Probabilistic NWP-Dependent Volcanic Ash Detection in Satellite Data

Timely and accurate detection of volcanic ash in satellite data is necessary for the issuing of reliable warnings to the aviation industry, for whom ash constitutes a very real danger and a great expense, and to other sectors such as the farming and health sectors (increased particulate matter in the lower atmosphere can increase respiratory problems for both people and livestock). The properties of volcanic ash are highly variable between eruptions, and sometimes also between different stages of the same eruption. The material of which the ash is constituted, the shape of the particles, the size of the particles and the concentration of particles all affect how the ash is seen from space, all vary greatly, and are often not known in advance, particularly for poorly monitored volcanoes. In order to detect ash as reliably as possible, it is therefore necessary to constrain ‘non-ash' observations as tightly as possible, and this can be done through exploitation of time- and space-specific Numerical Weather Prediction (NWP) data. A method for combining NWP-dependent simulated clear sky observations with simulated non-ash cloud and volcanic ash cloud observations (also NWP-dependent) to calculate a posterior probability for each of the three states is presented. The probabilistic result has the advantage that it inherently contains information on the certainty of the classification at each point within it, which can aid forecasters in their interpretation of the evidence that forms the basis of the warnings they are responsible for issuing. Warnings to aviation are based on retrieved ash properties, which are computationally expensive to derive, and so any useful detection technique must be relatively fast and should be applicable to problems of ash detection from any volcano, at any time. This method achieves this, and has the added advantage of being (in principle) generic enough to adapt to imagery from a range of sensors, potentially giving forecasters a generic tool to apply to much of the imagery they are asked to consider when issuing volcanic ash warnings.