Decision Theory Based Classification of Atmospheric Hazards

A method of using decision theory to classify atmospheric states by the lowest associated risk, rather than as the most probable state, is presented. This provides a mathematical framework to make the best decision under uncertainty. Such an approach is desirable for detecting atmospheric hazards using satellite remote sensing where there is a loss associated with an incorrect classification of the atmospheric state. The only thresholds that must be set are those set by the end user, whether that be a scientist or decision maker. For detecting atmospheric hazards in an operational setting, this removes the responsibility of making decisions affecting identification of hazardous areas from the scientist and transfers it to the decision maker. The use of decision theory is demonstrated using an example of flagging volcanic ash hazards by taking the action of classifying the atmosphere as safe or unsafe using the reverse absorption technique and observations from the SEVIRI sensor. The method is able to successfully differentiate between volcanic ash and desert dust - both of which exhibit a similar spectral signature. The shortcomings of the method highlight that decision theory can only aid decision making under uncertainty within the capabilities of the detection algorithm used.