Frost heave is estimated from the total amount

20 cm from 340 cm (11.2 ft) to 400 cm (13.1 ft).

of ice segregation in the frozen zone by:

The upper boundary pore water pressure was

chosen to be computer-generated, as follows.

(

)

θs = θi - θo - θn

(6)

When the profile is completely thawed and down-

where θs = volumetric segregated ice content (%)

water boundary condition is modeled by

θo = porosity (%)

θn = residual unfrozen water content (%).

=0

(7)

If θs is greater than 0, ice segregation has occurred

and the frost heave is computed by multiplying θs

which means that the velocity flux across this

by the zone thickness. The θn parameter estab-

boundary is zero. The upper-boundary condition

is set to 0 cm of water when the upper-boundary

lishes the pore water stress at the freezing front

temperature is above 0C and frozen regions re-

for the solution of the moisture transport equa-

tion. In this study, θn was obtained by assuming a

main in the column. When the surface tempera-

ture is below 0C, a specified constant upper-

moisture tension of 800 cm of water and solving

boundary pore pressure is used. To be consistent

eq 2. The use of the 800 cm of water condition

with previous studies, a value of 300 cm of water

stems from that being the highest tension mea-

was used.

sured in various field studies. Thaw settlement

The lower-boundary pore pressure condition of

from ice melting is the reverse process of that

FROST is set by specifying discrete pore water

described above for ice segregation.

pressures that relate to the water table elevation at

To conduct the calculations described above,

times when these conditions occur. At intermedi-

FROST requires the following input for each ma-

ate times, lower-boundary pore water pressures

terial: 1) Gardner's coefficients for soil moisture

are linearly interpolated. For all the Mn/DOT cases,

characteristics, 2) Gardner's coefficients for hy-

we set the lower boundary pore pressure to pro-

draulic conductivity characteristics, 3) porosity and

duce a constant water table depth throughout the

density of the soil, 4) thermal conductivity and

simulation. We simulated the water table in each

volumetric heat capacity of the dry soil, and 5) the

test section at the depth determined by field mea-

surements to be representative of the on-site con-

FROST also requires the following input for

ditions. Where the measured water table varied

initial and boundary conditions: 1) element lengths,

through a test section, we conducted two simula-

2) upper- and lower-boundary pore water pres-

tions using the deepest and shallowest values.

sures, 3) upper- and lower-boundary temperatures,

Input for the upper boundary temperature con-

4) initial temperature, pore pressure and ice con-

dition consists of a set of specified times and tem-

tent distributions with depth, 5) surcharge pres-

peratures that are implemented as step changes.

sure, 6) freezing point depression and 7) modifier

Values were input in 24-hr increments using the

of the upper node during thaw.

mean daily air temperature. Conditions at Buf-

In all cases, the pavement structure was simu-

falo, Minnesota, were simulated, since this is the

lated as a column with its upper boundary at the

nearest station to the Mn/ROAD facility (16 km;

pavement surface and extending down to 400 cm

10 mi) with a reasonably long record of meteoro-

(13.1 ft) using 99 elements. The length of ele-

logical data. The time period simulated was 1

ments within the expected zone of freezing

October 1959 to 14 November 1960. If the sever-

(down to 110 cm or 3.6 ft) was about 2 cm (0.8

ity of a winter is judged by its air freezing index,

in.). These lengths were adjusted for individual

the 19591960 winter is very near the average

cases to provide nodes positioned exactly at the

value for the 28-year period ending in 1987. The

depths where the interface between materials

distribution of freezing indices at Buffalo during

are located in the Mn/ROAD test sections. The

this time is shown in Figure 3. Starting the simu-

deeper soil was modeled with element lengths

lations on 1 October gave 30 days of computa-

of 4 cm between 110 and 230 cm (3.67.6 ft), 10

tions before the first freeze event, allowing the

cm between 230 and 340 cm (7.611.2 ft), and

4