Evaluation of the Heliosat-II method using daily irradiation data for four stations in Iran

Abstract

Efficient use of solar radiation needs detailed knowledge of its spatial and temporal variations. Such information can be achieved using interpolating measured irradiance by ground stations. But more reliable results can be obtained by processing geostationary satellite images. Heliosat is an algorithm which has been developed to estimate global horizontal irradiance at ground level from images taken in the visible band by the Meteosat satellites.

The aim of this study was to evaluate the Heliosat-II model by using daily global solar irradiation data measured at the four radiometric stations in Iran as well as Meteosat-5 images which are recorded by a spacecraft over 63°E. Mean RMSD% and MBD% for all stations were 11.7% and 1.9%, respectively. The mean values of intercept, slope and correlation coefficient were 0.82 (kWh m⁻²), 1.05 and 0.93, respectively. Seasonally, the maximum RMSD occurs in autumn (22.1%) and the minimum is experienced in spring (8.4%). This accuracy is a great achievement for producing a high quality solar radiation atlas in a country such as Iran with very sparse radiometric network and frequently unreliable measured irradiation data.

Introduction

The estimation of fluxes of incident shortwave radiation at the earth’s surface is important for a number of applications including climate monitoring, regional solar energy availability assessment for heating and electrical power generation purposes and for the evaluation of the cloud and radiation parameterization used in weather and climate models (Diak et al., 1996).

The availability of solar radiation data is limited by the sparsity of the existing networks. Zelenka et al. (1999) and Perez et al. (1997) state that estimating solar radiation data from satellite images is more accurate than interpolating data measured by a modern radiometric network. In daily basis, the error resulted from estimating incoming solar radiation by use of satellite images is equal to the error related to interpolating measured data in a radiometric network with 50 km distance between stations (Fig. 1). The error of interpolation increases with the distance between stations; for a radiometric network such as the one existing in Iran with average distance more than 500 km this error gets larger than 50%. In addition, the quality of the insolation fields produced from this network relies on the goodness of calibration, regular maintenance and the persistence of the functioning of all instruments. These requisites are not always met over the entire network.

On the other hand, there exist several methods for determining solar radiation at ground level from images of geostationary satellites. Furthermore, as reported by Rigollier and Wald (1998), the estimates made from satellite data always offer relatively equal quality within the observed area. It has been shown that the accuracy of the estimates does not depend on the geographical area, except for highly variable mountainous areas. In these complex mountainous areas, the quality of the results depends strongly on the spatial resolution of the satellite data: the better the resolution, the more accurate the estimates (Zelenka, 1994).

Models capable of exploiting satellite observations to generate irradiances range from physically vigorous to purely empirical. At one end, physical models (e.g. Raschke and Preuss, 1979, Gautier et al., 1980, Schillings et al., 2004a, Schillings et al., 2004b) attempt to explain the observed earth radiances (the brightness seen by the satellite) by solving radiation transfer equations which requires absolute satellite calibration knowledge and precise information on the composition of the atmosphere. At the other end, empirical/statistical models (Trapely, 1979, Cano et al., 1986, Fontoynont et al., 1998) may consist of a simple regression between satellite counts and corresponding measurements at the earth’s surface. Hybrid models (Rigollier et al., 2004) use a simple physical modeling approach and some degree of fitting to observations (Perez et al., 2002).

The first attempts to estimate solar radiation from satellite images are found in the works of US Department of Commerce (Hanson, 1971), Vonder Haar, 1973, Vonder Haar and Ellis, 1975, Vonder Haar and Ellis, 1978, Trapely, 1979 and Moser and Raschke (1983). The basic procedure of the Heliosat method was developed by Cano (1982) and Cano et al. (1986), and partly modified by several researchers (Diabate et al., 1988, Diabate et al., 1989, Moussu et al., 1989, Beyer et al., 1996, Hammer et al., 1997, Fontoynont et al., 1998, Rigollier and Wald, 1998; Rigollier et al., 2004). This method was developed to derive solar irradiance maps from Meteosat images. The basic idea of the Heliosat method is that the amount of global solar radiation over an area is statistically related to the cloud cover. The method relies on the prior determination of a reference albedo map and is statistical in nature and avoids the problem of satellite data calibration. In addition to the albedo, the final results are computed according to the linear relationship between the hourly transmission factor measured at the ground and the cloud cover index computed from the Meteosat data.

Diabate et al. (1989) have used the Heliosat version developed by Diabate et al. (1988) and Moussu et al. (1989) for on-line mapping of global solar radiation from the satellite data directly received at ground by a HF receiver. They used three hourly images to compute daily mean map of solar radiation. Rigollier and Wald (1998) improved the Heliosat method to produce a climatological database of solar radiation over Europe, Africa, and Atlantic Ocean based on the satellite observations.

Rigollier et al. (2004) introduced a new version of the Heliosat method entitled “Heliosat-II”. This new version integrates the knowledge gained by various exploitations of the original Heliosat method and its variants in a coherent and thorough way. It is based on the same physical principles but the input to the method, instead of the digital output from the sensor, is the calibrated radiances. This takes into account the changes of the sensor and removes the need for empirically defined parameters and pyranometric measurements to tune them. Fore more information about the calibration of the Meteosat images refer to Lefevre et al., 2000, Rigollier et al., 2002, Govaerts et al., 2004, EUMETSAT, 2005a, EUMETSAT, 2005b. In the new version of the Heliosat method, the assessment of the ground albedo and the cloud albedo is based on the explicit formulations of the path radiance and the transmittance of the atmosphere. For daily values, RMSDs computed from the measured data of 35 European stations were 20%, 16% and 10% for January, April and July, respectively. They noted that, the use of a block of 3 by 3 pixels would lead to a decrease of the standard deviation by 3, thus decreasing the above RMSDs to 8%, 7% and 4%, respectively. As shown in Table 1, the error of the Heliosat method is far less than the other methods.

All these studies confirm the credibility of the Heliosat-II method for estimating the solar irradiation from satellite images. Therefore, the aim of this study is to evaluate the Heliosat-II method by using the measured irradiation data of four Iranian radiometric stations and the high resolution visible images of Meteosat-5 located over 63°E and equator.