PVGIS radiation databases

PVGIS needs data on solar radiation in order to make estimates of the performance of PV systems and to do the other calculations possible in the web application. There exist a number of different sources of solar radiation data, but none of them are perfect, so it is important to understand the strengths and weaknesses of each data source. In the new version of PVGIS (autumn 2010), we have for the first time included a choice of solar radiation databases for some regions. For this reason we will here describe each of the databases that the user can choose.

NOTE: Below we will talk about the differences between the old version (PVGIS-3) and the new version (PVGIS-CMSAF). However, the old database is still available in the user interface. In some regions the old database is still the only choice.

Types of solar radiation data sources

• Ground measurements
• Calculations based on satellite data

Ground measurements of solar radiation

Direct measurements of the solar radiation at ground level can be made with a number of different instruments. One widely used instrument is the pyranometer. Typically, the instrument measures all the radiation coming from the sun and from the sky or clouds. When you want to know the solar radiation at a specific place, ground station measurements give the best results. It is also possible to measure with a high time resolution, typically every minute or even more often.

Possible problems with the measurements, apart from failure in the measurement system itself, is that the sensor may be covered with dirt, frost, or snow, or that the sensor is shadowed by nearby trees or buildings for some of the time during the year. These problems can be removed by careful siting and maintenance, but it makes it more uncertain to use data where you don't have direct experience with the measurements. Most of these potential problems will cause the measurement readings to be too low.

When there are no direct measurements at a given place, it is still possible to estimate the solar radiation from mesurements made nearby. Of course the quality of the estimate will decrease as the distance to the measurement site increases. It is also possible to combine data from several different measurement locations to make an estimate for the solar radiation in a place somewhere between the measurement sites. This method is used in the original PVGIS solar radiation database for Europe. The methods used in PVGIS are described in references 1&2 (see below).

Solar radiation estimates from satellite

There are a number of methods to estimate the solar radiation at ground level using data from satellites. Typically the satellites measure the light (visible or infrared) coming from the Earth. This light is mainly the light reflected from the ground or from clouds. The calculation of the solar radiation at ground level must therefore be able to take into account the radiation absorbed by the atmosphere as well as that reflected by clouds.

Different types of satellites can be used to estimate solar radiation. Geostationary weather satllites take pictures of the Earth at short intervals (every 15 or 30 minutes) so have a very good time resolution. However, each pixel in the picture typically represents a rectangle a few km on each side, so the estimate of solar radiation for each pixel will be the average of such an area. Polar-orbiting satellites fly closer to the Earth, so the space resolution is better. However, they don't stay permanently above a particular area, so they are normally able to take only a couple of pictures a day of a given area. The data used for PVGIS come mainly from geostationary satellites.

The main advantage of satellite-based methods is that they give a fairly uniform coverage of large areas while ground stations are often very far apart. On the other hand, there are potential problems also with the satellite methods:

• Snow on the ground is a special problem for satellite methods, since snow will look very much like clouds in the satellite images. There are methods to overcome this problem, but the uncertainty is higher in areas with snow.
• In mountain areas one pixel may cover an area with strongly varying altitude. The solar radiation dependence on altitude is not well represented in the satellite-based calculations.
• When the sun is very low in the sky the calculation from satellite data becomes very difficult. This can cause problems, in particular in winter at high latitudes.

Radiation databases for PVGIS

The existing PVGIS databases are based on the following types of data:

Original PVGIS Europe

The original PV database for Europe is based on an interpolation of ground station measurements as described in Ref. 2. The ground station measurement data are long-term monthly average of global and diffuse irradiation on a horizontal plane. The data were originally part of the European Solar Radiation Atlas (Ref. 3). The time period of data is 10 years, from 1981 to 1990.

Original PVGIS Africa

This database is from satellite-based calculations performed at MINES ParisTech, France, using data from the first generation of the Meteosat series of satellites. The spatial resolution of the original calculation is 15 arc-minutes, or about 28km right below the satellite (at the equator, 0° W). The data cover the period 1985-2004.

New CM-SAF - PVGIS database for Europe and Northern Africa

These data are based on calculations from satellite images performed by CM-SAF (see also references 4 and 5). The database represents a total of 12 years of data. From the first generation of Meteosat satellites (Meteosat 5-7), known as MFG, there are data from 1998 to 2005 and from the second-generation Meteosat satellites (known as MSG) there are data from June 2006 to May 2010. The spatial resolution is 1.5 arc-minutes (about 3km right below the satellite at 0° N, 0° W). The coverage extends from 0° N (equator) to 58° N and from 15° W to 35° E.

PVGIS-3 to PVGIS-CMSAF: change in the radiation values from the old databases to the new CM-SAF database

The change in global horizontal irradiation from PVGIS-3 to PVGIS-CMSAF is shown in Fig. 1 (for Europe) and Fig. 2 (for Africa). The differences are in percent between PVGIS-CMSAF and PVGIS-3. Yellow and red means the PVGIS-CMSAF gives higher values, blue means PVGIS-CMSAF gives lower values than the older PVGIS-3. The map only gives a genral overview. To see the difference of the results from PVGIS-3 to PVGIS-CMSAF, you can always try it out for the point you are interested in.

Fig. 1. Relative difference (in percent) between the new CMSAF based database and the traditional PVGIS database for Europe.

Fig. 2. Relative difference (in percent) between the new CMSAF based database and the traditional PVGIS database for Africa.

So which is more accurate?

The new CM-SAF data set has been tested extensively against high-quality measurements on the ground(see documentation at the CM-SAF site). Generally the overall error for the whole year is quite small. A list of ground stations and the yearly error (bias) is shown in the table below. In nearly all places the error is less than 5%. In many places the difference between PVGIS-3 and PVGIS-CMSAF are larger than that. For this reason we are convinced that in most places the new data set is an improvement.

Why is the old data set then wrong?

For Europe, the PVGIS-3 data set is based on measurements made on the ground which are then interpolated between points to get radiation values at any point. As we see in Fig. 1, the old values are generally lower than the new values. The interpolation procedure is not perfect, but it will not give values that are too low or high over large areas such as we see in Fig. 1. So the difference must be in the ground station measurements themselves. There are then two possibilities:

• The measurements were wrong
• The climate has changed so the irradiation is higher for the new data set (1998-2010) than for the old data set (1981-1990)

It is likely that the overall difference is caused by both these effects. When you make radiation measurements there are a number of things that can go wrong. Most of these faults will make the measured value too low. For instance, the sensor can be covered with dirt, snow or frost. There may also be shadows from trees and buildings, and the shadows from trees tend to get larger over the years as the trees grow. Overall, this could lead to too low values in many places. Still, it is unlikely that the effect would be so similar for many stations.

On the other hand, research has shown that the amount of solar radiation has increased over Europe in the last 30 years(see for example Ref. 6). This agrees quite well with the differences between PVGIS-3 and PVGIS-CMSAF.

For the new database for Africa the number of solar radiation measurement stations is very low. We have tried to check both the old and the new database against data from 4 different stations. The results are shown in Table 2. Note that two of these stations are outside the area of the new database as implemented in PVGIS right now, though these areas (Africa south of the Equator, the Arabian Peninsula) will become available in the next few months. From this comparison, we can see that while both databases do reasonably well for the two stations in Africa, the old PVGIS-3 database shows significant underestimation for the two stations in the Middle East. The difference between PVGIS-3 and PVGIS-CMSAF is shown in Fig. 2. The most clear difference is in southwest Sahara where the new PVGIS-CMSAF has significantly higher values than the older PVGIS-3. It is not completely clear why this is the case, and there are no measurement stations in the area. However, we think that it may be due to a problem with the calculation method used for the old data set. The ground in this area is very bright and it may have caused the calculation to mistake the white ground for clouds, so calculating too low radiation values. The same effect may be the reason why the PVGIS-3 values are too low in desert areas of the Middle East.

References

1. Šúri M., Hofierka J., A New GIS-based Solar Radiation Model and Its Application for Photovoltaic Assessments. Transactions in GIS, 8, 2, 175-190 (2004)
2. Šúri M., Huld T.A., Dunlop E.D., PVGIS: a web-based solar radiation database for the calculation of PV potential in Europe. International Journal of Sustainable Energy, 24, 2, 55-67 (2005)
3. Scharmer, Greif (Eds.)
4. Müller R., Matsoukas C., Gratzki A., Behr H.D., Hollmann R., The CM-SAF operational scheme for the satellite based retrieval ofsolar surface irradiance - A LUT badsed eigenvector hybrid approach. Remote Sensing of Environment, 113, 1012-1024 (2009)
5. Ineichen P., Barroso C.S., Geiger B., Hollmann, R., Marsouin A., Müller R., Satellite Application Facilities irradiance products: hourly time step comparison and validation over Europe', International Journal of Remote Sensing, 30, 5549-5571 (2009)
6. Wild M., Global dimming and brightening: A review, Journal of Geophysical Research, 114, D00D16 (2009)