Previous: Data and method description Next: EW irradiance variations due to ozone and cloudiness Up: Ext. Abst.

3. Validation of the method for retrieving UV irradiance variations

The proposed method was verified on the base of two datasets with a lot of ancillary information: broadband UVB measurements in Moscow (Meteorological Observatory of Moscow State University, MO MSU) and NSF UV spectroradiometer network data.

3.1 Validation according to Moscow broadband UVB data.

During 1994-1996 period the EW irradiance was measured by temperature stabilized Biometer 501 with a solar angle dependent calibration factor obtained in Finland in 1995 during WMO/STUK intercomparisons. Since 1999 the UVB measurements were resumed by UVB-1 Pyranometer (Yankee Environmental Systems Inc.) which has been checked at Colorado UV facility which is guided by the joint efforts of Natural Resource Ecology Laboratory of Colorado State University and NOAA. The calibration of UVB-1 signals in the units of EW irradiance was implemented using the method proposed in Lantz [1999]. Fig.1 shows a good agreement between the estimated and measured monthly means EW irradiance in snow-free conditions for both periods; correlation coefficient comprises R²=73%.

We also use TOMS data to compare the retrievals of EW irradiance over Moscow. Fig.2 shows a coincidence between the EW irradiance obtained from satellite and by using the proposed method. R²=87% if we use all ancillary data including the aerosol and cloud optical thickness to retrieve EW irradiance (see violet line in Fig.2) and R²=78% if we use only ozone and CQ_g parameter (see black line with squares). A good agreement in Fig. 1 and in Fig.2, i.e. between the measured EW irradiance and the EW estimated from ground and satellite data may be explained by a good agreement in input parameters –cloudiness and ozone. It is known that there is a good agreement between ground and satellite retrievals of total ozone. That is why we focused on the examination of cloud parameter. We used the data of 19 stations in different geographical areas to check the similarity in interannual variations of CQ_g parameter, total cloud amount (NA) and the CQ obtained from satellite. Fig.3 illustrates these comparisons over different geographical regions. There is a good agreement in relative year-to-year changes between CQ TOMS and ground CQ_g values, but CQ TOMS are systematically 5-20% lower due to aerosol attenuation, which is considered together with cloud attenuation according to the TOMS algorithm. Table 1 shows determination coefficients, R²_, which are calculated between CQ TOMS and NA, between CQ TOMS and CQ_g values. We received a better agreement with the CQ_g values and in some cases the difference in R² is very high due to prevailing thin upper cloudiness in several regions, which has been accounted for while calculating of CQ_g. On the other hand, a good correlation between interannual variation of CQ TOMS and CQ_g obtained as a weighted cloud amount parameter is an evidence, that it is cloud amount variation that plays a main role in CQ TOMS variability while variations in aerosol and cloud optical thickness make up only a few percents (R² =83% if we add the effects of cloud optical thickness and R² =87% if we add aerosol fluctuations (see Table 1)). At the same time we understand that other atmospheric factors may be of more importance if day-to-day variability of CQ TOMS values are discussed.

Fig.1. Measured and estimated EW irradiance variability. Snow-free conditions, Moscow 1995-2000.

Fig.2. Comparisons between EW irradiance variations obtained from TOMS data and estimated using ground ancillary information. May-September period, Moscow.

Table 1

3.2 Validation according to NSF spectral dataset.

In order to compare UVB retrievals in other geographical conditions we used the NSF spectral dataset which is supported with a lot of ancillary information and is available via bsimail.biospherical.com/nsf. The analysis was made for CIE EW irradiance and for Setlow dose in snow free conditions over Ushuaia, San Diego and Barrow sites. Fig.3 illustrates a good agreement in estimated and measured EW irradiance variability (R²=89% for Ushuaia and R²=82% for San Diego). For Setlow dose the correlation is less (R² is about 60%) presumably due to more sensitivity to other atmospheric parameters. Figure 4 illustrates an agreement between interannual variations in CIE EW irradiance and in Setlow dose as well as distinct upward trends obtained from both TOMS data and the NSF measurements mainly due to ozone loss over Ushuaia. The effect of ozone loss was enhanced by growth in CQ transmittance during the last years (see Fig.4a). There is also a good agreement in TOMS and NSF data for Barrow and San Diego sites. The analysis of year-to-year variations in EW irradiance and Setlow doze for these sites has not revealed any trend.

Fig.3. Measured and calculated variability of CIE EW irradiance in Ushuaia (a, austral summer) and in San Diego (b, boreal warm period).

Fig. 4. Variability of biologically active irradiance according to NSF measurements and TOMS data (in black) in Ushuaia (a,b) and San Diego( c,d). CIE EW irradiance is shown by red line with triangles (a,c) and Setlow weighted irradiance - by yellow line with squares (b,d). EW variability due to cloudiness (blue line) and due to ozone variations (pink line) are shown in 5a,c. Summer periods.