Professional Documents
Culture Documents
Abstract: Net radiation is a key variable in hydrological studies. However, measured net radiation data are rarely available and are often
subject to error because of equipment calibration or failure. Additionally, point measurements of net radiation do not represent the diversity of
the regional net radiation values, which are needed for many physical and biological processes such as climate monitoring and evapotran-
spiration mapping. The authors present a methodology to estimate daytime net radiation using a combination of remote sensing and climatic
data. This paper expands on previous original research by extending the estimation of net radiation to 24 h under all sky conditions. The
procedure estimates daytime and nighttime net radiation and combines the results to calculate the 24-h net radiation values. The methodology
combines information from satellite and local weather stations to estimate net radiation values. The procedure can estimate net radiation under
all sky conditions using measured or estimated solar radiation. Two different methods are presented to estimate net radiation. Comparisons
between measured and predicted daytime and 24-h net radiation using the two methods resulted in an average ratio ranging from 0.98–1.0 and
a standard error of estimate ranging from 0.83–1.62 MJ m−2 day−1 . Satellite data were used for the calculation of leaf area index, albedo,
and ground temperature. Although satellite data are scarce, the methodology is not limited to satellite imagery. One can easily use ground
level measurements of leaf area index, albedo, and temperature to estimate daytime and 24-h net radiation using the proposed methodology.
DOI: 10.1061/(ASCE)IR.1943-4774.0000539. © 2013 American Society of Civil Engineers.
CE Database subject headings: Evapotranspiration; Remote sensing; Solar radiation; Climates.
Author keywords: Evapotranspiration; Remote sensing; Solar radiation; Net radiation.
ð1Þ location of the sun in the sky. At solar noon, LST = 12 p.m.,
and the sun is at its highest point in the sky. Local solar time is
where Rni = instantaneous net radiation (in units of watts per calculated from the following equation:
square meter); Rsi = incident incoming short wave radiation (also
in units of watts per square meter); σ is the Stephan Boltzmann 1
LST ¼ t þ ðL − Lloc Þ þ Sc − DT ð9Þ
constant, 5.67 × 10−8 Wm−2 K−4 ; Ta and Ts = air and surface 15 std
temperature in degrees Celsius; respectively; and εs = surface emis- where t = local civil time (Pacific Standard Time, Eastern Standard
sivity, calculated using the leaf area index (LAI), i.e., the ratio of Time, and so on); and Lstd = longitude (degree) of the standard
total leaf area for the surface per unit of ground area. According meridian in the local time zone (degrees west of Greenwich).
to Bastiaanssen (1995), εs ¼ 0.95 þ 0.01 LAI when LAI < 3 and For example, Lstd ¼ 75, 90, 105, and 120° for the Eastern, Central,
εs ¼ 0.98 when LAI ≥ 3. εa is the atmospheric emissivity, and Rocky Mountain, and Pacific time zones in the United States. Lloc =
was calculated after Bastiaanssen (1995) from the following local longitude (degrees) west of Greenwich; DT ¼ 1 if daylight
equation: savings time is in effect, DT ¼ 0 otherwise; and Sc = correction
(hours), which accounts for the perturbation in earth’s rotation rate,
εa ¼ 0.85 × ð− ln τ sw Þ0.09 ð2Þ
and is calculated from the following equation:
Atmospheric transmissivity (τ sw ) was calculated from elevation Sc ¼ 0.1645 sinð2bÞ − 0.1255 cosðbÞ − 0.025 sinðbÞ ð10Þ
above sea level (Allen et al. 1998), as follows:
and
τ sw ¼ 0.75 þ 2 × 10−5 ðzÞ ð3Þ
2πðJ − 81Þ
b¼ ð11Þ
where z = elevation (in units of meters). Incident short wave radi- 364
ation, Rsi (in units of watts per square meter), can be measured
For flat surfaces, in which β is zero, Eq. (6) can be simplified to
locally using a standard pyranometer. If local Rsi data are not avail-
able, for clear sky conditions only, it can be estimated using the the following equation:
following equation: cosðθÞ ¼ sinðδÞ sinðϕÞ þ cosðδÞ cosðϕÞ cosðωÞ ð12Þ
Rsi ¼ Gsc × cos θ × dr × τ sw ð4Þ The solar angle for flat surfaces can also be calculated directly
using the sun elevation angle provided in the header file of Landsat
where Gsc is the solar constant (1367 Wm−2 ); θ = solar incidence images, using the following equation:
angle; and dr = inverse relative earth-sun distance, calculated from
the following equation (Allen et al. 1998): π
cosðθÞ ¼ cos −φ ð13Þ
2
2π
dr ¼ 1 þ 0.033 cos J ð5Þ
365 where φ = sun elevation angle in radians.
If measured short wave solar radiation data are available, then
where J = Julian day of the year. they should be used directly in Eq. (1) to calculate Rni . Values of
Incident surface temperature (Tsi ) can be measured either from a albedo, surface temperature, and LAI can be measured at the site
satellite or on the ground. The solar incident angle, θ in Eq. (4), can or obtained from periodic satellite images.
be calculated from the following equation (Duffie and Beckman Rnd was calculated using the following equation by Samani
1980): et al. (2007):
cosðθÞ ¼ sinðδÞ sinðϕÞ cosðβÞ − sinðδÞ cosðϕÞ sinðβÞ cosðγÞ Rs Ta þ 273.15 4
Rnd ¼ Rni ð14Þ
Rsi Tai þ 273.15
þ cosðδÞ cosðϕÞ cosðβÞcosðωÞ
þ cosðδÞ sinðϕÞ sinðβÞ cosðγÞ cosðωÞ where Rs (Mjm−2 ) = daily short wave radiation; Tai = incident air
temperature; and Ta = average of daily maximum and minimum air
þ cosðδÞ sinðβÞ sinðγÞ sinðωÞ ð6Þ temperatures in degrees Celsius. Comparisons of predicted and
measured daytime net radiation over a pecan canopy for 2002
where δ = solar declination in radians, which is calculated from the and 2003 using Eq. (14) are shown in Figs. 1 and 2. The pecan
following equation (Allen et al. 1998): orchard is located approximately 13 km south of Las Cruces,
15
Rnik ¼ εa δðTa þ 273.15Þ4 − εs ðTs þ 273.15Þ4 ð16Þ
10 Ts in Eq. (16) was calculated from the air temperature (Ta) using
the slope of saturation vapor pressure—temperature curve (Δ)
Pecan Orchard, 2002
5 Average Ratio = 1.00 and by iteratively solving for Ts in Eq. (17), as follows (Allen
-2
SEE = 1.37 MJ m day
-1 et al. 2005):
es − ea
0 Δ¼ ð17Þ
0 5 10 15 20 25 Ts − Ta
Downloaded from ascelibrary.org by New York University on 05/16/15. Copyright ASCE. For personal use only; all rights reserved.
-2 -1
Rnd measured, MJ m day
where es and ea (kiloPascals) are saturated vapor pressures at sur-
Fig. 1. Predicted and measured daytime net radiation (Rnd) over a face and air temperature and are functions of Ts and Ta (Celsius),
pecan orchard using Eq. (14) for the year 2002 (the average ratio is respectively (Allen et al. 2005). The slope of the saturation vapor
calculated as predicted/measured, and SEE is the standard error of pressure-temperature curve (Δ) is calculated from the following
the estimate) equation:
17.27Ta
2503 expðTaþ237.3 Þ
Δ¼ ð18Þ
30 ðTa þ 237.3Þ2
25 1:1
Saturated vapor pressure as a function of surface or air temper-
ature, eðTÞ, is calculated from the following equation (Allen
-1
et al. 2005):
Rnd predicted, MJ m day
20
-2
17.27T
eðTÞ ¼ 0.6108 exp ð19Þ
15 T þ 237.3
10 Combining Eqs. (14) and (15), the 24-h net radiation value
Pecan Orchard, 2003 (Rn24) can be calculated from the following equation:
Average Ratio = 1.04
5 -2
SEE = 1.85 MJ m day
-1
-1
-1
Rs predicted, MJ m day
Rnd predicted, MJ m day
-2
-2
25
15
20
10 15
0 5 10 15 20 25 -2 -1
-2 -1 Rs measured, MJ m day
Rnd measured, MJ m day
Fig. 3. Predicted and measured daytime net radiation (Rnd) over a Fig. 6. Predicted and measured solar radiation (Rs) for the year 2003
pecan orchard based on method A (the average ratio is calculated as (the average ratio is calculated as predicted/measured, and SEE is the
predicted/measured, and SEE is the standard error of the estimate) standard error of the estimate)
25
20
Rnd-2003
18 Rn24-2003 1:1
Rnd-2002
Rn24-2002 20
16
-1
-1
Rn 24 predicted, MJ m day
-2
12 15
10
8 10
Fig. 4. Predicted and measured 24-h net radiation (Rn24) over a pecan Fig. 7. Predicted and measured daytime net radiation (Rnd) over a
orchard based on method A (the average ratio is calculated as predicted/ pecan orchard based on method B (the average ratio is calculated as
measured, and SEE is the standard error of the estimate) predicted/measured, and SEE is the standard error of the estimate)
35 20
1:1 18 Rn24-2003
30 Rn24-2002
16
-1
Rn 24 predicted, MJ m day
-1
Rs predicted, MJ m day
25 14
-2
-2
12
20
10
15 8
6
10 Pecan Orchard, 2002 Pecan Orchard, 2002 and 2003
Average Ratio = 0.99 4 Method ‘b’
-2
SEE = 2.49 MJ m day
-1 Average Ratio = 0.98
5 -2 -1
2 SEE = 1.62 MJ m day
0 0
0 5 10 15 20 25 30 35 0 2 4 6 8 10 12 14 16 18 20
-2 -1 -2 -1
Rs measured, MJ m day Rn 24 measured, MJ m day
Fig. 5. Predicted and measured solar radiation (Rs) for the year 2002 Fig. 8. Predicted and measured 24-h net radiation (Rn24) over a pecan
(the average ratio is calculated as predicted/measured, and SEE is the orchard based on method B (the average ratio is calculated as predicted/
standard error of the estimate) measured, and SEE is the standard error of the estimate)
estimate.
1;440
Ra ¼ G × dr½ωs sinðϕÞ sinðδÞ þ cosðϕÞ cosðδÞ sinðωs Þ
π
ð27Þ
α ¼ 0.484ρ1 þ 0.335ρ3 − 0.324ρ5 þ 0.551ρ6 þ 0.305ρ8
where G is the solar constant (0.082 MJ m−2 min−1 ); dr = inverse
− 0.367ρ9 − 0.0015 ð23Þ relative distance from the Earth to the sun; ϕ = latitude; ωs = sunset
hour angle (radians); dr and δ are calculated from Eqs. (5) and (7),
where ρi = reflectance in band i. For Landsat data, α was calculated
respectively; and ωs is calculated from the following equation:
using the following equation, also described by Liang (2001):
ωs ¼ arccos½−tanðϕÞ tanðδÞ ð28Þ
α ¼ 0.356ρ1 þ 0.13ρ3 þ 0.373ρ4 þ 0.085ρ7 − 0.0018 ð24Þ
Allen (1995) suggested calculating K r from the following
Because of the variation related to pixel amplitude in satellite
equation:
sensors, the spatial resolutions of the calculated parameters are
limited to 90 and 60 m in ASTER and Landsat-5, respectively. 0.5
P
K r ¼ K ra ð29Þ
P0
Results and Discussion where P = mean atmospheric pressure at the site, in units of
This paper describes two different methods to estimate daytime and kiloPascals; P0 = mean atmospheric pressure at sea level
24-h net radiation. (101.3 kPa); and K ra is an empirical coefficient that is equal to
• Method A calculates Rnd from satellite measurements of albe- 0.17 and 0.2 for interior and coastal regions, respectively. Eq. (26)
do, normalized difference vegetation index (NDVI), surface was used to estimate the daily solar radiation (Rs) values for the
temperature (Ts), ground measurements of short wave radiation years 2002 and 2003. Figs. 5 and 6 show the comparison between
(Rsi and Rs), and air temperature (Tai and Ta) using Eqs. (1) and measured and predicted Rs values. The average ratio of predicted
(14). Rnn is calculated using Eq. (15). over measured values in Figs. 5 and 6 were 0.99 and 1.04, respec-
• Method B calculates Rnd and Rn24 in the absence of ground tively, and the SEE were 2.49 and 2.88 MJ m−2 day−1 , respectively.
measurements of short wave radiation (Rsi and Rs). In this Using method B, the daytime Rnd and Rn24 values were
method, Rnd is calculated from satellite measurements of albe- estimated for a pecan orchard for the years 2002 and 2003. The
do, NDVI, and surface temperature using Eqs. (1), (4), and (14), comparisons are shown in Figs. 7 and 8. The average ratio of
and estimated values of daily solar radiation (Rs) based on predicted over measured net radiation shown in Figs. 7 and 8 were
daily maximum and minimum air temperature (Hargreaves and 1.0 and 0.98, respectively, and the SEE values were 1.4 and
Samani 1982). This method is only applicable for short-term 1.62 MJ m−2 day−1 , respectively. The results are summarized in
clear sky conditions in which Rsi can be estimated using Eq. (4). Table 1. Figs. 5 and 6 show that method B over-predicts at the
lower end of the solar radiation values, which occur early in the
year. The over-prediction can potentially result in overestimation
Method A of Rnd values. The over-prediction could be attributable to the tem-
Figs. 3 and 4 compare the predicted and measured values of Rnd poral variation of the calibration coefficient in Eq. (26), which can
and Rn24 for 10 days during 2002 and 8 days during 2003 over a be corrected by adjusting the calibration coefficient for the early
pecan canopy based on method A. The ratio of predicted over mea- period of the year.
sured values in Figs. 3 and 4 were both 0.99 and the standard errors The nighttime net radiation in this study was calculated using
of estimate (SEE) were 0.83 and 1.28 MJ m−2 day−1 , respectively. estimates of ground temperature (Ts), which were obtained through
The SEE, which is the dispersion of the observed values about the iterative solution of Eq. (17), assuming a constant slope for the sa-
regression line or a measure of accuracy of prediction, was calcu- turated vapor pressure curve. The results can be further improved
lated as follows: by directly measuring the ground temperature values.
rPffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðY − Y 0 Þ2
SEE ¼ ð25Þ Conclusions
n−1
where SEE = standard error of estimate; Y = measured value The results of the comparison show that the methodologies pre-
(e.g., Rn measured); Y 0 = predicted value (e.g., Rn predicted); sented in this paper can estimate daytime and 24-h net radiation
and n = number of observations. with reasonable accuracy, considering the paucity of net radiation
Gautier, C., Diak, G., and Masse, S. (1980). “A simple physical model to
Allen, R. G. (1995). “Evaluation of procedures for estimating mean
estimate incident solar radiation at the surface from GOES satellite
monthly solar radiation from air temperature.” Rep. Prepared for the
data.” J. Appl. Meteorol., 19(8), 1005–1012.
Food and Agricultural Organization (FAO) of the United Nations,
Hargreaves, G. H., and Samani, Z. A. (1982). “Estimating potential evapo-
Rome, Italy.
transpiration.” J. Irrig. Drain. Eng., 108(3), 225–230.
Allen, R. G., et al., eds. (2005). The ASCE standardized reference evapo-
transpiration equation, ASCE, Reston, VA. Liang, S. (2001). “Narrowband to broadband conversion of land surface
Allen, R. G., Pereira, L. S., Raes, D., and Smith, M. (1998). “Crop evapo- albedo. I. Algorithm.” Remote Sens. Environ., 76(2), 213–238.
transpiration. Guidelines for computing crop water requirements.” FAO Ma, Y., Su, Z., Koike, T., and Menenti, M. (2002). “Determination of
Irrigation Paper No. 56, Food and Agricultural Organization of the regional net radiation and soil heat flux over a heterogeneous landscape
United Nations, Rome, Italy. of the Tibetan Plateau.” Hydrol. Process., 16(15), 2963–2971.
Bastiaanssen, W. G. M. (1995). “Regionalization of surface flux densities Payero, J. O., Neale, C. M. U., and Wright, J. L. (2006). “ Near-noon albedo
and moisture indicators in composite terrain: A remote sensing ap- values of alfalfa and tall fescue grass derived from multispectral data.”
proach under clear skies in Mediterranean climates.” Ph.D. dissertation, Int. J. Remote Sens., 27(3–4), 569–586.
CIP Data Koninklijke Bibliotheek, Den Haag, The Netherlands. Rosenberg, N. J., Blad, B. L., and Verma, S. B. (1983). Microclimate—The
Bastiaanssen, W. G. M. (1998). Remote sensing in water resources biological environment, Wiley, New York.
management: The state of the art, International Water Management Samani, Z., Bawazir, A. S., Bleiweiss, M., Skaggs, R., and Tran, V. D.
Institute, Colombia, Sri Lanka. (2007). “Estimating daily net radiation over vegetation canopy through
Campbell, G. S. (1977). An introduction to environmental biophysics, remote sensing and climate data.” J. Irrig. Drain. Eng., 133(4),
Springer, New York. 291–297.