Journal of Thermal Biology 85 (2019) 102408

Contents lists available at ScienceDirect

Journal of Thermal Biology

journal homepage: www.elsevier.com/locate/jtherbio

Black globe temperature from meteorological data and a bioclimatic T

analysis of the Brazilian Northeast for Saanen goats
Thieres George Freire da Silvaa,c,∗, Girlene Cordeiro Lima Santosa, Ana Maria Cabral Duartea,
Sílvia Helena Nogueira Turcob, José Francisco da Cruz Netoc,
Alexandre Maniçoba da Rosa Ferraz Jardimb, Thalyta Soares dos Santosb
a
Academic Unit of Serra Talhada, Federal Rural University of Pernambuco, Serra Talhada, Pernambuco, Brazil
b
Department of Agricultural Engineering, Federal University of the São Francisco Valley, Juazeiro, Bahia, Brazil
c
Department of Agricultural Engineering, Federal Rural University of Pernambuco, Recife, Pernambuco, Brazil

A R T I C LE I N FO A B S T R A C T

Keywords: The black globe temperature (BGT) is not a common measurement for weather station networks, despite having
Time scales great relevance to bioclimatic studies. The aim of this study was to propose equations for estimating the BGT,
Indices and statistical errors using meteorological data for different time scales and a bioclimatic evaluation of the Brazilian Northeast for
Regression breeding Saanen dairy goats. The data used in elaborating the equations were collected between 1 November
BGHI
2014 and 31 October 2017. Data for BGT, incident global solar radiation (SR), air temperature (AT), relative
humidity (RH) and wind speed were handled on a daytime, night-time, daily and monthly scale. One half of the
database was used to adjust the equations and the other half in the evaluation. The bioclimatic diagnosis of the
Brazilian Northeast was carried out based on mean monthly values of the black globe temperature and humidity
index (BGHI) estimated for the four seasons of the year. For the daytime scale, an equation based on AT
(BGT = 1.3897.AT-5.4421, r2 = 0.80) and a multiplicative model combining the effects of AT and SR
(BGT = [1.3897.AT-5.4421] (0.0384.ln(SR)+0.7935], r2 = 0.91) were obtained. AT adjusted well for BGT on
the night-time scale (BGT = 0.995.AT-0.6964, r2 = 0.99), daily scale (BGT = 1.1641.AT-1.5941, r2 = 0.97) and
monthly scale (BGT = 1.1550.AT-1.3498, r2 = 0.98). The BGT can therefore be calculated from AT and/or SR
for the daytime scale, and from AT only for the night-time, daily and monthly scales. In general, the west and
centre-south of the state of Bahia offer the animals the most thermal comfort during each season of the year. In
the state of Maranhão, heat stress occurs throughout the year, with the BGTI predominately in the range of
85–95. As such, strategies to combat heat stress should be encouraged to minimise the negative effects of climate
on milk production in Saanen goats, and favour the milk-production chain in the northeast of Brazil.

1. Introduction The black globe temperature index (BGTI) is a good example of an

index applied to bioclimatic diagnosis, but which requires data of the
Thermal stress is caused by a combination of environmental factors black globe temperature (BGT) (Lemke and Kjellstrom, 2012;
that affect the physiological parameters of animal intake, reproduction, Pezzopane et al., 2019). The BGT is an indirect measurement of the
lactation and production (Zobel et al., 2015; Das et al., 2016; Ribeiro radiant heat load of the environment obtained with a thermometer
et al., 2016; Curtis et al., 2017; Sejian et al., 2018). Bioclimatic indices installed inside a hollow copper sphere painted matte black, 0.05 m
have been developed to express the comfort/discomfort of the en- thick and 0.15 m in diameter, that allows the thermal sensation of an
vironment to animals (Lemke and Kjellstrom, 2012; Hajizadeh et al., animal to be inferred (Abreu et al., 2011; Hajizadeh et al., 2017).
2017), so that they can be used for improving agricultural planning and The absence of black globe thermometers in the weather stations of
decision-making (Silva et al., 2008; da Silva et al., 2010; McManus the meteorological monitoring networks of public State and Federal
et al., 2016; Pezzopane et al., 2019). institutions and private companies in Brazil makes it difficult to assess

∗
Corresponding author.Academic Unit of Serra Talhada, Federal Rural University of Pernambuco, Serra Talhada, Pernambuco, Brazil.
E-mail addresses: [email protected] (T.G.F. da Silva), [email protected] (G.C.L. Santos), [email protected] (A.M.C. Duarte),
[email protected] (S.H.N. Turco), [email protected] (J.F.d. Cruz Neto), [email protected] (A.M.d.R.F. Jardim),
[email protected] (T.S. dos Santos).

https://doi.org/10.1016/j.jtherbio.2019.102408
Received 28 April 2019; Received in revised form 25 August 2019; Accepted 25 August 2019
Available online 26 August 2019
0306-4565/ © 2019 Elsevier Ltd. All rights reserved.
T.G.F. da Silva, et al. Journal of Thermal Biology 85 (2019) 102408

the bioclimatic risk of different regions. Developing equations for es- intervals. Two globe thermometers were installed at the station (Type T
timating the BGT from meteorological variables can be useful in cal- copper-constantan thermocouple, Campbell Scientific, Logan, Utah,
culating bioclimatic indices in different regions (Silva et al., 2008; da USA, inserted inside a black-painted copper globe of high absorptivity,
Silva et al., 2010; Dimiceli et al., 2012; Mendes et al., 2014). Besides, diameter 150 mm, Testo, Inc., West Chester, PA, USA) at a height of 1.5
the black globe temperature sensor can be costly, making it difficult to and 2.5 m above the ground to measure the BGT in oC. Two aspirated
install in many locations (Dimiceli et al., 2012). Even more-expressive psychrometers, comprising copper-constantan thermocouples (type T,
results can be achieved with equations for different time scales, since Campbell Scientific, Logan, Utah, USA), measured the dry-bulb tem-
the data supplied by the institutions varies greatly with regard to time perature (DBT) (equal to the air temperature, AT) and the wet-bulb
(i.e. hourly, daily and monthly). The application of these equations in temperature (WBT), at the same height as the two globe thermometers,
estimating the BGT is very important for evaluating daytime and night- with readings output in oC. A pyranometer (CS 300 Pyranometer,
time energy exchanges in the interface animal-environment, as well as Campbell Scientific, Logan, Utah, USA) at a height of 2.5 m was used to
assess most promising producing regions and species. In addition, it measure incident global solar radiation (SR), with output values in W
allows the level of environmental modification required for animal m−2. An anemometer (03002 R.M. Young Wind Sentry Set, Campbell
production to be defined, and measure production costs (Das et al., Scientific, Logan, Utah, USA), placed at a height of 2 m, measured the
2016; Curtis et al., 2017; Sejian et al., 2018). wind speed in m s−1.
One activity that is being increasingly exploited is dairy goat The mean value of the DBT and WBT data from the two psy-
farming, due to the high nutritional value of the milk and its derivatives chrometers, measured at 1.5 and 2.5 m, was used to estimate the re-
(Irano et al., 2012; Zambom et al., 2013). Although important, espe- lative humidity (RH,%) by means of Eq. (1) (Pereira et al., 2002).
cially in the northeast of Brazil, its inclusion in agro-industry depends
e= (RH/eS) 100 (1)
on the use of activities that alter the low yield of the animals in hot
environments, i.e. bioclimatic analysis and the adoption of genetically where: e − the current water vapor pressure in air, kPa; and, eS - sa-
more-productive species (Lôbo and Silva, 2005; Pequeno et al., 2017; turation water vapor pressure in air, kPa.
Arnal et al., 2018). As Brazil is a tropical country, many native Eur- The value for “e” was calculated with Eq. (2) (Pereira et al., 2002).
opean animals from a temperate climate do not adapt well, i.e. Saanen
e= eSW − A Pa (DBT − WBI) (2)
goats, which specialise in milk production. Exotic breeds with good
dairy performance are concentrated in the south and southeast of where: eSW - saturation water vapor pressure in air based on the WBT,
Brazil, and exploiting them in the northeast therefore depends on a kPa; A - psychrometric constant (8.0-10−4 °C-1); and, Pa - estimated
bioclimatic diagnosis. Pequeno et al. (2017) developed a mathematical atmospheric pressure based on the local altitude, kPa.
model based on maximum air temperature to estimate milk production The values for “eS” and “eSU” were calculated using the mean va-
during peak lactation of the animals, and mapped the areas most ex- lues for DBT and WBT respectively (Pereira et al., 2002):
posed to heat stress in the Brazilian Northeast. However, these authors
eS = {[ 0.6108 exp(17.27 DBT )]/[ DBT + 237.3 ]} (3)
did not consider the BGTI, the index that best represents the thermal
environment of the animals. eSU = {[ 0.6108 exp(17.27 WBT )]/[ WBT + 237.3 ]} (4)
In the Northeast of Brazil, the goat herd amounts to 9,092,724 head,
composed of different races and aptitudes, meat and dairy, which re- The data for BGT, SR, AT, RH and WS of the 1096 days (157,824
sults in a large discrepancy in the goat milk yield between the states of values by variable) were processed every 10 min (separating them into
this region (IBGE, 2017). The wide diversity of climate (Alvares et al., daytime and night-time), daily (from all 144 data for each day, with
2014) and predominant adoption of an pastoral/extensive farming average of BGT, AT, RH and WS values, and integration of the SR data)
system causes heat stress in most of the animals (Silva et al., 2008; da and monthly (mean value of all the daily data for each month). For the
Silva et al., 2010; Mendes et al., 2014), and good performance depends daily and monthly scales, the unit for each variable was expressed in oC,
on bioclimatic assessments together with correct management. This MJ m−2 day−1, oC, % and m s−1 respectively.
type of action can boost local dairy farming, with added income for the Half of the database for each time scale, comprising alternate days
population, and encourage producers to remain on the farm. or months, was used to establish linear and non-linear regressions be-
The aim of this study therefore, was to propose equations for esti- tween the BGT (dependent variable) and the meteorological elements
mating the BGT using meteorological data for different time scales SR, AT, RH and WS (independent variables). The regressions were
(daytime, night-time, daily and monthly) and a bioclimatic evaluation carried out using the Sigmaplot v14 (Systat software, Inc., 2019). The
of the Brazilian Northeast for breeding Saanen goats, with a view to significance of the equations and their respective coefficients were
guiding practices for improving animal welfare and enhancing the evaluated by Student's F-test and Student's t-test respectively, at a level
production of goat milk. of 5%, and the coefficient of determination (r2) was calculated. Based
on r2, the variables were then ordered as to their importance in ex-
2. Material and methods plaining the BGT: 1st, 2nd, 3rd and 4th. The equations with higher
values for r2 were used in elaborating a multiplicative model. This
2.1. Estimating the BGT from meteorological data model comprised two equations, as per a structure described by Silva
et al. (2007):
The data used in elaborating and evaluating mathematical equa-
Y= f (a) f (b) (5)
tions for estimating the BGT were collected between 1 November 2014
and 31 October 2017 (1096 days) from a micro-weather station in- where: Y - dependent variable; f(a) - dependent function of x with linear
stalled in the district of Floresta (8.3°S, 38.5°W at 381°m), in the state of or non-linear fluctuation; and f(b) - dependent function of x with linear
Pernambuco, Brazil. The micro-weather station of 4 m-high with cou- or non-linear fluctuation;
pled sensors (black globe thermometers, aspirated psychrometers, In this step, f(a) was represented by the equation with the greatest r2
pyranometer and anemometer) was installed in the centre of an area of generated between the BGT and one of the independent variables (SR,
flat bare ground with no obstacles nearby, to avoid affecting the sensor AT, RH or WS), i.e. the first most important equation in explaining the
readings. Each sensor was connected to a multiplexer (AM16/32B-ST- BGT. This function f(a) was used in calculating the BGT (estimated
SW, Campbell Scientific, Logan, Utah, USA) and a data acquisition black globe temperature, BGTe), and the ratio was then obtained from
system (CR1000-ST-SW-NC, Campbell Scientific, Logan, Utah, USA), the actual observed values (observed black globe temperature, BGTo),
which registered data every 60 s, and stored the mean values at 10-min giving BGTo/BGTe. This ratio is the residual unexplained by the first

2
T.G.F. da Silva, et al. Journal of Thermal Biology 85 (2019) 102408

independent variable. The BGTo/BGTe ratio was correlated with the Therefore:
second independent variable having the greatest r2. In cases where
TPO = {[( 237.3 log(e/0.6108 )]/[ 7.5 − log(e/0.6108 )] (8)
insertion of the variable did not contribute expressively to the multi-
plicative model, the third followed by the fourth most important vari- The BGTIX values for the Brazilian Northeast were compared to the
able was used to form f(b). This procedure was adopted to obtain the ideal range of BGTI values for Saanen dairy goats. This range was es-
variable that explains the variation in BGT which was not explained by tablished from information on the critical air temperature for Saanen
the first variable. dairy goats quoted by Pequeno et al. (2017), and on RH for domestic
The generated equations and multiplicative model were evaluated animals quoted in Baêta and Souza (2010) and in Silva (2000).
from their application to the other half of the data set, with the results According to Pequeno et al. (2017), Saanen goats significantly re-
compared to the observed data by means of the mean bias error (MBE) duce milk production during the period of maximum lactation when TX
and root mean square error (RMSE), as well as the statistical indices of is greater than 33 °C. In turn, Baêta and Souza (2010) and Silva (2000)
precision (Pearson's correlation coefficient, r), accuracy (Willmott state that the ideal RH for domestic animals is, on average, 75%. These
agreement index) and confidence (coefficient of performance, c) data were applied in quantifying the BGTI with Eq. (6) and defining the
(Willmott, 1981, 1982; Camargo and Sentelhas, 1997). The MBE, ob- critical value to be used in bioclimatic evaluations for breeding Saanen
tained in oC, was used to measure the underestimation or over- goats.
estimation of the BGT equations in the long term. In turn, the RMSE, in BGTIX maps for the Brazilian Northeast were prepared for the four
o
C, gives the mean error in the short term. The closer to zero the MBE seasons of the year (summer, January to March; autumn, April to June;
and RMSE, the lower the error in estimation by the equations. The winter, July to September; and spring, October to December) from the
coefficient “r” indicates the degree of dispersion of the generated data Quantum GIS, version 2.18.12. (QGIS, 2009). The data were spatialised
in relation to the mean (straight), ranging from −1 to +1, so that the using spatial modelling, as described by Alvares et al. (2013), submit-
closer to the extremes, the better the accuracy of the equation. The ting the BGTIX, latitude, longitude and altitude to multivariate non-
signs indicate whether the correlation is negative or positive. The “d” linear regressions. In applying the multivariate equations, the digital
index indicates the degree of accuracy, the closer to +1, the greater the elevation model (Shuttle Radar Topography Mission, 90 × 90 m)
accuracy, i.e. the estimated values are the same as the observed values. (Miranda, 2019) was used as an independent variable. Latitude and
The “c” index, as suggested by Camargo and Sentelhas (1997), calcu- longitude maps were obtained by means of the trend method of spa-
lated as the product of the values for “r” and “d”, was used to classify tialization, from which, together with the MDE, the BGTIX maps were
the performance of the equations in estimating the BGT. Student's t-test generated.
was also applied between the estimated and observed data for non-re-
jection of the hypothesis H0 (p ≥ 0.05), i.e. that the mean value be-
3. Results
tween them did not differ.
3.1. Equations for estimating the BGT from meteorological data
2.2. Bioclimatic evaluation of the Brazilian Northeast for Saanen dairy
goats
According to Table 1, there was a highly significant correlation
(p < 0.0001) between BGT and the meteorological elements, except
The evaluation was performed to ascertain the thermal comfort
for WS, which had no effect on BGT at any of the time scales under
conditions of the Brazilian Northeast for breeding Saanen goats, which
study; therefore, WS was not adopted in generating the multiplicative
are animals specialised in dairy production. Monthly observed and
model. On the daytime scale, the correlation between BGT and the
monthly estimated historical data for maximum air temperature (TX,
o
meteorological elements was significant for SR, RH and AT (Table 1).
C) and minimum relative humidity (RHN,%) (data for mid-afternoon)
AT explained 80% of the variability in the BGT, followed by RH (65%)
from 2102 stations (Fig. 1A) were used to calculate the maximum BGT
and SR (45%). As AT explained most of the variability in the BGT
and maximum BGTI (BGTIX) (Buffington et al., 1981):
(80%), when the 20% residual not explained by AT (i.e. the BGTo/BGTe
BGTIX = BGT + 0.36 TDP + 41.5 (6) ratio) was correlated, first with RH and then with SR, the latter variable
showed a higher value for r2 (Fig. 2), and was therefore used in gen-
where: TDP = dew point temperature, in °C.
erating the multiplicative model. The resulting multiplicative model
The observed data were obtained from the Meteorological Database
was written as the product of the equation, BGT versus AT (Table 1),
for Teaching and Research of the National Meteorological Institute
with the equation, BGTo/BGTe versus SR (Fig. 2B):
(www.inmet.gov.br) (74 weather stations), and the estimated data
(2028 points), calculated from the models proposed by Medeiros et al. BGT = [( 1.3897 AT − 5.4421) (0.0384 ln(SR) + 0.7935 )], with r 2
(2005) and Silva et al. (2007) for TX and minimum relative humidity (9)
= 0.91
(RHN) respectively. Silva et al. (2007) state that, from the data for local
longitude (λ, in degrees), moisture index (IM, dimensionless) and mean Combining AT and SR accounted for 91% of the variability in the
monthly air temperature (T, oC), it is possible to estimate the mean BGT, which is greater than explained using AT only (80%).
relative humidity, as per Eq. (7). According to Fig. 3, the linear equation as a function of AT only has
a low long-term error (MBE = 0.02 °C), however in the short term, the
RH = a exp{− exp[− (IM − b)/ c]} [ d+ e (T λ )] (7)
mean error is 3.34 °C (RMSE) (Fig. 2A). The “r” of 0.90 showed the high
where: a, b, c, d, e − coefficients of the equation. precision of the model, i.e. that 90% of the estimated data were very
However, in this study, in order to estimate the RHN, the values for close to the mean. On the other hand, the “d” index was 0.94, indicating
T were replaced by TX. This procedure was adopted in order to obtain optimal accuracy, i.e. 94% of the estimated data were very close to
the best estimates and ensure more-accurate spatialization of the BGTIX observed data. As such, according to the confidence index (“c”), the
data, since most of the stations used have no RHN data. For the IM, performance of this equation was classified as “Very Good” and, based
obtained by IM = [(P/PET)-1].100, monthly values of precipitation (P) on Student's t-test at 5% significance (t = 0.16), the estimated values
and potential evapotranspiration (PET) are necessary. The PET was did not differ from the observed values. For the multiplicative model
calculated from the data for effective temperature (TEF) as described in (Eq. (9)), the performance was even better, with high precision
Pereira et al. (2002): TEF = 0.36 (3 TX - TN), where TN is the minimum (r = 0.95) and accuracy (d = 0.98), and low errors (MBE = 0.15 °C,
monthly air temperature. TPO was calculated by means of TX, in “eS” RMSE = 2.38 °C), and was considered “Excellent” (c = 0.93); Student's
(Eq. (3)) and by means of RHN, used in quantificating “e” by Eq. (1). t-test was also significant (t = 0.43 > 0.05).

3
T.G.F. da Silva, et al. Journal of Thermal Biology 85 (2019) 102408

Fig. 1. (A) Location of the data points in the Brazilian Northeast. (B) Amount of goat milk, in litres, by state (IBGE, 2017). AL - Alagoas; BA - Bahia; CE - Ceará; PB -
Paraíba; MA - Maranhão; PE - Pernmabuco; PI - Piauí; RN - Rio Grande do Norte; SE - Sergipe.

For the other time scales, AT was the only variable adopted in ex- 1.5941, r2 = 0.97, Table 1). The long-term (MBE = 0.0 °C) and short-
plaining the BGT, and it was not necessary to generate a multiplicative term (RMSE = 0.6 °C) errors were greatly reduced (Fig. 4A). The “r” of
model to include a new explanatory variable. On the night-time scale, 0.98 indicated the high precision, and the “d” of 0.99, the optimal ac-
the AT explained 99% of the variation in BGT (BGT = 0.995 AT - curacy of this equation. Performance, therefore, was classified as “Ex-
0.6964, r2 = 0.99, Table 1). From Fig. 3C, the low errors cellent”, and the t-test showed that there was no difference between the
(MBE = −0.02 °C and RMSE = 0.44 °C) and the “Excellent” perfor- estimated and observed data (t = 0.23 > 0.05). On the monthly scale,
mance (c = 0.99) can be seen, with the estimated values not differing AT explains 98% of the variability in BGT (BGT = 1.1550 AT - 1.3498).
from the observed values (t = 0.88 > 0.05). On the daily scale, AT The long-term (MBE = 0.01 °C) and short-term (RMSE = 0.30 °C) esti-
explained 97% of the fluctuation in BGT data (BGT = 1.1641 AT - mation errors were even lower when compared to the other time scales,

Table 1
Descriptive statistics and regression between the black globe temperature (Y = BGT, °C, dependent variable) and meteorological elements (X, independent variables)
at different time scales (daytime, night-time, daily and monthly). Logarithmic equation Y = a.ln (X) + b; and Linear equation Y = a (X) + b.
Time scale aBGT (oC) X Unit Minimum μ Maximum Equation r2 p n

−1
Daytime (SR > 0) 12.0–55.2 WS ms 0.0 2.0 6.9 ns – > 0.05 35436
Daytime (SR > 0) SR W m−2 0.1 427 1233 BGT = 2.8646.ln(X)+20.4247 0.45 < 0.0001 35436
Daytime (SR > 0) RH % 19.4 58.1 100.0 BGT = −0.3416.(X)+55.5556 0.65 < 0.0001 35436
o
Daytime (SR > 0) AT C 12.7 29.6 40.6 BGT = 1.3897.(X)-5.4421 0.80 < 0.0001 35436
Night-time (SR < 0) 12.0–47.9 WS m s−1 0.0 1.2 5.9 ns > 0.05 30861
Night-time (SR < 0) RH % 27.3 75.4 100.0 BGT = −0.1764.(X)+ 37.1534 0.61 < 0.0001 30861
o
Night-time (SR < 0) AT C 12.8 24.7 37.1 BGT = 0.9950.(X)-0.6964 0.99 < 0.0001 30861
Daily 21.8–36.4 WS m s−1 0.5 1.6 3.0 ns > 0.05 461
Daily SR MJ m−2 day−1 3.1 19.7 27.1 BGT = 0.4618.(X)+21.1379 0.47 < 0.0001 461
Daily RH % 43.4 65.9 89.9 BGT = −0.1935.(X)+42.9984 0.46 < 0.0001 461
o
Daily AT C 20.4 27.3 32.9 BGT=1.1641.(X)-1.5941 0.97 < 0.0001 461
Monthly 23.4–33.9 WS m s−1 1.0 1.6 2.6 ns > 0.05 30
Monthly SR MJ m−2 day−1 13.4 19.7 24.0 BGT = 0.7331.(X)+15.7701 0.63 < 0.0001 30
Monthly RH % 54.2 66.2 80.5 BGT = −0.2152.(X)+44.4487 0.43 < 0.0001 30
o
Monthly AT C 22.1 27.3 30.3 BGT = 1.1550.(X)-1.3498 0.98 < 0.0001 30

SR - incident global solar radiation, RH – relative humidity, AT – air temperature, WS = wind speed, r2 = coeficient of determination.
aBGT – variation in the values of the black globe temperature.
μ - mean.
n – number of observations.
ns – no significant.

4
T.G.F. da Silva, et al. Journal of Thermal Biology 85 (2019) 102408

Fig. 2. Correlation of the residual TGNo/TGNe not explained by air temperature (AT) (A) with relative humidity and (B) with incident global solar radiation.

with high precision (r = 0.98) and accuracy (d = 1.00), and a perfor- performance, animal health, and production costs (Das et al., 2016;
mance considered “Excellent” (Fig. 4B). According to Student's t-test, Ribeiro et al., 2016; Curtis et al., 2017). Therefore, bioclimatic eva-
the estimated values again did not differ from the observed data luation should always be included in planning by the producer. In this
(t = 0.42 > 0.05). analysis it is essential to compare the climate conditions of the area
with the requirements of the animals. In bioclimatic assessments, in-
dices of thermal comfort/discomfort are used; the BGTI is a good ex-
3.2. Bioclimatic evaluation of the Brazilian Northeast for Saanen dairy
ample, however, it requires BGT data. The absence of black globe
goats
thermometers in weather stations makes bioclimatic evaluation difficult
in the various regions (Dimiceli et al., 2012).
In the Northeast of Brazil, the BGTI varies between 75 and 95
Therefore, in the present study, equations for estimating the BGT
(Fig. 5). The largest range of territory with ideal thermal conditions
were first obtained by means of meteorological data for different time
(BGTI < 85) for breeding Saanen dairy goats occurs during the summer
scales (daytime, night-time, daily and monthly). The BGT showed a
and autumn (Fig. 5A and B), when maximum air temperatures are high
correlation with all the possible variables, with the exception of wind
(~31 °C), but lower than found at the end of winter (September) and
speed. This variable affects the injection or removal of heat to/from the
during the spring. During the winter and spring (Fig. 5C and D), areas
environment, but its contribution was low when compared to the other
with a BGTI < 85 are reduced, predominating in the state of Bahia and
variables (global solar radiation, temperature and relative humidity).
in the high-altitude areas of the other states. In general, the west and
The correlation between the BGT and SR fit a logarithmic equation,
centre-south of the state of Bahia offer the animals the most thermal
while the other equations were linear (Table 1). The positive signs of
comfort during each season of the year. In Maranhão, heat stress occurs
the angular coefficient of the equations (“a”) show that an increase in
throughout the year, with the BGTI predominately in the range of
one variable increases the values of the BGT. Only the correlation be-
85–95.
tween BGT and RH generated a negative “a” coefficient. A similar result
was noted by Hajizadeh et al. (2017).
4. Discussion For all time scales, the most important variable was air temperature
(AT), as noted by Lemke and Kjellstrom (2012) and Hajizadeh et al.
4.1. Estimating the BGT from meteorological data (2017); however, these authors also found correlation of the BGT with
solar radiation and wind speed. In the present study, global solar ra-
The climate is a major factor in animal farming, especially in tro- diation (SR) gained importance on the daytime scale. When SR > 0,
pical and subtropical regions. Furthermore, the climate defines the level two equations can be used to estimate the BGT, a simple linear equation
of modification of the breeding environment, management,

Fig. 3. Evaluation of the performance of the equation for estimating the daytime black globe temperature (BGT) (A) based only on air temperature (AT), (B) of the
multiplicative model combining daytime air temperature and incident global solar radiation (SR), and (C) of the equation for estimating the night-time black globe
temperature based on the night-time temperature.

5
T.G.F. da Silva, et al. Journal of Thermal Biology 85 (2019) 102408

Fig. 4. Evaluation of the performance of the equation for estimating the black globe temperature (BGT) (A) daily and (B) monthly based on mean air temperature
(AT).

based on AT only (BGT = 1.3897 AT -5.4421, r2 = 0.80, Table 1) and 2008; Mendes et al., 2014). However, for a good result it is important to
the other resulting from the product of two equations, generating the use indicators that unify the effects of the climate factors (Wang et al.,
multiplicative model, which combines the effect of AT and SR (Eq. (9)). 2018), since these are directly related to the physiological and beha-
This result shows that, depending on applicability, estimating the BGT vioural responses of the animals (Neves et al., 2009; McManus et al.,
from AT alone may be acceptable. However, for better estimates on a 2016). Because of the diversity of data cited in the literature on the
daytime scale, the multiplicative model is a good choice. These results condition of thermal comfort/discomfort of the Saanen breed, in the
corroborate those presented by Turco et al. (2008), who combined the present study, a critical value for BGTI was defined, to be used in the
effects of AT and SR on the BGT in a multiplicative model bioclimatic evaluation of various regions for breeding the animals.
(BGT = [1360 AT -2.358] (0.075 ln(SR) + 0.562]), and found that Combining the values for maximum temperature and relative humidity
these variables explained 87.7% of the variability in the BGT. Hajizadeh cited by Baêta and Souza (2010), Pequeno et al. (2017) and Silva
et al. (2017) propose a multiple linear equation as a function of three (2000) resulted in a critical limit for the BGTI of 85. Above this value,
variables (BGT = 0.01498 SR + 1.184 AT - 0.0789 RH - 2.739). The the reduction in milk production by these goats is considered quite
good BGT estimate from the meteorological data is of extreme im- pronounced, so that good animal performance is impaired; modifying
portance for improving zootechnical planning and decision-making the zootechnical structure and adjusting the management can sig-
(Silva et al., 2008; da Silva et al., 2010). nificantly affect production costs. A BGTI of 85 is within the mean
For the night-time scale, AT explained 99% of the variation in BGT, range of values normally found during the afternoon (82.50 in the
so it should be the only variable used to estimate the BGT (BGT = 0.995 shade and 91.30 in full sun, a mean of 86.90) in the Brazilian semi-arid
AT - 0.6964, r2 = 0.99, Table 1), it not being necessary to generate a region, according to studies carried out to evaluate the adaptation of
multiplicative model to include a new variable. After the regression Saanen goats in this region [Roberto et al. (2010): 87 in the shade and
analysis, AT was also the variable that most explained the fluctuation in 94 in full sun; Silva et al. (2010): 82 in the shade and 94 in full sun;
BGT data on the daily scale (BGT = 1.1641 AT - 1.5941, r2 = 0.97, Silva et al. (2011): 81 in the shade and 91 in full sun; and Silva et al.
Table 1). Good results for this time scale were also quoted by Abreu (2014): 80 in the shade and 87 in full sun].
et al. (2011), when relating the BGT and AT measured inside and In the northeast of Brazil, Bahia offers the best thermal environment
outside poultry sheds. From the result obtained here, it was found that for breeding Saanen dairy goats, with the west and centre-south of the
the daily AT explained 97% of the variability in the BGT, so it was not state being most important. Although the largest herd of goats (with
necessary to generate a multiplicative model to include another vari- different breeds and aptitudes, dairy and meat) is located in the north of
able. The same procedure was adopted for the monthly scale, since AT the state, where there are higher values for air temperature and lower
was again the variable that explained most of the variability in the BGT values for relative humidity, favourable conditions can be seen during
(98%) (BGT = 1.1550, AT - 1.3498). the first six months of the year, i.e. during the summer and autumn.
Though the data have been collected from a single meteorological During the other months (winter and spring), the BGTI is slightly higher
station, the equations for estimating the BGT can be applied to other (from 85 to 87) than the critical limit. This same seasonality in the
regions with different climate conditions. This application is possible thermal condition of the environment was found for Alagoas, Sergipe,
because of the wide values range collected in this study (Table 1) and Pernambuco and a large part of the states of Rio Grande do Norte, Ceará
linear and constant response characteristic of BGT replicated by the and Piauí, where the lowest heat stress is found during the summer and
models based AT and SR (Fig. 3B) or only AT (Fig. 4A and B), which autumn. The rainy period in the region is included in these two seasons,
allows the trend extrapolation. Other works showed this linear and which continues until July along the coastal areas. The high amount of
constant response of BGT to different meteorological variables, espe- cloud reduces the incidence of solar radiation, lowering the tempera-
cially air temperature (Lemke and Kjellstrom, 2012; Hajizadeh et al., ture, which combined with higher values for relative humidity (~75%)
2017). promotes thermal comfort. During the second half of the year, part of
winter and the whole of spring, the low cloud cover, high incidence of
4.2. Bioclimatic evaluation of the Brazilian Northeast for Saanen dairy radiation and smaller rainfall events increase the BGTI, making the
goats environment uncomfortable for the animals. The state of Maranhão has
the worst thermal conditions for raising Saanen dairy animals. During
Bioclimatic evaluation allows the species most adapted to a specific 2017, the state had the lowest production of goat milk among the states
climate condition to be identified and, therefore, the most promising of the Brazilian Northeast (IBGE, 2017) (Fig. 1B). For breeding dairy
livestock activity to be recommended (Turco et al., 2006; Silva et al., cows, Turco et al. (2006) cite appropriate conditions for the state of

6
T.G.F. da Silva, et al. Journal of Thermal Biology 85 (2019) 102408

Fig. 5. Bioclimatic evaluation of the Brazilian Northeast for breeding Saanen dairy goats. (A) BGHIX for spring (October to December); (B) BGHIX for summer
(January to March); (C) BGHIX for autumn (April to June); and, (D) BGHIX for winter (July to September). Critical BGHIX for the race equal to 85.

Bahia during the coldest months only (June to August). Silva et al. therefore reduced (Hill and Wall, 2017; Sejian et al., 2018); in addition,
(2008) report that during the hottest months (December to February), the high temperatures influence reproduction, due to changes in cel-
81% of the territory of Pernambuco has conditions that are stressful to lular function and in various organs of the reproductive system (Das
dairy cattle. Whereas, Mendes et al. (2014), in a bioclimatic study for et al., 2016; Pantoja et al., 2017). Sejian et al. (2018) recommend four
breeding Dorper sheep in the state, found that throughout the year, the groups of strategies to combat heat stress in livestock: management of
area of Agreste has the best climate conditions compared to the coastal, the facilities; nutritional, genetic and reproductive changes; and animal
forest and semi-arid regions. health and management.
Under conditions of heat stress, the level of production and the In the face of moderate stress conditions (BGTI from 85 to 87), al-
quality of the animals' milk decrease (Zobel et al., 2015; Ramón et al., tering the environment by such practices as a more regular and ba-
2016). According to Das et al. (2016), the sensitivity of the dairy animal lanced water and food supply, and adopting simple artificial or natural
increases as its production level rises because of increased metabolic shade, can reduce the inappropriate effects of heat (Ben Salem and
heat. In exotic breeds, the loss of yield is more intense, since the genes Smith, 2008; Sejian et al., 2018). Good nutritional management allows
of these animals are less able to adapt to hot climates. The high tem- the animals to remain in homeostasis or prevents nutritional defi-
perature negatively affects the hypothalamus, and food intake is ciencies caused by heat stress (Das et al., 2016). In turn, Pezzopane

7
T.G.F. da Silva, et al. Journal of Thermal Biology 85 (2019) 102408

et al. (2019) report that, when properly arranged, shading in a silvo- Buffington, D.E., Collazo-Arocho, A., Canton, G.H., et al., 1981. Black globe humidity
pastoral system decreases the radiant heat load on the animal by 22%. index (BGHI) as comfort equation for dairy cows. Trans. ASAE (Am. Soc. Agric. Eng.)
24, 711–714. https://doi.org/10.13031/2013.34325.
However, under more severe and persistent heat stress (BGTI in the Camargo, A.P., Sentelhas, P.C., 1997. Avaliação do desempenho de diferentes métodos de
range of 87–95), more intense practices for physically changing the estimativas da evapotranspiração potential no Estado de São Paulo, Brasil. Rev. Bras.
environment should be adopted, for example, the installation of air Agrometeorol. 5, 89–97. https://doi.org/10.1590/S1415-43662007000600006.
Curtis, A.K., Scharf, B., Eichen, P.A., Spiers, D.E., 2017. Relationships between ambient
conditioning systems (i.e. an evaporative cooling system with water in conditions, thermal status, and feed intake of cattle during summer heat stress with
the form of droplets, mist or spray, natural or forced air movement and access to shade. J. Therm. Biol. 63, 104–111. https://doi.org/10.1016/j.jtherbio.
cooling tanks) (Das et al., 2016). The use of a choice of floor-coverings 2016.11.015.
da Silva, T.G.F., de Moura, M.S.B., Sá, I.I.S., et al., 2010. Cenários de mudanças climáticas
also improves the thermal quality of the facility, quickly changing the e seus impactos na produção leiteira em estados nordestinos. Rev. Bras. Eng. Agrícola
microclimate and animal behaviour, encouraging the animal to ex- Ambient. 14, 863–870.
perience different environmental conditions over time (Sutherland Das, R., Sailo, L., Verma, N., et al., 2016. Impact of heat stress on health and performance
of dairy animals: a review. Vet. World 9, 260–268. https://doi.org/10.14202/
et al., 2017). By crossing more-productive exotic species with native
vetworld.2016.260-268.
species, it is also possible to obtain races that are less sensitive to heat Dimiceli, V.E., Piltz, S.F., Amburn, S.A., 2012. Black globe temperature estimate for the
stress (Ben Salem and Smith, 2008; Das et al., 2016; Sejian et al., 2018). WBGT index. AIP Conf. Proc. 170, 323–334. https://doi.org/10.1007/978-94-007-
4786-9_26.
Hajizadeh, R., Farhang Dehghan, S., Golbabaei, F., et al., 2017. Offering a model for
5. Conclusions estimating black globe temperature according to meteorological measurements.
Meteorol. Appl. 24, 303–307. https://doi.org/10.1002/met.1631.
The proposal of equations for estimating the black globe tempera- Hill, D.L., Wall, E., 2017. Weather influences feed intake and feed efficiency in a tem-
perate climate. J. Dairy Sci. 100, 2240–2257. https://doi.org/10.3168/jds.2016-
ture (BGT) on different time scales (daytime, night-time, daily and 11047.
monthly) was advanced in this study for application in meteorological IBGE, 2017. Produção e Venda de leite de cabra no ano. In: Censo Agropecuário, .
databases and the identification of suitable climate zones in the https://sidra.ibge.gov.br/tabela/963, Accessed date: 19 November 2017.
Irano, N., Bignardi, A.B., Rey, F.S.B., et al., 2012. Parâmetros genéticos para a produção
Northeast of Brazil for breeding Saanen dairy goats. On a daytime scale, de leite em caprinos das raças Saanen e Alpina. Rev. Cienc. Agron. 43, 376–381.
the BGT can be calculated from a simple linear equation and air tem- Lemke, B., Kjellstrom, T., 2012. Calculating workplace WBGT from meteorological data: a
perature (AT) data, but for even better results, combining the AT data tool for climate change assessment. Ind. Health 50, 267–278. https://doi.org/10.
2486/indhealth.ms1352.
with solar radiation (SR) data by means of a multiplicative model is Lôbo, R.N.B., Silva, F.L.R., 2005. Parâmetros genéticos para características de interesse
even more recommended. To calculate the BGT on night-time, daily and econômico em cabras das raças Saanen e Anglo-nubiana. Rev. Cienc. Agron. 36,
monthly scales, it is enough to apply AT data, again with simple linear 104–110.
McManus, C., Dallago, B.S.L., Lehugeur, C., et al., 2016. Patterns of heat tolerance in
equations. In the bioclimatic evaluation of the Brazilian Northeast for
different sheep breeds in Brazil. Small Rumin. Res. 144, 290–299. https://doi.org/10.
breeding Saanen dairy goats, a critical limit of 85 was established for 1016/j.smallrumres.2016.10.004.
the black globe temperature index (BGTI). In general, the west and Medeiros, S.D.S., Cecílio, R.A., Júnior, J.C.F.D.M., et al., 2005. Estimativa e espacializ
centre-south of the state of Bahia offer the animals the most thermal ação Estimativa espacialização das temper atur as do ar mínimas temperatur aturas e
máximas na Região Nordeste do Brasil. Rev. Bras. Eng. Agrícola Ambient. 9,
comfort during each season of the year. Alagoas, Paraíba, Pernambuco, 247–255.
Piauí, Rio Grande do Norte and Sergipe have suitable thermal condi- Mendes, AM. de P., Azevedo, M de, Lopes, P.M.O., Moura, GB. de A., 2014. Zoneamento
tions during the summer and autumn. In the state of Maranhão heat bioclimático para a raça ovina Dorper no Estado de Pernambuco. Pesqui. Agropecu.
Bras. 49, 986–993. https://doi.org/10.1590/S0100-204X2014001200009.
stress occurs throughout the year, with the BGTI predominately in the Miranda E., E., 2019. In: Miranda E.E. (Ed.), Embrapa Monitoramento por Satélite, 2005,
range of 85–95. As such, strategies to combat heat stress should be Campinas. http://www.relevobr.cnpm.embrapa.br, Accessed date: 15 January 2019.
encouraged to reduce the negative effects of climate on milk production Neves, M.L.M.W., De Azevedo, M., Da Costa, L.A.B., et al., 2009. Níveis críticos do Índice
de Conforto Térmico para ovinos da raça Santa Inês criados a pasto no agreste do
in Saanen goats, and favour the production chain in the northeast of Estado de Pernambuco. Acta Sci. Anim. Sci. 31, 167–175. https://doi.org/10.4025/
Brazil. actascianimsci.v31i2.3766.
Pantoja, MH. de A., Esteves, S.N., Jacinto, M.A.C., et al., 2017. Thermoregulation of male
sheep of indigenous or exotic breeds in a tropical environment. J. Therm. Biol. 69,
Acknowledgments
302–310. https://doi.org/10.1016/j.jtherbio.2017.09.002.
Pequeno, I.D., Turco, S.H.N., da Silva, T.G.F., Facó, O., 2017. Dairy production of ‘saanen’
The authors wish to thank the Conselho Nacional de goats based on meteorological variables and future climate scenarios. J. Braz. Assoc.
Agric. 37, 226–235. https://doi.org/10.1590/1809-4430-Eng.Agric.v37n2p226-235/
Desenvolvimento Científico e Tecnológico (No 305286/2015-3) and the
2017.
Fundação de Amparo à Ciência e Tecnologia do Estado de Pernambuco Pereira, A.R., Angelocci, L.R., Sentelhas, P.C., 2002. Agrometeorologia: Fundamentos e
(No APQ-1159-1.07/14). Aplicações Práticas. Agropecuária, Guiabá.
Pezzopane, J.R.M., Nicodemo, M.L.F., Bosi, C., et al., 2019. Animal thermal comfort in-
dexes in silvopastoral systems with different tree arrangements. J. Therm. Biol. 79,
Appendix A. Supplementary data 103–111. https://doi.org/10.1016/j.jtherbio.2018.12.015.
QGIS Development Team, 2009. QGIS Geographic Information System. Open Source
Supplementary data to this article can be found online at https:// Geospatial Foundation Project. http://qgis.osgeo.org.
Ramón, M., Díaz, C., Pérez-Guzman, M.D., Carabaño, M.J., 2016. Effect of exposure to
doi.org/10.1016/j.jtherbio.2019.102408. adverse climatic conditions on production in Manchega dairy sheep. J. Dairy Sci. 99,
5764–5779. https://doi.org/10.3168/jds.2016-10909.
References Ribeiro, N.L., Costa, R.G., Pimenta Filho, E.C., et al., 2016. Adaptive profile of Garfagnina
goat breed assessed through physiological, haematological, biochemical and hor-
monal parameters. Small Rumin. Res. 144, 236–241. https://doi.org/10.1016/j.
Abreu, PG de, Abreu, V.M.N., Franciscon, L., et al., 2011. Estimativa da temperatura de smallrumres.2016.10.001.
globo negro a partir da temperatura do bulbo seco. Eng na Agric 19, 557–563. Roberto, J.V.B., Souza, B.B., Silva, A.L.N., et al., 2010. Parâmetros hematológicos de
Alvares, C.A., Stape, J.L., Sentelhas, P.C., de Moraes Gonçalves, J.L., 2013. Modeling caprinos de corte submetidos a diferentes níveis de suplementação no semi-árido
monthly mean air temperature for Brazil. Theor. Appl. Climatol. 113, 407–427. paraibano. Rev Caatinga 23, 127–132.
https://doi.org/10.1007/s00704-012-0796-6. Sejian, V., Bhatta, R., Gaughan, J.B., et al., 2018. Review: adaptation of animals to heat
Alvares, C.A., Stape, J.L., Sentelhas, P.C., Gonçalves, JL. de M., 2014. Köppen’s climate stress. Animal 12, S431–S444. https://doi.org/10.1017/S1751731118001945.
classification map for Brazil. Meteorol. Z. 22, 711–728. https://doi.org/10.1127/ Silva, R.G., 2000. Introdução à bioclimatologia animal. Nobel, São Paulo.
0941-2948/2013/0507. Silva, TGF da, Zolnier, S., Moura, MSB de, Sediyama, G.C., 2007. Estimativa e
Arnal, M., Robert-Granié, C., Larroque, H., 2018. Diversity of dairy goat lactation curves espacialização da umidade relativa do ar para os estados de Alagoas, Bahia e Sergipe.
in France. J. Dairy Sci. 101, 11040–11051. https://doi.org/10.3168/jds.2018-14980. Rev. Bras. Agrometeorol. 15, 1–9.
Baêta, F.C., Souza, C.F., 2010. Ambiência Em Edificações Rurais - Conforto Animal, 2.ed. Silva, TGF da, Turco, S.H.N., Zolnier, S., et al., 2008. Variação regional do declínio na
UFV, Viçosa, MG. produção de leite durante o verão no estado de pernambuco. Eng na Agric 16,
Ben Salem, H., Smith, T., 2008. Feeding strategies to increase small ruminant production 109–123.
in dry environments. Small Rumin. Res. 77, 174–194. https://doi.org/10.1016/j. Silva, EMN da, Souza, BB de, Sousa, OB de, et al., 2010. Avaliação da adaptabilidade de
smallrumres.2008.03.008. caprinos ao semiárido através de parâmetros fisiológicos e estruturas do tegumento.

8
T.G.F. da Silva, et al. Journal of Thermal Biology 85 (2019) 102408

Rev Caatinga 23, 142–148. the Eighth ASABE International Livestock Environment Symposium. St. Joseph, MI,
Silva, C.M.B.D.A., Souza, BB de, Brandão, P.D.A., et al., 2011. Efeito das condições USA. ASABE, Foz do Iguaçu, Brasil, pp. 783–788.
climáticas do semiárido sobre o comportamento fisiológico de caprinos mestiços F1 Wang, X., Bjerg, B.S., Choi, C.Y., et al., 2018. A review and quantitative assessment of
Saanen x Boer. Rev Caatinga 24, 195–199. cattle-related thermal indices. J. Therm. Biol. 77, 24–37. https://doi.org/10.1016/j.
Silva, E.M.N., Souza, B.B., Silva, G.A., et al., 2014. Avaliação da Adaptabilidade de jtherbio.2018.08.005.
Caprinos Leiteiros no Semiárido Brasileiro com Auxílio da Termografia Willmott, C.J., 1981. On the validation of models. Phys. Geogr. 2 (2), 184–194. https://
Infravermelha. J. Anim. Behav. Biometeorol. 2, 95–101. https://doi.org/10.14269/ doi.org/10.1080/02723646.1981.10642213.
2318-1265/jabb.v2n3p95-101. Willmott, C.J., 1982. Some comments on the evaluation of model performance. Bull. Am.
Sutherland, M.A., Lowe, G.L., Watson, T.J., et al., 2017. Dairy goats prefer to use different Meteorol. Soc. 63, 1309–1313. https://doi.org/10.1175/1520-0477(1982)
flooring types to perform different behaviours. Appl. Anim. Behav. Sci. 197, 24–31. 063<1309:SCOTEO>2.0.CO;2.
https://doi.org/10.1016/j.applanim.2017.09.004. Zambom, M.A., Alcalde, C.R., Martins, E.N., et al., 2013. Produção, composição do leite e
SYSTAT SOFTWARE Inc, 2019. Sigmaplot for windows, version 14. Available at: https:// variação do custo e da receita de produção de leite de cabras Saanen recebendo
systatsoftware.com/. rações com casca de soja em substituição ao milho1. Semin. Agrar. 34, 1313–1326.
Turco, S.H.N., SILVA, T.G., SANTOS, L.F.D.O.S., et al., 2006. Zoneamento bioclimático https://doi.org/10.5433/1679-0359.2013v34n3p1313.
para vacas leiteiras no estado da Bahia. Eng. Agrícola 26, 20–27. https://doi.org/10. Zobel, G., Weary, D.M., Leslie, K.E., von Keyserlingk, M.A.G., 2015. Invited review:
1007/s13398-014-0173-7.2. cessation of lactation: effects on animal welfare. J. Dairy Sci. 98, 8263–8277. https://
Turco, S.H.N., Silva, TGF da, Oliveira, GM de, et al., 2008. Estimating black globe tem- doi.org/10.3168/jds.2015-9617.
perature based on meteorological data. In: Livestock Environments, Proceedings of