You are on page 1of 11

Agricultural and Forest Meteorology 100 (2000) 231–241

Estimation of mean monthly solar global radiation as


a function of temperature
Francisco Meza a,∗ , Eduardo Varasb,1
aDepartamento de Ciencias de Recursos Naturales, Pontificia Universidad Católica de Chile, Casilla 306, Correo 22, Santiago, Chile
b Departamento de Ingenierı́a Hidráulica y Ambiental, Pontificia Universidad Católica de Chile, Casilla 306, Correo 22, Santiago, Chile
Received 14 December 1998; received in revised form 11 August 1999; accepted 13 August 1999

Abstract
Solar radiation is the primary energy source for all physical and biochemical processes that take place on earth. Energy
balances are a key feature of processes such as temperature changes, snow melt, carbon fixation through photosynthesis in
plants, evaporation, wind intensity and other biophysical processes. Solar radiation level is sometimes recorded, but generally
it needs to be estimated by empirical models based on frequently available meteorological records such as hours of sunshine
or temperature.
This paper evaluates the behavior of two empirical models based on the difference between maximum and minimum
temperatures and compares results with a model based on sunshine hours. This work concludes that empirical models based
on temperature have a larger coefficient of determination than the model based on cloud cover for the normal conditions of
Chile. These models are easy to use in any location if the parameters are correctly adjusted. In addition, probability distribution
functions and confidence intervals for solar radiation estimates using stochastic modeling of temperature differences were
calculated. ©2000 Published by Elsevier Science B.V. All rights reserved.
Keywords: Solar radiation; Temperature; Random variable; Fourier series

1. Introduction uates proposed mathematical models to estimate so-


lar radiation as a function of temperature differences
In some cases a record of global solar radiation (RG ) and compares their performance with models based
using instruments such as pyranometers or actinome- on sunshine hours.
ters is available, however, there are many meteorolog- Solar radiation is the principal energy source for
ical stations which do not measure solar radiation, but physical, biological and chemical processes, such as,
do register other variables such as precipitation, pres- snow melt, plant photosynthesis, evaporation, crop
sure and temperature. For this reason, this paper eval- growth and is also a variable needed for biophysical
models to evaluate risk of forest fires, hydrological
∗ Corresponding author. Fax: +56-2-553-92-31.
simulation models and mathematical models of natu-
E-mail addresses: fmeza@puc.cl (F. Meza), evaras@ing.puc.cl
ral processes. Hence, in many occasions, a record of
(E. Varas). observed solar radiation or an estimate of radiation is
1 Fax +56-2-686-58-76. required.

0168-1923/00/$ – see front matter ©2000 Published by Elsevier Science B.V. All rights reserved.
PII: S 0 1 6 8 - 1 9 2 3 ( 9 9 ) 0 0 0 9 0 - 8
232 F. Meza, E. Varas / Agricultural and Forest Meteorology 100 (2000) 231–241

Table 1
2. Model description
Angström coefficients (a and b) recommended for Chilean locali-
ties. (Castillo and Santibáñez, 1981)
Extra-terrestrial solar radiation, also known as An-
Locality a b Latitude Longitude Altitude
got radiation (RA , MJ m−2 day−1 ) can be calculated (◦ S) (◦ W) (m)
as a function of the distance from the sun to earth (d,
Arica 0.28 0.57 18.29 70.19 035
km), the mean distance sun–earth (dm , km), latitude
Iquique 0.23 0.47 20.13 70.09 008
(φ, rad), solar declination (δ, rad) and solar angle at Antofagasta 0.23 0.47 23.28 70.20 122
sunrise (sunset) (Hs , rad) using the following expres- Copiapó 0.26 0.51 27.21 70.20 283
sion (Romo and Arteaga, 1983): Vallenar 0.22 0.46 28.35 70.46 469
  La Serena 0.29 0.57 29.54 71.15 032
(86400)(1360) dm 2 La Paloma 0.22 0.46 30.41 71.02 320
RA = Quintero 0.22 0.45 32.47 71.32 002
π d
Valparaiso 0.22 0.55 33.01 70.38 041
× [(Hs )sin(φ)sin(δ) + cos(φ)cos(δ)sin(Hs )] Santiago 0.22 0.44 33.27 70.42 520
Curicó 0.23 0.47 34.58 71.13 227
(1)
Constitución 0.22 0.45 35.20 72.26 007
Using the preceding relationship, solar radiation can Chillan 0.23 0.47 36.36 72.02 124
Concepción 0.26 0.51 36.47 73.07 009
be calculated for any point in the earth’s outer atmo- Temuco 0.23 0.47 38.46 72.39 114
sphere for each day of the year as a function of latitude Osorno 0.23 0.47 40.35 73.09 027
and solar declination. However, gases and clouds in- Puerto Montt 0.26 0.51 41.28 72.56 110
troduce changes to both magnitude and spectral com- Ancud 0.26 0.51 41.54 73.48 020
position of solar radiation. Puerto Aysén 0.26 0.51 45.24 72.42 010
Balmaceda 0.26 0.51 45.54 71.43 520
Punta Arenas 0.26 0.52 53.10 70.54 008
2.1. Angström model, 1924

Since the beginning of the century, efforts have Monteith (1966), Penman (1948), and Turc (1961)
been made to estimate solar radiation as a function have calibrated this expression for different places.
of extra-terrestrial solar radiation and the state of the Coefficients can vary significantly as Doorenbos
atmosphere (Castillo and Santibáñez, 1981). The pa- and Pruitt (1975) show. In Chile, Castillo and San-
rameter most commonly used is hours of sunshine. tibáñez (1981), have recommended the values given in
Usually, the ratio of global solar radiation to Angot ra- Table 1.
diation is correlated to the ratio of effective sunshine
hours to total possible sunshine hours.
Effective sunshine hours (n) are measured with a 2.2. Bristow–Campbell model, 1984
heliograph (Martı́nez-Lozano et al., 1984). Although
this instrument has a threshold, under which sunshine Incoming solar radiation is determined by the state
is not recorded, this error is not significant when esti- of the atmosphere. However, the dynamics of the
mating daily solar radiation. atmosphere is very difficult to predict. Considering
Angström (1924), suggested a simple linear re- transformations experienced by solar radiation, one
lationship to estimate global solar radiation (RG , can expect to find a relationship to express solar
MJ m−2 day−1 ) as a function of Angot radiation, radiation as a function of meteorological variables
actual sunshine hours (n) and potential or theoretical commonly registered at climatological stations. When
sunshine hours (N). solar radiations reaches the earth surface, part of it
is reflected and part is absorbed. The same occurs
RG n
=a+b (2) with long-wave radiation that each body emits as a
RA N function of its temperature. As Chang (1968), reports,
Angström suggested values of 0.2 and 0.5 for there is usually a good relation between net radiation
empirical coefficients a and b respectively. Other and global solar radiation, since the latter one is the
authors, such as Bennett (1962), Davies (1965), principal source of energy.
F. Meza, E. Varas / Agricultural and Forest Meteorology 100 (2000) 231–241 233

Furthermore, if the heat flow towards the soil is 3. Climatic data


neglected, one can find the ratio of sensible heat to
latent heat or Bowen ratio, on a daily basis (Campbell, In order to compare the behavior of the different
1977). Sensible heat is responsible for temperature models, monthly climatological data of 21 stations
variations, so it is possible to obtain a relationship representing different climatic regions of Chile were
between temperature differences and solar radiation, collected. Data ranged from Arica (latitude 18.3◦ S) to
being temperature a reflection of radiation balance. Punta Arenas (53.1◦ S) and was registered between the
Using this argument, Bristow and Campbell (1984), years 1971 and 1992.
suggested the following relationship for daily RG , as Selected meteorological variables were TM , Tm , P,
a function of daily RA and the difference between mean monthly degree of cloud cover (x) and RG .
maximum and minimum temperatures (1T, ◦ C): For the locations mentioned in Table 1, monthly val-
RG h i ues of maximum and minimum temperatures, cloud
= A 1 − exp(−B1T C ) (3) cover and atmospheric pressure for each year in the
RA
period 1971 to 1992, were available. Unfortunately,
Athough coefficients A, B and C are empirical, they for global solar radiation only the average value for
have some physical meaning. Coefficient A represents each month in that period was available and monthly
the maximum radiation that can be expected on a clear radiation values for each year were impossible to ob-
day. Coefficients B and C control the rate at which A tain from Dirección Meteorológica de Chile.
is approached as the temperature difference increases. In addition to the above, data from La Paloma sta-
Values most frequently reported for these coefficients tion was collected to compare the behaviour of models
are 0.7 for A, the range 0.004 to 0.010 for B and 2.4 based on temperature differences when they are ap-
for C. plied to estimate monthly global solar radiation. The
Since clear days present large temperature differ- selected meteorological variables in this case were TM ,
ences A tends to be the ratio between global solar radi- Tm , P, and RG between the years 1971 and 1978.
ation and Angot radiation, hence the sum of Angström Finally, data from Santiago station was used to
coefficients a and b tends to be similar to A. compare the behaviour of Bristow–Campbell and
Allen models when they are applied to estimate daily
2.3. Allen model, 1997 global solar radiation. The meteorological variables
were daily TM , Tm , P, and RG .
Allen (1997), suggested the use of a self-calibrating
model to estimate mean monthly global solar radiation
following the work of Hargreaves and Samani (1982). 4. Models applied to mean monthly data
He suggested that the mean daily RG can be estimated
as a function of RA , mean monthly maximum(TM , ◦ C) The extension of the reviewed models to apply them
and minimum temperatures (Tm , ◦ C). to monthly averages requires some explanation. The
RG Angström model was originally derived for daily so-
= Kr (TM − Tm )0.5 (4) lar radiation and hours of sunshine. Nonetheless, be-
RA
ing a linear function it can be readily applied to mean
Previously, Allen (1995), had expressed the empiri- monthly data since the expected value of a sum is
cal coefficient (Kr ) as a function of the ratio of atmo- equal to the summation of the expected values. Allen’s
spheric pressure at the site (P, kPa) and at sea level model was derived for monthly data so it can readily
(P0 , 101.3 kPa) as follows: be used. However, the Bristow–Campbell model is de-
 0.5 fined for daily data and has no evident extrapolation
P
Kr = Kra (5) to mean monthly values. For this reason, one can ex-
P0 pect to find a new set of coefficients when the same
In his work, Allen suggested values of 0.17 for interior expression is applied to monthly data.
regions and 0.20 for coastal regions for the empirical With the values of temperature, atmospheric pres-
coefficient Kra . sure and sunshine hours, mean monthly global solar
234 F. Meza, E. Varas / Agricultural and Forest Meteorology 100 (2000) 231–241

Table 3
radiation was calculated at each site, using the ex-
Regression between calculated and observed mean monthly global
pressions and empirical coefficients suggested by solar radiation using adjusted parameters of 20 Chilean localities
Angström (1924), Bristow and Campbell (1984), and
Model Slope Upper Lower R2
Allen (1995). Results show that models using the
limit. (95%) limit. (95%)
coefficients proposed in the literature do not esti-
mate correctly the historical average in each location. Angström 0.959 0.970 0.939 0.892
Allen 0.999 1.010 0.990 0.961
The slope of the relationship between calculated
Bristow–Campbell 1.152 1.170 1.138 0.928
and observed radiation is significantly different from
unity. This is especially notorious in the case of the
Bristow–Campbell model, although this result was ex- errors. The available data made it impossible to study
pected since the coefficients suggested by the authors the contribution of coefficients A and C. However, A
are applicable to daily data. represents the maximum radiation on a clear day and
Given the results, it was necessary to change the its value represents the observed data reasonably well.
Allen and Bristow–Campbell model coefficients to Moreover, a change in coefficient C does not affect
obtain a better fit, following the idea suggested by significantly the calculated global solar radiation.
Castillo and Santibáñez (1981) for the Angström Observed and calculated values for different
model. Least squares coefficients, which minimize the locations and models are shown in Fig. 1. In this
sum of square errors for each location were calculated figure the improvement in the relationships when us-
and included in Table 2. ing locally calibrated coefficients can be appreciated.
Due to the fact that monthly solar radiation values, The Angström model results using the coefficients
were not available for each year, as mentioned in the proposed by Castillo and Santibáñez (1981) are also
section about climatic data, the A and C coefficients included for comparison. Slopes of the different mod-
of Bristow–Campbell model were assumed fixed and els and the coefficients of determination are given in
the B coefficient was adjusted to minimize the square Table 3.
Allen’s model presents the best relationship. It has
a higher coefficient of determination and the slope
Table 2 is equal to unity with 90% confidence interval. The
Adjusted coefficients (Kra and B) of Allen and Bristow–Campbell
Bristow–Campbell model tends to under-estimate
models
global solar radiation but explains a large proportion
Locality Kra B
of sample variance. The Angström model fit the data
Arica 0.3354 0.01354 poorer than the other two.
Iquique 0.2854 0.01619
Antofagasta 0.4717 0.01944
Copiapó 0.2577 0.00203
4.1. Models applied to monthly data.
Vallenar 0.3457 0.00200
La Serena 0.2697 0.00677 Since the available data of global solar radiation
La Paloma 0.1589 0.00347 for most stations is only the average value for each
Quintero 0.2731 0.00589
month, it was necessary to examine if the relationships
Valparaı́so 0.0114 0.01144
Santiago 0.2593 0.00202 with the adjusted coefficients represent accurately the
Curicó 0.4348 0.00152 monthly values for each year. One station available
Constitución 0.2423 0.00555 with monthly global solar radiation data, is La Paloma.
Chillan 0.2316 0.00159 In this case the models with the adjusted coefficients
Concepción 0.3402 0.00242
derived with the average monthly values were used to
Temuco 0.2583 0.00154
Osorno 0.3756 0.00150 estimate monthly global solar radiation for each year.
Puerto Montt 0.3252 0.00290 A comparison between estimated monthly values for
Ancud 0.2820 0.00493 La Paloma, compared to observed monthly values is
Puerto Aysén 0.2870 0.00463 shown in Table 4.
Balmaceda 0.3058 0.00348
Results show that monthly global radiation for each
Punta Arenas 0.3471 0.00389
year can be adequately estimated with the derived
F. Meza, E. Varas / Agricultural and Forest Meteorology 100 (2000) 231–241 235

Fig. 1. (a) Comparison between observed and measured mean monthly global solar radiation using Angström parameters from the literature
(see text); (b) Comparison between observed and measured mean monthly global solar radiation using Angström adjusted parameters;
(c) Comparison between observed and measured mean monthly global solar radiation using Allen parameters from the literature; (d)
Comparison between observed and measured mean monthly global solar radiation using Allen adjusted parameters; (e) Comparison between
observed and measured mean monthly global solar radiation using Bristow-Campbell parameters from the literature; and (f) Comparison
between observed and measured mean monthly global solar radiation using Bristow–Campbell adjusted parameters.
236 F. Meza, E. Varas / Agricultural and Forest Meteorology 100 (2000) 231–241

Fig. 1 (Continued).
F. Meza, E. Varas / Agricultural and Forest Meteorology 100 (2000) 231–241 237

Table 4 Table 5
Regression between calculated and observed monthly global solar Average temperatures (1Ti ) and Fourier series coefficients Ci and
radiation at La Paloma station Di of 20 Chilean localities
Model Slope Upper Lower R2 Locality 1Ti Ci Di
limit. (95%) limit. (95%)
Arica 06.344 0.722 1.412
Allen 1.000 1.010 0.990 0.97 Iquique 05.629 0.629 1.162
Bristow–Campbell 0.994 1.006 0.982 0.96 Antofagasta 06.489 0.269 0.945
Copiapó 14.545 0.640 −0.292
Vallenar 13.193 1.264 0.177
models. Allen’s model presents the best relationship La Serena 07.856 0.220 −0.044
between observed and calculated monthly solar global La Paloma 14.143 0.330 0.345
Quintero 08.366 0.670 0.495
radiation because it explains a large proportion of the Valparaı́so 05.549 0.804 0.330
sample variance. In both models the slope is equal to Santiago 13.917 2.539 1.830
unity with 90 % confidence interval. This verifies that Curicó 14.612 4.235 2.851
the models can be used to estimate monthly values for Constitución 08.397 0.723 0.188
different years. Chillan 13.802 3.991 2.910
Concepción 10.073 2.134 1.365
Temuco 11.494 3.052 2.111
4.2. Global solar radiation distribution functions Osorno 11.031 3.140 1.592
Puerto Montt 08.592 1.862 0.773
Ancud 07.255 1.636 1.051
A probability distribution function for global solar Puerto Aysén 06.823 1.427 0.345
radiation was obtained as a derived distribution, when Balmaceda 09.078 2.130 1.151
radiation is expressed as a function of temperature dif- Punta Arenas 07.019 1.829 0.665
ferences and temperature differences are expressed as
a Fourier series with a random component. This ran-
 
dom error was found to be a random variable with nor- 2πj
mal distribution. This hypothesis was tested in both for 1Tij = 1Ti + Ci cos
12
the Bristow–Campbell and the Allen models using the  
2πj
Anderson–Darling test for normal distribution. Once +Di sin + Eij (6)
12
a distribution model for solar radiation is calculated,
confidence intervals for estimates can be computed. The coefficients Ci and Di are given in Table 5 for the
sites used in this work.
4.3. Temperature difference modelling.

Temperature has a marked seasonal variation due to 4.4. Probability distribution functions
periodicity in the earth’s orbit about the sun. For this
reason temperature variations can be represented us- If X is a continuous random variable with a proba-
ing mathematical cyclic functions. In this paper, dif- bility density function f(x) and Y is a monotonic func-
ferences between maximum and minimum tempera- tion of X, then the probability function of Y can be
tures were modelled using a Fourier series once the obtained multiplying the inverse function by the abso-
stationary component was removed, as suggested by lute value of the Jacobian of the transformation (J) or
Van Wijk and De Vries (1966) and Campbell and Nor- determinant of the first derivative of w(y) with respect
man (1997). These authors applied Fourier series with to X (Walpole and Myers, 1992):
one term to represent air temperatures.
The 1T in location i and month j (1Tij , ◦ C) can be g(y) = f [w(y)]|J | (7)
expressed as a function of mean annual 1T in location
i (1Ti , ◦ C), Fourier series coefficients at location i (Ci , Using this procedure probability density and proba-
Di ) and an error or residual in location i and month j bility distribution functions for RG estimated by Allen
(Eij , ◦ C) as follows: and Bristow–Campbell models were derived.
238 F. Meza, E. Varas / Agricultural and Forest Meteorology 100 (2000) 231–241

4.5. Bristow–Campbell model each location and the corresponding Fourier series co-
efficients (Eq. (6)). Residuals in this case were also
In this case the distribution function is calculated found to be well represented by a normal distribution
using Eq. (3) and replacing 1Tij for its expression in model, so the probability distribution of errors was
terms of annual 1T in each location and the corre- assumed known.
sponding Fourier series coefficients. Combining both The probability density function for solar radiation
the expressions, an equation for the residuals is ob- is shown in Eq. (8).The residuals are given in this case
tained. Residuals were found to be well represented by Eq. (11) and the first derivative by Eq. (12):
by a normal distribution model, so the probability dis-  
tribution of the errors was assumed known. The distri- RGij
Eij = − 1Ti
bution hypothesis was tested using Anderson–Darling (RAij )Krai (P /P0 )0.5
   
test. 2πj 2πj
The probability density function for solar radiation −Ci cos − Di sin (11)
12 12
following Eq. (7), is equal to the product of the normal
density function evaluated at the residuals for location The Jacobian is:
i and month j and the absolute value of the transfor-
mation Jacobian (Eq. (8)). The residuals are given in 2RGij (P0 )
|J | = (12)
this case by Eq. (9) and the first derivative by Eq. (10). Kra
2 (R
i Aij ) P
2
 
g(RGij ) = [J ]f (RGij ) (8)
The CDF is obtained integrating the probability den-
The residuals are given by the following equation ex- sity function. It was evaluated numerically to define
pressed as a function of terms already defined: confidence intervals for global solar radiation. Re-
 !1/2.4 sults for Arica and Vallenar are shown graphically
−ln 1 − RGij /0.7RAij in Fig. 2(c,d). The expected value for global solar
Eij = − 1Ti
Bi radiation given by the CDF using Allen’s model
    are higher than the Angot radiation because the
2πj 2πj
−Ci cos − Di sin (9) limits of integration derived in this case were zero
12 12 and infinite. On the other hand, the CDF using
The first derivative is: Bristow-Campbell model have clear and defined limits
 1/2.4   −1.4/2.4 which are zero and A times the Angot radiation. For
1 1 RGij this reason the CDF obtained with Bristow–Campbell
|J | = −ln 1 −
Bi 6.8 0.7RAij model is more accurate and has smaller confidence
  intervals.
1 1
× (10)
1 − RGij /0.7RAij RAij
The cumulative distribution function (CDF) is ob-
tained by integrating the probability density function. 5. Models applied to daily data
The CDF was evaluated numerically, using very small
intervals and the trapezoidal integration method, to 5.1. Allen’s model
define confidence intervals for global solar radia-
tion. Results for two locations Arica and Vallenar are Allen’s model, 1997 includes a correction term for
shown graphically in Fig. 2 (a,b). barometric pressure which in fact represents the alti-
tude of the station above sea level, since the pressure
4.6. Allen’s model as a function of elevation can be expressed in terms of
the pressure at sea level, the temperature gradient, the
Similarly, for Allen’s model, 1997, the probability temperature at station elevation and the Avogadro air
density function is obtained using Eq. (4) and replac- constant. This correction term is small compared to the
ing 1Tij for its expression in terms of annual 1T in influence of the temperature difference on radiation.
F. Meza, E. Varas / Agricultural and Forest Meteorology 100 (2000) 231–241 239

Fig. 2. (a) Expected values and confidence limits (5 and 95%) of daily mean global radiation using Bristow-Campbell model and Angot
radiation for Arica; (b) Expected value and confidence limits (5 and 95 %) of daily mean global radiation using Bristow-Campbell model
and Angot radiation for Vallenar; (c) Expected value and confidence limits (5 and 95 %) of daily mean global radiation using Allen model
and Angot radiation for Arica; (d) Expected value and confidence limits (5 and 95%) of daily mean global radiation using Allen model
and Angot radiation for Vallenar.

This model tends to over estimate global solar ra- Even though Allen’s model has a larger coeffi-
diation in a daily basis, and frequently estimates radi- cient of determination, the slope is clearly less than
ation in excess of the extra-terrestrial radiation, since unity, indicating that the model over-estimates solar
the condition expressed by Eq. (13) is fulfilled. This radiation.
model does not have a limit for the estimated solar
radiation. 5.2. Bristow–Campbell model
P0
1T > (13) This model is defined solely in terms of temperature
(Kra )2 P
differences and is thus simpler to apply. The value for
This condition is frequently true when the model is A coefficient is 0.7, which is a reasonable value for
applied to points located in interior regions which clear days. This type of day usually is associated to
usually experience large daily temperature variations. large temperature differences.
240 F. Meza, E. Varas / Agricultural and Forest Meteorology 100 (2000) 231–241

Fig. 2 (Continued).

Table 6
6. Conclusions
Regression between calculated and observed daily global solar
radiation at Santiago station
Empirical models to estimate global solar radia-
Model Slope Upper Lower R2
limit. (95%) limit. (95%)
tion are a convenient tool if the parameters can be
calibrated for different locations. These models have
Allen 0.561 0.549 0.571 0.85 the advantage of using meteorological data which are
Bristow–Campbell 1.090 0.979 1.202 0.79
commonly available.
For Chile, the models proposed by Allen and
Bristow–Campbell are adequate and allow estimates of
The behavior of the Bristow–Campbell model is mean average global solar radiation as a function of air
more consistent and reliable, since it has an upper limit temperature variation. Allen’s model has a larger coef-
given by parameter A. The regression analysis shows ficient of determination but requires both atmospheric
that the Bristow–Campbell model performs better pressure and temperature variation measurements.
(Table 6). On the other hand, Bristow–Campbell Models were calibrated for 20 locations in Chile which
model gives consistently a better estimate when represent a wide variation in climatic characteristics
applied to daily data. and hence the procedure described is considered to be
F. Meza, E. Varas / Agricultural and Forest Meteorology 100 (2000) 231–241 241

of general application. Temperature variation can be estimar radiación solar global diaria. Agricultura Técnica 41,
modelled by Fourier series and confidence intervals for 145–152.
global solar radiation estimates can be obtained using Campbell, G., 1977. An Introduction to Environmental Biophysics.
Springer, New York.
derived distribution procedures. Both the models have Campbell, G., Norman, J., 1997. An Introduction to Environmental
limitations when applied to daily data. Solar radiation Biophysics. Soils/AOS 532. Pullman WA.
at locations with large temperature differences are Chang, J.-H., 1968. Climate and Agriculture. Aldine Publishing.
not correctly modelled using Allen procedure and the Chicago.
Davies, J.A., 1965. Estimation of insolation for West Africa. Q.J.R.
Bristow–Campbell model had a better performance.
Meteorol. Soc. 91, 359–363.
Doorenbos, J., Pruitt, W., 1975. Crop water requierements,
Irrigation and drainage paper, 24. FAO, Rome.
References Hargreaves, G., Samani, Z., 1982. Estimating potential
evapotranspiration. J. Irrig. Drain. Eng. ASCE. 108, 225–230.
Allen, R., 1995. Evaluation of procedures of estimating mean Martı́nez-Lozano, J., Tena, F., Onrubia, J., De la Rubia, J.,
monthly solar radiation from air temperature. FAO, Rome. 1984. The historical evolution of Angström formula and its
Allen, R., 1997. Self-calibrating method for estimating solar modifications: review and bibliography. Agric. For. Meteorol.
radiation from air temperature. J. Hydrol. Eng. 2, 56–67. 33, 109–128.
Angström, A., 1924. Solar and terrestrial radiation. Q.J.R. Monteith, J.L., 1966. Local differences in the attenuation of solar
Meteorol. Soc. 50, 121–125. radiation over Britain. Q.J.R. Meteorol. Soc. 92, 254–262.
Bennett, I., 1962. A method of preparing maps of mean daily Penman, H.L., 1948. Natural evaporation from open water, bare
global radiation. Arch. Meteorol. Geophys. Bioklimatol Ser. B soil and grass. Proc. R. Soc. London Ser. A. 193, 120–145.
13, 216–248. Romo, J., Arteaga, R., 1983. Meteorologı́a Aplicada. Universidad
Bristow, K., Campbell, G., 1984. On the relationship between Autónoma de Chapingo. 442 pp.
incoming solar radiation and daily maximum and minimum Turc, L., 1961. Evaluation des besoins en eau d’irrigation.
temperature. Agric. For. Meteorol. 31, 159–166. Evapotranspiration potentielle. Ann. Agron. 12, 13–49.
Castillo, H., Santibáñez, F., 1981. Evaluación de la radiación solar Van Wijk, W., De Vries, D., (1966). Physics of Plant Environment,
global y luminosidad en Chile I. Calibración de fórmulas para 2nd ed. North-Holland, The Netherlands, 382 pp.

You might also like