Validation of solar irradiance data

Overview

Solar irradiance is the primary determinant of the energy available to PV systems, with Global Horizontal Irradiance (GHI) and Direct Normal Irradiance (DNI) serving as its key components. Both play a critical role in photovoltaic (PV) simulations for the calculation of total incident energy on the power plant and simulation of system performance.

Accurate GHI and DNI data are essential for PV simulations as they determine the solar energy available for power generation, directly influencing energy yield predictions and system performance assessments. Reliable solar irradiance data ensures precise modeling of site-specific solar conditions, reducing uncertainties in system design, energy production estimates, and financial modeling.

To validate the accuracy of the Solargis satellite model data, sub-hourly values from the model have been compared with ground-measured data from high-quality, publicly available stations worldwide.

The ground reference data are sourced exclusively from high-accuracy instruments, which undergo a rigorous quality assessment to ensure their uncertainties remain within the tolerance limits of the instruments. This process ensures that the validation comparisons are reliable and robust.

GHI validation statistics

GHI represents the total solar radiation received per unit area on a horizontal surface. As the most widely used parameter in PV simulations, GHI is critical for solar power plants. It serves as the baseline measure of the solar resource available at a site, directly influencing energy yield predictions.

The table below summarizes the accuracy statistics of Solargis GHI data, validated against high-quality ground measurements from over 300 sites representing diverse climate conditions worldwide. You can learn more about how different statistics are calculated in this other article.

A mean bias of 0.5% for GHI indicates near-perfect agreement between satellite-based GHI data and ground measurements, showcasing the model's ability to accurately reflect total solar irradiance across diverse regions.

A standard deviation of 3.0% for GHI bias values signifies minimal variability, reinforcing the dataset's reliability and consistency.

A more informative way to present the statistics is by grouping the reference sites according to climate zones.

*Only one site available for this climate zone.

On the other hand, the calculation of the Root Mean Square Deviation (RMSD) indicates consistent model performance, with values decreasing as the data is aggregated over longer time scales. In other words, RMSD is higher for hourly values compared to daily values, and higher for daily values compared to monthly ones. This behavior is expected for satellite-based models and can be explained by the differing nature of the data: satellite imagery provides averages over areas of several square kilometers, whereas pyranometers and pyrheliometers record measurements at a single point.

It is important to examine both sets of statistics: Bias and RMSD. While Bias identifies systematic deviations between model and measurements, RMSD captures non-systematic deviations that may otherwise be hidden in Bias due to cancellation effects. You can learn more about how different statistics are calculated in this other article.

*Only one site available for this climate zone.

The map below shows the sites at which the Solargis GHI time series was validated against the ground-measured data. By clicking on a site, its details can be displayed, including the basic site characteristics, and the validation statistics - bias, Root Mean Square Deviation (RMSD), and the number of valid data pairs.

DNI validation statistics

DNI measures the solar radiation received per unit area on a surface perpendicular to the sun’s rays, excluding diffuse radiation. Accurate and precise DNI data is essential for PV simulations, as it, together with GHI, enables the calculation of GTI. It is a specially critical parameter for specific PV technologies and applications that depend heavily on direct sunlight like CPV and CSP.

The table below shows the summary of the accuracy statistics of Solargis DNI data compared to high-quality ground measurements at more than 200 sites across all types of climates:

A mean bias of 2.2% for DNI shows good agreement between satellite-based DNI data and ground measurements, highlighting the model's reliability in capturing direct solar irradiance.

The 6.0% standard deviation for DNI bias values reflects moderate variability across sites, likely due to regional atmospheric complexities such as aerosols or cloud cover.

*Only one site available for this climate zone

For DNI, the calculation of the Root Mean Square Deviation (RMSD) indicates consistent model performance as well, with values decreasing as the data is aggregated over longer time scales.

*Only one site available for this climate zone.

The map below shows the sites at which the Solargis DNI time series was validated against the ground-measured data. By clicking on a site, its details can be displayed, including the basic site characteristics, and the validation statistics - bias, Root Mean Square Deviation (RMSD), and the number of valid data pairs.

Conclusions

• The validation exercise confirms that satellite-based GHI and DNI data are accurate and reliable for use in PV simulations, providing confidence in their ability to support the design, simulation, and optimization of PV systems.

• With validation conducted at over 330 sites globally, the reference data demonstrates very good geographic and climatic coverage, enhancing the reliability and applicability of the results across diverse regions and conditions. The consistency across varying climates and regions ensures confidence in the data's suitability for PV system design and energy yield predictions worldwide.

• The consistent RMSD (Root Mean Square Deviation) for hourly, daily, and monthly values reinforces the reliability of satellite-based data across temporal resolutions, supporting both short-term operational assessments and long-term energy yield predictions.

• In the tropical zone, deviations are driven by the frequent presence of broken clouds and the model’s limited ability to accurately determine their optical properties from satellite data.

• In the arid zone, the dominant factor is the correct representation of aerosols, which strongly influence solar radiation.

• In the temperate and cold zones, both cloud and aerosol effects act together, making model performance more complex to assess.

• In the cold and polar zones, further difficulties arise from extreme sun–satellite angles, distortions of the satellite grid cells, and the presence of snow cover, all of which hinder accurate retrieval of cloud properties.

Example site data

For several of our validation sites in different climates, we provide the original ground-measured data, link to Solargis satellite model time series, and a report with quality control results and the validation statistics. You can find this data below, but also tagged to each of the validation sites in the map above.

Detailed report for further analysis

You can examine and analyze the validation further in the detailed World Bank report. Please note that due to rounding, there may be small differences in the data displayed on the map and the data presented in the report.