Report. Reference. Helioclim 3 and Meteonorm 6 short wave irradiance validation over Africa. INEICHEN, Pierre

Transcription

1 Report Helioclim and Meteonorm short wave irradiance validation over Africa INEICHEN, Pierre Abstract Downward short wave incoming irradiances play a key role in the radiation budget at the earth surface. The monitoring of this parameter is essential for the understanding of the basic mechanisms involved in climate change, such as the greenhouse effect, the global dimming, the change in cloud cover and precipitations, etc. The use of geostationary satellite observations becomes crucial, since they allow the retrieval of irradiance at the surface, with the best spatial and temporal coverage. This study presents a common validation of two radiation products (Helioclim and Eumetsat Ocean and Sea Ice Facility) and one software product (Meteonorm ) against ground data from 9 stations covering up to one year of measurements over Europe and Africa. The overall conclusion is that the products are comparable in terms of bias and standard deviation, with lower bias and standard deviation over Europe. The surface solar irradiance is retrieved over the African continent with an average standard deviation of % and a negligible bias. If the atmospheric aerosol load becomes a better known input parameter, it remains sparlsy [...] Reference INEICHEN, Pierre. Helioclim and Meteonorm short wave irradiance validation over Africa. Geneva :, 7 p. Available at: Disclaimer: layout of this document may differ from the published version.

2 Helioclim and Meteonorm short wave irradiance validation over Africa Pierre Ineichen University of Geneva March Abstract Downward short wave incoming irradiances play a key role in the radiation budget at the earth surface. The monitoring of this parameter is essential for the understanding of the basic mechanisms involved in climate change, such as the greenhouse effect, the global dimming, the change in cloud cover and precipitations, etc. The use of geostationary satellite observations becomes crucial, since they allow the retrieval of irradiance at the surface, with the best spatial and temporal coverage. This study presents a common validation of two radiation products (Helioclim and Eumetsat Ocean and Sea Ice Facility) and one software product (Meteonorm ) against ground data from 9 stations covering up to one year of measurements over Europe and Africa. The overall conclusion is that the products are comparable in terms of bias and standard deviation, with lower bias and standard deviation over Europe. The surface solar irradiance is retrieved over the African continent with an average standard deviation of % and a negligible bias. If the atmospheric aerosol load becomes a better known input parameter, it remains sparlsy acquired over the African continent. On the other hand, the present validation shows that the atmospheric water vapor, retrieved from ground temperature and relative humidity, has a non-negligible effect on the results and should be better evaluated.

3

4 . Introduction Anthropogenic activities became an important factor in the climate change and a continuous monitoring of the solar irradiance reaching the ground is essential to understand the impact of such changes on the environment (Cutforth 7, Stanhill ). Unfortunately, the density of the ground measurement network is insufficient, especially on continents like Africa, or countries in the Near East. To circumvent this lack of measured data, the meteorological satellites are of great help and models converting the satellite images into the different radiation components become increasingly performing. If these converting models are well validated over the United States and Europe, it is not the case over the African continent. The present study will evaluate the global irradiance produced from satellite images by the Soda service ( and by the Ocean and Sea Ice Satellite Application Facilities (OSI-SAF) created by Eumetsat in 999, and based on cooperation between several offices, hosted by a national meteorological service ( The Meteonorm software is also evaluated, even if the production scheme is different (part of the data are also produced from satellite images).. Ground data To do the comparison and the validation, data from 9 ground stations situated in Spain, Israel and part of the African continent are used. Unfortunately, ground data for the African continent are very difficult to obtain, only a few regions are covered. Due to the lack of data, the quality control is also hard to perform, and neighbouring stations are used to verify the absolute calibration of the sensors. The work is done on data covering up to one year (here ) of data depending on the sites. Only a few of the ground sites acquire other data than the SSI. When the beam (DNI) or the diffuse component is available, these are used for the quality control. The water vapor w is used as input to clear sky models, it is evaluated from the ambiant temperature (T a ) and relative humidity (HR) by the use of Atwater model (Atwater 97); when the temperature and humidity measurements are missing, the data from a neighbouring station (if applicable) are used, the water vapor from the aeronet network, or at the last possibility, monthly average climatic data retrieved from Meteonorm. (Meteonorm 9) and Helioclim (Helioclim 9) are used. The climate, latitude, longitude and altitude of the stations, their corresponding data availability are given in Table I. - -

5 Station climate latitude longitude altitude m operated by Europe & Northern Africa El Saler (Spain) semi arid, warm summer FluxNet Geneva (CH) semi-continental.. CIE Las Majadas (Spain) semi arid, warm summer FluxNet Sede Boqer (Israel) dry steppe BSRN Tamanrasset (Algerie) hot, dry desert.78. BSRN Val Alinya (Spain) warm temperate, humid.. 77 FluxNet Yatir Forest (Israel) hot arid.. FluxNet Africa Agoufou (Mali) hot, arid desert. -.8 AMMA Bamba (Mali) hot, arid desert AMMA Banizoumbou (Niger) hot arid.. AMMA Bira (Benin) Winter dry, equatorial AMMA Djougou (Benin) Winter dry, equatorial AMMA Hedgerit (Mali) hot, arid desert. -.9 AMMA Kelema (Mali) hot, arid desert AMMA M'bour (Senegal) hot steppe arid AMMA Mt Kenya Warm humid GAW Nalohou (Benin) Winter dry, equatorial AMMA Tchizaloumou (Congo) Winter dry equatorial -.9. FluxNet Wankama (Niger) hot steppe, arid.. 8 AMMA AMMA BSRN CIE FluxNet Gaw African Monsoon Multidisciplinary Analyses Baseline Surface Radiation Network Commission Internationale pour l'eclairage Flux Network Global Atmosphere Watch parameters Station SSI DNI DLI w aod Agoufou (Mali) x x x Bamba (Mali) x x Banizoumbou (Niger) x x x Bira (Benin) x x Djougou (Benin) x x x El Saler (Spain) x x Geneva (Switzerland) x x x Hedgerit (Mali) x x Kelema (Mali) x x x Las Majadas (Spain) x x M'bour (Senegal) x x x Mt Kenya (kenya) x x Nalohou (Benin) x x x Sede Boqer (Israel) x x x x x Tamanrasset (Algeria) x x x x Tchizaloumou (Congo) x x Val Alinya (Spain) x x Wankama (Niger) x x x Yatir Forest (Israel) x x water vapor from Djougou full month aerosol optical depth from Dakar station partial month partial values no data water vapor from Banizoumbou Table I List of the stations, data and parameters availability. - -

6 . Satellite products Products from two services and one software will be compared and validated against ground measurements: - The SoDa Service offers a one-stop access to a large set of information relating to solar radiation and its use. It builds links to other services (also called resources) that are located in various countries. To answer a request, the SoDa service invokes several resources to elaborate the appropriate answer and ensures the flow and exchange of information between the services and itself, as well as with the customer. The global irradiance is one of the products offered by the service in the form of the Heliclim data base. - The Ocean and Sea Ice Satellite Application Facility (OSI-SAF) is an answer to the common requirements of meteorology and oceanography for a comprehensive information on the ocean-atmosphere interface. One of the objectives of the osi- SAF is to produce, control and distribute operationally in near real-time products using available satellite data with the necessary users support activities. Meteo- France is hosting the osi-saf ( - Meteonorm is a comprehensive meteorological reference, incorporating a catalogue of meteorological data and calculation procedures for solar applications and system design at any desired location in the world. It is based on over years of experience in the development of meteorological databases for energy applications. Meteonorm addresses engineers, architects, teachers, planners and anyone interested in solar energy and climatology. Figure a The global horizontal and normal beam irradiences are represented versus the sinus of the solar elevation for one clear day. Figure b The global clearness index Kt is reprsented separately for the morning and the afternoon data, versus the solar elevation angle for one year in hourly values. - -

7 . Data quality control For all the stations, the first quality control consist of an assessment of the acquisition time. To point out a possible time shift in the data, the symmetry in solar time of the irradiance for clear days is visually checked. The horizontal global and if available, the normal beam irradiances are plotted versus the sinus of the solar elevation angle, and the morning and afternoon curves should be mingled as visualized on Figure a. If this test is positive, a verification can be done with the help of the global clearness index K t defined as: SSI K t = = I sin(h) o SSI SSI TOA () where I o is the solar constant, h the solar elevation angle and SSI TOA, the solar surface irradiance at the top of the atmosphere. The clearness index is plotted for the morning and the afternoon data separately. The upper limit, representative of clear sky conditions, should also be mingled for the morning and the afternoon data as represented on Figure b for one year of data acquired at M Bour (Senegal). When these two conditions are fulfilled, the solar geometry can be precisely calculated. As mentioned above, the ground measurements stations are situated in only a few regions, and some of them very near one from the other, at the same altitude and similar climatic conditions. This is used to assess the calibration of the sensors. Three groups of sites can be used for this quality control: - Bira-Djougou-Nalohou are situated in a triangle in the Benin, and approximately at km one from the other, - Kelema-Agoufou-Hedgerit (Mali) are aligned with about km from Kelema to Hedgerit, - Banizoumbou-Wankama (Niger) are at about km of distance. It is then possible to assume that the average surface solar irradiance should be comparable in a absolute value. Figure is an illustration af the raw SSI measurements for the Bira-Djougou-Nalohou triangle. It can be seen on the figure that the calibration of the sensors is clearly not the same from one site to the other (upper left graph for Djougou and Nalohou), with higher values for Djougou and lower values for Nalohou. To assess the absolute value of calibration, aerosol optical depth (aod) measurements are used. These are done at the Agoufou, Djougou, Banizoumbou, Dakar (M Bour), Tamanrasset and Sede Boqer sites within the aeronet network. The aod measurements are acquired every ~ minutes when direct sun is available; the values are then averaged to give a daily value and used with the Solis clear sky model (Müller, Ineichen 8) to evaluate hourly clear sky SSI. These are reported on the three other graphs on Figure. A correction factor is then applied on each data sets to make them coherent with the clear sky values. This is illustrated on Figure for the same stations. Following - -

8 Solar surface irradiance hourly value [Wh/m h] Solar surface irradiance hourly value [Wh/m h] 8 8 Djougou Nalohou Clear sky (aeronet Djougou) Bira Solar surface irradiance hourly value [Wh/m h] Solar surface irradiance hourly value [Wh/m h] 8 8 Clear sky (aeronet Djougou) Nalohou Clear sky (aeronet Djougou) Djougou Figure Surface solar irradiance reported versus the day of the year for the three neighbouring stations Djougou, Bira and Nalohou in the Benin Solar surface irradiance hourly value [Wh/m h] Solar surface irradiance hourly value [Wh/m h] 8 8 Djougou Nalohou Clear sky (aeronet Djougou) Bira Solar surface irradiance hourly value [Wh/m h] Solar surface irradiance hourly value [Wh/m h] 8 8 Clear sky (aeronet Djougou) Nalohou Clear sky (aeronet Djougou) Djougou Figure Surface solar irradiance reported versus the day of the year for the three neighbouring stations Djougou, Bira and Nalohou in the Benin; the calibration factors are adapted to match to the clear sky SSI values. - -

9 Figure Daily highest value of the surface solar irradiance reported versus the day of the year for the station of Tamanrasset, for the measurements and the clear sky SSI evaluated from the aod and the Solis model. The clear sky index is also represented. this procedure, no correction is applied to the data from Hedgerit, Sede Boqer and Tamanrasset, and a positive or negative correction to the other concerned stations: Banizoumbou (-%), Djougou (+%), Kelema (+7%), M Bour (+%), Nalohou (+8%) and Wankama (+%). A visual control is illustrated on Figure where for each day of the year, the highest solar surface irradiance is represented versus the day of the year, for the ground measurements and the corresponding clear sky value SSI c evaluated from the aeronet network aod. The daily clear sky index K c defined as SSI SSI K c = () c is also represented, and should remain around the unity for correctly calibrated ground measurements. Similar graphs for all the concerned stations are given in the Annex.. Satellite derived data. Heliosat- algorithm The Helioclim data bank is produced with the Heliosat- method that converts observations made by geostationary meteorological satellites into estimates of the global irradiation at ground level. This version integrates the knowledge gained by various exploitations of the original method Heliosat and its varieties in a coherent and thorough way. It is based upon the same physical principles but the inputs to the method are calibrated radiances, instead of the digital counts output from the sensor. This change opens the possibilities of using known models of the physical processes in atmospheric optics, thus removing the need for empirically defined parameters and of pyranometric measurements to tune them. The ESRA models (ESRA ) are used for modeling the - -

10 clear-sky irradiation. The assessment of the ground albedo and the cloud albedo is based upon explicit formulations of the path radiance and the transmittance of the atmosphere. The turbidity is based on climatic monthly Linke Turbidity coefficients data banks.. OSI-SAF algorithm The algorithm is based on a physical parameterization using the visible channel of an imaging radiometer as main input. The visible count is calibrated and converted into TOA broadband albedo, using a narrow to broadband correction based on the well-calibrated broadband radiometer CERES (Clouds and Earth s Radiant Energy System) and an anisotropy correction from Manalo- Smith (998). The SSI parameterization, described in Brisson (999), can be summarized as follows: If clear SSI = SSI TOA τ a If cloudy SSI = SSI TOA τ τ cl (α c, α s ) α TAO = α ray + α c τ + α s τ (α c, α s ) In the first equation, the TOA SSI is attenuated by τ a, sun-to-surface clear atmosphere transmittance. The second equation introduces another attenuation term τ cl, sun-tosurface cloudy factor, which mainly depends on the cloud albedo α c and, to a lesser extent, on the surface albedo α s ; τ is consistent with τ a, but without multiple scattering, which is taken into account in τ cl. In the third equation, the TOA albedo, α TOA, is the sum of three terms: the contribution of the Rayleigh scattering α ray, the radiation reflected by the cloud α c attenuated by τ, sun-to-cloud-to-satellite clear atmosphere transmittance and the radiation reflected by the surface α s attenuated by τ, which accounts for a cloud and atmospheric effect. The cloud classification of the Nowcasting (NWC) SAF is used to separate between clear and cloudy cases. For clear cases, only the first equation is used; for cloudy cases, the cloud albedo is derived from the TOA albedo by solving the third equation, which then allows to calculate the SSI by the second one. The surface albedo is obtained, over land by monthly climatologic fields and, over sea by Briegleb (98) formulas. All transmittance terms are calculated by analytical formulas, mainly from Frouin (99), using as inputs: the water vapor content of the atmosphere predicted by the Numerical Weather Predicting (NWP) model ARPEGE of Météo-France and monthly climatologic fields of ozone content of the atmosphere, and horizontal visibility (tabulated)

11 . Meteonorm process The generation of hourly surface solar irradiance values at any desired location is done with stochastic models. The models generate intermediate data having the same statistical properties than the measured data, i.e. average value, variance, and characteristic sequence (autocorrelation). The generated data approximates the natural characteristics as far as possible. Recent research shows that data generated in this way can be used satisfactorily in place of long-term measured data (Gansler 99). The model of Aguiar (988) provides the starting point for this methodology. It calculates daily values of the surface solar irradiance with monthly mean values as inputs. The whole system of the matrices was changed from a clearness index K t to a clear sky index K c basis. Formulated like this, the maximum value of K c must correspond automatically to the clear sky model predictions used. As for the helioclim algorithm, the ESRA clear sky model is used (Rigollier, ESRA ). The basic SoDa mapped resource of monthly mean Linke turbidity factors is used to drive the clear sky model using algorithmic to obtain the required monthly mean daily values of the surface solar irradiance needed to calculate K c for any selected month and any point. This change required the daily Markov transition matrices tables to be completely revised to match the new formulation. The generation of hourly values is based on the model of Aguiar and Collares-Pereira (99) (TAGmodel: Time dependent, Autoregressive, Gaussian model). This model consists of two parts: the first part calculates an average daily profile; the second part simulates the intermittent hourly variations by superimposing an autoregressive procedure of the first order (Box 99).. The atmospheric turbidity The evaluation of the surface solar irradiance from satellite data is subject to the knowledge of the corresponding highest possible irradiance value for the considered time and location; it is used as normalization value. This radiation component is dependent of the atmospheric turbidity which is expressed with the help of the aerosol optical depth aod, the Linke turbidity T L (Linke (9), Kasten (98), Ineichen ()), the angström coefficients α and β (Angström 9) or the visibility (King 979). The clear sky models use these turbidity coefficients as input parameter. Radiative transfer (RTM) based model like Solis (Mueller (), Ineichen (8)) need the aerosol optical depth and the total atmospheric water vapor content w, ESRA (ESRA ) clear sky model the Linke turbidity coefficient (which is a combination of the aod and w, (Rigollier, Geiger 7), other the visibility (Gueymard 989) or the angström coefficients. The atmospheric turbidity has the highest effect on the transmission, it is also the most - 8 -

12 difficult to obtain or to retrieve. If tables and atlases exist, its variability is often non negligible from day to day, and even during the day. If the clear sky surface solar irradiance SSI c can be evaluated with the knowledge of the atmospheric turbidity, a back evaluation of the turbidity from measurements can also be done in different ways: - when the beam component DNI is known, with the use of Kasten pyrheliometric formula (Kasten 98) for an air mass M = (or Ineichen s air mass independent formulation (Ineichen )) the Linke turbidity coefficient T L can be obtained for clear days on a daily basis. Then from the Linke turbidity coefficient T L and the water vapor content of the atmosphere w, it is possible to evaluate the aerosol optical depth using a model developped by (Ineichen & 8). - from the SSI and/or the DNI, the aod can be evaluated following the procedure described in (Ineichen ) using the Solis clear sky model. The above methods are validated with spectral and broadband beam measurements acquired in Payerne during the year (Vuilleumier ) and in Bonville (US) for the year (SurfRad 9). This is illustrated on Figure (right) where the aod obtained from T L is plotted against the spectral aod measurements from Payerne. Considering that the values are daily averages, and that the method cumulates several models, the correlation between the aerosol optical depth obtained from the normal beam irradiance and from spectral measurements is satisfying. The second method is illustrated on Figure (left) with data acquired at Bondville (US) with a Multifiter Rotating Shadowband (MFRSR, Harisson 99). Here also, it can be seen that the correspondance is satisfactory, even if the dispersion increases for high aod values, due to the instability of the irradiance conditions for higher turbidity. Figure Aerosol optical depth obtained through the the DNI with the help of T L with Payerne (CH) measurements, and Solis model at the site of Bondville (US) The advantage of this method is that it doesn t need the normal beam irradiance (DNI) as input parameter, but, because of the relative low dependence of the surface global irradiance with the atmospheric aerosol optical depth, the method is very sensitive to - 9 -

13 the quality of the SSI measurements 7. Comparison and evaluation procedure In terms of validation, when evaluating satellite derived parameters with the same time step, the comparison can be done by means of scatter plots; these give a visual evaluation of the capability of the model to reproduce the measurements. An illustration is given on Figure for Helioclim product and applied to Nalohou and Yatir Forest sites. It can be seen here that the model gives very good results for Yatir Forest (Israel); on contrary, for data from Nalohou (Benin), a very high dispersion can be seen. This can be an issue as well from the model as the ground measurements. Concerning Meteonorm product, due to the synthetic generation of the hourly values, only the mean bias difference can be calculated and compared. Figure Scatter plots for the surface solar irradiance produced by Helioclim for two sites: Nalohou (Benin) and Yatir Forest (Israel). The statistical parameters like the mean bias difference (MBD), the root mean square difference (RM), the standard deviation () and the determination coefficient (R ) represent a quantification of the model dispersion. These statistical parameters include dispersions introduced by: - the retrieval procedure, - the comparison of point measurements (ground data) with aera measurements, - the comparison of instantaneous measurements with minutes integrated values. In the field of solar radiation and natural light, the comparison is often done in term of frequency of occurrence. The obtained graph is a line (or a bar chart) representative of the relative frequency of occurrence of a given parameter. This is illustrated on Figure 7 - -

14 Figure 7 Relative frequency of occurrence of the clearness index Kt for data from M Bour (Senegal) and Yatir Forest (Israel). Figure 8 Relative frequency of occurrence for the surface solar irradiance for data from Nalohou (Benin) and Yatir Forest (Israel). for the clearness K t. On the same graph, the frequency of occurrence of the ground measurements are represented as gray bars, and the different models in color lines. If the right graph (Yatir Forest) shows a good agreement between ground measurements and satellite models, the left graph (M Bour) shows clear discrepancies, particularly for high values of clearness index, that is clear conditions. A second order statistic, the Kolmogorov-Smirnov test (Espinar 9), is also applied to the data. It represents the capability of the model to follow the frequency of occurrence at each of the irradiance level. A visualisation is given on Figure 8 where the irradiance cumulated frequency of occurrence is represented against the SSI for the two same sites than above. Here, produced SSI values are in good agreement for the stations of Yatir Forest, but not for Nalohou. The Kolmogorov Smirnov test Integral (KSI) is defined as: KSI = SSI SSI max min G(SSI) M(SSI) dssi where G(SSI) and M(SSI) are respectively the ground measurements and the modelled cumulated frequencies of occurrence. - -

15 Gh [W/m] nb Helioclim OSI-SAF Meteororm mbd rmsd sd R KSI mbd rmsd sd R KSI mbd rmsd sd R KSI Agoufou (Mali) Bamba (Mali) Banizoumbou (Niger) Bira (Benin) Djougou (Benin) El Saler (Spain) Geneva (Switzerland) Hedgerit (Mali) Kelema (Mali) Las Majadas (Spain) M'Bour (Senegal) Mt Kenya (Kenya) Nalohou (Benin) Sede Boqer (Israel) Tamanrasset (Algeria) Tchizaloumou (Congo) Val d'alinya (Spain) Wankama (Niger) Yatir Forest (Israel) Europe & Northern Africa Africa Europe & Africa Table II Absolute statistical results obtained for each site and each product, overall results for the two groups of sites and for all sites together (except Nalohou and Tchizaloumou). Figure 9 Absolute MBD and RM obtained for each site and each product, overall results for the two groups of sites and for all sites together (except Nalohou and Tchizaloumou). - -

16 Gh [W/m] nb Helioclim OSI-SAF Meteororm mbd % rmsd % sd % R KSI mbd % rmsd % sd % R KSI mbd % rmsd % sd % R KSI Agoufou (Mali) Bamba (Mali) Banizoumbou (Niger) Bira (Benin) Djougou (Benin) El Saler (Spain) Geneva (Switzerland) Hedgerit (Mali) Kelema (Mali) Las Majadas (Spain) M'Bour (Senegal) Mt Kenya (Kenya) Nalohou (Benin) Sede Boqer (Israel) Tamanrasset (Algeria) Tchizaloumou (Congo) Val d'alinya (Spain) Wankama (Niger) Yatir Forest (Israel) Europe & Northern Africa Africa Europe & Africa Table III Relative statistical results obtained for each site and each product, overall results for the two groups of sites and for all sites together (except Nalohou and Tchizaloumou). Figure Relative MBD and RM obtained for each site and each product, overall results for the two groups of sites and for all sites together (except Nalohou and Tchizaloumou). - -

17 8. Results To ensure a correct and comparable validation of the different products, the following method was used to merge the products and the ground measurements: for each generated value, the nearest time stamped corresponding ground value is searched in the data base; this means that the satellite image was taken within the ground integration period. Then, only hours for which ground and both generated products are present are taken into account in the validation procedure; so the mainly the ground data availability restrict the number of points in the comparison. Furthermore, OSI-SAF produces only value for a solar elevation greater than, this will therefore be the lower limit for the whole comparison. The results of the validation are given on Table II and Table III respectively in absolute and relative values. The corresponding graphs are given on Figure 9 and Figure. The first established fact is that the sites of Nalohou and Tchizaloumou give very bad results. For Naholou, the scatter plot for the Helioclim product is represented on Figure : it can be seen that a very high dispersion is present. Referring to chapter, the SSI ground measurements were corrected by 8% to be coherent with the corresponding clear sky. Figure shows that the frequency of occurrence of the modelled surface solar irradiance clearness index is not in agreement with the corresponding measurements (in grey bars on the graph); at that site, during the summer, the highest daily value never reaches the clear sky value (see the graphs in the Annex ). Figure Surface solar irradiance clearness index frequency of occurrence for the different products. The corresponding ground values are represented in grey bars. For the site of Tchizaloumou in Congo, all the models give higher values than the ground measurements. This is illustrated on Figure. On the left graph, the scatter plot show a model overestimation (or measurements undersetimation) over the whole range of SSI. The right graph, where the SSI clearness index is plotted against the solar elevation angle, show that the measurements do very rarely reach a relatively high turbid clear sky model (in yellow). Furthermore, all the models give higher values. The frequency of occurrence of the clearness index shows also a high discrepancy with the measurements for all the models (see Annex ). The data for Tchizaloumou cover a half year from July to december which is more or less the dry season; it is the only station at that latitude. As no aod mesurements are done in the region of Tchizaloumou, it is not possible to give a further explanation to this discrepancy without other ground measurements. - -

18 For these reasons, the sites of Nalohou and Tchizaloumou are not included in the overall statistics given on Table II and III, and on Figures 9 and. On the previous Tables and Figures, it appears that the validation results are better for the Europe and Northern Africa sites, in terms of mean bias difference and dispersion. Considering the lack of ground measurements for the African continent and the difficulty to apply a quality control on these data, it is at this stage not possible to point out if the higher bias and dispersion issue from the ground data or the models. But the importance of the aerosol climatology in the process is a known fact, and, as it is better known over Europe, it certainly contributes to smaller biases and dispersions. Figure Site of Tchizaloumou. On the left, Helioclim modelled SSI against corresponding ground measurements. On the right, the clearness index against the solar elevation angle. Figure Site of Agoufou: aerosol optical depth retrieved from SSI for the ground measurements and the three products. - -

19 It can also be noted that if the OSI-SAF product shows a positive bias for all the stations (except Tamanrasset and Val d Alinya), Helioclim has highly variable biases depending on the site, and Meteonorm is more or less underestimating for the majority of the stations. To better understand the effect of the turbidity on the results, the aerosol optical depths are retro-calculated from SSI ground measurements and for the products with the procedure cited in chapter and described in (Ineichen ). The method permits to obtain daily turbidity values that are averaged to monthly and yearly values. The results are illustrated on Figure for the station of Agoufou (Mali). The difference between the Helioclim and OSI-SAF aerosol climatology appears clearly on this Figure. Then the annual deviation of the aerosol optical depth from ground measurements (i.e. retrieved from ground data of the solar surface irradiance) are represented against the model bias on Figure, where a tendency best fit is also represented. A non negligible correlation can be seen between the aod deviation and the bias. It has to be noted that not only the aerosol optical depth is responsible to the bias and the dispersion, but also the water vapor column and ground albedo. Figure Annual aerosol optical depth deviation from measurements (i.e. retrieved from ground SSI data) against the bodel bias In the Annex, for the site of Agoufou, Hedgerit and Kelema, the clears sky values show for Helioclim and Meteonorm a underestimation in the spring period (February to May), that induces a shift in the frequency of occurrence to low values of K t (see Annex ). For Helioclim, the clearness index graph against the solar elevation angle shows that the clear sky model is always to low in comparison to the measurements; less marked, the effect is also present for Meteonorm. The monthly analysis of the aerosol optical depth as shown on Figure for the site of Agoufou presents a clear overestimation of the aod. Curiously, even if the aod shows the same overestimation, the models don t show the same pattern for the site of Kelema; due to the short measurements period, the conclusion is difficult to draw. - -

20 The aerosol optical depth is acquired in Banizoumbou, and is used also for the site of Wankama, situated at km. For the two sites, the models underestimate the SSI in the summer period (June to September). Here again, the clearness index for the Helioclim product is to low for all the solar elevation angles (Annex ). In the Djougou region, the clear sky models undersetimate in the summer period (see Annexe ), even if the aod used in the model seem similar to the measurements (Figure ). The water vapor remains stable during the summer at about [cm], in contrary to the surface solar irradiance. The same pattern can be seen on the data from Bira and Nalohou. Figure Yearly and monthly aerosol optical depths retrieved from SSI values for the site of Agoufou and Djougou. Unfortunately, due to the lack of ground measurements (particularly, only a few sites cover a complete seasonnal variation period), and the difficulty to assess their calibration and quality, the statistics are to low to draw further conclusions. One of the main difficulty in such a validation is the high disparity of the data availability; the comparison can be correctly done only if the ground data cover the same period. As stated above, the results are better over Europe and Northern Africa. An interresting fact can be pointed out from the clearness index graphs versus the solar velevation angle Figure ): for the African sites, the Helioclim clearness index is to low for all the Figure Hourly clearness index against the solar elevation angle

21 solar elevation angles, while for the European sites, the agreement with the ground measurements is better for high solar elevation angles. The monthly mean bias and root mean square differences are illustrated on Figure 7 for the stations of Bamba and Banizoumbou; on the top of the graphs, the monthly and yearly SSI average and the number of hours included in the statistics are also given. Tables and graphs for all the sites are given in Annex. Figure 7 Monthly and yearly relative MBD and for the sites of Banizoumbou and Bamba. The bars are the mean bias differences, surrounded by ± one standard deviation. On the top of the graphs, the monthly and yearly SSI average and the number of hours included in the statistics are given. 9. Conclusions The lack of ground data over the African continent and the difficulty to obtain and to assess them is a major problem in validation projects in these region. For example, some data had to be corrected up to 8% to be coherent one with the other, and to match independent measurements. The main conclusions of the present study are the validation results over Europe and Northern Africa, and over West Africa. These are a bias from -% to % (- to 7 W/m ) and a dispersion from % to 7% ( to W/m ), with slightly better results over Europe. The higher dispersion on the African continent comes from a lower quality of the ground data and a poorer knowledge of the aerosol optical depths over these regions. The over and underestimation against the season is not systematic. For all the models, and the majority of the sites, the frequency of occurrence of high clearness indices is overestimated. The interannual variability of the irradiance conditions, the lack of ground and aerosol data conduct to a relatively high variability of the validation results from one year to the other and it is difficult to draw general conclusions based on one specific year (or even a partial year)

22 . Acknowledgements This comparative study could be performed thanks to the data provided by the FluxNet community ( for the Spain sites, Yatir Forest and Tchizaloumou, and the AMMA project ( for providing the other African sites except Mt Kenya provided by the GAW project, the BSRN and the Aeronet networks for the aerosol optical depths and for Tamanrasset and Sede Boqer ground data

23 . Nomenclature and abbreviations I o solar constant in W/m DNI direct normal incidence irradiance in W/m SSI surface solar global irradiance on a horizontal plane in W/m SSI c clear sky surface solar irradiance in W/m SSI TOA extra-atmospheric solar irradiance on a horizontal plane in W/m K t K c DLI global or surface solar irradiance clearness index clear sky index downward longwave irradiance on a horizontal plane and expressed in W/m h solar elevation or altitude angle in degrees ( ) M optical air mass RTM TL α β aod radiative transfer model Linke turbidity Angström turbidity coefficients aerosol optical depth T a T d T e T cb HR w w bc σ surface (ambient) temperature dew point temperature atmospheric emitting temperature cloud base temperature relative humidity atmospheric total water vapour column atmospheric total water vapour column below the cloud Stefan-Boltzman constant α TOA α c α s α ray top of the atmopshere albedo cloud albedo surface albedo Rayleigh scattering contribution to the albedo τ a τ cl atmospheric transmittance cloud transmittance MBD AMBD RM mean bias difference absolute mean bias difference root mean square difference standard deviation - -

24 - -

25 . References AGUIAR, R. AND M. COLLARES-PEREIRA (988) A simple procedure for generating sequences of daily radiation values using a library of markov transition matrices. Solar Energy, Vol., No., pp AGUIAR, R. AND M. COLLARES-PEREIRA (99) TAG: A time-dependent autoregressive, Gaussian model. Solar Energy, Vol. 9, No., pp ANGSTROM A. (9) Techniques of determining the turbidity of the atmosphere. Tellus, -. ATWATER M.A., BALL J.T. (97) Comparison of radiation computations using observed and estimated precipitable water. Appl. Meteorol., 9- BOX, G. O., G.M. JENKINS AND G.C. REINSEL. (99) Time series analysis: Forecasting and control. rd edition. Prentice Hall, Englewood Cliffs, NJ. BRIEGLEB B. P., MINNIS P., RAMANATHAN V., HARRISON E. (98). Comparison on regional clear-sky albedos inferred from satellite observations and model computations, Journal of Climate and Applied Meteorology,, pp. -. BRISSON A., Le BORGNE P., MARSOUIN A. (999) Surface Solar Irradiance retrieval from GOES data in the framework of the Ocean and Sea Ice Satellite Application Facility, Proceedings of the EUMETSAT Meteorological Satellite Data Users Conference, Copenhagen, - September. CUTHFORTH H.W., JUDIESH D. (7). Long-term changes to incoming solar energy om the Canadian Prairie. Agr. Forest Met., 7-7. ESPINAR B., RAMIREZ L., DREWS A., BEYER H-G., ZARZALEJO L.F., POLO J., MARTIN L. (9) Analysis of different comparison parameters applied to solar radiation data from satellite and German radiometric stations. Solar Energy 8, 8 ESRA () The european solar radiation atlas. Coordinators : K. SCHARMER, J. GREIF, ISBN : FROUIN R. A., CHERTOCK B., (99). A technique for global monitoring of net solar irradiance at the ocean surface. Part : Model, Journal of Applied meteorology,,pp

26 GANSLER, R.A., S.A. KLEIN AND W.A. BECKMAN (99) Assessment of the accuracy of generated meteorological data for use in solar energy simulation studies. Solar Energy, Vol., No., pp GEIGER B., MEUREY C., LAJAS D., FRANCHISTÉGUY L., CARRER D., and ROUJEAN J.-L., (7) Near real-time provision of downwelling shortwave radiation estimates derived from satellite observations, Meteorological Applications. to be submitted GUEYMARD, C. (989). A two-band model for the calculation of clear sky solar irradiance, illuminance and photosynthetically active radiation at the earth surface. Solar Energy (),. HARRISON L., MICHALSKY J., BERNDT J. (99) Automated multifilter shadowband radiometer: an instrument for optical depth and radiation measurements. Appl. Opt., pp8-. HELIOCLIM (9) Sun radiation atlas based on processing of Meteosat data. < HOYER-CLICK C. (9). Solemi scheme < INEICHEN P, PEREZ R. () A new airmass independent formulation for the Linke turbidity coefficient. Solar Energy 7, No., 7 INEICHEN P. () Comparison of eight clear sky broadband models against independent data banks. Solar Energy 8, 8 78 INEICHEN P. (8). Comparison and validation of three global-to-beam irradiance models against ground measurements. Sol. Energy, doi:./j.solener.7.. INEICHEN P. (8) A broadband simplified version of the Solis clear sky model, Solar Energy 8, pp 78 7 INEICHEN P. (8) Conversion function between the Linke turbidity and the atmospheric water vapor and aerosol content, Solar Energy 8, pp INEICHEN P.,BARROSOC.,GEIGERB.,HOLLMANNR.,MARSOUINA.,MUELLERR.(9) Satellite Application Facilities irradiance products: hourly time step comparison and validation over Europe. International Journal of Remote Sensing, Vol., No., pp

27 INEICHEN P., PEREZ R. () Aerosol quantification and characterization from global and beam irradiance measurements. Submitted to Solar Energy KASTEN F. (98) A simple parameterization of two pyrheliometric formulae for determining the Linke turbidity factor. Meteor. Rdsch., 7. KING R., BUCKIUS R.O. (979) Direct solar transmittance for clear sky. Solar Energy,pp 97- KINNE, S. et al () An AeroCom initial assessment optical properties in aerosol component modules of global models Atmos. Chem. Phys. Discuss.,, -. LINKE F. (9) Transmissions-koeffizient und Trübungsfaktor. Beitr. Phys. Fr. Atmos., pp 9-. LORENZ E., (9). EnMetSol scheme. IAE Task XXXVI < MANALO-SMITH N., SMITH G.L., TIWARI S.N., STAYLOR W.F., (998). Analytic forms of bidirectional reflectance functions for application to Earth radiation budget studies, Journal of Geophysical Research, Vol., D, pp MUELLER R.W. et al, () Rethinking satellite based solar irradiance modelling The SOLIS clear sky module. Remote Sensing of Environment,9, -7. Meteonorm. (9) Global Meteorological Database for Solar Energy and Applied Meteorology. < RIGOLLIER C., BAUER O., WALD L. () On the Clear Sky Model of the ESRA - european Solar Radiation Atlas - with Respect to the Heliosat Method. Solar Energy 8 (), -8. RIGOLLIER C., LEFÈVRE M, WALD L. () The method heliosat- for deriving shortwave solar irradiance radiation from satellite imagessolar Energy, 77(), 9-9, SATEL-LIGHT (). < STANHILL, G., COHEN, S.,(). Global dimming: a review of the evidence for a widespread and significant reduction in global radiation with discussion of its probable causes and possible agricultural consequences. Agric. Forest Meteorol. 7,

28 STUHLMANN, R., M. RIELAND AND E. RASCHKE, (99) An improvement of the IGMK Model to derive total and diffuse solar radiation at the surface from satellite data, Journal of Applied Meteorology 9, 8-. SURFRAD Network (9), Monitoring Surface Radiation in the Continental United States. NOAA, Surface Radiation Research Branch. Available from: < VUILLEUMIER L. (). Spectral and broadband beam measurements from the MeteoSwiss Payerne station of the Baseline Surface Radiation Network. MeteoSwiss, private communication. - -

29 - -

30 Helioclim and Meteonorm short wave irradiance validation over Africa Annex Modelled SSI against the ground measurements for the three products. Surface solar irradiance clearness index versus the solar elevation angle. In yellow the ground measurements and in blue the different products. Surface solar irradiance clearness index frequency of occurrence for the different products. The corresponding ground values are represented in grey bars. Surface solar irradiance cumulated frequency of occurrence for the different products and for the ground measurements

31 Hourly horizontal global irradiance [Wh/m h] Helioclim (Agoufou ). Measurements (AMMA Agoufou ) Helioclim (Agoufou ) Measurements (AMMA Agoufou ) Measurements (AMMA Agoufou ) Meteonorm (Agoufou ).8... Hourly horizontal global irradiance [Wh/m h] OSI-SAF (Agoufou ). Measurements (AMMA Agoufou ) OSI-SAF (Agoufou ) Measurements (AMMA Agoufou ) 8 Modelled SSI against the ground measurements for the three products. Surface solar irradiance clearness index versus the solar elevation angle. In yellow the ground measurements and in blue the different products. Relative frequency of occurrence of the clearness index K t Measurements (AMMA Agoufou ) Helioclim (Agoufou ) Meteonorm (Agoufou ) OSI-SAF (Agoufou ) % 8% % Cumulated frequency of occurrence of the global irradiance G h % K t = G h / I oh Surface solar irradiance clearness index frequency of occurrence for the different products. The corresponding ground values are represented in grey bars. % Measurements (AMMA Agoufou ) Helioclim (Agoufou ) Meteonorm (Agoufou ) OSI-SAF (Agoufou ) % 8 Global irradiance [W/m ] Surface solar irradiance cumulated frequency of occurrence for the different products and for the ground measurements

32 Hourly horizontal global irradiance [Wh/m h] Helioclim (Bamba ). Measurements (AMMA Bamba ) Helioclim (Bamba ) Measurements (AMMA Bamba ) Measurements (AMMA Bamba ) Meteonorm (Bamba ).8... Hourly horizontal global irradiance [Wh/m h] OSI-SAF (Bamba ). Measurements (AMMA Bamba ) OSI-SAF (Bamba ) Measurements (AMMA Bamba ) 8 Modelled SSI against the ground measurements for the three products. Surface solar irradiance clearness index versus the solar elevation angle. In yellow the ground measurements and in blue the different products. Relative frequency of occurrence of the clearness index K t Measurements (AMMA Bamba ) Helioclim (Bamba ) Meteonorm (Bamba ) OSI-SAF (Bamba ) % 8% % Cumulated frequency of occurrence of the global irradiance G h % K t = G h / I oh Surface solar irradiance clearness index frequency of occurrence for the different products. The corresponding ground values are represented in grey bars. % Measurements (AMMA Bamba ) Helioclim (Bamba ) Meteonorm (Bamba ) OSI-SAF (Bamba ) % 8 Global irradiance [W/m ] Surface solar irradiance cumulated frequency of occurrence for the different products and for the ground measurements

33 Hourly horizontal global irradiance [Wh/m h] Helioclim (Banizoumbou ). Measurements (AMMA Banizoumbou ) Helioclim (Banizoumbou ) Measurements (AMMA Banizoumbou ) Measurements (AMMA Banizoumbou ) Meteonorm (Banizoumbou ).8... Hourly horizontal global irradiance [Wh/m h] OSI-SAF (Banizoumbou ). Measurements (AMMA Banizoumbou ) OSI-SAF (Banizoumbou ) Measurements (AMMA Banizoumbou ) 8 Modelled SSI against the ground measurements for the three products. Relative frequency of occurrence of the clearness index K t Measurements (AMMA Banizoumbou ) Helioclim (Banizoumbou ) Meteonorm (Banizoumbou ) OSI-SAF (Banizoumbou ) % 8% % Surface solar irradiance clearness index versus the solar elevation angle. In yellow the ground measurements and in blue the different products. Cumulated frequency of occurrence of the global irradiance G h % K t = G h / I oh Measurements (AMMA Banizoumbou ) % Helioclim (Banizoumbou ) Meteonorm (Banizoumbou ) OSI-SAF (Banizoumbou ) % 8 Global irradiance [W/m ] Surface solar irradiance clearness index frequency of occurrence for the different products. The corresponding ground values are represented in grey bars. - - Surface solar irradiance cumulated frequency of occurrence for the different products and for the ground measurements.

34 Hourly horizontal global irradiance [Wh/m h] Helioclim (Bira ). Measurements (AMMA Bira ) Helioclim (Bira ) Measurements (AMMA Bira ) Measurements (AMMA Bira ) Meteonorm (x ).8... Hourly horizontal global irradiance [Wh/m h] OSI-SAF (Bira ). Measurements (AMMA Bira ) OSI-SAF (Bira ) Measurements (AMMA Bira ) 8 Modelled SSI against the ground measurements for the three products. Surface solar irradiance clearness index versus the solar elevation angle. In yellow the ground measurements and in blue the different products. Relative frequency of occurrence of the clearness index K t Measurements (AMMA Bira ) Helioclim (Bira ) Meteonorm (Bira ) OSI-SAF (Bira ) % 8% % Cumulated frequency of occurrence of the global irradiance G h % K t = G h / I oh Surface solar irradiance clearness index frequency of occurrence for the different products. The corresponding ground values are represented in grey bars. % Measurements (AMMA Bira ) Helioclim (Bira ) Meteonorm (Bira ) OSI-SAF (Bira ) % 8 Global irradiance [W/m ] Surface solar irradiance cumulated frequency of occurrence for the different products and for the ground measurements. - -