Toward quantitative forecasts of volcanic ash dispersal: Using satellite retrievals for optimal estimation of source terms

Abstract

Airborne volcanic ash is a hazard to aviation. There is an increasing demand for quantitative forecasts of ash properties such as ash mass load to allow airline operators to better manage the risks of flying through airspace likely to be contaminated by ash. In this paper we show how satellite‐derived mass load information at times prior to the issuance of the latest forecast can be used to estimate various model parameters that are not easily obtained by other means such as the distribution of mass of the ash column at the volcano. This in turn leads to better forecasts of ash mass load. We demonstrate the efficacy of this approach using several case studies.

Estimation of optimal dispersion model source parameters using satellite detections of volcanic ash

“…The ability to make use of ash polygons based on various sources of information about ash presence (including inputs from skilled forecasters) and demonstrated reliability make this method ideal for operational use. The sources considered in this paper have not been optimized to make quantitative forecasts of mass loadings, which are in increasing demand; this issue is addressed in another paper [ Zidikheri et al , ]. However, while volcanic ash advisories are still issued as polygons defining ash boundaries only, the method presented in this paper is likely to be of relevance to volcanic ash forecasters.…”

Quantitative Verification and Calibration of Volcanic Ash Ensemble Forecasts Using Satellite Data

“…In this paper we extend the inverse modeling methodology of Zidikheri and Potts () and Zidikheri et al (, , ) for this purpose. Using this method, the authors demonstrated that various source parameters such as the height, diameter, start time, vertical, temporal, and particle size mass distributions can be inferred and lead to better forecasts.…”

“…This is conventionally a uniform line source with the same emission rate specified at all altitudes from the summit up to a maximum height as determined by various ash cloud height estimates (Witham et al, ). For large eruptions with significant umbrella cloud growth, this approach is inappropriate and must be replaced with both a broader source mass distribution in the horizontal and a source mass distribution in the vertical biased toward higher altitudes (Zidikheri et al, , ). However, for the purpose of analyzing ash detections only, once the maximum height has been specified, the vertical mass distribution below the maximum height is irrelevant and a vertically uniform and horizontally broad cylindrical distribution is adequate to represent the umbrella cloud effects.…”

“…Several studies (Boichu et al, , ; Chai et al, ; Eckhardt et al, ; Kristiansen et al, , ; Moxnes et al, ; Seibert et al, ; Stohl et al, ) have addressed this problem by making use of variational techniques to estimate the emission profile subject to satellite‐based quantitative retrievals of ash properties. Zidikheri et al () also addressed this problem by extending the methodology of Zidikheri and Potts () and Zidikheri et al (, ) to include the estimation of vertical mass emission profile, time variation, and particle size distribution subject to satellite retrievals using the SECO algorithm and ACCESS‐G global meteorological model wind fields. In this paper we further extend the methodology of Zidikheri and Potts () and Zidikheri et al (, , ) by calculating the aforementioned pattern correlation threshold parameter self‐consistently.…”

“…Zidikheri et al () also addressed this problem by extending the methodology of Zidikheri and Potts () and Zidikheri et al (, ) to include the estimation of vertical mass emission profile, time variation, and particle size distribution subject to satellite retrievals using the SECO algorithm and ACCESS‐G global meteorological model wind fields. In this paper we further extend the methodology of Zidikheri and Potts () and Zidikheri et al (, , ) by calculating the aforementioned pattern correlation threshold parameter self‐consistently. This is accomplished by maximizing the Brier skill score (which is related to the Brier score) with respect to the threshold parameter.…”