Estimating spectral irradiance from measurements in seven spectral bands

Accurate measurement and characterisation of fluctuations in the irradiance environment is important for many areas of optical remote sensing. This paper reports a method of estimating spectral irradiance over the VNIR region (400 1100nm) from the radiance of a calibrated reference panel, measured in seven narrow (10nm) spectral bands. Earlier work established the potential for estimating spectral irradiance from multi-band data using a neural network technique (Milton et al., 2000). The approach described here uses linear regression analysis to regenerate the irradiance spectrum from data in seven reference wavelengths. The method was tested using data from a specially designed multiband radiometer – the INdependent SPectral IRradiance Estimator (INSPIRE). The irradiance spectrum was partitioned into a number of distinct regions within each of which the spectral irradiance was estimated from irradiance measured at one of the reference wavelengths. The precision of the method was found to be better than ±5% over most wavelengths from 400nm to 1100nm. Furthermore, the slope coefficients of the individual regression models were found to be sensitive to the sky radiance conditions, especially over the region 600760nm, and improvement in the precision of the predicted spectrum (to within ±3%) was obtained by taking the diffuse-to-global (D:G) irradiance ratio at the time of measurement into account.