Sie sind auf Seite 1von 12

Solar Energy 153 (2017) 611–622

Contents lists available at ScienceDirect

Solar Energy
journal homepage: www.elsevier.com/locate/solener

Estimation of photosynthetically active radiation using solar radiation in


the UV–visible spectral band
Zhian Sun a,⇑, Liang Hong b, Liu Jingmiao c,d, Shi Guoping e
a
Environment & Research Division, Australian Bureau of Meteorology, Melbourne, Victoria 3001, Australia
b
Meteorological Observational Centre, China Meteorological Administration, Beijing 100081, China
c
Institute of Atmospheric Environment, China Meteorological Administration, Shenyang 110016, China
d
Chinese Academy of Meteorological Sciences, Beijing 100081, China
e
Institute of Geographic and Remote Sensing, Nanjing University of Information Science and Technology, Nanjing, China

a r t i c l e i n f o a b s t r a c t

Article history: In this paper, we discuss a method using the UV–visible band flux to estimate photosynthetically active
Received 6 September 2016 radiation (PAR). We use radiative transfer models to calculate the ratio of PAR to the UV–visible flux and
Received in revised form 2 June 2017 confirm that this ratio is close to constant regardless of atmospheric conditions. The best ratio identified
Accepted 5 June 2017
in this study is 0.8644. We further introduce a UV–visible band model and use it to calculate the PAR flux
using observational datasets, mainly taken from the US SURFRAD network. The results are compared with
the observations. It is seen that the PAR values determined using this method are in good agreement with
Keywords:
observations. The relative mean biases (rMB) for both clear- and all-sky conditions evaluated for seven
Photosynthetically active radiation
UV–visible band flux
SURFRAD stations are 1.04% and 4.85%, respectively. The corresponding relative root mean square
Global horizontal irradiance errors (rRMS) are 6.22% and 17.97%, respectively.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction ratio of 4.57. This ratio has been used by many researchers (e.g. Hu
and Wang, 2014; Xia et al., 2008) for the conversion between the
Photosynthetically active radiation (PAR) is defined as the two quantities. In this paper, the unit of irradiance for PAR is used
radiative flux contained in the spectral regions between 400 and for the consistency with the other radiative quantities. The results
700 nm. This is because only the energy that the solar photons pos- can be converted into the photon flux using McCree’s ratio.
sess in this spectral region can be converted into chemical energy Solar energy in the PAR spectrum is a necessary input in appli-
in the plant photosynthesis processes. PAR can be expressed in cations dealing with plant physiology, biomass production, and
terms of solar irradiance (W m2) or photosynthetic photon flux natural illumination in greenhouses (Alados et al., 1996). Unlike
(lmol m2 s1). A monochromatical relationship between these global horizontal irradiance (GHI), which is a standard variable
two variables can be expressed by measured at the Baseline Surface Radiation Network (BSRN) and
most other solar radiation measurement stations, PAR is not rou-
Fkk
Pk ¼ ; ð1Þ tinely measured at these stations. Therefore, it has to be estimated
NA hc
using models with other measurements. The PAR estimation meth-
where P k represents the photon flux, F k denotes the solar irradiance, ods can be classified into three categories: physically-based mod-
k is the wavelength, NA is the Avocado constant, h is the Planck con- els, empirically-based models, and satellite-based models.
stant, c is the speed of light. It is seen that the monochromatic pho- Gueymard (1989a, 1989b) developed two physical models for esti-
ton flux depends on the wavelength explicitly and therefore the mating PAR. One model designed specifically for the PAR spectral
integrated broad-band photon flux cannot be derived directly from band (Gueymard, 1989a) uses the surface pressure, ozone amount,
the broad-band irradiance. Nevertheless, a statistic relationship Angstrom turbidity coefficients, single-scattering albedo, surface
between them exists. McCree (1972) analysed a ratio of Q p /RF , albedo and solar zenith angle as input parameters to calculate
where Q p and RF are the integrated photon flux and irradiance in PAR directly. The other model CPCR2 (Code for Physical Computa-
400–700 nm, using spectral measurements and obtained a constant tion of Radiation, 2 bands) (Gueymard, 1989b) is similar but based
on two spectral intervals, UV–visible region (0.29–0.7 lm) and
⇑ Corresponding author. near infrared (0.7–2.7 lm). The ratio of PAR to the UV–visible band
E-mail address: zhian.sun@bom.gov.au (Z. Sun).
irradiance is parametrized using the air mass. Both models apply to

http://dx.doi.org/10.1016/j.solener.2017.06.007
0038-092X/Ó 2017 Elsevier Ltd. All rights reserved.
612 Z. Sun et al. / Solar Energy 153 (2017) 611–622

clear-sky conditions. A high-performance solar radiation model band radiation measurements or models. Moon (1940) proposed a
REST2 was developed later by Gueymard (2008) which uses the constant PAR/GHI ratio 0.44. Hu et al. (2007) analyzed observations
same two-band schemes as in the previous CPCR2, but with a num- of PAR and GHI from 36 stations distributed across diverse ecosys-
ber of improvements. Qin et al. (2012) developed a physically- tems in China and found that the PAR/GHI ratios have significant
based PAR band model for estimating daily mean clear-sky PAR spatial, diurnal, and seasonal variations. Many field measurements
values. The effect of clouds was considered using sunshine dura- around the world have also demonstrated that the PAR/GHI ratio
tion data to work out the daily mean all-sky PAR values. The has diurnal, seasonal and special variations. It is also a function
advantage of the physically-based models is that these models of atmospheric variables such as water vapour, ozone, aerosol
were normally developed on the basis of the physical relationship loading and cloud conditions (Zhou et al., 1984; Jacovides et al.,
between PAR and dependent variables. Therefore they are not 2003; López et al., 2001; Wang et al., 2014; Bai, 2015). Ma et al.
restricted by local climatic or geographical conditions, and have (2007) summarized results for this ratio obtained from various
relatively good universal characteristics. locations around the world and found that its values vary between
A number of empirical models for estimation of PAR have been 0.35 and 0.58.
developed based on observational data of PAR and GHI. These mod- The above literature reviews indicate that both the empirical
els usually set Q p or Q p /GHI as a function of the broadband radia- and satellite-based methods have some restrictions on applica-
tion measurements, clearness index and solar zenith angle. The tions, whereas the physically-based models are relatively free of
clearness index is defined as a ratio of the observed GHI to the restrictions and therefore should constitute the preferred
extraterrestrial GHI. It can thus represent the effect of the atmo- approach. In this paper, we introduce a new physics-based param-
sphere on radiation at the surface, including absorbing gases, eterization to estimate PAR. In our earlier studies, we developed an
clouds, and aerosols. Alados et al. (1996) proposed such a model. accurate physical model to estimate the global and direct solar
Xia et al. (2008) and Hu et al. (2010) followed the same idea but radiation in the spectral regions of UV–visible (200–700 nm) and
developed different regression equations. Based on these near infrared (700–4000 nm). We intend to use our UV–visible
researches, Hu and Wang (2014) proposed three empirical models band model to estimate PAR in the 400–700 nm spectrum. The
for estimating PAR and tested their accuracy. Yu et al. (2015) pro- advantage of this approach is that the separate effects on PAR by
posed and tested nine empirical models for estimations of the major gasses, aerosols and clouds can be explicitly described. We
daily-mean PAR. All these models are performed well when evalu- conducted theoretical investigations to demonstrate that the Q p /
ated using local observations. However, these models have not GHI ratio is not a constant whereas Q p /(UV–visible-flux) ratio is
been properly tested over other climatic or geographical regions close to a constant. We then use this constant ratio and the UV–vis-
and thus it is unclear whether the same accuracy can be main- ible band irradiance determined by our band model to determine
tained in these regions. An exceptional case may be the study con- PAR and test the results against PAR observations obtained from
ducted by Zempila et al. (2016) who proposed a statistical model to seven stations in the United States.
link PAR with GHI using a multi-channel singular-spectrum analy- The paper is arranged in the following order: Section 2 intro-
sis so that PAR can be estimated from GHI and vice versa. Their duces the UV–visible band model. Section 3 describes the data
model has a unique feature in that the model established using used in this study. Section 4 presents investigation of the best ratio
data at one site can produce the PAR estimations with similar level of PAR to UV–visible irradiance. Section 5 shows the evaluation
of accuracy at all other sites. results. Section 6 conducts a sensitivity study, and Section 7 sum-
In general, the empirical models have advantages of simple marizes results and draws conclusions.
forms and easy to use. They can also produce relatively accurate
PAR estimations in a local area. The disadvantage of these models
2. UV–visible band model
is that they are usually restricted by local climate conditions due to
their dependency on local observational data. Another limitation of
As mentioned above, a physical band model for the estimation
these empirical models is that they cannot be used in the areas
of global and direct solar radiation in the spectral region 200–
where these measurements are not available.
700 nm has been developed in our earlier study. The details of this
PAR can also be estimated from satellite measurements. Rubio
band model have been described by (Sun and Liu, 2013; Sun et al.,
et al. (2005) used the measurements of the Meteosat Satellite to
2012; 2013). Only a brief description is given here. The band model
estimate the GHI. PAR was subsequently estimated using the GHI
was named SUNFLUX and developed based on the detailed radia-
with an empirical model proposed by Alados-Arboledas et al.
tion scheme SES2 (Sun, 2011). The SES2 scheme is a modified ver-
(2000). Zheng et al. (2008) proposed a method of using GOES visi-
sion of the Edwards and Slingo scheme originally developed at the
ble imagery to retrieve PAR. The advantages of Zheng’s method are
UK Met Office (Edwards and Slingo, 1996). The accuracy of the
that both surface reflectance and atmospheric parameters can be
SES2 scheme has been tested against line-by-line calculations,
simultaneously derived from observed radiances, without the
and the bias of the GHI at the surface is less than 1 W m2 relative
requirement for external knowledge of surface or atmospheric
to the corresponding line-by-line references. The bias against
parameters. Su et al. (2007) described a method that retrieves
observations is less than 4 W m2.
PAR and its direct and diffuse components from the Surface and
The transmission data for various atmospheric conditions
Atmospheric Radiation Budget (SARB) product of Cloud and the
including absorbing gasses, Rayleigh scattering, aerosols, and
Earth’s Radiant Energy System (CERES). Liu et al. (2008) developed
clouds were generated using the SES2 scheme and the results are
a method to estimate PAR in China using the MODIS data. There are
parameterized in simplified functions. For clear sky, the UV–visible
many other similar studies (e.g. Li et al., 2010; Pickett-Heaps et al.,
band flux can be calculated by
2014; Pankaew et al., 2014). The retrieval of PAR from satellite
measurements has an advantage of uniform spatial and temporal Q clr ¼ lS0 E0 T clr ; ð2Þ
coverage, which is valuable for remote areas where the data
required for determining PAR are not available. However, the where l is the cosine of the solar zenith angle; S0 is the solar con-
2
satellite-based methods cannot be used in an NWP or climate stant, E0 ¼ rr0 is the eccentricity correction factor of the Earth’s
model for PAR simulations in land-surface processes. orbit, r is the sun-earth distance and r 0 is the mean sun-earth dis-
Some studies focused on investigations into the ratio of PAR to tance, and Tclr is the total clear-sky atmospheric transmittance
the broadband solar radiation in order to estimate PAR from broad- which is further determined by
Z. Sun et al. / Solar Energy 153 (2017) 611–622 613

T g T a ð1  a" Þ measurements. Pyrgeometers are calibrated twice against stan-


T clr ¼ ; ð3Þ
1  Að1  ð1  Ra Þð1  a# ÞÞ dards traceable to the World Radiation Centre in Davos, Switzer-
land before going into the field. The first calibration is conducted
where Tg is the transmittance due to gaseous absorption; a# repre- at the National Renewable Energy Laboratory in Golden, Colorado
sents the Rayleigh albedo for the upward transfer radiation; a" and the second at Table Mountain near Boulder, Colorado where
denotes the Rayleigh albedo for the downward radiation; A is the the Surface radiation Research Branch of NOAA’s Air Resources lab-
surface albedo; Ra is the aerosol albedo and T a is the aerosol trans- oratory maintains three reference instruments of each type that
mittance. The methods for determining these variables are the SURFRAD instruments are checked against before and after
described in Sun and Liu (2013) and Sun et al. (2013). For cloudy deployment. The Quantum sensors for PAR measurements are sent
sky conditions, to the factory for calibration before they are used again. The instru-
Q c ¼ Q clr T c ; ð4Þ ments are calibrated annually and these calibrations ensure that
90% of the measurements are within 11 W m2, and 99% are within
where, T c is the cloudy sky transmittance and is determined by 15 W m2 of the standards. In addition to the radiation data, ancil-
lary measurements are taken to support research; these include
TwT i
Tc ¼ ; ð5Þ meteorological variables, solar spectral radiation measurements
ð1  Arw Þð1  Ari Þ
and a fraction of sky cover due to clouds. The aerosol optical depths
where Tw and Ti represent the transmittances of water and ice (AOD) at five spectral regions are retrieved from the solar spectral
clouds, respectively, and rw and ri denote the albedos of water measurements (Augustine et al., 2003). The cloud optical depths
and ice clouds. The all-sky fluxes are then determined by are retrieved using the shortwave broadband fluxes (Barnard
et al., 2008). The SUNFLUX scheme requires the cloud asymmetry
Q ¼ Q clr ð1  nÞ þ Q c n; ð6Þ factor for determination of cloud albedo. Since the asymmetry fac-
where n is cloud fraction. Again, the methods for determining these tor of 0.87 for water clouds (Charles N. Long, personal communica-
transmittances and albedos are provided in Sun et al. (2012). The tion, Atmospheric Radiation Measurement Program, Pacific
method has been evaluated using solar radiation observations and Northwest National Laboratory, Richland, Washington) and 0.8
the accuracy of the calculated solar flux under clear-sky conditions for ice clouds (Barnard et al., 2008) are used in the retrieval of
is better than 5%, and that for all-sky conditions is better than 7%. the cloud optical depth at the SURFRAD stations, these two values
Note that the Sunflux scheme was further modified recently with are used in the PAR calculations as well. It should be noted that the
a new aerosol parameterization and a new treatment of overlapping asymmetry factor varies with spectral wavelength and cloud parti-
absorptions between clouds and aerosols. The original aerosol cle sizes. Fu and Takano (1994) and Fu (2007) have shown that the
scheme only considers the effect of total aerosol optical depth. asymmetry factor for ice clouds can vary between 0.74 and 0.85. In
The new scheme considers five aerosol species separately and Section 6, we have conducted sensitivity studies to test the effects
therefore leads to a relatively better representation of regional dif- of uncertainties in the input variables. The results show that 20%
ferences in the aerosol radiative effects. The consideration of over- of the possible variability in asymmetry factors causes PAR simula-
lapping absorption between clouds and aerosols further improves tion error of less than 3%. This variability is larger than the range of
the accuracy of total GHI and UV–visible band GHI. A paper 0.74–0.85 and therefore the asymmetry factors used in this study
describes these modifications and presents the evaluation results are appropriate. The total precipitable water content can be deter-
has been submitted and will be shown elsewhere. mined using radiosonde data. However, the use of radiosonde data
will significantly reduce the number of samples, because the radio-
sonde data are only available twice a day while radiation data are
3. Data available at 1-min or 3-min intervals. To use the most available
data, an empirical relationship between precipitable water content
We use observational datasets from the surface radiation bud- and water vapour pressure at the surface at these SURFRAD sta-
get network (SURFRAD) in the United States to evaluate the tions has been developed so the former can be estimated from
method. The SURFRAD is NOAA’s major US radiation network and the latter. An empirical equation derived using all sounding and
consists of seven stations. Table 1 lists the information on these vapour pressure data at the SURFRAD stations is given below
stations. The earliest radiation measurements started in 1995
 
and more than 20 years observational data are available in four P
stations. Li-COR Quantum Sensors are used to measure PAR that
W ¼ 0:2968 þ 0:1539 ev ; ð7Þ
P0
has a response function that sharply turns on and cuts off at 400
and 700 nm (John Augustine, personal communication, National where W represents the precipitable water, p and p0 denote the sur-
Oceanic and Atmospheric Administration/Air Resources Labora- face pressure and standard pressure (1013.25 hPa), ev is the water
tory/Surface Radiation Research Branch, Boulder, Colorado,). The vapour pressure. The correlation coefficient between the W from
SURFRAD has adopted the standards for measurement set by the radiosonde and Eq. (7) is 0.904. The relative mean bias and rms
Baseline Surface Radiation Network for the instruments calibra- error are 0.018% and 21.886%, respectively. This equation is used
tions (Augustine et al., 2000). It is 15 W m2 for broadband solar to estimate precipitable water. Since the water vapour absorption

Table 1
Information of SURFRAD stations. The data from start date to July 2016 at each station are used in this paper.

Station code Station name Latitude Longitude Height (m) Surface type Start date
BND Bondville, Illinois 40.05 N 88.37 W 213 Grass 1 Jan 1995
SXF Sioux Falls, South Dakota 43.73 N 96.62 W 473 Grass 15 June 2003
DRA Desert Rock, Nevada 36.63 N 116.02 W 1007 Desert sand 16 Mar 1998
FPK Fort Peck, Montana 48.31 N 105.10 W 634 Grass 28 Jan 1995
GWN Goodwin Creek, Mississippi 34.25 N 89.87 W 98 Grass 1 Jan 1995
PSU Pen Penn. State Uni. Pennsylvania 40.72 N 77.93 W 376 Grass 29 June 1998
TBL Tab Table Mountain, Boulder, Colorado 40.13 N 105.24 W 1689 Grass 19 June 1995
614 Z. Sun et al. / Solar Energy 153 (2017) 611–622

concentrates in the near infrared band, the uncertainties in Eq. (7) all uncertainties due to absorption and scattering by air molecules,
should not introduce large errors in PAR calculations and this has aerosol and cloud particles in the spectra beyond 700 nm will
been confirmed in Section 6. affect the PAR results. By comparison, if the PAR/UV–visible-flux
ratio is used instead, the uncertainties can be significantly reduced.
4. Investigation of the best ratio of PAR to broadband radiation In order to confirm this argument, we perform theoretical cal-
culations using a radiative transfer scheme under various atmo-
As mentioned in the introduction, most previous studies have spheric conditions. For convenience, we define the PAR/UV–
been focused on the ratio of PAR to global solar radiation. This is visible ratio as RV and the PAR/GHI ratio as RG. We then use the
because global solar radiation is available from most radiation GENLN2 model to calculate these two ratios for various solar
measurement stations. Although this approach is simple and con- zenith angles and atmospheric profiles. Five McClatchey et al.
venient, it may not be a good option. In order to demonstrate this (1972) standard atmospheres (mid-latitude summer (mls), mid-
point and investigate the best PAR ratio, we conducted theoretical latitude winter (mlw), sub-arctic summer (sas), sub-arctic winter
calculations using a general line-by-line atmospheric transmit- (saw) and tropics (tro)) and 13 solar zenith angles ranging from
tance and radiance model (GENLN2) (Edwards, 1992) and broad- 0 to 89° are used in this investigation. Gaseous absorptions due
band radiative transfer scheme SES2 (Sun, 2011). We use these to H2O, CO2, O3, CH4, N2O and O2 are included in the calculations.
two models to calculate PAR flux, UV–visible band flux, and GHI The evaluations for the effects of clouds and aerosols on the RV
for a number of influencing variables. The PAR/GHI ratio and and RG ratios are conducted using the broadband radiative transfer
PAR/UV–visible-flux ratio are compared to identify the best ratio scheme SES2 (Sun, 2011). The spectral band limits of the SES2
for use in the PAR estimation. The variables tested include absorb- scheme have been amended such that the spectral interval of
ing gases, aerosol, water and ice clouds. GENLN2 is a high spectral bands 2–4 is the same as the PAR spectrum (400–700 nm). The cor-
resolution line-by-line scheme and provides a foundation and ref- responding fractions of the solar flux at the top of the atmosphere,
erence for developing broadband radiative transfer schemes the Rayleigh scattering coefficients and the gaseous absorption
(Edwards and Slingo, 1996; Sun and Rikus, 1999; Sun, 2011). It is coefficients in these spectral bands were also modified accordingly.
accurate for radiative flux calculations due to gaseous absorption The calculations are conducted for the various ranges of the atmo-
and therefore appropriate for use in evaluating the effects of gas- spheric variables listed in Table 2. The variables in Table 2 are
eous absorption on the PAR ratios. The HITRAN 2008 molecular major factors influencing PAR values. The ranges of these variables
spectroscopic database (Rothman et al., 2009) is used in the calcu- cover the possible values occurring in the real atmosphere. There-
lations. The calculations using GENLN2 assume a mid-latitude fore, the test results may have a global application. The mid-
summer atmosphere, a solar zenith angle of 30°and a surface latitude summer atmosphere is used as a background, and the solar
albedo of 0.2. zenith angle is fixed to 30°. An ice cloud layer is placed in a pres-
Fig. 1 shows the spectral solar irradiance at the top of the atmo- sure layer between 260 and 300 hPa, and a water cloud between
sphere (TOA) and at the earth surface determined using GENLN2. 915 and 940 hPa. The calculations for each variable are nested into
The irradiances at the surface were determined for water vapour, four loops so that the interactions between them are included.
carbon dioxide and ozone separately, thus reflecting the absorption To consider the effect of different aerosol types, the above cal-
spectra of each of these absorbing species. It is seen that within the culations are repeated five times, one for each of organic carbon,
PAR spectral region, only ozone is a strong absorber. Beyond black carbon, dust, sulphate, and sea salt aerosols. We use an annu-
700 nm, the absorption by water vapour and carbon dioxide dom- ally averaged aerosol mixing ratio profiles simulated by the Spec-
inates and the structures of these absorption spectra are complex. tral Radiation-Transport Model for Aerosol Species (SPRINTARS)
Furthermore, the cloud absorption spectra overlap with these (Takemura et al., 2009). SPRINTARS is a numerical model devel-
molecular spectra, making the treatment in a radiation model even oped by the Climate Change Science Section, Research Institute
more complicated. If the PAR/GHI ratio is used to estimate the PAR, for Applied Mechanics, Kyushu University for simulating the
effects of atmospheric aerosols on the climate system and air pol-
lution at global scale. The model data used in this study were from
SPRINTARS simulations for 2006–2008. The annual mean mixing
ratios of the SPRINTARS dust aerosol climatology at BSRN Saharan
Tamanrasset station; organic carbon, black carbon and sulphate
aerosols at East-Asia region; sea salt aerosols at Hainan Island
coast in China are selected as background aerosol profiles in the
SES2 model calculations.
To incorporate the ranges of varied data listed in Table 2 into
the SES2 model calculations, the following scaling equation is used

X i ¼ sxi ; ð8Þ

where xi represents the background atmospheric mixing ratio pro-


files of water vapour, ozone, aerosol and cloud; s is a scaling factor
defined as a ratio the corresponding varied variables listed in the

Table 2
Ranges of variables used in RV and RG ratio calculations using SES2 scheme. AOD:
aerosol optical depth; IOD: ice cloud optical depth; WOD: water cloud optical depth.
Fig. 1. Solar spectral irradiance at the top of the atmosphere (black curve) and at
the surface attenuated by water vapour (red curve), carbon dioxide (green) and Variable Range Samples
ozone (blue), respectively. Shaded area indicates the PAR spectral region. The
O3 (DU) 50–500 9
results are obtained using the GENLN2 model with a mid-latitude summer
AOD 0–4 20
atmosphere and a solar zenith angle of 30°. (For interpretation of the references
IOD 0–15 18
to colour in this figure legend, the reader is referred to the web version of this
WOD 0–250 20
article.)
Z. Sun et al. / Solar Energy 153 (2017) 611–622 615

Table 2 to the integrated values of xi . For ozone, s is the ratio of total therefore, confirm that the use of RV to estimate PAR can consider-
ozone absorber amounts. For aerosol and cloud it is the ratio of the ably reduce the uncertainties due to changes in atmospheric
optical depth. The scaled profile values X i are used in the calcula- conditions.
tions. When X i profiles are integrated, the results are equal to the The above results indicate that a constant RV value together
varied values given in Table 2. with the UV–visible band flux can be used to estimate PAR with
The results from the line-by-line and SES2 calculations are better accuracy than if using a constant RG. Therefore, we need
shown in Fig. 2, where the PAR ratio anomalies defined as a differ- to work out a best RV value for this approach. From our modelled
ence between the PAR ratio and its mean are plotted. Fig. 2a and b results, we calculate the mean values and standard deviations of
display the results determined using the GENLN2 line-by-line RV and RG and list the results in Table 3. It is seen that the aver-
scheme for the five standard atmospheric profiles with a solar aged values of RV and RG determined by the GENLN2 model
zenith angle of 30°. It is seen that the RV ratios are almost con- (shown as LBL in Table 3) are slightly smaller than those of the
stants, i.e. anomalies are close to zero for all five of the atmo- SES2 scheme. The line-by-line calculations only include gaseous
spheres, whereas the RG ratios vary with the atmosphere model. absorption, whereas the SES2 results include both absorption and
Fig. 2c and d further show a variation of the PAR ratios with the scattering. Therefore, in addition to the difference in the number
solar zenith angle determined using the line-by-line model. Again, of samples, the inclusion of the effects of scattering in the SES2
the RV anomalies are very close to zero except when the solar model calculations may be the main reason responsible for these
zenith angle approaches 90°, whereas the RG anomalies show differences. To confirm this explanation, we ran SES2 for gaseous
noticeable variation with zenith angle. absorption only (no scattering) and presented the results from this
The results from the SES2 calculations for various ozone con- calculation in the last two rows in Table 3. The results show that
centrations, dust aerosol, and cloud optical depths are shown in the mean values of RV and RG from this run are very close to those
Fig. 2e and f where the PAR ratios’ anomalies are plotted as a func- from GENLN2. These comparisons indicate that the scattering
tion of total visible optical depth. There are 65070 scatter data effect due to atmospheric particles can increase PAR ratios in the
points, and the different results are clustered for each total visible two spectral intervals. These results make sense because aerosols
optical depth. Note that we only show the results when the dust and clouds have strong scattering effects in the visible spectral
aerosols are included. The results with the rest four aerosol species region. It is also evident that the mean RV value including the scat-
are very similar to those shown in Fig. 2e and f and thus they are tering effect should be used in this study. This value is 0.8644.
not included. For RV anomalies, the values are spread around the Another issue which has been mentioned in the introduction is
zero line without a clear trend, whereas those of RG show a clear the conversion between PAR irradiance and PAR photon flux.
increasing trend with the increase in total visible optical depth. McCree (1972) has worked out a constant ratio of 4.57 based on
These results clearly demonstrate that the RV ratio is close to a observations. Here we employed the spectral model of flux,
constant, whereas the RG ratio has a considerable variability. We, SMARTS (Simple Model of the Radiative Transfer of Sunshine)

Fig. 2. PAR ratio RV and RG anomalies determined using the GENLN2 line-by-line scheme and SES2 broadband radiation scheme for various atmospheric conditions. Panels a
and b show line-by-line results for five standard atmospheres; Panels c and d display the line-by-line results as a function of solar zenith angle for the mid-latitude summer
atmosphere; Panels e and f present the SES2 broadband results as a function of total visible optical depth.
616 Z. Sun et al. / Solar Energy 153 (2017) 611–622

Table 3
Statistical results for PAR ratios RV and RG.

Mean Standard deviation Min Max Sample No


RV(LBL) 0.8576 9.91E04 0.8560 0.8590 10
RG(LBL) 0.4569 1.76E02 0.4266 0.4692 10
RV(SES2) 0.8644 7.95E03 0.8415 0.8868 65070
RG(SES2) 0.5840 3.12E02 0.4558 0.6300 65070
RV(SES2 no scattering) 0.8586 4.75E04 0.8480 0.8638 150
RG(SES2 no scattering) 0.4603 6.44E03 0.4588 0.4610 150

developed by Gueymard (Gueymard, 2001) to calculate the spec- 1750 nm, and 10 nm from 1750 to 4000 nm. This spectral resolu-
tral irradiance and photon flux. The SMARTS has a spectral resolu- tion is high enough for determining the spectral photon flux and
tion of 0.5 nm from 280 nm to 400 nm, 1 nm from 400 nm to spectral band total photon flux. The scheme includes the effects

Table 4
Statistical test results for PAR calculated using the ratio RV and UV–visible band flux. The PAR results determined by the SES2 model are used as the references. rMB: relative
mean bias; rRMS: relative-root-mean-squared difference; Maxdiff: maximum difference between model and observation; CC: correlation coefficient; r: standard deviation; a:
intercept; b: slope.

rMB (%) rRMS (%) Maxdiff (Wm2) CC r A b


RVmax 1.4 2.0 5.3 0.999 0.6 0.4 1.003
RVmin 3.3 6.7 20.4 0.999 2.4 0.4 0.957
RVmean 0.7 2.8 10.0 1.0 1.1 0.4 0.982

Fig. 3. Comparison of modelled and observed GHI at seven SURFRAD stations for entire available minutes observations. The blue lines represent linear regressions between
modelled and observed results and black lines are 1:1 diagonal lines. The left panels show the density of scattered data points and right panels present a histogram of
difference between the modelled and observed GHI. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this
article.)
Z. Sun et al. / Solar Energy 153 (2017) 611–622 617

Fig. 4. Same as Fig. 3 but with the extreme GHI values greater than the corresponding extraterrestrial values removed.

of gaseous and aerosol absorption and scattering. We use SMARTS the evaluation of radiation models. The intercept a and slope b of
to calculate both the PAR irradiance and photon flux for a number the least-squares linear fit is used to measure the correlation
of water vapour amounts and aerosol optical depths and then cal- between the two datasets. Table 4 lists the results. The maximum
culate the ratios of Q p /RF . For water vapour ranges between 0.1 and PAR difference between the estimated and referenced values is also
12 cm, this ratio ranges between 4.55 and 4.57. For AOD ranges included in Table 4. One can see that even when the highest and
between 0.01 and 3.0, the ratio varies from 4.57 to 4.65. These lowest RV ratios are used, the mean relative biases are only
modelled values are very close to 4.57 of McCree estimation and 1.346% and 3.26%, respectively; and relative RMS errors are
therefore our model result in irradiance unit from this study can 1.95% and 6.69%, respectively. In general, the statistical results
be safely converted to the photon flux with this constant factor. from the mean RV ratio are better. These test results provide high
Now we test the accuracy of the PAR values determined using confidence for the use of mean RV ratios to determine PAR.
this approach. The PAR values for all SES2 model calculations are
recovered using the three RV ratios (mean, min and max in Table 3)
and UV–visible band flux, and the results are compared with PAR 5. Comparison of modelled PAR with observations
flux determined using the SES2 real PAR band. The UV–visible band
flux is determined by the sum of flux from the SES2 band 1 to band We use the data described in Section 3 and the SUNFLUX
4 which covers the spectral interval of 200 to 700 nm. The SES2 scheme introduced in Section 2 to calculate the UV–visible band
real PAR flux is determined by the sum of flux from the band 2 flux and total GHI and then determine the PAR flux using two
to band 4 which covers the interval of 400 to 700 nm. The reason methods. The first is using the RV ratio and the UV–visible band
we use the maximum and minimum RV ratios in the test calcula- flux and the second using the RG ratio and total GHI. We have used
tions is to see if the variation of RV ratio causes an unacceptable the entire observational data available for download (from the
error in the PAR calculations. We calculate the relative-mean- start dates listed in Table 1 for each station until July 2016) from
bias (rMB), standard deviation (r), the relative-root-mean- the SURFRAD website. Since the number of samples is enormous,
squared difference (rRMS) and correlation coefficient (CC). These it may be more meaningful to plot the binned data point density
parameters are defined by (Beyer et al., 2009), as a benchmark than the scattered points themselves. Fig. 3 shows such a density
for radiation products and are used widely in the community for plot for a comparison between modelled and observed GHIs under
618 Z. Sun et al. / Solar Energy 153 (2017) 611–622

Fig. 5. Comparison of clear-sky PAR flux determined using the clear-sky UV–visible band flux and the RV ratio with observations at the SURFRAD stations. Blue lines represent
the linear regressions between modelled and observed results and black lines are 1:1 diagonal lines. (For interpretation of the references to colour in this figure legend, the
reader is referred to the web version of this article.)

all-sky conditions at these seven stations. The reason we present mented by albedo enhancement during winter (Gueymard, 2017).
the GHI results here is for supporting the PAR results later. In order Such a situation is extremely complex and current models are
to evaluate the model performance at each location, the compar- unable to handle it correctly. The discrepancies between the mod-
ison results are plotted for each of the SURFRAD stations sepa- elled results and observations at the up-right end of the diagram
rately. It is seen that the higher number density distributes along are mainly the results from these multiple interactions. In order
the 1:1 diagonal line which demonstrates that the modelled GHIs to confirm this argument, we removed all-sky GHIs that are greater
are in reasonable agreement with the observations. The blue lines than the extraterrestrial GHIs and plot the new results in Fig. 4. It is
indicate linear regressions between the modelled and observed seen that the overestimation at the up-right end of the Fig. 3 disp-
GHIs. These lines lie slightly above the diagonal lines, indicating pares. One can also note that the results are very similar at all
the slight overestimations of the modelled GHIs. The linear corre- seven stations. In particular no clear difference is found at the
lation coefficients between the modelled and observed values are DRA station (desert environment) or at TBL (a high mountain sta-
all greater than 0.98. The relative mean bias is 2.06% which implies tion). This suggests that the SUNFLUX scheme is of a universal
that the model has a small positive bias. The histogram in the character.
right-hand panels shows that the maximum frequency corre- Fig. 5 shows the same number density plot but for the modelled
sponds to a difference of 10 W m2. The model negative biases clear-sky PAR using products of the RV ratio and the UV–visible
are noticeably large, however, when the observed GHIs are large. band flux. It is seen that the agreement between modelled results
This requires further scrutiny. Note that the maximum observed and observations is very encouraging. Most data points are close to
GHI in Fig. 3 is greater than the solar constant. Indeed, such large the diagonal line. The linear fit lines are almost overlayed on the
values are common at all seven SURFRAD stations. After discussion diagonal lines. There are some scattered points spreading above
with the SURFRAD team, we trust that those large values are real, the diagonal line which are most likely the result of cloud contam-
and they are the result of cloud enhancement effects, possibly aug- ination. The statistical test parameters for the clear-sky results are
Z. Sun et al. / Solar Energy 153 (2017) 611–622 619

listed in Table 5, which shows that the relative mean bias at each terns of the PAR density are similar to those of the GHI, and these
station is less than 4%, the relative RMS error is less than 7.5%, discrepancies are clearly related to the multiple interactions of
and the correlation coefficient is greater than 0.99 for all seven sta- radiation between the surface and clouds as discussed previously
tions. The rMB and rRMS for all stations together are 1.044% and for the GHI comparison. The maximum PAR value at the top of
6.224%, respectively. the atmosphere (TOA) among the seven SURFRAD stations is
Fig. 6 shows the all-sky PAR determined using the RV ratio and 505 W m2. The modelled PARs are all smaller than this value
the UV–visible band flux. Apparently, the PAR values are underes- while the maximum observed PAR values at all seven stations
timated when the observed values are large. The distribution pat- are all greater than this value. Consequently, the modelled results

Table 5
Statistical test results for the modelled clear-sky PAR against observations of SURFRAD. rMB: relative mean bias; rRMS: relative-root-mean-squared difference; Maxdiff:
maximum difference between model and observation; CC: correlation coefficient; r: standard deviation; a: intercept; b: slope.

Station code rMB (%) rRMS (%) Maxdiff (Wm2) CC r (Wm2) a (Wm2) b Samples
BND 1.1 7.3 230.5 0.995 12.9 2.0 1.023 315514
DRA 3.9 5.6 146.6 0.998 8.8 5.4 0.985 1044592
FPK 2.2 5.6 147.6 0.998 8.2 4.3 1.006 424413
GWN 1.9 6.4 276.1 0.997 10.8 3.3 1.038 410643
PSU 0.1 6.9 231.5 0.996 11.3 2.8 1.019 98359
SXF 2.2 5.6 174.3 0.998 8.4 4.9 1.008 331786
TBL 2.1 6.1 146.2 0.997 10.4 3.2 0.995 484590
ALL 1.0 6.2 276.1 0.997 10.1 3.7 1.011 3109897

Fig. 6. Same as Fig. 5 but for all-sky results.


620 Z. Sun et al. / Solar Energy 153 (2017) 611–622

are only compared well with observations for PAR values ratio. On average over the seven stations, these values become
<400 W m2. For PAR values >400 W m2, the modelled results 4.9% and 18%, respectively.
are systematically underestimated.
The PAR results determined using the RG ratio and the total
GHIs are shown in Fig. 7. Unlike the results using the RV ratio, 6. Sensitivity study
the modelled PARs using the RG ratio are systematically overesti-
mated even if the GHIs are underestimated. The reason for the In this section, we perform a sensitivity test to investigate the
overestimation by this method is that the RG ratio is not a constant error responses of the modelled PAR to the uncertainty of the input
as discussed previously. The mean RG used in the calculations is variables. Six input variables are investigated. The PAR results at
too high compared with the results reported in the literature. This the seven SURFRAD stations determined by the normal input data
implies that an appropriate parameterization for the RG ratio are used as the benchmarks. The input variables of water vapour,
should be developed if the total GHI is used to estimate the PAR ozone, surface albedo, AOD, water and ice cloud optical depths
values. and asymmetry factors are then perturbed by 20%, respectively.
Table 6 lists the statistical test parameters for the all-sky PAR The results obtained using the perturbed input data are compared
results determined using both the RV and RG ratios, respectively. with the benchmarks. The statistical test results from these sensi-
One can see that the relative mean biases rMB for the RV rows tivity calculations are listed in Table 7. It is seen that the modelled
are all negative and those for the RG rows are positive, but the PAR is relatively sensitive to the surface albedo, ice cloud asymme-
absolute rMB values from the RV rows are smaller than those of try factor, ice and water cloud optical depths. 20% uncertainties
RG rows. For other parameters, the results from using the RV ratio in water cloud optical depth can cause 5.5% to 8.31% of errors
are clearly better than those from using the RG ratio. At the SUR- in the PAR calculations. The errors due to uncertainty in the surface
FRAD stations, rMB varies between 0.3% and 9.4%, and rRMS albedo are 4.17% to 6.11%. The uncertainties in ice cloud optical
varies between 17% and 19% when PAR is calculated from the RV depth and asymmetry factor lead to similar errors of 1.76% to

Fig. 7. Same as Fig. 6 but using RG ratio and total GHI.


Z. Sun et al. / Solar Energy 153 (2017) 611–622 621

Table 6
Statistical test parameters for the modelled PAR against observations of SURFRAD. The results determined using both RV and RG are presented for each of the stations. rMB:
relative mean bias; rRMS: relative-root-mean-squared difference; Maxdiff: maximum difference between model and observation; CC: correlation coefficient; r: standard
deviation; a: intercept; b: slope.

rMB (%) rRMS (%) Maxdiff (Wm2) CC r (Wm2) a (Wm2) b Samples


BND RV 2.9 18.2 222.4 0.98 16.5 4.9 1.01 800568
RG 21.8 45.6 330.8 0.97 36.4 14.2 1.37
DRA RV 9.4 17.0 232.8 0.98 17.2 4.7 0.97 250181
RG 21.4 40.4 305.5 0.98 41.1 21.4 1.39
FPK RV 8.0 17.9 187.3 0.98 16.1 4.8 0.97 693177
RG 13.3 36.2 305.9 0.97 33.1 19.6 1.33
GWN RV 0.3 17.3 188.4 0.98 15.3 3.3 1.04 712460
RG 25.8 48.6 319.7 0.98 36.7 10.9 1.38
PSU RV 1.8 17.6 176.1 0.98 15.2 4.3 1.03 481318
RG 24.3 49.6 425.9 0.98 37.0 14.6 1.41
SXF RV 6.2 18.9 211.2 0.97 17.4 5.6 1.0 578381
RG 20.1 43.1 324.1 0.97 37.0 16.7 1.37
TBL RV 5.5 18.9 234.6 0.98 19.8 9.9 0.98 652917
RG 19.1 41.2 425.9 0.97 17.3 2.1 0.96
ALL RV 4.9 18.0 234.6 0.98 16.8 5.4 0.99 4169002

Table 7 observations. The relative mean bias and rRMS for all-sky condi-
Sensitivity tests of modelled PAR to the uncertainties of input data. tions are 4.85% and 17.97%, respectively.
Variable Relative error % Error & SD Wm2 We have also shown that the PAR results determined using the
RG ratio and the total GHI are significantly overestimated. The
O3 0.81 to 0.80 0.76 ± 0.53 to 0.74 ± 0.52
H2O 0.01 to 0.01 0.01 ± 0.03 to 0.01 ± 0.03
results also show that the model underestimates both GHI and
AODa 0.49 to 0.84 0.46 ± 1.94 to 0.79 ± 2.21 PAR values modelled using the RV ratio under all-sky conditions,
gi 2.56 to 2.85 2.35 ± 4.91 to 3.1 ± 4.25 especially when the observations are large. In the extreme cases,
gw 0.80 to 1.52 0.77 ± 2.59 to 1.46 ± 4.56 the observed GHI and PAR values can be greater than those at
sw 8.31 to 6.50 7.63 ± 7.39 to 5.96 ± 5.97
the TOA due to the multiple reflections between the surface and
si 2.65 to 1.76 2.01 ± 4.39 to 1.72 ± 3.75
Rg 4.17 to 6.11 3.85 ± 5.75 to 5.60 ± 12.44 atmosphere. The current model is unable to simulate such a case
accurately. Although the modelled PAR irradiances based on the
a
AOD: aerosol optical depth; si : ice cloud optical depth; sw : water optical cloud RG ratio are larger than these extreme observations, they cannot
depth; H2O: precipitable water; Rg : surface albedo.
be regarded as an ability to simulate these cases. This is because
these large values are not derived from physical simulations;
2.85%. The scheme is not sensitive to the water vapour, ozone, and rather, they result from the use of an inappropriate constant ratio.
aerosol optical depth: 20% errors in these input data lead to less The underestimation of PAR by the current model is highly related
than 1% errors in the PAR calculations. to this deficiency. Although the interactions between the surface
and scattering agents in the atmosphere have been considered in
the SUNFLUX scheme, a further modification is necessary to
7. Conclusion improve the treatment of these processes. We are currently col-
lecting high-quality data in order to investigate possible key fac-
In this paper, we discussed the advantages of using the ratio of tors that may influence the extreme GHI values and hope to
PAR to the UV–visible spectral band flux to estimate the PAR flux. develop a correction scheme to improve the modelled results cor-
We have used detailed radiative transfer schemes to perform the responding to these processes.
investigations and demonstrated that the ratio of PAR to UV–visi- Our method needs further evaluation using observational data
ble band flux is close to a constant, whereas the ratio of PAR to total over other regions. The constant ratio of PAR to UV–visible flux
GHI varies with atmospheric conditions. Consequently, the use of needs further verification from observations. We have collected a
the RV ratio can considerably reduce the uncertainty in the deter- new dataset from the Chinese flux observation and research net-
mination of the PAR flux compared with the use of the RG ratio. work and try to collect even more data from different regions in
The best RV ratio identified in this work is 0.8644. the world to fulfill these evaluations.
We used the SUNFLUX band model and observational data at
seven SURFRAD stations in the US to calculate the PAR fluxes and Acknowledgments
compared the results with the PAR observations. The results show
that the clear-sky PAR fluxes determined using the RV ratio with Funding: This work was supported by the China Meteorological
the UV–visible band GHI are in good agreement with the observa- Administration for the projects on the observing and modelling
tions. The relative mean biases and rRMS for all seven SURFRAD study on spatial and temporal variation of radiation budget and
stations are less than 1.04% and 6.22%, respectively. The correla- PAR in regional scale and the evaluation of streamline agro-
tion coefficients between the modelled PAR and observations are climatic potential productivity and effects of climate change in
all greater than 0.99. By comparison, the modelled PAR accuracy Northeast China under grant CCSF201313.
under all-sky conditions is slightly worse than under the clear- ESA Aerosol optical depth was downloaded from the website at
sky conditions. These are expected results because the uncertain- http://www.icare.univ-lille1.fr/archive/?dir=CCI-Aerosols. The
ties of the cloud optical properties are usually large. Nevertheless, SURFRAD data were downloaded at http://www.esrl.noaa.gov/
the all-sky results are still in a reasonably good agreement with the gmd/grad/surfrad/.
622 Z. Sun et al. / Solar Energy 153 (2017) 611–622

Dr. John Augustine is thanked for useful discussions on the use Agric. For. Meteorol. 107, 279–291. http://dx.doi.org/10.1016/S0168-1923(01)
00217-9.
of SURFRAD data.
Ma, J.Y., Liu, J.M., Li, S.K., Liang, H., Jiang, C.Y., Wang, B.Z., 2007. Study on the features
of the photosynthetic active radiation (PAR) with experimentations and
measurements. J. Nat. Resources 22 (5), 673–682. http://dx.doi.org/10.11849/
References zrzyxb.2007.05.001.
McClatchey, R.A., Fenn, R.W., Selby, J.A., Volz, F.E., Garing, J.S., 1972. Optical
Alados, I., Foyo-Moreno, I., Alados-Arboledas, L., 1996. Photosynthetically active properties of the atmosphere (No. AFCRL-72-0497). AIR FORCE CAMBRIDGE
radiation: measurements and modelling. Agric. For. Meteorol. 78, 121–131. RESEARCH LABS HANSCOM AFB MA.
http://dx.doi.org/10.1016/0168-1923(95)02245-7. McCree, K.J., 1972. Test of current definitions of photosynthetically active radiation
Alados-Arboledas, L., Olmo, F.J., Alados, I., Perez, M., 2000. Parametric models to against leaf photosynthesis data. Agric. Meteorol. 10, 443–453. http://dx.doi.
estimate photosynthetically active radiation in Spain. Agric. For. Meteorol. 101, org/10.1016/0002-1571(72)90045-3.
187–201. http://dx.doi.org/10.1016/S0168-1923(99)00163-X. Moon, P., 1940. Proposed standard solar-radiation curves for engineering use. J.
Augustine, J.A., DeLuisi, J.J., Long, C.N., 2000. SURFRAD–A national surface radiation Franklin Inst. 230, 583–617. http://dx.doi.org/10.1016/S0016-0032(40)90364-
budget network for atmospheric research. Bull. Am. Meteor. Soc. 81, 2341– 7.
2357. http://dx.doi.org/10.1175/1520-0477(2000)081<2341:SANSRB>2.3.CO;2. Pankaew, P., Milton, E.J., Dash, J., 2014. Estimating hourly variation in
Augustine, J.A., Cornwall, C.R., Hodges, G.B., Long, C.N., Medina, C.I., DeLuisi, J.J., photosynthetically active radiation across the UK using MSG SEVIRI data. IOP
2003. An automated method of MFRSR calibration for aerosol optical depth Conference Series: Earth and Environmental Science, vol. 17, pp. 1–7. http://dx.
analysis with application to an Asian dust outbreak over the United States. J. doi.org/10.1088/1755-1315/17/1/012069.
Appl. Meteorol. 42, 266–278. http://dx.doi.org/10.1175/1520-0450(2003) Pickett-Heaps, C.A., Canadell, J.G., Briggs, P.R., Gobron, N., Haverd, V., Paget, M.J.,
042<0266:AAMOMC>2.0.CO;2. Pinty, B., Raupach, M.R., 2014. Evaluation of six satellite-derived Fraction of
Bai, J., 2015. Study on solar spectral radiation and calculating method of Absorbed Photosynthetic Active Radiation (FAPAR) products across the
photosynthetically active radiation at Baikal Lake. Adv. Geosci. 5, 33–42. Australian continent. Remote Sens. Environ. 140, 241–256. http://dx.doi.org/
http://dx.doi.org/10.12677/ag.2015.52005. 10.1016/j.rse.2013.08.037.
Barnard, J.C., Long, C.N., Kassianov, E.I., McFarlane, S.a., Comstock, J.M., Freer, M., Qin, J., Yang, K., Liang, S., Tang, W., 2012. Estimation of daily mean
McFarquhar, G.M., 2008. Development and evaluation of a simple algorithm to photosynthetically active radiation under all-sky conditions based on relative
find cloud optical depth with emphasis on thin ice clouds. Open Atmos. Sci. J. 2, sunshine data. J. Appl. Meteorol. Climatol. 51, 150–160. http://dx.doi.org/
46–55. http://dx.doi.org/10.2174/1874282300802010046. 10.1175/JAMC-D-10-05018.1.
Beyer, H.G., Martinez, J.P., Suri, M., Torres, J.L., Lorenz, E., Müller, S.C., Hoyer-klick, C., Rothman, L.S., Gordon, I.E., Barbe, A., Benner, D.C., Bernath, P.F., Birk, M., Boudon, V.,
Ineichen, P., 2009. Report on Benchmarking of Radiation Products. MESOR Rep. Brown, L.R., Campargue, A., Champion, J.P., Chance, K., Coudert, L.H., Dana, V.,
D 1.1.3. Devi, V.M., Fally, S., Flaud, J.M., Gamache, R.R., Goldman, A., Jacquemart, D.,
Edwards, D. P. D.P., 1992. GENLN2: A general line-by-line atmospheric Kleiner, I., Lacome, N., Lafferty, W.J., Mandin, J.Y., Massie, S.T., Mikhailenko, S.N.,
transmittance and radiance model. Version 3.0: Description and users guide. Miller, C.E., Moazzen-Ahmadi, N., Naumenko, O.V., Nikitin, A.V., Orphal, J.,
NCAR Technical Note NCAR/TN-367+STR, http://dx.doi.org/10.5065/D6W37T86. Perevalov, V.I., Perrin, A., Predoi-Cross, A., Rinsland, C.P., Rotger, M., Šimečková,
Edwards, J.M., Slingo, A., 1996. Studies with a flexible new radiation code. I: M., Smith, M.A.H., Sung, K., Tashkun, S.A., Tennyson, J., Toth, R.A., Vandaele, A.C.,
Choosing a configuration for a large-scale model. Q. J. R. Meteorol. Soc. 122, Vander Auwera, J., 2009. The HITRAN 2008 molecular spectroscopic database. J.
689–719. http://dx.doi.org/10.1002/qj.49712253107. Quant. Spectrosc. Radiat. Transf. 110, 533–572. http://dx.doi.org/10.1016/j.
Fu, Q., Takano, Y., 1994. On the limitation of using asymmetry factor for radiative jqsrt.2009.02.013.
transfer involving cirrus clouds. Atmos. Res. 34 (1–4), 299–308. http://dx.doi. Rubio, M.A., Lopez, G., Tovar, J., Pozo, D., Batlles, F.J., 2005. The use of satellite
org/10.1016/0169-8095(94)90098-1. measurements to estimate photosynthetically active radiation. Phys. Chem.
Fu, Q., 2007. A new parameterization of an asymmetry factor of cirrus clouds for Earth, Parts A/B/C 30, 159–164. http://dx.doi.org/10.1016/j.pce.2004.08.029.
climate models. J. Atmosph. Sci. 64, 4140–4150. http://dx.doi.org/10.1175/ Su, W., Charlock, T.P., Rose, F.G., Rutan, D., 2007. Photosynthetically active radiation
2007JAS2289.1. from Clouds and the Earth’s Radiant Energy System (CERES) products. J.
Gueymard, C., 1989a. An atmospheric transmittance model for the calculation of the Geophys. Res.: Biogeosci. 112 (G2). http://dx.doi.org/10.1029/2006JG000290.
clear sky beam, diffuse and global photosynthetically active radiation. Agric. Sun, Z., 2011. Improving transmission calculations for the Edwards-Slingo radiation
For. Meteorol. 45, 215–229. http://dx.doi.org/10.1016/0168-1923(89)90045-2. scheme using a correlated-k distribution method. Q. J. R. Meteorol. Soc. 137,
Gueymard, C., 1989b. A two-band model for the calculation of clear sky solar 2138–2148. http://dx.doi.org/10.1002/qj.880.
irradiance, illuminance, and photosynthetically active radiation at the earth’s Sun, Z., Liu, A., 2013. Fast scheme for estimation of instantaneous direct solar
surface. Sol. Energy 43, 253–265. http://dx.doi.org/10.1016/0038-092X(89) irradiance at the earth’s surface. Sol. Energy 98, 125–137. http://dx.doi.org/
90113-8. 10.1016/j.solener.2012.12.013.
Gueymard, C.A., 2001. Parameterized transmittance model for direct beam and Sun, Z., Liu, J., Zeng, X., Liang, H., 2012. Parameterization of instantaneous global
circumsolar spectral irradiance. Sol. Energy 71, 325–346. http://dx.doi.org/ horizontal irradiance: Cloudy-sky component. J. Geophys. Res. Atmos. 117, 1–
10.1016/S0038-092X(01)00054-8. 10. http://dx.doi.org/10.1029/2012JD017557.
Gueymard, C.A., 2008. REST2: High-performance solar radiation model for Sun, Z., Rikus, L., 1999. Parametrization of effective sizes of cirrus-cloud particles
cloudless-sky irradiance, illuminance, and photosynthetically active and its verification against observations. Q. J. R. Meteorol. Soc. 125, 3037–3055.
radiation–Validation with a benchmark dataset. Sol. Energy 82, 272–285. http://dx.doi.org/10.1002/qj.49712556012.
http://dx.doi.org/10.1016/j.solener.2007.04.008. Sun, Z., Zeng, X., Liu, J., Liang, H., Li, J., 2013. Parametrization of instantaneous global
Gueymard, C.A., 2017. Cloud and albedo enhancement impacts on solar irradiance horizontal irradiance: Clear-sky component. Q. J. R. Meteorol. Soc. 140, 267–
using high-frequency measurements from thermopile and photodiode 280. http://dx.doi.org/10.1002/qj.2126.
radiometers. Part 1: Impacts on global horizontal irradiance. Sol. Energy. Takemura, T., Egashira, M., Matsuzawa, K., Ichijo, H., O’Ishi, R., Abe-Ouchi, A., 2009.
http://dx.doi.org/10.1016/j.solener.2017.05.004. A simulation of the global distribution and radiative forcing of soil dust aerosols
Hu, B., Wang, Y., Liu, G., 2007. Spatiotemporal characteristics of photosynthetically at the Last Glacial Maximum. Atmos. Chem. Phys. 9, 3061–3073. http://dx.doi.
active radiation in China. J. Geophys. Res. 112 (D14). http://dx.doi.org/10.1029/ org/10.5194/acp-9-3061-2009.
2006JD007965. Wang, L., Gong, W., Ma, Y., Hu, B., Zhang, M., 2014. Photosynthetically active
Hu, B., Wang, Y., Liu, G., 2010. Long-term trends in photosynthetically active radiation and its relationship with global solar radiation in Central China. Int. J.
radiation in Beijing. Adv. Atmos. Sci. 27, 1380–1388. http://dx.doi.org/10.1007/ Biometeorol. 58 (6), 1265–1277. http://dx.doi.org/10.1007/s00484-013-0690-7.
s00376-010-9204-2. Xia, X., Li, Z., Wang, P., Cribb, M., Chen, H., Zhao, Y., 2008. Analysis of photosynthetic
Hu, B., Wang, Y., 2014. Comparison of multi-empirical estimation models of photon flux density and its parameterization in Northern China. Agric. For.
photosynthetically active radiation under all sky conditions in Northeast China. Meteorol. 148 (6), 1101–1108. http://dx.doi.org/10.1016/j.
Theor. Appl. Climatol. 116, 119–129. http://dx.doi.org/10.1007/s00704-013- agrformet.2008.02.008.
0941-x. Yu, X., Wu, Z., Jiang, W., Guo, X., 2015. Predicting daily photosynthetically active
Jacovides, C.P., Tymvios, F.S., Asimakopoulos, D.N., Theofilou, K.M., Pashiardes, S., radiation from global solar radiation in the Contiguous United States. Energy
2003. Global photosynthetically active radiation and its relationship with global Convers. Manage. 89, 71–82. http://dx.doi.org/10.1016/j.
solar radiation in the Eastern Mediterranean basin. Theoret. Appl. Climatol. 74 enconman.2014.09.038.
(3–4), 227–233. http://dx.doi.org/10.1007/s00704-002-0685-5. Zempila, M.-M., Taylor, M., Bais, A., Kazadzis, S., 2016. Modeling the relationship
Li, L., Xin, X., Su, G., Liu, Q., 2010. Photosynthetically active radiation retrieval based between photosynthetically active radiation and global horizontal irradiance
on HJ-1A/B satellite data Sci. China Earth Sci. 53, 81–91. http://dx.doi.org/ using singular spectrum analysis. J. Quant. Spectrosc. Radiat. Transf. http://dx.
10.1007/s11430-010-4142-5. doi.org/10.1016/j.jqsrt.2016.06.003.
Liu, R., Liang, S., He, H., Liu, J., Zheng, T., 2008. Mapping incident photosynthetically Zheng, T., Liang, S., Wang, K., 2008. Estimation of incident photosynthetically active
active radiation from MODIS data over China. Remote Sens. Environ. 112, 998– radiation from GOES visible imagery. J. Appl. Meteorol. Climatol. 47, 853–868.
1009. http://dx.doi.org/10.1016/j.rse.2007.07.021. http://dx.doi.org/10.1175/2007JAMC1475.1.
López, G., Rubio, M.A., Martı́nez, M., Batlles, F.J., 2001. Estimation of hourly global Zhou, Y.H., Xiang, Y.Q., Shan, F.Z., 1984. A climatological study on the
photosynthetically active radiation using artificial neural network models. photosynthetically active radiation. Acta Meteorol. Sin. 42, 387–397.

Das könnte Ihnen auch gefallen