Automatic detection of volcanic ash from Himawari – 8 satellite using artificial neural network

Volcanic ash is a significant phenomenon towards aviation safety and capacity of influencing climate change. Therefore, accurate information of the volcanic ash spatial distribution in the atmosphere possessed a fundamental role in the community. Polar type satellites such as Terra / Aqua equipped w...

Full description

Saved in:
Bibliographic Details
Main Authors: Putra, Richard Mahendra, Saputro, Adhi Harmoko, Arazak, Laiza, Kharisma, Sulton
Format: Conference Proceeding
Language:English
Subjects:
Air safety
Artificial neural networks
Atmospheric models
MODIS
Neural networks
Satellite imagery
Satellite observation
Satellites
Sensors
Spatial distribution
Synchronous satellites
Training
Volcanic ash
Volcanic eruptions
Volcanoes
Citations: Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Volcanic ash is a significant phenomenon towards aviation safety and capacity of influencing climate change. Therefore, accurate information of the volcanic ash spatial distribution in the atmosphere possessed a fundamental role in the community. Polar type satellites such as Terra / Aqua equipped with MODIS sensors are capable of providing vivid imagery of the volcanic ash spatial distribution. However, the deficiency of this satellite is unable to perform optimal imagery for real-time monitoring due to its limitation require to be located above the volcanic ash site. Therefore, a geostationary satellite is a feasible solution to solve this issue hence its capability to observe specified fixed location continuously. Despite its capability, this type of satellite also performs designated weaknesses hence non-absolute perpendicularity observation angles on certain conditions towards the observed objects from the satellite fixed position. The purpose of this study is to create an automatic detection system using Himawari-8 satellite observations data by applying Artificial Neural Network (ANN) with training datasets and utilizing Terra / Aqua polar type satellite with MODIS sensors as validator. The input variation based on previous research references were using three bands, all bands, and four variations of satellite bands. The result of the study justifies the models established using all bands and four variation bands can produce good performances in training data, although less consistent if applied towards other cases. While the single-pixel model with three-band input well suited over Mt. Merapi volcanic eruption event on June 1st, 2018 with 93.71% accuracy
ISSN:0094-243X
1551-7616
DOI:10.1063/1.5141725