Aquaculture Research, 2003, 34, 581^584
SHORT COMMUNICATION
The thermal growth coefficient (TGC) model of fish
growth: a cautionary note
Malcolm Jobling
NFH, University of Troms,Troms, Norway
Correspondence: Malcolm Jobling, NFH, University of Troms, 9037 Troms, Norway. E-mail: malcolmj@nfh.uit.no
Since the time it was proposed in 1981 (Iwama &
Tautz 1981), the thermal growth coe⁄cient model
(TGC model) has been widely used for predictive pur-
poses in production planning (discussed by Iwama
1996; Alan
r
, Kadri & Paspatis 2001). The popular-
ity of the model relates to its ease of use and £exibil-
ity; growth data collected for ¢sh of given size at one
temperature can be used to predict the growth of ¢sh
of a di¡erent size when held at other temperatures.
This will not, however, apply across the full range of
conditions over which farmed ¢sh may be reared,
and uncritical use of the TGC model can lead to ser-
ious projection errors being made. The purpose of
this paper is to point out some of the limitations of
the model, and to emphasize conditions under which
caution should be exercised in use of the TGC for
growth prediction. The TGC model has the following
basic form,
pffiffiffiffiffi 3pffiffiffiffiffi
3
W t ¼ W 0 þ ½ðT=1000Þ
t
where T is temperature in 1C, and t is time in
days. This means that, for ¢sh reared at constant
temperature, a plot of cube root of weight against
time gives a straight line. The thermal growth coe⁄-
cient (TGC) is calculated in relation to ‘degree-days’
(T
t):
pffiffiffiffiffi pffiffiffiffiffi
TGC ¼ ½ð3 W t 3 W 0 Þ=ðT
tÞ
1000
and, when a growth prediction is made, the formula
becomes:
pffiffiffiffiffi
Wt ¼ f3 W 0 þ ½ðTGC=1000Þ
ðT
tÞg3
r 2003 Blackwell Publishing Ltd
When the model takes this form, there are three basic
assumptions:
 growth increases in a steady and predictable man-
ner with increasing temperature;
 the length (L)^weight (W) relationship is
WpL3;
 growth in length (for any temperature) is constant
over time (i.e. L increases linearly over time).
All these assumptions may be violated under
some of the conditions to which farmed ¢sh are
exposed.
The ¢rst major problem is that growth does
not show a steady increase with increasing tempera-
ture, but the rate^temperature relationship for
growth is a‘bell-shaped’curve (Jobling1994). In other
words, at the lower end of the temperature range,
growth increases with increasing temperature, but
growth rate reaches a peak at an intermediate tem-
perature and then starts to decline. The conse-
quences of this for the TGC can be illustrated using
the growth data for Baltic salmon Salmo salar L., pre-
sented by Koskela, Pirhonen & Jobling (1997) (Fig.1a).
The salmon held at 11 1C grew from E40 g to 66.5 g
in 42 days, giving an estimate for the TGC of 1.361. If
this TGC is used to predict the growth of the salmon
at 20.5 1C, the ¢sh are estimated to grow from E40 g
to 97 g over the 42-day period. Examination of the
rate^temperature curve in Fig. 1a shows that rates of
growth are the same at 11 1C and 20.5 1C, so the
growth prediction based upon the TGC calculated for
¢sh held at 11 1C (TGC 51.361) results in a consider-
able overestimation (estimated 97 g vs. achieved
66.5 g).
581
TGC model of ¢sh growth M Jobling
Figure1 In£uence of temperature on the growth of
Baltic salmon Salmo salar L. (a) The growth rate^tem-
perature relationship. The dashed line indicates that the
speci¢c growth rate (SGR) at 11 1C is the same as that
at 20.5 1C {SGR 5 [(ln Wt  ln W0) t^1]
100}. (b) The
change in the thermal growth coe⁄cient (TGC) with in-
creasing temperature. The arrow indicates the estimate
of the TGC at 20.5 1C (E0.75) [calculated from data in
Koskela et al. (1997)].
The TGC for salmon growing from E40 g to 66.5 g
over 42 days at 20.5 1C is 0.73. This is close to the va-
lue indicated on the curve constructed by plotting
the TGC^temperature relationship calculated using
the data given by Koskela et al. (1997) (Fig. 1b). This
example indicates that the TGC is not independent of
temperature over the full thermal range. TGC de-
clines at the higher end of the temperature range.
An additional example of this can be seen in the
growth data for small Atlantic cod Gadus morhua L.
(E6^50 g) presented by Bj˛rnsson, Steinarsson &
Oddgeirsson (2001). The results of three experiments
are given, covering the temperature range 7.5^16 1C.
When TGC is calculated and plotted against tempera-
ture (Fig. 2), it can be seen that TGC is independent of
temperature within the range 7.5^12.5 1C, but TGC
582 r 2003 Blackwell Publishing Ltd, Aquaculture Research, 34, 581^584
Aquaculture Research, 2003, 34, 581^584
Figure 2 In£uence of temperature on the thermal
growth coe⁄cient (TGC) of Atlantic cod Gadus morhua
L. (weight range E6^50 g) [calculated from data given
in Bj˛rnsson et al. (2001)].
Figure 3 In£uence of temperature on the thermal
growth coe⁄cient (TGC) of Atlantic cod Gadus morhua
L. (weight E2000 g) [calculated from data given in
Bj˛rnsson et al. (2001)].
declines above this range. This means that the TGC
calculated for the ¢sh held at 7.5 1C can be used for
growth prediction up to 12.5 1C, but erroneous re-
sults will be obtained if this TGC is used for prediction
of growth of ¢sh held at higher temperatures.
The ‘bell-shaped’ form of the growth rate^tempera-
ture relationship also means that TGC is high at
low temperature. For example, the TGC reported
for Atlantic salmon parr by Bendiksen, Jobling &
Arnesen (2002) was higher at 2 1C than at 8 1C. The
phenomenon can also be illustrated using the
growth data for large Atlantic cod (E2000 g) pre-
sented by Bj˛rnsson et al. (2001). The in£uence of
temperature on the growth of large cod was exam-
ined over the temperature range 1^16 1C.When TGC
is calculated and plotted against temperature (Fig. 3),
Aquaculture Research, 2003, 34, 581^584
it can be seen that TGC is highest at lowest tempera-
ture, declines and is independent of temperature
within the range 5^9.5 1C and then starts to decline
again above this range.
These examples illustrate the following points:
 aTGC calculated under one set of temperature con-
ditions cannot be used to make growth predictions
across the full range of temperatures encompassed
by the growth^temperature curve;
 the TGC calculated at one temperature can only be
used to make estimates at another provided that
the two temperatures fall on part of the growth^
temperature curve that ful¢ls the criterion of ‘y.
growth increases in a steady and predictable man-
ner with an increase in temperature’;
 TGC is at its most versatile for prediction of growth
over part of the ascending limb of the growth^
temperature curve. Here, a single TGC can be ap-
plied over a range of several temperature units.
The second problem with the TGC model relates to
the assumption regarding the length^weight rela-
tionship. The assumption that W p L3 means that
the TGC should only be used for making growth pre-
dictions when the condition factor [K 5 (W L^3)

100] of the ¢sh remains stable over time. Under
rearing conditions in which the increase in body
mass becomes uncoupled from the increase in
length, and there is a change in K over time, these
changes will be re£ected in the estimate of the TGC.
This problem can be illustrated using growth data
taken from a study designed to examine the compen-
satory growth response of Atlantic cod (Jobling,
Mely, dos Santos & Christiansen 1994). Groups of
cod were established using length (o60 and
60^70 cm) and condition factor (Ko0.8, 0.8^0.9 and
40.9) as the selection criteria. The initial weights of
the ‘small, high K’ ¢sh (o 60 cm, 40.9: E1650 g) and
the‘large, low K’¢sh (60^70 cm, o0.8: E1680 g) were
similar. According to the assumptions of the model,
these ¢sh should have similar TGCs; this is because
their initial weights were very similar. The growth re-
sponses of the ¢sh within each category were exam-
ined.Within each length group, the ¢sh that were in
the poorest condition, i.e. had the lowest initial K, dis-
played the most rapid rates of weight gain. This was
re£ected in the estimate of the TGC for each group
(Fig. 4): TGC was higher for the ¢sh in poor condition
than for those with a higher initial K.
The TGC of the ‘small, high K’ ¢sh (o 60 cm, 40.9)
was 4.57 and that of the ‘large, low K’ ¢sh (60^70 cm,
o0.8) was 6.92, and the cod within these groups dif-
fered in weight by E500 g after 6 weeks of growth.
r 2003 Blackwell Publishing Ltd, Aquaculture Research, 34, 581^584
TGC model of ¢sh growth M Jobling
Figure 4 In£uence of body size (length in cm) and
condition [K 5 (W L^3)
100] on the thermal growth
coe⁄cient (TGC) of Atlantic cod Gadus morhua L. under-
going catch-up or compensatory, growth [calculated
from Jobling et al. (1994)].
This contradicts the model assumption and demon-
strates the uncoupling of length growth and the in-
crease in body mass that may occur in ¢sh that are
undergoing periods of catch-up or compensatory
growth. This type of growth may be observed in ¢sh
that are recovering from a voluntary or involuntary
fast, from periods of restricted feeding or are in the
phase of recovery after spawning (Jobling 1994).
From the above, it can be deduced that growth of ¢sh
undergoing catch-up growth would be underesti-
mated if projections were made based upon a TGC cal-
culated for the growth displayed by well-fed ¢sh in
‘good condition’ (high K) and growing ‘normally’.
To conclude, the TGC model is a simple and £exible
model that can be used for growth prediction and
production planning purposes under a variety of con-
ditions, but the model should not be used uncritically.
Extrapolations should not be made outside the tem-
perature range that ful¢ls the criterion of ‘y.growth
increases in a steady and predictable manner with an
increase in temperature’, and spurious results will
also be generated if growth projections are attempted
for ¢sh that are experiencing marked changes in
body condition, i.e. increase in body mass is un-
coupled from growth in length.
References
Alan
r
A., Kadri S. & Paspatis M. (2001) Feeding manage-
ment. In: Food Intake in Fish (ed. by D Houlihan,T. Boujard
& M. Jobling), pp. 332^353. Blackwell Science, Oxford.
Bendiksen E.—., Jobling M. & Arnesen A.M. (2002) Feed in-
take of Atlantic salmon parr Salmo salar L. in relation to
temperature and feed composition. Aquaculture Research
33, 525^532.
583
TGC model of ¢sh growth M Jobling
Bj˛rnsson B., Steinarsson A. & Oddgeirsson M. (2001) Opti-
mal temperature for growth and feed conversion of imma-
ture cod (Gadus morhua L.). ICES Journal of Marine Science
58, 29^38.
Iwama G.K. (1996) Growth of salmonids. In: Principles of
Salmonid Culture (ed. by W. Pennell & B.A. Barton),
pp. 467^515. Elsevier, Amsterdam.
Iwama G.K. & Tautz A. (1981) A simple growth model for
salmonids in hatcheries. Canadian Journal of Fisheries
and Aquatic Sciences 38, 649^656.
Jobling M. (1994) Fish Bioenergetics. Chapman & Hall, London.
584 r 2003 Blackwell Publishing Ltd, Aquaculture Research, 34, 581^584
Aquaculture Research, 2003, 34, 581^584
Jobling M., Mely O.H., dos Santos J. & Christiansen B. (1994)
The compensatory growth response of the Atlantic cod:
e¡ects of nutritional history. Aquaculture International 2,
75^90.
Koskela J., Pirhonen J. & Jobling M. (1997) Feed intake,
growth rate and body composition of juvenile Baltic
salmon exposed to di¡erent constant temperatures.
Aquaculture International 5, 351^360.
Keywords: growth, temperature, condition factor,
Atlantic salmon, Atlantic cod, production planning