Changes to the physical properties of
soils puddled for rice during drying
A.J. Ringrose-Voase
a,*, J.M. Kirby
a, Gunomo Djoyowasito
b,
W.B. Sanidad
c, C. Serrano
c, Tabran M. Lando
daCSIRO Land and Water, GPO Box 1666, Canberra, ACT 2601, Australia bDepartment of Agricultural Technology, Brawijaya University, Malang, East Java, Indonesia cBureau of Soils and Water Management, Elliptical Road cnr. Visayas Avenue, Quezon City, Philippines dResearch Institute for Maize and Other Non-rice Cereal Crops, PO Box 173, Ujung Pandang, South Sulawesi, Indonesia
Abstract
Puddling is used to prepare soil for irrigated rice (Oryza sativaL.) throughout SE Asia creating a soft mud often over a plough pan. Whilst these conditions are favourable for the rice crop, they are less so for any dryland crop, such as mungbean (Vigna radiata(L.) Wilzek) or soybean (Glycine maxL. Merr.), grown in the dry season (DS) after rice harvest leading to low or erratic yields. The physical properties of soils puddled for rice were measured as they changed in the period after draining ¯ood water for rice harvest. The experiments were run at four sites in Indonesia and the Philippines for 4±6 weeks, during which they were kept free from weeds. Apart from one site, where there were heavy rain showers, the sites received no effective rainfall during the experiment. Soil moisture content and hydraulic potential in the upper 40 cm were measured regularly in ®ve replicate plots. Evaporation from the soil was measured using mini-lysimeters. Strength properties were measured using a penetrometer. Cracks were measured using an intercept technique. Hydraulic conductivity was calculated using a modi®cation of the instantaneous pro®le technique. The sites behaved in a similar fashion, with initial loss of water by drainage, followed by loss by evaporation from the surface. However, the low conductivities limited the upwards supply of water for evaporation so that evaporation from the soil surface decreased. This resulted in strong drying of the upper 5±10 cm, but much smaller decreases in moisture content lower down. The strength of the upper layers increased as they dried. The low conductivities suggest that waterlogging due to heavy rain is likely to cause problems for dry season cropping. However, the preservation of water in the sub-soil may allow ¯exibility in the sowing dates for dry season crops to avoid periods of heavy rain or shortage of labour.#2000 Published by Elsevier Science B.V.
Keywords:Soil puddling; Water movement; Shrink/swell soils; Soil strength; Rice; Dry season cropping
1. Introduction
Throughout SE Asia puddling is used to prepare soil for lowland rice. It involves cultivation of the soil after
it has been softened by ¯ooding for several days and creates a layer of soft mud which often overlies a dense plough pan. Puddling and the maintenance of ¯ooded conditions create favourable conditions for rice by aiding weed control; making seedling trans-planting easier; creating reduced conditions, which improve soil fertility, and reducing deep percolation of the standing water in which rice is grown (De Datta, 1982). In terms of the degree of alteration of soil
*Corresponding author. Tel.:61-2-6246-5956;
fax:61-3-6246-5965.
E-mail address: [email protected] (A.J. Ringrose-Voase).
structure, puddling is a rather extreme form of tillage, because it results in aggregate breakdown and the destruction of macropores. Where there is suf®cient irrigation water to grow rice all year, this does not matter. However, where there is insuf®cient water to maintain ¯ooded conditions in the dry season (DS), the soil conditions created by puddling make it dif®-cult to grow a dryland crop during the DS using moisture stored in the soil pro®le.
Typically, standing water is drained about 14 days before rice harvest to allow the soil to dry and harden. As the soil dries during this time and after rice harvest, soil strength increases and the soil shrinks and cracks. Whilst there may be suf®cient residual moisture to grow a post-rice, DS crop such as mungbean or soy-bean, the poor structural condition of both the puddled layer and the plough pan often results in poor crop establishment and root growth leading to low or erratic yields. A particular problem is that heavy rain in the early part of the dry season can cause waterlogging because of low hydraulic conductivity. Conversely, in climates where there is insuf®cient DS rainfall for the crop to survive with its roots solely in the surface layer, crops can fail because their roots are unable to penetrate the sub-soil to extract water, because of the plough pan. The aim of the experiments described in this paper was to provide basic data about the changing physical, mechanical and structural conditions in a rice soil during the period following drainage of ¯ood water in preparation for rice harvest. The data should improve understanding how the soil water pro®le changes in such situations and how these changes drive changes in structural and strength properties. This may allow development of a model of the soil water pro®le after ®eld drainage for various climatic conditions and the accompanying changes in the structural and mechan-ical properties. Such a model would allow investiga-tion of how frequently soil condiinvestiga-tions are likely to cause failure of a post-rice crop for particular soil/ climate combinations.
2. Methods
2.1. Soils
The experiments were conducted at four sites of the project on `Management of clay soils for lowland, rice-based cropping systems' (So and Ringrose-Voase,
2000): an experimental station at Barangay Buena-vista, San Ildefonso, Bulacan Province, Philippines; a farmer's ®eld in Barangay Calmay, Manaoag, Panga-sinan Province, Philippines; a farmer's ®eld at Alle-polea near Maros, S Sulawesi, Indonesia and an experimental station at Ngale, near Ngawi, E Java, Indonesia. These sites have been described in detail by Schafer and Kirchhof (2000).
Ringrose-Voase et al. (1995) present some analyses of the soil (Table 1). The Ngale soil is a heavy clay with 800 g kgÿ1 <2mm. Maros and Manaoag have very similar particle size distributions with 500± 550 g kgÿ1<2mm and are both silty clays. San Ilde-fonso has the least clay (350±400 g kgÿ1<2mm) and is a clay loam to clay. Unlike the other soils it has a signi®cant sand fraction (350 g kgÿ1>50mm) in the top soil. The clay fractions show a range of miner-alogical compositions. San Ildefonso is dominated by smectite (63%) and kaolinite (35%); Manaoag by smectite (70%), vermiculite (15%) and kaolinite (10%); Maros by kaolinite (50%), smectite (25%) and illite (15%) and Ngale by smectite (95%). The range of clay minerals and clay contents gives a wide range in the contents of shrink/swell clays (Table 1). This is also re¯ected in the range of modi®ed linear shrinkage (LSmod) values from 5 to 19% measured
using the method of McKenzie et al. (1994). (LSmodis
measured on air-dry soil ground to <2 mm and then wet to a potential ofÿ100 cm H2O, whereas standard
linear shrinkage is measured on a paste of remoulded soil.) In particular, the Maros and Manaoag soils have similar particle size distributions but have quite dif-ferent clay minerals and shrink/swell potentials.
Ringrose-Voase et al. (1995) found that the soils are mainly neutral with Maros and San Ildefonso showing some acidity in the surface. Electrical conductivity of the 1:5 extracts indicate that none of the soils is saline. The cation exchange capacities vary from about 20 cmol (p) kgÿ1 at Maros to nearly 70 cmol (p) kgÿ1at Ngale. The differences are due to miner-alogy and clay content. However the exchange com-plexes of all the soils are dominated by divalent cations (Ca and Mg) and none are sodic.
2.2. Experiment
The experiment was conducted at Maros in April± May 1993 for a duration of 34 days. At San Ildefonso
and Manaoag it was conducted in January 1994 over periods of 26 and 28 days, respectively, and at Ngale in May±June 1994 for 40 days.
The ®rst experiment at Maros was carried out at the beginning of the dry season during the period when rice would be harvested and dry season crops planted. However, this period proved prone to occasional heavy rain, which considerably slowed drying of the soil. The other experiments were carried out in the middle of the dry season to ensure the maximum possible drying and none received any substantial rain. The Ngale site was covered by a temporary rain shelter as a precaution, but this proved unnecessary.
All sites except Manaoag were prepared several months before the start of the experiment by puddl-ing involvpuddl-ing two animal-drawn ploughpuddl-ings and harrowings and were transplanted with rice, which was nearly ready for harvesting at the start of the experiment. During the rice crop the ¯ood water had receded on several occasions resulting in some consolidation of the puddled layer. The rice was harvested and the soil re¯ooded brie¯y. The following day was considered to be the ®rst day after drainage (DAD).
The Manaoag soil is relatively well drained which meant that ¯ooded conditions could not be maintained
into the middle of the dry season. Consequently, this site had been sown to mungbean after the wet season rice crop. Shortly before the experiment began, the mungbean was harvested and the site was ¯ooded and puddled by two cattle-drawn ploughings and harrow-ings to a depth of about 10 cm. The site was allowed to drain overnight before measurements were started the following morning.
During the experiments the sites were kept free of vegetation by regular weeding and cutting of rice ratoons.
2.3. Measurements
The sites were divided into seven plots (18 m4 m at San Ildefonso; 14 m2 m at Manaoag; 18 m3 m at Maros and 6 m1.7 m at Ngale). The second and ®fth plots were used for crack measurements and the other ®ve for other measurements.
Measurements and sampling were carried out at various intervals over the course of the experiments as follows:
Particle size distributions, proportions (on whole soil basis) of swelling clay minerals (smectite and vermiculite) and modi®ed linear shrinkages (LSmod) for the E3 sitesa
Site Depth (cm) g kgÿ1 Whole soil (%) LS
mod
Sand 50±2000mm Silt 2±50mm Clay <2mm Smectitevermiculite
San Ildefonso (Ustic Epiaquert)
The interval between sampling days was generally shorter at the start of the experiments when conditions were changing rapidly and longer towards the end of the experiments.
2.3.1. Evaporation from the soil surface
Evaporation from the soil surface was measured using a single mini-lysimeter in each of the ®ve plots. These consisted of cores of 10 cm diameter taken from the soil surface in metal rings of 5±10 cm depth. The bases of the cores were sealed using plastic to prevent water movement between the cores and the surround-ing soil. The cores were weighed and placed back in the soil until the next sampling day, when they were recovered and re-weighed to determine the volume of water lost through evaporation. In addition the moist-ure content of the mini-lysimeter was determined. Fresh mini-lysimeters were set up on each sampling day to minimise errors caused by cutting off the ¯ux of water from below.
2.3.2. Surface cracks
Surface cracks were measured in two plots using transects consisting of a series of six linked semicir-cles of 1 m diameter (Ringrose-Voase and Sanidad, 1996). The length of crack per unit area, LA, was
estimated from the number of intercepts with cracks. The depth and width distributions were determined by measuring the ®rst ®ve cracks intercepted by each semi-circle using a ¯exible ruler (i.e. 60 pairs of mea-surements in total). These data were used to estimate mean cross-sectional area, X, assuming a triangular cross-section. The crack volume per unit area,VA, can be estimated as LAX. The crack volume fraction,
VV(d), pro®le can also be calculated from the data.
2.3.3. Soil moisture content and bulk density
On each sampling day soil moisture content, air-®lled porosity and bulk density were measured using a single 10 cm diameter core taken from each of the ®ve plots using a long coring tube and an electric jack hammer. The cores were divided into volumetric samples at depths of 0±5, 5±10, 10±15, 15±20, 25± 30 and 35±40 cm. Care was taken when positioning the corer to avoid any large surface cracks, since it was impossible to sample the crack pattern adequately. Since the volume of cracks was not included, the core samples tended to overestimate the moisture content
and bulk density. The volumetric contents were cor-rected by multiplying by (1ÿVV(d)), whereVV(d) is the volume fraction of cracks in a given layer. For example for volumetric moisture content,y:
yy measured 1ÿVV d (1)
From here on y refers to the volumetric moisture content corrected in this way.
2.3.4. Soil moisture potential
Soil moisture potential was measured using arrays of six tensiometers installed in the centres of the ®ve plots at depths of 0±5, 5±10, 10±15, 15±20, 25±30 and 35±40 cm. These were read by a pressure transducer. When the potential decreased beyond tensiometer range later in the experiment, cores (generally 10 cm diameter5 cm depth) were taken and the matric potential measured in the laboratory using the ®lter paper method (Greacen et al., 1989).
The moisture retention characteristic was also determined in the laboratory on ®ve replicate cores (100 mm diameter50 mm depth) from the 0±10, 15± 20 and 30±35 cm layers at each site using a combina-tion of tension and pressure plates (McIntyre, 1974). The relationship between gravimetric moisture con-tent and pocon-tential was summarised by ®tting a Camp-bell (1985) function with Hutson and Cass (1987) smoothing to the data in a two stage process. This was necessary because the ®eld data were too noisy to determine the shape of the curve, but the degree of saturation in the ®eld was different to that in the laboratory. First the function was ®tted to the labora-tory data to determine the shape of the curve (i.e. the slope,b, and air-entry potential,ce). Next the function
was ®tted to the ®eld data to determine the position of the curve (i.e. the saturated moisture content, ygsat)
using ®xed values ofbandcedetermined in the ®rst
stage.
2.3.5. Soil strength
Soil strength was measured using a Rimik recording penetrometer at 1.5 cm depth intervals to 50 cm depth. Measurements were made in triplicate within each of the ®ve plots (i.e. 15 replicates in total).
2.3.6. Plot layout
Within each plot the volumetric samples, mini-lysimeters and soil strength measurements were all
taken in close proximity to each other. These measure-ments are destructive and cause substantial distur-bance in their proximity, especially in soft, puddled soil. In order not to disturb parts of the plots reserved for later sampling, they were not located randomly within each plot on any one day. Instead they were located near one end of the plot on the ®rst sampling day and then progressed across the plot on subsequent sampling days. However, to ensure some randomisa-tion, the locations for the ®rst day were alternated between opposite end of the plots. This sampling strategy also meant that tensiometer measurements, which were located in the centre of each plot, were not necessarily more closely related to the other measure-ments in the same plot than to those in the other plots. As a result, the hydraulic conductivity calculations (see below) were made using values for each para-meter which had been averaged for the ®ve replicates. Hence the calculations were performed only once for the whole experiment, rather separately for each plot, which would have given some estimate of variability.
2.3.7. Hydraulic conductivity
Unsaturated hydraulic conductivities relative to the solid material, km(c), were estimated for each site
using a modi®cation of the instantaneous pro®le method (e.g. Hillel, 1980, pp. 217±221, 227±232). The method was modi®ed by allowing evaporation from the surface and measuring the amount of eva-poration using mini-lysimeters, instead of preventing evaporation and setting the ¯ux at the surface to zero. In addition, calculations were made relative to mate-rial depth (depth of solid) rather than actual depth to allow for changes in the depth of each soil layer due to shrinkage. The calculations are given in the Appendix A.
3. Results and discussion
3.1. Bulk density
The bulk density (corrected for cracks) changed over time especially in the surface layers (Fig. 1). The expectation was for bulk density to increase (i.e. volume decrease) as the soil dried due to both the shrinkage of shrink/swell clay minerals and to con-solidation of material previously unconsolidated by
puddling. However, in most cases the bulk density appeared to decrease with drying rather than increase. This apparent volume expansion was most probably caused by greater compression occurring during sam-pling immediately after draining when the soil was wet and soft. The soil appeared to go through three distinct stages with respect to sampling. On the ®rst sampling, the soil was extremely soft and it was dif®cult to keep cores within the coring tube as they were removed from the ground. It was likely that the cores were compressed during extraction from the coring tube, especially at Manaoag where the soil was freshly puddled. On one or two subsequent sam-pling occasions, the soil was ®rmer and the cores easier to remove from the ground. However, the soil was generally sticky and was probably compressed both during insertion of the coring tube into the ground and during extraction of the cores. For the remainder of the experiment, the soil was more plastic and tube insertion and core extraction were much easier. During this period compression was probably minimal.
Sample compression appears to be almost unavoid-able in puddled soils so any measurements relying on volumetric samples must be treated with suspicion, i.e. volumetric moisture content, bulk density, porosity and air-®lled porosity. Nevertheless, it was still pos-sible to analyse moisture content changes by assuming that compression mainly affected the gaseous compo-nent of the soil leaving the quantity of water relative to a unit quantity of solid (i.e. gravimetric moisture content) largely unchanged. Water movement was tracked as a function of the soil particles or material depth rather than the actual depth. Indeed such an analysis is necessary whether volume change is caused by the degree of compression during sampling varying on different sampling dates or by shrinkage due to drying.
3.2. Moisture retention characteristic
The Campbell function parameters describing the moisture retention in the soils are shown in Table 2. In most cases the air-entry potential, at which moisture content starts to decrease more rapidly, is quite large (<ÿ100 cm H2O), especially in the puddled layer.
macropores by puddling. The only exceptions are the sub-soils at Manaoag and Maros.
Plant available water held at potentials between
ÿ100 cm H2O (approximately ®eld capacity) and ÿ15,000 cm H2O (approximately permanent wilting
point) was greatest in the puddled layer at each site. In the sub-soils, plant available water was particularly high at Ngale because of its high clay content. It was least at Manaoag and Maros where drainable water was greatest.
3.3. Hydraulic conductivity
Since hydraulic conductivity, km(c) (Fig. 2), was
calculated as a function of material depth, problems associated with volume change during sampling were avoided. For ease of comparison, functions ®tted to the data for each site are shown in Fig. 3. San Ildefonso has the smallest saturated hydraulic conductivity (0.5 mm per day) and has the smallest conductivity over much of the matric potential range. Manaoag has the greatest unsaturated conductivity which accords with experience of the site being very well draining.
Unfortunately, its saturated conductivity could not be estimated from the experimental data. Maros and Ngale have similar saturated conductivities of 3.0 and 3.5 mm per day but differ in the manner in which
km(c) decreases with potential. The relative
magni-tudes of the hydraulic conductivities appear to have little relation to the drainable water content shown in Table 2.
3.4. Evaporation
A concern with mini-lysimeters is the effect of cutting off upward ¯uxes from underlying soil. This results in soil in the mini-lysimeter becoming drier than the surrounding soil that it is meant to represent, thereby underestimating the rate of evaporation. Attempts were made to minimise the discrepancy in moisture contents by taking fresh mini-lysimeters on each sampling day. Each line segment in Fig. 4 shows the mean moisture content of the mini-lysimeters at the start and ®nish of a measurement period of several days. The discrepancy between the moisture content at the end of one period and the start of the next was
Table 2
Campbell (1985) function parameters describing the moisture retention characteristica
Site Depth
0±10 0.467 ÿ106 6.56 0.011 0.237
10±20 0.308 ÿ80 9.96 0.009 0.118
25±40 0.344 ÿ162 6.95 0.003 0.162
Manaoag
0±10 0.460 ÿ134 7.73 0.006 0.205
10±20 0.430 ÿ78 10.38 0.012 0.159
25±40 0.420 ÿ17 14.70 0.048 0.108
Maros
0±10 0.473 ÿ140 8.74 0.005 0.192
10±20 0.377 ÿ37 13.36 0.027 0.110
25±40 0.355 ÿ30 15.74 0.026 0.090
Ngale
0±10 0.724 ÿ193 5.61 0.006 0.385
10±20 0.692 ÿ272 5.49 0.003 0.357
25±40 0.693 ÿ231 6.34 0.004 0.331
aSaturated moisture content,y
gsat; air-entry potential,ce; slope,b.
bDrainable water content is calculated as the difference between that at the water contents at saturation andÿ100 cm H
2O, using the
Campbell function with Hutson and Cass smoothing.
Fig. 2. Hydraulic conductivity (in material space) Ð matric potential relationships at the four sites. Lines show functions ®tted to the data for all depths. Two dry points at Ngale were considered outliers and excluded from the ®t.
caused by two factors. The ®rst was cutting off the ¯ux at the bottom of the mini-lysimeter. The second was differences in the material depths of the mini-lysi-meters due to shrinkage of the soil or differences in the lengths of the rings used. Shrinkage will increase the material depth of soil in the mini-lysimeter for the next period assuming the ring is inserted to the same depth. If material depth increased between measurement periods it is likely to create a discrepancy in the moisture contents because the extra soil included for the second period would generally be wetter due to the steep moisture content gradients that develop as the soil dries. The horizontal bars in Fig. 4 show particularly large moisture content discrepancies occurred when there were large changes in material depths. However, the majority of the discrepancies were probably a result of cutting off the ¯ux at the bottom of the mini-lysimeter, resulting in an under-estimate of evaporation. The discrepancies were gen-erally fairly small at San Ildenfonso and Ngale, moderate at Manaoag and quite large at Maros.
Notwithstanding possible errors, Fig. 5 shows the rate of evaporation,Ejÿ1!j/(tjÿtjÿ1), from the surface
as measured using the mini-lysimeters, whereEjÿ1!j
is the net evaporation between samplingsjÿ1 andjat timestjÿ1andtj. The rate of evaporation declined as
the surface soil dried and water movement from below was restricted by low hydraulic conductivity. Mana-oag and Maros generally had higher rates than San Ildefonso and Ngale. The differences in magnitude were at least partially driven by greater evaporative demand at Manaoag and Maros. In addition, rain at Maros frequently rewet the surface and slowed the
reduction in soil evaporation due to drying. However, differences in hydraulic conductivity also had an effect on evaporation as the soil dried. The very low rates at Ngale were partially because its conduc-tivity declines very rapidly once the soil dries past a potential ofÿ50 cm. Conversely, Manaoag maintains a greater conductivity as it dries and can supply water for a greater evaporation rate.
3.5. Moisture loss
The change in the moisture contents,yg, at each site
are shown in Fig. 6. Fig. 7 shows the ¯uxes through the pro®le,qi,jÿ1!j, over time calculated using Eq. (A.6).
At the start of the experiment the soil was driest at Maros. However, heavy rain the day before the second sampling 9 DAD wetted the pro®le. Subsequent rain, including a heavy shower 25 DAD meant that drying was very slow. This made it dif®cult to determine the ¯ux at the surface, necessitating the omission ofkm(c) estimates calculated using such periods.
There was no signi®cant rain at the other sites during the experiments resulting in much greater drying of the surface layers. Fluxes at these sites generally followed a similar pattern, with relatively large downward ¯uxes as the soil returned to ®eld capacity immediately after re-¯ooding. However, even in this early period, there was a zero-¯ux plane a few centimetres below the surface as water in the surface soil was removed by evaporation. This zero-¯ux plane moved down the pro®les very rapidly as the pro®les drained. This was followed by a period of moderate upward ¯uxes supplying the evaporative demand at the surface. The upward ¯uxes then declined as removal of water decreased the matric potential and
km(c).
At San Ildefonso and Manaoag the water used to re-¯ood the sites drained from the upper 40 cm between 1 and 5 DAD. Thereafter, there was less water loss from the sub-surface layers, presumably because they had reached ®eld capacity. Drying occurred within the upper 10 cm at San Ildefonso and 20 cm at Manaoag. At Ngale, drying was restricted to the upper 10 cm and virtually none occurred below this depth, despite the experiment being run for 40 days. In general, the depth to which drying occurred was greatest in soils with the greatest unsaturated hydraulic conductivity. Below this depth, the sub-soils remained near ®eld capacity.
Fig. 4. Gravimetric moisture content of the mini-lysimeters used to estimate evaporation between different days after drainage. Horizontal bars show the material depth (right axis) of the mini-lysimeters used for each interval. 95% CI: 95% con®dence interval.
Fig. 6. Moisture content pro®les on different days after drainage (DAD). 95% CI: 95% con®dence interval. Dotted lines (from left): moisture contents at potentials ofÿ15,000 andÿ100 cm water and saturation derived from the moisture characteristic (in some cases the latter two may be almost superimposed).
3.6. Crack development
At all sites except Maros, the length of crack per unit area,LA, initially rose rapidly to a maximum of 30±35 mmÿ2(Fig. 8). The freshly puddled nature of the soil at Manaoag meant that there were no cracks to start with, so the maximum was reached later. The very rapid increase in crack length at Ngale was probably due to the self-mulching nature of the Ngale soil. Self-mulching is `the tendency to form a loose, granular (dry) surface mulch as the result of freezing and thawing, or wetting and drying' (Soil Survey Staff, 1960). After reaching a maximum the length of cracks decreased slightly as some cracks closed. This phe-nomenon was also observed by Hallaire (1984) and ourselves on earlier occasions (unpublished data).
The rate of increase in crack volume,VV, however, was much faster at Manaoag than at the other sites
because water loss was rapid (Fig. 9). The total crack volume reached was also different at different sites. Crack volume depends upon the extent of drying and its interaction with the proportion of clay minerals with shrink/swell properties and the potential of the puddled soil to consolidate. Hence at Maros, there was little development of cracks because there was little drying. At the other sites the relative roles of clay mineralogy and consolidation can be investiga-ted by comparing the relative magnitudes of the actual volume of cracks generated per unit loss of water in the 0±5 cm layer,DVV/Dyg(as measured in the ®eld)
with the volume of cracks that would be expected on the basis of LSmod expressed on an area basis
(2LSmod±LSmod2) between a potential of ÿ100 cm
of water and oven dry (Table 3). AlthoughDVV/Dyg
and LSmod are not strictly equivalent since the
former is per unit loss of moisture content and the
Fig. 8. Development of crack length at the four sites: (&) San Ildefonso; (^) Manaoag; (~) Maros; (*) Ngale.
Fig. 9. Development of crack volume at the four sites: (&) San Ildefonso; (^) Manaoag; (~) Maros; (*) Ngale.
Table 3
Shrinkage characteristics of the 0±5 cm layera
Site Dyg DVV/Dyg(A) Crack volume calculated as areal
shrinkage2 LSmod±LSmod2(B)
A/B
San Ildefonso 0.395 0.282 (r20.97) 0.137 2.06
Manaoag 0.442 0.638 (r20.86) 0.183 3.49
Maros 0.255 0.141 (r20.87) 0.098 1.45
Ngale 0.544 0.499 (r20.85) 0.347 1.44
aDy
g(g gÿ1) is the change in gravimetric moisture content over the experiment.DVV/Dygis the change in crack volume per unit loss of
moisture calculated as the regression coef®cient obtained by regressing crack volume (m3mÿ3) against loss of moisture content (g gÿ1) over
the course of the experiment. Areal shrinkage is the area of cracks expected from the modi®ed linear shrinkage (LSmod, see Table 1). The ratio
of these two indicates how much of the cracking measured in the ®eld is explained by LSmod.
latter over a ®xed matric potential range, the ratios of the two parameters indicate whether differences in cracking between sites is accounted for by the differences in LSmod. Since LSmod is largely
depen-dent on the proportion of shrink/swell clay (Table 1), small ratios will occur where shrink/swell clays played a greater role relative to other factors such as consolidation. The highest ratio (3.49) was found at Manaoag indicating that consolidation of the puddled layer played a greater role in shrinkage than at the other sites, which is consistent with the fact that the site was freshly puddled and had no opportunity to dry and consolidate before the experiment started. The low ratios (1.45) at Maros and Ngale suggest that the proportion of shrink/ swell clay was relatively more important at these sites than at the others. Consolidation was less important at these sites probably because the puddled layer had both settled and been subjected to periods of drying during the rice crop. The intermediate ratio for San Ildefonso suggests that less consolidation had occurred during the preceding rice crop than at Maros and Ngale.
Thus the small volume of cracks at Maros was due to lack of drying; a small proportion of shrink/swell clays and a low potential for consolidation presumably because consolidation had occurred beforehand dur-ing the rice crop. Despite a lack of potential for consolidation at Ngale, there was a large volume of cracks due to a greater loss of water combined with a high proportion of shrink/swell clays. The large volume of cracks at Manaoag was probably due to reasonably high water loss combined with intermedi-ate LSmod and a high potential for consolidation. At
San Ildefonso moderate drying combined with lower LSmodand some consolidation to give an intermediate
volume of cracks.
3.7. Air-®lled porosity (inter-crack)
The air-®lled porosity (AFP) discussed here includes only the AFP between the cracks. Changes in AFP are shown in Fig. 10. As discussed above, AFP values must be interpreted with care because compression during sampling would tend to affect AFP more than other soil components and would be greatest at the start of the experiment when the soil was softest.
Except at Manaoag, which had been freshly puddled, there was reasonable AFP in the surface layers at the start of the experiment indicating that ®eld saturation was probably incomplete. The increase in AFP as moisture content decreased was tempered by increases in the crack volume. Thus at Ngale and Manaoag, which showed the greatest increase in crack volume per unit loss of moisture content (Table 3) there was correspondingly less increase in AFP. At San Ildefonso where there was less crack development, there was greatest development in AFP.
3.8. Soil strength
Fig. 11 shows the penetration resistance pro®les at the four sites as the pro®les dried. The strengths of the surface layers increased as they dried. The strength of the upper 2 cm was usually weaker than immediately below. This was because of the formation of a dry crust about 2 cm deep that cracked extensively. Although the solid material between the cracks was probably stronger than that below, the cracking reduced the penetration resistance. In the surface layers, Ngale and Manaoag tended to be stronger than San Ildefonso and Maros.
Below 15 cm depth, strength changed slowly because of the slow rate of drying. At Manaoag the sub-soil tended to be stronger, because of the degree to which the pro®le had dried before it was re-¯ooded for the experiment. Ngale tended to be weakest in the sub-soil. The only pro®le to show any evidence of a plough pan was Maros, where there was an increase in strength at about 15 cm depth. In none of the soils did the strength of the sub-soil increase to more than 2 MPa, which is widely quoted as limiting root growth. Given that the sub-soils only dry very slowly in the absence of roots to extract water, it is likely that their strength will remain low until after roots have penetrated a layer. Hence, it is possible that sub-soil strength may not be limiting for post-rice crops.
3.8.1. Relationship between strength and moisture content Ð puddled layer
relationship between penetration resistance (PR) and moisture content (yg) for the surface soil at the four
sites. The relationships are approximately linear (although at Maros there is some evidence that it may be curvilinear) so a regression equation of the form PRÿmygcwas ®tted to the data for each site.
The regression analyses are shown in Table 4. The ®ts were highly signi®cant as shown by the F-statistic probabilities and account for 75±83% of the variation in penetration resistance.
San Ildefonso had lower penetration resistances than the other sites over most of the moisture content range with smaller values of bothcandm. Ngale was stronger than the other sites at most moisture contents. However, themvalue (slope) for Maros was greater than that for Ngale and Manaoag so that its strength was similar to that at Ngale at low water contents (i.e. itscvalue is similar to Ngale). The greater strength of the Ngale soil at a given moisture content was because
the soil has a large amount of tightly held water due to its high smectite content.
3.8.2. Relationship between penetration resistance and moisture content Ð sub-soil
Unfortunately, the lack of sub-soil drying during the experiment meant that penetration resistance was only measured over a small range of moisture con-tents. The regression analyses accounted for less than 50% of the variation in resistance and were mostly insigni®cant.
In summary, differences in the penetration resis-tances at various depths in the four soils result from an interaction of the different relationships between strength and moisture content at each site and the different rates at which the pro®les dry. The latter itself results from the interaction between moisture content, hydraulic potential and hydraulic conductiv-ity, which are also different in each soil. The regres-sion analyses probably allow prediction of penetration resistance from moisture content at least in the puddled layer. Therefore a reasonable model of the water balance for the soils could be used to predict soil strength as well.
4. Conclusion
Detailed measurements of the physical conditions in puddled soil as it dried after ®eld drainage showed several features common to all four sites:
The hydraulic conductivities of the soils were all relatively small and never exceeded 10 mm per day even near saturation. As the water content
Fig. 12. Regressions of penetration resistance for the 0±5 and 5± 10 cm layers against moisture content. Symbols are data points, lines are ®tted by linear regression as in Table 4.
Table 4
Regression analyses of the form PRÿmygcfor the 0±5 cm and 5±10 cm layers of penetration resistanceaagainst moisture contentbfor the
four sites
Site m(S.E.) c(S.E.) r2 F-statistic probability
San Ildefonso 0±10 cm 2459 (324) 1541 (126) 0.827 6.510ÿ6
Manaoag 0±10 cm 3860 (502) 2238 (194) 0.831 5.610ÿ6
Maros 0±10 cm 4412 (677) 2525 (311) 0.752 1.410ÿ5
Ngale 0±10 cm 3152 (423) 2583 (273) 0.776 1.410ÿ6
aPR, means of 3 depths/layer3 replicates/plot5 plots. by
g, means of 5 replicate plots.
decreased, hydraulic conductivity fell to very low values.
During the first few days water movement in the profile was downwards. However, a zero-flux plane rapidly developed and moved down the profile so that most of the time water movement in the top 40 cm was upwards, driven by evaporation from the surface.
Once the surface layer had dried, water loss by evaporation was limited by the falling hydraulic conductivity. The surface layers dried rapidly, with soils with greater unsaturated hydraulic conductiv-ity tending to dry to greater depths. Sub-soils tended to remain near field capacity.
Drying of the surface layers was accompanied by cracking and an increase in strength. The amount of cracking was controlled by the content of shrink/ swell clay and the potential for consolidation of the puddled layer. Drying of the puddled layer earlier in the rice crop was presumably accompanied by consolidation, reducing the potential for consolida-tion during the drying experiment.
The slow rate of drying of the sub-soil meant that sub-soil strength remained low. It is unlikely to become limiting to root growth until after roots have already penetrated a layer and begun to extract water.
The low hydraulic conductivities encountered sug-gest that waterlogging after heavy rain could cause problems during crop establishment. For example, the risk of a late typhoon in the early part of the dry season is high in some parts of the Philippines. On the other hand, in the absence of rain, the availability of moist-ure for post-rice DS cropping is likely to be controlled more by the available water capacity of the sub-soil than by drying. This should give a measure of ¯ex-ibility for sowing dates, without the need to sow rapidly after rice harvest. If the plant available moist-ure capacity of the sub-soil is suf®cient for a crop, sowing can take place when there is suf®cient surface moisture for crop establishment. This could be when the soil is still moist immediately after rice harvest, or later when there has been suf®cient rainfall to rewet the surface layer. This might allow sowing to be delayed to periods in the dry season when the prob-ability of heavy rain is less or when other resources such as labour are more available.
These data will allow soil water and strength to be modelled for a range of soil/climate combinations. This will help to show where post-rice legumes can be established with a reasonable probability of success and allow different species and sowing dates to be tested. To do this it will be crucial to have better data on the critical soil physical conditions for suc-cessful establishment of DS legume species in terms of the length of time for which they can tolerate waterlogging; the driest soil they are able to tolerate and the greatest soil strength their roots are able to penetrate.
Acknowledgements
The authors thank the Australian Centre for Inter-national Agricultural Research (ACIAR) for funding this project and the following:
Bureau of Soils and Water Management, Philip-pines: Messrs Oscar Manaois, Ulyses Latoza and Jojo Ponce for help with ®eld work and Director Godofredo Alcasid, Dr Perfecto Evangelista and Mrs Esperanza Dacanay for their support of the work.
Mr and Mrs Severo Arellano for use of their ®eld (Manaoag) and for accommodation at their house.
Research Institute for Maize and Other Non-rice Cereal Crops, Maros, Indonesia: Messrs Mustari Basir, Rahmat Hanif Anasiru and members of the Agricultural Engineering Department for help with ®eld work and Drs A. Hasanuddin and Bambang Prastowo for supporting the work.
Research Institute for Legume and Tuber Crops, Malang, Indonesia: Mr and Mrs Djarwo for providing accommodation and meals at the Ngale site; Messrs Prakoso and Suharto for help with ®eld work and Dr Titis Adisarwanto for supporting the work.
Brawijaya University, Malang, Indonesia: Messrs Daniel Budiyono, Udji Dewantoro, Sugeng Riyadi, Riyanto and Budi Santoso for help with ®eld work.
Appendix A. Calculation of hydraulic conductivity
As discussed above, the layout of the experiment meant that it was not possible to perform the calcula-tions for each plot independently in order to gain an estimate of variability. Instead the calculations were performed on the mean bulk densities, moisture con-tents and matric potentials for each sampling interval on each sampling day. Values for unsampled layers (20±25 and 30±35 cm) were estimated by linear inter-polation between the adjacent layers. The bulk density and moisture content of the layer below 40 cm depth was assumed to be the same as the 35±40 cm layer. Bulk densities and moisture contents were corrected for the volume of unsampled cracks using Eq. (1). In addition, mean evaporation from the soil surface was calculated from the mini-lysimeters.
In rigid soils, ¯uxes are calculated at ®xed points in the soil, usually the boundaries between sampling layers. In soils that shrink and swell such points move as the soil shrinks and swells. This means that a point coinciding with the boundary between two sampling layers on one day will not necessarily coincide with that boundary on another day when the moisture content has changed, assuming that sampling is car-ried out at the same intervals down the pro®le. In order to measure ¯uxes at points ®xed relative to the solid component, it is necessary to de®ne material layers. These layers have a constant thickness of solid or material thickness,m. The boundaries between mate-rial layers occur at ®xed matemate-rial depths,M, which are equal to the cumulative material thicknesses of the overlying layers (i.e. the depth of solid between the surface and the boundary). The real depths,Z, of the boundaries will vary as the overlying layers change thickness. Here the material layers were de®ned as being the same as the sampling layers on the ®rst sampling date, i.e. the boundary under thekth material layer has the same material depth,Mk,1, and real depth, Zk,1, as theith sampling layer,Mi,1andZi,1, whereki.
On a subsequent sampling,j, thekth material bound-ary will have the same material depth (Mk,jMk,1) but
a different real depth (Zk,j6Zk,1). Conversely, the ith
sampling boundary will have the same real depth (Zi,jZi,1) but different material depth (Mi,j6Mi,1).
The real depths of the material boundaries,Zk,j, were
tracked using the following steps.
The material thickness, mi,j, of each sampling
layer, i, on each sampling day,j, was calculated from the bulk density of the layer,rbi;j, the particle density,rs, and the thickness,z, which in this case
was 5 cm
mi;jz
rbi;j
rs (A.1)
The material depth of the boundary under theith sampling layer,Mi,j, was the cumulative material
thickness of all the sampling layers above it
Mi;j
Xi
I1
mI;j (A.2)
The sampling layer, i, in which the kth material boundary occurs on thejth sampling day was found by comparingMk,1 with the values ofMi,j
calcu-lated in Eq. (A.2) for thesamplingboundaries (i.e. the sampling layer for which Miÿ1,j<Mk,1<Mi,j).
The real depth,Zk,j, of the material boundary was
calculated by linear interpolation between the sam-pling boundaries above and below it
Zk;jZiÿ1;1 Zi;1ÿZiÿ1;1
surface and the lower boundary of eachsampling
layer,i, on each sampling date,j, was calculated as
Vi;j
X
i
I1
yI;jz (A.4)
whereyI,jis the volumetric water content in layerI
on sampling datejandzis the sampling interval, in this case 50 mm.
The cumulative moisture content,Vk,j, above each
material boundary,k, on each sampling day,j, was obtained by linear interpolation between Viÿ1,j
and Vi,j for the sampling boundaries above and
below it
where i is the sampling layer in which the kth material boundary currently lies.
The flux of water, qk,jÿ1!j, across the kth
material boundary between sampling dates jÿ1 andjwas
poration from the soil surface over the same period (measured using the mini-lysimeters) andtjÿ1andtj
are the number of days since drainage on sampling datesjÿ1 andj. Note that positive fluxes indicate downward movement and negative fluxes upward movement.
The fluxes calculated relate to the period between sampling days. The flux on a sampling day, qk,j,
was calculated by linearly interpolating between the fluxes of the sampling periods before and after (i.e.qk,jÿ1!jandqk,j!j1) assuming these relate to
Clearly,qk,jcould not be calculated for the first and
last sampling days.
It was assumed that the tensiometers moved with the material layers so that the measured matric potentials, ck,j, referred the mid-points of the
material layers. The matric potential on the kth material boundary,ck0.5,j, is calculated by linear
interpolation between the measured potentials above, ck,j, and below, ck1,j, at depths of
The total potential (matricgravitational),Ck,j, at
the mid-point of thekth layer on thejth sampling was
The potential gradient in material space, (dC/ dM)k,j, across the lower boundary of the kth
material layer on thejth sampling date was dC
dMk;j
2 Ck;jÿCk1;j Mk1;jÿMkÿ1;j
(A.10)
Positive values indicate that potential decreases with depth (i.e. water will move downwards).
The hydraulic conductivity in material space,
km(c)k,j, at the bottom of the kth material layer
on thejth sampling date was calculated as
km ck;j qk;j
dC=dZk;j (A.11)
wherecis the matric potential,ck0.5,j, at thekth
material boundary on thejth sampling date from Eq. (A.8). Negativekmvalues sometimes resulted
where the direction of the flux and the potential gradient were opposite. This was caused by field variation because moisture content and potential were not measured in the same location. Such values were ignored. Note thatkmin material space
is related tokin real space:
km c k c
rb
rs (A.12)
References
Campbell, G.S., 1985. Soil Physics with BASIC. Elsevier, New York. De Datta, S.K., 1982. Principles and Practice of Rice Production.
Wiley, New York, 618 pp.
Greacen, E.L., Walker, G.R., Cook, P.G., 1989. Procedure for the ®lter paper method of measuring soil water suction. CSIRO (Australia) Division of Soils, Divisional Report No. 108. Hallaire, V., 1984. Evolution of crack networks during shrinkage of
a clay soil under grass and winter wheat crops. In: Bouma, J., Raats, P.A.C. (Eds.), Proceedings of the ISSS Symposium on Water and Solute Movement in Heavy Clay Soils. International Institute for Land Reclamation and Improvement Publication 37, Wageningen, pp. 49±53.
Hillel, D., 1980. Fundamentals of Soil Physics. Academic Press, New York.
Hutson, J.L., Cass, A., 1987. A retentivity function for use in soil-water simulation models. J. Soil Sci. 38, 105±113.
McIntyre, D.S., 1974. Water retention and the moisture character-istic. In: Loveday, J. (Ed.), Methods for Analysis of Irrigated Soils. Commonwealth Bureau of Soils Technical Communica-tion 54. Commonwealth Agricultural Bureaux, Farnham Royal, UK, pp. 43±62.
Ringrose-Voase, A.J., Sanidad, W.B., 1996. A method for measuring the development of surface cracks in soils: application to crack development after lowland rice. Geoderma 71, 245±261.
Ringrose-Voase, A.J., Hutka, J., Beatty, J., Raven, M.D., Rath, H., 1995. Analyses of some Indonesian and Philippine soils for ACIAR project 8938 on `Management of clay soils for lowland, rice-based cropping systems'. CSIRO (Australia) Division of Soils Technical Report 11/1995.
Schafer, B.M., Kirchhof, G., 2000. The soil and climate characterisation of benchmark sites for lowland rice-based cropping systems in the Philippines and Indonesia. Soil Tillage Res. 56, 15±35.
So, H.B., Ringrose-Voase, A.J., 2000. Management of clay soils for rainfed lowland rice-based cropping systems: an overview. Soil Tillage Res. 56, 3±14.
Soil Survey Staff, 1960. Soil classi®cation: a comprehensive system, the 7th approximation. US Department Agriculture, Washington, DC.
