World Renewable Energy Congress VII, Cologne, Germany, 29 June - 5 July, 2002

SOLAR RESOURCE ASSESSMENT AND SITE EVALUATION

USING REMOTE SENSING METHODS

Carsten Hoyer, Christoph Schillings*,

Detlev Heinemann, Hermann Mannstein**, Franz Trieb*

University of Oldenburg, Faculty of Physics, Department of Energy and Semiconductor Research

D-26111 Oldenburg, Phone +49 441 798 3929, Fax +49 441 798 3326, Email: hoyer@[Link]
*
German Aerospace Center (DLR), Institute of Technical Thermodynamics, Stuttgart, Germany
**
German Aerospace Center (DLR), Institute of Atmospheric Physics, Oberpfaffenhofen, Germany

ABSTRACT

In this paper we describe and compare two methods used by the University of Oldenburg and DLR Stuttgart
to calculate surface solar irradiance from METEOSAT 7 imagery. We examine the consistency of the two
methods and compare the results with ground measured data. As a result we show the differences between
the methods and the errors relative to ground measurements. The benefit of such satellite derived solar
irradiance is shown by a brief presentation of two example applications. Finally an outlook is given into
potential improvements resulting from the enhanced possibilities of the METEOSAT Second Generation
(MSG) satellite which is due to start in 2002.

INTRODUCTION

An efficient world wide expansion of solar power requires reliable data on the solar energy resource. The
assessment of the available solar irradiance is of particular importance for planning and performance
evaluation. Ground based measurements are expensive and are rarely available for most sites, especially in
the solar belt. Geostationary satellites like METEOSAT 7 provide the opportunity to derive information on
the solar irradiance for a large area at a temporal resolution of up to 30 minutes and a spatial resolution of up
to 2.5 km.

THE METHODS

The general idea of both methods is to deal with atmospheric and cloud extinction separately. In a first step
the clear-sky irradiance for a given location and time is calculated. The methods use different models,
assumptions and input data to account for the atmospheric state. Details will be given for each of the
methods below. In a second step a cloud index is derived from METEOSAT imagery. This step uses the fact
that the planetary albedo measured by the satellite is proportional to the amount of cloudiness. The so
derived cloud index is then correlated to the cloud transmission. The clear-sky irradiance is then diminished
by the cloud transmission to infer the ground irradiance.

The HELIOSAT Method

The HELIOSAT-Method was originally proposed by Cano [1] and later modified by Beyer et al. [2] and
Hammer [3]. For the calculation of the clear-sky irradiance it uses the direct irradiance model of Page [4]
 World Renewable Energy Congress VII, Cologne, Germany, 29 June - 5 July, 2002

and diffuse irradiance model of Dumortier [5]. Both use the Linke turbidity factor for atmospheric
extinction. The direct irradiance is:
G direct ,clearsky = G 0 ⋅ ε ⋅ e −0.8662⋅TL ( 2 )⋅δ R ( m )⋅m (1)

where G0 is the extraterrestrial irradiance, ε the eccentricity correction, TL the Linke turbidiy factor for
airmass 2, δR(m) the Rayleigh optical thickness and m the airmass. The diffuse irradiance is an empirical fit
by Dumortier [5]:

Gdiffus ,clearsky = G0 ⋅ ε ⋅ (0.0065 + ( −0.045 + 0.0646 ⋅ TL ( 2)) ⋅ cos θ z + (0.014 − 0.0327 ⋅ TL ( 2)) ⋅ cos θ z2 (2)

Here θz is the solar zenith angle. Since there is no information on the atmospheric turbidity from
METOSAT, a climatological relation is used:

 2π   2π 
TL ( 2) = T0 + u ⋅ cos  + v ⋅ sin  (3)
 365 ⋅ J   365 ⋅ J 

T0, u and v are site specific fit parameters, J is the day of the year. A map with the parameters for Europe has
been set up during the EU-funded Satel-Light project [6].

For the derivation of the minimum and maximum values of the cloud index albedo values are needed. The
minimum ρmin corresponds to the reflectance of the ground and the maximum ρmax to optically thick clouds.
For the minimum ground albedo maps are computed on a monthly basis. The maximum is derived from a
statistical analysis of satellite images and is commonly done only once for each satellite sensor. The cloud
index n is then given by

ρ − ρ min
n= . (4)
ρ max − ρ min

This cloud index is then empirically correlated to the clear sky index kT*. This relationship is basically kT* =
1 – n with minor modifications for n→0 and n→1.

The ground irradiance is obtained from

G = kT* ⋅ (Gdirect ,clearsky ⋅ cos θ z + Gdiffus,clearsky ) (5)

The direct and diffuse component of the ground irradiance is then calculated using a statistical model of
Skartveit/Olseth [7].

The DLR Method for Direct Normal Irradiance (DNI)

The method for the DNI is developed by DLR using the clear sky parameterization model by Bird [8] that
uses transmittances for the individual atmospheric constituents. We added a transmission coefficient to take
into account the attenuation of clouds. DNI is defined as follows:

G direct = G 0 ⋅ ε ⋅ τ R ⋅ τ Gas ⋅ τ Ozon ⋅ τ WV ⋅ τ Ae ⋅ τ Cl (6)

with the eccentricity corrected extraterrestrial solar irradiance (G0) and the transmittance functions τi of the
Rayleigh atmosphere, mixed gases (CO2 and O2), O3, water vapor, aerosols and clouds, respectively. This
method needs information about atmospheric ozone, water vapor and aerosol. The cloud index τCl is
calculated by using the IR and VIS channels from the METEOSAT satellite based on self adjusting, local
thresholds which represent the daily variation of the surface properties [9]. The reference land surface
temperature is described as a function of time for every pixel:
 World Renewable Energy Congress VII, Cologne, Germany, 29 June - 5 July, 2002

T = a0 + a1 (cos(χ − a3 + sin (a2 )× sin (χ − a3 )) + 0.1× sin (χ − a3 )) (7)

with χ = t / 24 × 2π and t = decimal hours of the satellite (UTC). a0 gives the daily mean temperature, a1 the
temperature amplitude, a2 influences the width and steepness of the daily temperature wave and a3 gives the
phase shift, which is dominated by the local solar time.

COMPARISON TO GROUND DATA

We have compared results of both methods to ground data from Girona, located in north-eastern Spain, 30
km off the Mediterranean Sea. The data is global and diffuse horizontal irradiance from 1st Jan. 1998 to 15th
Nov. 1998. We have computed the root mean square error (RMSE) and the mean bias (MBE). The relative
errors are with respect to the ground measured values. The ground data averages were calculated separately
for each method, since they show different gaps in the satellite image series. This results in different relative
RMSE values even for similar RMSEs. For the computation of the errors night-time values, ground
measurement errors and missing satellite images have been eliminated. The errors are given in table 1.

Direct Irradiance
We start with the comparison of the direct
irradiance calculated by both methods. Figure
1 shows the results of the methods compared
Satellite Gdirect,hor(W/m2)

with ground measured data. Both methods

show consistent results. The DLR method
shows a slight overestimation, whereas the
HELIOSAT derived data are slightly below the
ground data. The scatter of the DLR method is
larger than of the HELIOSAT method.

To compare the clear sky models we have

chosen clear sky periods. The selection criteria
were, that the direct irradiance had to be at
least twice as large as the diffuse irradiance
and there had to be at least four consecutive Ground Gdirect,hor(W/m2)
values which fulfil these criteria. This leads to
Figure 1: Comparison of ground measured and calculated direct
about 1000 values for each of the methods. The
horizontal irradiance, all days.
results are shown in Figure 2. The scatter is
significantly smaller for both methods, the
HELIOSAT method has a larger negative bias,
many of the points below 500 W/m² are below
the diagonal, whereas the DLR method is
Satellite Gdirect,hor(W/m2)

nicely centered around the line.

Ground Gdirect,hor(W/m2)

Figure 2: Comparison of ground measured and calculated direct

horizontal irradiance, clearsky periods
 World Renewable Energy Congress VII, Cologne, Germany, 29 June - 5 July, 2002

TABLE 1
ERROR VALUES: HORIZONTAL IRRADIANCE
Method RMSE rRMSE MBE rMBE
DLR direct 130.6 44.4% 14.6 5.6%
HELIOSAT direct 79.7 28.6% -2.7 -0.6%
DLR direct (clear sky) 81.8 17.7% 4.5 1.0%
HELIOSAT direct (clear sky) 73.8 17.8% -47.1 -10.1%
HELIOSAT global 62.8 16.6% -2.3 -0.6%

Gobal Irradiance
Since only the HELIOSAT method calculates
global irradiance, no comparison is made for
the DLR method. Figure 3 shows a scatter

Satellite Gglobal,hor(W/m2)
plot for global irradiance. The scatter is
smaller than for the direct irradiance.

Conclusion
Both methods show good results for
calculating direct and global irradiance with a
high temporal and spatial resolution. The
quality of the derived values is comparable to
a ground station at 25 km distance [10]. The
HELIOSAT method performs better in all-
day cases, whereas the DLR method is
preferable in clear sky situations. This is Ground Ggloabl,hor(W/m2)
probably due to the different focus settinggs Figure 3: Comparison of global irradiance for the HELISOAT
of the two methods. The HELIOSAT method method, all days.
focuses on all-day global irradiance. The
DLR method is designed for solar thermal power plant evaluation, where clear sky direct irradiance is most
important. Here the DLR method can use the advantage of a much more detailed clear sky model using
atmospheric input data for ozone, water vapor and aerosol.

APPLICATIONS

Remote Performance Check of PV Systems: PVSAT

The idea of the PVSAT procedure is to evaluate the performance of grid connected photovoltaic (PV)
systems on a monthly basis and to detect failures, which lead to losses in the energy yield. The yield strongly
depends on the incoming solar irradiance making location and time specific irradiance information
absolutely necessary. Remote sensing techniques have a strong advantage over ground based measurements
due to its high spatial coverage. Ground based measurement networks are very expensive and satellite
derived data have shown to be preferable to ground measurements of more than 25 km distance [10]. The
PVSAT procedure calculates a hourly time series of global irradiance using the HELIOSAT method. This
time series is used to simulate the yield of the PV system. The so calculated yield can be compared to the
yield of the actual system. Significant differences between both figures are a strong indication of a
malfunction of the PV system.

Site Evaluation for Solar Thermal Power Plants: STEPS

Solar thermal power plants will provide a major share of the renewable energy sources needed in the future.
STEPS [11,12], an Evaluation system for Solar Thermal Power Stations, was designed to calculate the
performance of such power stations as function of solar direct irradiance, geographical conditions (land
slope, land cover, distance from cooling water resources, etc.), infrastructure (pipelines, electricity grids,
streets etc.) and the configuration and performance of a selected solar thermal power plant concept. A
 World Renewable Energy Congress VII, Cologne, Germany, 29 June - 5 July, 2002
potential
geographic information system (GIS) is used to sites

combine and process all the parameters for site direct

assessment as shown in Figure 4. One of the most irradiation

important input parameter is the solar direct Infrastructure levelised electricity

irradiance. To provide geographically continuous costs cost in Morocco
information over several years about the solar insurance
costs
resource in remote regions, the irradiance is
calculated using remote sensing methods described
above. Results were obtained with high spatial and
temporal resolution.

OUTLOOK: PERSPECTIVES OF MSG

Figure 4: Combining satellite derived direct irradiance with
The new European meteorological satellite MSG geographical, technical and economical data to calculate the
levelised electricity costs, for example in Morocco.
(METOSAT Second Generation) which is due to
start in summer 2002 offers additional spectral channels and a higher spatial and temporal resolution. The
added channels will allow for a better of the atmospheric state, especially regarding clouds. This enables the
use of radiative transfer models instead of parameterized models for e. g. the turbidity in HELIOSAT. A
further benefit is the determination of atmospheric constituents (e.g. water vapor, ozone, ...) from only one
satellite platform which reduces the dependency of external data. Overall a significant increase in accuracy
is expected.

ACKNOWLEDGEMENTS

The ground measurements of Girona are kindly provided by Josep-Abel González, Departament de Física,
Escola Politècnica Superior, Girona (Spain).

REFERENCES

1. Cano, D., Monget, J.M., Albuisson, M., Guillard, H., Regas, N. and Wald, L. (1986): “A method for the
determination of the global solar radiation from meteorological satellite data”, Solar Energy, 37, 31-29.
2. Beyer, H.G., Constanzo, C., Heinemann, D. (1996): “Modifications of the Heliosat procedure for
irradiance estimates from satellite data”, Solar Energy, 56, 207-212.
3. Hammer, A. (2000): “Anwendungsspezifische Solarstrahlungsinformationen aus Meteosat-Daten”.
Ph.D. Thesis, Oldenburg.
4. Page, J. (1996): “Algorithms for the Satellight programme”, Technical Report.
5. Dumortier, D. (1995): “Modelling global and diffuse horizontal irradiance under cloudless skies with
different turbidities”, Daylight II, JOUR2-CT92-0144, Final Report Vol 2.
6. [Link]
7. Skartveit, A., Olseth, J.A., Tuft, M.E. (1998): “An hourly diffuse fraction model with correction for
variability and surface albedo”. Solar Energy, 63, 173-183.
8. Bird, R.E. (1984): “A simple, solar spectral model for direct-normal and diffuse horizontal irradiance”,
Solar Energy, 32, 461-471.
9. Mannstein H., H. Broesamle, C. Schillings, F. Trieb (1999): “Using a meteosat cloud index to model the
performance of solar thermal power stations”, Procs. Eumetsat Conf., 239-246, Copenhagen.
10. Zelenka, A., Perez, R., Seals, R. and Renné, D. (1999): “Effective accuracy of satellite-derived hourly
irradiances”, Theor. Appl. Climatol., 62, 199-207.
11. Broesamle, H., H. Mannstein, C. Schillings and F. Trieb (2001): “Assessment of solar electricity
potentials in North Africa based on satellite data and a geographic information system”, Solar Energy,
70 (1), 1-12.
12. [Link]/steps