Comparison of several BHT correction methods: a case

study on an Australian data set

Bruno Goutorbe, Francis Lucazeau, Alain Bonneville

To cite this version:

Bruno Goutorbe, Francis Lucazeau, Alain Bonneville. Comparison of several BHT correction methods:
a case study on an Australian data set. 2007. �hal-00136398�

HAL Id: hal-00136398

https://hal.archives-ouvertes.fr/hal-00136398
Submitted on 13 Mar 2007

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
Manuscript submitted to Geophys. J. Int.

Comparison of several BHT correction methods: a case study

on an Australian data set

Bruno Goutorbe∗ , Francis Lucazeau and Alain Bonneville

Institut de Physique du Globe de Paris, France
∗ Corresponding author. E-mail: [email protected]

7th February 2007

SUMMARY
Bottom-hole temperatures (BHT) from oil exploration provide useful constraints on the sub-
surface thermal regime, but they need to be corrected to obtain the equilibrium temperature.
In this work we introduce several BHT correction methods and compare them using a large
Australian data set of more than 650 groups of multiple BHT measurements in about 300
oil exploration boreholes. Existing and suggested corrections are classified within a coherent
framework, in which methods are divided into: line/cylinder source; instantaneous/continuous
heat extraction; one/two component(s). Comparisons with reservoir test temperatures show
that most of the corrections lead to reliable estimates of the formation equilibrium temperature
within ± 10◦ C, but too few data exist to perform an inter-comparison of the models based
on this criterion. As expected, the Horner method diverges from its parent models for small
elapsed times (or equivalently large radii). The mathematical expression of line source models
suffers from an unphysical delay time that also restrains their domain of applicability. The mo-
del that takes into account the difference of thermal properties between circulating mud and
surrounding rocks – that is the two-component model – is delicate to use because of its high
complexity. For these reasons, our preferred correction methods are the cylindrical source mo-
dels. We show that mud circulation time below 10 hours has a negligible effect. The cylindrical
source models rely on one parameter depending on the thermal diffusivity and the borehole ra-
2 B. Goutorbe, F. Lucazeau and A. Bonneville

dius, which are poorly constrained, but the induced uncertainty on the extrapolations remains
reasonably low.

Key words: Borehole, Geothermics, Bottom-hole temperatures, BHT correction methods

1 INTRODUCTION

It is important in many domains to have a proper knowledge of undisturbed subsurface tempera-

ture. For example, it is used in oil exploration to constrain the temperature history of sedimentary

basins in order to estimate hydrocarbon maturation. In geothermal energy, it is needed to quan-

tify the heat content of reservoirs. In addition, the geothermal gradient must be known to estimate

terrestrial heat flow, which is a fundamental parameter in Earth sciences for the study of mantle

convection, lithospheric deformation, hydrothermal circulation, radiogenic heat production and

cooling of the Earth. Temperatures measured in deep boreholes after drilling form a vast data set,

but it is well known that they are altered by the drilling process, mainly because of the cooling

effect of mud circulation. Temperature logs are usually not used because they are too perturbed

by the complex drilling history. Only recordings of the bottom-hole temperature (BHT) can be

corrected, and then only if an adequate time series is collected..

A multitude of methods to extrapolate the undisturbed temperature have been developed, cor-

responding to various assumptions about the cooling effect of the circulating mud, the borehole

geometry and the thermal properties of the borehole/surrounding rock system. A review of existing

corrections can be found in Hermanrud et al. (1990), but few studies have tried to classify and com-

pare the methods using a large real-world BHT data set. In most of the cases, inter-comparisons

were limited to a few data sets and/or a few correction methods (Hermanrud et al. 1990, Luheshi

1983, Shen & Beck 1986).

The aim of this paper is to describe several correction methods and compare them using a

large data set of BHT measurements from Australian oil exploration wells. In the next section

we gradually introduce a class of BHT corrections, namely analytical methods in which mud

circulation is modelled as an infinite line or cylindrical sink of heat. New models are suggested

within the framework of existing methods, which we classify by their underlying assumptions
 BHT correction methods 3

(line/cylinder source, instantaneous/continuous heat extraction, one or two component(s)). Finally,

we describe the Australian data set, apply the different methods to it, and discuss the specificities,

weaknesses and domains of applicability of each method.

2 MATHEMATICAL FORM OF BHT CORRECTIONS

We develop the theoretical temperature evolution at the bottom of the borehole, predicted by the

different correction methods. Parameters needed for each correction are listed in Table 1.

2.1 Line source methods

Most of the models assume that the mud circulation acts as a heat sink. An instantaneous line

source/sink induces the following temperature perturbation (Carslaw & Jaeger 1959):
µ ¶
QILS r2
δTILS (te ) = exp − , (1)
4πκr te 4κr te

where te is the elapsed time after heat release, r the distance to the source, κr the rock thermal

diffusivity and QILS the thermal strength per unit of source length, that is, the temperature to which

a unit length of source would raise a unit volume of rock (QILS < 0 as we consider a heat sink).

If we choose not to neglect the circulation time tc , and consider that heat is continuously

relaxed, we derive for t < tc the expression for a continuous line source by integrating Eq. (1)

over time:
Z t µ ¶
QCLS r2
δTCLS (t) = exp − dt′ (2a)
0 4πκ r (t − t ′) 4κr (t − t′)
µ 2 ¶
QCLS r
= E1 , (2b)
4πκr 4κr t
R∞ e−u
with QCLS the thermal strength per unit length and time of the line source and E1 (x) = x u
du

the exponential integral. Applying the superposition principle, the temperature after the end of

mud circulation is
· µ ¶ µ 2 ¶¸
QCLS r2 r
δTCLS (tc , te ) = E1 − E1 , (3)
4πκr 4κr (tc + te ) 4κr te

where te is the elapsed time since the end of mud circulation. This expression was first derived

by Bullard (1947) and was used by, e.g., Funnell et al. (1996), Townend (1999), Zschocke (2005),
4 B. Goutorbe, F. Lucazeau and A. Bonneville

with r = a (radius of the borehole). Under the assumption a2 /4κr te ≪ 1, and using the following

series expansion near 0:

µ ¶
x1
E1 (x2 ) − E1 (x1 ) = ln + x2 − x1 + O(x21 , x22 ), (4)
x2

Eq. (3) is approximated by the well-known Horner formula (Dowdle & Cobb 1975):
µ ¶
QCLS tc
δTHorn (tc , te ) ≈ ln 1 + , (5)
4πκr te

which is the most widely used correction method.

If, instead of an infinite line source, one considers a more realistic semi-infinite problem, it is
1
easy to show from symmetry considerations that Eqs. (1), (3) and (5) still hold to a factor 2
on the

plane located at the end of the line and perpendicular to it.

2.2 Cylinder source methods

We now turn to the cases where the borehole radius a is not neglected. At the centre of the borehole,

the effect of an instantaneous cylinder source can be found by integrating Eq. (1) over the radius:
Z a µ ¶
QICS r2
δTICS (te ) = exp − 2πrdr (6a)
0 4πκr te 4κr te
· µ ¶¸
a2
= QICS 1 − exp − , (6b)
4κr te

where QICS is the thermal strength per unit volume of source, which is simply the initial tem-

perature perturbation in the cylinder. This solution was introduced by Leblanc et al. (1981) and

Middleton (1982). Taking into account the circulation time, tc , we derive for t < tc the continuous

cylinder source expression (still at the centre of the borehole) by integrating Eq. (6) over time:
Z t · µ ¶¸
a2
δTCCS (t) = QCCS 1 − exp − dt′ (7a)
0 4κ r (t − t′)
µ 2 ¶
QCCS · a2 a
= QCCS · t − E2 , (7b)
4κr 4κr t
R ∞ −u
with QCCS the thermal strength per unit volume and time of source, and E2 (x) = x eu2 du

the second-order exponential integral. Applying the superposition principle, after the end of mud
 BHT correction methods 5

circulation:
· µ ¶ µ 2 ¶¸
QCCS · a2 a2 a
δTCCS (tc , te ) = QCCS · tc − E2 − E2 . (8)
4κr 4κr (tc + te ) 4κr te

Again, using the series expansion near 0:

µ ¶
1 1 x2 x1 − x2
E2 (x2 ) − E2 (x1 ) = − + ln + + O(x21 , x22 ) (9)
x2 x1 x1 2

and the identity QCLS ⇔ QCCS · πa2 , the first non-null terms of the series expansions of the CCS

[Eq. (8)] and CLS [Eq. (3)] corrections, as a2 /4κr te → 0, are equal and correspond to the Horner

approximation [Eq. (5)]. However, the next terms differ, showing that the line source model is not

a higher-order approximation of the cylinder source method.

As in the line source case, Eqs. (6) and (8) just have to be divided by two in the semi-infinite

case, as the observation point is at the bottom of the borehole.

2.3 Two-component model

A more complex model, which takes into account the difference in thermal properties between

the borehole mud and the surrounding rocks, was studied by Luheshi (1983), using numerical

methods. Shen & Beck (1986) proposed analytical solutions for this problem based on Laplace

transforms. We define some dimensionless parameters:

r r
κr te κr tc ρcm κm λm κr r
τe = 2 , τ c = 2 , β = , δ= , σ= , R= (10)
a a ρcr κr λr κm a

where the subscripts m and r refer respectively to mud and rock properties, ρc is the volume heat

capacity and λ the thermal conductivity (the other symbols are identical to previous sections). As-

suming that mud circulates at constant temperature, Shen & Beck (1986) show that the temperature

perturbation takes the form:

Z ∞ ¡ ¢ ¡ ¢
δT2-comp (τc , τe ) 4 J0 wδ J0 R wδ dw
=1− 2 G1 (τc , τe , w) , (11a)
∆T0 π w=0 φ (w) + ψ (w)
2 2 w
Z ∞ −w2 τe 2 2
4 e − e−u τe e−u τc du
with G1 (τc , τe , w) = 2 2 2
, (11b)
π u=0 u −w
2 2 J0 (u) + Y0 (u) u
³w´ ³w´
φ(w) = σY0 (w)J1 − Y1 (w)J0 , (11c)
δ δ
³w´ ³w´
ψ(w) = σJ0 (w)J1 − J1 (w)J0 , (11d)
δ δ
6 B. Goutorbe, F. Lucazeau and A. Bonneville

with ∆T0 the initial difference of temperature between the mud and rock and Ji , Yi the Bessel

functions of the first and second kind. If, on the other hand, mud circulation is modelled as a

constant heat sink per unit time and length Q2-comp (< 0), then:

λr
δT2-comp (τc , τe ) = H(τc + τe ) − H(τe )
Q2-comp
Z ∞ £ β ¡ w ¢ σ ¡ w ¢¤ ¡ w ¢
2 J
2 0 δ
− w J1 δ J0 R δ dw
− 3 G2 (τc , τe , w) , (12a)
π w=0 φ (w) + ψ (w)
2 2 w
Z ∞ −w2 τe 2 2
4 e − e−u τe 1 − e−u τc du
with G2 (τc , τe , w) = 2 , (12b)
π u=0 u −w
2 2 φ∞ (u) + ψ∞ (u) u
2 2
Z ∞ 2
1 1 − e−u τ J0 (Ru) φ∞ (u) − Y0 (Ru) ψ∞ (u)
H (τ ) = 2 du, (12c)
π u=0 u2 φ2∞ (u) + ψ∞ 2 (u)

βu
φ∞ (u) = Y0 (u) − Y1 (u), (12d)
2
βu
ψ∞ (u) = J0 (u) − J1 (u). (12e)
2

As for the line source method, the above formulae are used with r = a (R = 1) for BHT

correction.

3 APPLICATION TO THE AUSTRALIAN TEMPERATURE SET

3.1 Data set and methodology

We had access to oil exploration data in ∼ 300 wells from Wiltshire Geological Services°.
R Most

of the wells are located in Australia and a few in New-Zealand (Fig. 1). The data set accounts for

about 650 groups of multiple {TBHT , te } measurements, a part of them also having the circulation

time tc (we take a default value tc of 3 hours for the others). Half of the groups are made of two

measurements, one-third of three measurements, and the rest of four or more. There are also com-

plete sets of geophysical well logs – including caliper, sonic, density, neutron, electrical resistivity

and gamma-ray. Finally, around 100 temperatures TDST from drillstem tests are available from 18

of the wells. TDST corresponds to the reservoir fluid temperature, which is supposed to be in equi-

librium with surrounding rocks, so it is considered to represent the undisturbed rock temperature.

All available temperature data are shown in Fig. 2.

Since a BHT is a perturbed measurement, TBHT = T∞ + δT ([tc , ]te ) with T∞ the undisturbed
 BHT correction methods 7

rock temperature to estimate and δT the chosen model in the list of Table 1. To our knowledge,

the ILS and CCS corrections have never been presented, but as they appear naturally within the

framework, we shall keep them in the following inter-comparison. The parameters Q and ∆T0

(see previous section) are considered to be unknown, so at least two TBHT and their elapsed time

te must be available to extrapolate T∞ , which we do using a classical least-square linear regres-

sion. Extrapolations that yield Q or ∆T0 > 0 (that is, a T∞ lower than BHT measurements) are

discarded.

The other parameters (Table 1) are estimated with the help of the geophysical well logs. The

radius a is available from the caliper log after some smoothing. Unlike usual approaches, we do not

assign a constant thermal diffusivity to the rocks, as it depends on the in situ temperature, porosity

and rock type. The rock thermal conductivity λr is predicted from the well logs (sonic, density,

neutron, electrical resistivity and gamma-ray) using a recently developed neural network method

(Goutorbe et al. 2006). The volume heat capacity of the rock matrix ρcmatrix hardly varies from

one rock type to another (Beck 1988) and depends on the temperature T . The rock heat capacity

ρcr is the harmonic mean of ρcmatrix (T ) and the heat capacity of water ρcwater (T ), weighted with

volume proportions. ρcmatrix (T ) is inferred from Vosteen & Schellschmidt (2003), ρcwater (T )

from Lide (2004). The volume proportions are calculated from the neutron porosity log. T is

actually the T∞ we seek, but as temperature only has a second-order effect on heat capacity, a

rough estimate is sufficient: we do this by adding to each TBHT measured at depth zBHT the quantity

TbDST (zBHT ) − TbBHT (zBHT ), where TbDST (.) and TbBHT (.) denote respectively the linear regression

of all TDST and all TBHT against their depth of measurement. Rock thermal diffusivity is then

κr = λr /ρcr . The mean values of λr , ρcr and κr in our data set are:

λr = 1.9 ± 0.6 W.m−1 .K−1 , (13a)

ρcr = 3 ± 0.2 · 106 J.K−1 .m−3 , (13b)

κr = 6.8 ± 2 · 10−7 m2 .s−1 . (13c)

Circulating mud properties obviously vary depending on mud type and operating conditions,

however due to lack of information it is difficult to have proper estimates. Hence as a crude ap-
8 B. Goutorbe, F. Lucazeau and A. Bonneville

proximation we have taken constant values, suggested by Luheshi (1983):

λm = 0.7 W.m−1 .K−1 , (14a)

ρcm = 5 · 106 J.K−1 .m−3 , (14b)

λm
κm = = 1.4 · 10−7 m2 .s−1 . (14c)
ρcm

3.2 Results

3.2.1 T∞ versus TDST

As TDST are supposed to be undisturbed rocks temperatures (see previous section) we compared

them with T∞ resulting from correction of BHT measurements, in wells where both types of

temperature are available (Fig. 3). Too few data exist to perform quantitative statistics, and com-

parison is difficult as temperatures are usually not at same depths, therefore only qualitative re-

marks can be made. In most of the wells, the results are quite close from each other, except for

the two-component model which seems to give systematically higher predictions, and a few cor-

rected temperatures using the line-source models that are clearly out of tendency. T∞ are usually in

agreement with the TDST geotherm within ±5-10◦ C, which is the accuracy expected by a number of

authors (e.g. Brigaud 1989). In wells showing inconsistencies (e.g., well Montague 1), no model

is in agreement with the TDST geotherm, which questions the quality of TBHT or TDST measure-

ments rather than the correction methods. In well Wanaea 1, below the only TDST measurement,

the temperatures corrected with the two-component model seem to be in better agreement with

TDST than the other corrections. However the discrepancy of the two-component model predic-

tions with respect to the other corrections, and also with respect to the other predictions from the

same correction at lower depths, tend to suggest that the TDST measurement is the one question-

able. Hence from these comparisons it is difficult to establish a rating of the different correction

methods, but it can be seen that in most cases they give “reasonable” predictions.
 BHT correction methods 9

3.2.2 Horner versus CLS/CCS

The Horner model is the first order development of CLS and CCS corrections to the limit a2 /4κte →

0 (see section 2.1). As expected, Horner predictions of T∞ diverge from its “parents” models when

the underlying approximation do not hold any more (Fig. 4), Horner extrapolations being system-

atically lower. Therefore, as discussed by several authors (e.g., Shen & Beck 1986), one should

avoid the use of Horner correction when a2 /4κte & 1 (too large radius or, equivalently, too small

elapsed time).

3.2.3 Line source versus cylinder source

As pointed out by Luheshi (1983), when the effect of the mud is modelled as a line source, the

borehole radius is taken into account by considering the perturbation at the distance r = a of the

line (see section 2.1). As a consequence the mathematical form of the thermal evolution (Eqs. 1

and 3) predicts that the temperature continues to decrease after the end of mud circulation (Fig.

5a), as some time is necessary for heat to propagate to the distance r = a from the line source.

This theoretical “delay time” td is obviously unphysical, as one expects borehole temperature to

increase immediately after the end of mud circulation. If a TBHT in a group has an associated

elapsed time te that is lower than td (thus supposedly belonging to the decreasing part of the

thermal evolution), then the model predicts a completely unrealistic T∞ (see example in Fig. 5a).

Therefore the line source methods cannot be used on TBHT having te < td . By solving the equations
¯ ¯
∂δTILS ¯ ∂δTCLS ¯
∂te ¯ = 0 and ∂te ¯
= 0, it is easy to see that td = a2 /4κr for the ILS model, and
te =td te =td

numerical resolution shows that td ∼ a2 /4κr for the CLS correction if tc is not too large (Fig. 5b).

So the line source methods cannot be applied when:

te < td , (15a)
a2
that is > 1, (15b)
4κr te

which is confirmed by the large discrepancy between line source and cylinder source models as

a2 /4κr te & 1 (Fig. 5c). Therefore our conclusion is that the line source models do not necessarily

have a larger applicability domain than the Horner correction.

10 B. Goutorbe, F. Lucazeau and A. Bonneville

3.2.4 Instantaneous source versus continuous source

Forward modelling has shown that the circulation time tc has a non negligible influence on the

theoretical temperature perturbation δT (e.g., Luheshi 1983). However, the opposite is not neces-

sarily true, i.e., tc may not have such an influence when extrapolating T∞ from TBHT measurements

– as noticed by, e.g., Funnell et al. (1996). On our data, differences between continuous and ins-

tantaneous corrections are usually not larger than a few percent for tc < 10 hours (Fig. 6). Beyond

10 hours, tc seems to have more effect on the predictions, at least for the cylinder source models,

but too few data are available to draw a firm conclusion.

3.2.5 Two-component model versus cylinder source models

As pointed out by Shen & Beck (1986), the constant temperature and constant heat supply versions

of the two-component model give virtually identical results in practical applications (Fig. 7a). On

the other hand, the two-component model extrapolate equilibrium temperatures that are largely

higher than the other corrections (Fig. 7b). Such discrepancies are puzzling, and close inspection

shows that the two-component model often extrapolate unrealistic values of T∞ , as this can be seen

on the examples of Fig. 7c. Hence, although the theoretical background of the method is certainly

more accurate, its complexity and the number of parameters to estimate may actually make it less

robust on real-world cases.

3.2.6 Sensitivity with respect to r2 /4κr

As can be seen in Eqs. (1), (3), (6) and (8), the main parameter of the line and cylinder source mo-

dels is τ ≡ a2 /4κr . As we have only indirect estimations, it is important to quantify the sensitivity

of the extrapolations with respect to τ . The relative uncertainty on τ is

sµ ¶2 µ ¶2
δτ δa δκr
= 2 + . (16)
τ a κr

Assuming that a is relatively well constrained thanks to direct log measurement (δa/a ∼ 5-10%)

and the indirect estimate of κr much less reliable (δκr /κr ∼ 15-25%), we have δτ /τ ∼ 15-30%.

This generates in turn some variability on the extrapolated T∞ . The sensitivity increases with τ ,
 BHT correction methods 11

staying in most cases below ±3% for a relative uncertainty on τ of 15%, and reaching ±5% with

δτ /τ = 30% (Fig. 8). Therefore it seems that the sensitivity of the corrections remains reasonably

low, even with thermal diffusivity κr poorly constrained.

4 SUMMARY AND CONCLUSIONS

We have presented and classified several analytical BHT corrections of various complexity levels.

By performing inter-comparisons on a real-world oil exploration data set from Australia, we have

drawn the following conclusions:

(1) Extrapolated undisturbed temperatures from all methods are qualitatively in agreement

with measurements from drillstem tests (within 5-10◦ C) in most of the wells.

(2) As expected, the widely used Horner method breaks down when its underlying assumption

(a2 /4κr te ≪ 1) is not valid.

(3) The line source models, which are sometimes used instead of the Horner method, suffer

from an unphysical delay time that actually restrain their applicability domain.

(4) It seems that the circulation time cannot be neglected beyond 10 hours, at least for the

cylinder models, but this conclusion does not lie on a firm statistical basis.

(5) Taking into account the contrast of thermal properties between circulating mud and sur-

rounding rocks is the most realistic way of modelling the problem. However the mathematical

form of the correction reaches a level of complexity, and requires a large number of parameters,

making it delicate to use on practical cases.

The domains of applicability of the correction methods are summarized in Table 2. The cylin-

der source models are our preferred corrections, as they take into account some geometrical and

thermal characteristics of the circulation process, have few theoretical restrictions regarding their

applicability domain and keep a simple analytical form for practical use.

5 ACKNOWLEDGEMENTS

Michael Wiltshire, the director of the company Wiltshire Geological Services°,

R provided us oil

exploration data from Australia, making this work possible.

12 B. Goutorbe, F. Lucazeau and A. Bonneville

References

Beck, A., 1988. Methods for determining thermal conductivity and thermal diffusivity, in Hand-

book of terrestrial heat-flow density determination, edited by R. Haenel, L. Rybach, & L. Ste-

gena, Kluwer Academic Publishers.

Brigaud, F., 1989. Conductivité thermique et champ de température dans les bassins

sédimentaires à partir des données de puits, Ph.D. thesis, Centre Géologique et Géophysique,

Université des Sciences et Techniques du Languedoc, Montpellier, France, 414 p.

Bullard, E., 1947. The time taken for a borehole to attain temperature equilibrium, Mon. Not. R.

Astron. Soc., Geophys. Suppl. 5, 127–130.

Carslaw, H. & Jaeger, J., 1959. Conduction of heat in solids, Clarendon, Oxford, 510 p.

Cramer, W. & Leemans, R., pers. comm. The climate database version 2.1, available from World

Wide Web: http://portal.pik-potsdam.de/members/cramer/climate.html.

Dowdle, W. L. & Cobb, W., 1975. Static formation temperature from well logs – an empirical

method, J. Petr. Tech., pp. 1326–1330.

Funnell, R., Chapman, D., Allis, R., & Armstrong, P., 1996. Thermal state of the Taranaki Basin,

New Zealand, J. Geophys. Res., 101(B11), 25197–25215.

Goutorbe, B., Lucazeau, F., & Bonneville, A., 2006. Using neural networks to predict thermal

conductivity from geophysical well logs, Geophys. J. Int., 166, 115–125, doi: 10.1111/j.1365-

246X.2006.02924.x.

Hermanrud, C., Cao, S., & Lerche, I., 1990. Estimates of virgin rock temperature derived from

BHT measurements: Bias and error, Geophysics, 55(7), 924–931.

Leblanc, V., Pascoe, L., & Jones, F., 1981. The temperature stabilization of a borehole, Geo-

physics, 46(9).

Levitus, S. & Boyer, T., 1994. World Ocean Atlas 1994, Volume 4: Temperature, NOAA ATLAS

NESDIS 4.

Lide, D., 2004. CRC Handbook of chemistry and physics, 85th Edition, CRC Press.

Luheshi, M., 1983. Estimation of formation temperature from borehole measurements, Geophys.

J. R. Astr. Soc., 74, 747–776.

 BHT correction methods 13

Middleton, M., 1982. Bottom-hole temperature stabilization with continued circulation of drilling

mud, Geophysics, 47(12), 1716–1723.

Shen, P. & Beck, A., 1986. Stabilization of bottom-hole temperature with finite circulation time

and fluid flow, Geophys. J. R. Astr. Soc., 86, 63–90.

Townend, J., 1999. Heat flow through the West Coast, South Island, New Zealand, New Zeal. J.

Geol. Geophys., 42, 21–31.

Vosteen, H. & Schellschmidt, R., 2003. Influence of temperature on thermal conductivity, thermal

capacity and thermal diffusivity for different types of rock, Phys. Chem. Earth, 28, 499–509.

Zschocke, A., 2005. Correction of non-equilibrated temperature logs and implications for

geothermal investigations, J. Geophys. Eng., 2, 364–371, doi: 10.1088/1742-2132/2/4/S10.

14 B. Goutorbe, F. Lucazeau and A. Bonneville

LIST OF FIGURES

1 Location of oil exploration wells with temperature data

2 Available temperature measurements versus depth

3 Surface temperatures, TDST and T∞ versus depth

4 Relative differences between CLS/CCS and Horner predictions

5 (a) Example of T∞ extrapolation using CLS and CCS models; (b) Delay time td as

a function of a2 /4κr ; (c) Relative differences between line source and cylinder source

models

6 Relative differences between predictions from continuous and instantaneous models

7 (a) T∞ from the constant temperature version versus T∞ from the constant supply

of heat version of the two-component model; (b)Relative differences of predicted T∞

between: two-component and ICS, two-component and CCS; (c)T∞ extrapolation from

series of TBHT measurements taken from wells Drummer 1 and Warb 1A, using ICS and

two-component model

8 Relative differences between predictions with modified τ and initial τ , versus ini-

tial τ , using ICS model

 BHT correction methods 15

Table 1. List of parameters needed for BHT corrections (ILS: instantaneous line source; CLS: conti-
nuous line source; ICS: instantaneous cylinder source; CCS: continuous cylinder source; 2-comp: two-
component).

Par. Description Horner ILS CLS ICS CCS 2-comp

tc Mud circulation time × × × ×

a Borehole radius × × × × ×
κr Rock thermal diffusivity × × × × ×
λr Rock thermal conductivity ×
ρcr Rock vol. heat capacity ×
κm Mud thermal diffusivity ×
λm Mud thermal conductivity ×
ρcm Mud vol. heat capacity ×

Table 2. Summary of the restrictions for the correction methods. The two-component model has not been
included because its complexity is a serious obstacle to its use.

Correction Restrictions

Horner a2 /4κte ≪ 1
ILS a2 /4κte < 1
CLS a2 /4κte < 1
ICS tc < 10 hours?
CCS −
16 B. Goutorbe, F. Lucazeau and A. Bonneville
110°E 155°E
8°S

Elevation / water depth (m)

-7000 0 3000
44°S

Figure 1. Location of oil exploration wells in which temperature data are available. Crosses: wells having
one or two groups of multiple BHT measurements; open triangles: three or four groups; open circles: five
groups or more.
 BHT correction methods 17

Temperature (°C)
50 100 150

0
1000
2000
Depth (mbsf)
3000
4000
5000

Figure 2. Available temperature measurements versus depth. TBHT : grey crosses (te < 30 hours) and open
circles (te ≥ 30 hours); TDST and surface temperatures: black dots. Surface temperatures from Levitus &
Boyer (1994) (offshore) and Cramer & Leemans (pers. comm.) (onshore) databases. As can be expected,
the geotherm defined by BHT measurements with a high elapsed time te is closer to the TDST geotherm.
18 B. Goutorbe, F. Lucazeau and A. Bonneville

Temperature (°C)
0 100 200 300 400 500
0

TDST / surface temp.

Horner correction
ILS correction
1000

CLS correction
ICS correction
CCS correction
Two-c. (T=cst) corr.
Depth (mbsf)

Two-c. (Q=cst) corr.

2000

Barnett 1
Elder 1 Hampton 1
3000

Blina 1
Chervil 1
Blacktip 1
Goodwyn 1
Echo 1 Lesueur 1
4000

0 100 200 300 400 500

0
1000
Depth (mbsf)
2000

Naccowlah 1
Thylacine 2
3000

Woodada 1
North
Rankin 5
Tidepole 1 Yarrada 1
4000

Montague 1 Petrel 4 Wanaea 1

Figure 3. Surface temperatures, TDST and T∞ (from correction of TBHT ) versus depth, in wells where both
TBHT and TDST are available. Temperatures of a given well are shifted of +50◦ C with respect to previous
well. The size of symbols for T∞ corresponds roughly to an uncertainty of ±5◦ C. The two versions of
the two-component model correspond respectively to the constant temperature and constant supply of heat
assumptions (see section 2.3). At some depths there are missing methods, as they yielded Q or ∆T0 > 0
(see section 3.1).
 BHT correction methods 19

30
Relative difference (%)
20
10
0

0.01 0.1 1 10
a²/4κrtemin

Figure 4. Relative differences of predicted T∞ between: CLS and Horner corrections (grey dots), CCS and
Horner corrections (open circles), versus a2 /4κr te (with smallest te of the TBHT series taken).

50
20

a) b) c)
100

Relative difference (%)

40
15
Temperature (°C)

1h
t c=
30

h
10
80

t d (h)

t c=
10

td
{ Tt =8h=51°C
20

BHT
e tc=5h
60

10

{ tT=4.5h
e
BHT =48.3°C
40

0
0

0 10 20 30 40 0 5 10 15 20 0.01 0.1 1 10
t e (h) a²/4κr (h) a²/4κrtemin

Figure 5. (a) Example of T∞ extrapolation from a couple of TBHT measurements (black dots) taken from
well Bridgewater Bay 1. tc = 3 hours, a = 0.21 m, κr = 4.49 · 10−7 m2 .s−1 . The CLS model (solid line)
predicts an unrealistic T∞ = 141◦ C, while the CCS model (dashed line) predicts T∞ = 59◦ C. (b) Delay
time td of the CLS model as a function of a2 /4κr , for various tc . (c) Relative differences of predicted T∞
between: ILS and ICS corrections (grey dots), CLS and CCS corrections (open circles), versus a2 /4κr te
(with smallest te of the TBHT series taken).
20 B. Goutorbe, F. Lucazeau and A. Bonneville

10
8
Relative difference (%)
6
4
2
0

0.1 1 10 100
t c (h)

Figure 6. Relative differences of predicted T∞ between: CLS and ILS corrections (grey dots), CCS and
ICS corrections (open circles), versus mud circulation time tc . There are less data than other comparisons
because we have only considered temperatures with available tc .
 BHT correction methods 21

200
200

80
Temp. (°C) - Two-c. model, T=cst

a) b) c)

Relative difference (%)

150

60

Temperature (°C)
Drummer 1

100
100

40
20
50

Warb 1A

50
0
0

0 50 100 150 200 0.01 0.1 1 10 0 20 40 60 80

Temp. (°C) - Two-c. model, Q=cst a²/4κrtemin t e (h)

Figure 7. (a) T∞ from the constant temperature version versus T∞ from the constant supply of heat version
of the two-component model (open circles), with ±5% relative difference shown (dashed lines); (b) Relative
differences of predicted T∞ between: two-component and ICS corrections (grey dots), two-component and
CCS corrections (open circles), versus a2 /4κr te (with smallest te of the TBHT series taken); (c) Example of
T∞ extrapolation from series of TBHT measurements (black dots) taken from wells Drummer 1 and Warb
1A. Parameters for Drummer 1: tc = 1.25 hours, a = 0.16 m, λr = 2.16 W.m.−1 .K−1 , ρcr = 2.8 · 106
J.m−3 .K−1 , κr = 7.8 · 10−7 m2 .s−1 . Parameters for Warb 1A: tc = 2.5 hours, a = 0.23 m, λr = 2.22
W.m.−1 .K−1 , ρcr = 2.9 · 106 J.m−3 .K−1 , κr = 7.7 · 10−7 m2 .s−1 . The two-component model (solid
lines) predicts unrealistic T∞ of 266◦ C for Drummer 1 (≈ TBHT + 180◦ C!) and 89◦ C for Warb 1A (though
last measured TBHT = 65◦ C with te ≫ tc ), while the ICS model (dashed lines) predicts respectively
T∞ = 161◦ C and 68◦ C.
22 B. Goutorbe, F. Lucazeau and A. Bonneville

10
Relative difference (%)
5
0
-5
-10

0 5 10
a²/4κr (h)

Figure 8. Relative differences between predictions with modified τ (say, τ ′ ) and initial τ , versus initial τ ,
using ICS model. Filled triangles: τ ′ = 0.7τ ; filled triangles: τ ′ = 0.85τ ; open circles: τ ′ = 1.15τ ; open
circles: τ ′ = 1.30τ . The sensitivity of the other models (line and cylinder source) is similar.