Sei sulla pagina 1di 10

Geosciences Journal

Vol. 12, No. 3, p. 255 − 264, September 2008


DOI 10.1007/s12303-008-0026-5
ⓒ The Association of Korean Geoscience Societies and Springer 2008

Estimating apparent thermal diffusivity using temperature time series: A com-


parison of temperature data measured in KMA boreholes and NGMN wells

Min-Ho Koo* Department of Geoenvironmental Sciences, Kongju National University, Chungnam 314-701, Korea
Yoonho Song Groundwater and Geothermal Resources Division, Korea Institute of Geoscience and Mineral Resources
(KIGAM), Daejeon 305-350, Korea

ABSTRACT: Two different time series data sets, shallow ground comprehensive review on the application of temperature to
temperatures of 58 synoptic stations of the Korea Meteorological a variety of hydrogeological settings. Recently, ground tem-
Administration (KMA) and groundwater temperatures of 67 wells
perature data have been vigorously used in reconstructing
of the National Groundwater Monitoring Network (NGMN), were
analyzed to estimate the apparent thermal diffusivity by using the past climate changes (Lachenbruch and Marshall, 1986;
analytical solution of the one-dimensional heat conduction equa- Veliciu and Safanda, 1998; Huang et al., 2000; Cermak and
tion. The KMA temperature data measured at 1-5 m depths illus- Bodri, 2001; Dorofeeva et al., 2002; Gosselin and Mare-
trated values of the phase delay and the amplitude decay coincident schal, 2003).
with their theoretical relationship, indicating that the conductive Temperature measurements in shallow vadose environ-
heat transport should prevail over the nonconductive processes. ments are mainly used in agronomical studies to character-
On the contrary, some of the estimates from temperatures at a
depth of 0.5 m were away from the theoretical values. It is most ize soil properties (Passerat de Silans et al., 1996; Verhoef
likely that the deviation would be caused by the effects of latent et al., 1996) and the thermal regime of the active layer
heat associated with freezing and thawing of the near ground sur- above permafrost (Hinkel, 1997; Hinkel et al., 2001). Tem-
face. In contrast to KMA data, results obtained from the NGMN perature time series data measured in soils at a shallow
data highly deviated from the theoretical ones, and thereafter depth generally show annual cyclic variations, which can be
yielded unacceptably high values of thermal diffusivities as com- effectively described by the one-dimensional heat conduc-
pared to the representative values of soils and rocks. Implication
of the discrepancy between two data sets was discussed in con- tion model. The thermal diffusivity is the key parameter
junction with perturbation of the conductive heat transport by that controls the temperature in soils near the ground sur-
free convection of water and air occurring in large diameter wells face where periodic temperature change is progressively
as well as the convective heat transport by groundwater flow. attenuated and delayed with depth. The investigation of soil
temperature variations has practical applications in agricul-
Key words: apparent thermal diffusivity, temperature time series, heat
tural engineering, since it can lead to the evaluation of opti-
conduction model, critical geothermal gradient, KMA, NGMN
mum conditions for plant growth and development, and
1. INTRODUCTION also can be used for the control of the thermal-moisture
regime of soil (Usowicz et al., 1996). It is also an important
Temperature measurements in boreholes are quite simple parameter in designing the geothermal heat pump system,
and can be easily performed to get vertical profiles or time in which extraction of thermal energy is accomplished by
series data. Furthermore, equipments for measuring temper- using a ground heat exchanger (GHE). The heat transfer
ature in boreholes are readily available within the accuracy between the GHE and the surrounding geological forma-
of ± 0.01°C. There have been many attempts to use the sub- tions occurs primarily by conduction. Thus, the effective
surface temperature with a variety of applications in hydro- thermal diffusivity of the ground is the fundamental param-
geology. It has been used as a tracer for analyzing the eter in determining a required length of the GHE (Kavanaugh
concurrent flow of heat and water along vertical profiles of and Rafferty, 1997). Lee (2006) analyzed characteristics of
the subsurface medium to elucidate vertical groundwater ground and groundwater temperature data measured in
movement (Bredehoeft and Papdopulos, 1965), groundwa- Seoul, a metropolitan city of Korea, as a basic feasibility
ter recharge (Taniguchi and Sharma, 1993; Tabbagh et al., study for geothermal heat pumps.
1999), percolation rates in the vadose zone (Constantz et The thermal diffusivity of soils and rocks can be obtained
al., 2003) and the pattern of groundwater discharge in a from laboratory experiments (Moench and Evans, 1970;
streambed (Conant, 2004). Anderson (2005) presented a Stephenson, 1987, Bristow et al., 1994; Kluitenberg et al.,
1995) which, however, often yield unsatisfactory results
*Corresponding author: koo@kongju.ac.kr since they usually deal with a small size sample in spite of
256 Min-Ho Koo and Yoonho Song

heterogeneities of the real field. Furthermore, in the case of is concerned with the underground medium where both con-
soils, it is difficult to take undisturbed samples as preserved ductive and nonconductive heat transfer processes occur, α
as the natural condition of the materials. Therefore, field- is usually called the ‘apparent’ or ‘effective’ thermal diffu-
measured ground temperature data have been widely used sivity. The nonconductive heat transfer includes convection
as an alternative to estimate the thermal diffusivity (Adams driven by water and vapor transport, phase changes and
et al., 1976; Asrar and Kanemasu, 1983). Horton et al. (1983) associated latent heat. Although heat transfer in the subsur-
analyzed the limitations of the different methods to calcu- face is dominated by conduction, it is recognized that the
late the thermal diffusivity in terms of the calculated results nonconductive processes also play a comparably important
and for the quantity and quality of data required to make the role particularly in the near ground surface. Thus, the esti-
calculations. mated apparent thermal diffusivity can be very different
Most approaches in the literature to use subsurface tem- from the real value associated with conduction.
peratures assume that temperatures measured in a borehole Furthermore, if the vadose zone is concerned, the con-
should be in equilibrium with the surrounding media. How- ductive process becomes quiet complicated in itself, being
ever, it has long been recognized that the fluids within open affected by presence of water. The vadose zone undergoes
boreholes, air and water, are subject to being unstable by temporal variation of the thermal diffusivity due to associ-
free convection, provided that the thermal gradient exceeds ated variation of its water content driven by precipitation
a certain critical value (Krige, 1939). Convection within the and evapotranspiration. Soils with higher water contents
borehole causes borehole temperatures to be different from tend to have higher thermal diffusivities. Therefore, the heat
the surrounding media, and thereafter misleads the inter- conduction model based on the assumption of time-invari-
pretation of temperature data. Diment (1967) and Gretener ant diffusivity can only give approximate results of temper-
(1967) presented observational evidence that water in large ature variation. Conversely, the estimated thermal diffusivity
diameter wells was unstable under the normal geothermal based on the model of time-invariant diffusivity should be
gradient as predicted by the Krige’s formulation. Thus, it interpreted as an average value over the period of measure-
might be inappropriate to use temperature data measured in ments. Thus, the unsaturated flow leads to a complicated
large diameter wells to interpret the hydrothermal process process of heat transfer in the vadose zone by causing spa-
occurring in the subsurface, since the condition of thermal tiotemporal variation of thermal properties as well as con-
equilibrium between the borehole and the surrounding vective heat transfer.
medium could be invalidated by convection. Based on temperature time series data, the apparent ther-
The primary objective of this paper is to analyze the mal diffusivity can be determined by analytical or numer-
apparent thermal diffusivity from two different time series ical solutions of Equation (1) with the assumption of constant
data sets: shallow ground temperatures of the Korea Mete- diffusivity. Horton et al. (1983) investigated several meth-
orological Administration (KMA) measured in small diam- ods available for determining the apparent thermal diffusiv-
eter boreholes and groundwater temperatures of the National ity of soil near the ground surface from soil temperatures.
Groundwater Monitoring Network (NGMN) measured in Zhang and Osterkamp (1995) discussed some consider-
large diameter wells. Estimation of the thermal diffusivity ations of finite difference methods in determining the ther-
in the literature has been performed mostly by using soil mal diffusivity from a temperature time series. Analytical
temperature data in the unsaturated zone, but seldom by methods are used in this study to estimate the apparent ther-
groundwater temperature data. Thus, the motivation of this mal diffusivity.
study is to compare the results of two different data sets and In order to get the analytical solution of Equation (1), the
further discuss their implications in association with free periodic oscillations of the ground surface temperature,
convection of fluids within boreholes as well as the con- expressed as a Fourier series, can be used for a boundary
vective heat transport due to groundwater flow. condition;
M
2. METHODS AND MATERIALS Ts ( t ) = T0 + ∑ An sin ( nωt + φn ) (2)
n=1

2.1. Theoretical background where T0 is the average ground surface temperature, ω is the
The governing equation describing one-dimensional con- fundamental angular frequency, An is the amplitude and φn is
ductive heat transfer in a homogeneous medium is the phase. The analytical solution of Equation (1) with a
2 time-periodic surface temperature of Equation (2) in a semi-
∂T ∂ T
------ = α -------2- (1) infinite half-space is;
∂t ∂z
M
– z nω ⁄ 2α
where T is temperature, t is time, z is the depth from the T ( z, t ) = T0 + ∑ An e sin ( nωt + φ n – z nω ⁄ 2α )
ground surface taken as positive downward and α is the n=1 (3)
thermal diffusivity of the bulk medium. When the equation It appears from Equation (3) that there are some analyt-
Estimating apparent thermal diffusivity using temperature time series 257

ical expressions available to determine the apparent thermal to nullify the phase delay (Koo et al., 2003). The last method
diffusivity: the phase equation, the amplitude equation and is used in this study.
the logarithmic equation. Figure 1 is the calculated results from Equations (4) and
Under the assumption of a periodic surface temperature (5) illustrating how the amplitudes of the temperature vari-
with the fundamental angular frequency (M=1), the phase ations decay and how the phases are delayed with depth
equation to determine apparent thermal diffusivity can be depending on the thermal diffusivity. Representative values
derived explicitly from the argument of the trigonometric of the thermal diffusivity for various soils and rocks range
term in Equation (3); between 0.2 and 1 mm2/sec (Carslaw and Jaeger, 1959).
2 Thus, it is highly likely that the ground temperatures up to
1 z2 – z1
α = ------- ------------
- (4) a depth of 10 m would exhibit discernible annual fluctua-
2ω δt tions, and could be utilized to determine the apparent ther-
where δt is the phase difference between temperature vari- mal diffusivity.
ations at the two depths z1 and z2. Three methods are avail- Combining Equations (4) and (5) yields the relationship
able to calculate the phase difference from two temperature between the phase delay and the amplitude ratio;
time series measured at different depths. First, the phase
delay can be calculated from the sinusoidal functions best- ln ( ΔT2 ⁄ ΔT1 ) = –ω δt (6)
fitted to the measured temperatures by applying the least
square method. Secondly, it can be obtained from the time The phase delay and the amplitude ratio estimated from
interval between measured occurrences of the maximum two temperature time series at different depths should fol-
temperature (Horton et al., 1983). Lastly, a cross-correlation low the theoretical relationship as given in Equation (6),
analysis, which is used in this study, can also yield the phase provided that the temperature field is not highly disturbed
difference of two temperature time series (Koo et al., 2003). by the nonconductive processes of heat transfer. Using both
Under the assumption made above for the phase equation, synthetic and measured time series of temperature data,
the amplitude equation to determine apparent thermal dif- Koo et al. (2003) analyzed effects of the vertical water flow
fusivity can be derived explicitly from the argument of the in the vadose zone on the estimates of thermal diffusivity.
exponential term in Equation (3); In their numerical results, effects of the water flow were
2 evidently reflected in the amplitude decay to increase the
ω z2 – z1
α = ---- ------------------------------
- (5) apparent thermal diffusivity. On the contrary, the estimates
2 ln ( ΔT1 ⁄ ΔT2 ) obtained from the phase equation gave a remarkably accu-
where ΔT1 is the amplitude at z1 and ΔT2 is the amplitude at rate result regardless of occurrence of the water flow.
z2. Similar methods are also available to calculate the ampli-
tudes. First, the sinusoidal functions best-fitted to the mea- 2.2. Materials
sured temperatures can be also used to get the amplitudes.
Secondly, they can be approximated from the minimum and The apparent thermal diffusivities of the shallow ground
maximum temperature readings over a period of the funda- in Korea are analyzed by using ground temperature data
mental cycle (Hinkel, 1997). Lastly, the amplitude ratio can measured at 58 synoptic stations of KMA and groundwater
be directly obtained from a linear regression analysis of two temperature data measured at 264 wells of NGMN consist-
temperature time series in which the phase difference, deter- ing of 169 bedrock wells and 95 alluvial wells (Fig. 2).
mined by the cross-correlation analysis, is taken into account The ground temperature, a surface observatory element of

Fig. 1. Annual variations of the sub-


surface temperature calculated from
the solution of one-dimensional heat
conduction equation: (a) the amplitude
decay and (b) phase delay with depth.
258 Min-Ho Koo and Yoonho Song

Fig. 2. Location maps of (a) surface synoptic stations of KMA and (b) monitoring stations of NGMN: closed circles represent the mon-
itoring stations with both alluvial and bedrock wells, and open circles represent the stations with a bedrock well only.

meteorological parameters, is being measured at the syn- soils at various depths of KMA stations, the phase and
optic stations of KMA. It is measured at several depths in amplitude equations are applied to time series data of
steel boreholes with an inner diameter of 4.2 cm. The anal- ground temperatures. Figure 3 shows temporal variation of
ysis was made for the temperature data over the period from ground temperatures at various depths measured at the
1981 to 2002 measured at depths of 0, 0.5, 1, 1.5, 3 and 5 Chuncheon station. A clear illustration of the amplitude
m on a daily basis. All of the data were preprocessed to pro- decay and the phase delay is observed in the figure, indi-
duce temperature time series with a sampling interval of 24 cating that conduction is the dominant mechanism of heat
hours. A prior process of quality control for raw data was transfer in the shallow ground. As discussed above, the
conducted to eliminate some unacceptable values which phase delay between two temperature time series is deter-
were thought to be associated with wrong readings or inputs mined by a cross-correlation analysis (Fig. 4a), and the
by mistake. The elimination was performed by a simple apparent thermal diffusivity is calculated by Equation (4).
numerical scheme where the temperature under inspection The amplitude decay is also determined by a linear regres-
is regarded as a bad data, if it is higher or lower than the sion analysis of two time series of which the phase delay is
temperature of the previous day by more than 15 °C. nullified (Fig. 4b), and the apparent thermal diffusivity is
Under the ‘Master plan for groundwater management’ of calculated by Equation (5).
the Ministry of Construction and Transportation (MOCT) in Figure 5 shows estimates of the phase delay and the
1996, the Korea Water Resources Corporation (KOWACO) amplitude decay between temperatures at the ground sur-
has constructed NGMN to measure and compile nationwide face (z1) and temperatures at the depths of 0.5, 1, 1.5, 3 and
data on water level, temperature and electrical conductivity 5 m (z2) determined for 58 KMA stations. The solid line in
of groundwater. NGMN operates an automated measuring Figure 5 represents the theoretical relationship between the
system on a 6-hour basis. All the data are being collected, phase delay and the amplitude decay as given in Equation
analyzed and provided by the National Groundwater Infor- (6) derived from the solution of one-dimensional heat con-
mation Management and Service Center (GIMS) in KOWACO duction equation. Most of the estimates for temperatures at
which was established in 2003. Temperature data of 264 depths equal to or greater than 1 m fall closely on the the-
monitoring wells which had been installed from 1995 to oretical curve, indicating that the conductive process of heat
2000 were obtained from GIMS, and the data measured transfer prevails over the nonconductive processes in the
from the beginning of measurements to December of 2001 associated depth intervals. Therefore, the estimated appar-
were used for the analysis. ent diffusivity should be quite close to the intrinsic diffu-
sivity.
3. RESULTS AND DISCUSSION On the contrary, some of the estimates for temperatures
near the ground surface (0.5 m) are far away from the the-
3.1. Analysis of KMA temperature data oretical curve. The most likely possibility is that the non-
conductive heat transfer would occur actively near the
In order to estimate the apparent thermal diffusivity of ground surface, and thereby could cause a temperature vari-
Estimating apparent thermal diffusivity using temperature time series 259

Fig. 3. Variations of ground tempera-


ture measured at various depths of the
Chuncheon station.

Fig. 4. Determination of the phase


delay and the amplitude ratio of two
temperature time series by: (a) a cross
correlation analysis and (b) a linear
regression of the two data sets.

ation to deviate from the theoretical one based on the heat


conduction model. As would be expected, soils near the
ground surface are highly vulnerable to freezing and thaw-
ing, convective heat transfer due to infiltrated water and
temporal variation of thermal properties. As illustrated in
Figure 5, the deviation associated with the nonconductive
heat transfer seems to lead to a reduction of the phase delay.
Although not directly explored in this paper, the link
between the nonconductive heat transfer and reduction of
the phase delay could be examined by a numerical model of
heat transfer incorporating convection, phase changes and
associated latent heat.
Figure 6 shows a clear evidence of the nonconductive
heat transfer, illustrating variation of the ground tempera-
ture affected by latent heat associated with freezing and
thawing. During the months of the year when freezing and
thawing of the ground surface occur (January – March), the
ground temperature at a depth of 0.5 m is immune to
change being affected by the latent heat released or
absorbed during phase changes. Thus, the effect of latent Fig. 5. Estimated phase delay and amplitude damping with depth:
heat in the cold season highly disturbs the conductive tem- the solid line represents theoretical values calculated from the
perature variation observed in the other seasons. The per- solution of one-dimensional heat conduction equation.
260 Min-Ho Koo and Yoonho Song

the phase and amplitude equations based on the simplified


heat conduction model to determine the apparent thermal
diffusivity may be limited particularly for soils near the
ground surface in Korea. In consideration of the results dis-
cussed above, it is likely that the limited availability would
be associated with the freezing and thawing process as well
as the convective heat transfer of water and air which is not
directly accounted for in this paper. These nonconductive
processes generally are more prominent in the top soils
within 0.5 m depth below the ground surface.
Figure 7 shows variations of the apparent thermal diffu-
sivity of KMA soils with depth determined by the ampli-
tude equation. Figure 7a represents thermal diffusivities for
the depth intervals illustrated by arrow lines, and Figure 7b
represents thermal diffusivities for the whole intervals from
the ground surface to the depths given in y-axis. The results
clearly demonstrate that soils at greater depths would have
Fig. 6. Variations of surface and ground temperatures showing the
effect of latent heat associated with freezing and thawing of the higher thermal diffusivities. Particularly the drastic increase
near ground surface (Chuncheon KMA station). can be observed at a depth of 1 m. As discussed in Koo et
al. (2003), variability of the apparent thermal diffusivity
should be attributed to the distinctive differences in the
sistence of a constant temperature close to zero degree is porosity, the water content and the organic content of soils.
directly related to soil water content. It is intuitive that soils Thus, it is likely that low diffusivities observed in the top
with higher water contents would undergo the latent heat soils would be caused by high porosities due to less com-
effect for longer period of time. Soils with smaller particles paction and development of the aggregate structure, low
generally have higher water contents due to their higher water contents driven by evaporation and high organic con-
field capacity and lower permeability. Thus, it is probable tents as compared to the underlying soils. The estimated
that the latent heat effect would be more pronounced in diffusivity is regarded as an outlier if it is higher than the
fine-grained soils. In addition to the latent heat effect, freez- upper quartile by more than 1.5 times of the interquartile
ing and thawing of soils also can affect the process of con- range (IQR). It is not clear whether the outliers shown at
ductive heat transfer by contrasting thermal diffusivities of shallow depths represent actual diffusivities or not. How-
water (0.14 mm2/sec) and ice (1.2 mm2/sec). Thus, the heat ever, the outliers plotted in Figure 7a are much higher than
transfer in soils near the ground surface is a very compli- the representative values of soils in the literature (Carslaw
cated process, being affected by the presence of water and and Jaeger, 1959).
its phase changes in the cold season. As mentioned earlier,
the heat conduction model based on the assumptions of 3.2. Analysis of groundwater temperature data of NGMN
constant diffusivity and no heat source/sink only gives
approximate results, which sometimes lead to erroneous 3.2.1. Analyzing quality of data
predictions depending on the degree of discrepancies between Monitoring data of groundwater temperature in 169 bed-
the assumed and the real field situations. rock wells and 95 alluvial wells of NGMN are used to esti-
Conclusively, it appears from Figure 5 that availability of mate the apparent thermal diffusivity. The average installation

Fig. 7. Variations of the estimated ther-


mal diffusivity with depth: each box-
whisker plot represents thermal diffu-
sivities for (a) the depth interval illus-
trated by arrow lines and (b) the
interval from the ground surface to the
depth given in y-axis.
Estimating apparent thermal diffusivity using temperature time series 261

depths of bedrock and alluvial wells are 70 m and 20 m,


respectively. Data loggers of the alluvial wells are mostly
placed at depths from 5 m to 10 m below the ground sur-
face, depending on the well depth and the water level. Most
data loggers of the bedrock wells are placed at a depth of
20 m. In consideration of the depths of temperature mea-
surements, it is expected from Figure 1 that a periodic tem-
perature variation with discernible amplitude would be
observed in the alluvial wells; meanwhile it may or may not
be noticeable in the bedrock wells. Thus, if the conductive
heat transfer prevails in the subsurface of monitoring wells,
groundwater temperatures would be periodic or practically
constant depending on the depth of measurements.
However, the expectation of steady periodicity or con-
stancy has failed in many temperature data sets of NGMN.
The failure is thought to be mainly caused by malfunction
of the data logger for a certain period of time. In spite of
regular inspection of monitoring stations and prompt
replacement of malfunctioning devices in the data logger,
abnormal values or outliers often occur due to intrinsic lim-
itations of automatic monitoring and transmission (Yi et al., Fig. 8. Patterns of temperature variation observed in the monitor-
2005). Yi et al. (2005) analyzed NGMN data sets of mon- ing wells of NGMN: (a) normal (periodic or constant), (b) abnor-
itoring wells in the Han River basin and found that the most mal, and (c) mixed.
prominent patterns of the outliers were rapid decline in
water level, no variation in temperature and steep decline in 171 (64.8%), 27 (10.2%), and 66 stations (25%), respectively.
electrical conductivity. However, inspection of the table The stations with normal variations comprise 85 stations
summarizing the outliers in Yi et al. (2005) reveals that no with periodic variation and 86 stations with no variation.
variation in temperature is mostly found in the bedrock The stations with mixed variations also show periodic vari-
wells. As mentioned above, it is expected that the bedrock ation (10 stations) as well as no variation (56 stations), if
wells would show a periodic temperature variation with the fractions with abnormal variations are ignored.
small amplitudes or a constant temperature due to a great
depth of measurements. Thus, consideration of no variation 3.2.2. Determination of the apparent thermal diffusivity
as one of criteria to recognize outliers in temperature may Temperature data of the 95 monitoring stations showing
lead to a wrong judgment, unless the depth of measure- periodic variations are used to estimate the apparent thermal
ments is taken into account in the analysis. diffusivity. For 10 data sets which contain abnormal vari-
In this study, temperature variations observed in 264 ations in a partial period of time, the only fractions of data
monitoring wells of NGMN are classified into three pat- showing steady periodicity are used for the analysis. Deter-
terns. First, based on the speculation discussed above, tem- mination of the apparent thermal diffusivity requires another
perature data sets showing periodicity or constancy for the data set measured at a different depth, which is not avail-
whole period of measurements are considered normal or able in the monitoring stations of NGMN. Thus, in spite of
natural (Fig. 8a). Secondly, date sets showing irregular or a crude analogy, the ground surface temperature measured
abrupt changes and steady increase or decrease are consid- at the nearest KMA station was inevitably used as an alter-
ered abnormal or artificial (Fig. 8b). It is likely that abnor- native. As would be expected, the discrepancy should lead
mal variations occurred mainly due to malfunction of the to results with varying degrees of errors depending on the
data logger and partially due to artificial impacts such as distance between the monitoring well and the nearest KMA
groundwater contamination and nearby construction accom- station. The apparent thermal diffusivity is determined by
panying ground excavation. Finally, date sets having abnor- the same procedure as the analysis of KMA data. In spite of
mal variations for a certain period of time are considered the prior elimination of data with abnormal temperature
mixed (Fig. 8c). The spikes in temperature observed in Fig- measurements, the estimated results need a further refine-
ure 8a and 8c are related to short-term disturbances by reg- ment since they are flawed by the use of KMA data as an
ular inspection of the data logger and pumping for well- alternative to the surface ground temperature and some of
maintenance and water sampling. The analysis for temper- data sets still contain less reliable measurements. Thus, in
ature data sets observed in 264 monitoring wells illustrates order to enhance quality of the analysis, some of data sets
that the normal, abnormal and mixed variations are found in are additionally discarded if the correlation coefficient is
262 Min-Ho Koo and Yoonho Song

Fig. 9. Comparisons of (a) estimated


values of the phase delay and the
amplitude damping with the theoreti-
cal curve (solid line) and (b) the ther-
mal diffusivities obtained from the
phase and amplitude equations.

less than 0.8 in the linear regression analysis of two tem- groundwater temperature. When the analysis specifically
perature time series to obtain the amplitude ratio. Finally, requires natural subsurface temperature undisturbed by the
67 data sets, which comprise 53 alluvial wells and 14 bed- well, the measured temperature should represent the tem-
rock wells, are selected and used for estimating the apparent perature of the surrounding medium adjacent to the well.
thermal diffusivity. This is likely to occur when there is no free convection of
Estimates of the phase delay and the amplitude decay water within the well.
determined for 67 data sets are shown in Figure 9a where Krige (1939) presented the following equation to calcu-
the solid line represents the theoretical relationship. In con- late the critical thermal gradient of a fluid-filled column
trast to the results of KMA temperature data as given in required for the onset of free convection;
Figure 5, many of the estimates highly deviate from the the-
gaT Cνα
oretical curve, falling mostly on the zone below the theo- Gc = --------- + ----------4- (7)
cp gar
retical curve due to underestimation of the phase delay or
the amplitude decay. The discrepancy indicates that the where g is the acceleration due to gravity, a is the volume
nonconductive heat transfer prevails over conduction in coefficient of thermal expansion, T is the absolute temper-
groundwater of the wells. Figure 9b shows the apparent ature, cp is the specific heat, C is a constant which has the
thermal diffusivities estimated by the phase and amplitude value 216 in cgs units, ν is the kinetic viscosity, α is the ther-
equations. In the case of alluvial wells (open circles), some of mal diffusivity and r is the radius of the fluid-filled column.
the estimates are within the range of the representative values As shown in the equation, thermal stability of the water in
for soils and rocks, while others are not. On the contrary, all boreholes is primarily associated with the borehole diameter.
of the estimates for bedrock wells ranging between 2.2 and Figure 10 shows the critical thermal gradients for water and
58 mm2/sec are much higher than the representative values. air as a function of well radius, calculated by Eq. (7) using
Consequently, the results suggested that the use of a peri- the data given in Table 1. Air has higher values of the critical
odic temperature time series to determine the thermal dif- thermal gradient than water due to higher values of ν and α.
fusivity should be highly limited when applied to the For the 200 mm diameter of NGMN wells, the critical ther-
NGMN data, while it produced fairly reliable estimates for mal gradient is calculated to be 0.027 °C/m for air and
the KMA data. The inconsistency may be attributed to two 0.00034 °C/m for water at 15 °C around which temperature
distinctive features of the KMA and the NGMN data. First, of most groundwater in Korea falls on. Therefore, convec-
the KMA data was measured at shallow depths in the tion of water can easily occur in NGMN wells, since the
vadose zone, while the NGMN data was measured at geothermal gradient generally ranges 0.01-0.03 °C/m in
depths below the water table in the saturated zone. It is well Korea. Vertical mixing of groundwater in the well due to
known that the influence of groundwater flow plays a sig- free convection could be indirectly inferred from Figure 11
nificant role on the temperature distribution of the subsur- showing that the groundwater temperature is generally
face. It also can disturb the patterns of periodic variations higher in the bedrock wells than in the alluvial wells,
observed in the shallow underground. Thus, the NGMN although the difference of measurement depths is not great.
data should be influenced by the convective heat transfer Convection of air also can occur within the well, especially
due to groundwater flow. Secondly, there is a great differ- at shallow depths below the ground surface, since the sub-
ence in the diameter of KMA boreholes and MGMN wells. surface near the ground surface can have a strongly positive
For a large diameter well, thermal stability of the water col- geothermal gradient in winter.
umn within the well could be a great concern in analyzing Thus, as in the case of NGMN, either convective heat
Estimating apparent thermal diffusivity using temperature time series 263

attributed to perturbation of the conductive heat transport


system by free convection of air and water in the well as
well as convective heat transport by groundwater flow.

4. SUMMARY AND CONCLUSIONS

Temperature time series data measured in boreholes can


be effectively used for estimating the apparent thermal dif-
fusivity based on the heat conduction equation. In this
paper, two temperature data sets, shallow ground tempera-
tures of KMA and groundwater temperatures of NGMN,
were analyzed to estimate the apparent thermal diffusivity.
The KMA temperature data were greatly fitted to the heat
conduction model illustrating the phase delay and the
amplitude decay coincident with their theoretical relation-
ship. Thus, it is likely that the annual thermal signal of the
Fig. 10. Variations of the critical thermal gradient for water and air ground surface in KMA stations should be propagated into
with well radius calculated by the Krige’s equation. the ground mainly by conduction. Some of the data near the
ground surface, however, showed perturbation of the con-
Table 1. Constants used for calculating the critical thermal gradient ductive heat transfer due to the effects of latent heat asso-
of water and air at 15 °C ciated with freezing and thawing of water in the ground. As
Fluid a (°C-1) T (K) c (ergs/g °C) ν (cm2/s) α (cm2/s) discussed earlier, the vadose zone undergoes temporal vari-
Water 1.50E-04 288 4.19E+07 1.14E-02 1.44E-03 ation of hydrothermal properties of the ground due to
Air 3.47E-03 288 1.01E+07 1.47E-01 1.87E-01 change of the water content. Koo and Kim (2008) presented
a numerical model that could estimate seasonal variation of
transport by groundwater flow or free convection of water the vertical water flux as well as the thermal diffusivity
and air within large diameter wells could invalidate the using temperature data. The suggested model, as compared
availability of periodic temperature data to determine the to the conventional heat conduction model, can elucidate
apparent thermal diffusivity. It is not possible to quantita- the hydrothermal processes occurring in the vadose zone
tively evaluate their effects on estimates of the apparent with better precision.
thermal diffusivity calculated by the heat conduction model. In contrast to the KMA data, the NGMN temperature
However, the results of this study clearly demonstrate that data were not fitted to the heat conduction model with the
the apparent thermal diffusivity should be overestimated, if phase delay and the amplitude decay highly deviated from
it is calculated from groundwater temperature data mea- their theoretical relationship. Furthermore, the estimated
sured in a large diameter well. The overestimation can be apparent thermal diffusivities were unacceptably high as
compared to the representative values of soils and rocks.
Inconsistency of the KMA and NGMN data could be attrib-
uted to the distinctive differences of two data sets in the
depth of measurements and the diameter of wells. The
KMA data were measured from small diameter boreholes
within the vadose zone, while the NGMN data were mea-
sured in groundwater of large diameter wells. Consideration
of these differences leads to presumption that overestima-
tion of the apparent thermal diffusivity illustrated in the
NGMN data could be associated with the convective heat
transport by groundwater flow and free convection occurring
possibly within large diameter wells. These observations
impose a limitation on the use of groundwater temperature
data, especially measured in large diameter wells, to esti-
mate the apparent thermal diffusivity of the subsurface
medium. Ignoring the effects of convection on the thermal
system around the well could result in erroneous interpre-
Fig. 11. Comparison of mean groundwater temperatures of alluvial tations of the apparent thermal diffusivity. Thus, it would be
and bedrock wells. challenging to develop a numerical model to quantitatively
264 Min-Ho Koo and Yoonho Song

analyze the effects of the convective heat transfer on the over the last five centuries reconstructed from borehole temper-
estimates of the apparent thermal diffusivity. The authors are atures. Nature, 403, 756–758.
currently working on laboratory experiments of undisturbed Kavanaugh, S.P. and Rafferty, K., 1997, Ground-source heat pump:
Design of geothermal systems for commercial and institutional
soils collected from the KMA stations to investigate the effects buildings. American Society of Heating, Refrigerating and Air-
of porosity, water content, grain size distribution and organic Conditioning Engineers, Inc., Atlanta, 167 p.
content on the thermal diffusivity. This research is expected Kluitenberg, G.J., Das, B.S., and Bristow, K.L., 1995, Error analysis
to reveal how the estimated apparent thermal diffusivity var- of the heat pulse method for measuring soil volumetric heat
ies in dependent upon the physical properties of soils. capacity, diffusivity, and conductivity. Soil Science Society of
America journal, 59, 719−726.
ACKNOWLEDGMENTS: This study was financially supported by Koo, M. and Kim, Y., 2008, Modeling of water flow and heat trans-
the Korea Energy Management Corporation (KEMCO). The authors port in the vadose zone: Numerical demonstration of variability
wish to thank Dr. Myoung-Seok Suh at Kongju National University of local groundwater recharge in response to monsoon rainfall in
and In-Ok Kang in KOWACO for providing data of this work. Korea. Geosciences Journal, 12, 123−137.
Koo, M., Kim, Y., Suh, M.C., and Suh, M.S., 2003, Estimating ther-
REFERENCES mal diffusivity of soils in Korea using temperature time series
data. Journal of the Geological Society of Korea, 39, 301−317 (in
Adams, W.M., Watts, G., and Mason, G., 1976, Estimation of thermal Korean with English abstract).
diffusivity from field observations of temperature as a function of Krige, L.J., 1939, Borehole temperatures in the Transvaal and Orage
time and depth. American Mineralogist, 61, 560−568. Free State. Proceedings of the Royal Society of London, A173,
Anderson, M.P., 2005, Heat as a ground water tracer. Ground Water, 450−474.
43, 951−968. Lachenbruch, A.H. and Marshall, B.V., 1986, Changing climate:
Asrar, G. and Kanemasu, E.T., 1983, Estimating thermal diffusivity Geothermal evidence from permafrost in the Alaskan Arctic. Sci-
near the soil surface using Laplace transform: Uniform initial ence, 234, 689−696.
conditions. Soil Science Society of America Journal, 47, 397−401. Lee, J.Y., 2006, Characteristics of ground and groundwater temper-
Bredehoeft, J.D. and Papadopulos, I.S., 1965. Rates of vertical atures in a metropolitan city, Korea: considerations for geother-
groundwater movement estimated from earth’s thermal profile. mal heat pumps. Geosciences Journal, 10, 165−175.
Water Resources Research, 1, 325−328. Moench, A.F. and Evans, D.D., 1970, Thermal conductivity and dif-
Bristow, K.L., Kluitenberg, G.J., and Horton, R., 1994, Measurement fusivity of soil using a cylindrical heat source. Soil Science Soci-
of soil thermal properties with a dual-probe heat-pulse technique. ety of America journal, 34, 377−381.
Soil Science Society of America journal, 58, 1288−1294. Passerat de Silans, A., Monteny B.A., and Lhomme, J.P., 1996,
Carslaw, H.S. and Jaeger, J.C., 1959, Conduction of heat in solids. 2nd Apparent soil thermal diffusivity, a case study: HAPEX-Sahel
edition, Oxford University Press, London, 510 p. experiment. Agricultural and Forest Meteorology, 81, 201−216.
Conant, B.J., 2004, Delineating and quantifying ground water discharge Stephenson, D.G., 1987, A procedure for determining the thermal dif-
zones using streambed temperature. Ground Water, 42, 243−257. fusivity of materials. Journal of Building Physics, 10, 236−242.
Constantz, J., Tyler, S., and Kwicklis, E., 2003, Temperature-profile Tabbagh, A., Bendjoudi, H., and Benderitter, Y., 1999, Determination
methods for estimating percolation rates in arid environments. of recharge in unsaturated soils using temperature monitoring.
Vadose Zone Journal, 2, 12−24. Water Resources Research, 35, 2439−2446.
Croft, D.R., Cermak, V. and Bodri, L., 2001, Climate reconstruction Taniguchi, M. and Sharma, M.L., 1993, Determination of ground-
from subsurface temperatures demonstrated on example of Cuba. water recharge using the change in soil temperature. Journal of
Physics of the Earth and Planetary Interiors, 126, 295−310. Hydrology, 148, 219−229.
Diment, W.H., 1967, Thermal regime of a large diameter borehole: Usowicz, B., Kossowski, J., and Baranowski, P., 1996, Spatial vari-
Instability of the water column and comparison of air- and water- ability of soil thermal properties in cultivated fields. Soil & Till-
filled conditions. Geophysics, 32, 720−726. age Research, 39, 85−100.
Dorofeeva, R.P., Shen, P.Y., and Shapova, M.V., 2002, Ground sur- Veliciu, S. and Safanda, J., 1998, Ground temperature history in
face temperature histories inferred from deep borehole tempera- Romania inferred from borehole temperature data. Tectonophys-
ture-depth data in Eastern Siberia. Earth and Planetary Science ics, 291, 277−286.
Letters, 203, 1059−1071. Verhoef, A., van den Hurk, B., Jacobs A.F.G., and Heusinkveld, B.G.,
Gretener, P.E., 1967, On the thermal instability of large diameter 1996, Thermal soil properties for vineyard (EFEDA-I) and savanna
wells - An observational report. Geophysics, 32, 727−238. (HAPEX-Sahel) sites. Agricultural and Forest Meteorology, 78, 1−18.
Hinkel, K.M., 1997, Estimating seasonal values of thermal diffusivity Yi, M.J., Lee, J.Y., Kim, G.B., and Won, J.H., 2005, Analysis of
in thawed and frozen soils using temperature time series. Cold abnormal values obtained from National Groundwater Monitor-
Regions Science and Technology, 26, 1−15. ing Stations. Journal of Korean Society of Soil and Groundwater
Hinkel, K.M., Paetzold, F., Nelson, F.E., and Bockheim, J.G., 2001, Environment, 10, 65−74 (in Korean with English abstract).
Patterns of soil temperature and moisture in the active layer and Zhang, T. and Osterkamp, T.E., 1995, Considerations in determining
upper permafrost at Barrow, Alaska: 1993-1999. Global and Plan- thermal diffusivity from temperature time series using finite differ-
etary Change, 29, 293–309. ence methods. Cold Regions Science and Technology, 23, 333−341.
Horton, R., Wierenga, P.J., and Nielson, D.R., 1983, Evaluation of
methods for determining apparent thermal diffusivity of soil near Manuscript received April 26, 2008
the surface. Soil Science Society of America Journal, 47, 23−32. Manuscript accepted July 9, 2008
Huang, S., Pollack, H.N., and Shen, P.Y., 2000, Temperature trends

Potrebbero piacerti anche