The study presented in this paper dealt with exploring utility of multispectral image dataset for estimating geotechnical characteristics of expansive soils and mapping variations in their expansion potential. A geotechnical parameter (weighted plasticity index) and soil spectra derived from Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) imagery were linked. A multivariate statistical calibration, partial least squares regression analysis is used to establish the link. A coefficient of determination, R2 of 0.71 coupled with low root mean square error of prediction, low standard error of performance, negligible bias and small offsets are obtained. These model performance indices indicate a strong relationship between weighted plasticity indices and soil reflectance spectra. Measured and predicted values of weighted plasticity indices show a similar spatial trend of variation. Results indicate capability of ASTER data for estimating and thereby mapping variation in magnitude of soil expansiveness. The presented analytical approach can significantly contribute to geotechnical applications. Especially to obtain an indication of soil expansiveness in a reconnaissance and technical feasibility studies, preliminary site investigation schemes and basic assumption of parameters which in turn influence preliminary choice of possible road alignment, structures and associated cost estimates.
Get your grade
or your money back
using our Essay Writing Service!
Keywords: Expansive soil, swell-shrink, weighted plasticity index, ASTER, PLS.
The first stage of any major civil engineering project generally involves a reconnaissance study of a project site followed by detailed geotechnical investigation. Primary aim of such a study is to collect data concerning ground conditions in order to assess their likely influence or vice versa on design, construction and performance of infrastructures. Potential problems that could affect design, construction, performance and life time of infrastructures are best determined during pre-design phase when compromises can be made between structural, architectural, mechanical, and other aspects of design without disrupting design processes. Changes during design phase or construction will probably delay activities and pose economic disadvantages. It is therefore critical to ensure that material conditions are properly assessed in a geotechnical investigation scheme. One of these is identifying and characterizing expansive soils. Since expansive soils change their geotechnical properties with variation in moisture content, possible heave prediction in such soils is needed. A number of qualitative and quantitative, direct and indirect, in-situ and laboratory testing procedures are available to identify and characterize expansive soils. Due to their simplicity and good correlation among other engineering properties with soil expansion potential, consistency limits are common indicators of soil expansiveness. The more soil testing is done before hand, the easier it is to reduce risk in design of infrastructure. However, it is impractical to analyze many samples over short distances for it is costly and time consuming. Remote sensing can potentially provide with a continuous representation of a site under investigation, other than discrete sampling points. Soil expansion potential depends among other things on clay content and mineralogy of soil (Nelson and Miller, 1992) which also control their spectral characteristics (Van der Meer, 1999: Chabrillat et al., 2002; Kariuki et al., 2003).
Research in soil science well explored the basic relationships between spectral response and soil characteristics; established role of remote sensing for characterizing and mapping soil and soil properties. Stoner and Baumgardner (1981) presented spectral reflectance and variations in spectral reflectance characteristics of different soils. Ben-Dor and Banin (1994) showed potential of near infrared spectroscopy for deriving soil properties. Shepherd et al (2005) used a multivariate calibration technique, partial least squares regression analysis (PLSR) to predict soil properties from their reflectance spectra. Van der Meer (1999) outlined potential of remote sensing for mapping soils susceptible to volume changes based on diagnostic clay mineral spectral signatures. Chabrillat et al. (2002) demonstrated capability of hyperspectral remote sensing in detecting and mapping expansive clay minerals. Bourguisnon et al. (2007) mapped different clay mineral species (kaolinite, illite, smectite) from an ASTER image. One-to-one relationships between selected engineering parameters and laboratory acquired soil reflectance spectra were established by Kariuki et al. (2003). They presented empirical relations using known specific clay mineral diagnostic absorption features (~1400 nanometer, ~1900 nanometer and ~2200 nanometer wavelengths) parameters and soil engineering characteristics. Waiser et al., (2007) predicted soil clay content from visible near infrared spectra of soils using PLSR models. Ben-Dor et al., (2002) mapped soil properties (organic matter, soil saturated moisture and soil salinity) from a hyperspectral image data. Rainey et al., (2003) mapped clay and sand content of intertidal environments of estuarine from airborne remote sensed data with multivariate regression techniques. Yitagesu et al. (2009a, 2009b) reported significance of laboratory spectroscopy for quantifying geotechnical parameters of expansive soils through a multivariate statistical regression analysis. Instead of using atmospheric absorption bands, they used wavelengths that fall within atmospheric windows for possible extension of the approach to optical remote sensing. Extension of the technique to optical remote sensing for deriving estimates of geotechnical characteristics of expansive soils over a large area on the other hand, can provide a significant input to geotechnical applications.
Always on Time
Marked to Standard
The purpose in the current study is to evaluate utility of multispectral remote sensing data (ASTER) for estimating and mapping variation in soil expansiveness in terms of selected geotechnical parameter (weighted plasticity index) of soils. A multivariate statistical analysis is used to explore relationship between ASTER derived soil reflectance spectra and weighted plasticity indices (PIw). We propose a simple method of estimating soil expansiveness from ASTER reflectance spectra of soils, and thereby map variations in magnitude of soil expansiveness.
2.1 Study area
The study was carried out in an area located (Figure 1) in the central part of Ethiopia, in the upper valley of the Awash River which drains the northern part of the Rift Valley. Topography ranges from a relatively flat to hilly, undulating and steep mountainous terrain. Elevation ranges from 1500 to 2500 meters above sea level. Conical-shaped isolated hills of scoria formed during the late stages of volcanism are common in the study area. Climate is moderate to wet with mean annual rainfall of 1200 millimeters in Addis Ababa and areas close by, and 870 millimeters around the town of Nazret. Temperature ranges from 8 oc to 25 oc twenty five degree centigrade. While topography controls the ease with which the soils are drained, heavy rainy periods followed by prolonged dry periods contribute to susceptibility of soils to volume changes.
Geology (Abebe et al., 1999) around TuluDimtu (Figure 1 shows names of the towns) consists of Tertiary to Quaternary volcanic formations which include alkaline basalts, spatter and cinder cones, ignimbrites, rhyolitic flows and domes, and trachyte. Near Debre Zeyt, alluvial and lacustrine deposits dominate which include sand, silt and clay. From Debre Zeyt to Modjo town lacustrine deposits, and after Modjo town fall and poorly welded pyroclastic deposits dominate with ryolitic and trachytic formations intercalations.
Soils in the study area can be classified into vertisols, luvisols, leptosols, phaeozems and andosols (Figure 1). According to FAO (1998) vertisols are clay rich (smectitic) expanding soils that swell and shrink with fluctuation in moisture content. Luvisols are common soil types in flat or gently sloping land, derived from a variety of unconsolidated material including alluvial, colluvial and eolian deposits. Leptosols are very shallow soils over hard rock or in unconsolidated gravely material that are common in mountainous areas. Phaeozems are soils that are predominantly derived from basic material and are rich in organic matter. Andosols are young soils in volcanic regions that are usually associated with pyroclastic parent materials. From engineering perspective, soils that are predominately black and contain highly expansive clay (vertisol family, commonly termed as black cotton soils) are found in the section from Addis Ababa to Modjo covering an extensive area. According to Abebe et al., (1999), the black cotton soils are of alluvial, lacustrine and colluvial origin. The hilly and mountainous terrains are mainly covered with fresh to partially weathered basalts.
Natural vegetation cover is in general poor since most of the area is farmland, hence provides enough soil exposures for soil remote sensing in dry periods. Built-up areas follow the existing road alignment connecting Addis Ababa to Nazret. Kaliti, Akaki, Dukem, Debre Zeyt, Modjo and Nazret are the major built-up areas. Deeply incised drainage patterns and gully erosions are common features in the area particularly past Modjo town towards Nazret.
2.2 Soil sampling
Much of the new Addis Ababa - Nazret expressway route traverses on expansive soils mainly black cotton soil. Figure 1 also shows that majority of the route passes on vertisol soil classes (scale of soil map in figure 1 is coarse, so generalized). Possible soil expansion potential should be predicted to eliminate or minimize its detrimental effect on the highway subgrade and associated adverse impacts on the adjoining environment.
Soil samples were collected along the new expressway route from its starting point near TuluDimtu to its end at the town of Nazret. The sampling was part of a comprehensive geotechnical investigation and testing scheme for assessing suitability and quality of subgrade materials. Samples were recovered from shallow trial pits of 1meter depth which is commonly the depth at which shallowly founded structures are laid. Samples were taken at every 500 meters intervals. Additional trial pits of about 3meters deep were dug between 3 to 5 kilometers intervals with an aim of determining vertical extent of potentially expansive soils.
2.3 Geotechnical testing
This Essay is
a Student's Work
This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.Examples of our work
Identification and prediction of soil expansiveness was based on consistency limits. This method has an advantage of using parameters that are relatively easy to measure. Consistency limits (liquid limits (LL), plasticity limits (PL) and plasticity indices (PI)) were determined in accordance with ASTM D4318-05 standard test method. Weighted plasticity indices (PIw) were calculated from laboratory determined soil plasticity indices and percent fraction of each soil sample passing 0.425 millimeter ASTM sieve as follows:
PIw=PI * (% material passing 0.425 millimeter sieve)/100 ã€”1ã€•
PIw, therefore compensate for the effect of coarser grained material that is not included in testing plasticity indices of soil samples. Soil plasticity is influenced by content and mineralogy of clay fractions in soil, hence is an indirect measure of their expansiveness. Plasticity and soil expansiveness are directly proportional. In accordance with the Ethiopian Roads Authority Site Investigation Manual (ERA, 2002), soil with weighted plasticity index of greater than twenty percent are categorized as potentially expansive. Such soils may cause major problems when construction is undertaken unless their actual expansiveness is quantified and proper mitigation measures are formulated accordingly during design of infrastructures. Encountering soil with weighted plasticity index of more than twenty percent warrants additional detailed investigation of soil potential swell-shrink characteristics.
Particle size distributions tests were conducted in accordance with ASTM D6913-04e1 standard test method using sieve analysis (for the fraction passing through 2 millimeter, 0.425 millimeter and 0.075 millimeter ASTM sieve openings). Grading of soils finer than 0.075 millimeter ASTM sieve is determined by hydrometer analysis in accordance with ASTM D422-63(2007) standard test method.
2.4 Mineralogical analysis
Mineralogical composition of soil samples were examined using a Siemens D5000 X-ray diffractometer (XRD) analysis. The analysis was done on bulk soil samples to determine overall constituents. Clay fractions were analyzed to quantify major, minor and trace composition of clay species in the soil samples. X-ray florescence (XRF) analysis was used for determining the oxides in the soil samples.
2.5 Multispectral image analysis
The Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) covers visible near infrared (VNIR, 400-1000 nanometer), short wave infrared (SWIR, 1000-2500 nanometer) and thermal infrared (TIR, 8000-12000 nanometer) regions of the electromagnetic spectrum. ASTER has nine bands in the VNIR and SWIR regions of the electromagnetic spectrum, and five bands situated in the TIR region. Some of the ASTER bands in the SWIR wavelength region are situated in wavelength regions that are well known to be related with characteristic absorption features of clay minerals (Clark, 1999). Soil spectral signatures measured by many, narrow and contiguous bands of high spectral resolution instruments show well-resolved spectral features that are important in visual qualitative identification of clay mineral assemblage (Chabrillat et al., 2002; Kariuki et al., 2003). However, ASTER band widths are coarse in comparison to laboratory spectrometers. Range of wavelengths regions covered by ASTER sensor coupled with their spatial resolutions are summarized in Table 1.
Use of ASTER data in mineral mapping and lithologic discrimination has become a common practice in recent years (Bourguisnon et al., 2007; Hubbard and Crowley, 2005; Rowan and Mars, 2003) for its optimal placing of bands that are sensitive to minerals, its low acquisition costs while covering broad and inaccessible areas. Apart from availability of large archives of ASTER data, upcoming similar missions (ENMAP, the new LANDSAT etc) would be of interest for future application with respect to mapping soil engineering properties.
Two ASTER level (1B) scenes covering the study area, acquired in January 2008 were obtained from the EROS Data Center (EDC), South Dakota, U.S.A. Geo-metric correction and geo-referencing was done by the image provider. The ASTER scenes were preprocessed including co-registration of the 30 meter spatial resolution SWIR bands with the 15 meter spatial resolution of VNIR bands. Internal average relative (IAR) reflectance calibration was used to retrieve scene reflectance values from the ASTER radiance data. The two ASTER scenes were subsequently mosaiced and a spatial subset to the extent of the study area was then created.
Spectral responses of natural materials recorded by imaging devices are rarely homogeneous or continuous (Van der Meer, 2004). In the case of this study, land cover other than soil and variation in topography are some sources of heterogeneity. Supervised classification of the imagery was performed using ground truth data having soil, vegetation, water bodies and built-up areas as surface cover classes. Spectral angle mapper (SAM) classification technique was used to stratify the image into these surface cover classes. SAM is a physically based supervised classification method (Kruse et al., 1993) where image spectra is compared and matched with reference spectra. It compares the angle between image spectra with that of reference spectra, in which smaller angles represent closer matches and larger angles represent dissimilarities. The advantage of using SAM is its insensitivity to illumination and albedo effects (Kruse et al., 1993; Mather, 1999) which are present in the ASTER images. An overall accuracy of 70.5 % with kappa coefficient of 0.6231 is obtained. Non soil surfaces classes (built-up areas, vegetation, water bodies) were masked out and only the soil classes are used for further analysis. Thermal infrared bands were excluded since the interest in this study was on soil reflectance characteristics. Spectral reflectance curves of soil samples were collected from the soil class of SAM stratified ASTER image. A total of 92 spectra were extracted from locations where soil samples were taken for a geotechnical characterization.
2.6 Multivariate calibration
Partial Least Squares Regression (PLSR) analysis generalizes and combines features from Principal Component Regression (PCR) and Multiple Linear Regression (MLR) analysis. In PLSR, new variables called PLS components, that are linear combinations of the original explanatory variables (Druilhet and Mom, 2006) will be created by decomposing explanatory variables. Apart from handling a large number of variables and avoiding problem of collinearity, PLSR takes information content of a response variable into account while decomposing sets of explanatory variables (Martens and Naes, 1989; Wold et al., 2001) into PLS components. More information on these three multivariate calibration methods, their algorithms and differences can be found in Brereton (2000), Martens and Naes (1989), Wold et al. (2001), Yeniay and Goktas (2002).
PLSR analysis was used to explore the possible relationship between ASTER derived soil reflectance spectra and weighted plasticity indices. After examining the distribution of variables, appropriate transformations were carried out on variables that showed skewed distributions to make their distribution symmetrical (Wold et al., 2001). Different spectral data preprocessing techniques (Martens and Naes, 1989; Selige et al., 2006) were applied on the soil spectra prior to performing the multivariate regression analysis. This enhances spectral features so as to obtain accurate input data for the PLSR analysis and improve models prediction ability and hence reduce prediction errors (Martens and Naes, 1989). According to Martens and Naes (1989), preprocessing is important to avoid or reduce irrelevant variation in the explanatory variables that can arise from instrument noise or physical properties unrelated to the phenomena of interest. Spectral data normalization was carried out to normalize spectral input data in order to remove uncontrollable scale variations. Normalization does not induce any change into the dataset except for simply rescaling them. Multiplicative scatter correction (MSC) was done to avoid scatter effects from the soils texture, grain size and porosity (Dhanoa et al., 1994). In MSC, the scatter for each sample is estimated relative to that of a reference sample which can be a mean spectrum calculated from all spectra in the calibration model. A full cross validation method is used to calibrate and validate the prediction model (Wold et al., 2001; Martens and Naes, 1989). This method is based on a leave one out principle where one sample will be left out at a time and the model is calibrated on the remaining sample. This will be repeated N times until every sample is left out once and the model is computed on the remaining samples, and the left out sample is predicted. Model performance indices; coefficient of determination (R2), root mean square of prediction (RMSEP), standard error of performance (SEP), bias and offset are used to evaluate model performance.
3. Results and Discussions
3.1 Geotechnical Characteristics
The soils exhibited a wide range of plasticity and grading character: liquid limit of 27-110 %, plasticity indices of 5-70 %, weighted plasticity indices of 2-67 %. Majority of the soil along the road alignment are fine grained. Percentage by weight passing the ASTM 0.075 millimeter sieve ranges from 8 % in coarser soils to a 100 % in most black cotton soils. Hydrometer analysis conducted on selected soil samples showed a high amount of clay fraction especially in those obtained from the first sixty kilometers of the expressway alignment. Clay content in the tested soil samples range from 10-60 % by weight, with higher proportions recorded in black cotton soils. Selected particle size distribution curves are presented in Figure 2.
Figure 3 shows the variation in weighted plasticity indices of soil samples with station from the start of the route near Tulu Dimitu to its end at Nazret town. As illustrated in the Figure, weighted plasticity indices generally tend to be high from kilometer 0-60 of the expressway route with majority of samples falling into weighted plasticity indices of greater than 20 %. This means that expansiveness of soils along the first 60 kilometer of the alignment is high with highest peaks recorded for soil samples obtained at the beginning of the route (5-11 km). Weighted plasticity indices decrease from kilometer 60 onwards till the route ends at Nazret town.
Values of weighted Plasticity Indices recorded for samples obtained from deeper trial pits are plotted in Figure 4. As the Figure depicted weighted plasticity indices of soils do not vary much between surface soil samples and those recovered from deep test pits. There is no systematic increase or decrease in weighted plasticity indices of soils with depth for the whole samples. Generally high weighted plasticity indices (that is > 20 %) are exhibited by most of the soil samples. Comparable expansiveness is manifested by both surface soils and those from deeper pits at least to a depth of 3meters below natural ground level. Accordingly, the geotechnical nature in terms of swell-shrink potential of surface soil samples and those obtained from deeper test pits seem mostly similar.
Organic matter content test indicated that some soils are rich in organic matter (Table 2 bottom). A content of up to twenty one percent by weight of organic matter was recorded.
3.2 Mineralogical assemblage
In the X-ray diffraction (XRD) analysis conducted on some soil samples, smectite particularly montmorillonite and nontronite are found in expansive soils of the study area (Figure 5). The results show presence of these minerals in majority of tested soil samples as major constituents, comprising more than 30 % by weight. Illite/montmorillonite mixed clay minerals, illite, kaolinite were also found in the soils ranging from major (> 30 %) to moderate (10-30 %), minor (2-10 %) and trace (< 2 %) amounts. Presence of quartz, feldspars and goethite are also indicated from the mineralogical assemblage analysis. Calcite was also found in trace amount in some of the soil samples. Formations of these minerals are favored by the geology of the study area, coupled with the tropical climatic conditions and topographic setting. XRD and chemical analysis results are presented in Table 2 Top and bottom.
3.3 Relation between soil expansiveness and ASTER reflectance spectra
A multivariate calibration model was developed linking ASTER image reflectance spectra and weighted plasticity indices of the soils. A coefficient of determination (R2) of 0.71 with root mean square error of prediction (RMSEP) of 6.3, standard error of performance (SEP) of 6.3, a bias of -0.004 and an offset of 5.9 are obtained (Figure 6). The coefficient of determination is high, indicating much of the variation in soil weighted plasticity indices can be accounted for by ASTER image derived soil reflectance spectra. According to the model performance indices, the model has good prediction ability over the range of samples used in this analysis.
3.4 Mapping soil expansiveness using ASTER
Equation obtained from multivariate prediction model is used to map variations in weighted plasticity indices of soils of the study area from ASTER imagery.
PIw=(8.214+(3.595*b1)+(-3.174*b3)+(10.675*b4)+(-3.771*b7)+(-7.029*b8)+(-5.541*b9)) Where the b1….b9 refer to ASTER bands 1…9. ã€”2ã€•
Among the significant predictors, while ASTER bands one and four are found to show positive loadings, bands three, seven, eight and nine show negative loadings. The positive loading from band one can be attributed to spectral signatures of organic matter and amorphous iron oxides (Ben-Dor and Banin, 1994) that are present within the studied soil samples. Organic matter and iron oxides both have amplifying effect on soil expansiveness while they show diagnostic spectral signatures in the VNIR wavelength regions of the electromagnetic spectrum. Negative loading from band three is probably related to sand related spectral features, as well as absorption caused by presence of ferrous iron (Rowan et al., 2003). Positive loading from band four on the other hand can be related to kaolinite (particularly halloysite varieties which can exhibit some degree of swell-shrink character) spectral characteristics. The short wave infrared bands (bands seven, eight and nine) negative loadings might be due to absorption of clay mineral (montmorrilonite, nontronite, illite/montmorillonite, illite) spectral characteristics in these wavelength regions. Rowan et al., (2003) noted that ASTER bands seven, eight and nine are dominated by Fe, Mg-OH absorption features.
Comparing direct measurement values of weighted plasticity indices (shown in Figure 3) with those obtained from PLSR prediction (depicted in Figure 7), a similar spatial trend of variation is observed. Corresponding with the measured values, samples obtained from the 0-60 kilometer stretch of the alignment show higher weighted plasticity indices mostly above 20 %. Most samples from kilometer 60 towards the end of the route on the other hand fall below weighted plasticity indices of 20 %. Although a general similarity in spatial pattern of variations in soil expansiveness is observed, the prediction underestimated weighted plasticity indices of soil samples from the 0-60 kilometer stretch of the route. On the other hand, it slightly overestimated the weighted plasticity indices of soil samples from kilometer 60-80 of the route. The underestimation and overestimation is depicted in the tables summarizing descriptive statistics of measurement and prediction respectively. Mean of predicted PIw is lower than that of measured in the 0-60 kilometer stretch of the route; on the other hand mean of predicted PIw is higher than that of measured in the 60-80 kilometer stretch of the route. Note also differences in the minimum and maximum values. Although there is underestimation in the 0-60 kilometer stretch of the route, standard deviation of the prediction is narrow. Standard deviation of prediction for the 60-80 kilometer stretch of the route is higher than that obtained for measured values.
Map of weighted plasticity indices of soil is presented in Figure 8. On this map the existing road with towns and the new expressway alignment with kilometer markers (showing the length in kilometers of the route) are superimposed to allow easy visualization and comparison with previously presented results (in Figures 3, 4 and 6). The map illustrates that most of areas in the 0-60 kilometer stretch of the expressway alignment show higher expansiveness in terms of weighted plasticity indices (denoted in red color). Areas in the last 20 kilometer stretch of the route on the other hand, show lower (denoted in blue color) weighted plasticity indices. The remaining areas fall in between high and low soil expansiveness (as denoted by the green, yellow to orange colors).
Blue pattern around Addis Ababa city towards Kaliti in the map indicates that soils in this vicinity are of lower expansiveness. However, these areas are extensively covered with black cotton soils and as previous studies indicate (GSE, 1990) exhibit high degree of swell-shrink potential. These areas are urbanized places where surface soils can be influenced or contaminated with imported materials (probably gravel and natural select materials, sand, crushed aggregates etc) for construction purposes. Perhaps this is a possible explanation as to why the map showed soils of lower expansiveness in these areas.
In this study area the vertical extent of expansive soils is also high and shows similar nature of expansiveness with those manifested by surface soil samples. This extends to the whole depth subjected to investigation; about 3meters for samples recovered from the first 60 kilometer stretch (Figure 4) and 1meter for samples recovered from the remaining stretch of the route. Map of weighted plasticity indices of soils of this area, which is modeled using spectral responses of surface soil recorded in ASTER image can be safely used as representative of the subsurface to depths mentioned above. However, this might not be the case everywhere as soil strata can vary from place to place. Similar is applicable for soil geotechnical characteristics.
3.5 Validation with a separate image dataset
Reflectance characteristics of soil recorded in image data can be affected by several factors that might interfere with soil spectral signatures (Sullivan et al., 2005). These factors include: observing conditions which are related with atmospheric conditions and topographic variation effects; instrument calibration and atmospheric calibration uncertainties which can contribute to difficulty of achieving accurate surface reflectance; physical conditions of soil cover (e.g. soil moisture, texture and surface conditions); mixed pixels which might contribute to impurity of spectral signatures more so in broad band imagery like ASTER; data quality or signal to noise ratio; difficulty of accounting for subtle spectral variations of soil forming minerals which can be the cumulative result of the aforementioned factors etc.
ASTER scenes of same area acquired at a different year (March, 2006) are used to test repeatability of the model. As described in section 2.5, same preprocessing and classification are applied on the ASTER scenes. SAM gave an overall accuracy of 65 % with kappa coefficient of 0.56. As in section 3.4, equation ã€”2ã€•is used to map weighted plasticity index.
The two weighted plasticity index maps outputted from different ASTER scenes are not in perfect agreement, similarities and differences are noted. As with Figure 8, majority of Figure 9 also falls in the high weighted plasticity index class (greater than 20 % PIw) denoted by red color. Spatial relationships between the two maps are presented in a profile (Figure 10) which indicates moderate correlation. The upper portions of both maps seem more similar than the lower portions. PIw values of samples from kilometer 50 onwards of those from the validation map are generally very low (non-plastic), while same samples in the calibration map exhibit some degree of plasticity although mostly less than 20 % PIw. Apart from differences arising from surface cover variations in the two scenes, other possible sources of discrepancy can be factors that might interfere with soil spectral signatures recorded by imaging device.
Aggregating PIW values from the two maps shown in Figure 8 and 9 using soil map units from Figure 1 clearly indicated differences in magnitude of PIw in the calibration and validation maps respectively (Figure 11). Figure 11 is used only for the sake of illustrating similarities and dissimilarities in the two PIw maps, but has no implication on plasticity characteristics of different soil units due to the coarse scale of the soil map and hence generalization of soil map units. Andosols in both calibration and validation maps tend to exhibit lower plasticity (almost non-plastic), but they show larger variability in the calibration than in the validation map. Luvisols tend to exhibit higher plasticity in both cases, but with larger variability and lower mean in the validation map. Mean PIw in vertisols from calibration map is higher than those in the validation map, while lesser variability is apparent in the former; the variability towards larger PIw values is high in the later. Although the trends of variations of soil PIw seem broadly similar in both maps, differences in magnitudes and variability signify effect of factors that may have influence on soil spectral signatures recorded in image data.
The objective in this study was to explore the relationship between spectral response of soils derived from ASTER and their respective expansiveness. The use of ASTER data for mapping variation in magnitude of soil expansiveness is evaluated. High coefficient of determination R2 of 0.71 coupled with low root mean square error of prediction, small standard error of performance, negligible bias and small offsets are obtained. These model performance indices indicate strong correlation between PIw and ASTER derived soil reflectance spectra, and good prediction ability in the multivariate calibration. From a geotechnical point of view, the presented analytical approach can be of significance. This is in a reconnaissance and technical feasibility studies, preliminary site investigation schemes and basic assumption of parameters which in turn influence preliminary choice of possible road alignments, structures and associated cost estimates. In summary based on our experimental results:
Information in variation of geotechnical characteristics of soil is obscured in areas where land cover is non-soil (e.g. highly vegetated areas). This is probably one major setback in utilizing remotely sensed data for mapping geotechnical characteristics of expansive soils especially in areas where there exist little or no soil outcrops.
Since spectral signatures of soil that are recorded in ASTER imagery are those of surface soil, representativeness of maps developed with the presented analytical approach should be checked prior to taking assumption regarding subsurface soil geotechnical characteristics.
ASTER VNIR and SWIR bands can be used for mapping geotechnical characteristics of soil that show dependence on clay mineralogical assemblage. A continuous surface showing quantitative variation in a geotechnical parameter of interest (PIw) rather than set of discrete point values are recorded in the map. However, mapping spatial variability in soil geotechnical characteristics from multispectral remote sensing data seem to be confounded by different factors interfering in spectral response of soils recorded in image datasets. Hence, further research is required to investigate factors influencing soil spectral characteristics at broad band image scale data; and thereby finding way of optimizing prediction of geotechnical properties and mapping their spatial variability.