Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
Five satellite products derivingbeam and global irradiance validation
on data from 23 ground stations
Pierre IneichenUniversity of Geneva
February 2011
Irradiance map produced by 3Tier
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 39 -
Five satellite products derivingbeam and global irradiance validation
on data from 23 ground stations
Pierre IneichenUniversity of Geneva
February 2011
Abstract
Models converting satellite images into the different radiation components becomeincreasingly performing and give often better estimation of the solar irradiance availabilitythan ground measurements if the station is not situated in the near vicinity of theapplication.
Five different satellite products deriving both global and beam irradiance are validatedagainst data from 23 ground sites. The main conclusions are:
• the global irradiance is retrieved with a negligible bias and an average standarddeviation around 16% for the best algorithm. For the beam irradiance, the bias isaround several percents, and the standard deviation around 35%,
• the main deviation comes from the knowledge of the aerosol optical depth,
• the high latitude sites give not poorer results than the other sites,
The interannual variability of the irradiance conditions, the lack of independent groundmeasurements such as aerosol data, the difficulty to assess the exact calibration of theground data, and the choice of a specific year to carry out the validation, conduct toresults that give good indications, but from which it is difficult to draw general conclu-sions.
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 1 -
1. Introduction
Models converting satellite images into the different radiation components becomeincreasingly performing and give often better estimation of the solar irradiance availabilitythan ground measurements if the station is not situated in the near vicinity of theapplication (Zelenka, 1999). If the global irradiance can be derived with a good accuracy,it is more difficult for the beam component and the dispersion of the models is higher.
The aim of the present study is to validate and compare five different products derivingthe global and beam irradiance components from meteorological satellite images. It is acomplement to a previous study conducted by the author (Ineichen 2009) on productsfrom Eumetsat Satellite Application Facilities (SAF). These algorithmes are derived byGeoModel in Bratislava (SolarGis), Helioclim Soda (heliosat 3v3), 3Tier company in theUnited States, University of Oldenburg (EnMetSol-Solis and EnMetSol-Dumortier) andthe IrSolAv company in Spain.
2. Ground data
The ground data used in the study are acquired at stations part of networks such asBaseline Solar Radiation Network (BSRN), Commission International de l’Eclairage (CIE),FluxNet network, Swiss Institute of Meteorology (ISM-Anetz) and World Radiation DataCenter (WRDC). Data from 23 ground sites situated mainly on the European continentare used. The work is done on data covering the year 2006.
Beside the global horizontal irradiance Gh, half of the sites acquire the normal beam
irradiance Bn. High precision instruments (WMO 2008) such as Kipp and Zonen CM10
Table I List of the stations, data and parameters availability.
Station Country Climate Dh Bn latitude ° longitude ° altitude m
Cabauw The Netehrlands temperate maritime x 51.970 4.930 2 BSRN - KNMICamborne United Kingdom temperate maritime x 50.220 -5.310 88 BSRN - Met OfficeCarpentras France mediterranean x 44.080 5.060 100 BSRN - Météo FranceDavos Dorf Switzerland semi-continental alpin x 46.810 9.840 1610 WRDC - Met OfficeEl Saler Spain semi arid, warm summer 39.346 -0.319 10 FluxNetGeneva Switzerland semi-continental x 46.199 6.131 420 CIE - UNIGEJungfraujoch Switzerland high alpine x 46.550 7.980 3571 CLIMAP - Météo SuisseLas Majadas Spain semi arid, warm summer 39.942 -5.773 260 FluxNetLerwick United Kingdom cold oceanic x 60.130 -1.180 82 BSRN - Met OfficeLocarno Switzerland warm temperate, humid 46.170 8.780 367 ANETZ - Météo SuisseNantes France oceanic x 47.150 -1.330 30 CIE - CSTBPayerne Switzerland moderate maritime/continental x 46.820 6.950 490 BSRN - Météo SuisseSede Boqer Israel dry steppe 30.867 34.767 457 BSRN - Met OfficeSion Switzerland dry alpine 46.220 7.330 489 ANETZ - Météo SuisseSonnblick Austria temperate alpine x 47.050 12.950 3105 WRDC - ZAMGTamanrasset Algeria hot, dry desert 22.780 5.520 1400 BSRN - Met OfficeThessaloniki Greece mediterranean temperate x 40.630 22.970 60 WRDC - Met OfficeToravere Estonia cold humid x 58.270 26.470 70 BSRN - EMHIVal Alinya Spain warm temperate, humid 42.152 1.449 1770 FluxNetVaulx-en-Velin France semi-continental x 45.780 4.930 170 CIE - ENTPEWien / Hohe Warte Austria continental x 48.250 16.350 203 WRDC - ZAMGYatir Forest Israel hot arid 31.347 35.052 650 FluxNetZürich Switzerland temperate atlantic 47.475 8.530 558 ANETZ - Météo Suisse
ANETZ MeteoSwiss network CUEPE Energy Group, UniGe
BSRN Baseline Surface Radiation Network EHMI The Estonian Meteorological and Hydrological Institute
CIE Commission Internationale pour l'Eclairage ENTPE Ecole Nationale des Mines de Paris
CLIMAP MeteoSwiss Climate Maping application KNMI The Netherlands Institute of Meteorology
CSTB Centre Scientifique et Technique du Bâtiment UNIGE University of Geneva
ZAMG Zentralanstalt für Meteorologie und Geophysik/Geodynamik
operated by
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 2 -
and Eppley PSP pyranometers, and Eppley NIP pyrheliometers, are used to acquire thedata. A stringent calibration, characterization and quality control was applied on all thedata by the person in charge of the measurements, the coherence of the data for allthe stations was verified by the author and is described in the following section. For twosites, the aerosol optical depth aod and the water vapor column w is independentlyacquired and is used as input to clear sky models in order to assess the instrumentscalibration factors.
The climate, latitude, longitude and altitude of the stations are given in Table I.
3. Data quality control
For all the stations, the first quality control consist of an assessment of the acquisitiontime stamp. To point out a possible time shift in the data, the symmetry in solar time ofthe irradiance for very clear days is visually checked. The horizontal global and if available,the normal beam irradiances are plotted versus the sinus of the solar elevation anglefor specific clear days. If the time stamp is correct, the afternoon curve should lay overthe morning curve as visualized on Figure 1a.
If this test is positive, a verification can be done with the help of the global clearnessindex K
t defined as:
)sin(hIG
Ko
ht ⋅
=
where Gh is the horizontal global irradiance, I
o is the solar constant, and h the solar
elevation angle. The clearness index is plotted for the morning and the afternoon datain a separate color. The upper limit, representative of clear sky conditions, should lay
Figure 1a The global horizon-tal and normal beamirradiances are representedversus the sinus of the solarelevation angle for a clear day.
Figure 1b The global clearness index Kt is representedseparately for the morning (green) and the afternoon(yellow) data, versus the solar elevation angle for oneyear in hourly values.Clear sky model data are represented in blue.
0
200
400
600
800
1000
0.0 0.2 0.4 0.6 0.8 1.0
Global and beam irradiances
sin (h)
GenevaJuly 18, 2006
0.0
0.2
0.4
0.6
0.8
1.0
0 20 40 60 80
Global clearness index Kt
Solar elevation angle h
Geneva 2006 (CH)
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 3 -
over for the morning and the afternoon data as represented on Figure 1b for one yearof data acquired at the site of Geneva for the year 2006. Hourly clear sky conditionvalues are plotted in light blue on the same graph. When these two conditions arefulfilled, the time stamp of the data bank is correct, and the solar geometry can beprecisely calculated. This test is very sensitive and a time shift of only a few minutes willconduct to an visible assymetry. A similar test can be done with the beam clearnessindex K
b defined as:
o
nb I
BK =
but this parameter is less sensitive to a possible time shift.
The coherence test between the two components can be verified with the help of theglobal and beam clearness indices (Ineichen 2010). The hourly beam clearness index isplotted versus the corresponding global index as illustrated on Figure 2 for the site ofCarpentras. On the same graph, the clear sky data evaluated with the Solis clear skymodel (Müller 2004, Ineichen 2008a) are represented for four different values of aerosoloptical depth (aod). The more usual corresponding Linke turbidity coefficient T
Lam2
retrieved from the beam irradiance:
)( 2 MATon
LamcdaeIB ⋅⋅−= δ
and evaluated at air mass AM = 2 is also given on the graph (Linke 1922, Ineichen2008b). δ
cda is the optical depth of a clean and dry atmosphere. An important deviation
from the clear sky lines can indicate calibration uncertainties, beam irradiancemissalignement or soiled sensors.
The absolute sensor calibration can be assessed with the help of a clear sky modelwhen the atmospheric aerosol optical depth and the water vapor column are known.These two parameters are normally retrieved from spectral measurements. When the
Figure 2 The beam clearness index is plottedagainst the global clearness index. On the samegraph, clear sky modelled values arerepresented for 4 different aerosol loads
0
0.2
0.4
0.6
0.8
0 0.2 0.4 0.6 0.8 1
ground measurements aod = 0.01 / Linke = 2.3 aod = 0.1 / Linke = 3.2 aod = 0.2 / Linke = 4.2 aod = 0.5 / Linke = 6.8
Water vapor = cm CarpentrasBeam clearness index Kb
Global clearness index Kt
2 aerosol type: rural
Figure 3 Atmospheric water vapor columnevaluated from the ambiant temperature andrelative humidity against the watre vaporretrieved from spectral measurements.
0
1
2
3
4
5
0 1 2 3 4 5Atmospheric water vapor column retrieved from aeronet spectral measurements
Atmospheric water vapor column retrieved from ground measurements with the Atwater model [cm]
Carpentras (F) 2003 - 2009
pointsslopembdsd R2
=====
1432104%
0.960.750.03
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 4 -
water vapor w is missing, it can be evaluated from the ground ambiant temperature(T
a) and relative humidity (HR) by the use of Atwater model (Atwater 1976) with a
good precision as illustrated on Figure 3 for the site of Carpentras and data acquiredfrom 2003 to 2009.
For some sites, the aod measurements retrieved from independent networks such asaeronet are acquired as soon as direct sun is available; these values are then averagedto give a daily value and used with the Solis clear sky model to evaluate hourly clear skyG
h and B
n values.
Day by day, the highest hourly value is then selected from the measurements andplotted against the day of the year on Figure 4. These points are representative of theclearest daily sky conditions. Based on the aod and water vapor content w of theatmopshere, the corresponding clear sky values are evaluated with the model. As thehighest values for each day is selected, the upper limit for these two series shoud laytogether if the two sets of measurements (irradiance and aod) are coherent. On thesame graph are also represented the daily clear sky indices defined as:
∑∑
=
dayhc
dayh
h G
GK and ∑
∑=
daync
dayn
hb B
BK
These values, if the data are coherent, should have an upper limit near of the unity.
4. The clearness index Kt and sky type classification
As it is the case for the majority of the national networks, the global irradiance is theonly available measured parameter concerning the solar radiation. Even if for half of thestations the beam component is available, the global irradiance and the corresponding
Figure 4 Daily highest value of the global irradiance reported versus the day of the yearfor the station of Carpentras, for the measurements and the corresponding clear sky evaluatedfrom the aod and the Solis model. The daily clear sky index is also represented.
0
200
400
600
800
1000
1200
0 31 61 92 122 153 183 214 244 275 305 336 3660
1
2
3
4
5
6
Clear sky (aeronet)
Ground measurements
Daily clear sky index
Day by day hourly global irradiance maximum value [W/m2]
Day of the year
Daily clear sky index Carpentras 2006
0
200
400
600
800
1000
1200
0 31 61 92 122 153 183 214 244 275 305 336 3660
1
2
3
4
5
6
Clear sky (aeronet)
Ground measurements
Daily clear sky index
Day of the year
Daily clear sky index Carpentras 2006Day by day hourly beam irradiance maximum value [W/m2]
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 5 -
clearness index Kt are key parameters in the field of irradiance modelization. The clearness
index Kt was introduced as a norm (Black 1954) to characterize the insulation condi-
tions at a given point in time when only the global component is known. Unfortunately,this parameter is not independent of the solar elevation angle as it is shown on the leftgraphe of Figure 5 where the clearness index K
t is plotted versus the solar elevation
angle for the site of Carpentras. It can be seen on this Figure that clear sky conditions,determined by the upper limit of the clearness index values, are not equally representedby K
t for the different solar elevation angles.
In order to use the clearness index as a reliable sky condition descriptor, Perez et al.(1990) modified this parameter to make it independent of the solar elevation angle.The formulation is the following:
( )( )( )1.0/4.99.0/4.1exp031.1'
++−⋅=
AMK
K tt
where AM is the optical air mass as defined by Kasten (1980). This modified clearnessindex is represented on Figure 5 (right graph) for the same points than above. It canclearly be seen on this Figure that even if some patterns are still present, the modifiedclearness index is relatively independent from the solar elevation angle. Therefore, it isnow possible to define three zones to characterize three sky types:
clear sky conditions 0.65 < K’t ≤ 1.00
intermediate sky conditions 0.30 < K’t ≤ 0.65
cloudy sky conditions 0.00 < K’t ≤ 0.30
These limits are arbitrary, but are coherent with other classifications, like for examplethe Cloud Free Index saturation (CFIsat) as defined by Dürr (2006):
CFIsat = 100 (CFI - 1) / z
where z and CFI are defined in Dürr (2004). CFIsat is independent from the irradiancemeasurements; it is a function of the downward surface longwave irradiance and thedry bulb temperature. The comparison between the CFIsat and K’
t is illustrated on
Figure 6 (right graph). The red dashed lines represent the limits used in the presentstudy and applied on the modified clearness index. In the CFIsat classification, clear skyconditions are defined by a CFIsat below 0%, and cloudy conditions above 50%. Thecorresponding limits are represented in blue dashed lines. It can be seen on Figure 6 thatthe majority of the points are situated in the intersections of the corresponding threezones.
Another assessement can be done with the cloud cover as illustrated on Figure 6 (leftgraph). Here also, the limits used in the present study are well correlated with the cloud
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 6 -
cover (0% cloud cover for the clear sky conditions and 100% cloud cover for theovercast sky conditions).
5. Satellite derived data
Five different products are validated in the present study. The methodology and theinput parameters are described in the following section. The product from University ofOldenburg (EnMetSol) is evaluated for two different aerosol climatologies.
5.1 SolarGis
The irradiance components are the results of a five steps process: a multi-spectralanalysis classifies the pixels, the lower boundary (LB) evaluation is done for each timeslot, a spatial variability is introduced for the upper boundary (UP) and the cloud indexdefinition, the Solis clears sky model is used as normalization, and a terrain disaggregationis finally applied.
Four MSG spectral channels are used in a classification scheme to distinguish cloudsfrom snow and no-snow cloud-free situations. Prior to the classification, calibrated pixelvalues were transformed to three indices: normalized difference snow index (Ruyter2007), cloud index (Derrien 2005), and temporal variability index. Exploiting the potential
0
0.2
0.4
0.6
0.8
1
0 10 20 30 40 50 60 70 80
Clearness index Kt = Gh / Io sin(h)
Solar elevation angle [°]
Measurements (BSRN Carpentras 2006)
0
0.2
0.4
0.6
0.8
1
0 10 20 30 40 50 60 70 80
Modified clearness index Kt'
Solar elevation angle [°]
Measurements (BSRN Carpentras 2006)
Figure 5 Global and modified clearness index versus the solar elevation angle.
Figure 6 Cloud cover and CFIsat coefficient versus the modified clearness index. The skycondition selection limits are represented in dashed red lines.
0
20
40
60
80
100
0.0 0.2 0.4 0.6 0.8 1.0
Cloud cover [%] Carpentras
Modified clearness index, K t'
overcast intermediate clear
-75%
-50%
-25%
0%
25%
50%
75%
100%
125%
0 0.2 0.4 0.6 0.8 1Modified clearness index Kt'
CFIsat Carpentras
overcast intermediate clear
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 7 -
of MSG spectral data for snow classification removed the need of additional ancillarysnow data and allowed using spectral cloud index information in cases of complexconditions such as clouds over high albedo snow areas.
In the original approach by Perez (2002), the identification of surface pseudo-albedo isbased on the use of a lower bound (LB), representing cloudless situations. This approachneglects diurnal variability of LB that is later corrected by statistical approach. Instead ofidentifying one value per day, LB is represented by smooth 2- dimensional surface (inday and time slot dimensions) that reflects diurnal and seasonal changes in LB andreduces probability of no cloudless situation.
Overcast conditions represented in the original Perez model by a fixed Upper Bound(UB) value were updated to account for spatial variability which is important especiallyin the higher latitudes. Calculation of cloud index was extended by incorporation ofsnow classification results.
The broadband simplified version of Solis model (Ineichen 2008a) was implemented. Asinput of this model, the climatology values from the NVAP water vapor database (Randl1996) and Atmospheric Optical Depth data by (Remund 2008) assimilated with Aeronetand Aerocom datasets are used.
Simplified Solis model was also implemented into the global to beam Dirindex algorithmsto calculate Direct Normal Irradiance component (Perez 1992, Ineichen 2008c). Diffuseirradiance for inclined surfaces is calculated by updated Perez model (1987).
Processing chain of the model includes post-processing terrain disaggregation algorithmbased on the approach by Ruiz-Arias (2010). The disaggregation is limited to shadowingeffect only, as it represents most significant local effect of terrain. The algorithm useslocal terrain horizon information with spatial resolution of 100 m. Direct and circumsolardiffuse components of global irradiance were corrected for terrain shadowing.
5.2 Heliosat-2 algorithm
The Helioclim 3 data bank is produced with the Heliosat-2 method that converts obser-vations made by geostationary meteorological satellites into estimates of the globalirradiation at ground level. This version integrates the knowledge gained by variousexploitations of the original Heliosat method and its varieties in a coherent and thoroughway.
It is based upon the same physical principles but the inputs to the method are calibratedradiances, instead of the digital counts output from the sensor. This change opens thepossibilities of using known models of the physical processes in atmospheric optics,thus removing the need for empirically defined parameters and of pyranometricmeasurements to tune them. The ESRA models (ESRA 2000, Rigollier 2000 and 2004)are used for modeling the clear-sky irradiation. The assessment of the ground albedo
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 8 -
and the cloud albedo is based upon explicit formulations of the path radiance and thetransmittance of the atmosphere. The turbidity is based on climatic monthly Linke Turbiditycoefficients data banks.
The Liu and Jordan (1960) model is used to split the global irradiance into the diffuseand beam components.
5.3 3Tier algorithm
Satellite-based time series of reflected sunlight are used to determine a cloud index timeseries for every land surface worldwide. A satellite based daily snow cover dataset isused to aid in distinguishing snow from clouds. In addition, the global horizontal clearskyradiation G
hc is modeled based on the surface elevation of each location, the local time,
and the measure of turbidity in the atmosphere. 3Tier opted to use a satellite-based,monthly time series of aerosol optical depth and water vapor derived from the ModerateResolution Imaging Spectroradiometer (MODIS). This dataset was combined with anotherturbidity dataset that includes both surface and satellite observations to provide a turbiditymeasure that spans the period of our satellite dataset and is complete for all landsurfaces. The cloud index n and the clear sky irradiance G
hc are then combined to model
the global horizontal irradiance Gh. This component of the process is calibrated for each
satellite based on a set of high-quality surface observations. Gh estimates are then
combined with other inputs to evaluate the other irradiance components Dh and B
n.
5.4 EnMetSol
The EnMetSol method is a technique for determining the global radiation at ground bythe use of data from a geostationary satellite (Beyer 1996, Hammer 2003). It is usedin combination with a clear sky model to evaluate the 3 irradiance parameters G
h, D
h
and Bn. The key parameter of the method is the cloud index n, which is estimated from
the satellite measurements and related to the transmissivity of the atmosphere via
Kc = 1 – n
where the transmissivity is expressed by the clear sky index Kc defined as the ratio of
global irradiance Gh and the corresponding clear sky irradiance G
hc:
hc
hc G
GK =
Two sets of data produced with the EnMetSol algorithm will be analyzed, correspondingto two different clear sky irradiance models:
• the model of Dumortier (Fontoynont 1998) with the Remund (2009)MeteonormHR high resolution data base for the turbidity input,
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 9 -
• and the original Solis clear sky model (Mueller 2004) with monthly averages ofAOD (Kinne 2005) and water vapour content (Kalnay 1996) as input parameters.
For the Dumortier clearsky, a diffuse fraction model (Lorenz 2007) is used to calculatethe all sky diffuse horizontal irradiance (via G
h-D
h). A recently developed beam fraction
model (Hammer 2009) is used to calculate the Bn for all sky conditions with the Solis
model.
5.5 IrSolAv
In the IrSolAv irradiance derivation scheme, the cloud index n is derived using themethodology developped by Dagestad and Olseth (Dagestad and Olseth, 2007) withsome modifications in the ground albedo determination. The ground albedo is computedfrom a forward and backward moving window of 14 days taking into account its evolutionduring the day, as function of the co-scattering angle.
The global horizontal irradiance Gh is then evaluated from the cloud index with the
model proposed by Zarzalejo (Zarzalejo et al., 2009); it uses as independent variablesthe cloud index, the 50-percentile of the cloud index for a given place, and the air massAM. The normal beam irradiance B
n is calculated from the global irradiance with the help
of Louche correlation (Louche et al., 1991).
In a second step, the clear sky conditions are indentified with the algorithm proposedby Polo (Polo et al., 2009a; Polo et al., 2009b); for these clear conditions, the irradiancesare evaluated with the ESRA clear sky model (Rigollier 2000), using the aerosol opticaldepth aod taken from Soda, MODIS or from a method proposed by Polo (Polo et al.,2009a) depending on their availability.
6. Comparison and evaluation procedure
In terms of validation, when evaluating satellite derived parameters with the same time
0
200
400
600
800
1000
0 200 400 600 800 1000
Measurements (ISM Zurich 2006)
EnMetSol Solis (Zurich 2006)Hourly horizontal global irradiance [Wh/m2h]
MBD =SD =R2 =
3%28%0.950
0
200
400
600
800
1000
0 200 400 600 800 1000
Measurements (BSRN Payerne 2006)
EnMetSol Dumortier (Payerne 2006)Hourly horizontal global irradiance [Wh/m2h]
MBD =SD =R2 =
4%18%0.978
Figure 8 Scatter plots for the global horizontal irradiance produced by EnMetSol for Zurichand Payerne.
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 10 -
step, the comparison can be done by means of scatter plots; these give a visual
evaluation of the capability of the model to reproduce the measurements. On these
graphs, the diagonal line is representative of an ideal model, and the points should lay
around this line. An illustration is given on Figure 8 for Zurich and Payerne, two sites that
showing different dispersions.
The statistical parameters like the mean bias difference (mbd), the root mean square
difference (rmsd), the standard deviation (sd) and the determination coefficient (R2)
represent a quantification of the model dispersion. These statistical parameters include
dispersions introduced by:
• the retrieval procedure,
• the comparison of point measurements (ground data) with aera
measurements,
• the comparison of the average of four instantaneous measurements
with 60 minutes integrated values.
In the field of solar radiation and natural light, the comparison is often done in term offrequency of occurrence: for the irradiance, it gives an indication of the repartition foreach level of radiation, and for the clearness index, that the level of radiation occures atthe right time during the day. The obtained graph is a line (or a bar chart) representativeof the relative frequency of occurrence of the considered parameter. This is illustratedon Figure 9 for the global irradiance and the corresponding clearness index K
t. On the
same graph, the frequency of occurrence of the ground measurements are representedas grey bars, and the different models in color lines.
A second order statistic, the Kolmogorov-Smirnov test (Espinar 2009), is also appliedto the data. It represents the capability of the model to reproduce the frequency ofoccurrence at each of the irradiance level. In order to avoid a peak at the zero level ofbeam irradiance, these values are excluded form the statistic. A visualisation is given onFigure 10 where the irradiance cumulated frequency of occurrence is represented against
Figure 9 Global irradiance and clearness index Kt relative frequency of occurrence for data
acquired in Vaulx-en-Velin (F). The grey bars are representative of the measurements.
0 200 400 600 800 1000
Measurements (CIE Vaulx-en-Velin 2006)
Geomodel (Vaulx-en-Velin 2006)
Helioclim3 (Vaulx-en-Velin 2006)
3Tier (Vaulx-en-Velin 2006)
IrSolAv (Vaulx-en-Velin 2006)
EnMetSol Solis (Vaulx-en-Velin 2006)
EnMetSol Dumortier (Vaulx-en-Velin 2006)
Global irradiance [W/m2]
Relative frequency of occurrence of the global irradiance Gh
0.0 0.2 0.4 0.6 0.8 1.0
Measurements (CIE Vaulx-en-Velin 2006)
Geomodel (Vaulx-en-Velin 2006)
Helioclim3 (Vaulx-en-Velin 2006)
3Tier (Vaulx-en-Velin 2006)
IrSolAv (Vaulx-en-Velin 2006)
EnMetSol Solis (Vaulx-en-Velin 2006)
EnMetSol Dumortier (Vaulx-en-Velin 2006)
Relative frequency of occurrence of the clearness index Kt
Clearness indexKt = Gh / Ioh
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 11 -
the irradiance for the same site than above. The quantitative value representative of theKolmogorov Smirnov test Integral (KSI) is defined as:
h
G
G hchc dGGFGFKSIh
h∫ ⋅−= max
min
)()( mod
where Fc(G
h) and F
c(G
hmod) are respectively the ground measurements and the
corresponding modelled cumulated frequencies of occurrence.
7. Global irradiance results
To ensure a correct and comparable validation of the different products, the followingmethod was used to merge the products and the ground measurements: for eachgenerated value, the nearest time stamped corresponding ground value is searched inthe data base; this means that the satellite image was taken within the ground integration
-100
0
100
200
300
400
Cab
auw
Cam
borne
Car
pentr
as
Dav
os
El S
aler
Gen
eva
Jun
gfrau
joch
Las
Maja
das
Ler
wick
Loc
arno
Nan
tes
Pay
erne
Sed
e Boq
er S
ion
Son
nblic
k
Tam
anras
set
The
ssalo
niki
Tora
vere
Val
d'Alin
ya
Vau
lx-en
-Veli
n
Wien
Yati
r For
est
Zuri
ch
All s
ites
Global irradiance average SolarGis Heliosat 3v3
3Tier EnMetSol (Solis) EnMetSol (Dumortier) IrSolAv
Global irradiance mean bias deviation [W/m2]
Figure 11 Average global irradiance and absolute mean bias difference
Figure 10 Relative frequency of occurrence for the global and the beam irradiance formeasurements at the site of Vaulx-en-Velin.
0%
20%
40%
60%
80%
100%
0 200 400 600 800 1000
Measurements (CIE Vaulx-en-Velin 2006) Geomodel (Vaulx-en-Velin 2006) Helioclim3 (Vaulx-en-Velin 2006) 3Tier (Vaulx-en-Velin 2006) IrSolAv (Vaulx-en-Velin 2006) EnMetSol Solis (Vaulx-en-Velin 2006) EnMetSol Dumortier (Vaulx-en-Velin 2006)
Global irradiance [W/m2]
Global irradiance cumulated frequency of occurrence
0%
20%
40%
60%
80%
100%
0 200 400 600 800 1000
Measurements (CIE Vaulx-en-Velin 2006) Geomodel (Vaulx-en-Velin 2006) Helioclim3 (Vaulx-en-Velin 2006) 3Tier (Vaulx-en-Velin 2006) IrSolAv (Vaulx-en-Velin 2006) EnMetSol Solis (Vaulx-en-Velin 2006) EnMetSol Dumortier (Vaulx-en-Velin 2006)
Normal beam irradiance [W/m2]
Normal beam irradiance cumulated frequency of occurrence
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 12 -
Table
II
Firs
t an
d s
econd o
rder
sta
tist
ics
in a
bso
lute
val
ues
for th
e glo
bal
horizo
nta
l irr
adia
nce
. The
site
s in
gre
y ar
e not ta
ken in
to a
ccount in
the
ove
rall
stat
istics
.
Table
III
Firs
t an
d s
econd o
rder
sta
tist
ics
in r
elat
ive
valu
es for
the
glo
bal
horizo
nta
l irr
adia
nce
. The
site
s in
gre
y ar
e not ta
ken in
to a
ccount in
the
ove
rall
stat
istics
.
Gh [
W/m
2 ]nb
R2
mbd
%sd
%KS
IR
2m
bd%
sd%
KSI
R2
mbd
%sd
%KS
IR
2m
bd%
sd%
KSI
R2
mbd
%sd
%KS
IR
2m
bd%
sd%
KSI
Cab
auw
258
4153
0.97
7-3
199
0.96
7-2
235
0.96
11
2416
0.98
2-1
176
0.98
1-1
178
Cam
born
e27
340
870.
976
-219
80.
966
223
90.
962
-124
150.
981
117
80.
981
017
8 C
arpe
ntra
s37
939
930.
985
113
40.
978
216
100.
980
-114
90.
987
212
90.
986
412
140.
901
-232
11 D
avos
326
4200
0.94
2-4
2719
0.91
59
3331
0.91
5-2
3321
0.94
6-1
527
510.
937
-928
330.
834
-845
34 E
l Sal
er41
336
000.
971
116
60.
971
116
90.
959
-119
110.
976
314
120.
976
214
80.
962
-318
19 G
enev
a31
736
220.
977
417
120.
974
019
40.
960
323
120.
983
515
110.
982
716
220.
943
227
11 J
ungf
rauj
och
399
3311
0.89
9-1
338
0.81
77
4349
0.81
15
4425
0.70
3-4
052
162
0.69
7-3
353
131
0.68
9-1
754
67 L
as M
ajad
as41
035
090.
976
515
210.
970
717
280.
966
618
230.
978
814
340.
979
614
240.
963
418
16 L
erw
ick
228
3306
0.95
8-1
2610
0.93
05
3418
0.92
04
3518
0.96
4-3
2413
0.96
4-4
2414
Loc
arno
314
4184
0.97
95
1715
0.95
58
2524
0.95
97
2423
0.98
16
1619
0.97
94
1714
0.95
51
2519
Nan
tes
292
4200
0.97
9-2
179
0.94
81
276
0.96
3-2
2214
0.98
41
157
0.98
3-2
1610
0.94
30
289
Pay
erne
304
4005
0.97
61
194
0.96
4-5
2314
0.95
82
2411
0.97
83
178
0.97
84
1813
0.94
30
2816
Sed
e Bo
qer
576
2890
0.98
60
911
0.98
2-6
1033
0.97
0-4
1327
0.98
5-3
918
0.98
6-4
925
0.97
4-5
1232
Sio
n32
542
800.
963
-323
210.
947
1327
430.
954
-325
260.
978
-118
160.
975
-219
170.
886
-439
33 S
onnb
lick
357
3933
0.85
8-1
3816
0.83
814
4050
0.80
63
4424
0.75
7-2
945
102
0.74
7-1
847
640.
765
148
17 T
aman
rass
et52
942
930.
985
011
50.
968
616
300.
970
315
150.
970
215
140.
972
214
15 T
hess
alon
iki
384
3675
0.98
21
143
0.97
8-4
1616
0.96
4-2
2016
0.97
5-2
1614
0.97
5-3
1716
0.95
9-1
2115
Tor
aver
e26
637
840.
971
-120
40.
940
130
160.
944
027
210.
973
-319
100.
972
-320
9 V
al d
'Alin
ya44
624
990.
968
217
110.
927
1226
520.
951
421
170.
965
-218
230.
966
218
170.
927
1026
45 V
aulx
-en-
Vel
in30
841
410.
980
516
140.
972
320
110.
965
622
180.
983
615
190.
983
515
140.
950
426
13 W
ien
288
4203
0.97
3-1
196
0.96
9-4
2113
0.96
3-1
2213
0.98
1-2
167
0.98
10
163
0.93
2-1
3011
Yat
ir Fo
rest
549
3221
0.98
4-1
107
0.98
0-3
1118
0.97
4-3
1217
0.98
8-4
923
0.98
7-4
921
0.97
2-3
1318
Zur
ich
274
4192
0.95
34
2712
0.94
01
3213
0.93
76
3119
0.95
03
289
0.95
06
2916
0.92
03
3510
All
site
s33
275
837
n/a
117
n/a
n/a
121
n/a
n/a
121
n/a
n/a
116
n/a
n/a
116
n/a
n/a
033
n/a
EnM
etSo
l (So
lis)
IrSol
Av
Sola
rGis
Hel
iosa
t 3v3
3Tie
rEn
Met
Sol (
Dum
ortie
r)
Gh [
W/m
2 ]nb
R2
mbd
sdKS
IR
2m
bdsd
KSI
R2
mbd
sdKS
IR
2m
bdsd
KSI
R2
mbd
sdKS
IR
2m
bdsd
KSI
Cab
auw
258
4153
0.97
7-6
499
0.96
7-4
585
0.96
13
6316
0.98
2-1
436
0.98
1-3
448
Cam
born
e27
340
870.
976
-652
80.
966
663
90.
962
-266
150.
981
147
80.
981
047
8 C
arpe
ntra
s37
939
930.
985
448
40.
978
759
100.
980
-455
90.
987
845
90.
986
1446
140.
901
-612
111
Dav
os32
642
000.
942
-13
8919
0.91
529
107
310.
915
-610
721
0.94
6-5
089
510.
937
-29
9233
0.83
4-2
814
734
El S
aler
413
3600
0.97
14
676
0.97
13
679
0.95
9-3
7811
0.97
612
6012
0.97
67
608
0.96
2-1
375
19 G
enev
a31
736
220.
977
1255
120.
974
159
40.
960
1072
120.
983
1647
110.
982
2250
220.
943
586
11 J
ungf
rauj
och
399
3311
0.89
9-4
132
80.
817
2917
049
0.81
118
177
250.
703
-161
208
162
0.69
7-1
3021
013
10.
689
-67
217
67 L
as M
ajad
as41
035
090.
976
2162
210.
970
2868
280.
966
2373
230.
978
3458
340.
979
2457
240.
963
1675
16 L
erw
ick
228
3306
0.95
8-2
5910
0.93
011
7818
0.92
010
8018
0.96
4-7
5413
0.96
4-1
054
14 L
ocar
no31
441
840.
979
1654
150.
955
2480
240.
959
2375
230.
981
1951
190.
979
1454
140.
955
279
19 N
ante
s29
242
000.
979
-749
90.
948
479
60.
963
-565
140.
984
444
70.
983
-745
100.
943
181
9 P
ayer
ne30
440
050.
976
357
40.
964
-14
6814
0.95
85
7411
0.97
88
538
0.97
814
5513
0.94
3-1
8516
Sed
e B
oqer
576
2890
0.98
61
5111
0.98
2-3
457
330.
970
-24
7327
0.98
5-1
553
180.
986
-23
5325
0.97
4-2
970
32 S
ion
325
4280
0.96
3-1
175
210.
947
4388
430.
954
-11
8326
0.97
8-3
5816
0.97
5-8
6117
0.88
6-1
412
633
Son
nblic
k35
739
330.
858
-313
416
0.83
850
143
500.
806
1115
924
0.75
7-1
0215
910
20.
747
-63
168
640.
765
217
217
Tam
anra
sset
529
4293
0.98
53
565
0.96
830
8530
0.97
015
8015
0.97
011
7814
0.97
210
7615
The
ssal
onik
i38
436
750.
982
353
30.
978
-16
6016
0.96
4-8
7616
0.97
5-9
6314
0.97
5-1
164
160.
959
-280
15 T
orav
ere
266
3784
0.97
1-4
534
0.94
03
8016
0.94
4-1
7221
0.97
3-8
5110
0.97
2-7
529
Val
d'A
linya
446
2499
0.96
89
7611
0.92
752
117
520.
951
1794
170.
965
-10
8023
0.96
67
7817
0.92
745
116
45 V
aulx
-en-
Vel
in30
841
410.
980
1551
140.
972
1160
110.
965
1866
180.
983
1946
190.
983
1447
140.
950
1179
13 W
ien
288
4203
0.97
3-3
556
0.96
9-1
361
130.
963
-264
130.
981
-547
70.
981
147
30.
932
-387
11 Y
atir
Fore
st54
932
210.
984
-453
70.
980
-16
5918
0.97
4-1
666
170.
988
-22
4723
0.98
7-2
047
210.
972
-18
6918
Zur
ich
274
4192
0.95
310
7412
0.94
03
8713
0.93
718
8519
0.95
08
769
0.95
016
7916
0.92
07
9510
All
site
s33
275
837
n/a
355
n/a
n/a
470
n/a
n/a
470
n/a
n/a
454
n/a
n/a
455
n/a
n/a
183
n/a
Sola
rGis
Hel
iosa
t 3v3
3Tie
rEn
Met
Sol (
Solis
)En
Met
Sol (
Dum
ortie
r)IrS
olA
v
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 13 -
0
20
40
60
Cab
auw
Cam
borne
Carp
entra
s
Dav
os
El S
aler
Gen
eva
Jun
gfrau
joch
Las
Maja
das
Lerw
ick
Loc
arno
Nan
tes
Pay
erne
Sed
e Boq
er S
ion
Son
nblic
k
Tam
anra
sset
The
ssalo
niki
Tora
vere
Val
d'Alin
ya
Vau
lx-en
-Veli
n
Wien
Yati
r Fore
st
Zuri
ch
All s
ites
SolarGis Heliosat 3v3 3Tier
EnMetSol (Solis) EnMetSol (Dumortier) IrSolAv
Global irradiance relative standard deviation [%]
Figure 12 Absolute standard deviation for the global irradiance. In red, the average Gh.
0
100
200
300
400
Cab
auw
Cam
born
e
Car
pentr
as
Dav
os
El S
aler
Gen
eva
Jun
gfraujo
ch
Las
Maja
das
Lerw
ick
Loc
arno
Nan
tes
Pay
erne
Sed
e Boq
er S
ion
Son
nblic
k
Tam
anras
set
The
ssalo
niki
Tora
vere
Val
d'Alin
ya
Vau
lx-en
-Veli
n
Wien
Yati
r Fore
st
Zuri
ch
All s
ites
Global irradiance average SolarGis Heliosat 3v3
3Tier EnMetSol (Solis) EnMetSol (Dumortier) IrSolAv
Global irradiance standard deviation [W/m2]
Figure 13 Relative standard deviation for the global irradiance.
0
20
40
60
Cab
auw
Cambo
rne
Carpe
ntras
Dav
os
El S
aler
Gen
eva
Jun
gfrau
joch
Las
Maja
das
Lerw
ick
Loc
arno
Nan
tes
Payer
ne
Sede B
oqer
Sion
Son
nblic
k
Taman
rass
et
The
ssalonik
i
Torav
ere
Val
d'Alin
ya
Vau
lx-en
-Veli
n
Wien
Yati
r Fore
st
Zuric
h
All s
ites
SolarGis Heliosat 3v3 3Tier
EnMetSol (Solis) EnMetSol (Dumortier) IrSolAv
Global irradiance second order statistic: Kolmogorof-Smirnov index KSI
Figure 14 Second order statistics KSI for the global irradiance
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 14 -
period. Then, only hours for which ground and both generated products are present aretaken into account in the validation procedure; so, mainly the ground data availabilityrestrict the number of points in the comparison.
The results of the validation are given on Table II and Table III respectively in absoluteand relative values. The corresponding graphs are given on Figure 11 to 14.
The first established fact is that the sites of Davos, Jungfraujoch and Sonnblick givemuch higher differences than the other stations. This is due to the snow cover during allor part of the year, which is not taken into account by all the models, and is difficult toevaluate precisely. This is clearly visible on frequency of occurrence plotted fir the clearnessindex as shown on Figure 15 for the site of Sonnblick. These site are not part of theoverall statistics. The site of Nantes is also represented on Figure 15 for comparisonpurpose.
The overall average mean bias deviation is very low for all the products, and except forheliosat 3, the sign of the mean bias deviation is in general the same for all the algorithms.It is interesting to note that for high latitude sites (Cabauw, Camborne, Lerwick andToravere), the bias is even lower than for the middle latitude sites. The site of Val d’Alyniain Spain shows slightly variable bias depending on the product, this can be explained by
0.0 0.2 0.4 0.6 0.8 1.0
Measurements (WRDC Sonnblick 2006)
Geomodel (Sonnblick 2006)
Helioclim3 (Sonnblick 2006)
3Tier (Sonnblick 2006)
IrSolAv (Sonnblick 2006)
EnMetSol Solis (Sonnblick 2006)
EnMetSol Dumortier (Sonnblick 2006)
Relative frequency of occurrence of the clearness index Kt
Clearness indexKt = Gh / Ioh
0.0 0.2 0.4 0.6 0.8 1.0
Measurements (CIE Nantes 2006)
Geomodel (Nantes 2006)
Helioclim3 (Nantes 2006)
3Tier (Nantes 2006)
IrSolAv (Nantes 2006)
EnMetSol Solis (Nantes 2006) EnMetSol Dumortier (Nantes 2006)
Relative frequency of occurrence of the clearness index Kt
Clearness indexKt = Gh / Ioh
Figure 15 Frequency of occurence of the clearness index for the site of Sonneblick (highaltitude and snow) and for Nantes.
0.0 0.2 0.4 0.6 0.8 1.0
Measurements (FluxNet Val d'Alinya 2006)
Geomodel (Val d'Alinya 2006)
Helioclim3 (Val d'Alinya 2006)
3Tier (Val d'Alinya 2006)
IrSolAv (Val d'Alinya 2006)
EnMetSol Solis (Val d'Alinya 2006)
EnMetSol Dumortier (Val d'Alinya 2006)
Relative frequency of occurrence of the clearness index Kt
Clearness indexKt = Gh / Ioh
Figure 16 Frequency of occurence of global irradiance and the clearness index for the site ofVal d’Alynia.
0 200 400 600 800 1000
Measurements (FluxNet Val d'Alinya 2006)
Geomodel (Val d'Alinya 2006)
Helioclim3 (Val d'Alinya 2006)
3Tier (Val d'Alinya 2006)
IrSolAv (Val d'Alinya 2006)
EnMetSol Solis (Val d'Alinya 2006)
EnMetSol Dumortier (Val d'Alinya 2006)
Global irradiance [W/m2]
Relative frequency of occurrence of the global irradiance Gh
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 15 -
the high altitude of the site (1770m) and the mountainous environment. The producedirradiance varie from one algorithm to the other; the corresponding graph is given onFigure 16 for the global irradiance and the clearness index. This can be due to the clearsky model used for the normalisation in conjunction with their input parameter(atmospheric water vapor content, aerosol load and linke turbidity).
In term of absolute standard deviation (due to the negligible bias, the root mean squaredifference and the standard deviation are equivalent), except for the high altitude sta-tions (Davos, Jungfraujoch, Sonnblick and Val d’Alynia), all the site show the same orderof magnitude, including the high latitude sites. Due to the high irradiance level for Car-pentras, Sede Boqer, Tamanrasset and Yatir forest, the sites show good relative stan-dard deviations.
The choice of a clear sky model is a key point in the satellite irradiance derivation, it willhave a direct influence on the output. The measured and modelled global clearness
0
0.2
0.4
0.6
0.8
1
1.2
0 10 20 30 40 50 60 70 80
Measurements (BSRN Carpentras 2006)
EnMetSol Dumortier (Carpentras 2006)
Solar elevation
Clearness index Kt = Gh / Io sin(h)
0
0.2
0.4
0.6
0.8
1
1.2
0 10 20 30 40 50 60 70 80
Measurements (BSRN Carpentras 2006)
EnMetSol Solis (Carpentras 2006)
Solar elevation
Clearness index Kt = Gh / Io sin(h)
0
0.2
0.4
0.6
0.8
1
1.2
0 10 20 30 40 50 60 70 80
Measurements (BSRN Carpentras 2006)
3Tier (Carpentras 2006)
Solar elevation
Clearness index Kt = Gh / Io sin(h)
0
0.2
0.4
0.6
0.8
1
1.2
0 10 20 30 40 50 60 70 80
Measurements (BSRN Carpentras 2006)
SolarGis (Carpentras 2006)
Solar elevation
Clearness index Kt = Gh / Io sin(h)
0
0.2
0.4
0.6
0.8
1
1.2
0 10 20 30 40 50 60 70 80
Measurements (BSRN Carpentras 2006)
Helioclim3 (Carpentras 2006)
Solar elevation
Clearness index Kt = Gh / Io sin(h)
Figure 17 The global clearness index Kt represented against the solar elevation angle for the
site of Carpentras. In yellow, the measurements and in blue the different products.
0
0.2
0.4
0.6
0.8
1
1.2
0 10 20 30 40 50 60 70 80
Measurements (BSRN Carpentras 2006)
IrSolAv (Carpentras 2006)
Solar elevation
Clearness index Kt = Gh / Io sin(h)
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 16 -
index is plotted against the solar elevation for all the products and the site of Carpentrason Figure 17. On these graphs, the clear sky conditions are represented by the upperboundary; it is interesting to point out that even if there is a great gap between themeasurements and the modelled values for Helioclim, the overall performance is of thesame order of magnitude than for the other products. It can also be noted here that forhalf of the products, the highest and the lowest modelled values of K
t are never reached.
To better understand the bias, a dependence analysis is done with the solar elevationangle and the water vapor column. To illustrate the results, two examples are given onFigure 18 where the bias is represented against the considered parameter, and a best fitis traced to underline the tendency. If the main pattern of all the models is to underestimatethe global irradiance for low water vapor column values and overestimate it for high w,the bias pattern is variable from one model to the other with the solar elevation angle.
The study was also done with the turbidity, but this parameter was not available exceptfor the sites of Carpentras and Sede Boqer, and on a daily basis. As for the water vaporcontent, the general tendency is a positive slope on the bias with the daily aerosoloptical depth. An illustration for EnMetSol is given on Figure 19.
The irradiance for clear sky conditions is directly derived from the clear sky model. For
Figure 18 Global irradiance mean bias difference against the solar elevation angle (left) andthe atmospheric water vapor content for two products and the site of Carpentras.
-200
-150
-100
-50
0
50
100
150
200
0 20 40 60 80
Helioclim3 (Carpentras 2006)
MBD =SD =
2%16%
Solar elevation angle [°]
Global irradiance bias [Wh/m2h]
-200
-150
-100
-50
0
50
100
150
200
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
EnMetSol Dumortier (Carpentras 2006)
MBD =SD =
4%12%
Atmospheric water vapor column [cm]
Global irradiance bias [Wh/m2h]
-200
-150
-100
-50
0
50
100
150
200
0.0 0.1 0.2 0.3 0.4 0.5
EnMetSol Solis (Sede Boqer 2006)
MBD =SD =
-3%9%
Daily aerosol optical depth
Global irradiance bias [Wh/m2h]
-200
-150
-100
-50
0
50
100
150
200
0.0 0.1 0.2 0.3 0.4 0.5
EnMetSol Solis (Carpentras 2006)
MBD =SD =
2%12%
Daily aerosol optical depth
Global irradiance bias [Wh/m2h]
Figure 19 Global irradiance mean bias difference against the daily aerosol load of theatmosphere for the two sites of Sede Boqer and Carpentras.
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 17 -
all sky condition, the clear sky model is combined with the cloud index. It is thereforeinteresting to differentiate the sky conditions following the rules defined in section 4.Table IV gives the overall results obtained for the three sky types. Without surprise, theclear sky conditions show the lowest standard deviation. All the algorithms have thesame tendency to underestimate for clear conditions, and overestimate the globalirradiance for intermediate and overcast conditions. Going more into details, the sametendency is also visible for all the sites. This is illustrated on Figure 20 for a cloudy(Lerwick) and a sunny (Carpentras) site. All the Figures and the complete Tables aregiven in the Annex.
The second order statistic given by the Kolmogorov-Smirnov index is a combination of
0%
20%
40%
60%
80%
100%
0 200 400 600 800 1000
Measurements (ISM Sion 2006) Geomodel (Sion 2006) Helioclim3 (Sion 2006) 3Tier (Sion 2006) IrSolAv (Sion 2006) EnMetSol Solis (Sion 2006) EnMetSol Dumortier (Sion 2006)
Global irradiance [W/m2]
Global irradiance cumulated frequency of occurrence
0%
20%
40%
60%
80%
100%
0 200 400 600 800 1000
Measurements (CIE Nantes 2006) Geomodel (Nantes 2006) Helioclim3 (Nantes 2006) 3Tier (Nantes 2006) IrSolAv (Nantes 2006) EnMetSol Solis (Nantes 2006) EnMetSol Dumortier (Nantes 2006)
Global irradiance [W/m2]
Global irradiance cumulated frequency of occurrence
Figure 21 Clearness index cumulated frequecy of occurrence for two sites with comparablestandard deviation: Sion and Nantes.
-200
-150
-100
-50
0
50
100
150
200
0 0.2 0.4 0.6 0.8 1
3Tier (Lerwick 2006)
MBD =SD =
4%35%
Modified clearness index Kt'
Global irradiance bias [Wh/m2h]
overcast intermediate clear
-200
-150
-100
-50
0
50
100
150
200
0 0.2 0.4 0.6 0.8 1
3Tier (Carpentras 2006)
MBD =SD =
-1%14%
Modified clearness index Kt'
Global irradiance bias [Wh/m2h]
overcast intermediate clear
Figure 20 Global irradiance mean bias difference against the modified clearness index for thetwo sites of Lerwick and Carpentras.
sky type Gh nb mbd sd mbd sd mbd sd mbd sd mbd sd mbd sd
-9 43 -23 49 -18 57 -16 41 -16 43 -25 66
-2% 9% -5% 10% -4% 11% -3% 8% -3% 9% -7% 17%
12 67 22 75 12 80 20 60 19 61 19 95
5% 27% 9% 30% 5% 32% 8% 24% 8% 25% 11% 53%
18 49 39 65 42 62 25 48 24 48 36 79
22% 60% 48% 80% 51% 75% 30% 58% 29% 59% 63% 138%
clear 498 35824
24096249intermediate
1591782overcast
EnMetSol (Dumortier)
IrSolAvSolarGis Heliosat 3v3 3TierEnMetSol
(Solis)
Table IV First order statistics in absolute and relative values for the global horizontal irradianceand for the three sky conditions. The absolute values are in [W/m2].
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 18 -
the bias, the standard deviation (dispersion) and the cumulated frequency of occurence.It is more sensitive and highlights smaller deviations than the first order indicators. Thiscan be seen for example for the sites of Sion or Sede Boqer, where the standarddeviation is similar to the other sites, but the clearness index frequency of occurenceshows descreapancies with the corresponding measurements. The cumulated frequencyof occurrence for Sion and Nantes are given on Figure 21. The deviation from themeasurements in Sion is visible for all the models at low and high irradiance levels; it canbe due to snow in the Rhône Valley.
In conclusion, except for the four high altitude sites (potentially with snow), the averagebias is around 1% (4 [W/m2]), positive for all the models. The best product derives theglobal irradiance with a standard deviation of 16% (55 [W/m2]).
Some models never reach the measured highest and lowest clearness index. The majorityof the products show a bias dependance with the solar elevation angle, the water vaporcolumn and when available, with the aerosol optical depth. In term of sky type, thegeneral pattern is to underestimate the global irradiance for clear conditions, and tooverestimate it for all the other conditions. Not all models take into account the snow,but even when included in the algorithm, the global irradiance for the concerned sites isnot satisfactory.
8. Beam irradiance results
The same methodology is used for the validation of the beam irradiance component,except that the slots with no direct irradiance are excluded from the validation in orderto avoid a bias in the overall statistic and a peak at zero irradiance in the frequency ofoccurrence.
Figure 22 Average beam irradiance and absolute mean bias difference
-100
0
100
200
300
400
Cab
auw
Cam
borne
Car
pentr
as
Dav
os
El S
aler
Gen
eva
Jun
gfrau
joch
Las
Maja
das
Ler
wick
Loc
arno
Nan
tes
Pay
erne
Sed
e Boq
er S
ion
Son
nblic
k
Tam
anras
set
The
ssalo
niki
Tora
vere
Val
d'Alin
ya
Vau
lx-en
-Veli
n
Wien
Yati
r For
est
Zuri
ch
All s
ites
Normal beam average SolarGis Heliosat 3v3
3Tier EnMetSol (Solis) EnMetSol (Dumortier) IrSolAv
Normal beam irradiance mean bias deviation [W/m2]
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 19 -
Table
VFi
rst
and s
econd o
rder
sta
tist
ics
in a
bso
lute
val
ues
for
the
norm
al b
eam
irr
adia
nce
. The
site
s in
gre
y ar
e not
take
n into
acc
ount
in t
he
ove
rall
stat
istics
.
Table
VI
Firs
t an
d s
econd o
rder
sta
tist
ics
in r
elat
ive
valu
es f
or
the
norm
al b
eam
irr
adia
nce
. The
site
s in
gre
y ar
e not
take
n into
acc
ount
in t
he
ove
rall
stat
istics
.
B n [W
/m2 ]
nbR
2m
bd%
sd%
KSI
R2
mbd
%sd
%KS
IR
2m
bd%
sd%
KSI
R2
mbd
%sd
%KS
IR
2m
bd%
sd%
KSI
R2
mbd
%sd
%KS
I
Cab
auw
220
4095
0.93
9-6
4228
0.87
715
5928
0.86
811
6214
0.93
1-4
4547
0.93
34
4439
Cam
born
e23
134
710.
942
342
280.
885
2958
260.
896
1056
200.
937
344
460.
940
1242
31 C
arpe
ntra
s44
839
930.
936
-426
450.
856
238
370.
905
-232
300.
924
-129
390.
923
729
200.
729
-153
28 D
avos
314
4200
0.86
411
6094
0.74
833
7910
80.
707
1186
113
0.73
5-3
682
276
0.75
7-7
7819
00.
663
492
153
El S
aler
Gen
eva
307
3622
0.91
82
4141
0.88
74
4739
0.86
910
5242
0.93
33
3735
0.93
715
3614
0.85
31
5434
Jun
gfra
ujoc
h18
533
110.
541
154
194
158
0.32
717
620
121
60.
461
167
206
174
0.31
9-4
818
957
80.
333
419
748
50.
399
6919
736
5 L
as M
ajad
as L
erw
ick
157
3119
0.89
06
7030
0.78
250
102
420.
764
3710
747
0.90
2-2
6741
0.90
68
6527
Loc
arno
Nan
tes
284
4199
0.94
3-1
134
370.
838
455
320.
890
-746
250.
945
-533
270.
944
-934
330.
851
-354
11 P
ayer
ne26
440
050.
889
653
320.
841
663
730.
832
1967
120.
887
954
480.
893
2254
300.
821
767
68 S
ede
Boq
er63
827
460.
887
-718
560.
763
-13
2610
10.
760
-826
630.
835
-11
2284
0.82
6-7
2362
0.70
0-7
2960
Sio
n S
onnb
lick
257
3933
0.70
866
107
320.
496
105
130
800.
487
6714
158
0.45
5-4
812
536
50.
450
013
524
60.
524
6514
085
Tam
anra
sset
545
4293
0.92
7-5
2561
0.75
617
4382
0.81
12
3862
0.81
95
3760
0.84
313
3571
The
ssal
onik
i T
orav
ere
286
3784
0.92
9-1
139
360.
807
464
470.
792
566
320.
890
-15
4984
0.89
7-5
4765
Val
d'A
linya
Vau
lx-e
n-V
elin
316
4060
0.93
9-2
3334
0.88
86
4330
0.89
310
4454
0.94
63
3123
0.95
04
3011
0.87
80
4621
Wie
n25
742
030.
900
-148
920.
880
653
129
0.85
810
5884
0.90
7-6
4714
30.
909
1146
116
0.83
41
6214
0 Y
atir
Fore
st Z
uric
h
All
site
s32
645
590
n/a
-435
n/a
n/a
850
n/a
n/a
549
n/a
n/a
-139
n/a
n/a
638
n/a
n/a
-154
n/a
EnM
etSo
l (So
lis)
IrSol
Av
EnM
etSo
l (D
umor
tier)
Sola
rGis
Hel
iosa
t 3v3
3Tie
r
B n [W
/m2 ]
nbR
2m
bdsd
KSI
R2
mbd
sdKS
IR
2m
bdsd
KSI
R2
mbd
sdKS
IR
2m
bdsd
KSI
R2
mbd
sdKS
I
Cab
auw
220
4095
0.93
9-1
393
280.
877
3213
028
0.86
824
136
140.
931
-810
047
0.93
310
9739
Cam
born
e23
134
710.
942
696
280.
885
6813
426
0.89
622
129
200.
937
710
146
0.94
027
9831
Car
pent
ras
448
3993
0.93
6-1
911
845
0.85
68
172
370.
905
-914
130
0.92
4-3
129
390.
923
3212
820
0.72
9-6
237
28 D
avos
314
4200
0.86
434
187
940.
748
104
247
108
0.70
735
271
113
0.73
5-1
1525
727
60.
757
-23
244
190
0.66
312
291
153
El S
aler
Gen
eva
307
3622
0.91
85
125
410.
887
1314
539
0.86
932
161
420.
933
1011
335
0.93
746
112
140.
853
316
534
Jun
gfra
ujoc
h18
533
110.
541
285
360
158
0.32
732
637
321
60.
461
310
381
174
0.31
9-8
835
057
80.
333
736
548
50.
399
127
365
365
Las
Maj
adas
Ler
wic
k15
731
190.
890
911
030
0.78
278
160
420.
764
5816
847
0.90
2-4
104
410.
906
1310
227
Loc
arno
Nan
tes
284
4199
0.94
3-3
195
370.
838
1215
732
0.89
0-1
913
225
0.94
5-1
494
270.
944
-27
9633
0.85
1-9
154
11 P
ayer
ne26
440
050.
889
1514
032
0.84
116
166
730.
832
4917
812
0.88
724
142
480.
893
5814
230
0.82
119
177
68 S
ede
Boq
er63
827
460.
887
-46
115
560.
763
-86
168
101
0.76
0-4
916
663
0.83
5-6
813
884
0.82
6-4
314
462
0.70
0-4
318
760
Sio
n S
onnb
lick
257
3933
0.70
817
127
632
0.49
627
033
480
0.48
717
336
358
0.45
5-1
2332
036
50.
450
034
624
60.
524
166
359
85 T
aman
rass
et54
542
930.
927
-27
138
610.
756
9423
582
0.81
114
208
620.
819
2720
460
0.84
368
191
71 T
hess
alon
iki
Tor
aver
e28
637
840.
929
-30
112
360.
807
1318
347
0.79
215
188
320.
890
-43
140
840.
897
-15
133
65 V
al d
'Alin
ya V
aulx
-en-
Vel
in31
640
600.
939
-610
334
0.88
819
137
300.
893
3014
054
0.94
611
9623
0.95
012
9411
0.87
80
144
21 W
ien
257
4203
0.90
0-3
124
920.
880
1513
512
90.
858
2615
084
0.90
7-1
512
114
30.
909
2711
911
60.
834
315
814
0 Y
atir
Fore
st Z
uric
h
All
site
s32
645
590
n/a
-11
115
n/a
n/a
2516
3n/
an/
a17
160
n/a
n/a
-512
8n/
an/
a19
125
n/a
n/a
-317
6n/
a
EnM
etSo
l (D
umor
tier)
Sola
rGis
Hel
iosa
t 3v3
3Tie
rEn
Met
Sol (
Solis
)IrS
olA
v
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 20 -
The first and second statistics are given on Table V and VI, and on Figures 22 to 24. Itcan be seen on Figure 22 that the absolute bias for the beam component is variablefrom one site to the other, and depends on the product: for example, it is positive inPayerne (15-60 W/m2) and highly negative in Sede Boqer (-70 to -110 W/m2). ForTamanrasset, it varies from -27 [W/m2] (Geomodel) to +68 [W/m2] (EnMetSol) and+94 [W/m2] (Heliosat), and for Lerwick, from -4 to +78 [W/m2].
The standard deviation varies from 20% to 100%; it is highly variable in absolute andrelative values. The site of Sede Boqer show the best values, but as it is a standarddeviation, it doesn’t take into account the its high mean bias. These deviation areillustrated on Figure 25, where on the left graph, the beam clearness index K
b is
represented versus the solar elevation angle. It can be seen that the modelled clear skyupper limit never reaches the corresponding measurements. On the right graph, the
0
20
40
60
80
Cab
auw
Cam
borne
Car
pentr
as
Dav
os
El S
aler
Gen
eva
Jun
gfrau
joch
Las
Maja
das
Lerw
ick
Loc
arno
Nan
tes
Pay
erne
Sed
e Boq
er S
ion
Son
nblic
k
Tam
anra
sset
The
ssalo
niki
Tor
aver
e
Val
d'Alin
ya
Vau
lx-en
-Veli
n
Wien
Yati
r For
est
Zur
ich
All s
ites
SolarGis Heliosat 3v3 3Tier
EnMetSol (Solis) EnMetSol (Dumortier) IrSolAv
Normal beam irradiance second order statistic: Kolmogorof-Smirnov index KSI
0
20
40
60
80
Cab
auw
Cam
borne
Car
pentr
as
Dav
os
El S
aler
Gen
eva
Jun
gfrau
joch
Las
Maja
das
Lerw
ick
Loc
arno
Nan
tes
Pay
erne
Sed
e Boq
er S
ion
Son
nblic
k
Tam
anra
sset
The
ssalo
niki
Tor
aver
e
Val
d'Alin
ya
Vau
lx-en
-Veli
n
Wien
Yati
r Fore
st
Zur
ich
All s
ites
SolarGis Heliosat 3v3 3Tier
EnMetSol (Solis) EnMetSol (Dumortier) IrSolAv
Normal beam irradiance relative standard deviation [%]
Figure 23 Relative standard deviation for the global irradiance.
Figure 24 Second order statistics KSI for the global irradiance
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 21 -
frequency of occurrence is represented: the clear sky peak is to low for all the products.Part of the discreapancy can be an calibration issue. In fact, when comparing the BSRNbeam ground measurements with the clear sky irradiance evaluated with solis fromaeronet aod measurements, a 4% difference is visible as illustrated on Figure 26 (left:uncorrected, right: corrected by 4%). Also, inspecting BSRN data from Sede Boqer, acommon 41 days sequence was found in data from 2005 and 2006. Comparison withsatellite evaluated data conducted to eliminate this sequence from the present valida-tion as it was much better reproduced with 2005 data. The 4% correction is allreadyincluded in the above results.
The relative standard deviation is comparable for all the sites, except for Lerwick, the
0.0 0.2 0.4 0.6 0.8 1.0
Measurements (BSRN Sede Boqer 2006)
Geomodel (Sede Boqer 2006) Helioclim3 (Sede Boqer 2006) 3Tier (Sede Boqer 2006)
IrSolAv (Sede Boqer 2006) EnMetSol Solis (Sede Boqer 2006) EnMetSol Dumortier (Sede Boqer 2006)
Relative frequency of occurrence of the beam clearness index Kb
Beam clearness indexKb = Bn / Io
0
0.2
0.4
0.6
0.8
1
1.2
0 10 20 30 40 50 60 70 80
Measurements (BSRN Sede Boqer 2006)
EnMetSol Dumortier (Sede Boqer 2006)
Solar elevation
Beam clearness index Kb = Bn / Io
Figure 25 Beam clearness index versus the solar elevation angle, and the corresponding rela-tive frequency of occurrence for the site of Sede Boqer.
0
200
400
600
800
1000
1200
0 31 61 92 122 153 183 214 244 275 305 336 3660
1
2
3
4
5
6
Clear sky (aeronet)
Ground measurements
Daily clear sky index
Day of the year
Daily clear sky index Sede Boqer 2006Day by day hourly beam irradiance maximum value [W/m2]
0
200
400
600
800
1000
1200
0 200 400 600 800 1000 1200
Normal beam irradiance evaluated with Solis and aeronet data [W/m2] Sede Boqer (2006)
Ground measurements normal beam irradiance [W/m2]
0
200
400
600
800
1000
1200
0 200 400 600 800 1000 1200
Normal beam irradiance evaluated with Solis and aeronet data [W/m2] Sede Boqer (2006)
Ground measurements normal beam irradiance [W/m2]
Ground measurements corrected by 4%
0
200
400
600
800
1000
1200
0 31 61 92 122 153 183 214 244 275 305 336 3660
1
2
3
4
5
6
Clear sky (aeronet)
Ground measurements
Daily clear sky index
Day of the year
Daily clear sky index Sede Boqer 2006Day by day hourly beam irradiance maximum value [W/m2]
Ground measurements corrected by 4%
Figure 26 Sede Boqer beam irradiance before and after correction. On top, beam irradianceand clear sky index; bottom, daily measurements and évaluated data.
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 22 -
highest site in term of latitude, and where the average beam irradiance is very low. Asfor the global component, the results are good for the high latitude sites, and, due tothe snow, not significative for the high altitude sites.
The same parameter dependance than for the global irradiance is conducted on thebeam component. The results are given on Figures 27 and 28, where the same sitesand models as for the global component are represented; they show very similar pat-tern (see Figures 18 and 19). This is a consequence of the method used to derive thebeam irradiance: all the algorithm derive in a first step the global component, and split itinto beam and diffuse with the help of a diffuse fraction model (Dirindex (Perez 1992)for SolarGis, Liu and Jordan (Liu 1960) for Helioclim, and in-house models for the otherproducts). This means that the conclusion drawn for the global are also valid for the
-300
-200
-100
0
100
200
300
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
EnMetSol Dumortier (Carpentras 2006)
MBD =SD =
4%12%
Atmospheric water vapor column [cm]
Beam irradiance bias [Wh/m2h]
-300
-200
-100
0
100
200
300
0 20 40 60 80
Helioclim3 (Carpentras 2006)
MBD =SD =
2%16%
Solar elevation angle [°]
Beam irradiance bias [Wh/m2h]
-300
-200
-100
0
100
200
300
0.0 0.1 0.2 0.3 0.4 0.5
EnMetSol Solis (Carpentras 2006)
MBD =SD =
2%12%
Daily aerosol optical depth
Beam irradiance bias [Wh/m2h]
-300
-200
-100
0
100
200
300
0.0 0.1 0.2 0.3 0.4 0.5
EnMetSol Solis (Sede Boqer 2006)
MBD =SD =
-3%9%
Daily aerosol optical depth
Beam irradiance bias [Wh/m2h]
Figure 27 Normal beam irradiance mean bias difference against the solar elevation angle(left) and the atmospheric water vapor content (right) for the site of Carpentras.
Figure 28 Normal beam irradiance mean bias difference against the daily aerosol load of theatmosphere for the two sites of Sede Boqer and Carpentras.
Table VII First order statistics in absolute and relative values for the normal beam irradiance andfor the three sky conditions. The absolute values are in [W/m2].
sky type Bn nb mbd sd mbd sd mbd sd mbd sd mbd sd mbd sd
-66 121 -61 163 -45 161 -66 138 -24 137 -83 179
-10% 19% -10% 26% -7% 26% -10% 22% -4% 22% -13% 29%
40 110 119 146 81 163 58 108 72 114 73 171
28% 78% 84% 103% 57% 115% 41% 76% 51% 81% 52% 120%
14 45 51 70 38 93 18 50 18 52 39 89
302% 983% 1117% 1522% 831% 2022% 394% 1084% 393% 1129% 845% 1927%overcast
intermediate 142 15173
IrSolAv
clear 627 20352
SolarGis Heliosat 3v3 3Tier
5 10065
EnMetSol (Solis)
EnMetSol (Dumortier)
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 23 -
beam irradiance, even if the effects are more important due to the chained models.
The validation results for the three categories of the sky type are given on Table VIIwhere the results for overcast conditions are included, even if they are not significative.Here again, comparable conclusions can be drawn, the bias is negative for clear condi-tions and positive for intermediate skies.
The above results are also visible on the second order statistic KSI, as illustrated onFigure 24 and 29, and given in Table V and VI. The fact that the beam is measured orretrieved from the global and the diffuse is not the main issue on the difference in thecumulated frequency of occurrence curves, it is more a site effect.
In conclusion, the beam irradiance is retrieved from the satellite images with an averagestandard deviation of 35% to 54% depending on the model. In term or irradiance value,it rages from 90 to 200 [W/m2] according to the model and the site. The mean biasdifference is slightly higher than for the global component, negative for clear conditionsand positive for intermediate skies.
As the beam component is derived from the global, the dependances with the differentparameters are similar but sharper than for the global irradiance.
9. Conclusions
The first conclusion is that the quality control is a key point in any model validation. Evenif the data are highly qualified by the organisation in charge of the acquisition, uncertaintiescan remain in the data and influence the validation. The best case is when independentdata such as aerosol optical depth are available.
The main conclusions of the present study are represented by the first order statistics:
• for latitude from 20° to 60°, altitude from sea level to 1800 m and various climate,the global irradiance is retrieved with a negligible bias and an average standard
0%
20%
40%
60%
80%
100%
0 200 400 600 800 1000
Measurements (CIE Nantes 2006) Geomodel (Nantes 2006) Helioclim3 (Nantes 2006) 3Tier (Nantes 2006) IrSolAv (Nantes 2006) EnMetSol Solis (Nantes 2006) EnMetSol Dumortier (Nantes 2006)
Normal beam irradiance [W/m2]
Normal beam irradiance cumulated frequency of occurrence
0%
20%
40%
60%
80%
100%
0 200 400 600 800 1000
Measurements (BSRN Payerne 2006) Geomodel (Payerne 2006) Helioclim3 (Payerne 2006) 3Tier (Payerne 2006) IrSolAv (Payerne 2006) EnMetSol Solis (Payerne 2006) EnMetSol Dumortier (Payerne 2006)
Normal beam irradiance [W/m2]
Normal beam irradiance cumulated frequency of occurrence
Figure 29 Beam clearness index cumulated frequecy of occurrence for Nantes where thebeam is retrieved from the diffuse, and Payerne where the beam componant is acquired.
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 24 -
deviation around 16% for the best algorithm. For the beam irradiance, the bias isaround several percents, and the standard deviation around 35%,
• as expected, the main dependance comes from the knowledge of the aerosoloptical depth. A lower dependance with the atmospheric water vapor column andthe solar elevation angle is pointed out,
• even if the snow cover is taken into account in the algorithm, the sites situated inhigh altitude such as Junfraujoch and Sonnblick give bad results and are not takeinto account in the overall statistic.
• the high latitude sites such as Cabauw, Camborne or Toravere give not poorerresults than the other sites, only Lerwick, the highest site in latitude (60°N) presentsmore difficulties. It has also very low levels of irradiance,
• for the majority of the sites, SolarGis and EnMetSol give the best statistics forboth of the components.
Neverteless, the interannual variability of the irradiance conditions (it has be shown in aprevious study that even for the same site and algorithm, the results can vary from oneyear to the other, Ineichen 2010a), the lack of independent ground measurements suchas aerosol data, the difficulty to assess the exact calibration of the ground data, and thechoice of a specific year to carry out the validation, conduct to results that give goodindications, but from which it is difficult to draw general conclusions.
10. Acknowledgements
This comparative study could be performed thanks to the data provided by the FluxNetcommunity (http://www.fluxdata.org), the Baseline Surface Radiation Network, theAeronet network, the CSTB of Nantes and ENTPE in Vaulx-en-Velin, the World RadiationData Center, and the Swiss Institute of Meteorology.
The derived products where kindly provided by Geomodel in Slovakia, 3Tier in Seattle,the Helioclim data bank in France, University of Oldenburg in Germany, and IrSolAvcompany in Spain.
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 25 -
11. Nomenclature and abbreviations
Io
solar constant in W/m2
Bn
direct normal incidence irradiance in W/m2
Gh
surface solar global irradiance on a horizontal plane in W/m2
Ghc
clear sky surface solar irradiance in W/m2
Dh
surface solar diffuse irradiance on a horizontal plane in W/m2
Kt
global or surface solar irradiance clearness indexK
bbeam clearness index
Kc
clear sky indexK
hdaily global clear sky index
Khb
daily beam clear sky indexn cloud index
Fc(G
h) cumulated frequecy of occurrence of the global irradiance
h solar elevation or altitude angle in degrees (°)AM optical air mass
RTM radiative transfer modelT
LLinke turbidity
aod aerosol optical depth, or atmospheric aerosol load
Ta
surface (ambient) temperatureHR relative humidityw atmospheric total water vapour column
mbd mean bias differencermsd root mean square differencesd standard deviationKSI Kolmogorov-Smirnov index, second order statistic
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 26 -
12. References
Aeronet (AErosol RObotic NETwork) program, http://aeronet.gsfc.nasa.gov/
Atwater M.A., Ball J.T. (1976) Comparison of radiation computations using observed
and estimated precipitable water. Appl. Meteorol. 15, 1319-1320
Beyer HG, Costanzo C, Heinemann D (1996): Modications of the Heliosat procedurefor irradiance estimates from satellite data. Solar Energy 56, 207–212.
Black J.N., Bonython C.W., Prescott J.A. (1954) Solar radiation and the duration ofsunshine. Quart. J. Roy. Met. Soc. 80, pp231-235
Dagestad, K. F. and Olseth, J. A., (2007). A modified algorithm for calculating the cloudindex. Solar Energy. 81, 280-289.
Derrien M., H. Gleau (2005) MSG/SEVIRI cloud mask and type from SAFNWC,International Journal of Remote Sensing, 26, 4707-4732.DUMORTIER D (1998): The Satellite model of turbidity variations in Europe. Technical
Report
Dürr B., Philipona R, (2004) Automatic cloud amount detection by surface longwavedownward radiation measurements. J. Geoph. Res., Vol 109
Dürr B. (2006) , personal communication
Espinar B, Ramirez L, Drews D, Beyer H.G, Zarzalejo L.F, Polo J, Martin L. (2009)
Analysis of different comparison parameters applied to solar radiation data from satel-
lite and German radiometric stations, Solar Energy 83 (2009) 118–125
ESRA (2000) The european solar radiation atlas. Coordinators : K. SCHARMER, J.
GREIF, ISBN : 2-911762-21-5
Fontoynont M, Dumortier D et al. (1998) Satellight: A WWW server which provides high
quality daylight and solar radiation data for Western and Central Europe. 9th Conference
on Satellite Meteorology and Oceanography, 434–437.Frouin R. A., Chertock B., (1992).
A technique for global monitoring of net solar irradiance at the ocean surface. Part 1 :
Model, Journal of Applied meteorology, 31,pp. 1056-1066.
Hammer A, Heinemann D, Hoyer C, Kuhlemann R, Lorenz E, Müller R, Beyer HG (2003):Solar energy assessment using remote sensing technologies. Remote Sensing of
Environment, 86, 423–432.
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 27 -
Hammer A, Lorenz E, Kemper A, Heinemann D, Beyer HG, Schumann K, Schwandt M(2009): 'Direct normal irradiance for CSP based on satellite images of Meteosat Second
Generation', SolarPACES 2009, Berlin
Helioclim (2009) Sun radiation atlas based on processing of Meteosat data.
<http://www.helioclim.net/>
Ineichen P, Perez R. (2002) A new airmass independent formulation for the Linke
turbidity coefficient. Solar Energy 73, No. 3, 151–157
Ineichen P. (2008c). Comparison and validation of three global-to-beam irradiance models
against ground measurements. Sol. Energy , doi:10.1016/j.solener.2007.12.006
Ineichen P. (2008a) A broadband simplified version of the Solis clear sky model, Solar
Energy 82, pp 758–762
Ineichen P. (2008b) Conversion function between the Linke turbidity and the atmospheric
water vapor and aerosol content, Solar Energy 82, pp. 1095–1097
Ineichen P. (2008c). Comparison and validation of three global-to-beam irradiance models
against ground measurements. Sol. Energy , doi:10.1016/j.solener.2007.12.006
Ineichen P., Barroso C., Geiger B., Hollmann R., Marsouin A., Mueller R. (2009) SatelliteApplication Facilities irradiance products: hourly time step comparison and validationover Europe. International Journal of Remote Sensing, Vol. 30, No. 21, pp. 5549–5571
Ineichen P., Perez R. (2010). Aerosol quantification based on global irradiance. Solar
paces proceedings, 21-24 sept. Perpignan, France. Available from:
http://www.cuepe.ch/html/biblio/pdf/ineichen_2010_aerosol_quantification(SP).pdf
Ineichen P. (2010). Interannual variability and global irradiance evaluation. IEA Task
XXXVI report. Available from:
http://www.cuepe.ch/html/biblio/pdf/ineichen_2010_05_interannual_variability.pdf
Kalnay E et al. (1996): The NMC/NCAR 40-year reanalysis project. Bull. Am. Meteorol.
Soc. 77 (3), 437–472.
Kasten F. (1980) A simple parameterization of two pyrheliometric formulae for determining
the Linke turbidity factor. Meteor. Rdsch. 33, 124–127.
Kinne, S. et al (2005) An AeroCom initial assessment optical properties in aerosol
component modules of global models Atmos. Chem. Phys. Discuss., 5, 1-46.
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 28 -
Linke F. (1922) Transmissions-koeffizient und Trübungsfaktor. Beitr. Phys. Fr. Atmos.
10, pp 91-103.
Liu B.Y.H, Jordan R.C (1960) The interrelationship and characteristic distribution of di-rect, diffuse and total solar radiation. Volume 4, Issue 3, Pages 1-19
Lorenz E. (2007): Improved diffuse radiation model, MSG. Report for the EC-project
PVSAT-2: Intelligent Performance Check of PV System Operation
Louche, A., Notton, G., Poggi, P. and Simonnot, G., (1991). Correlations for directnormal and global horizontal irradiation on a French Mediterranean site. Solar Energy.46, 261-266.
Mueller R.W. et al, (2004) Rethinking satellite based solar irradiance modelling – The
SOLIS clear sky module. Remote Sensing of Environment,91, 160-174.
Perez R., R. Seals, P. Ineichen, R. Stewart, D. Menicucci (1987) A New SimplifiedVersion of the Perez Diffuse Irradiance Model for Tilted Surfaces. DescriptionPerformance Validation. Solar Energy, 39, 221-232.
Perez R., Ineichen P., Seals R., Zelenka A. (1990) Making full use of the clearness indexfor parametrizing hourly insolation conditions. Solar Energy, Vol. 45, N° 2, pp. 111-114.
Perez R., P. Ineichen, E. Maxwell, R. Seals, A. Zelenka (1992) Dynamic global todirect irradiance conversion models. ASHARE Trans. Res. Series, 1992, 354-369
Perez R., P. Ineichen, K. Moore, M. Kmiecik, C. Chain, R. George, F. Vignola(2002) A New. Operational Satellite-to-Irradiance Model, Solar Energy, 73, 2002,307-317.
Polo, J., Zarzalejo, L. F., Martin, L., Navarro, A. A. and Marchante, R., (2009a) Estima-tion of daily Linke turbidity factor by using global irradiance measurements at solarnoon. Solar Energy. 83, 1177-1185.
Polo, J., Zarzalejo, L. F., Salvador, P. and Ramírez, L. (2009b) Angstrom turbidity andozone column estimations from spectral solar irradiance in a semi-desertic environmentin Spain. Solar Energy. 83, 257-263.
Randel D., T. H Vonder Haar, M.A. Ringerrud, G. L. Stephens, T.J. Greenwald C.L.Combs (1996) A New Global Water Vapour Dataset. Bulletin of the AmericanMeteorological Society, 77, 1233-1246.
Remund J. (2008), Updated Linke Turbidity and aerosol optical depth climatolo-gies, IEA SHC Task 36 expert meeting, June 2008, Wels, Austria.
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 29 -
Remund J. (2009): Aerosol optical depth and Linke turbidity climatology, Description forfinal report of IEA SHC Task 36, Meteotest Bern
Rigollier C., Bauer O., Wald L. (2000) On the Clear Sky Model of the ESRA - europeanSolar Radiation Atlas - with Respect to the Heliosat Method. Solar Energy 68 (1), 33-48.
Rigollier C., Lefèvre M, Wald L. (2004) The method heliosat-2 for deriving shortwavesolar irradiance radiation from satellite imagesSolar Energy, 77(2), 159-169
Ruiz-Arias J.A., T. Cebecauer, J. Tovar-Pescador, M. Šúri, (2010). Spatialdisaggregation of satellite-derived irradiance using a high resolution digitalelevation model, accepted to Solar Energy.
Ruyter de Wildt M., G. Seiz, A. Gruen (2007) Operational snow mapping usingmultitemporal Meteosat SEVIRI imagery, Remote Sensing of Environment, 109,29-41.
WMO Publication N° 8 (2008) Guide to Meteorological Instruments and Methods ofObservation, ISBN 978-92-63-100085
Zarzalejo, L. F., Polo, J., Martín, L., Ramírez, L. and Espinar, B. (2009). A new statisticalapproach for deriving global solar radiation from satellite images. Solar Energy. 83,480-484.
Zelenka A, Perez R, Seals R,Renne D (1999) Effective Accuracy of Satellite-DerivedHourly irradiance. Theor. Appl. Climatol. 62,199-207
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 30 -
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 31 -
Annexe: figures in the following pages:
For the clear, intermediate and overcast sky conditions:
• First and second order statistics in absolute values for the global hori-zontal irradiance. The sites in grey are not taken into account in theoverall statistics.
• First and second order statistics in relative values for the global horizon-tal irradiance. The sites in grey are not taken into account in the overallstatistics.
• First and second order statistics in absolute values for the normal beamirradiance. The sites in grey are not taken into account in the overallstatistics.
• First and second order statistics in relative values for the normal beamirradiance. The sites in grey are not taken into account in the overallstatistics.
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 32 -
Cle
ar s
ky c
onditio
ns:
first
and s
econd o
rder
sta
tist
ics
in a
bso
lute
val
ues
for
the
glo
bal
horizo
nta
l irra
dia
nce
. The
site
s in
gre
y ar
e not
take
n into
acc
ount
in t
he
ove
rall
stat
istics
.
Cle
ar s
ky c
onditio
ns:
first
and s
econd o
rder
sta
tist
ics
in rel
ativ
e va
lues
for th
e glo
bal
horizo
nta
l irr
adia
nce
. The
site
sin
gre
y ar
e not
take
n into
acc
ount
in t
he
ove
rall
stat
istics
.
GH
I [w
/m2]
nbR
2m
bd%
sd%
R2
mbd
%sd
%R
2m
bd%
sd%
R2
mbd
%sd
%R
2m
bd%
sd%
R2
mbd
%sd
%
Cab
auw
420
1374
0.98
0-4
120.
983
-812
0.97
6-6
130.
986
-510
0.98
5-6
10 C
ambo
rne
459
1362
0.98
5-5
100.
986
-710
0.97
6-8
130.
990
-68
0.99
0-6
8 C
arpe
ntra
s48
125
220.
993
-17
0.99
1-3
90.
988
-39
0.99
4-2
60.
993
07
0.92
1-7
22 D
avos
517
1775
0.95
4-7
150.
970
-912
0.91
5-1
021
0.94
8-2
016
0.93
5-1
418
0.83
2-1
629
El S
aler
549
2027
0.98
0-2
90.
987
-49
0.97
2-5
110.
985
-18
0.98
5-3
80.
978
-710
Gen
eva
508
1464
0.98
30
90.
983
-510
0.96
6-2
130.
987
08
0.98
62
90.
957
-515
Jun
gfra
ujoc
h60
015
580.
930
-618
0.93
2-1
716
0.83
1-1
128
0.68
5-5
033
0.67
5-4
434
0.76
8-3
130
Las
Maj
adas
539
2022
0.98
82
80.
990
08
0.98
32
100.
991
37
0.99
21
70.
982
-19
Ler
wic
k40
287
40.
970
-10
140.
972
-11
150.
950
-11
190.
977
-10
130.
976
-11
13 L
ocar
no51
218
130.
982
-19
0.98
5-5
100.
968
-212
0.98
4-1
80.
980
-29
0.97
2-8
11 N
ante
s45
615
420.
982
-411
0.97
8-8
120.
972
-713
0.98
8-4
90.
987
-79
0.96
4-6
15 P
ayer
ne50
415
390.
977
-211
0.97
2-9
130.
964
-414
0.98
0-3
100.
979
-111
0.95
8-7
15 S
ede
Boq
er66
222
850.
989
-16
0.98
8-7
60.
975
-59
0.98
9-4
60.
988
-56
0.97
9-7
8 S
ion
535
1851
0.95
7-7
130.
977
-211
0.94
1-9
160.
979
-69
0.97
2-7
110.
846
-11
25 S
onnb
lick
481
1837
0.92
1-7
210.
946
-817
0.87
0-1
128
0.79
0-3
831
0.77
4-2
934
0.84
7-1
029
Tam
anra
sset
621
3154
0.99
4-2
60.
990
18
0.98
50
90.
992
-26
0.99
2-2
6 T
hess
alon
iki
536
2025
0.98
9-1
70.
988
-68
0.97
1-6
120.
978
-510
0.97
7-6
100.
973
-511
Tor
aver
e40
015
450.
985
-410
0.96
8-8
170.
968
-714
0.98
3-6
110.
979
-612
Val
d'A
linya
578
1532
0.98
5-3
80.
985
010
0.97
9-2
100.
985
-89
0.98
6-5
80.
966
213
Vau
lx-e
n-V
elin
514
1468
0.98
50
80.
984
-49
0.97
30
110.
987
18
0.98
7-1
80.
973
-311
Wie
n46
015
510.
981
-310
0.98
3-7
110.
972
-513
0.98
6-5
90.
986
-39
0.95
6-6
16 Y
atir
Fore
st62
625
230.
993
-25
0.99
3-5
60.
987
-47
0.99
4-6
50.
994
-55
0.98
2-4
8 Z
uric
h46
513
510.
975
-213
0.97
3-7
150.
963
-316
0.97
2-4
130.
971
-115
0.95
6-6
16
All
site
s49
835
824
-29
-510
-411
-38
-39
-717
3Tie
rEn
Met
Sol (
Dum
ortie
r)So
larG
isH
elio
sat 3
v3En
Met
Sol (
Solis
)IrS
olA
v
Gh [
w/m
2]nb
R2
mbd
sdR
2m
bdsd
R2
mbd
sdR
2m
bdsd
R2
mbd
sdR
2m
bdsd
Cab
auw
420
1374
0.98
0-1
750
0.98
3-3
350
0.97
6-2
455
0.98
6-2
342
0.98
5-2
544
Cam
born
e45
913
620.
985
-25
460.
986
-34
470.
976
-36
570.
990
-27
360.
990
-28
36 C
arpe
ntra
s48
125
220.
993
-532
0.99
1-1
544
0.98
8-1
642
0.99
4-8
310.
993
034
0.92
1-3
410
7 D
avos
517
1775
0.95
4-3
976
0.97
0-4
764
0.91
5-5
310
90.
948
-101
810.
935
-70
920.
832
-82
148
El S
aler
549
2027
0.98
0-1
152
0.98
7-2
448
0.97
2-2
560
0.98
5-7
440.
985
-14
440.
978
-40
52 G
enev
a50
814
640.
983
046
0.98
3-2
451
0.96
6-8
670.
987
240
0.98
612
450.
957
-25
74 J
ungf
rauj
och
600
1558
0.93
0-3
910
50.
932
-100
980.
831
-68
168
0.68
5-3
0119
90.
675
-262
206
0.76
8-1
8617
9 L
as M
ajad
as53
920
220.
988
1342
0.99
0-1
430.
983
1252
0.99
118
370.
992
635
0.98
2-3
51 L
erw
ick
402
874
0.97
0-4
058
0.97
2-4
560
0.95
0-4
375
0.97
7-3
951
0.97
6-4
452
Loc
arno
512
1813
0.98
2-3
440.
985
-23
520.
968
-963
0.98
4-5
420.
980
-12
470.
972
-39
55 N
ante
s45
615
420.
982
-20
480.
978
-38
560.
972
-31
590.
988
-17
400.
987
-33
420.
964
-28
68 P
ayer
ne50
415
390.
977
-13
550.
972
-46
640.
964
-22
700.
980
-14
520.
979
-456
0.95
8-3
673
Sed
e B
oqer
662
2285
0.98
9-6
380.
988
-46
410.
975
-34
590.
989
-27
410.
988
-35
430.
979
-43
55 S
ion
535
1851
0.95
7-3
571
0.97
7-9
580.
941
-46
830.
979
-31
500.
972
-38
570.
846
-60
133
Son
nblic
k48
118
370.
921
-35
990.
946
-38
830.
870
-53
134
0.79
0-1
8515
00.
774
-139
162
0.84
7-4
914
0 T
aman
rass
et62
131
540.
994
-11
340.
990
451
0.98
5-2
550.
992
-14
390.
992
-15
37 T
hess
alon
iki
536
2025
0.98
9-7
380.
988
-34
430.
971
-30
630.
978
-28
550.
977
-30
550.
973
-29
60 T
orav
ere
400
1545
0.98
5-1
641
0.96
8-3
368
0.96
8-2
758
0.98
3-2
543
0.97
9-2
448
Val
d'A
linya
578
1532
0.98
5-1
548
0.98
5-3
570.
979
-11
600.
985
-48
490.
986
-26
470.
966
1175
Vau
lx-e
n-Ve
lin51
414
680.
985
143
0.98
4-2
047
0.97
3-2
590.
987
340
0.98
7-4
400.
973
-16
57 W
ien
460
1551
0.98
1-1
248
0.98
3-3
348
0.97
2-2
259
0.98
6-2
241
0.98
6-1
242
0.95
6-2
973
Yat
ir Fo
rest
626
2523
0.99
3-1
233
0.99
3-3
140
0.98
7-2
845
0.99
4-3
429
0.99
4-3
131
0.98
2-2
851
Zur
ich
465
1351
0.97
5-9
600.
973
-32
680.
963
-14
740.
972
-19
620.
971
-568
0.95
6-2
676
All
site
s49
835
824
-943
-23
49-1
857
-16
41-1
643
-25
66
3Tie
rEn
Met
Sol (
Solis
)So
larG
isH
elio
sat 3
v3En
Met
Sol (
Dum
ortie
r)IrS
olA
vGlobal irradiance
Clear sky conditions: 0.65 < K’t ≤ 1.00
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 33 -
Beam irradiance
Clear sky conditions 0.65 < K’t ≤ 1.00
Bn [
w/m
2]nb
R2
mbd
%sd
%R
2m
bd%
sd%
R2
mbd
%sd
%R
2m
bd%
sd%
R2
mbd
%sd
%R
2m
bd%
sd%
Cab
auw
530
1335
0.88
4-1
221
0.82
9-1
026
0.82
0-7
270.
867
-15
220.
865
-622
Cam
born
e55
611
950.
894
-920
0.82
6-2
250.
824
-626
0.88
1-1
121
0.88
8-3
20 C
arpe
ntra
s65
225
220.
901
-10
150.
827
-11
220.
870
-918
0.86
7-8
180.
868
-118
0.68
4-1
431
Dav
os67
617
750.
742
-730
0.52
5-8
390.
399
-12
490.
423
-45
430.
463
-22
450.
393
-18
52 E
l Sal
er G
enev
a63
114
640.
805
-820
0.77
5-1
322
0.75
0-3
250.
815
-720
0.83
43
190.
724
-14
26 J
ungf
rauj
och
381
1558
0.39
395
117
0.32
554
117
0.44
370
116
0.28
6-6
711
70.
294
-34
121
0.34
312
119
Las
Maj
adas
Ler
wic
k49
179
00.
839
-14
280.
783
-635
0.71
2-6
390.
849
-20
270.
858
-10
27 L
ocar
no N
ante
s59
515
420.
875
-14
180.
781
-15
250.
794
-13
240.
868
-11
180.
861
-15
180.
748
-14
28 P
ayer
ne58
515
390.
709
-630
0.65
5-1
233
0.64
81
360.
662
-332
0.69
57
310.
629
-10
35 S
ede
Boqe
r75
722
470.
824
-13
130.
666
-19
200.
628
-14
200.
722
-16
170.
704
-12
180.
540
-14
23 S
ion
Son
nblic
k51
818
370.
595
2960
0.40
526
710.
372
1278
0.32
2-6
170
0.29
9-2
981
0.39
216
79 T
aman
rass
et70
631
540.
904
-10
160.
702
329
0.73
3-6
260.
756
-525
0.79
02
23 T
hess
alon
iki
Tor
aver
e58
215
450.
875
-14
210.
740
-17
330.
757
-13
290.
815
-23
230.
810
-14
25 V
al d
'Alin
ya V
aulx
-en-
Velin
643
1468
0.86
9-9
150.
801
-11
190.
796
-121
0.84
6-5
160.
866
-416
0.78
1-1
121
Wie
n54
115
510.
760
-729
0.78
6-1
028
0.72
1-2
320.
761
-14
280.
768
129
0.70
7-1
233
Yat
ir Fo
rest
Zur
ich
All
site
s62
720
352
-10
19-1
026
-726
-10
22-4
22-1
329
3Tie
rEn
Met
Sol (
Solis
)En
Met
Sol (
Dum
ortie
r)So
larG
isH
elio
sat 3
v3IrS
olA
v
Bn [
w/m
2]nb
R2
mbd
sdR
2m
bdsd
R2
mbd
sdR
2m
bdsd
R2
mbd
sdR
2m
bdsd
Cab
auw
530
1335
0.88
4-6
411
20.
829
-51
139
0.82
0-3
514
10.
867
-79
115
0.86
5-3
311
8 C
ambo
rne
556
1195
0.89
4-4
910
90.
826
-11
141
0.82
4-3
614
40.
881
-63
114
0.88
8-1
910
9 C
arpe
ntra
s65
225
220.
901
-63
100
0.82
7-7
014
10.
870
-57
119
0.86
7-5
211
50.
868
-811
60.
684
-91
202
Dav
os67
617
750.
742
-49
202
0.52
5-5
526
30.
399
-83
328
0.42
3-3
0728
80.
463
-149
306
0.39
3-1
2534
8 E
l Sal
er G
enev
a63
114
640.
805
-51
129
0.77
5-8
313
90.
750
-20
159
0.81
5-4
312
30.
834
2112
20.
724
-90
165
Jun
gfra
ujoc
h38
115
580.
393
363
445
0.32
520
644
60.
443
265
440
0.28
6-2
5444
40.
294
-130
460
0.34
346
452
Las
Maj
adas
Ler
wic
k49
179
00.
839
-68
138
0.78
3-2
917
00.
712
-31
190
0.84
9-9
913
40.
858
-49
130
Loc
arno
Nan
tes
595
1542
0.87
5-8
310
70.
781
-92
147
0.79
4-7
914
40.
868
-68
108
0.86
1-8
810
70.
748
-85
168
Pay
erne
585
1539
0.70
9-3
317
70.
655
-72
193
0.64
85
210
0.66
2-2
018
90.
695
4218
40.
629
-60
206
Sed
e Bo
qer
757
2247
0.82
4-9
710
10.
666
-141
153
0.62
8-1
0514
80.
722
-121
126
0.70
4-9
213
40.
540
-102
171
Sio
n S
onnb
lick
518
1837
0.59
515
031
20.
405
133
366
0.37
265
402
0.32
2-3
1836
40.
299
-148
418
0.39
282
409
Tam
anra
sset
706
3154
0.90
4-7
411
20.
702
2420
40.
733
-41
186
0.75
6-3
317
40.
790
1316
1 T
hess
alon
iki
Tor
aver
e58
215
450.
875
-84
119
0.74
0-9
819
40.
757
-78
167
0.81
5-1
3513
70.
810
-83
145
Val
d'A
linya
Vau
lx-e
n-Ve
lin64
314
680.
869
-57
990.
801
-71
120
0.79
6-9
134
0.84
6-3
310
50.
866
-28
101
0.78
1-7
413
6 W
ien
541
1551
0.76
0-4
015
90.
786
-56
151
0.72
1-1
317
60.
761
-75
154
0.76
86
155
0.70
7-6
518
0 Y
atir
Fore
st Z
uric
h
All
site
s62
720
352
-66
121
-61
163
-45
161
-66
138
-24
137
-83
179
EnM
etSo
l (So
lis)
EnM
etSo
l (D
umor
tier)
Sola
rGis
Hel
iosa
t 3v3
3Tie
rIrS
olA
v
Cle
ar s
ky c
onditio
ns:
first
and s
econd o
rder
sta
tist
ics
in a
bso
lute
val
ues
for
the
norm
al b
eam
irra
dia
nce
. The
site
sin
gre
y ar
e not
take
n into
acc
ount
in t
he
ove
rall
stat
istics
.
Cle
ar s
ky c
onditio
ns:
first
and s
econd o
rder
sta
tist
ics
in rel
ativ
e va
lues
for th
e norm
al b
eam
irra
dia
nce
. The
site
s in
gre
y ar
e not
take
n into
acc
ount
in t
he
ove
rall
stat
istics
.
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 34 -
Global irradiance
Intermediate sky conditions 0.30 < K’t ≤ 0.65
Inte
rmed
iate
sky
conditio
ns:
first
and s
econd o
rder
sta
tist
ics
in a
bso
lute
val
ues
for th
e glo
bal
horizo
nta
l irr
adia
nce
.The
site
s in
gre
y ar
e not
take
n into
acc
ount
in t
he
ove
rall
stat
istics
.
Inte
rmed
iate
sky
conditio
ns:
first
and s
econd o
rder
sta
tist
ics
in r
elat
ive
valu
es for
the
glo
bal
horizo
nta
l irr
adia
nce
.The
site
s in
gre
y ar
e not
take
n into
acc
ount
in t
he
ove
rall
stat
istics
.
Gh [
W/m
2 ]nb
R2
mbd
%sd
%R
2m
bd%
sd%
R2
mbd
%sd
%R
2m
bd%
sd%
R2
mbd
%sd
%R
2m
bd%
sd%
Cab
auw
252
1578
0.95
0-4
220.
951
024
0.91
90
270.
970
117
0.96
80
17 C
ambo
rne
249
1546
0.94
10
240.
963
623
0.91
31
290.
962
520
0.96
34
20 C
arpe
ntra
s26
798
60.
937
826
0.96
615
260.
929
426
0.96
413
210.
965
1522
0.77
910
51 D
avos
235
1571
0.85
1-2
430.
948
2635
0.86
17
430.
897
-11
340.
887
-537
0.69
3-4
60 E
l Sal
er30
410
830.
919
727
0.95
312
250.
908
328
0.94
211
230.
943
1023
0.89
85
29 G
enev
a26
612
080.
935
725
0.95
26
250.
898
431
0.95
810
210.
958
1122
0.86
87
36 J
ungf
rauj
och
270
1220
0.76
31
540.
917
4644
0.77
327
550.
739
-24
450.
734
-14
480.
564
470
Las
Maj
adas
292
1036
0.92
512
280.
957
2026
0.90
010
310.
947
1925
0.94
916
240.
893
1332
Ler
wic
k23
413
070.
933
024
0.93
78
310.
870
635
0.93
8-3
230.
938
-423
Loc
arno
258
1199
0.94
49
250.
958
1931
0.92
812
300.
963
1422
0.96
213
220.
901
933
Nan
tes
264
1595
0.95
8-1
210.
938
731
0.92
80
270.
973
517
0.97
32
170.
896
534
Pay
erne
247
1398
0.94
25
250.
933
-129
0.90
44
320.
957
823
0.95
710
240.
867
437
Sed
e Bo
qer
274
522
0.93
111
270.
946
228
0.87
81
320.
944
1025
0.94
47
240.
900
932
Sio
n22
414
370.
907
-133
0.96
428
330.
905
034
0.95
35
250.
950
426
0.77
11
52 S
onnb
lick
287
1570
0.76
53
540.
901
3744
0.77
719
550.
767
-18
450.
761
-549
0.66
110
64 T
aman
rass
et31
391
50.
954
1325
0.96
230
350.
921
1836
0.95
625
310.
957
2430
The
ssal
onik
i26
010
360.
944
526
0.95
03
290.
921
231
0.94
23
260.
940
326
0.90
09
35 T
orav
ere
233
1368
0.94
1-3
260.
945
633
0.87
7-1
340.
941
-125
0.94
6-1
24 V
al d
'Alin
ya30
863
70.
923
1431
0.94
037
400.
894
1336
0.93
912
260.
940
1727
0.83
830
46 V
aulx
-en-
Velin
266
1610
0.95
58
220.
957
824
0.92
38
280.
968
1219
0.96
810
190.
884
834
Wie
n25
415
700.
931
-127
0.93
7-2
290.
920
-129
0.95
80
210.
958
322
0.84
81
40 Y
atir
Fore
st32
550
50.
899
627
0.94
98
220.
865
230
0.94
35
190.
942
520
0.86
23
31 Z
uric
h25
915
600.
901
635
0.91
05
390.
884
737
0.90
66
350.
908
936
0.85
54
42
All
site
s24
924
096
527
930
532
824
825
1153
3Tie
rEn
Met
Sol (
Dum
ortie
r)So
larG
isH
elio
sat 3
v3En
Met
Sol (
Solis
)IrS
olA
v
Gh [
W/m
2 ]nb
R2
mbd
sdR
2m
bdsd
R2
mbd
sdR
2m
bdsd
R2
mbd
sdR
2m
bdsd
Cab
auw
252
1578
0.95
0-9
540.
951
-160
0.91
90
690.
970
343
0.96
80
44 C
ambo
rne
249
1546
0.94
11
600.
963
1658
0.91
32
720.
962
1250
0.96
310
49 C
arpe
ntra
s26
798
60.
937
2171
0.96
641
680.
929
1271
0.96
435
560.
965
3959
0.77
927
137
Dav
os23
515
710.
851
-510
00.
948
6282
0.86
115
101
0.89
7-2
681
0.88
7-1
287
0.69
3-8
141
El S
aler
304
1083
0.91
922
820.
953
3777
0.90
810
840.
942
3470
0.94
329
690.
898
1487
Gen
eva
266
1208
0.93
520
670.
952
1567
0.89
810
830.
958
2755
0.95
830
580.
868
1995
Jun
gfra
ujoc
h27
012
200.
763
414
60.
917
123
118
0.77
372
148
0.73
9-6
512
10.
734
-38
130
0.56
411
189
Las
Maj
adas
292
1036
0.92
534
810.
957
5775
0.90
028
900.
947
5674
0.94
947
710.
893
3893
Ler
wic
k23
413
070.
933
057
0.93
719
740.
870
1381
0.93
8-7
540.
938
-10
54 L
ocar
no25
811
990.
944
2464
0.95
850
800.
928
3277
0.96
337
560.
962
3357
0.90
123
84 N
ante
s26
415
950.
958
-155
0.93
818
810.
928
-172
0.97
314
460.
973
445
0.89
612
91 P
ayer
ne24
713
980.
942
1262
0.93
3-2
710.
904
1079
0.95
721
560.
957
2559
0.86
710
92 S
ede
Boqe
r27
452
20.
931
3075
0.94
65
770.
878
289
0.94
428
680.
944
1964
0.90
026
86 S
ion
224
1437
0.90
7-2
750.
964
6273
0.90
5-1
760.
953
1256
0.95
08
580.
771
311
6 S
onnb
lick
287
1570
0.76
59
156
0.90
110
612
50.
777
5415
80.
767
-51
128
0.76
1-1
514
10.
661
2918
3 T
aman
rass
et31
391
50.
954
4180
0.96
295
108
0.92
156
113
0.95
678
980.
957
7695
The
ssal
onik
i26
010
360.
944
1469
0.95
08
760.
921
480
0.94
28
690.
940
869
0.90
024
90 T
orav
ere
233
1368
0.94
1-7
610.
945
1577
0.87
7-2
800.
941
-258
0.94
6-1
56 V
al d
'Alin
ya30
863
70.
923
4295
0.94
011
412
30.
894
4011
00.
939
3880
0.94
051
840.
838
9414
3 V
aulx
-en-
Velin
266
1610
0.95
523
580.
957
2264
0.92
322
740.
968
3151
0.96
826
520.
884
2290
Wie
n25
415
700.
931
-268
0.93
7-6
730.
920
-272
0.95
80
530.
958
655
0.84
83
101
Yat
ir Fo
rest
325
505
0.89
919
880.
949
2572
0.86
56
970.
943
1763
0.94
215
640.
862
1110
2 Z
uric
h25
915
600.
901
1691
0.91
012
100
0.88
418
970.
906
1690
0.90
823
940.
855
1010
8
All
site
s24
924
096
1267
2275
1280
2060
1961
1995
3Tie
rEn
Met
Sol (
Solis
)So
larG
isH
elio
sat 3
v3En
Met
Sol (
Dum
ortie
r)IrS
olA
v
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 35 -
Beam irradiance
Intermediate sky conditions 0.30 < K’t ≤ 0.65
Bn [
W/m
2 ]nb
R2
mbd
%sd
%R
2m
bd%
sd%
R2
mbd
%sd
%R
2m
bd%
sd%
R2
mbd
%sd
%R
2m
bd%
sd%
Cab
auw
122
1568
0.76
412
740.
680
8010
40.
577
5612
10.
772
3173
0.75
437
80 C
ambo
rne
104
1293
0.81
652
900.
759
144
115
0.73
874
125
0.81
563
860.
810
7497
Car
pent
ras
143
986
0.72
252
850.
729
125
980.
679
6410
60.
782
7678
0.77
994
860.
419
111
167
Dav
os76
1571
0.56
114
621
10.
566
322
212
0.55
719
525
00.
688
3211
80.
675
106
167
0.47
416
325
9 E
l Sal
er G
enev
a15
312
080.
725
3984
0.73
569
840.
615
5711
70.
791
4368
0.80
560
750.
651
5610
1 J
ungf
rauj
och
1412
200.
269
1660
1743
0.11
733
0118
330.
203
2624
2311
0.25
046
179
00.
270
988
1322
0.16
015
4218
01 L
as M
ajad
as L
erw
ick
8212
300.
673
6412
90.
570
197
198
0.51
315
121
40.
692
4910
70.
662
6212
4 L
ocar
no N
ante
s16
315
950.
798
056
0.64
654
930.
684
1382
0.84
017
510.
836
950
0.64
628
94 P
ayer
ne11
113
980.
675
6011
10.
620
8812
40.
544
9716
00.
711
6899
0.71
793
111
0.53
180
142
Sed
e Bo
qer
272
450
0.65
914
520.
355
474
0.52
015
680.
541
859
0.51
511
620.
315
2982
Sio
n S
onnb
lick
3215
700.
401
680
786
0.40
213
1775
90.
395
969
920
0.45
117
835
20.
490
466
594
0.39
182
391
0 T
aman
rass
et12
491
50.
721
9399
0.57
825
217
10.
641
149
158
0.60
617
214
50.
637
200
140
The
ssal
onik
i T
orav
ere
131
1368
0.73
32
820.
726
8111
30.
411
5714
70.
633
1696
0.69
429
93 V
al d
'Alin
ya V
aulx
-en-
Velin
206
1592
0.80
213
530.
704
3867
0.69
931
770.
852
2246
0.86
022
460.
653
2275
Wie
n15
215
700.
748
1675
0.74
849
840.
674
4098
0.80
417
630.
792
3872
0.63
334
102
Yat
ir Fo
rest
Zur
ich
All
site
s14
215
173
2878
8410
357
115
4176
5181
5212
0
3Tie
rEn
Met
Sol (
Solis
)En
Met
Sol (
Dum
ortie
r)So
larG
isH
elio
sat 3
v3IrS
olA
v
Bn [
W/m
2 ]nb
R2
mbd
sdR
2m
bdsd
R2
mbd
sdR
2m
bdsd
R2
mbd
sdR
2m
bdsd
Cab
auw
122
1568
0.76
414
900.
680
9712
60.
577
6814
80.
772
3889
0.75
445
97 C
ambo
rne
104
1293
0.81
654
930.
759
150
120
0.73
877
130
0.81
566
890.
810
7710
1 C
arpe
ntra
s14
398
60.
722
7412
20.
729
179
141
0.67
992
152
0.78
210
911
10.
779
136
123
0.41
915
924
0 D
avos
7615
710.
561
111
161
0.56
624
516
20.
557
149
191
0.68
824
900.
675
8112
70.
474
124
197
El S
aler
Gen
eva
153
1208
0.72
560
129
0.73
510
612
80.
615
8817
90.
791
6610
40.
805
9311
50.
651
8515
5 J
ungf
rauj
och
1412
200.
269
232
243
0.11
746
125
60.
203
366
322
0.25
064
110
0.27
013
818
50.
160
215
251
Las
Maj
adas
Ler
wic
k82
1230
0.67
352
105
0.57
016
116
20.
513
124
175
0.69
240
870.
662
5110
1 L
ocar
no N
ante
s16
315
950.
798
091
0.64
688
152
0.68
422
134
0.84
027
830.
836
1482
0.64
646
154
Pay
erne
111
1398
0.67
566
123
0.62
098
138
0.54
410
817
80.
711
7611
00.
717
104
124
0.53
189
158
Sed
e Bo
qer
272
450
0.65
937
142
0.35
510
200
0.52
040
185
0.54
121
160
0.51
531
170
0.31
578
222
Sio
n S
onnb
lick
3215
700.
401
216
250
0.40
241
924
20.
395
308
293
0.45
157
112
0.49
014
818
90.
391
262
290
Tam
anra
sset
124
915
0.72
111
512
20.
578
312
212
0.64
118
419
50.
606
213
179
0.63
724
717
3 T
hess
alon
iki
Tor
aver
e13
113
680.
733
310
80.
726
106
148
0.41
175
193
0.63
321
125
0.69
438
122
Val
d'A
linya
Vau
lx-e
n-V
elin
206
1592
0.80
227
110
0.70
479
139
0.69
964
159
0.85
245
950.
860
4495
0.65
345
154
Wie
n15
215
700.
748
2411
40.
748
7412
70.
674
6114
90.
804
2696
0.79
257
109
0.63
351
155
Yat
ir Fo
rest
Zur
ich
All
site
s14
215
173
4011
011
914
681
163
5810
872
114
7317
1
EnM
etSo
l (So
lis)
EnM
etSo
l (D
umor
tier)
Sola
rGis
Hel
iosa
t 3v3
3Tie
rIrS
olA
v
Inte
rmed
iate
sky
conditio
ns:
first
and s
econd o
rder
sta
tist
ics
in a
bso
lute
val
ues
for
the
norm
al b
eam
irr
adia
nce
.The
site
s in
gre
y ar
e not
take
n into
acc
ount
in t
he
ove
rall
stat
istics
.
Inte
rmed
iate
sky
conditio
ns:
first
and s
econd o
rder
sta
tist
ics
in rel
ativ
e va
lues
for th
e norm
al b
eam
irra
dia
nce
. The
site
s in
gre
y ar
e not
take
n into
acc
ount
in t
he
ove
rall
stat
istics
.
Five satellite products deriving beam and global irradiance against ground data validationP. Ineichen
- 36 -
Global irradiance
Overcast sky conditions 0.00 < K’t ≤ 0.30
Gh [
W/m
2 ]nb
R2
mbd
%sd
%R
2m
bd%
sd%
R2
mbd
%sd
%R
2m
bd%
sd%
R2
mbd
%sd
%R
2m
bd%
sd%
Cab
auw
7912
010.
879
1241
0.85
231
600.
820
4856
0.90
723
390.
904
2140
Cam
born
e91
1179
0.86
69
430.
897
4365
0.84
834
490.
898
2241
0.90
022
40 C
arpe
ntra
s77
485
0.80
326
650.
906
6768
0.74
841
650.
879
4654
0.88
145
550.
673
9213
0 D
avos
9985
40.
704
2472
0.89
512
910
90.
775
5167
0.79
814
530.
791
2557
0.48
551
107
El S
aler
8649
00.
697
3179
0.80
348
740.
627
6610
60.
789
4971
0.78
947
710.
594
4789
Gen
eva
8995
00.
829
2452
0.89
326
490.
803
4359
0.87
029
450.
871
3046
0.65
338
84 J
ungf
rauj
och
105
533
0.63
273
123
0.90
018
414
70.
668
141
122
0.61
125
810.
594
4493
0.46
797
170
Las
Maj
adas
100
451
0.60
627
770.
821
8978
0.56
162
960.
749
5672
0.75
252
710.
647
4993
Ler
wic
k85
1125
0.86
831
490.
873
5383
0.79
256
640.
834
2048
0.83
919
48 L
ocar
no64
1172
0.85
055
750.
848
111
119
0.78
299
107
0.87
856
670.
877
5466
0.72
772
114
Nan
tes
9610
630.
856
338
0.81
746
780.
837
2645
0.90
719
340.
909
1432
0.65
027
71 P
ayer
ne88
1068
0.81
315
510.
831
1658
0.76
842
630.
881
2543
0.88
426
440.
654
3983
Sed
e Bo
qer
118
830.
579
1473
0.78
745
750.
437
7612
50.
697
2862
0.69
823
600.
443
2096
Sio
n81
992
0.69
028
820.
869
141
118
0.80
047
710.
840
3561
0.84
531
600.
555
5610
9 S
onnb
lick
138
526
0.73
552
990.
747
139
117
0.70
577
108
0.69
723
820.
686
4494
0.57
875
129
Tam
anra
sset
114
224
0.67
829
760.
865
120
107
0.71
773
980.
754
7913
10.
754
7813
1 T
hess
alon
iki
9161
40.
751
1865
0.75
63
680.
673
4687
0.77
426
620.
772
2562
0.61
243
98 T
orav
ere
7987
10.
830
2862
0.87
063
960.
668
6071
0.81
819
500.
825
1950
Val
d'A
linya
9733
00.
762
5894
0.80
319
214
80.
603
105
131
0.79
175
880.
790
8393
0.60
911
515
1 V
aulx
-en-
Vel
in86
1063
0.82
624
530.
854
4162
0.78
947
640.
869
2847
0.87
124
460.
627
3989
Wie
n90
1082
0.84
48
450.
848
753
0.81
130
500.
878
1240
0.87
714
410.
637
2882
Yat
ir Fo
rest
123
193
0.71
430
760.
858
6672
0.62
461
850.
736
3563
0.73
933
620.
574
3388
Zur
ich
9112
810.
760
2663
0.78
533
820.
740
5572
0.76
929
660.
770
3169
0.61
942
93
All
site
s82
1591
722
6048
8051
7530
5829
5963
138
3Tie
rEn
Met
Sol (
Dum
ortie
r)So
larG
isH
elio
sat 3
v3En
Met
Sol (
Solis
)IrS
olA
v
Gh [
W/m
2 ]nb
R2
mbd
sdR
2m
bdsd
R2
mbd
sdR
2m
bdsd
R2
mbd
sdR
2m
bdsd
Cab
auw
7912
010.
879
932
0.85
225
480.
820
3844
0.90
718
310.
904
1731
Cam
born
e91
1179
0.86
68
390.
897
3959
0.84
831
450.
898
2037
0.90
020
37 C
arpe
ntra
s77
485
0.80
320
500.
906
5252
0.74
832
500.
879
3542
0.88
135
420.
673
7110
0 D
avos
9985
40.
704
2472
0.89
512
810
80.
775
5166
0.79
814
530.
791
2557
0.48
550
107
El S
aler
8649
00.
697
2768
0.80
341
630.
627
5791
0.78
942
610.
789
4060
0.59
440
76 G
enev
a89
950
0.82
921
460.
893
2344
0.80
339
530.
870
2640
0.87
127
410.
653
3475
Jun
gfra
ujoc
h10
553
30.
632
7712
90.
900
193
155
0.66
814
812
90.
611
2785
0.59
447
980.
467
102
179
Las
Maj
adas
100
451
0.60
627
770.
821
8978
0.56
162
960.
749
5672
0.75
252
700.
647
4993
Ler
wic
k85
1125
0.86
827
410.
873
4671
0.79
248
550.
834
1741
0.83
916
41 L
ocar
no64
1172
0.85
035
480.
848
7176
0.78
263
690.
878
3643
0.87
735
430.
727
4673
Nan
tes
9610
630.
856
337
0.81
745
750.
837
2543
0.90
718
320.
909
1431
0.65
026
68 P
ayer
ne88
1068
0.81
313
450.
831
1451
0.76
837
550.
881
2238
0.88
423
380.
654
3473
Sed
e Bo
qer
118
830.
579
1786
0.78
753
890.
437
8914
80.
697
3373
0.69
827
710.
443
2411
4 S
ion
8199
20.
690
2267
0.86
911
495
0.80
038
570.
840
2849
0.84
526
480.
555
4688
Son
nblic
k13
852
60.
735
7113
70.
747
192
161
0.70
510
514
80.
697
3211
30.
686
6112
90.
578
103
178
Tam
anra
sset
114
224
0.67
833
870.
865
137
122
0.71
784
112
0.75
490
150
0.75
490
149
The
ssal
onik
i91
614
0.75
116
590.
756
361
0.67
342
780.
774
2456
0.77
223
560.
612
3989
Tor
aver
e79
871
0.83
022
490.
870
5075
0.66
847
560.
818
1540
0.82
515
39 V
al d
'Alin
ya97
330
0.76
256
910.
803
185
143
0.60
310
112
60.
791
7385
0.79
080
900.
609
111
146
Vau
lx-e
n-V
elin
8610
630.
826
2045
0.85
435
530.
789
4055
0.86
924
400.
871
2040
0.62
734
77 W
ien
9010
820.
844
740
0.84
86
480.
811
2745
0.87
810
360.
877
1236
0.63
725
73 Y
atir
Fore
st12
319
30.
714
3793
0.85
881
880.
624
7510
50.
736
4377
0.73
940
770.
574
4110
9 Z
uric
h91
1281
0.76
023
580.
785
3074
0.74
050
660.
769
2660
0.77
029
630.
619
3985
All
site
s82
1591
718
4939
6542
6225
4824
4836
79
3Tie
rEn
Met
Sol (
Solis
)So
larG
isH
elio
sat 3
v3En
Met
Sol (
Dum
ortie
r)IrS
olA
v
Ove
rcas
t sk
y co
nditio
ns:
first
and s
econd o
rder
sta
tist
ics
in a
bso
lute
val
ues
for th
e glo
bal
horizo
nta
l irr
adia
nce
. The
site
s in
gre
y ar
e not
take
n into
acc
ount
in t
he
ove
rall
stat
istics
.
Ove
rcas
t sk
y co
nditio
ns:
first
and s
econd o
rder
sta
tist
ics
in r
elat
ive
valu
es for
the
glo
bal
horizo
nta
l irr
adia
nce
. The
site
s in
gre
y ar
e not
take
n into
acc
ount
in t
he
ove
rall
stat
istics
.