Volcanic ash detection and retrievals from MODIS data by means of Neural Networks
Abstract. Volcanic ash clouds detection and retrieval represent a key issue for the aviation safety due to the harming effects they can provoke on aircrafts. A lesson learned from the recent Icelandic Eyjafjalla volcano eruption is the need to obtain accurate and reliable retrievals on a real time basis. The current most widely adopted procedures for ash detection and retrieval are based on the Brightness Temperature Difference (BTD) inversion observed at 11 and 12 μm that allows volcanic and meteo clouds discrimination. While ash cloud detection can be readily obtained, a reliable quantitative ash cloud retrieval can be so time consuming to prevent its utilization during the crisis phase. In this work a fast and accurate Neural Network (NN) approach to detect and retrieve volcanic ash cloud properties has been developed using multispectral IR measurements collected by the Moderate Resolution Imaging Spectroradiometer (MODIS) over Mt. Etna volcano during 2001, 2002 and 2006 eruptive events. The procedure consists in two separate steps: the ash detection and ash mass retrieval. The detection is reduced to a classification problem by identifying two classes of "ashy" and "non-ashy" pixels in the MODIS images. Then the ash mass is estimated by means of the NN, replicating the BTD-based model performances. The results obtained from the entire procedure are very encouraging; indeed the confusion matrix for the test set has an accuracy greater than 90 %. Both ash detection and retrieval show a good agreement when compared to the results achieved by the BTD-based procedure. Moreover, the NN procedure is so fast to be extremely attractive in all the cases when the quick response time of the system is a mandatory requirement.