Introduction

In Phase 3 of PROMOTE the surface UV algorithm developed during Phase 2 has been applied to the data covering period of July 1983 to June 2007. In Phase 3 Vitamin D weighted daily UV doses and Daily maximum UV index has been calculated in addition to Erythemal daily doses.

The algorithm uses the cloud data from the International Cloud Climatology Project (ISCCP). This data is available on a global 280 km equal area grid, providing cloud fraction and cloud optical depth, of the cloud-covered part of the pixel, with a temporal resolution of 3 h. The total ozone is taken from the Ozone Record Service by KNMI. Climatological aerosols as provided by the Global Aerosol Data Set (GADS) are used for aerosol input. More details on the algorithm are available here.

The Phase 3 final data set has been validated against high quality UV data by ground-based instruments at eight European stations. Seven of these stations (Sodankyla and Jokioinen in Finland, Norrkoping in Sweden, Bilthoven in the Netherlands, Lindenberg in Germany, Hradec Kralove in the Czech Republic, and Thessaloniki in Greece) were chosen because their data are conveniently available within the SCOUT-O3 EC funded project, where FMI is also participating. To improve areal scope, the data from Reading in UK was also used in the validation.

In practice, the aim of the validation is to determine typical bias and spread of the satellite-retrieved UV data as compared to the ground-based reference. Most reference stations have data starting from the early or mid 1990s. However, the station of Norrkoping has data available already since 1983, and will thus make an important contribution to validating the performance of the complete long-term satellite-retrieved UV record.

Because in 1980's very few UV measurements are available so-called reconstructed UV data produced in the SCOUT-O3 project (submitted manuscript: Reconstructing UV for the past 4 decades using models and measurements, P.N. den Outer et al.) has also been used in the validation. From this work so-called "Best Estime" of Erythemal daily doses based on five different reconstruction models has been used in the validation of Promote Phase 3 final data at above-mentioned seven stations included in the the SCOUT-O3 project.

The following table shows the availability of validation data:

Station	Erythemal Daily Doses	Reconstructed Erythemal Daily doses	Vitamin D Daily doses	Daily maximum UV index
Sodankyla	1990-2006	1983-2006	1990-2007	1990-2006
Jokioinen	1995-2006	1983-2006	1995-2007	1995-2006
Norrkoping	1983-2007	1983-2006	-	-
Bilthoven	1996-2006	1983-2006	-	1996-2006
Reading	2003-2005	-	2003-2007	-
Lindenberg	1995-2007	1983-2006	-	1995-2007
Hradec Kralove	1996-2006	1983-2006	-	-
Thessaloniki	1991-2006	1983-2006	1991-2004	1991-2006

Description of the validation method

For each station, we examined the following quantities in order to assess the performance of the satellite-retrieved data:

• mean bias = mean value of percentage deviation as compared to reference data, i.e., mean([est-ref]/ref*100%)
• median bias = median value of percentage deviation as compared to reference data, i.e., median([est-ref]/ref*100%)
• std = standard deviation of ratio of estimated to reference value, i.e., std(est/ref)
• W20 = amount (relative frequency in %) of values found within plus/minus 20% from the reference

Results

The validation results for the different UV products are presented here:

• Erythemal Daily Doses
• Vitamin D Daily Doses
• Daily Maximum UV Index

Discussion

The PROMOTE UV Record provides global and long-term information on the surface UV radiation. During the development of the data set the aim has been to ensure homogeneity when combining data from multiple satellite sensors. In order to meet this aim, we have chosen to use cloud data of the International Satellite Cloud Climatology Project (ISCCP) in combination with a multi-sensor, assimilated record of the total ozone column. The validation shows that the method is working reasonably, although there is a general tendency for overestimation when the data is compared with ground-based instruments.

Daily doses

At most sites the satellite UV daily doses have positive bias in comparison with the ground-based measurements and with the reconstructed UV data. It should be noted that the satellite-retrieved and ground-based dose values are not fully comparable, and also the ground-based data involves some uncertainty. Because of the relatively large ground pixel size (the ISCCP cloud data is originally in an 280km equal area grid), the satellite-retrieved dose values represent an average dose for a larger region than those based on the ground-based measurements. A more detailed analysis (not shown here) shows that indeed some of the overestimation in the satellite algorithm can be attributed to the cloud information in the ISCCP data, which gives too weak cloud attenuation in particular during days with heavy clouds according to the ground measurements. Another obvious factor for the positive bias is absorbing aerosols that are not currently taken into account by the surface UV retrieval algorithm.

Daily maximum UV Index

For a given location the Daily maximum UV Index is instantaneous quantity that can be a sunny moment during an otherwice cloudy day. If ground-based instrument makes measurements with high frequency the possibility of detection of such a sunny moment increases. High quality UV spectroradiometers are often slow-scanning instruments and one spectral scan can take up to 5-10 minutes. For practical reasons many instruments make UV scans with very different frequency. E.g. the data available in this work is measured roughly every 30-60 minutes in Jokioinen and Sodankyla, every 12 minutes in Bilthoven, and every 60 minutes in Thessaloniki. In Lindenberg during yearly years measurements were made roughly every 60 minutes, thereafter every 30 minutes, and since 2005 every minute.

The validation of Daily maximum UV Index has some extra challenges because of the nature of this quantity and because of the differences in the representativeness of satellite and ground-based data. In this work the satellite derived Daily maximum UV Index is calculated as a average of cloudy and cloud-free parts of the grid-box using time step of one hour. As a result the maximum UV Index is affected by cloudy part of the grid-box unless the grid-box is totally cloud-free. Usually this leads to underestimated satellite data under partly cloudy situations if the data is compared against ground-based data with high measurement frequency.

These differences are seen in the validation results. In general, the satellite data still has positive bias in Jokioinen, Sodankyla, and Thessaloniki where measurement frequency is low. However, in Bilthoven and Lindenberg with high measurement frequency the satellite data has negative bias.

Conclusions

Satellite derived Daily UV Doses (Erythemal and Vitamin D) are typically overestimates when compared against ground-based instruments. Part of the overestimation is due to the limited aerosol input, but part of it is related to the limitations of cloud input for the UV calculations. The comparision of satellite derived Daily maximum UV Index against ground-based measurement is difficult because of the nature of this quantity and the different field-of-view of satellite and ground-based instruments. Also differences in temporal resolution of ground-based data affect the results. The users of satellite data should take into account the representativeness of this data. A special care is needed if Daily maximum UV Index is applied locally.

Jussi Kaurola, 31 August 2009