Evaluation of methods to determine the spectral variations of aerosol optical thickness
MetadataShow full item record
The methods used to derive spectral variations of aerosol optical thickness, AOT are evaluated. For our analysis we have used the AOT measured using a hand held sunphotometer at the coastal station on the west coast of India, Dona-Paula, Goa for the year 2000 - 2006. These data show seasonal and diurnal variations of AOT and Angstrom exponent. The mean values of Angstrom exponents are 0.97 (SD= 0.32) for summer, 0.87(SD=0.32), during post-monsoon and 1.39 (SD=0.26) during winter. The seasonal values of extinction coefficients are 0.22 (SD=0.07) during summer, 0.22 (SD=0.09) for post-monsoon and relatively low value of 0.14 (SD=0.06) during winter. The mean value of Angstrom exponent, a for the period of observations 2000-2006 is 1.22 (SD = 0.38) and extinction coefficient b is 0.18 (SD=0.1). For our study we have used three methods that determine the spectral variations of AOT. These methods are Volz method using two wavelengths, linear fitting and polynomial fitting using spectral AOT. It is observed that the values of a obtained from the pair of wavelengths when compared with those obtained using linear fit are comparable for the wavelength pairs 675-870 and 440-870 (R sup(2) approx. 0.9). This indicates that alpha (Lambda 870) is nearly independent of wavelength and thus 870 nm is a better reference band. The values of alpha (440,500) and alpha (500,675) are found to be underestimated. Though it has been reported that the polynomial model provide values closer to the measured AOT, our observations indicate that spectral AOT values derived using the Angstrom exponent provide less error. We have evaluated the methods comparing the spectral values of AOT derived using the methods with the measured. RMSE values using linear fit (mean RMSE = 0.031) are lower than those of polynomial fit (mean RMSE = 0.042). RMSE values using polynomial are found to increase with the values of AOT, and such a trend is not seen using the AOT values derived using the Angstrom exponent values from linear fit. For lower wavelengths 440 and 500 nm, the negative of the second derivative alpha, is found to increase with tau sub(a) (Lambda). Thus it is observed that using Angstrom exponent derived using a simple linear fit could be used to determine the spectral variations of AOT.
- Scholarly literature