Uvmfrsr Data Processing

The UV-MFRSR uses independent interference filter-photodiode detectors and an automated rotating shadow band to measure the total horizontal and diffuse horizontal UV solar irradiance at seven wavelengths concurrently (Michalsky et al., 1988; Harrison et al., 1994). Direct normal irradiance is determined within the data logger by subtracting the diffuse from the total signal, followed by a correction for imperfect cosine response. The UV-MFRSR instruments provide synchronized measurements of spectral irradiances within the UV wavelength region. The direct normal measurement provides an input required for retrieval of atmospheric optical properties, including air mass components, such as the aerosol optical depths and columnar ozone content that may not be feasible from the total irradiance signal alone. Essential adjustments to the UV-MFRSR measurements include removal of dark current bias voltages and a correction for deviations of the sensors from an ideal cosine response. The corrected voltages are then converted into irradiances by multiplying by a calibration factor.

8.4.1.1 Dark Current Bias Removal

Certain extraneous voltages may be generated by the electronics of the observation system, including data loggers, amplifiers, circuits, and connecting wires. These signals are measured at night in the absence of solar radiation using voltages averaged from one hour prior to, and one hour after, the time of minimum solar elevation over the three day period preceding the current data processing day. For a more detailed discussion of the dark current voltage bias removal, please view the USDA UVMRP website (http://uvb.nrel.colostate.edu/UVB/index.jsf), under the "Monitoring Network," then the "Data Processing Procedures" links.

8.4.1.2 Cosine Correction

The direct beam irradiance measured on a horizontal surface is given by the product of the incident normal irradiance and the cosine of the solar zenith angle (SZA). However, the detectors of the UV-MFRSR do not have perfect cosine responses, especially at higher incidence angles. The correction for the deviation of the response from an ideal cosine response is known as cosine correction or angular correction. The non Lambertian response of the UV-MFRSR instruments has been described by Harrison et al. (1994). Characteristics of the cosine responses of the UV-MFRSR instruments used in the network are described by Bigelow et al. (1998). The actual angular responses of an instrument are a function of solar zenith and azimuth angles and were determined annually by YES (in the initial periods of network operation and by CUCF since 1997). The cosine response is determined through laboratory measurements based on an independent radiometric characterization of individual detectors made through the diffuser along two orthogonal planes (Michalsky et al., 1995). With knowledge of angular responses along north-south and east-west azimuths, correction factors for direct normal voltages are interpolated according to solar zenith and azimuth angles at the time of measurement.

An imperfect cosine response affects the measurement of the diffuse horizontal irradiance in addition to the direct normal component (Leszczynski et al., 1998). Experimental results show that the uncertainties in diffuse irradiance due to the assumption of a homogeneous sky radiance distribution are within +1.5% in the UV-B band for varying atmospheric and geographic conditions (Grobner et al., 1996). This assumption is employed in processing the diffuse irradiance values of the UV-MFRSR.

8.4.1.3 Out-of-Band Correction

In the original design of the UV-MFRSR the two channels at 300 nm and 305 nm used silicon-carbide (SiC) photodiodes and the remaining five channels used gallium-phosphide (GaP) photodiodes. By the end of 1997, this instrument design had been deployed at the 22 sites, which comprised the climatological network at that time. In December of 1997, the first of many failures of GaP photodiodes occurred ultimately resulting in a solution developed jointly by UVMRP and YES which involved the replacment of the GaP photodiodes with silicon (Si) versions. The GaP photodiodes were replaced with silicon in all UVMRP UV-MFRSRs by the end of 2002. Thus, the UV-MFRSRs now use Si photodiodes in the 311 nm through 368 nm channels and retain the original SiC photodiodes for the two channels centered at 300 nm and 305 nm (Janson et al., 2004).

Ideally, a radiometer should be designed to reject all radiation outside the designated wavelength region, i.e., in the 2 nm wide wavelength band of each of the seven channels. However, laboratory tests showed that the replacement Si photodiodes allowed out-of-band light from the NIST traceable tungsten-halogen lamps during the calibration procedure to contribute to the calibration signal, resulting in an inaccurate calibration (Lantz et al., 2005). Further studies revealed that the contribution of out-of-band light to the signal depended on several factors, including the channel, the individual instrument, and the lamp used for the laboratory calibration. Further analysis revealed that the out-of-band light contributing to the signal is mainly from wavelengths longer than 570 nm. This contribution of out-of-band light is commonly referred as "red light leakage" or "red leakage" for simplicity. To characterize the contribution of out-of-band light to the signal, each of the UV-MFRSR instruments were measured for out-of-band rejection using short-pass cutoff filters in the field with the sun as the radiation source and in the laboratory using the lamp as the radiation source. Results are given in Tables 8.2 and 8.3. Forty-seven UV-MFRSR instruments were tested in the field for out-of-band light and showed negligible signal within the detection limit of the measurements when the sun is used as the radiation source (Table 8.2). The largest average percent contribution is in the 317 nm channel. The average contribution of out-band-light is 0.4% of the total solar signal, with a maximum of 1.3% for this channel. The implications are that the Langley plot calibrations, which are described below, are not affected by red light leakage to any significant extent. All UV-MFRSR instruments were tested in the laboratory for out-of-band light using a FEL-type quartz-tungsten-halogen lamp as the radiation source and a 400 nm long pass filter. As shown in Table 8.3, there is a significant percentage of out-of-band light contributing to the signal when a lamp is used as the calibration source. The most significant contribution from the out-of-band light is in the 317 nm channel where the average is 22.1%. All of the channels with Si photodiodes have a detectable out-of-band contribution to the total lamp signal. The two channels using SiC photodiodes have no measurable out-of-band signal.

Table 8.2 Percent contributions of out-of-band light to the measured solar signal from 47 filter radiometers (from Lantz et al., 2005)

300 nm

305 nm

311 nm

317 nm

325 nm

332 nm

368 nm

Average

0.1%

0.0%

0.2%

0.4%

0.1%

0.1%

0.1%

Maximum

0.0%

0.3%

0.5%

1.3%

0.3%

0.4%

0.5%

Minimum

0.0%

0.0%

0.0%

0.0%

0.0%

0.0%

0.0%

Table 8.3 Percent contributions of out-of-band light to the measured lamp signal during calibration with a FEL-type quartz-tungsten-halogen lamp from 40 filter radiometers (from Lantz et al., 2005)

300 nm

305 nm

311 nm

317 nm

325 nm

332 nm

368 nm

Average

0.0%

0.0%

8.8%

22.1%

12.1%

11.2%

4.0%

Maximum

0.1%

0.1%

28.4%

35.7%

28.5%

23.1%

10.5%

Minimum

0.0%

0.0%

4.2%

6.6%

3.9%

4.9%

1.1%

The calibration factors corrected for out-of-band signals for the Si photodiodes are applied to the corrected voltages (after the dark current voltage bias and cosine correction) to convert them into irradiances, which are the lamp calibrated channel data provided to the public through an internet website. Earlier versions of the UV-MFRSR instruments that used the GaP photodiodes participated in several North American Interagency Inter-comparisons of Ultraviolet Spectroradiometers conducted during 1995 - 1997. The measurements of the UV-MFRSR instruments agreed with the filter-weighted irradiances of the participating spectroradiometers with a +5% uncertainty at local noon (Early et al., 1998a, 1998b; Lantz et al., 2002). Collocated measurements show that after the out-of-band corrections are applied, the irradiances measured by a representative UV-MFRSR instrument using Si photodiodes agree with the filter-weighted irradiance measurements from a U1000 spectroradiometer within the uncertainty of ± 1.5% (Lantz et al, 2005) for the seven channels. All lamp calibrated UV-MFRSR data accessed via the UVMRP website has now been corrected for the out-of-band radiation leakage.

Was this article helpful?

0 0

Post a comment