SAF/LAND/MF/PUM_DSSF/1.4. Issue: Version 1.4 DSSF Date: 19 October Product User Manual PUM DOWN-WELLING SURFACE SHORT-WAVE RADIATION FLUX

Transcription

1 Product User Manual DOWN-WELLING SURFACE SHORT-WAVE RADIATION FLUX Reference Number: SAF/LAND/MF/_DSSF/1.4 Issue/Revision Index: Issue 1.4 Last Change: 19/10/2006 1

2 DOCUMENT SIGNATURE TABLE Name Date Signature Prepared by : Météo-France / CNRM 19/10/2006 Approved by : Project Manager 19/10/2006 DOCUMENTATION CHANGE RECORD Issue / Revision Date Description : Version /10/2004 Preliminary version Version /11/2004 Version prepared for SIVVRR 2 Version /03/2005 Version prepared for Check-Point Meeting Version /06/2005 Version prepared for ORR-1 Version /01/2006 Version prepared for ORR-1 Close-Out Version /10/2006 Version prepared for OR-1 2

3 DISTRIBUTION LIST Consortium Internal Distribution Organisation Name No. Copies IM Carlos Direitinho Tavares IM Luís Pessanha IM Pedro Viterbo ICAT IM IM IM IM IM IM IM IMK IMK IMK MF MF MF MF RMI RMI RMI FMI UV UV UV UV Carlos da Camara Isabel Trigo Isabel Monteiro Sandra Coelho Carla Barroso Pedro Diegues Teresa Calado Benvinda Barbosa Folke-S. Olesen Andreas Schmidt Ewa Gajewska Jean-Louis Roujean Bernhard Geiger Dominique Carrer Catherine Meurey Françoise Meulenberghs Alirio Arboleda Nicolas Ghilain Niilo Siljamo Joaquín Meliá F. Javier García Haro Beatriz Martínez Fernando Camacho de Coca 3

4 Steering Group Distribution Nominated by: Name No. Copies IM Fátima Espírito Santo EUMETSAT Lorenzo Sarlo EUMETSAT Yves Govaerts EUMETSAT François Montagner UGM Luigi de Leonibus MF François Bouyssel RMI Alexandre Joukoff FMI Tapio Tuomi External Distribution Organisation Name No. Copies EUMETSAT Frédéric Gasiglia EUMETSAT Dominique Faucher EUMETSAT Lothar Schueller EDISOFT Teresa Cardoso EDISOFT Ricardo Pereira EDISOFT Paulo Carmo SKYSOFT Nuno Costa SKYSOFT Justino Sousa 4

5 Table of Contents 1 INTRODUCTION ALGORITHM Overview Definition Clear Sky Method Cloudy Sky Method PRODUCT DESCRIPTION Continental Windows File Format Temporal Reference Product Content Summary of Product Characteristics VALIDATION Daily Cycle Scatter Plots Validation Statistics REFERENCES

6 List of Figures Figure 1: The list of products generated by the LSA SAF system (including internal products)....9 Figure 2: The geographical areas for which the products are separately processed and distributed by the LSA SAF system Figure 3: Simplified flow chart of the DSSF algorithm Figure 4: Schematics illustrating some elements of the clear (left) and cloudy sky (right) DSSF estimation methods Figure 5: The time difference between the slot and the acquisition time of the image line as a function of the line number (left) and the latitude for a longitude of 0 (right)...17 Figure 6: Re-composed full-disk image of the down-welling surface short-wave radiation flux estimate for the European window on the 15 th of February 2006 at 13:30 UTC Figure 7: Quality flag corresponding to the DSSF estimate depicted in Figure 6. The signification of the bit codes is given in Table 4. xxx indicates that the most significant bits 5 to 7 were used for visualisation while xx represents the least significant bits 0 to Figure 8: Examples for daily time series of DSSF estimates and in-situ measurements at the ground validation stations. The colour code of the dots is the same as for the quality flag in Figure Figure 9: Scatter plots between the satellite estimates and the ground measurements. The colour code of the dots is the same as for the quality flag in the previous figures. The bottom right plot shows the daily averaged estimates as explained in the text. (Note that the shown range of values is different in this case.)...23 Figure 10: Temporal evolution of bias and standard deviation between the LSA SAF estimates and ground measurements for all validation sites combined (Carpentras, Roissy, Évora, and Toravere) for the period from October 2004 to August

7 List of Tables Table 1: Numerical values of the constants appearing in the parameterisations of Frouin et al. (1989) Table 2: Characteristics of the four LSA SAF geographical areas: Each region is defined by the corner positions relative to an MSG image of 3712 columns per 3712 lines with indices starting at 1 in the North and West Table 3: Content of the DSSF product files Table 4: DSSF quality flag information Table 5: General HDF5 attributes...27 Table 6: Dataset attributes

8 1 Introduction The main purpose of the Satellite Application Facility on Land Surface Analysis (LSA SAF) is to develop techniques to retrieve products related with land, landatmosphere interactions, and biospheric applications using data from the EUMETSAT satellite systems: Meteosat Second Generation (MSG-1 launched in August 2002 and MSG-2 in December 2005) and the meteorological operational polar satellites (the first MetOp scheduled for October 2006). Recent studies, namely those involving systematic comparisons of land surface parameterisation schemes, have stressed the role of land surface processes on weather forecasting and climate modelling. Numerical Weather Prediction (NWP) models have incorporated improved land surface representations that require sophisticated assimilation schemes of different types of land surface data that may include remote-sensed information (e.g. on LST, surface albedo, vegetation and soil moisture dynamics). Therefore the NWP community has been identified as having the greatest potential to fully exploit the LSA SAF products and the meteorological users have been assigned the highest priority during the phases of product design and development. However, the LSA SAF addresses a much broader community, including amongst others: Atmospheric reanalyses and climate modelling, which require detailed information on the nature and properties of land. Environmental management and land use, which require information on land cover type and land cover changes (e.g. provided by biophysical parameters or thermal characteristics). Agricultural and forestry applications, which require information on soil and vegetation properties. Renewable energy resources assessment, particularly biomass, which depends on biophysical parameters, and solar energy, which highly depends on down-welling short-wave radiation at the surface. Natural hazards management, which requires frequent observations of terrestrial surfaces in both the solar and thermal bands. Climatological applications and climate change detection. Products to be derived by the LSA SAF (Figure 1) will be based on data from satellites combined with data from different other sources. Data from EUMETSAT are extracted from Levels 1.0/1.5 for MSG and Levels 1.a/1.b for MetOp. MetOp data will be completed with information from other programmes like NOAA. Other sources than EUMETSAT satellite systems may also be used such as routine meteorological information. All the products will be computed within the area covered by the MSG disk or by EPS in the adjacent polar region over specific geographical regions, with the corresponding spatial and temporal resolution. 8

9 Figure 1: The list of products generated by the LSA SAF system (including internal products). The product generation for the MSG disk is split into four different geographical areas (Figure 2), to which different priorities are assigned: Europe (Euro) is the highest priority geographical area, covering all EUMETSAT member states; Africa is split into two geographical regions (NAfr and SAfr) and is given intermediate priority; the geographical area including South America (SAme) has the lowest priority. The LSA SAF system is fully centralised at IM in Lisbon and is able to operationally generate, archive, and disseminate the products. The monitoring and quality control of the operational products is performed automatically by the LSA SAF software, which provides quality information to be distributed with the products. The LSA SAF products are currently available from the LSA SAF web-site ( which contains real time examples of the products as well as updated information. This document is one of the product manuals dedicated to LSA SAF users and concerns the Down-welling Surface Short-wave Radiation Flux (DSSF) algorithm developed by Météo-France. The details of the methodological description refer to version 1.12, which was implemented in the operational system on 13 September

10 Europe N_Africa S_America S_Africa Figure 2: The geographical areas for which the products are separately processed and distributed by the LSA SAF system. 10

11 2 Algorithm 2.1 Overview The down-welling surface short-wave radiation flux (DSSF) refers to the radiative energy in the wavelength interval [0.3µm, 4.0µm] reaching the Earth's surface per time and surface unit. It essentially depends on the solar zenith angle, on cloud coverage, and to a lesser extent on atmospheric absorption and surface albedo. The method for the retrieval of DSSF that is implemented in the LSA SAF system largely follows previous developments achieved at Météo-France in the framework of the OSI SAF (Brisson et al., 1999; Ocean & Sea-Ice, 2002). The main differences of the LSA SAF product are the spatial and temporal resolution, the source of ancillary input data, and the use of three short-wave SEVIRI channels (0.6µm, 0.8µm, and 1.6µm). Figure 3 shows a simplified flow chart of the algorithm. For clear and cloudy pixels quite different parameterisations are applied. The cloud mask therefore represents an important piece of information for the execution of the algorithm. In the presence of clouds the down-welling radiation reaching the ground is considerably reduced. The DSSF is strongly anti-correlated with the observable top-of-atmosphere reflectances: The brighter the clouds appear on the satellite images, the more radiation is reflected by them and the less radiation reaches the ground. Clear Cloud Mask Cloudy Geometry Atmosphere TOA Reflectance Factor Broadband Conversion, BRDF Model TOA Albedo Simplified Radiation Transfer Model Atmospheric Transmittance Cloud Transmittance Solar Constant Down-welling Surface Short-wave Radiation Flux Figure 3: Simplified flow chart of the DSSF algorithm. 11

12 2.2 Definition The down-welling surface short-wave radiation flux the spectral irradiance F is defined as the integral of E (λ) over the wavelength interval [λ 1 =0.3µm, λ 2 =4µm]: λ 2 F = E( λ) dλ. (1) λ 1 The spectral irradiance is the hemispherical angular integral of the down-welling spectral radiance L ( λ, θ, φ) weighted by the cosine of the zenith angle: 2ππ / 2 E ( λ) = L( λ, θ, φ)cos( θ )sin( θ ) dθdφ. (2) 0 0 It includes contributions owing to the direct solar radiation attenuated by the atmosphere as well as diffuse radiation. In the applied retrieval scheme the DSSF is approximated as F = F v( t) cosθ T 0 s, (3) F 0 where is the solar constant (with minor corrections according to the restricted wavelength interval considered), θ s the solar zenith angle, and T an effective transmittance of the atmosphere or cloud-atmosphere system. The factor v( t) = cos(2π t / 365) (4) takes into account the varying distance of the sun as a function of the day t of the year (Iqbal, 1983). For the effective transmittance T different expressions are used depending on whether a given pixel is marked as clear or cloudy. The information on cloud cover is provided by the cloud mask software that was developed by the NWC SAF and which is integrated in the LSA SAF operational system. Figure 4: Schematics illustrating some elements of the clear (left) and cloudy sky (right) DSSF estimation methods. 12

13 2.3 Clear Sky Method In the case of clear pixels the factor T is specified as n TA T = TA + TA ( AS AA ) =. (5) n= 1 1 AS AA T A represents the transmittance of the atmosphere and quantifies the contribution to the surface flux by the direct radiation as well as the diffuse radiation after scattering by the atmosphere as illustrated in the left diagram of Figure 4. The flux contribution owing to multiple scattering of the light between the surface and the atmosphere is taken into account by the denominator on the right-hand side of equation (5). A S denotes the surface albedo and A A the spherical albedo of the atmosphere. As indicated in the equation, the final form of this expression results from a geometric series taking into account an infinite number of scattering orders between surface and atmosphere. The atmospheric transmittance is calculated after Frouin et al. (1989) as with τ H2O τ O3 τ Aer+ CO2+ O2 T A e e e b H 2O τ H2O = ah 2 O W / cos s ) = (6) O3 ( θ, τ O3 = ao3( U O3 / cos θ s ) 1 b (7) and τ Aer + CO2+ O2 = ( a + ). cos θs V W is the water vapour column density in g/cm 2, U O3 the total ozone amount in atm.cm, and V the visibility in km. In the operational system the water vapour estimate is obtained from ECMWF numerical weather model forecasts and the ozone amount is specified according to the TOMS climatology. The visibility is currently kept at a fixed value of 20 km and the values of the parameters a were chosen according to the continental aerosol type. The spherical albedo of the atmosphere A A = a + b / V is also parameterised as a function of visibility according to Frouin et al. (1989). The numerical values of the various constants are listed in the table below. The value used for F 0 is slightly lower than the solar constant since a restricted wavelength interval is considered in the definition of DSSF and in the radiative transfer calculations which served for setting up the optical thickness parameterisations (7). The surface albedo A s is taken from the LSA SAF near real time albedo product (, 2005). For the time being the bi-hemispherical variant AS bh is used and the diurnal cycle is approximated by using the functional form suggested by Dickinson (1983) and Briegleb et al. (1986) as 1+ d AS = AS bh with d = (8) 1+ 2d cosθ s b and b 13

14 Table 1: Numerical values of the constants appearing in the parameterisations of Frouin et al. (1989). F W/m 2 a H H O b O3 O a b 0.57 O3 a b a b Cloudy Sky Method For cloudy pixels the DSSF estimate relies on a simplified physical description of the radiation transfer in the cloud-atmosphere-surface system according to Gautier et al. (1980) and Brisson et al. (1999). It is assumed that the whole image pixel is covered by a homogeneous cloud layer. The effective transmittance factor T is now given by TATC T =. (9) 1 ASTbc AC Compared to the clear-sky case the enumerator additionally includes the cloud transmittance T C which is the decisive quantity in this expression. The denominator has a similar significance as in equation (5) and quantifies multiple scattering between the surface and the bottom of the cloud layer. denotes the cloud albedo and cloud. T bc represents the atmospheric transmittance between the surface and the The cloud transmittance TC and albedo AC may be highly variable on small time scales depending on the daily evolution of the clouds. Their instantaneous values are determined from the satellite measurements with the help of a simple physical model. For this purpose the measured spectral reflectances in the 0.6µm, 0.8µm, and 1.6µm SEVIRI channels are first transformed to broad-band top-of-atmosphere albedo A TOA by applying the spectral conversion relations proposed by Clerbaux et al. (2005) and the angular reflectance model of Manalo-Smith et al. (1998). As illustrated in the right diagram of Figure 4, the signal at the top of the atmosphere comprises contributions owing to Rayleigh scattering by the atmosphere above the cloud ( ), radiation reflected by the cloud which is attenuated by the atmosphere A R above ( A C T SunCloudSat ), and radiation reflected by the surface which is attenuated by the 2 atmosphere and the cloud ( A T T ): S SunSurfaceSat C A C 14

15 2 ASTSunSurfaceSatTC ATOA = AR + ACT SunCloudSat +. (10) 1 ASTbc AC As before, the significance of the denominator in the third term is to take into account multiple scattering between the surface and the bottom of the cloud layer. Following Brisson et al. (1999) the transmittance factors T, T, and T, as well as A R are calculated with parameterisations given by Lacis and Hansen (1974). The cloud transmittance T is expressed in terms of the cloud albedo A and the cloud absorption a C as C SunCloudSat SunSurfaceSat C bc T C = 1 A a = 1 A α A. (11) C C C C The cloud absorption is modelled as a linear function of the cloud albedo by introducing the cloud absorption factor α. The currently employed numerical value of α = was not derived from first principles, but has been adjusted by matching the final flux estimates with the help of a validation data base (Ocean & Sea-Ice, 2005). This parameter therefore mainly serves for absorbing the methodological approximations and uncertainties, rather than for quantifying the physical cloud properties. Combining the expressions (10) and (11) allows us to calculate the two unknowns TC and AC from the observable ATOA by solving a quadratic equation, and finally to obtain the DSSF estimate with equations (9) and (3). The equation system gives a physical solution for T and A unless the observed value for A TOA is beyond the following limits: C C A = A + A T min TOA R S SunSurfaceSat corresponding to T C = 1 (12) max TSunCloudSa t A TOA = AR + corresponding to T C =0. (13) 1+ α If the limiting conditions for the top-of-atmosphere albedo are violated, the cloud transmittance is set to the limiting value and a quality flag is set accordingly. 15

16 3 Product Description 3.1 Continental Windows In the LSA SAF operational system the four geographical regions depicted in Figure 2 are processed separately. The main characteristics of these windows are listed in Table 2. The projection and spatial resolution correspond to the characteristics of the Level 1.5 MSG/SEVIRI instrument data. Information on geo-location and data distribution are available from the LSA SAF website: Table 2: Characteristics of the four LSA SAF geographical areas: Each region is defined by the corner positions relative to an MSG image of 3712 columns per 3712 lines with indices starting at 1 in the North and West. Region Name Description Initial Column Final Column Initial Line Final Line Size in Columns Size in Lines Total Number of Pixels Euro Europe NAfr SAfr SAme Northern Africa Southern Africa Southern America File Format The DSSF algorithm generates a single type of external product output files. The file names for the DSSF products derived from MSG respect the convention HDF5_LSASAF_MSG_DSSF_Region_YYYYMMDDhhmm where Region, YYYY, MM, DD, hh, and mm, respectively, denote the region name, year, month, day, hour, and minute of the start of image data acquisition. For distribution via EumetCast the prefix "S-LSA_-" is added. The LSA SAF products are provided in the HDF5 format developed by the NCSA (National Center for Supercomputing Applications) at the University of Illinois. A comprehensive description as well as libraries for handling HDF5-files in Fortran and C are available at A user friendly graphical interface to open and view HDF5-files can be downloaded from The HDF5-format permits the definition of a set of attributes for providing relevant information. Each LSA SAF product file includes the general attributes listed in Table 5 of Appendix B. Within the HDF5-files the information is organised in the form of separate datasets. For each dataset a set of additional attributes is available (Table 6 of Appendix B). 16

17 3.3 Temporal Reference The DSSF product is calculated for every second slot of MSG input data at intervals of 30 minutes. The estimates are derived for the instantaneous acquisition time of each image line. The SEVIRI image scans are performed from South to North. Hence in Northern Europe the line acquisition time deviates from the slot time by approximately 12 minutes. The difference Δ t between the acquisition time and the slot time is a linear function of the line position. It can be approximated as Δ t = 12' 40" 0.2"Y, (14) where Y is the line number in the full disk image with indices starting in the North. This relation is depicted in Figure 5 which also shows the dependence of Δt on the latitude (for a longitude of 0 ). Figure 5: The time difference between the slot and the acquisition time of the image line as a function of the line number (left) and the latitude for a longitude of 0 (right). 3.4 Product Content The DSSF product files comprise two datasets: one for the actual flux estimates and one for the quality flag. The technical properties are summarised in Table 3. Figure 6 and Figure 7, respectively, depict examples for the DSSF estimate and the corresponding quality flag derived from one slot of MSG observations. Table 3: Content of the DSSF product files. Parameter Quality Flag Dataset Name Unit Range Variable Type Scale Factor F DSSF W/m 2 2-Byte [0, 1500] 10 Signed Integer 1-Byte DSSF_Q_Flag na [0, 255] na Unsigned Int. 17

18 Figure 6: Re-composed full-disk image of the down-welling surface short-wave radiation flux estimate for the European window on the 15 th of February 2006 at 13:30 UTC. Figure 7: Quality flag corresponding to the DSSF estimate depicted in Figure 6. The signification of the bit codes is given in Table 4. xxx indicates that the most significant bits 5 to 7 were used for visualisation while xx represents the least significant bits 0 to 1. 18

19 The signification of the numerical values of the quality flag is explained in Table 4. Bits 0 and 1 propagate the land/sea mask information. The bits 2 to 4 include the cloud mask flag as provided by the respective input file (cf. Nowcasting SAF, 2004). Finally bits 5 to 7 contain information about the method applied to calculate the DSSF estimate and possible problems encountered. Table 4: DSSF quality flag information. Bit Binary Code Description Bits 0-1 Land/Sea Mask 00 Ocean Land 10 Space (Outside of MSG disk) Continental Water Bits 2-4 Cloud Mask 000 Unprocessed 001 Clear 010 Contaminated 011 Cloud Filled 100 Snow/Ice 101 Undefined Bit 5-7 DSSF Algorithm 000 Cloudy Sky Method 001 Cloudy Sky Method, A TOA below lower limit of equation (12) 010 Cloudy Sky Method, A TOA above upper limit of equation (13) 011 Algorithm Failed 100 Clear Sky Method 101 Night 110 Beyond specified View Angle Limit 111 Not Processed (Cloud Mask undefined) ++ For ocean and space pixels the values of the bits 2 to 7 are zero. 19

20 3.5 Summary of Product Characteristics Product Name: Product Code: Down-welling Surface Short-wave Radiation Flux DSSF Product Level: Level 3 Product Parameters: Coverage: Packaging: Sampling: Spatial Resolution: Projection: Units: W/m 2 Range: Format: MSG full disk (Continental pixels) Europe, Northern Africa, Southern Africa, South America pixel by pixel basis MSG/SEVIRI full resolution (3km 3km at nadir) MSG/SEVIRI Level 1.5 data projection (technical upper limit) 16 bits signed integer (DSSF estimate) 8 bits (quality flag) Frequency of Generation: every 30 minutes Size of Product Files: Additional Information: 0.7Mb 3.5Mb (depending on the continental window and the compression efficiency) Identification of bands used in algorithm: MSG/SEVIRI VIS 0.6 MSG/SEVIRI NIR 0.8 MSG/SEVIRI SWIR 1.6 Identification of ancillary and auxiliary data: Land/Sea mask Cloud Mask (from NWC SAF software) View Azimuth and Zenith Angles (from LSA SAF System) Solar Azimuth and Zenith Angles (from LSA SAF System) Total Column Water Vapour (from ECMWF model) Ozone Content (from TOMS climatology) Land Surface Albedo (LSA SAF Product) 20

21 4 Validation Validation studies have been performed by comparing the product with ground measurements of the down-welling radiation flux. The presented results are based on the Baseline Surface Radiation Network (BSRN) stations of Carpentras (France) and Toravere (Estonia) for which data concomitant with the product time series were already available. In addition, the in-situ data from ground stations run by the LSA SAF project in Évora (Portugal) and by Météo-France in Roissy (France) have been used. Note that during the validation period from October 2004 to August 2006 successive algorithm versions were running in the operational system. 4.1 Daily Cycle In general a good agreement between the satellite estimates and the in-situ data can be observed when comparing the daily time series. A few examples are shown in Figure 8. The cases for Carpentras and Évora presented in the figure are typical for clear sky situations. For cloudy sky conditions the day chosen for Toravere represents a favourable example. In the unfavourable case depicted for Roissy with a rather large dispersion, the discrepancies cannot entirely be attributed to deficiencies of the retrieval method. The example also illustrates the limitation of the validation approach when the conditions are highly variable in space and time. At least part of the dispersion is a consequence of comparing a local measurement with an estimate for a rather extended image pixel. For a quantitative analysis, the exact acquisition times of the satellite measurements need to be taken into account (cf. Section 3.3). For the ground stations of Carpentras and Toravere the in-situ data were available with a high temporal resolution in the order of minutes. In this case the in-situ data were averaged over intervals of 15 minutes centred on the exact acquisition times of the respective satellite measurements as a function of the site coordinates. For the other sites the accessible data had already been averaged over certain time intervals. In this case the data points were linearly interpolated and re-sampled corresponding to the exact reference times of the satellite product. 21

22 Figure 8: Examples for daily time series of DSSF estimates and in-situ measurements at the ground validation stations. The colour code of the dots is the same as for the quality flag in Figure Scatter Plots Figure 9 depicts scatter plots of the LSA SAF estimates against the in-situ measurements for the whole validation period. The top left plot includes all data points for Carpentras for which the clear sky method was applied and the top right plot all cloudy data points for Roissy. As expected, the dispersion of the distribution is much smaller for the clear than for the cloudy data points. The biases are relatively small in both cases and there is no significant evidence for a dependence of the bias on the level of the DSSF estimate. A few outliers can be perceived in the scatter plot for clear sky. However, their number is quite small compared to the large amount of data points included in the graph. The outliers may be caused by small clouds which obstructed the direct solar radiation but could not be detected on the pixel scale. 22

23 Figure 9: Scatter plots between the satellite estimates and the ground measurements. The colour code of the dots is the same as for the quality flag in the previous figures. The bottom right plot shows the daily averaged estimates as explained in the text. (Note that the shown range of values is different in this case.) The bottom left plot of Figure 9 includes all available data points for the site of Évora. For validation purposes we also calculated daily averages of the LSA SAF DSSF product for the pixels corresponding to the validation sites. This is helpful for comparing the quantitative validation statistics to those of other products which are not available as instantaneous estimates. The daily values are determined by averaging all available (day-time) LSA SAF DSSF estimates for a given day. For comparison only the in-situ measurements corresponding to the product time slots actually used for the determination of the daily DSSF product are then also averaged to obtain the corresponding daily averaged in-situ measurement. (Note that this prescription is useful only for validation purposes, but not appropriate for generating a daily averaged DSSF product meant to be distributed and utilised. For this purpose the problems of temporal reference for the average and the treatment of missing data would have to be considered much more carefully.) The resulting data points for Évora are depicted in the bottom right plot of the figure. As expected the dispersion is much smaller than for the instantaneous radiation estimates. 23

24 4.3 Validation Statistics For quantifying the validation results, the bias defined as the average of the difference between the LSA SAF estimate and the in-situ measurement as well as the standard deviation of that difference are considered. The temporal evolution of these statistical quantities for all four stations combined over the whole validation period is shown in Figure 10. The position of the symbols in the graphs indicates the bias, and the length of the bars (from the centre to each end) corresponds to the standard deviation as defined above. Monthly sub-samples of the validation data points are considered in order to illustrate a possible temporal evolution of the product quality. From the top left to the bottom right the panels show the results for the data points processed with the clear sky method, for the cloudy sky method, for all processed day-time data points combined irrespective of the method applied, and for the daily averaged DSSF values which were calculated for validation purposes only as described above. The top left plot for clear sky also includes the bias values (but not the standard deviation) for morning and afternoon data points separately. Considering the whole validation period and all sites the standard deviation of the instantaneous validation data points is in the order of 40 Wm -2 for clear sky and 110 Wm -2 for cloudy sky. Combining all (instantaneous) data points the standard deviation amounts to roughly 85 Wm -2. It reduces to 30 Wm -2 for the daily averaged values. Taking into account the whole validation data set there is a small positive bias of about 5 Wm -2. However, a slight seasonal trend can be perceived especially for the clear sky case. The standard deviation of the monthly bias values depicted in Figure 10 is about 10 Wm -2 to 15 Wm -2 for the clear and cloudy cases and in the order of 5 Wm -2 for the quantities referring to all data points and the daily average. Figure 10: Temporal evolution of bias and standard deviation between the LSA SAF estimates and ground measurements for all validation sites combined (Carpentras, Roissy, Évora, and Toravere) for the period from October 2004 to August

25 5 References Briegleb B.P., P. Minnis, V. Ramanathan and E. Harrison, 1986, Comparison of Regional Clear-Sky Albedos Inferred from Satellite Observations and Model Computations, Journal of Applied Meteorology, 25, 2, Brisson A., Le Borgne P., Marsouin A., 1999, Development of Algorithms for Surface Solar Irradiance retrieval at O&SI SAF low and Mid Latitude, Météo-France/CMS, Lannion. Clerbaux N., Bertrand C., Caprion D., Depaepe B., Dewitte S., Gonzalez L., Ipe A., 2005, Narrowband-to-Broadband Conversions for SEVIRI, Proceedings of the 2005 EUMETSAT Meteorological Satellite Conference, Dubrovnik, Dickinson R.E., 1983, Land surface processes and climate Surface albedos and energy balance, Advances in Geophysics, 25, Frouin, R., D. W. Lingner, C. Gautier, K. S. Baker, and R. C. Smith, 1989, A simple analytical formula to compute clear sky total and photosynthetically available solar irradiance at the ocean surface, J. of Geophys. Res., 94, Gautier C., Diak G., Masse S., 1980, A simple physical model to estimate incident solar radiation at the surface from GOES satellite data, J. Climate Appl. Meteor., 19, Iqbal M., 1983, An Introduction to Solar Radiation, Academic Press, 390pp. Lacis A.A., J.E. Hansen, 1974, A parameterization for the absorption of solar radiation in the Earth's atmosphere, J. Atmos. Sci. 31, , 2005, Land Surface Albedo Product User Manual, Version 1.3. Manalo-Smith N., G.L. Smith, S.N. Tiwari, W. F. Staylor, 1998, Analytic forms of bidirectional reflectance functions for application to Earth radiation budget studies, Journal of Geophysical Research, Vol. 103, D16, pp. 19, , 751. Nowcasting SAF, 2004, User Manual for the PGE of the SAFNWC/MSG: Scientific Part, Version 1.0. (latest version: 1.2, 2005) Ocean & Sea Ice SAF, 2002, Surface Solar Irradiance Product Manual, Version 1.2. Ocean & Sea Ice SAF, 2005, Surface Solar Irradiance Product Manual, Version

26 Developers The development and implementation have been carried out under the responsibility of the Centre National de Recherches Météorologiques (CNRM) de Météo-France (MF). Authors: Bernhard Geiger, Catherine Meurey, Dulce Lajas, Laurent Franchistéguy, Dominique Carrer, and Jean-Louis Roujean Appendix A. AL: ATBD: AVHRR: BRDF: CNRM: cwv: DSSF: ECMWF: EPS: EUMETSAT: GOES: HDF: IM: NIR: METEOSAT: METOP: MISR: MF: MSG: MTR: na: NWC: NWP: PAR: OSI: SAF: SEVIRI: SMAC: SPR: SVPD: SVVP: TOC: TOA: TOMS: URD: Glossary Land Surface Albedo Product Algorithm Theoretical Basis Document Advanced Very High Resolution Radiometer Bi-directional Reflectance Distribution Function Centre National de Recherches Météorologiques column water vapour Down-welling Surface Short-wave Radiation European Centre for Medium-Range Weather Forecast EUMETSAT Polar System European Meteorological Satellite Organisation Geo-stationary Operational Environmental Satellite Hierarchical Data Format Instituto de Meteorologia (Portugal) Near Infrared Radiation Geostationary Meteorological Satellite Meteorological Operational polar satellites of EUMETSAT Multi-Angle Imaging Spectra-Radiometer Météo-France Meteosat Second Generation Mid Term Review not applicable NowCasting Numerical Weather Prediction Photosynthetically Active Radiation Ocean and Sea Ice Satellite Application Facility Spinning Enhanced Visible and Infrared Imager Simplified Method for the Atmospheric Correction Scientific Prototyping Report Scientific Validation Plan Document Software Validation & Verification Plan Top of Canopy Top of Atmosphere Total Ozone Mapping Spectrometer User Requirements Document 26

27 Appendix B. HDF5-Attributes The set of general attributes common for all LSA SAF files and their possible values presently assigned for the DSSF product are described in the table below. Table 5: General HDF5 attributes. Attribute Description Data Type Value for DSSF Product Files SAF SAF package String LSA CENTRE Institution (generating/disseminating data) String MF ARCHIVE_FACILITY Centre where the data is archived String IM-PT PRODUCT Defines the name of the product String DSSF PARENT_PRODUCT_NAME Array of up to 4 product names, upon which the product is based String Array(4) CMa, ALBEDO, WV, SZA/SAA SPECTRAL_CHANNEL_ID PRODUCT_ALGORITHM_VERSION Channel Identification (1 bit per channel, where LSB is HRV and MSB is IR13.4; values are 0 if not used, 1 if used.) Version of the Algorithm that produced the product Integer 14 String 1.12 CLOUD_COVERAGE Indicator of the cloud coverage in the product String NWC-CMa OVERALL_QUALITY_FLAG Overall quality flag for the product String OK or NOK ASSOCIATED_QUALITY_INFORMATION Several miscellaneous quality indicator for the product String - REGION_NAME Processed Region Name String COMPRESSION Compression Flag Integer Euro, NAfr, SAfr, or SAme 0 Uncompressed 1 Compressed FIELD_TYPE Data filed type String Product FORECAST_STEP Forecast Step in Hours Integer 0 NC Number of columns Integer Depends on Region NL Number of lines Integer Depends on Region NB_PARAMETERS Number of datasets Integer 2 NOMINAL_PRODUCT_TIME Production Time String YYMMDDhhmmss SATELLITE Platform identifier (mission and spacecraft the product originated from) String Array(10) MSG1 or MSG2 27

28 Attribute Description Data Type Value for DSSF Product Files INSTRUMENT_ID Instrument which acquired the product or data used by the product String Array(10) SEVI INSTRUMENT_MODE Scanning mode of the instrument at the time of the acquisition.satellite Identification String STATIC_VIEW IMAGE_ACQUISITION_TIME Image Acquisition Time (SEVIRI 1.5 Images) String YYMMDDhhmm ORBIT_TYPE Coverage of the product (only for EPS) String GEO PROJECTION_NAME Projection name and sub-satellite point String GEOS(+000.0) NOMINAL_LONG Satellite Nominal Longitude Real as in Level 1.5 data NOMINAL_LAT Satellite Nominal Latitude Real as in Level 1.5 data CFAC Column Scaling Factor (SEVIRI 1.5 Images) Integer as in Level 1.5 data LFAC Line Scaling Factor (SEVIRI 1.5 Images) Integer as in Level 1.5 data COFF Column Offset (SEVIRI 1.5 Images) Integer Depends on Region LOFF Line Offset (SEVIRI 1.5 Images) Integer Depends on Region START_ORBIT_NUMBER First of two orbit numbers in the EPS product, valid at the starting of the sensing, i.e, at the beginning of a dump Integer 0 END_ORBIT_NUMBER SUB_SATELLITE_POINT_START_LAT SUB_SATELLITE_POINT_START_LON SUB_SATELLITE_POINT_END_LAT SUB_SATELLITE_POINT_END_LON SENSING_START_TIME SENSING_END_TIME Final of the orbit numbers in the EPS product, valid at the ascending node crossing, i.e. towards the end of a dump Latitude of sub-satellite at start of acquisition Longitude of sub-satellite at start of acquisition Latitude of sub-satellite at end of acquisition Longitude of sub-satellite at end of acquisition UTC date & time at acquisitions start of the product UTC date & time at acquisition end of the product Integer 0 Real 0.0 Real 0.0 Real 0.0 Real 0.0 String - String - PIXEL_SIZE For image products, size of pixel at nadir. For meteorological products resolution/accuracy String 3.1km GRANULE_TYPE Type description of the item String DP 28

29 Attribute Description Data Type Value for DSSF Product Files PROCESSING_LEVEL Processing Level Applied for generation of the product String 3 PRODUCT_TYPE Abbreviation name for the product type rather product category String LSA DSSF PRODUCT_ACTUAL_SIZE Actual size of the product String Depends on Region PROCESSING_MODE Processing mode for generation of the product String N, B, R, or V DISPOSITION_FLAG Disposition status indicator of the product, as set by the UMARF operator String O, T, or C TIME_RANGE Temporal Resolution String frequency: every 30 min STATISTIC_TYPE Statistic Type String instantaneous LSB Lower Significant Bit MSB Most Significant Bit YY - Year; MM-Month; DD Day; hh Hour; mm Minute; ss Second String => Character (len=80) Integer => Integer (kind=4) Real => Real (kind=8) 29

30 The attributes for each dataset of the HDF5-files are described in the following table. Table 6: Dataset attributes. Attribute Description Data Type Value for Dataset DSSF Value for Dataset DSSF_Q_Flag CLASS Dataset type String Data Data PRODUCT PRODUCT_ID Defines the name of the product Product identification accordingly with WMO tables String DSSF Q-Flag Integer N_ COLS Number of columns Integer N_ LINES Number of lines Integer Depends on Region Depends on Region Depends on Region Depends on Region NB_BYTES SCALING_FACTOR OFFSET Number of bytes per pixel Scaling factor for the parameter Offset of the scaling factor Integer 2 1 Real Real MISS_VALUE Missing value Integer UNITS Parameter Units Integer W/m 2 N/A CAL_SLOPE Calibration Constant Real CAL_OFFSET Calibration Constant Real