Synthesized solar spectral irradiance in the 3.7 µm band for Earth remote sensing applications

Since the late 1970's launch of the first Advanced Very High Resolution Radiometer (AVHRR) instrument aboard the TIROS-N polar orbiter, measurements in the 3.7 µm atmospheric window have been exploited for use in cloud detection and screening, cloud phase and surface snow/ice discrimination, quantitative cloud particle size retrievals, and fire detection. The utility of the band has led to the incorporation of a 3.7 µm channel on a number of low Earth orbit and Geosynchronous imagers, as well as on future operational imagers for weather and climate applications. Daytime 3.7 µm band observations include both reflected solar and thermal emission energy. It is the observed cloud bidirectional reflectance, or closely associated cloud emissivity, and not the radiance itself, that is the fundamental quantity relevant for cloud microphysical retrievals. Since 3.7 µm channels are calibrated to a radiance scale (via onboard blackbodies), knowledge of the top-of-atmosphere solar irradiance in the spectral region is required to infer reflectance. In this presentation, we examine the fundamental uncertainty in the observed 3.7 µm solar reflectance that comes from uncertainty in the knowledge of the absolute solar spectral irradiance.

Despite the ubiquity of a 3.7 µm channel, absolute solar spectral irradiance data comes from either a single observation made in 1969 (Thekaekara, 1974), or several popular synthetic spectra (e.g., Kurucz, 1995). In this presentation, we compare the historical 3.7 µm spectral irradiance data sets with the recent semi-empirical solar model of Fontenla el al. (2006), which is constructed to match available observations in the visible and shortwave infrared spectral regions. The model accounts for active sun features, and is expected to have uncertainties of less than 2% in the 3.7 µm spectral region. We find that typical instrument channel-averaged spectral irradiances using the observations reported by Thekaekara are 3-4% less than those derived from the Fontenla model; the Kurucz spectrum, as included in the MODTRAN4 distribution, gives irradiances about 10% smaller than the Fontenla model. Because the slope and curvature of the irradiance spectrum is significant in this spectral region, accurate characterization of the instrument spectral bandpass is important.