The method Heliosat-2 for deriving shortwave solar radiation from satellite images
Introduction
Several authors have shown the potentialities of the images of the Earth taken by the meteorological satellites for the mapping of the global irradiation impinging on a horizontal surface at the ground level. Zelenka et al., 1992, Zelenka et al., 1999 and Perez et al. (1997) demonstrate that for the best methods and for pixels of size of 10 km or so, the irradiation assessed by satellite is better than that estimated by the means of an interpolation technique applied to measurements performed at meteorological stations as soon as the distance to the stations is greater than 34 km for the hourly irradiation and 50 km for the daily irradiation.
Many methods were developed especially in the '80s. Among them was the method called Heliosat-1 (Cano, 1982; Cano et al., 1986; Michaud-Regas, 1986; Diabaté, 1989), one the most accurate as shown by several authors (Grüter et al., 1986; Raschke et al., 1991). It was simple enough to widely disseminate in the world (Diabaté et al., 1988a, Diabaté et al., 1988b, Diabaté, 1989; Wald et al., 1992). It became popular and several modifications were proposed (Moussu et al., 1989; Obrecht, 1990; Zelenka et al., 1992, Zelenka et al., 1999; Beyer et al., 1996; Fontoynont et al., 1997; Iehlé et al., 1997; Ineichen and Perez, 1999).
There are several empirical parameters in the method Heliosat-1, especially in the computation of the apparent albedoes of the ground and clouds and normalization of the digital counts. The relationship between the cloud index and the clearness index is empirically defined and its parameters are computed by the means of a comparison between the cloud index and measurements made by meteorological stations in the area under concern. All these parameters were well tuned during the construction of the method or of its varieties using ground-based measurements and this explains the good results attained by the authors. Table 1 gives the root mean square error (RMSE) reported by authors having developed a variety of the Heliosat-1 method and using measurements for tuning parameters (Rigollier, 2000). In all cases, adjustment is made with a null bias. The errors are obtained by subtracting satellite-derived assessments from ground measurements performed in the meteorological network at a coincident location and coincident time.
However, when applied to other areas, or other periods, the accuracy, expressed as bias and RMSE in percentage of the mean irradiation, is usually lower than claimed by the inventors (Table 2). For example, one may note that the biases are not negligible in many cases. One may also remark in the first two rows that the biases found by Diabaté for two different periods over the same area differ largely: −7% in 1983 and −1% for 1984 and 1985. Using the same parameters than Diabaté, Obrecht and Raschke found biases in monthly means of daily irradiation that range from −8% to −29% for various parts of Sahel. This table clearly shows that modifications of the method Heliosat-1 are necessary to ensure that any correct implementation should lead to similar performances.
The purpose of the present paper is to present a new version, called Heliosat-2, which integrates the knowledge gained by these various exploitations of the original method and its varieties in a coherent and thorough way. The various empirical parameters present in the method Heliosat-1 are now expressed using physical laws and there is no need for coincident pyranometric measurements to tune these parameters. The major motivations for creating this new version were to improve the capabilities of the method to process any type of data taken in the broadband visible range by geostationary meteorological satellites, including large time-series of images taken by different sensors, and to improve the implementation of the method by reducing the number of empirical parameters.
Section snippets
The new method Heliosat-2
Given the good fundamentals of the method Heliosat-1, it was decided to keep it principle, that is the construction of a “cloud index” resulting from a comparison of what is observed by the sensor to what should be observed over that pixel if the sky were clear, which is related to the “clearness” of the atmosphere. Actually, this principle is commonly adopted when the only inputs are images taken in the visible broad range (Pastre, 1981; Möser and Raschke, 1983, Möser and Raschke, 1984; Cano
Comparison between retrieved values and station measurements
Similarly to previous works, hourly and daily irradiations derived from satellite are compared to measurements performed at ground level by pyranometers in meteorological stations (Table 3). The 35 stations were selected in flat areas, in order to avoid the specific errors encountered in mountainous areas. We only used measurements of hourly irradiation greater than 10 W h m−2, a value typical of the diffuse hourly irradiation for the sunset and sunrise under clear-sky at 60°N. For these hours of
Conclusion
The method Heliosat-2 meets the objectives set up before its development. It is more physically sound than the previous one. All parameters that needed to be tuned for each implementation of the method Heliosat-1 have been removed, set up to constant values, or automatically determined. No ground measurement is used for the development, contrary to the method Heliosat-1 and others. This should ensure a worldwide application of the method Heliosat-2.
The first comparison with ground measurements
Acknowledgements
We are grateful to Hans-Georg Beyer, Dominique Dumortier, Gerhard Gesell, Annette Hammer, Detlev Heinemann, Pierre Ineichen, John Page, Christian Perrin de Brichambaut, Richard Perez, Corrado Ratto, Christian Reise, Antoine Zelenka for the numerous discussions we had on the method Heliosat and its various improvements. We thank the two referees for their help in clarifying the content of the paper. The meteorological offices from France, Germany, Hungary, Spain, The Netherlands and United
References (51)
- et al.
Modifications of the Heliosat procedure for irradiance estimates from satellite images
Solar Energy
(1996) - et al.
A method for the determination of the global solar radiation from meteorological satellite data
Solar Energy
(1986) - et al.
Quick use of Wefax images from Meteosat to determine daily solar radiation in France
Solar Energy
(1992) - et al.
A modification of the Heliosat method to improve its performance
Solar Energy
(1999) - et al.
A web service for controlling the quality of measurements of global solar irradiation
Solar Energy
(2002) - et al.
Comparing satellite remote sensing and ground network measurements for the production of site/time specific irradiance data
Solar Energy
(1997) - et al.
A new operational model for satellite-derived irradiances: description and validation
Solar Energy
(2002) - et al.
On the clear sky model of the 4th European Solar Radiation Atlas with respect to the Heliosat method
Solar Energy
(2000) - Angles, J., Menard, L., Bauer, O., Rigollier, C., Wald, L., 1999. A climatological database of the Linke turbidity...
- et al.
A Web server for accessing a database on solar radiation parameters
Utilization of satellite data for the assessment of large scale PV grid integration
Estimating incident solar radiation at the surface from images of the Earth transmitted by geostationary satellites: the Heliosat Project
International Journal of Solar Energy
Derivation of cloud index from geostationary satellites and application to the production of solar irradiance and daylight illuminance data
Theoretical and Applied Climatology
Cited by (405)
QIENet: Quantitative irradiance estimation network using recurrent neural network based on satellite remote sensing data
2024, International Journal of Applied Earth Observation and GeoinformationComposition of feature selection techniques for improving the global horizontal irradiance estimation via machine learning models
2024, Thermal Science and Engineering ProgressIntra-day solar irradiation forecast using machine learning with satellite data
2023, Sustainable Energy, Grids and NetworksSPARTA: Solar parameterization for the radiative transfer of the cloudless atmosphere
2023, Renewable and Sustainable Energy Reviews
- 1
Christelle Rigollier is now with Silogic, Toulouse, France.