INTRODUCTION
Since, the early 1990’s, in the lower Lusatian lignitemining district
glacial till is used to restore arable land, which has been destroyed due to
largescale lignite mining operations. The glacial till in its natural state
before excavation is highly compacted and shows a dense structure, bulk density
is in between 1.9 and 2.1 g cm^{3} (Narra, 2005).
The high bulk density values are regardless, whether, these substrates are used
for agriculture or for nature protection purposes.
Investigations on metastable arrangements of particles, e.g. by Hartmann
et al. (2002) point up that loose arrangements of particles, as it
is created by tillage operations, or in this case by mining operation steps
like excavation, transport and dumping, are unstable. If wetted, the soil tends
to compact rapidly associated with the negative effects on root development
of crops (Lesturgez, 2005). This waterinduced compaction
phenomenon is well known from the loess soil in Lybia (Assallay
et al., 1996, 1997) and is further reported
for agriculturally used soil in Germany (Horn and Rosteck,
2000). This phenomenon in addition describes the increase in bulk density
without additional loads, i.e., only by the weight of the soil itself, when
the soil or the particle arrangement is wetted (Narra, 2005).
The process behind this phenomenon is that the additional water modifies the
stress situation between the soil particles. Associated with this, the equilibrium
of the inner forces in the soil becomes unstable and is shifted to a new one
(Narra, 2005). The studies concerning the waterinduced
compaction are primarily focused on the formation of dense layers, socalled
fragipans, in agriculturally used soil (Bryant, 1989;
Franzmeier et al., 1989; Assallay
et al., 1998; Lesturgez, 2005; Narra,
2005).
Studies of waterinduced compaction behavior under laboratory conditions are
investigated using bench Oedometers with continuously increasing loads. Drained
or airdried soil samples are loaded stepwise to a certain defined stress and
are then wetted. When the subsidence in the sample was completed, then the load
was further increased. It was found that the disturbed metastable soil due
to its lower packing state undergoes a greater waterinduced compaction compared
to undisturbed metastable soil previously subjected to both vertical stresses
and wetting and drying cycles (Assallay et al., 1997).
Due to the necessity to understand the waterinduced compaction process consolidation tests are carried out on natural soil and artificial substrates. The approach of these tests is to reproduce the situation occurring when the soil show waterinduced compaction behaviour. The tests consisted of the determination of the dry and saturated consolidation curve of the used soil and the substrate, respectively and then a certain number of combined consolidation tests are carried out. The combined consolidation tests comprised the following steps: (1) loading the dry sample to a certain load, (2) flooding the sample at this load level and (3) the samples are drained to pF1.8 and the saturated samples are loaded stepwise from 0 to 2500 kPa to the maximum stress value.
The goal of this study is to prove the approach to predict the internal stress after waterinduced compaction (hydrocollapse) on the basis of consolidation tests with disturbed glacial till samples, which are formerly subjected to defined stress values representing defined depths of the soil. Further, consolidation tests with undisturbed wet and disturbed dry samples shall prove the changes in the internal stress due to mining operations.
MATERIALS AND METHODS
The Saalian glacial till is from the lower Lusatian lignitemining district located about 100 km Southeast of Berlin, Germany. The test material was collected in October 2006 from the precut section at the Northernmost lignite opencast mining pit Jänschwalde (51°49' N, 14°34' E) from depths of 3 to 10 m at an interval of 1 m. The sediment layer stretches exclusively over the mining area having a mean thickness of approximately 20 m. The study was carried out from Oct 2006 to May 2007. The glacial till is of sandy texture, coherent structure (69.7% sand, 25.8% silt, 4.5% clay) and slightly calcareous (4.3% of CaCO_{3}). The glacial till is classified as loamy sand (USDA soil texture classification).
Void ratio (e) values are obtained from bulk density (ρ_{soil})
and particle density (ρ_{particle}) with the help of Eq.
1. The measured particle density of 2.65 mg m^{3} (Stock,
2005) was used in the calculation.
The carbonate content was determined using a ScheiblerDietrich apparatus.
The Scheibler method involves the determination of the carbonate content of
a soil based on a volumetric method. The carbonates contained in the sample
are converted into CO_{2} by adding hydrochloric acid to the sample.
As a result of the pressure of the CO_{2} released, the water in a burette
that is deaerated rises. The difference in level measured is an indication
for the released quantity of CO_{2}, from which the carbonate content
can be calculated. The carbonate content is expressed as equivalent calcium
carbonate content. The equation for the calculation of amount of calcium carbonate
and the conversion tables are taken from Barsch and Billwitz
(1990).
The impact of disturbance on the stability (preconsolidation load) of the glacial till is evaluated using three substeps. The first step comprised of determination of the stressstrain behaviour in undisturbed soil samples reflecting the undisturbed conditions. In the second step, evaluation of the stressstrain behaviour on dry soil samples was carried out, which reflects the situation during summer when the glacial till dries due to reduction in ground water level. With the specific saturation of dry preloaded soil samples, in the third step, the effect of saturation and recompaction by rain was simulated. The goal of this study was to model the effective stress in the rewetted soil based on the stressstrain behavior of the soil. It shall be shown that the stress obtained corresponds to the overburden load. The above described approach is exemplarily carried out for undisturbed sediment samples and for disturbed dry sediment samples from two depths (4 and 8 m). The two different depths (4 and 8 m) chosen for evaluating water induced compaction are based on the content of carbonates present. The carbonate content divided the profile clearly into 2 different layers. First layer (up to 4 m) is free of carbonate having a mean value between 0.07 and 0.1%. The second deeper layer (from 8 m) characterizes the carbonate content to be about 5.3 to 5.4%. The alteration of carbonate content was at 6 m depth.
The preconsolidation load is estimated for both the dry and wetted soil using
the Casagrande (1936) and Silva (1970)
methods, whereas the point of maximum curvature and virgin compression line
are mathematically determined using the zeropoint of 2nd derivation and 1st
zeropoint of the 3rd derivation of Eq. 2 as proposed by Baumgartl
and Kock (2004). Equation 2 describes the hydraulic model
of van Genuchten (1980) in the following modified way:
where, e_{max} and e_{min} being the maximum and minimum void ratio values, when the applied pressure α→4 and α, n and m are the empirical parameters. The parameters e_{min}, α and n are free fitted, whereas m is calculated as displayed in Eq. 2.
The undisturbed soil samples are saturated and drained at a water tension of 63.1 hPa (pF 1.8) on suction plates. The samples are then placed in Oedometer and the different stresses are applied. The resultant height difference of the samples at each stress level has been recorded. In the case of disturbed dry samples, the samples are subjected to different stresses in the Oedometer and for the disturbed saturated samples, the samples are placed on the suction plates with different applied preloads (0, 5, 10 and 20 kPa) at a water tension of 63.1 hPa. Then, the samples are placed in Oedometer and are subjected to different stresses.
RESULTS
Water Retention Characteristics
The profile of the precut section can be dissected into two sections from
the curves water content vs. water tension: (1) curve obtained from 4 m depth
and (2) curve obtained from 8 m depth. It can be seen that the samples from
4 m do not have coarse pores as the drainage of the samples start from the point
above field capacity, whereas the samples from 8 m have coarser pores as the
drainage starts below the field capacity value (Fig. 1a, b).
At 4 m depth the curve showed a steeper gradient and the difference in the
water content values of the saturated and the point of permanent wilting point
is about 0.15 cm^{3} cm^{3}. At 8 m depth the samples displayed
maximum saturation water content and the increase in drainage to be gradual.
On the basis of water retention curves, the saturated water content (1 hPa or
pF0), field capacity (63.1 hPa or pF1.8) and permanent wilting point (15849
hPa or pF4.2) are calculated and are given in Table 1.

Fig. 1: 
Water retention curves (pFCurve) at (a) 4 and (b) 8 m depths
of the precut section in lignite mining pit Jänschwalde (5 replicates
from each depth) 
Table 1: 
Values of saturated water content, field capacity, permanent
wilting point at different depths 

Table 2: 
Magnitude of waterinduced compaction based on the differences
in void ratios at 10, 100 and 1000 kPa 

P_{4} = 4 m depth, P_{8} = 8 m depth 
The results showed that the water content measured at 4 and 8 m depths was
around 20% volume and 30% volume higher compared to water content measured at
permanent wilting point. The huge amount of water content can cause in sudden
collapses noted as waterinduced compaction. This collapse would be plastic
and is not reversible. This high water content also resulted in swelling and
shrinkage of the profile. Swelling of the samples has been observed when the
samples are completely saturated on the suction plate, especially from the 8
m depth samples. Shrinkage has been observed in the samples from 4 m depth when
the samples are completely drained at pF4.2.
WaterInduced Compaction
The waterinduced compaction of the metastable soil is evaluated at three
different stress levels (10, 100 and 1000 kPa). The dry test soil displays a
considerable capacity to restrict compaction (Fig. 2). The
load capacity of soil decreases abruptly with wetting and the soil collapses
to a certain value of void ratio described by the compression curves of the
wet soil. As a result, the waterinduced compaction rates were higher in between
10 and 100 kPa with a maximum at 100 kPa (Table 2). The relation
between the added quantity of water and the voidratio changes during the waterinduced
compaction for stress levels of 10, 100 and 1000 kPa can be observed in Fig.
2a and b.

Fig. 2: 
Preconsolidation curves plotted for undisturbed, disturbed
dry and disturbed wet samples from (a) 4 and (b) 8 m depths 
Table 3: 
Maximum and minimum void ratio values obtained from 4 and
8 m depths for undisturbed, disturbed dry and disturbed saturated samples
with preloads 

From Fig. 2 it becomes evident, that the waterinduced compaction
already commenced when the soil is still in the unsaturated state. At this point,
about 20% of the pore space is filled with water. Disturbed saturated samples
from both 4 and 8 m depths showed a decrease in the initial void ratio value
with the increase in the preloaded stress (0, 5, 10, 20 kPa). The final void
ratio values of all the disturbed wet curves were almost equal. The observations
showed that with the increase in the preloads, the curves get closer to each
other and assume a flatter shape similar to the shape of the curve obtained
from undisturbed samples.
The difference between the void ratio values of the disturbed dry and the disturbed saturated curves gives the magnitude of waterinduced compaction. Initial void ratio and final void ratio values are given in Table 3. The difference between the initial void ratio and the final void ratio values were less in the case of 8 m samples compared to that of 4 m samples. This difference in void ratio endangers the intensity of waterinduced collapse. Intensity of collapse was high with higher void ratio difference. The disturbed saturated curve preloaded with 20 kPa stress has a similar shape as that of the undisturbed samples. This indicates that the disturbed saturated samples are taking the shape of the undisturbed samples with increasing preloads. So to find out the value at which the initial void ratio value of the disturbed wet curve will be same as that of the initial void ratio value of the undisturbed curve. The stress required was 900 kPa for 4 m depth samples and 600 kPa for 8 m depth samples. The undisturbed samples are taken from the precut section which is with out any external loads acting over it, but still we saw that the soil particles are ideally packed to have such high stresses, this can be due to the oscillation of the glacial till under ice glaciation.
Table 4: 
Preconsolidation loads calculated for undisturbed, disturbed
dry and disturbed saturated samples with preloads from 4 and 8 m depths 

Overestimation: *Two fold higher, **Three fold higher. C:
Casagrande (1936), S: Silva (1970) 
PreConsolidation Load Before and after Wetting
The preconsolidation load before and after waterinduced compaction is
mathematically estimated from the recompression data using Eq.
2. For the dry samples, all modeled curves fit the recompression data, although,
the curves underestimate the measured void ratio values in the range of the
transition between the recompression and virgin compression curve in all cases
(Fig. 2).
Preconsolidation load was calculated using the inflection point and the point of maximum curvature, reproducing overburdenpressure values distinctly higher than the stress from which the soil has been unloaded. The point of maximum curvature also gave values that are distinctly different to the applied stress levels.
For wetted samples, all modeled curves fit the recompression data with high accuracy, although for Pw = 10 kPa, an underestimation in the range of the transition between the recompression and virgin compression curve was observed. Table 4 gives the values of the preconsolidation loads measured from 4 and 8 m depths. The preconsolidation loads are overestimated in the case of 4 m depth samples; whereas, the samples from 8 m depth are in close correlation to the simulated overburden pressures (preloads). However, the point of maximum curvature gave values that are more realistic to the simulated overburden pressures.
DISCUSSION
The results of the experiments on waterinduced compaction displayed in Fig. 2 show that the collapse rates for the investigated glacial till can be easily estimated from the dry and wetcompaction curves. This waterinduced compaction is useful in determining the collapse potential of naturally and artificially compacted soils. The range of stresses that usually occur when the glacial till is subjected to cover spoil heaps are in between 10 and 100 kPa, which is equivalent to a coverlayer thickness of 2 to 10 m.
Under laboratory conditions, the maximum voidratio changes and therewith the
highest potential of the metastable soil to collapse has been detected in the
range of 10 to 1000 kPa. The congruence of the stresses induced by the overburden
pressure after deposition of the glacial till and the stress range of maximum
collapse rates show that the glacial till is highly sensitive to wetting under
the naturally occurring field conditions. This phenomenon of collapse due to
the induced water in the glacial till has also been observed by Stock
(2005). The sensitivity of the glacial till material to wetting is due to
the treatment of soil in which it is transformed into a metastable soil structure
after sieving. It should be noted that this treatment is comparable to the field
conditions in which excavation, transport and deposition activities disrupt
the soil structure (Narra, 2005). The assumptions with
the disruption of soil structure are consistent with model experiments on loess
investigated by Assallay et al. (1997) and for
sandy soil investigated by Lesturgez (2005).
Assallay et al. (1997) and Lesturgez (2005)
demonstrated that the disturbances in the natural soil will be decreasing with
wetting and is related to their load bearing capacity, while the undisturbed
loess soil have their maximum collapse at stresses >1000 kPa as also observed
by Assallay et al. (1996).
From Fig. 2, the initiation of waterinduced compaction is seen to occur before the soil is saturated. This is interesting since, it shows that the risk of waterinduced collapse begins at a very early state of saturation or in other words, the threshold for critical water content appears to be approximately 20% or more waterfilled pore space.
However, this is only theoretical since, it is not clearly known, whether,
the water is homogeneously distributed throughout the soil or only at the beginning
of the collapse event, i.e., only in the top few millimeters of the soil (Stock,
2005; Narra, 2005). Further, it became clear that
the collapse occasionally continued to occur after the soil has been saturated,
especially at 100 kPa, where the soil showed the highest rate of collapse (Table
2). The collapse continues over a longer period than that of the actual
saturation, since the collapse rate is slower than the saturation process.
Baumgartl and Kock (2004) have shown that the preconsolidation
load can be estimated purely mathematically using the sigmoidal fit function
described by Van Genuchten (1980) to derive the parameters
needed for the Casagrande (1936) and Silva
(1970) methods. The sigmoidal Van Genuchten (1980)
model presumes a continuous run of the data, which needs to be fitted. The Casagrande
(1936) method is highly subjective and varies from person to person in finding
out the point of maximum curvature. This problem is solved by the Silva
(1970) method and the estimated preconsolidation values are not subjected
to interpretation. Therefore, the value obtained will not be prone to errors
(Narra, 2009).
CONCLUSION
The waterinduced compaction rates of the metastable test soil derived using the dry and wet compression curves are found to be highest between 10 and 1000 kPa. The simulated waterinduced compaction with both disturbed and disturbed saturated conditions indicate a irreversible (plastic) deformation of the metastable soil. The preconsolidation load provides a useful prediction of the overburden pressure after deposition under field conditions. The derived overburdenpressure (preloads) values are more realistic after wetting for 8 m depth samples compared to 4 m depth samples.