Solar irradiance site adaptation

Overview

Solar irradiance retrieved from satellite-based models has lower spatial and temporal resolution compared to on-site measurements. While satellite data is excellent for long-term consistency, it may not fully capture local microclimates or specific atmospheric conditions at a specific point location.

Site adaptation methods adapt satellite-derived Global Horizontal Irradiance (GHI) and Direct Normal Irradiance (DNI) datasets (and derived parameters) to local conditions that are not fully recorded in the original satellite and atmospheric inputs. The Solargis site adaptation process pursues two objectives:

• Improvement of the overall bias (removal of systematic deviations)

• Improvement of the fit of the frequency distribution of values.

Two main approaches are typically employed. The first involves the adaptation of satellite-based GHI and DNI values (indirectly also adapting Global Tilted Irradiance - GTI), correcting the bias and aligning the cumulative distribution functions. The second focuses on adapting the input parameters of the solar radiation model, where more complex parameters such as Aerosol Optical Depth and/or Cloud Index are adjusted.

This adaptation is possible when measured data meet certain requirements regarding length and quality. The outcome is a multi-year solar dataset with improved accuracy and decreased uncertainty.

Site adaptation challenges

Spatial mismatch between satellite and ground data

A satellite radiometer integrates signal over an area of several square kilometres, while a ground station measures at a single point. In complex regions, such as narrow valleys with frequent fog or coastal areas, a single satellite pixel may represent a mix of conditions that differ from what the ground station, located at a specific point, actually experiences.

This mismatch - known as the nugget effect - accounts for nearly half of the hourly Root Mean Square Deviation (RMSD) for GHI and DNI, and is most pronounced during intermittent cloud cover and changing aerosol conditions. Coarse spatial resolution in atmospheric databases, such as aerosol and water vapour inputs, further limits the ability to capture local atmospheric patterns.

DNI sensitivity

DNI is particularly sensitive to variability in cloud cover, aerosols, water vapour, and terrain shading. The relationship between GHI and DNI uncertainty is nonlinear and a small error in GHI can correspond to a significantly larger error in DNI.

Site adaptation methodology

Data preparation

Time resolution - Solargis satellite data is available in 15-minute time steps, while ground measurements are typically available in time steps between 1-minute and 60-minute. To reduce the effect of time resolution differences and the conceptual differences between point and pixel measurements, all metrics are calculated on aggregated hourly data.

Exclusion of extreme events - The methodology avoids matching the model to extreme, non-prevailing cases such as exceptional dust storms or volcanic ash outbreaks. This ensures that short-term, rare anomalies do not skew the long-term adapted results.

Accuracy metrics

Adaptation parameters are derived from the overlapping period of ground measurements and modelled values, assuming that systematic differences between the two are stable over a one- to two-year period. The following metrics are used to assess the improvement due to site adaptation:

Mean Bias and RMSD: Calculated from all hourly daytime data pairs, in absolute and relative form (divided by the daytime mean GHI value).

KSI (Kolmogorov-Smirnov test Integral): Quantifies differences between the cumulative distribution functions of two datasets. The normalized KSI is defined as:

where is the number of data pairs.

Since KSI depends on sample size, it is only valid for the relative comparison of cumulative irradiance distribution fits.

Adaptation of satellite-based GHI and DNI values

This approach is effective when deviations between ground-measured and satellite-derived data are small and stable, but becomes unreliable when significant inconsistencies exist. Two methods are used:

Ratio method

Adjusts long-term monthly and annual averages of DNI and GHI by correcting systematic bias. Ratios between ground-measured and satellite-derived solar radiation are calculated for the overlap period and applied to recalibrate the long-term satellite dataset. This method is simple to implement but only addresses mean bias, does not account for distribution differences, and does not fully utilize the information from ground measurements.

Fitting cumulative distribution function

Aligns the cumulative distribution functions (CDF) of satellite-derived and ground-measured data, improving the representation of both typical and extreme values. The mean value of the satellite data is also adjusted to match the ground-measured mean. This approach typically reduces RMSD and enhances the accuracy of satellite-based data for local applications.

Note: Site adaptation methods must be applied carefully. Inappropriate application to non-systematic deviations, or the use of less accurate ground data, can degrade rather than improve the primary satellite-derived dataset.

Adaptation of input parameters of solar radiation model

This approach adapts the input parameters of solar radiation models to improve the accuracy of GHI, DIF, and DNI outputs. Two methods are employed: Adaptation of clearness index and Adaptation of model input data.

Adaptation of clearness index

The clearness index (the ratio of surface radiation to top-of-atmosphere radiation) is used as the key model input. It is derived from ground measurements using a solar radiation model, then calibrated using least-square regression to align hourly or daily differences between ground-measured and satellite-based parameters. The site-adapted clearness index is then used to recalculate GHI, DIF, and DNI. This method effectively reduces mean bias, RMSD, and KSI, but its use of a single parameter limits its capacity to handle complex atmospheric effects.

Adaptation of model input data

Substitutes the clearness index with more detailed parameters such as Aerosol Optical Depth (AOD) or Cloud Index, and recalculates the entire model to ensure consistency among GHI, DNI, and DIF components. By addressing seasonal and regional inaccuracies in AOD and cloud descriptions, this method reduces mean bias to within the expected uncertainty of measurement instruments, while also minimizing RMSD and KSI. It is particularly valuable in regions where aerosols or cloud cover significantly affect solar radiation, such as semi-arid and desert areas.

Requirements for measured data

The capability to perform site adaptation depends on the quality and duration of the ground measurements.

Measurement duration

The following table summarizes the suitability of different measurement periods:

Usage in the Solargis platform

Site adaptation is an essential part of the Solargis Evaluate application to simulate the data. Additionally, we use it when generating TMY and TS data available via API and as a part of the consultancy services.

Further reading

• "Site adaptation of satellite-based DNI and GHI time series: Overview and SolarGIS approach": Cebecauer, T.; Suri, M. AIP Conf. Proc. 1734, 150002 (2016).

• "Analysis of different comparison parameters applied to solar radiation data from satellite and German radiometric stations": Espinar, B.; Ramírez, L.; Drews, A.; Beyer, H.G.; Zarzalejo, L.F.; Polo, J.; Martín, L. Solar Energy, 83(1), 118–125 (2009).

• "Effective accuracy of satellite-derived hourly irradiances": Zelenka, A.; Perez, R.; Seals, R.; Renne, D. Theoretical and Applied Climatology, 62, 199–207 (1999).