Abstract

Key Words

Raised beds, image analysis, structure, porosity

Introduction

In the past, crop production in southwest Victoria has been restricted on the predominantly duplex soil types found in this area Gardner et al. (1992). The major drawbacks with these soils are their susceptibility to waterlogging and their hardsetting nature. Poor crop yields have been associated with hostile soil physical conditions. These soils typically have slow internal drainage and little plant available water (Robinson et al. 2003).

Waterlogging is the principal concern and it occurs seasonally due to a combination of factors; the relative abundant winter rainfall, small evaporation rates and the hydraulic properties of duplex soils. In particular the slow permeability of the medium to heavy clay B horizons prevents the soil from freely draining. A hydraulic throttle develops at the interface between the A and B horizons which is caused by an abrupt change in texture. This characteristic often leads to the formation of a saturated region which corresponds to the root zone of growing crops.

The use of raised beds (RB) in cropping systems has recently become popular in southwest Victoria. Farmers have adopted RB as a convenient form of surface drainage that reduces the risk of waterlogging and improves the potential for cropping. It is not clearly understood which soil properties of RB provide improved growing conditions compared to conventional cultivation systems.

There has been limited research on RB in southwest Victoria. Recent work on RB has focused on their agronomic performance with only brief attention paid to soil properties. Peries et al. (2001) reported improvements in structure of RB after measuring reduced soil bulk density. Peries et al. (2003) postulated that controlled traffic may be the cause of changes to structure on RB. Well-structured soil contains sufficient space (such as macropores) to allow roots access to water and nutrients (Passioura, 1991). Furthermore, the presence of macropores can improve the infiltration rate of the soil, although the magnitude of increase depends on many factors including macropore shape and size as well as antecedent soil water content (White, 1985).

There have been several studies that have used image analysis to assess structure and provide useful information on different cultivation systems. Shipitalo and Protz (1987) found significantly greater macroporosity in a tilled treatment compared to a no-till plot. Associated with this the mean pore diameter was smaller and pore shape was significantly different. In contrast, a study with similar treatments by Moran and McBratney (1992) showed that cultivation severed the connection of surface pores to the subsoil below. A different study by Moran et al. (1988) measured both porosity and hydraulic conductivity. Greater macroporosity was correlated with greater unsaturated hydraulic conductivity (at 10 mm tension but not 40 mm), which suggests a soil structural benefit was detected. Recently, Vervoort and Cattle (2003) investigated relationships between soil structural parameters (derived from image analysis) and hydraulic conductivity. Strong correlations between saturated hydraulic conductivity with porosity and pore connectivity (as measured by pore genus) were detected. In addition, they estimated pore tortuosity to provide an improved framework on the relationship of pore parameters (mean pore size) with important soil properties such as hydraulic conductivity. The association between structure and infiltration rate is of functional importance for the ability of the soil to drain. Improved water flow through the soil will reduce the likelihood of waterlogging and improve cropping potential.

It is hypothesized that the soil structure of RB is better, by providing a more aerated and improved environment for plant growth. The objective of this paper is to directly assess the soil pore structure of RB in comparison to CC. A resin-impregnated image analysis method was chosen to investigate pore structure. Quantification of porosity, the pore size distribution and several parameters of pore structure were obtained to investigate this. Exploration of the relationships between structural parameters of each tillage system was undertaken to predict the functional performance of RB. Prior to and in addition to this work, hydraulic conductivity and moisture retention characteristic data were collected, but no significant differences were found (data not shown).

Material and Methods

The experimental site was 30 km west of Geelong in southwest Victoria. The experiment was a randomised block design of tillage treatments (0.2 ha plots) replicated three times. Soil was sampled from two plots including one from each treatment, raised beds (RB) and conventional cultivation (CC). Average annual rainfall for the district is 525 mm. The soil is a grey Sodosol (Isbell, 1996), with a sandy clay loam Ap horizon (up to 15 cm) overlying a gravelly clay B horizon. The depth of the Ap horizon is distinctly variable across the site. An indication of this is shown by the clay content in the top 10 cm, which ranges from 19 to 34 %. The structural condition of the soil is strongly influenced by these and other soil properties. A selection of soil properties for the site (Table 1) provides further background information.

Table 1. Selected soil properties at the experimental site.

Horizon	Depth (cm)	pH (water)	EC (dS/m)	Exchangeable cations (meq/ 100g)	ESP (%)	Clay (%)
Ap	0 – 10	6.2	0.21	16.1	5.0	28.5
B21cg	10 - 75	7.4	0.15	22.6	12.1	48.5
B22	75 - 110	9.1	0.23	18.2	23.6	32.5

Undisturbed soil samples were collected using aluminium soil cores (100 mm outside diameter, 75 mm height). The cores were pushed slowly into the soil by hand to ensure that minimal disturbance was caused during excavation. The location of the excavation sites was mid-way between the furrows of the RB to avoid compacted areas. Samples were taken from the surface and from 75 mm depth which was immediately below the top core. Sampling was done at the beginning of the growing season in June 2003 (T1) and post-harvest in February 2004 (T2). Due to the hard nature of the soil in the CC treatment the lower core was pushed in with the aid of a hydraulic cylinder at T1. At T2 a dripper system was set up to slowly wet up the soil and soften it.

In the laboratory, cores were put into an oven (40^o C) for at least 24 hours. The method to impregnate the soil is described by Salins and Ringrose-Voase (1994). In summary, the samples were submerged in a tub of polyester resin mixture (CR64 polyester embedding resin, methyl methacrylate monomer and a hydroperoxide catalyst). This was then placed in a vacuum chamber to remove entrapped air. A fluorescent dye (UVtex OB fluorescent dye – Ciba-Geigy Pty Ltd) was added to the resin to illuminate the pore space under ultraviolet (UV) light. Once hardened the cores were cut off and each sample was sectioned horizontally. Cutting was done at 10 mm intervals with a circular diamond saw bathed in kerosene. This cutting method was slow (10 min per cut), but it produced high quality sections with well-defined, smooth faces. Seven sections were prepared from each core starting from the surface at 10 mm depth, 20 mm, 30 mm and so on through the lower core down to 140 mm. Polyester resin was chosen because of its ability to produce well-impregnated soil samples, as it is sufficiently viscous to penetrate most pore space in soils. Its slow curing nature (usually up to six weeks) ensured that structural integrity is maintained as samples do not shrink (Murphy, 1986).

The sections were photographed under ultraviolet light using a digital camera (Olympus C-4040). A UV filter was fitted to the camera to ensure that only light fluorescing from the subject reached the camera lens. The photographs were stored in a TIF format to preserve maximum detail and avoid image compression.

After photographing the sections the first stage of image processing was to segment the solid and pore components. Initial observation of the photographs often revealed a brightly illuminated pore space as a network of pores and fine cracks. The soil matrix was clearly identified, often in two shades of colour with the darker areas containing greater clay content. In many sections this was interspersed with numerous stones which appeared black. This wide range of colours and shades produced a randomly distributed histogram with several peaks of intensity (Figure 1).

Figure 1. Pixel frequency histogram of a typical image showing the full spectrum of light including the red, green and blue channels.

Small luminance values correspond to pixels that represent solid areas and large values to pixels in pore spaces. The grey area between these extremes is the area of interest that requires differentiation for image analysis. This processing operation is called image segmentation. Care was taken to ensure that the segmented image accurately reflected pores in the photograph. Ringrose-Voase (1996) warns that small differences in threshold value can cause considerable changes in the number of pixels that are designated as either pore or solid space. A simplified procedure based upon the watershed extraction method was adopted using OPTIMAS software (Media-Cybernetics, 1999). Reith and Mayhew (1988) defined the watershed extraction method as detecting the set of all points on the surface of a grey tone function that represent minima.

The intensity of pore pixels close to solids was recorded randomly across seven images to determine the most representative value to threshold the image. Images were viewed at close magnification to easily detect individual pixels. The five smallest pore pixel intensity values were recorded, to allow the minimum pore pixel intensity to be estimated. Thresholding was performed using this minimum value to convert every grey level image into a binary format. The calculated luminance values were 135 for T1 and 142 for T2 samples. Lastly, the image was inverted so that pore space represented black and solid space white. The binary image was saved in an 8-bit grey level format and loaded into SOLICON software Cattle et al. (2000) for batch processing. The output parameters calculated were porosity, surface area (SA), pore star length (PSL), solid star length (SSL), mean pore intercept length (MPIL), mean solid intercept length (MSIL), pore genus (PG), solid genus (SG), mean pore star area (MPSA) and mean pore star shape (MPSS). Porosity is the proportion of pixels representing pore space of the whole image, SA measures the pore: solid interface, and PSL or SSL is the expected continuous length of a pore or solid respectively. MPIL and MSIL are effectively the mean pore or mean solid diameter, PG and SG measure of pore or solid connectivity, MPSA is the pore area averaged over the whole image and MPSS indicates pore shape, with circles given a value of 1 and long thin objects a value close to 0. For further details on these structural parameters see the SOLICON user manual Cattle et al. (2000) or McBratney and Moran (1990). Image selections were circular and mostly between 1200 and 1300 pixels in diameter. The actual size of each image selection was 90 to 93 mm; therefore, each pixel was approximately 70 to 75 µm width. Statistical analysis was done using GenStat software (VSN-International, 2003).

Results and Discussion

The proportion of pores within specified class sizes was averaged for all depths (Figure 2). There was very little difference between the two treatments for the pore classes or the solid components.

Figure 2. Pore size distribution by treatment at the second sampling time.

The structural parameters calculated using SOLICON are summarized (for all depths) for both treatments and sampling times (Table 2). At the first sampling time RB samples had greater porosity than CC samples. This combined with a larger PG indicates a better connected and larger network of pore space. Consequently, the RB samples have a smaller MSIL indicating that the aggregates are smaller, which may allow improved root exploration. SA was much higher for RB samples which is expected with the greater porosity. The SA measurement is the area of solid soil where resin has flowed into the soil matrix. Functionally this suggests a larger pore network where water might flow through when the soil drains. PSL and MPIL are different descriptors of the pores and both were larger in CC than in RB. MPSA was also larger which suggests a larger pore network in the CC. Shape (MPSS) was similar for each treatment across both times.

At the second sampling time there were smaller differences between the treatments and the values of many parameters were close to those of the CC samples at the first sampling time. Values of parameters which indicate the soil is well-structured (porosity – the amount of pore space, MPIL – the size of pores and PG – effective pore connectivity) all decreased from the first to the second sampling time for RB samples. This is surprising as the second sampling time was post-harvest and it was expected that the soil would have dried out and some cracking might have developed.

Table 2. Image analysis parameters by treatment and sampling time (Standard Errors are shown in brackets).

Image analysis parameter	Raised bed, T1 (n = 38)	Conventional cultivation, T1 (n = 35)	Raised bed, T2 (n = 70)	Conventional cultivation, T2 (n = 69)
Porosity (mm3 mm-3)	0.35 (<0.01)	0.32 (<0.01)	0.31 (<0.01)	0.31 (<0.01)
SA (mm2 mm-3)	1.78 (0.04)	1.28 (0.03)	1.40 (0.02)	1.31 (0.02)
PSL (mm)	0.80 (0.02)	1.00 (0.04)	0.75 (0.03)	0.90 (0.05)
SSL (mm)	6.23 (0.40)	6.77 (0.18)	7.20 (0.18)	7.45 (0.31)
MPIL (mm)	0.31 (<0.01)	0.35 (<0.01)	0.29 (<0.01)	0.31 (<0.01)
MSIL (mm)	1.90 (0.16)	2.31 (0.06)	2.38 (0.05)	2.43 (0.06)
PG ( x 10-2 mm-2)	0.12 (<0.01)	0.06 (<0.01)	0.05 (<0.01)	0.05 (<0.01)
SG ( x 10-2 mm-2)	0.86 (0.02)	0.67 (0.02)	0.87 (0.02)	0.79 (0.01)
MPSA	228.3 (2.80)	241.7 (1.99)	212.4 (1.31)	218.3 (1.22)
MPSS	0.49 (<0.01)	0.50 (<0.01)	0.50 (<0.01)	0.50 (<0.01)

Trends with depth were investigated for each parameter at separate sampling times. Overall, the response to depth was mixed with the values of many parameters apparently random and insensitive to depth. However, there were some that did correlate with depth, such as the weak positive trend observed for porosity (Figure 3). The greater porosity at 80 mm is thought to be an artefact of the sampling division between cores. It is at this depth that the first section of the second (lower) core is taken.

Figure 3. Average porosity values with depth for the second sampling time

An aggregated (both sampling times together) correlation matrix (Table 3) displays the strength of structural parameter relationships for each treatment. For many parameters the correlations are similar. However in RB treatment porosity and PG have a stronger positive correlation (0.78) than in the CC treatment (0.18). This indicates that as RB samples increased in porosity they also became more connected. PG has much higher correlations with SA, PSL, MPIL and MPSS in the RB treatment than in the CC treatment. This indicates the more connected nature of RB porosity was associated with greater measures of pore parameters (PSL and MPIL). No significance test was done to compare these correlations, although it is interesting to note that whilst the values are similar the RB values often show a stronger relationship than the CC values. MPIL, the effective size of pores, is more positively correlated with porosity of the RB soil. Not surprisingly, this corresponds inversely with MSIL, so as porosity decreases the size of solid components (effectively aggregates) increases. The correlation at the first sampling time only reveals that the effective pore diameter (MPIL) has a much stronger relationship with porosity for the RB treatment (0.77) than the CC treatment (0.46). With sampling times combined, this is hidden. In general though, most correlations did not change substantially when both sampling times were brought together.

Linear regression analysis was performed with porosity as the response variable and all the other structural parameters as explanatory variables, to determine the relationship between these descriptors of structure. Slopes were compared for the four time by treatment combinations, i.e. RB T1, CC T1, RB T2 and CC T2. This analysis highlighted that most of the significant differences detected were with RB at T1 and the other three treatment/time combinations.

SA, which was significantly different between T1 and T2 for RB samples, is of particular interest as it suggests the extent of the pore network has declined with time. Furthermore, the relationship between porosity and SA is important as it explains 64 % of the variation in porosity. MPIL explained a 42 % variation in porosity, but was the only pore related parameter that was found to be significantly different for each treatment/time combination (Figure 4).

Table 3. Correlations between structural parameters for raised beds (RB) and conventional cultivation (CC) systems from both sampling times.

	Treatment	Por	SA	PSL	SSL	MPIL	MSIL	MPSA	MPSS	PG
SA	RB	0.81	1.00	-	-	-	-	-	-	-
	CC	0.76
PSL	RB	0.39	-0.09	1.00	-	-	-	-	-	-
	CC	0.40	-0.18
SSL	RB	-0.60	-0.85	0.17	1.00	-	-	-	-	-
	CC	-0.41	-0.71	0.33
MPIL	RB	0.58	0.05	0.86	0.11	1.00	-	-	-	-
	CC	0.53	0.09	0.88	0.24
MSIL	RB	-0.77	-0.92	0.01	0.87	-0.10	1.00	-	-	-
	CC	-0.70	-0.90	0.13	0.85	0.05
MPSA	RB	-0.56	-0.60	-0.21	0.55	-0.23	0.60	1.00	-	-
	CC	-0.71	-0.66	-0.19	0.41	-0.24	0.61
MPSS	RB	-0.75	-0.29	-0.65	0.05	-0.83	0.25	0.18	1.00	-
	CC	-0.58	-0.05	-0.75	0.34	-0.78	0.00	0.25
PG	RB	0.78	0.53	0.44	-0.25	0.54	-0.46	-0.24	-0.78	1.00
	CC	0.18	0.16	-0.01	0.07	0.04	-0.10	-0.07	-0.16
SG	RB	0.00	0.42	-0.40	-0.47	-0.49	-0.42	-0.33	0.46	-0.17
	CC	0.10	0.40	-0.13	-0.39	-0.26	-0.38	-0.26	0.19	-0.26

Figure 4. Linear regression between porosity and mean pore intercept length (MPIL) at the first sampling time (a) and the second sampling time (b) for raised bed (♦) and conventional cultivation ().

In contrast, assessing the equivalent solid component of the soil was more informative. MSIL accounted for 58 % change in porosity and was significantly different between RB T1 and the other time/ treatment combinations. PG, which has an important functional meaning as a structural parameter, could provide only 32 % variation in porosity, although large significant differences were found with other time/ treatment combinations. Significant differences that were found with MPSS and MPSA were not considered to be relevant because of the small actual size of these differences (e.g. MPSS – see Table 2). Each parameter was significantly different with at least one time or treatment combination except for SG which only accounted for 15 % of porosity variation.

Conclusion

This study suggests that the RB contained a greater total number of small pores that are well-connected. In contrast, the CC is made up of a fewer number of larger pores, which indicates that RB topsoils are slightly better structured than CC topsoils. The largest measured differences were found between RB samples at T1 and the other treatment by time combinations. The pore size distribution was similar for each treatment and depth trends were either weak or not found. Pore connectivity (PG) is positively correlated more strongly with other structural parameters for RB than CC treatment. Therefore, it is predicted RB have improved soil drainage potential. However, caution should be exercised, as at times (particularly at T2) there appeared to be only small differences in the structural attributes between the RB and CC samples. Furthermore, the structure of the RB samples declined noticeably from T1 to T2. This deterioration may be due to aggregate collapse leading to a reduction in pathways connecting pores through the matrix. The performance of RB in maintaining structure over time remains uncertain. Nevertheless, this work provides a good basis for future work to further explore the soil structure of RB with a fully replicated experiment that takes account of spatial variability and temporal changes.

Acknowledgements

Impregnation and sectioning of the soil samples was done in the Butler Laboratory, CSIRO Land and Water, Canberra. The authors would like to thank A. Ringrose-Voase and I. Salins for their assistance in allowing this work to be done. Statistical advice was provided by G. Hepworth of the Statistical Consulting Centre, University of Melbourne.

References

Cattle SR, Farrell RA, McBratney AB, Moran CJ, Roesner EA, Koppi AJ (2000) Solicon ©. The University of Sydney, Cotton Research and Development Corporation.

Gardner WK, Fawcett RG, Steed GR, Pratley JE, Whitfield DM, van Rees H (1992) Crop production on duplex soils in south-eastern Australia. Australian Journal of Experimental Agriculture 32, 915-27.

Isbell RF (1996) 'The Australian Soil Classification.' CSIRO Publishing.

McBratney AB, Moran CJ (1990) A rapid method of analysis for macropore structure: 2 Stereological model, statistical analysis, and interpretation. Soil Science Society of America Journal 54, 509-515.

Media-Cybernetics (1999) Optimas Image Analysis Software. In. www.optimas.com/

Moran CJ, Koppi AJ, Murphy BW, McBratney AB (1988) Comparison of the macropore structure of a sandy loam surface soil horizon subjected to two tillage treatments. Soil Use and Management 4, 96-102.

Moran CJ, McBratney AB (1992) Acquisition and analysis of three-component digital images of soil pore structure. 2. Application to seed beds in a fallow management trial. Journal of Soil Science 43, 551-566.

Murphy CP (1986) 'Thin section preparation of soils and sediments.' A B Academic

Passioura JB (1991) Soil Structure and Plant Growth. Australian Journal of Soil Research 29, 717-28.

Peries R, Bluett C, Wightman B, Rab MA, Islam N (2003) Is Controlled Traffic Changing Soil Structure under Raised Beds in Southern Victoria. In 'International Soil Tillage Research Organisation'. Brisbane, Australia

Peries R, Johnston T, Bluett C, Wightman B (2001) Raised bed cropping leading the way in high rainfall southern Australia. In '10th Australian Agronomy Conference'. (Hobart)

Reith A, Mayhew TM (1988) 'Stereology and Morphometry in Electron Microscopy - Problems and Solutions.' Hemisphere Publishing Corporation: New York.

Ringrose-Voase AJ (1996) Measurement of macropore geometry by image analysis of sections through impregnated soil. Plant and Soil 183, 27-47.

Robinson N, Rees D, Reynard K, MacEwan RJ, Dahlhaus PG, Imhof M, Boyle G, Baxter N (2003) 'A land resource assessment of the Corangamite region.' Department of Primary Industries, Victoria.

Salins I, Ringrose-Voase AJ (1994) 'Impregnation techniques for soils and clay materials - The problems and overcoming them.' CSIRO: Australia.

Shipitalo MJ, Protz R (1987) Comparison of morphology and porosity of a soil under conventional and zero tillage. Canadian Journal of Soil Science 67, 445-456.

Vervoort RW, Cattle SR (2003) Linking hydraulic conductivity and tortuosity parameters to pore space geometry and pore-size distribution. Journal of Hydrology 272, 36-49.

VSN-International (2003) GenStat Seventh Edition. Lawes Agricultural Trust

White RE (1985) The influence of macropores on the transport of dissolved and suspended matter through soil. In 'Advances in Soil Science'. Ed. BA Stewart pp. 95-120. Springer Verlag