Premium
Spectral derivative analysis of solar spectroradiometric measurements: Theoretical basis
Author(s) -
Hansell R. A.,
Tsay S.C.,
Pantina P.,
Lewis J. R.,
Ji Q.,
Herman J. R.
Publication year - 2014
Publication title -
journal of geophysical research: atmospheres
Language(s) - English
Resource type - Journals
eISSN - 2169-8996
pISSN - 2169-897X
DOI - 10.1002/2013jd021423
Subject(s) - cirrus , lidar , hyperspectral imaging , remote sensing , environmental science , optical depth , aerosol , atmospheric sciences , spectral line , spectroscopy , spectral bands , refractive index , derivative (finance) , materials science , optics , physics , meteorology , geology , quantum mechanics , astronomy , financial economics , economics
Spectral derivative analysis, a commonly used tool in analytical spectroscopy, is described for studying cirrus clouds and aerosols using hyperspectral, remote sensing data. The methodology employs spectral measurements from the 2006 Biomass‐burning Aerosols in Southeast Asia field study to demonstrate the approach. Spectral peaks associated with the first two derivatives of measured/modeled transmitted spectral fluxes are examined in terms of their shapes, magnitudes, and positions from 350 to 750 nm, where variability is largest. Differences in spectral features between media are mainly associated with particle size and imaginary term of the complex refractive index. Differences in derivative spectra permit cirrus to be conservatively detected at optical depths near the optical thin limit of ~0.03 and yield valuable insight into the composition and hygroscopic nature of aerosols. Biomass‐burning smoke aerosols/cirrus generally exhibit positive/negative slopes, respectively, across the 500–700 nm spectral band. The effect of cirrus in combined media is to increase/decrease the slope as cloud optical thickness decreases/increases. For thick cirrus, the slope tends to 0. An algorithm is also presented which employs a two model fit of derivative spectra for determining relative contributions of aerosols/clouds to measured data, thus enabling the optical thickness of the media to be partitioned. For the cases examined, aerosols/clouds explain ~83%/17% of the spectral signatures, respectively, yielding a mean cirrus cloud optical thickness of 0.08 ± 0.03, which compared reasonably well with those retrieved from a collocated Micropulse Lidar Network Instrument (0.09 ± 0.04). This method permits extracting the maximum informational content from hyperspectral data for atmospheric remote sensing applications.