Satellite-based irradiance

Overview

Solar irradiance models must account for the various atmospheric processes that attenuate solar radiation as it travels from space to the Earth's surface. These processes include the effects of water vapor, aerosols, and clouds.

To obtain historical and recent solar irradiance data, Solargis employs a semi-empirical solar radiation model. This model is termed "semi-empirical" because it integrates satellite data to detect cloud properties through advanced algorithms while incorporating key physical parameters.

Semi-empirical models offer reliable results and effectively replicate real-world conditions. They strike a balance between fully physical and empirical models. Fully physical models attempt to explain Earth’s radiance by solving radiative-transfer equations, but they require detailed knowledge of the atmosphere's composition, which can be a limitation. Empirical models, by contrast, rely on simple regressions between satellite-visible channel data and ground station measurements, often lacking the precision required for solar energy applications.

The Solargis semi-empirical satellite-based model for solar irradiance consists of two main components:

• clear-sky model

• cloud model

Together, these models provide highly precise and accurate estimates of non-shaded global solar irradiance on a horizontal surface for specific locations.

The final output of the model is all-sky irradiance, which accounts for all atmospheric factors, including cloud cover. However, the model calculates non-shaded irradiance, meaning it does not consider shadows from the horizon, nearby objects, or the plant’s own infrastructure. These shading effects are applied later when calculating the incident irradiance on the power plant's surface, referred to as Global Tilted Irradiance (GTI) or irradiation.

All-sky irradiance corresponds to global irradiance, which must later be divided into its two main solar irradiance components: direct and diffuse irradiance. In addition, the calculation of GTI will incorporate contributions from ground-reflected irradiance, which requires evaluating the ground albedo at the power plant’s location.

The result of semi-empirical models can also be used in energy forecasting applications by means of cloud motion vector (CMV) models. These models estimate short-term cloud movement to predict solar irradiance for the next few hours.

Clear sky model

Clear-sky irradiance represents the solar energy reaching the Earth’s surface at a specific location and time, assuming no clouds are present. In other words, it is the theoretical maximum irradiance, upon which the actual reduction due to cloud cover is superimposed.

Cloud model

Satellite data from several geostationary satellites is used to quantify cloud attenuation through cloud index calculations. This process relies on high-resolution imagery and radiometric data from multiple spectral bands provided by radiometer instruments on geostationary satellites covering different regions of the globe.

Since 1994, radiometers on satellites like Meteosat, GOES, and Himawari have advanced significantly:

Enhancement of spatial resolution

The spatial resolution of the satellite data used—Meteosat, GOES, and MTSAT—is approximately 3 km at the sub-satellite point (see table below for further details). However, actual pixel resolution may vary depending on the time period and the site’s latitude, with higher latitudes experiencing more pixel distortion due to their distance from the equator.

Using resampling techniques and digital terrain models, the spatial resolution of data products can be enhanced to as high as 3 arc-seconds, which corresponds to about 90 meters at the equator, with slightly lower resolution as you move towards the poles.

Cloud detection in high surface albedo areas

In areas with high surface albedo, such as snow-covered regions, salt flats, or white-sand areas, the cloud index is derived from infrared channels in satellite data.

Additionally, snow depth and air temperature data from Numerical Weather Prediction (NWP) models are incorporated to classify each data point into one of the following categories: snow-covered, snow-free land, water, cloud, or unclassified.

Occasionally, snow may not be detected, particularly in winter months, which can lead to higher model uncertainty in snow-prone regions.

Model update frequency

Final atmospheric data are not available at the time of solar radiation calculation. After some time, usually with a delay of 12 hours, we receive updated, final atmospheric data inputs. When we receive updated atmospheric data, we recalculate solar radiation data using the updated inputs.

The typical timeline of calculation of solar radiation data is as follows:

1. Every day, solar radiation data is calculated for DAY-1 and DAY-2.

2. At the beginning of each month, solar radiation data for the previous month is re-calculated using final atmospheric data inputs. Atmospheric data are homogenized with historical data records to avoid abrupt changes due to changes in atmospheric models.

Further reading

• SPARTA: Solar parameterization for the radiative transfer of the cloudless atmosphere. José A. Ruiz-Arias, 2023, Renewable and Sustainable Energy Reviews, vol. 188, p. 113833, Dec. 2023.