The impact of satellite sensor viewing geometry on time-series analysis of volcanic emissions

Abstract

Time-series analysis techniques are being increasingly used to process satellite observations of volcanic gas emissions and heat flux, with the aim of identifying cyclic behaviour that could inform hazard assessment or elucidate volcanic processes. However, it can be difficult to distinguish cyclic variations due to geophysical processes from those that are artefacts of the observation technique. Here, we conduct a comprehensive investigation into the origin of cyclicity in volcanic observations by analysing daily, global satellite measurements of volcanic SO2 loading by the Ozone Monitoring Instrument (OMI) and thermal infrared anomalies detected by the Moderate Resolution Imaging Spectroradiometer (MODIS). We use a fast Fourier Transform (FFT) multi-taper method (MTM) to analyse multiple phases of activity at 32 target volcanoes, utilising measurements obtained from three NASA satellite instruments (Aura – OMI, Aqua – MODIS and Terra – MODIS), and identify a common cycle (period of ~ 2.3 days), which is not considered to be of volcanic origin. Based on the presence of this cycle in multiple satellite datasets, we attribute it to variations in viewing angle during the 16-day orbit repeat cycle of sun-synchronous satellites maintained in Low Earth Orbit (LEO). A 5-point averaging correction procedure is tested on satellite observations from Kilauea volcano, Hawaii, and is found to reduce the impact of higher frequency cycles and reveal the presence of longer-period geophysical signals. In addition to the identification of a signal common to different measurement techniques, an underlying cyclical pattern was found in the OMI SO2 observations (periods of ~ 7.9 and 3.2 days) generated by the OMI Row Anomaly (ORA). We conclude that identification of the presence and magnitude of non-geophysical cyclic behaviour, which can suppress natural cycles in time-series data, and implementation of appropriate corrections, is crucial for accurate interpretation of satellite observations. The use of data averaging to suppress non-geophysical cycles imposes limits on the length of natural cycles that can be confidently identified in moderate resolution satellite observations from polar-orbiting spacecraft.