Which dual-band infrared indices are optimum for identifying aerosol compositional change using Himawari-8 data?. (15th November 2020)

• Record Type: Journal Article
• Authors: Sowden, M.; Blake, D.
• Is Part Of: Atmospheric environment. Volume 241(2020)
• Publication Date: 2020-11-15
• Subjects: Dust indices -- Aerosol -- Infrared absorption -- Himawari-8; Air -- Pollution -- Periodicals; Air -- Pollution -- Meteorological aspects -- Periodicals; 551.51
• DOI: 10.1016/j.atmosenv.2020.117620
• Languages: English
• ISSNs: 1352-2310

Abstract

Abstract: Aerosol optical depth algorithms predominantly use the visible portion of the electromagnetic spectrum. However, quantifying sporadic dust events throughout the full 24-h period requires using continuous wavelengths such as infrared (IR). Identifying aerosols, using IR from geostationary data, has relied on subtraction indices rather than normalised differences. Limited attention has been given to determining which IR indices could be suitable for identifying aerosol compositional change. Suitable IR indices could potentially result in multi-spectral data from geostationary satellites, such as Himawari, being used to separate dust from other types of aerosols. This study evaluated three index types: subtraction (brightness temperature difference (BTD)), normalised differences, and division (i.e. quotient). The effectiveness of these three indices were assessed against three sample matrix types: (i) pure soil spectra from the USGS Spectral Library Version 7 database; (ii) potential aerosol plumes, estimated from data relating to the formation and dissipation of a dust storm; and (iii) annual variance, at a surface monitoring site. Absorbance values from the USGS spectral database were aggregated into the spectrally broad Himawari-8 infrared bands. Potential plumes (i.e. transient over a small area) were identified from daily variances in the Himawari-8 satellite data. Principal component analysis was used to determine the variance explained by the first principal component for each of the sample and index types and to evaluate the effectiveness of the indices for detecting dust events. Simple subtraction indices explained more of the variance than normalised differences or division for all sample types. Of the 45 BTD indices analysed, only seven resolved the cloud and aerosol plumes into separate groups. Of these seven indices, BTD3.9–6.2 μm and BTD11–12 μm had the least correlation with the other indices and were chosen as the best indices to identify aerosol compositional change. A further three indices BTD9.6–13 μm, BTD8.6–10 μm, and BTD6.9–7.3 μm were selected based on low correlations between other indices and ensuring that all ten IR wavelengths were utilised. This study indicates that the combination of these five indices, rather than a single index, may optimise the identification of aerosol compositional change. Highlights: BTD marginally better than NDDI and simple quotient. BTD3.9–6.2 μm is a moisture index. BTD11–12 μm is a particle size index. BTD6.9–7.3 μm is an atmospheric stability index. Using five indices optimises aerosol compositional change detection.