ITALIAN AIR FORCE METEOROLOGICAL SERVICE AND MÉTÉO FRANCE UPDATING AND DEVELOPMENT OF METHODS FOR WORLDWIDE ACCURATE MEASUREMENTS OF SUNSHINE DURATION WMO TECO, Brussels, Belgium, October 2012 Vuerich Emanuele (IMS) 1

Contents  Motivation and purpose  Procedures and methods  Data analysis and results  Conclusions 2 Objective To provide the scientific community with the results of an all-seasons bilateral inter-comparison on sunshine duration (SD) measurement methods that has been recently conducted by the Italian Meteorological Service (IMS-AF) and Météo France (MF) during the period

Motivation and purpose 1. Motivation Need for cost-effective solutions with suitable accuracy for the climatology and solar energy applications; a transition from no-more-recommended SD methods to methods with improved achievable accuracy; a standardized calibration method for SD measurements is not yet available and agreed in the community need to verify the applicability of the CIMO Guide requirements: CIMO Guide being SD PERIOD [h day -1 ] =ΣSD SUB-PERIODS if I ≥ I THR = 120 W m -2, than: 1.a “threshold accuracy” of 20% is admitted ( I Є [96;144] W m -2 ); 2.SD uncertainty should be ±0.1 h day -1 3.SD resolution should be 0.1 h Are such requirements applicable? 3

Motivation and purpose 2. Intercomparison objectives IMS and MF started a bilateral intercomparison that has been carried out in two different climatological locations (Vigna di Valle, Italy and Carpentras, France) to achieve the following objectives: Primary. Evaluation of the daily achievable uncertainty of SD measurements: pyranometric methods (global irradiance),1-min- averaged direct solar irradiance and network detectors. Secondary. Global use of a pyranometric method for SD by using BSRN (Baseline Surface Radiation Network) data ( for time reason this secondary argument is not treated here: see Vuerich et al. – TECO2012 and Morel et al. - BSRN meeting, Postdam, 1-3 August 2012 ) 4

Procedures and methods 1. Intercomparison references and instruments 5 Carpentras, France (2011) Messrs Morel and Mevel at work Reference: Pyrheliometer: Epply NIP Instruments : Pyranometer: K&Z CM11 Network SD sensor: C&S recorder (burn method) Reference: Pyrheliometer: K&Z CH1 Instruments : Pyranometer: K&Z CM11 Network SD sensor: CE181 (scanning by optical fiber) Primary reference: absolute TMI pyrheliometer (RRC Carpentras)

Procedures and methods 2. SD measurements methods (1/3) Reference method (SD REF ): pyrheliometric method based on 1 second direct irradiance (I) measurements (to reduce the uncertainty due to questionable sunshine minutes in case of using 1 minute averages of I) SD using 1 min avg of I and applying the “threshold accuracy” Three SD measurements by the pyrheliometer: (1) SDpyrh 1m, minutes of SD if the 1 min average of I ≥ 120 W m -2 ; (2)SDthr 96, seconds of SD if I ≥ 96 W m -2 ; (3)SDthr 144, seconds of SD if I ≥ 144 W m -2 ; Pyranometric methods by using global irradiance, G:  Step Algorithm (SA): minutes of SD using 1-min averaged G compared with a “rough” threshold (G THR = 0.4 Go, with Go = I 0 sin(h) ). 6 If G ≥ G THR then SD SA = 1 minute, otherwise SD SA = 0 minute

Procedures and methods 2. SD measurements methods (2/3)  Carpentras method (Olivieri,1998) or Météo-France Algorithm (MFA): minutes of SD through the measurement of 1-min averaged global irradiance (G) compared with an accurate threshold value. 7 Carpentras, France (2009) Messers Morel, Olivieri, Didier, Vuerich If G ≥ G Seuil then SD MFA = 1 minute, otherwise SD MFA = 0 minute where: - G Seuil = F c 1080 (sin(h)) 1.25 (model) - h ≥ 3° (data filtering) -F c = A + Bcos(2πd/365) with: - F c representing a fraction of global irradiance in clear sky in mean conditions of atmospheric turbidity; - h being the elevation angle of the sun in degrees; - d being the day number of the annual sequence; F c factor depends on the climatic conditions of the location and A,B coefficients can be empirically calculated through a long term comparison with SD measurements by means of a pyrheliometer. For Vigna di Valle and Carpentras: A=0.73 and B=0.06

Procedures and methods 2. SD measurements methods (3/3)  Slob and Monna Algorithm (SM): algorithm to calculate daily SD from the sum of 10 minutes SD which implies the use of 10 minutes average of G and the use of its maximum and minimum values during the 10 min interval. The procedure to apply the algorithm is documented in the Annex of Chapter 8 of the CIMO Guide (WMO, 2008) and by Hinssen and Knap (2007).  SD network detectors: SD sensors respectively operated by the Italian Met Service (IMS) and Météo-France (MF) radiation networks. 8

Data analysis and results 9 The data have been analyzed by: - scatter plots of SD x versus SD REF (where x is the measuring principle used) with linear fits, - dispersion plots of SD daily differences (SD x – SD REF ) with normal distribution fits, and by determing pertinent parameters for estimating the daily achievable uncertainty of each method and the applicability of CIMO guide requirements BIAS (trueness) St.dev (precision, repeat) R 2 (goodness of fitting) Skewness (symmetry)

10 SD pyrh 1m versus SD REF (SD pyrh 1s )

11 SD trh96 and SD trh144 versus SD REF (SD pyrh 1s )

12 SD SENSOR CE181 (MF) SD SENSOR C&S (IMS)

13 SD SM versus SD REF (SD pyrh 1s )

14 SD SA versus SD REF (SD pyrh 1s )

15 SD MFA versus SD REF (SD pyrh 1s )

RESULTS (Carpentras) 16 METHODUncertainty* [hday -1 ]Npts [%] in range {-0,1 ; +0,1} [h/d] SD REF (1s)± 0.06/ SD pyrh1 min± % Threshold accuracy± 0.42Not applicable CE181 sensor± % MFA (best G method) ± % SA[-1.37; +1.71]21.5% Slob&Monna (Worse G method) [-1.41; +2.07]17,4% * Calculated at 95% of confidence level (normal distribution), otherwise the interval with the 95% of samples CIMO Guide target uncertainty

RESULTS (Vigna di Valle) 17 METHODUncertainty* [hday -1 ]Npts [%] in range {-0,1 ; +0,1} [h/d] SD REF (1s)± 0.06/ SD pyrh1 min± % Threshold accuracy± 0.45Not applicable C&S recorder[-2.35; +1.69]15.5% MFA (Best G method) ± % SA[-1.36; +1.68]18.2% Slob&Monna (Worse G method) [-1.45; +2.43]13,8% * Calculated at 95% of confidence level (normal distribution), otherwise the interval with the 95% of samples CIMO Guide target uncertainty

Summary and conclusions The SD REF must be calculated by 1s I instead of 1-min averaged I; The achievable uncertainties of pyranometric methods are not comparable with the CIMO Guide target uncertainty and the threshold tolerance must be reviewed; The CIMO Guide requirement is not an appropriate target for routine operational measurements and should be reviewed; The CIMO Guide should be reviewed for taking into account additional algorithms for estimating SD (such as MFA) and for updating the typical achievable uncertainty of those methods; An IOM report on this bilateral intercomparison is expected very soon for extensively providing the results and the achievements. ET on II is investigating the possibility to organize a large and all- seasons intercomparison for all classes of radiation instruments in Europe with the cooperation of PMOD/WRC and sponsorship of interested countries. Is the time to seriously think about that? 18

19 … any successful enterprise starts with a group photo and around table! Italy and France Contacts:

Extra data analysis and results 20 Improvement and global use of MFA by BSRN data: - Concept: MFA algorithm was also extended to nine BSRN stations by using 1-min average global and direct irradiances for at least 4 consecutive years. - Purpose: determination of the best set of A and B coefficient that minimize the total relative error of SD over a long period of time (years) and a method for an universal application of the MFA for estimating SD from global irradiance at all latitudes. - Technique: consists in an empirical method that permits to select the A,B from the plot of the cumulative difference between the SD from MFA and SD from 1 min average of direct irradiance (assumed as reference because available from BSRN data). - Results: recently presented at the BSRN meeting in Postdam, 1-3 August 2012.

21 BSRN STATIONLatitudeABErr % MOMOTE -20,670-0,1 TAMANRASSET 220,770-0,2 TATENO 360,730,05-0,2 BOULDER 400,670,060,1 CARPENTRAS 440,710,050,1 PAYERNE 470,750,060,4 PALAISEAU 480,750,040,2 CABAUW 520,770,060,2 TORAVERE 580,740,060,4