Spectrally Enhanced Cloud Objects—A generalized framework for automated detection of volcanic ash and dust clouds using passive satellite measurements: 1. Multispectral analysis

Abstract

While satellites are a proven resource for detecting and tracking volcanic ash and dust clouds, existing algorithms for automatically detecting volcanic ash and dust either exhibit poor overall skill or can only be applied to a limited number of sensors and/or geographic regions. As such, existing techniques are not optimized for use in real-time applications like volcanic eruption alerting and data assimilation. In an effort to significantly improve upon existing capabilities, the Spectrally Enhanced Cloud Objects (SECO) algorithm was developed. The SECO algorithm utilizes a combination of radiative transfer theory, a statistical model, and image processing techniques to identify volcanic ash and dust clouds in satellite imagery with a very low false alarm rate. This fully automated technique is globally applicable (day and night) and can be adapted to a wide range of low earth orbit and geostationary satellite sensors or even combinations of satellite sensors. The SECO algorithm consists of four primary components: conversion of satellite measurements into robust spectral metrics, application of a Bayesian method to estimate the probability that a given satellite pixel contains volcanic ash and/or dust, construction of cloud objects, and the selection of cloud objects deemed to have the physical attributes consistent with volcanic ash and/or dust clouds. The first two components of the SECO algorithm are described in this paper, while the final two components are described in a companion paper.

Spectrally Enhanced Cloud Objects—A generalized framework for automated detection of volcanic ash and dust clouds using passive satellite measurements: 2. Cloud object analysis and global application

“…The volcanic ash/dust probability determined using the naïve Bayesian approach described in Pavolonis et al . [], in combination with results from a cloud property retrieval algorithm [ Pavolonis et al ., ], is used to identify satellite pixels that might contain volcanic ash and/or dust. All pixels that potentially contain ash or dust are sorted into cloud objects.…”

“…In Part 1 of this paper [ Pavolonis et al ., ], several sophisticated multispectral satellite metrics, derived from measurements with central wavelengths of approximately 0.65, 3.9, 7.3, 8.5, 11, and 12 µm, were utilized in a naïve Bayesian procedure to determine the probability that a given satellite pixel contains volcanic ash and/or nonvolcanic dust. The Bayesian method can utilize all of those spectral channels or several different channel subsets as dictated by sensor capabilities, solar zenith angle, and/or intellectual curiosity [see Pavolonis et al ., , Table 1]. The Bayesian method was trained empirically using a very large Moderate Resolution Imaging Spectroradiometer (MODIS)‐based data set where the horizontal bounds of volcanic ash and dust clouds were manually analyzed.…”

Monitoring the Agung (Indonesia) Ash Plume of November 2017 by Means of Infrared Himawari 8 Data

“…The split window technique identifies airborne ash by means of a fixed threshold test applied to difference of Brightness Temperatures (BT) measured at the aforementioned wavelengths, i.e., BT 11 − BT 12 [29]. Advanced detection methods, minimizing the impact of atmospheric water vapor on the above-mentioned brightness temperature difference (BTD) or analyzing signals measured in other spectral bands such as MIR (Medium Infrared) and/or VIS (Visible) ones, have shown a higher efficiency in identifying ash clouds (e.g., [30][31][32][33][34][35][36][37][38]). RST ASH [16] is an ash detection method, based on the Robust Satellite Technique (RST) multi-temporal approach [39], running on both polar [14,15,40] and geostationary [17] satellite data.…”

Improving volcanic ash predictions with the HYSPLIT dispersion model by assimilating MODIS satellite retrievals

“…They include ash mass loading, cloud top height, and effective particle radius. Pavolonis et al (2013Pavolonis et al ( , 2015a described the details of the retrieval methodology and how the ash cloud observations are derived from the retrieved parameters, such as radiative temperature and emissivity. Here, volcanic ash observations from the 2008 Kasatochi eruption at five different instances are utilized.…”

“…An automated volcanic ash cloud detection system has been developed and continuously improved (Pavolonis et al, 2006(Pavolonis et al, , 2013(Pavolonis et al, , 2015a. In addition to detecting and monitoring ash clouds, satellite measurements allow many ash cloud characteristics to be quantified.…”