00–4.75 mm size fraction, of which around 1000 g was obtained. The 2.00–4.75 mm size fraction was selected for investigation as it is considered to be the desired aggregate size range for production of packet salad.

2.3 Measures of aggregate stability

Aggregate stability was determined on the 2.00–4.75 mm aggregate fraction for each of the three replicates. Aggregate stability was determined by wet sieving (WS) using an Eijkelkamp wet sieving apparatus, in which 4 g of aggregates was slowly immersed in distilled water and mechanically raised and lowered for 3 min on top of a 250-μm sieve. Aggregate stability was determined as the proportion of aggregates retained on the sieve after removing of coarse particles by ultrasonic disruption and re-sieving.

Aggregate stability was also determined by rainfall simulation (RS) using a Cornel sprinkle infiltrometer. Approximately 30 g of air dry 2.00–4.75 mm aggregates was placed on the top of 250-μm sieve (200 mm diameter size) mounted within a large funnel containing a 380-mm diameter filter paper (code 415) to catch the <250 μm particles as they passed through the sieve. Tap water (EC = 70 μs/m) was applied from the Cornel infiltrometer from 1.84 m height, at 0.7 kPa head, from 130 needles over a 314 cm2 area, at a constant rate of 36 mm/h. The mean weight diameter MWD of the droplets was determined by the flour pellet method to be 3.09 ± 0.14 mm (Laws and Parsons 1943). The proportion of aggregates dislodged from the sieve surface by rainfall impact was ignored. Aggregate stability was determined as the proportion of aggregates retained on the sieve after removing of coarse particles by ultrasonic disruption and re-sieving.

Clay dispersion was measured following the spontaneous dispersion method by Rengasamy et al. (1984), in which 20 g of soil aggregates was weighed into a 120-ml transparent jar, 100 ml distilled water was slowly added down the sides of the jar, then shaken end-over-end at 30 rpm for 1 h and left for 4 h to settle. The amount of suspended clay was determined by extracting a 10 ml aliquot of the suspension extracted from a depth of 5 cm and frying at 105 °C.

2.4 Chemical, physical and mineral analyses

Chemical and physical analyses were conducted on the 2.00–4.75 mm soil aggregate fraction for each of the three replicates for each site. Soil organic carbon (SOC) was determined by wet oxidation (Walkley and Black 1934) by CSBP laboratories. Labile carbon was measured by both cold and hot water extractions as described by Ghani et al. (2003), in which 3 g of air dry aggregates was placed in a 50-ml polypropylene centrifuge tube with 30 ml of distilled water, which was shaken end-over-end at 30 rpm for 30 min and then centrifuged at 3500 rpm for 20 min for determination of cold water extracted carbon. The supernatant was then filtered through a 0.45 μm cellulose nitrate membrane. The remaining sediment was re-suspended in 30 ml of distilled water and placed in a hot water bath for 16 h at 80 °C. The sediment was re-suspended then centrifuged and the supernatant extracted through a 0.45-μm cellulose nitrate membrane filter to determine the hot water extracted carbon (HWC). Total organic and inorganic carbons for both hot and cold water extracted supernatants were measured in duplicate using a Shimadzu total organic carbon analyser.

Selected soil chemical properties were determined for the 2.00–4.75 mm aggregates by CSBP laboratories according to Rayment and Lyons (2011). Electrical conductivity and soil pH in water were measured using a soil to solution ratio of 1:5, and pH in CaCl2 was measured after adding calcium chloride solution to soil solution (4A1, 4B3, 3A1). Soluble and exchangeable cations were determined by using a soil solution ratio of 1:5 (5A4, 15E1). The exchangeable acidities Al+3 and H+ were measured by titration with NaOH and HPWl following extraction with 1 m KCl in a 1:5 ratio for 1 h (15G1). Calcium carbonate percentage was determined by using dilute hydrochloric acid (19B2). Reactive Al3+ and Fe3+ were determined by Tamms reagent (oxalic acid/ammonium oxalate). CEC was measured as the effective cation exchangeable capacity (ECEC) as the sum of cations (Ca2+, Mg2+, Na+, K+ and Al3+). Sodium adsorption ratio (SAR), exchangeable sodium percentage (ESP), exchangeable potassium percentage (EPP), the cation ratio of soil structural stability (CROSS), exchangeable cation ratio (ECR) and monovalent cations of adsorption ratio (MCAR) were calculated according to Rengasamy and Marchuk (2011) and Marchuk et al. (2014).

As particle size and clay mineralogy are uniform over short distances (20 m), a single bulked sample from all three replicates was used for analysis of particle size and clay mineralogy of each site. Particle size of the 2.00–4.75 mm soil aggregate fraction was measured by mid-infrared (MIR) by CSBP (Rayment and Lyons 2011). Clay mineralogy was determined by the Mineral Resources Tasmania using an automated Philips X-Ray diffractometer system using nickel-filtered copper radiation at 40 kV/30 mA and a sample spinning proportional detector.

2.5 Statistical analyses

A number of statistical procedures were used to explore the relationships between soil chemical, physical and mineralogical properties vs aggregate stability and clay dispersion. These included (i) Spearman correlation in SPSS, (ii) forward linear regression in SAS Enterprise guide version 6.1 and (iii) regression tree analysis in JMP 10. The use of all three statistical procedures was desired to overcome assumptions and limitations associated with each type of analysis, differences in normality between soil attributes, the large number of soil attributes, and interest in creating a flow diagram based model of what influences aggregate stability for farmers. The regression tree analysis was conducted as an alternative data analysis method to nonlinear regression in which values are partitioned into smaller groups, where the interactions are explicit. Partition analysis recursively partitions data to form a tree of decision rules until the desired fit is reached (SAS Institute 2014). Decision trees were constructed manually, where splits were required to exceed a minimum logworth value of 2.0 (p = 0.01). Results are presented for significantly (logworth > 2.0) related properties. Spearman correlation and regression tree included all soil properties, while linear regression included only the correlated soil properties identified by the Spearman correlation in which clay and exchangeable Al3+ were excluded to prevent analysis of dependant variables, i.e. sand, silt and clay.

3 Results

3.1 Aggregate stability and clay dispersion verses soil properties (Spearman correlation)

The influence of soil properties on dry aggregate stability varied between both the methods used for measuring aggregate stability and each of the three statistical procedures. The 2.00–4.75 mm aggregate fraction was dominated by quartz (43–73 %) and smectite or smectite-kaolinite (8–30 %). Sand content ranged from 56 to 70 %, clay content ranged from 16 to 33 % and silt content ranged from 8 to 15 %. The mean of SOC of the 60 samples was 2.1 %, which ranged from 1.45 to 3.87 %, while the mean hot water extractable carbon was 78.7 mg/l, which ranged from 52.5 to 117.1 mg/l. Exchangeable Ca2+ ranged from 6.1 to 26.8 meq/100 g, exchangeable Mg2+ ranged from 2.3 to 15.3 meq/100 g, exchangeable K+ ranged from 0.66 to 1.65 meq/100 g and exchangeable Na+ ranged from 0.13 to 1.35 meq/100 g.

Aggregate stability determined by RS and WS was both significantly correlated with the proportion of quartz, smectite, sand and silt (Table 1). The highest correlated soil property was smectite content for aggregate stability determined by RS, compared to quartz content for aggregate stability determined by WS. Neither measure of aggregate stability was significantly correlated with clay content. Aggregate stabilities determined by RS and WS were both significantly correlated with SOC determined by wet oxidation; R 2 was 0.46 and 0.59 respectively. Clay dispersion was significantly correlated with the proportion of quartz, sand and clay respectively (Table 1), but unlike the other two measures of aggregate stability, clay dispersion was not significantly related to smectite content or any measure of soil carbon.

Table 1 Correlation between aggregate stability determined by rainfall simulation, wet sieving and clay dispersion vs soil properties
Full size table

For other soil properties, aggregate stability (RS, WS) was highly correlated with ECEC, followed by exchangeable Ca+, exchangeable Mg2+, EPP, ECR and MCAR (Table 1). Exchangeable sodium appeared to have little to no influence on aggregate stability as either Na+ or SAR were correlated to any measure of aggregate stability, while ESP and CROSS were the lowest ranked soil properties to be significantly correlated with aggregate stability (RS). Aggregate stability determined by RS was significantly correlated with pH, but no correlation existed between pH and aggregate stability determined by WS. Aggregate stability (RS + WS) was not significantly related to any of the soluble cations.

Correlation between aggregate stability determined by clay dispersion and soil properties was notably different to aggregate stability determined by RS and WS. Clay dispersion was highly correlated with pH and then a range of similarly correlated variables including Ca2+, Na+, SAR, ESP, CROSS and ECR (Table 1). No correlation existed between clay dispersion and exchangeable Mg+2, K+, ECEC, EPP, MCAR or soluble cations.

3.2 Soil properties vs aggregate stability (linear regression)

3.2.1 Rainfall simulation

Linear regression demonstrated that 70 % of variability in aggregate stability could be explained by exchangeable Ca2+ (Table 2). Sand, quartz, organic carbon (wet oxidation) and silt contents explained between 1 and 9 % of the variance in aggregate stability. No correlation existed between aggregate stability (RS) and cation ratios such as SAR, EPP and CROSS or organic carbon determined by hot water extraction.

Table 2 Forward linear regression between aggregate stability (rainfall simulation) and soil properties
Full size table

3.2.2 Wet sieving

The relationship between aggregate stability determined by WS and soil properties differed substantially to that determined by RS. Around 46 % of the variance in aggregate stability (WS) was explained by quartz content, 18 % by sand content, while 11 % was explained by exchangeable Ca2+ (Table 3). Smectite content and organic carbon (wet oxidation) were explained between 2 and 4 % of variance in aggregate stability (WS). Aggregate stability (WS) was not significantly related to SAR, CROSS or ESP.

Table 3 Forward linear regression between aggregate stability (wet sieving) and soil properties
Full size table

3.2.3 Clay dispersion

Clay dispersion was significantly related to pH (CaCl2), quartz content, sand content, exchangeable Na+, reactive Al3+, ESP and ECEC (Table 4). Around 45 % of the variance in clay dispersion was explained by pH, while 13 % was explained by quartz content. Other related soil properties explained only 2–5 % of variance in clay dispersion. Clay dispersion was not linearly related to soil organic carbon or soluble cations.

Table 4 Forward linear regression between clay dispersion and soil properties
Full size table

3.3 Soil properties related to aggregate stability (RS, WS) and clay dispersion according to regression tree analysis

Results from the regression tree analysis were profoundly influenced by methodology used to determine aggregate stability. Decision tree analysis indicated that aggregate stability determined by RS was most closely associated with ECEC, in which ECEC values less than 28 meq/100 g were then most closely related to reactive Al3+ followed by smectite and then ESP for soils that had reactive Al3+ less than 858 mg/kg, or EPP when reactive Al3+ was more than 858 mg/kg (Fig. 2a). When exchangeable ECEC was greater than 28 meq/100 g, aggregate stability determined by RS was most closely related to exchangeable Ca2+.

Fig. 2
figure 2

Decision tree analysis of the influence of soil properties on aggregate stability determined by rainfall simulation (a), wet sieving (b) and clay dispersion (c) (logworth > 2 = P < 0.01)

Full size image

Decision tree analysis indicated that aggregate stability determined by WS was most closely related to quartz content, followed by MCAR (quartz > 0.73) and silt (quartz < 0.73) as shown in Fig. 2b. When MCAR was greater than 0.45, aggregate stability determined by WS was influenced by sand and by reactive Al3+ (sand < 66), then by ESP (reactive Al3+ > 783).

Clay dispersion was most closely associated with pH (CaCl2), followed by quartz content and then silt content when pH was less than 6.5 (Fig. 2c). SAR contributed to dispersion when pH was greater than 6.5.

3.4 Summary of analysis: the highest five soil properties related to aggregate stability

No single soil property was consistently related to all three measures of aggregate stability and all three statistical procedures. Soil properties associated with aggregate stability differed considerably between the three measures of aggregate stability, whereas within each measure of aggregate stability, the three statistical procedures tended to identify similar soil properties. Overall, the soil properties most commonly related to aggregate stability were quartz content, followed by exchangeable Ca2+, then sand content, smectite content, pH and then ECEC (Table 5). For RS, aggregate stability had the highest association with ECEC, exchangeable Ca2+ and then to a lesser extent, EPP and smectite content. For WS, aggregate stability had the highest association with quartz content and then to a lesser extent, exchangeable Ca2+, sand content and organic carbon. Aggregate stability determined by clay dispersion was most closely related to pH, then quartz content and then to a lesser extent, sand content and SAR. Soil properties which were either not ranked or infrequently ranked in the five most related soil properties with aggregate stability (RS + WS) and clay dispersion include ESP, CROSS, exchangeable Mg2+ and labile carbon (all not listed); ECR and MCAR (both listed once); and SAR and clay content (both listed only twice).

Table 5 Presenting of the five highest ranked soil properties to be significantly related to aggregate stability by analysis type
Full size table

4 Discussion

The type of soil properties associated with aggregate stability differed with each of the three methods used for measuring aggregate stability. Aggregate stability determined by RS was related to soil properties associated with aggregation, namely ECEC and the proportion of polyvalent cations (Ca2+, Al3+), while aggregate stability determined by WS was more closely related to soil properties associated with disaggregation, namely sand and quartz content and to a lesser extent the proportion of monovalent cations (MCAR, EPP) and organic carbon. Clay dispersion was related to factors active in re-aggregation and flocculation, namely pH, quartz content and different measures of particle size. These differences emphasise that the importance of different soil properties to stability/instability of soil aggregates differs according to the method used to measure aggregate stability as also reported by Haynes (1993) and Le Bissonnais (1996). For RS, aggregates were bombarded with water that rapidly drained through the sieve such that disaggregation mostly resulted from raindrop impact and thus mechanical failure of the aggregates rather than slaking or dispersion. For WS, the aggregates were immersed and moved through a column of water such that disaggregation resulted from slaking by differential swelling and air compression, and to a lesser extent dispersion. Disaggregation during the clay dispersion test also involved slaking and dispersion but was also followed by a period of settling which allowed for flocculation.

Overall, aggregate stability determined by RS was positively related to ECEC, the proportion of polyvalent cations, and to a lesser extent, the presence of smectite (Table 5). The strength of the association with ECEC was unexpected, which was attributed to ECEC being composed of around 60 % Ca2+. Positively charged cations Ca+2 and Al+3 help bond negatively charged soil clay particles into more stable micro-aggregates (Chan and Heenan 1999; Igwe et al. 2009). This is due to polyvalent cations having greater charge density than monovalent cations, resulting in flocculation (Six et al. 1999; Wuddivira and Camps-Roach 2007). Numerous studies have shown aggregate stability increases with Ca2+, and thus the use of gypsum for promoting aggregate stability and preventing dispersion (Bennett et al. 2014, Hanay et al. 2004).

Previous studies have reported that aggregate stability is negatively related to smectite content due to these clays undergoing swelling, which occurs in soils with high ESP and low electrolyte concentration (Reichert et al. 2009, Singer 1994, Stern et al. 1991). In our study, the presence of smectite increased rather than decreased aggregate stability (RS, WS), which we attributed to the high sand content of the soil, such that presence of any clay but especially the large external surface area of smectite increased opportunity for primary particles to bond into aggregates.

Aggregate stability determined by WS was negatively related to quartz and sand contents (Table 5), which is consistent with many other studies (Chaney and Swift 1984; Kemper and Koch 1966). This is due to the large size and low surface area of sand, and because sand does not have negative charges like clay that are easily held together by aggregating agents such as metal cations or organic molecules (Bazzoffi et al. 1995).

Numerous studies have shown that aggregate stability is negatively related to various cation ratios such as SAR, ESP, CROSS and ECR (Laurenson et al. 2011; Rengasamy and Marchuk 2011), especially when disaggregation occurs by differential swelling or dispersion. In this study, aggregate stability determined by RS and WS was negatively correlated with EPP, ECR and MCAR. These cation ratios indicated that exchangeable K+ in addition to Na+ contributed to aggregate breakdown. Exchangeable K+ was found to be as, if not more important than exchangeable Na+ for disaggregation of these soils. Disaggregation due to K+ has been shown by Laurenson et al. (2011) in soil with low exchangeable Na+ and high exchangeable K+. Other studies also report that soil aggregates exposed to rainfall or overhead irrigation undergo breakdown due to high exchangeable K+ (Levy and Feigenbaum 1996; Rengasamy 2010).

Results demonstrate that clay dispersion was highly correlated with pH and quartz content, and to lesser extent, SAR and clay content (Table 5). All statistical tests ranked pH as the most closely associated parameter with clay dispersion (Table 5). Voelkner et al. (2015) also reported that clay dispersion was negatively correlated with pH. The relationship between clay dispersion and pH was attributed to Al3+ becoming more available at low pH (soil pH ranged from 4.4 to 7.4) and thus active in flocculation, while at higher pH, Na+ and K+ were likely to become available and aid in dispersion and swelling. According to the decision tree analysis and Spearman correlation, SAR was significantly related to clay dispersion, while abundance of Na+ was linearly related to clay dispersion. However, associations with Na+ were generally less important than the influence of pH and particle size on dispersion.

Unexpectedly, SOC was not consistently or strongly associated with aggregate stability. SOC was correlated with aggregate stability determined by rainfall simulation (R 2 = 0.46) and wet sieving (R 2 = 0.59) but not clay dispersion. However, linear regression indicated that only 2 % of the variance in aggregate stability determined by RS was explained by SOC. Furthermore, labile carbon was only indicated by Spearman correlation to be moderately correlated with aggregate stability determined by WS and not related to any measure of aggregate stability when analysed by linear regression or the decision trees. Overall, SOC was moderately and inconsistently associated with aggregate stability determined by RS and WS, while labile carbon was poorly associated with aggregate stability determined by WS. This finding is in contrast to the extensive literature that strongly relates aggregate stability to SOC or measures of labile carbon (Tisdall and Oades 1982; Chenu et al. 2000; Loveland and Webb 2003). In most soils, aggregate stability is known to be directly related to soil carbon content, while in some soils, this relationship has been shown to be threshold dependent (Ahmad and Roblin 1971; Carter 1992), in which organic carbon content needs to be exceeded before a relationship between aggregate stability and SOC is established. For example, Boix-Fayos et al. (2001) showed that a threshold of 3–3.5 % SOC had to be attained to achieve increases in aggregate stability, in which carbon content had no effect on aggregate stability below this threshold. We postulate the lack of a relationship between labile carbon and aggregate stability determined by RS, and the poor to moderate relationship between SOC and aggregate stability determined by RS and WS was due to the soil carbon levels at most sites being below or near a soil carbon-aggregate stability threshold. Thus it appears that carbon loss was so extensive at most sites, and the remaining soil carbon no longer makes a substantial contribution to aggregate stability. Further research is required to have better understanding the nature of the remaining recalcitrant carbon and its potential role in aggregation.

5 Conclusions

Aggregate stability was found to be related to different soil properties depending on the means by which aggregate stability was determined and to a lesser extent type of statistical analysis. Overall, RS demonstrated aggregate stability was related to soil properties that promote aggregation and flocculation such as Ca2+ and ECEC, while WS demonstrated aggregate stability was related to soil properties that promote disaggregation and dispersion including sand content, quartz content and to lesser extent the proportion of monovalent cations. Clay dispersion was closely related to pH possibly as a surrogate for Al3+ at low pH, and Na+ at high pH. Curiously, aggregate stability determined by RS and WS was only moderately correlated with SOC; however, neither aggregate stability (RS, WS) nor clay dispersion was closely associated with labile carbon or the abundance/proportion of Na+. For aggregate stability determined by RS and WS, the amount of K+ appeared to be as important if not more important than the amount Na+ in promoting disaggregation.

Management options for improving aggregation appear limited as aggregate stability was mostly related to inherent soil properties such as sand/quartz and smectite content. However, the positive relationship between aggregate stability determined by RS and polyvalent cations such as exchangeable Ca2+ and Al3+ may provide some opportunity to improve aggregate stability through application of products that are rich in Ca2+ such as gypsum, or Al3+ such as alum. We found little evidence that aggregate stability determined by RS and WS was influenced by total carbon or labile carbon; as such, it is likely that the remaining carbon in these soils is recalcitrant and not actively involved in aggregation. However, this does not preclude the addition of new carbon for promoting aggregation. Field trials are required to determine if application of compost, manure or crop residues would supply or enable production of organic compounds that do promote aggregation.