Provider comparison
PVsyst gives access to many weather data sources.
These show that the available weather data are far from being an exact science. There are big discrepancies between these databases, and it is very difficult to estimate which one is the best suited for a given project or location, and what is the probable error.
We have performed a comparison between several sources for different locations to provide a better understanding of how weather data can vary across sources.
Comparisons cannot be made rigorously, because of the variety of conditions:
- Not all sources are available for every spot on the globe. Some of them are for given locations, other ones perform interpolations or are for discrete grids of variable sizes.
- Climate variability: the sources apply for measurements of given years, or averaged periods which differ from one source to another (or even for one location to another, depending on historical measurements availability)
- Measurements: The analysis and validation of ground station data or satellite images involve sophisticated models which are constantly evolving and improving. The methods and techniques may vary from source to source.
For the comparison, we have chosen as reference the annual available irradiation [kWh/m²/year]. This parameter is relevant for PV grid systems, as the PV output is quasi-linear with the solar energy input.
Comparison between several data sources for different continents
The following figures present a comparison of four free weather data sources available in PVsyst across nine locations on different continents. Multiple versions of each data provider are included, highlighting not only the differences between providers but also the variations that can occur between versions from the same provider. Weather data providers continuously update and refine their databases and algorithms, as well as expand the temporal coverage of their datasets.
The graph shows the deviation from the average at each location in percent. The average has been performed over all sources without any weighting. Each site’s climate classification (KGPV) is indicated next to the city name.
Comparison of different data sources for several sites in Europe.
Comparison of different data sources for several sites in Africa.
Comparison of different data sources for several sites in North, Central America and Caribbean.
Comparison of different data sources for several sites in South America.
Comparison of different data sources for several sites in Asia.
Comparison of different data sources for several sites in the South West Pacific.
A few conclusions can be drawn from the graph:
- Different sources typically concur within 10% of the mean value, demonstrating a high degree of agreement in most cases.
- It is challenging to determine which source most accurately represents reality, especially given that no one can reliably predict future climate changes.
- The variability and uncertainties in data from each source exhibit regional dependencies, suggesting that geographic and environmental factors significantly influence measurement discrepancies.
<!--## Changes within the data of the same provider
The providers of weather data keep updating and improving their databases and algorithms. As an example, the following graph shows values for the twelve sites of the European comparison for different Meteonorm Versions. Almost all values stay the same within a few percent, but for Barcelona and Sevilla one sees that significant corrections were performed between V4 and V6, and V6.1 and V7 respectively.
Changes like this can also be observed for other providers like NASA or PVGIS, although for different sites. In the previous plot, the difference between the PVGIS 'classical' values and PVGIS CM SAF is particularly striking. Although these values are based on two different measurement techniques (terrestrial measurements and satellite images), the provider PVGIS attributes part of this difference to a climatic change that took place between the end of the 20th and beginning of the 21st century. For the detailed explanation check this PVGIS page.
In the case of Meteonorm's synthetic data, the intrinsic variability of the generated data leads to a 1% variation in the energy yield and can be neglected.
Comparison for different Meteonorm Versions
Yearly variations and climatic evolution
The following graph is based on a homogeneous sample of continuous measurements from the same source (ISM - Swiss Institute for Meteorology) for Geneva, from 1981 to 2012.
It shows that in Geneva, the annual variation of the global horizontal irradiance stayed well below 5% with only a few exceptions during 20 years. But then the average increased significantly since 2003, staying 10% above the previous value. This is not necessarily valid for other sites in Europe!
Comparison with other sources
These ISM data, which we can consider as reliable due to the fact that they are weather data standard measurements, performed using calibrated pyranometers and corroborated with our own “scientific” measurements in Geneva since 25 years [P. Ineichen], are compared with the Satel-light (satellite) and Helioclim (from terrestrial measurements) data.
We can observe that the Satel-light data overestimate the ISM data by around 5%, while the Helioclim results are more chaotic (over-estimate the good years and underestimate the bad ones). The Helioclim-2 2005 values are 7.5% over the ISM measurements.
Satel-light data
For other sites in Europe, the Satel-light data are always far over the Meteonorm ones, with one exception in Berlin. This exception is not attributable to the Meteonorm value; as we can see on the global comparison plot above, the Satel-light data for Berlin are significantly below the other Satellight data. We don’t have any explanation for that.
Yearly variations for Satellight data