Journal of Hydrology, 107 (1989) 99-111
99
Elsevier Science Publishers B.V., Amsterdam - - Printed in The Netherlands
[11
THE UTILITY OF CLIMATIC STATION DATA IN MAKING HOURLY POTENTIAL E V A P O T R A N S P I R A T I O N ESTIMATES FOR REMOTE SENSING STUDIES
MANFRED OWE
Hydrological Sciences Branch, Laboratory for Terrestrial Physics, NASA]Goddard Space Flight Center, Greenbelt, MD 20771 (U.S.A.) (Received May 9, 1988; accepted after revision August 22, 1988)
ABSTRACT Owe, M., 1989. The utility of climatic station data in making hourly potential evapotranspiration estimates for remote sensing studies. J. Hydrol., 107: 99-111. Regional estimates of evapotranspiration and soil moisture are often highly inaccurate due to a lack of representative meteorological data with which to calculate the moisture and energy fluxes. A procedure is outlined which uses available hourly climatic station data to estimate these moisture and energy fluxes. These hourly climatic station observations ancl synthesized radiation were compared to detailed meteorological and radiation flux measurements taken up to 100 km away. The two data sets compared well in both trend and magnitude. Solar and net radiation were modelled by using vapor pressure and observed cloud cover and compared well to values observed at the study sites. Hourly potential evapotranspiration estimates were made with both the micromet measurements and the climatic station data for each region, and alsG compared well. This study was performed for two areas of different climatic conditions, a natural prairie in Kansas and a bare soil in Beltsville, Maryland. Large-area estimates of hourly potential evapotranspiration made with climatic station data may be useful when combined with additional midday satellite observations in the thermal, visible, and near-infrared bands to estimate relevant vegetation parameters, actual evapotranspiration, and soil moisture.
INTRODUCTION
Satellite remote sensing techniques have provided an opportunity to make large area observations of a variety of surface phenomena. Since these measurements reflect the average surface conditions integrated over an entire instantaneous field of view (IFOV) or footprint~ they can be correlated with or used to infer a variety of areally representative surface fluxes and physical parameters. The remote measurement may often be a more accurate indicator of large area surface condition than areally averaged point measurements. Unfortunately, in order to derive the proper relationships, one generally needs to acquire a variety of surface data with which to conduct validations. Obtaining reliable and representative ground data is often difficult, though, unless costly collection efforts have been specifically established for a given
0022-1694/89/$03.50
© 1989 Elsevier Science Publishers B.V.
100 study. Variables such as evapotranspiration and soil moisture may be modelled, but estimates are often highly inaccurate due to a lack of meteorological data with which to calculate the moisture and energy fluxes. This becomes especially evident when working at regional scales. Investigators must often work with potential evapotranspiration (PEt) alone or as the primary input to moisture index models. This is not altogether bad, however, if one uses radiation-based PEt models rather than empirical models relying solely on temperature. In this study, synthesized radiation and routine weather station observations made at various distances are compared to detailed micrometeorological measurements made at two sites. Estimates of potential evapotranspiration are also made and compared, using the two coincident data sets. Potential evapotranspiration (PEt) has been widely used since its introduction, and its applicability in many areas of climatological research such as evapotranspiration estimates, water budgets, or comparing the climatic regimes of different regions has been widely demonstrated. The decision to use PEt is often based on the lack of the many multilayer micrometeorological observations usually required for more sophisticated physical Et models. Most PEt models, such as Thornthwaite (1955), Hamon (1963), and Blaney and Criddle (1950) are temperature-based models, relying on mean daily or monthly temperature. Generally, these techniques are highly empirical and very site specific, in that they were developed for and function best in a given region or climatic regime. These techniques continue to be popular because temperature is the easiest and most commonly measured atmospheric parameter. Radiation-based methods are generally more physically sound, although many of these techniques are also, to a large extent empirical (Jensen and Haise, 1963; Nicks and Harp, 1980). The Penman (1948) formula yields reasonable estimates for PEt, although some empiricism remains with the original calculation of the aerodynamic term. Van Bavel (1966) introduced a method whereby the aerodynamic resistance component is replaced with a physically based vapor transfer coefficient. In that study, soil moisture conditions were maintained at a level such that actual Et would equal PEt throughout the experiment. Van Bavel (1966) subsequently developed a more useful and widely applicable definition for ,potential evapotranspiration. Even though micrometeorological measurements may not ~_ave been made at the study site, or within the region of interest, climatic station data are often available and are generally adequate for regional application at a daily time scale, and may even be sufficiently representative for more local hourly applications. Since radiation is a major driver of the evapotranspiration process, reasonable estimates of this parmneter should account for most of the requirements affecting the vapor transfer process. Data is presented demonstrating that synthesized radiation together with routine weather station observations may be used to yield reasonable estimates of hourly potential evapotranspiration.
101 TABLE 1 Parameters used in the analysis either measured or estimated at the study sites and climatic stations Location
Study site Climatic station
Atmospheric parameters Incoming solar
Net radiation
Air temp.
Air pressure
Vapor* pressure
Wind speed
% cloud cover
M E
M E
M M
E M
M M
M M
M
*ea for climatic stations is calculated from dew point temperature; E = estimated or modelled parameter; M = measured or observed parameter.
APPROACH
Micrometeorological data for several 24h periods from two different climatic regimes, the Konza Prairie area near Manhattan, Kansas, and the Beltsville Agricultural Research Center (BARC) in Maryland were used as a baseline. These measurements were originally made during unrelated field studies in the summer of 1985. Potential Et was calculated by using these data in a modified Penman approach described below. Atmospheric parameters were then replaced by measurements obtained from the nearby climatic stations, while radiation values were synthesized by a procedure also described below. Hourly meteorological measurements taken at Concordia, Kansas and the Marshall Army Airfield, Fort Riley, Kansas were used to replace the data from the Konza Prairie study site, while measurements from Tipton Army Airfield, Fort Meade, Maryland were used to replace the Beltsville observations. The climatic stations are approximately 100, 20 and 10 km respectively from the two original field sites. Parameters measured at the study sites and the climatic stations are indicated in Table 1.
Input models ~vaporation is primarily a function of two processes. The first is described in terms of radiation, and depicts the energy available to convert water from a liquid to a vapor phase. The second is an aerodynamic term which describes the sink strength or the ease with which vapor is transferred to the atmosphere. Evaporation is the net upward rate, and is directly proportional to the difference in vapor pressures between the surface and the atmosphere. Penman's (1948) original expression for PEt has been derived in a variety of forms (Sellers, 1965; Monteith, 1973; Mather, 1978) and can be shown to take the
102
form:
LE
=
A -'(R. + G)+ LE, ~ A -+1
(1)
where L is the latent heat of vaporization, E is the amount of water evaporated or transpired, A is the slope of the saturated vapor pressure vs. temperature curve, ~ is the psychrometer constant, Rn is the net; radiation, and G is the soil heat flux. The term E, is the mass transfer term which wa:~ determined empirically by Penman (1948) from the Dalton equation. Van Bavel (1966) introduced an expression based on the turbulent transfer of water vapor between a vegetated surface and the atmosphere, taking the form: E, =
p~k 2ua (es - ea)
(2)
ea Iln/'Za z0 d / 1 2 where e is the ratio of the molecular weight of water to that of air or 0.622, k is the von Karman constant and taken as 0.41, u, is the wind velocity, P~ is the air pressure, and p is the air density and may be calculated from the ideal gas equation p = P , / R T , in which R is the specific gas constant for air and T i s the absolute temperature. The terms es and e, are the saturated vapor pressure at the ambient air temperature and the measured vapor pressure respectively, z, is the measurement height taken as 160cm and d is the displacement height which was ~et to zero. Difficulty is usually encountered in selecting an appropriate roughness parameter, z0. A technique suggested by Tanner and Pelton (1960), based on an estimated uniform vegetation height, h, and taking the form: Z0
--
(3)
h 0"964/7.638
was used. An average uniform vegetation height of 10cm was used for the Konza Prairie (z0 - 1.2cm), while a vegetation height 0.01 cm (z0 - 0.0015) was used with the Bel~sville data to reflect bare soil conditions. Radiation estimation
Radiation measurements are rarely taken on a routine continuous basis at public climatic stations. Reasonable estimates, though, may be calculated, and compare well with the measured values of the base data. Th~:, PEt model requires net radiation (Rn) as an input parameter, and is defined as: Rn =
Rsd-
Rsu + R i d -
Rlu
(4)
103
where Rsd and R~, are the downward and reflected solar radiation respectively, and R~a and R~u are the downward and upward terrestrial radiation. Incoming solar radiation at the earth's surface for clear sky conditions was calculated as follows in a manner similar to Eagleson (1970), such that:
R~
= blo sin08un e x p ( - n/sinOs~n)
(5)
where I0 is the average global insolation at the top of the atmosphere, b is a fitting parameter found to be 0.8, and n is an air turbidity factor which is described as a function of the vapor pressure. Both were derived by optimization with observed data for a wide range of atmospheric conditions. The term 0~,n is the angle of the sun above the horizon. This last term is a function of latitude, time of day, and day of the year, and is determined as follows: sin0~n = sinasin~ + cosacos~cos~
(6)
where ¢ is the latitude, a is the solar declination angle and ~ is the hour angle of the sun. The last two parameters may be estimated from: a = 23.45sin[0.9863(k - 81)]
(7)
and: = u [ 1 2 ( t - 12)
(8)
where k is the day of the year and t is the hour of the day. Cloud cover can have a significant effect on incoming solar r a d i a t i o ~ so clear-sky estimates must be adjusted accordingly. A high altitude, relatively thin cloud layer may reflect less than 20% of the incoming shortwave ~adiation, while a low thick layer of stratus or stratocumulus may reflect over 80%. Lumb (1964) presents data on the fraction of total solar radiation transmitted through the atmosphere for different cloud types. Studies have also related daily solar radiation to sunshine hours (Schultz, 1976; Rietveld, 1978). A partial cloud cover may cause no reduction in the downward component, and may even exceed the flux beneath a clear sky for short periods due to reflection at and around the edges. Data presented by Cunniff (1958) indicates no noticeable reduction in the downward shortwave flux for a broken cloud cover of up to 30%. Cloud cover, however is very much a local feature at any given time. The sun may shine through an opening in the clouds for long periods, but may also be obscured for extended periods by a single slow moving cloud. Sky conditions in the form of percent total cloud cover and ceiling height are reported along with hourly atmospheric measurements at climatic stations, and may be used to adjust incoming solar radiation. Average data relating percent cloud cover to solar radiation incident at the surface (Cunniff, 1958) was used to derive the relationship in Fig. 1. The term, fl, is then used to define the ratio Rsd/R*sd. Figures 2a-e compare the observed incoming solar radiation flux to that calculated. A variety of cloud cover conditions are represented, and although some variability exists between the observed and calculated values, agreement
104 I
I
I
I
100
-
~ C O V E R
8O ,HIGH
A o o
60
1=
40
RsdlRsd*
/
,il,;y I
I - , ~ . " - " LOW CLOUDS
e¢-
20
/$ = 113.7 - 79.8e -2"52(1"c) 0
I
I
I
I
0.20
0.40
0.60
0.80
1.00
FRACTION CLEAR SKY (1-c)
Fig. 1. The relationship between percent cloud cover and fl, the ratio of cloudy sky incoming solar radiation (R~d) to clear sky incoming solar radiation (R~).
appears to be quite good. It should be noted that the same calibration coefficients have been used in each case. Hourly cloud cover observations for each of the days represented are given in Table 2. Upward solar radiation is estimated as: Rsu
=
(9)
~Rsd
where = is the average surface albedo. Average values for typical surfaces may be found in a variety of sources (Sellers, 1965; Monteith, 1973). A value of 0.2 (a)
KONZA PRAIRIE ~l/r~,.~. INCOMING SOLAR 19 JUNE 1 9 8 5 i Y | " \X AND NET
?
(b)
I
I
BELTSVILLE ~ 11 JULY 1985 /
800
800
600
600
4O0
400
2oo
2O0
I
I
INCOMING SOLAR ~ AND NET
E ffl t-
Z Z O
< IE
~/,:f" 0
8
12
16
TIME OF DAY (HOURS)
I 20
I
8
.........
'/UuI'LLI'u 12
Mn , 16
°'~'~l 20
TIME OF DAY (HOURS)
Fig. 2. Comparison of measured and synthesized solar and net radiation for a wide range of cloud cover conditions at the two test areas.
105
(c)
I 8OO
(dl
I ¢] 'NCOMIDNAGTSoOLAR -~
=
- 1:EjL~yV'II'~I 5
= ~ . _ _
!
I
700
/[
E 600 (/) I-,,. I- 500
400
Z Z
-
,_i/
"
\
-
I
_o
8
12
16
20
600 z 400
I RI~LI'SVILLE = AUGUST 1988~
800
I
-!;l / I 8
I 12
I 16
I
2O
TIME OF DAY (HOURS)
TIME OF DAY (HOURS) (e)
\%
E
2OO
100
-
800
C)
I-,< 300 Q <~ ¢1:: 200
|
INCOMING SOLAR RADIATION /,
?
!
BELTSVILLE 13 JULY 1985
I I 'NCOM'NG SOLAR RADIATION
-
E 600 Z
400
2O0
0
0
8
12
16
20
TIME OF DAY (HOURS)
was used for the Konza Prairie site, while 0.15 was used for the Beltsville site. While albedo varies according to surface character, it is also a function of incidence angle (Geiger, 1961). This dependence is primarily observed over plane surfaces, and is less a factor over vegetation (Kuhn and Suomi, 1958). Longwave or terrestrial radiation is the radiant flux due to the emission from land surfaces and atmospheric gases. It is an extremely difficult phenomenon to measure, and it is more practical and in many instances easier to model from other measureable parameters. A variety of techniques are available which will model upward and downward terrestrial radiation adequately under both clear and cloudy atmospheric conditions (Brutsaert, 1982). The Brunt Equation (Anderson, 1954) is often used to estimate net
106 TABLE 2 H o u r l y cloud cover o b s e r v a t i o n s at the i n d i c a t e d climatic s t a t i o n s in t e n t h s of t o t a l sky cover H o u r l y cloud cover o b s e r v a t i o n s 06
07
08
09
10
11
12
13
14
15
16
17
18
19
20
0 0
0 0
0 0
1 0
1 0
3 2
4 2
4 3
4 3
3 1
3 2
3 3
2 3
1 0
1 0
8 10 10
7 10 0 10
6 l0 0 10
2 10 1 10
1 10 0 10
2 10 1 9
1 l0 1 8
1 9 1 6
1 9 1 6
1 5 1 9
06.19-85 Riley Concordia
Meade 07-11-85 07-12-85 07-13-85 08-02-85
terrestrial radiation, and is defined as: Rln
=
a[T:
-
(c + d x / e a ) T ~ ] ( 1
-
aC)
(10)
where Ts and Ta are the surface and air temperature, C is the fractional cloud cover, a is a cloud height parameter, and c and d are empirical constants which may vary geographically. Representative values have been compiled by Anderson (1954) or may be calculated from available data. Although empirical, this equation incorporates a number of key parameters that affect longwave radiation and gives reasonable estimates. In the absence of surface temperature Chang (1968) set Ta = Ts, resulting in the following expression which was used in the study: Rln
--
a ~ ( 0 . 5 6 - 0.08x/e~)(1 - aC)
(11)
A value for a of 0.6, indicating a medium height cloud cover was used. Since this equation uses only air temperature, it may either under- or overestimate depending on the surface cover characteristics. If the surface is vegetated, air temperature is generally greater than the surface and thus the calculated net longwave component may be overestimated. Conversely, air temperature is generally less than that of a bare surface, and the resulting net longwave values may be underestimated. Both cases were observed in the study. Figures 2a and b compare the estimated net radiation values to those observed at the two sites. The soil heat flux, G, is difficult to estimate. It may be measured with sophisticated instrumentation, but this is impractical for most studies, especially large area investigations. The soil heat flux may also be estimated from temperature gradients or theoretical calculations. For regional applications, the parameters required for accurate soil heat flux estimates are generally unavailable. Simplified empirical relationships may be useful in some situations. The soil heat flux was expressed as a fraction of net radiation, such
107 that:
G
mRn
=
(12)
where m is an empirical constant. Brutsaert (1982) cites a broad range of values for the constant based on a variety of studies. For bare soil conditions, an average value of 0.,3 is suggested and subsequently used in this case, while for vegetated soils, values from zero to 0.1 may be appropriate. A value of 0.1 was subsequently used for the Konza Prairie site. Choudhury et al. (1987) describe a more rational approach for estimating the soil heat flux by applying an exponential function to net radiation based on leaf area index. This relationship states that: G
~ e x p ( - flL)Rn
=
(13)
where ~ and fl are empirical constants which must be determined from data and L is the leaf area index.
Comparison of meteorological measurements Typically, the meteorological parameters of the base data compared well with the coincident measurements from the nearby climatic stations. Though some differences do exist with regard to the magnitude of maxima and minima and their respective times of occurrence, these do not appear to be great. While the difference in location and subsequent microclimate may account for some of the disparity, much is undoubtedly also due to the fact that the climatic
I
I
!
I
AIR TEMPERATURE
.
KONZA PRAIRIE
/'J
|
i
306
~
~AIR
.....X.~..~ ,~.;...." "~.'\
19 JUNE 1985
r,/
295
!
304
I
"
302
TEMPERATURE |
"'u<'IW.-"/ -
\
l
I,M
uJ L~ Lu a Z 290
/
I,M m
t ""'. o.
uJ I,-
/l/ //,"
l
285 - ~
300
'
298
I
"''
296
I:
//~"
"'"".
....
"/7 ~.%~.~lf"
..............
KONZA PRAIRIE FORT RILEY CONCORDtA
~ .....
-
BELTSVILLE FORT MEADE
294
I
I
I
I
n
I
I
I
I
4
8
12
16
20
24
8
12
16
TIME OF DAY (HOURS)
2O
TIME OF DAY (HOURS)
Fig. 3. Comparison of vapor pressures measured at the test sites to those of the nearby climatic stations. Climatic station pressures were derived from calculated relative humidity and saturated vapor pressure.
108
17
I
I
i
I
KONZA
I
I
PRAIRIE
I
I
:"
I
!"
I1'~
I-"
tdJ /
I
I
I
I
I
i
4
i
/",,
'"
,
.
I VAPOR
PRESSURE
:l "
:.:
I
e
~
/
i
12
',. t:'
~
/ I
/ t
~
/
--
/
BELTSVILLE
~
11 JULY 1985
i
20
%4
/
~
/ I
16
-t
A;-
•/ I
< , o
t\
; "V,'~ ":
i
8
,,,... .......
/~ /
/ .... KONZA PRAIRIE~ / ......... FORT RILEY / -CONCORDIA ~ I
,
l
/
"lli~":
.....
I/
24 F"
"1
?'.
k
,,
I-.
11
ij
J "
//' ~ ,
z
i
19 JUNE 1985
I"
is
i
I
PRESSURE
t,,"...
16
3
I
VAPOR
- BELTSVl - FORT MEADE
14r
ii
....
|
24
TIME OF DAY (HOURS)
V
I
I
I
8
12
16
20
TIME OF DAY (HOURS)
Comparison of air temperatures measured at the test sites to those of the nearby climatic stations. F i g . 4.
station data are instantaneous readings while the field data are an average of continuous measurements taken over 30 min periods. Figures 3-5 compare the atmospheric parameters observed at each study site to the hourly data from the respective nearby climatic station(s). The paired parameters appear to be in reasonably close agreement, although wilid speed demonstrates the great~st difference in the comparisons. This is especially noticeable in the wind speed measurements from the Concordia station. This is I
KONZA
I
|
I
PRAIRIE
I
|
WINDSPEED
^
19 JUNE 1985
!
/'~
i\.
.
BELTSV|LLE
500
300
¢n 400 e
250
I
I
f ~
WINDSPEED
11 J U L I~ , /i ~~ ~ _ ~/ ~
'T
E tJ
Z C3 300 u.I uJ 13.. U3 ZEl 20C
...... KONZAPRAIRIE | . . . . . . . . .
I k
I
100 ,+i#'........ i I:
0.0
FORT RILEY CONCORDIA
:
~: i
4
l',
t"
i "~... r '.~
8
TIME
/'~f |.:'"~,l
I
12
150
."! /!
~ "...:
200
::i
..l"
:~1
;il
U
/
t "
/
~. "
:~
-
I
16
OF DAY (HOURS)
100
i
,
/
FORT MEADE
"
I
I
20
I
24
50
~
,
,
,
8
12
16
~-/ 20
T I M E OF DAY (HOURS)
Fig. 5. Comparison of wind speeds measured at the testsitesto those of the nearby climatic stations.
109
not unreasonable, since wind speed is highly variable in space, and by far the most temporally variable of the measured parameters. A factor which must be given some consideration when using climatic station data is the representativeness of the area surrounding the observation station with respect to the study region. Differences in vegetation, elevation and topographic chr~racteristics greatly affect the microclimate and resultant instrument readings. Although smaller basins rarely contain more than one clim~.tic Station, when making regional Et estimates it would be appropriate to use as many observation stations as available to generate either areal average values or to use a partial area approach. RESULTS
Hourly estimates of potential evapotranspiration (PE t) were made using the original measurements from the study sites. These data were then replaced with meteorological observations from the nearby climatic stations. The radiation measurements were replaced by the synthesized data which were shown to be reasonably accurate in earlier figures. Potential Et estimates were then made again with the replacement data. The estimates appear in reaso,~able agreement with those resulting from the original data. Figure 6 compares PEt estimates resulting from each study site. The greatest deviations occur during periods of highly variable cloud cover. The estimates appear to be best under clear sky conditions (less than 30%). Estimates are also good during periods of denser cloud cover as long as the distribution is relatively uniform across the sky and the clouds are moving steadily. Variable cloud cover conditions will certainly affect the incoming radiation, and will subsequently have a significant effect on the PEt estimates.
!
1.0
198~~
I
I
I
K O N Z A PRAIRIE 1 9 JUNE
A
19~'
I
i
I
0.9
I
POTENTIAL
I
BELTSVILLE 11 JULY
Et 3.8
.
I
POTENTIAL
Et
0.7 0.8
0.6
re" "1"
/ ""
',.
\
0.6
0.5
_z
-
f
-
0.4
Z . o.
LU
/,,,"
0.4
,..,...',, \ ,~
/.(/
0.2
/ ! . " // /
/,'
....
h' O.0
I
"... '.. ~,
........ KONZA PRAIRIE
I 8
~ORT ,ILEY
"..".
\'-.
co.coRo,A
II
I 12
I
I 16
TIME OF D~Y (HOURS)
I ........ 20
0.3 0.2 O.1
8
12
16
20
TIME OF DAY (HOURS)
Fig. 6. Comparison of potential evapotranspiration calculated with testsiteand observed radiation data to that calculated wlth climatic station data and synthesized radiation.
110
The large difference in calculated P E t seen with the Konza and Concordia data is due to the difference in wind speeds and vapor pressures observed a~ the two sites. Summary and conclusions
Meteorological parameters from two areas of differing climatic regimes were compared to hourly climatic station data taken up to 100 km away. The two data sets compared remarkably well in both trend and magnitude. Modelled radiation, adjusted for cloud cover, also compared well to the flux measurements of the base data. Ho:~rly observations of percent total sky cover were obtained from the climatic stations and are critical in making the radiation estimates. Potential evapotranspiration estimates were made with both the micrometeorological measurements and the hourly data for each region and also compared well. Since the radiation term is the strongest parameter of the potential Et model, good estimates of net radiation are the most essential component to making accurate model calculations. As incoming radiation is highly affected by cloud cover, observations of sky conditions greatly enhance the accuracy of the net radiation and subsequent PEt estimates. Obtaining ground measurements for large-scale remote sensing studies is difficult for some highly spatially variable parameters, especially evapotranspiration and surface moisture. Long-term data sets are even more difficult to assemble. Meteorological measurements are more readily available and generally more areally representative than point measurements of surface parameters and when used in energy balance models may be used to generate estimated values of these parameters. ModeUed radiation together with commonly available hourly climatic station observations may be used in place of flux and micromet measurements, even if taken at considerable distances, and have been shown to yield comparable estimates of potential evapotranspiration. One should be aware of the representativeness of the area surrounding the climatic station relative to the entire region of interest, and if more than one climatic station exists with a region, then areal estimates would most likely be improved by the combined use of all available data. Partial area techniques or even simple interpolation between known points may yield more representative areal estimates of surface parameters. Hourly estimates of potential moisture loss may be even more useful when combined with daily midday remote observations in the thermal, visible and near infrared bands to estimate actual evapotranspiration, soil moisture, and relevant vegetation parameters. ACKNOWLEDGEMENTS
The author is indebted to Peter Camillo for his reviews and unselfish assistance with the modelling. The helpful reviews and suggestions of R.J.
111
Gurney and W.J. Shuttleworth are gratefully appreciated. Personnel of the cited climatic stations are also thanked for providing the hourly observations used in the study. REFERENCES Anderson, E.R., 1954. Energy budget studies. U.S. Geol. Surv., Prof. Pap., No. 269. Blaney, H.F. and Criddle, W.D., 1950. Determining water requiremev~ in irrigated areas from climatological and irrigation data. Soil Conserv. Serv. Tech. Pap., bo. Brutsaert, W.H., 1982. Evaporation into the Atmosphere: Theory, History and Applications. Reidel, Boston, Mass., 299 pp. Chang, J.H., 1968. Climate and Agricult~Jre: An Ecological Survey. Aldine, Chicago, Ill., 296 pp. Choudhury, B.J., Idso, S.B. and Reginato, R. J., 1987. Analysis of an empirical model for soil heat flux under a growing wheat crop for estimating evaporation by an ;,r.frared temperature based energy balance equation. Agric. For. Meteorol., 39: 283-297. Cunniif, C.V., 1958. Solar radiation on walls facing east and west. Air Cond. Heat. Vent., 55(10): 82-88. Eagleson, P.S., 1970. Dynamic Hydrology. McGraw-Hill, New York, N.Y., 461 pp. Geiger, R., 1961. Das Klima der bodennahen Luftschicht. Vieweg, Braunschweig (Engl. transl.: Scripta Technica, Inc., 1965, The Climate near the Ground, Harvard University Press, Cambridge). Hamon, W.R., 1963. Computation of direct runoff amounts from storm rainfall. Int. Assoc. Sci. Hydrol. Publ., 63: 52-62. Jensen, M.E. and Haise, H.R., 1963. Estimating evapotranspiration from solar radiation data. J. Irrig. Drain. Div., Proc. ASCE, 89: 15-41. Kuhn, P.M. and Suomi, V.E., 1958. Airborne observations of albedo with a beam reflector. J. Meteorol., 15: 172-174. Lumb, F.E., 1964. The influence of cover on hourly amounts of total solar radiation at the sea surface. Q. J. R. Meteorol. Soc., 90: 43-56. Mather, J.R., 1978. The Climatic Water Budget in Environmental Analysis. Heath, fornnto, Ont., 239 pp. Monteith, J.L., 1973. Principles of Environmental Physics. Elsevier, New York, N.Y., 241 pp. Nicks, A.D. and Harp, J.F., 1980. Stochastic generation of temperatures and radiation. J. Hydrc,!., 48: 1-17. Penman, H.L., 1948. Natural evaporation from open water, bare soil and grass. Proc. R. Soc. London, A 193. Rietveld, M.R., 1978. A new method for estimating the regression coefficients in the formula relating solar radiation to sunshine. Agric. Meteorol., 19: 24~-252. Schultz, R.E., 1976. A physically based method of estimating solar radiation from suncards. Agric. Meteorol., 16: 85-101. Sellers, W.D., 1965. Physical Climatology. Univ. Chicago Press, Chicago, Ill., 272 pp. Tanner, C.B. and Pelton, W.L., 1960. Potential evapotranspiration estimates by the approximate energy balance method of Penman. J. Geophys. Res., 65: 3391-3413. Thornthwaite, C.W. and Mather, J.R., 1955. The Water Balance. Lab. Clim. Publ., No. 8., Centerton, N.J. Van Bavel, C. M. H., 1966. Potential evapotranspiration: The combination concept and its experimental verification. Water Resour. Res., 2:455 437.