Abstract: Responsiveness of soils to volume change at all moisture statuses affect infrastructural development of soils. Volume Shrinkage (VS) was estimated from Coefficient of Linear Extensibility (COLE) values, using 150 soil samples from 30 soil profile pits in southeastern Nigeria. The objective of this study was to use soil properties related to shrinkage in predicting and monitoring volume changes in these soils in southeastern region of Nigeria. The value of liquid limit (26-30%) was higher than that of plastic limit (11-13%). Plasticity index was lower (16-17%) in the upper soil layer (0-40 cm) than in the lower horizons (41-80 cm layer) by 24%. Volume shrinkage value (11.20-11.80%) was least at the topmost horizon (0-40 cm depth) and increased to between 14-18% in the lower part of the profile (40-100 cm). Soils at all depths showed low to moderate shrink-swell properties having coefficient of linear extensibility of 0.036-0.057 cm cm-1 with middle layers showing higher shrink-swell characteristics. These results show moderate shrinkage (VS = 10-20) although intra-pedal variability was prominent. Multiple regression analysis showed good predictive ability (R2 = 0.76): p = l.05; n = 150) of Atterbergs limits, texture, bulk density and organic matter (OM) in determining shrinkage properties. Accuracy of prediction was defined by a low Root Mean Square Error (RMSE) value of 0.01 while the model overestimated slightly (bias±0.00001).
INTRODUCTION
Models of agro-ecological systems provide means of prediction beyond the bounds of experience or experimentation. Extrapolation may involve simulation of the expected behaviour for different climate and/or soil conditions, perhaps with alternative management or for a longer time scale than observed experimentally (Probert and Keating, 2000). Models in soil studies are used to collect data in a systematic manner, describe system or to predict relationships and behaviour in order to extend our knowledge of soils. They represent a series of approximations towards the truth. Soil landscape models provide descriptions of the gradually varying contiguity of soil properties in the landscape. Soil landscape models hypothesize that there is a close relationship between soil attributes and landscape position (Pennock et al., 1987; Irvin, 1996). Soil modelling is applied in pedogenesis (Hoosbeek and Bryant, 1994; Suarez and Goldberg, 1994) and the development of quantitative pedogenic process models have improved our understanding of pedogenic and geomorphic processes, especially with advances in geostatistics and computer power. Gertsev and Gertseva (2004) observed that homomorphic models are imperfect representation of reality in ecological models. Ecological models are classified into two, namely isomorphic and homomorphic models based on degree of complexity with the latter describing relationships in soil groups as a whole and not as individual phenomenon.
Shrinkage properties of soils guide in estimating the strength of soil resources in terms of its ability to sustain farm buildings, farm structures, farm road and general agricultural mechanization and engineering. Swelling and shrinking clay soils change in volume with water content changes. These volume changes depend on the amount and type of clay minerals and are characterized by their magnitude and geometry. They result in the occurrence of shrinkage cracks and surface subsidence (Cornelis et al., 2006; MacNabb and Boersma, 1996) and this influences sustained use of soils. Now, it is common to hear of cracked and collapsed agricultural and non-agricultural engineering structures in southern Nigeria hence Igwe (2003) called for the inclusion of shrink-swell potential of soils in soil survey reports.
In Southeastern Nigeria, soils are formed from six major parent materials, namely Coastal Plain Sands, Shale, Upper Coal Measures, Lower Coal Measures, False-bedded Sandstones and Alluvium. Although the pedogenic processes in the area may seem to be similar, wider variations in soils including volume shrinkage properties vary as a result of differences in parent materials of soils. For instance, soils having greater clay content may have greater shrinkage capacity, implying that sandy soils have lower shrinkage values.
A number of researchers (Igwe, 2003; Mbagwu and Abeh, 1998) have associated the soil moisture volume changes with shrink-swell phenomena to changes in water content, clay content, mineralogy and type of cation present on the soil exchange site, soil structure, organic matter, density, sesquioxides, overburden pressure and interactions among these properties. In addition to this, bulk density influences shrinkage behaviour of soils. Igwe (2003) postulated that soils in which montmorillonite was a major component had a wide range of COLE values, indicating that differences in clay content may be the primary factor controlling the degree of shrinkage. On the other hand, soils having equal amounts of kaolinite and montmorillonite behave like montmorillonitic soils. Although shrink-swell potential is recommended for inclusion in soil survey reports, this important parameter is absent in all existing survey reports for areas where very intensive agricultural operation is on-going in the study area. Thomas et al. (2000) showed that the shrink- swell potential in kaolinitic and mixed mineralogy soils and acid montmorillonitic soils is often more difficult to predict. Presently, no one method of soil analysis estimates shrink-swell potential accurately for all soils. Soil scientists recognize that shrink-swell behavior can best be predicted by examining a combination of physical, chemical and mineralogical soil properties. A protocol that integrates these properties and then establishes a shrink-swell model that can be extrapolated across the same or similar parent materials is needed.
The objective of this study was to use soil properties related to shrinkage [Plastic Index (PI), Total Clay (TC), Total Sand (TS), Organic Matter (OM) and Bulk Density (BD)] in predicting and monitoring volume changes (represented by Volume Shrinkage (VS)) in these soils in southeastern region of Nigeria.
An understanding of the relationship between soil properties and swell-shrink characteristics of soils will aid a priori prediction of the effects of these soil properties on building foundations, septic tank, subsurface absorption systems, roads, dams and other structures in contact with the soil. Structural damage to homes and agricultural structures (i.e., walls, irrigation channels and foundations) due to expansive soils is costly to repair and may be somewhat avoidable if soil properties, such as clay content and the Coefficient of Linear Extensibility (COLE), are investigated (Vaught et al., 2006).
MATERIALS AND METHODS
Description of Study Site
The study was conducted at the floodplains of Otamiri River in Federal
University of Technology, Owerri in southeastern Nigeria during the 2004 rainy
season. The area is located between latitudes 5°30' and 6°00'N and longitudes
7°00' and 7°30'E. Soils of the study site are derived from alluvial
deposits and are below 50 m above sea level. The topography is almost flat having
a slope gradient of between 0 and 2%. Six soil groups classified based on their
parent material, namely Coastal Plain Sands, Alluvium, Shale, Lower Coal Measures,
Upper Coal Measures and False-bedded Sandstones were identified at the site.
The mean annual rainfall is 2250 mm with an annual temperature range of 26-29°C.
The soil has isohyperthermic soil temperature regime (Anikwe et al.,
2003).
Field Studies
A free survey method was used (observation points that are representative
of the site are chosen by the surveyors based on personal judgment and experience)
(Mulla and McBratney, 2000). Auger and core samples were collected from 150
pedogenic horizons of 30 soil profile pits dug at the site. Five profile pits
were dug on each of the 6 soil groups classified based on their parent material,
namely Coastal Plain Sands, Alluvium, Shale, Lower Coal Measures, Upper Coal
Measures, False-bedded Sandstones. Core samples (for determining soil dry bulk
density) were collected using open-faced coring tube (area, 19.5 cm3
and height, 5 cm from Eijkelkamp Agrisearch Equipment) at the selected depths.
Auger samples were collected at fixed depths of 0-20, 21-40, 41-60, 61-80 and
81-100 cm using a hand-pushed auger (Push Probe, 23 mm diameter). For auger
samples visible roots, twigs and leaves were manually removed in the screen
house and the samples air-dried at ambient temperature for 72 h and sieved (using
2 mm sieves). Core samples were analyzed and mean results from each depth used
whereas the auger samples (collected from each depth) were mixed and composite
sub-samples (from each depth) used for the analyses. All soil profile pits were
geo-referenced using a Global Positioning System (GPS) (Garmin Ltd. Kansas,
USA).
Laboratory Analyses
Particle size distribution of less than 2 mm fine earth fractions was measured
by hydrometer method as described by Gee and Or (2002). Bulk density was estimated
by the method of Grossman and Reinsch (2002). Atterbergs limits, namely
Liquid Limit (LL) and Plastic Limit (PL) were estimated using Cassagrande method
as described by Smith (1990). Plasticity Index (PI) was calculated as a difference
between LL and PL.
The Coefficient of Linear Extensibility (COLE) being a measure of shrink-swell behaviour was calculated as follows using the method described by Foth (1984):
COLE
= Length of moist sample/Length of dry sample 1 | (1) |
Volumetric Shrinkage (VS) was calculated from COLE according to Schafer and Singer (1976) as follows:
VS = {(COLE + 1)3 } 1x100 | (2) |
Total carbon was measured by combustion with an Elemental CNS analyzer (Elemental America, Inc. Mt. Laurel, NJ) according to the methods described by Nelson and Sommers (1996). Organic matter was estimated as total carbon content multiplied by a factor of 1.724.
Statistical Analyses
Soil data were subjected to Analysis of Variance (ANOVA) using the PROC
mix-model of SAS (2000). Means were separated using Standard Error of the Difference
(SED) at 5% level of probability. Multiple regression was used to calculate
the variance associated with the best fitting linear combination of the variables
according to the model below as proposed by Steel and Torrie (1980).
Y = a+b1x1+ b2x2+b3x3+
+ bnxn |
Where, | ||
Y | = | Volume Shrinkage (VS) of soils |
A | = | Intercept |
bs | = | Slopes corresponding to the Xs |
xs | = | Independent soil variables |
n | = | Number of variables used in the model. |
In order to evaluate average prediction uncertainty of the model, Root Mean Square Error (RMSE) was estimated.
(4) |
Where, | ||
di | = | Difference between the predicted and measured values at a given level |
n | = | Number in the data set. |
The bias of the model was calculated to evaluate model overestimation or underestimation.
(5) |
Di is defined as in Eq. 4
RESULTS AND DISCUSSION
Particles Size Distribution
Intrinsic soil properties varied significantly with depth at 5% level of
probability. Sand and clay-sized particles had higher values possibly due to
the parent materials from which they were derived. Lower clay content (17%)
was found in the upper 0-20 cm of the study soils. However, clay content increased
at lower depths (21-100 cm) to 20-30%. Conversely, silt and sand content decreased
with depth. The extent of shrinking and swelling is influenced by the amount
and kind of clay in the soil (Table 1).
Table 1: | Particle size distribution of soil at different depths |
Table 2a: | Atterbergs limits, bulk density and COLE |
Table 2b: | Shrink-swell hazard ratings |
(Source: Schafer and Singer, 1976) |
Table 3: | Correlation coefficients of Volume Shrinkage (VS) and some soil properties |
*: Significant-p<0.05, NS: Non Significant |
Atterbergs Limits, Bulk Density and VS
Atterbergs limits present the values of liquid limit, plastic limit
and plasticity index for the soils (Table 2a). The value of
liquid limit (26-30%) was higher than that of plastic limit (11-13%). Plasticity
index was lower (16-17%) in the upper soil layer (0-40 cm) than in the lower
horizons (41-80 cm layer) by 24%. The value of bulk density was 1.36 mg m-3
at the upper horizons (0-20 cm). This increased to 1.40-1.58 mg m-3
in the lower part of the profile (21-100 cm). Similarly, volume shrinkage value
(11.20-11.80%) was least at the topmost horizon (0-40 cm depth) and increased
to between 14-18% in the lower part of the profile (40-100 cm). Conversely,
soil organic matter with a value of 1.85% was higher at the upper horizon (0-20
cm) when compared with values of between 0.09-1.16% at the lower part of the
profile (21-100 cm). Soils at all depths showed low to moderate shrink-swell
properties (Table 2b) with middle layers showing higher shrink-swell
characteristics. These results show that soils of the study area in their natural
state exhibit shrink-swell potentials. Higher shrink-swell values at the middle
layers of the soil profile may be attributable to high concentration of illuvial
clays. Soil shrinkage has been related to a number of different soil properties.
Some of these include clay percentage (De Jong et al., 1992) and organic
carbon (Reeve et al., 1980). Of all the soil properties, soil clay content
has most often been shown to be related to soil shrinkage potential. For example,
Simon et al. (1987) showed that 42-71% of the variation of COLE could
be explained by soil clay percentage. Other studies have indicated shrinkage
indices such as COLE were more strongly related to the fine clay fraction (Dasog
et al., 1988).
Relationship Between VS and Selected Soil Properties
The Volume Shrinkage (VS) showed positive significant correlation (p<0.05,
n = 150) with Atterbergs limit, clay and OM. The highest correlation was
found with plastic limit (R = 0.64) and clay (R = 0.61) while medium to low
correlation was found between VS and liquid limit (R = 0.59), plastic index
(R = 0.50) and organic matter (R = 0.47) (Table 3). Gray and
Allbrook (2002) found no significant relationship between organic carbon and
COLE.
Table 4: | Volume shrinkage model attributes (p<0.0001) |
They concluded that there appears to be no consensus on the role of organic matter in soil shrinkage in soils. While Reeve et al. (1980) found a positive correlation between organic carbon and shrinkage in both topsoil and subsoil samples, the majority of studies have found no relationship with this soil property (Gray and Allbrook, 2002).
No significant relationship was found between soil bulk density and VS. Flowers and Lal (1999) found that field measurements of soil shrinkage did not show change in the total shrinkage volume due to the lack of changes in the soil bulk density. They found that field bulk density values ranged between 1.4 and 1.6 mg m-3 and no differences in shrinkage volume were found in this range. Igwe (2003) made similar findings in the floodplain soils along the banks of River Niger. Earlier, Mbagwu and Abeh (1998) and Thomas et al. (2000) found strong linear relationships between COLE and plastic limit. Therefore, it will be concluded that those Atterbergs limits and clay content contribute significantly to volume shrinkage and eventually shrink-swell hazards of soils of the area.
Modelling
Using five independent variables, namely, total clay, total sand, OM and
BD, VS model was derived as follows:
VS = 0.190 + 0.040 BD + 0.010 OM 0.001 PI
0.002 Total clay 0.003 Total Sand. |
These predictor variables were highly associated with VS (R2 = 0.76; p<0.0001; n = 150). The prediction was highly accurate (RMSE = 0.01) with slight overestimation (bias = + 0.00001). These trends occur although soils were formed from different parent materials. These mean that the variables used in the model had direct relationship with VS and could be used to predict the swell-shrink behaviour of soils in the area of study (Table 4).
CONCLUSION
Soils of the study site exhibit varying degrees of volume shrinkage with depth. Generally, these soils showed moderate shrink-swell hazard rating although undisturbed soil samples were used in the analysis.
Volume Shrinkage (VS) showed strong relationship with Atterbergs limits and clay and as a consequence, these properties plus BD, texture and OM were used as predictors in VS modelling.
The volume shrinkage model resulted to a good prediction of load-bearing strength of soils at p≤0.05 with little I-R2 value using 150 soil samples. This model will be useful in sustainable use of soils for engineering purposes, especially at planning stages of project implementation.