JPSS Volcanic Hazards Initiative: VOLCAT SO2 Alerts

As described in a previous blog post, the Joint Polar Satellite System (JPSS) Volcanic Hazards Initiative seeks to develop volcanic cloud products from the JPSS series satellites (Suomi-NPP, NOAA-20) addressing current and future needs of the International Civil Aviation Organization (ICAO) International Airways Volcano Watch (IAVW). A major component of this work has been the development of a JPSS multisensor product for detecting and characterizing volcanic sulfur dioxide (SO2), an emerging need of aviation stakeholders. This multisensor product combines the high spectral sensitivity of the Cross-Track Infrared Sounder (CrIS, a hyperspectral infrared spectrometer) with the high spatial resolution of the VIIRS imager within the framework of the NOAA/UW-CIMSS VOLcanic Cloud Anaylsis Toolkit (VOLCAT). As these instruments are aboard S-NPP and NOAA-20 and can operate day and night, this product provides 4 times daily global coverage of SO2 clouds.

The product is driven by an advanced algorithm designed for the CrIS instrument which can estimate the total column concentration of SO2 and the altitude of the SO2 layer. The CrIS algorithm operates in two passes over new measurements. The first stage determines which CrIS measurements have spectral features indicative of a statistically significant amount of SO2 compared with the background atmosphere. The second stage generates estimates of SO2 column loading and height for the significant detections in the first stage. Since VIIRS is also weakly sensitive to SO2, a fused product is possible which interpolates some of the CrIS data to the VIIRS imager swath.

This post will focus primarily on the generation of VOLCAT alerts for SO2 and a subsequent companion post will focus on the estimated SO2 cloud properties. 

Because the alerts are ultimately driven by CrIS measurements, they are to a large extent controlled by the footprint of CrIS measurements.  CrIS  scans  between ±48.3 degrees. Each  scan  contains  30  fields  of regard (FOR), each of which contains 9 circular fields of view (FOV) arranged as a 3×3 array.  As the scanning mirror moves, the FOR rotates slightly and consequently, FOVs transition from 14 km-diameter circles at nadir, to 43.6 km x 23.2 km (major and minor axes) ellipses on the edge of scan leading to a swath width of approximately 2200 km with approximately 30% gaps and an average sampling distance of 16 km. The ground pattern of these FOVs is shown in Fig. 1.  Because CrIS FOVs are so large, fusion with VIIRS (750 m resolution) is so valuable. 

Figure 1: VIIRS long-wave IR (band I5) scene of the Red Sea with 4 full scans of CrIS FOV footprints plotted showing the pattern made by CrIS measurements. From Wang et al. (2013). 

As described in the previous post, the basic information of VOLCAT SO2 alerts are (Fig. 2):

  1. Information on the time and date of the alert (note the typical 2-3 hr latency) and the primary satellite and instrument responsible. 
  2. A contextual VIIRS false color image of the scene with a blue polygon representing the location of the likely SO2 cloud.
  3. Typical VOLCAT alert information including volcanic region, radiative center location, and likely source volcanoes. 
  4. The maximum SO2 layer height detected and the 90th percentile of all heights detected. The difference between the 90th percentile height and the maximum height is an important measure of uncertainty in these alerts when many CrIS FOVs contain SO2 – the larger this difference, the more overall uncertainty for the cloud. Additionally, the mean tropopause height is included for context, specifically if the 90th percentile height is higher than the tropopause height, the same cloud is likely to remain a hazard for longer, perhaps several overpasses.  

Under the “Show More” drop down, additional alert information is provided, most important of which is the estimate of total cloud mass and of total pixel area (Fig. 2).  These provide a sense of magnitude for the alert. Typically SO2 clouds of more than several kilotons of SO2 (1 kt = 0.001 Tg) are noteworthy.  As the ICAO/IAVW guidance on SO2 is not currently fully in place, all SO2 clouds are assigned a “Volcanic Event Priority Ranking”  of 3, the lowest priority group generated by VOLCAT. 

Figure 2: Example VOLCAT SO2 Alert derived from S-NPP CrIS/VIIRS data on 31 July, 2020 for the ongoing eruption of Nishinoshima, Japan. The scan start position in the image has affected the distance to nearby volcanoes, especially the true source volcano in this case. (https://volcano.ssec.wisc.edu/alert/report/127034)

There are many interesting features of these new alerts, though only a few are addressed here. Because CrIS spectra and VIIRS multispectral imagery are sensitive to different absorption bands of SO2, it is routine that alerts are generated without any SO2 apparent in VIIRS imagery.  In the case of the example cloud in Fig. 2, colocated volcanic ash from the Nishinoshima has reduced the VIIRS SO2 sensitivity. The SO2 absorption feature measured by CrIS is minimally sensitive to ash, so a robust alert is still possible under these conditions. Additionally, because CrIS is so sensitive to SO2, it may be used successfully as a proxy for some ash clouds, especially in the vicinity of coeval ash alerts where there is a faint multispectral ash signal but no retrieved ash properties (Fig. 3). In this case from Nishinoshima, there is a clear, but faint multispectral signal of ash outside the brown “Ash/Dust Cloud” polygon assigned by VOLCAT, but within the CrIS/VIIRS SO2 detection. 

Figure 3: Example VOLCAT SO2 (left) and Ash (right) Alert on 31 July, 2020 for the ongoing eruption of Nishinoshima, Japan. Note the faint pink ash/dust signal extending west-southwest from the main ash cloud. Colocated SO2 in this region is evidence that dilute ash may be present. SO2: https://volcano.ssec.wisc.edu/alert/individual/148692 Ash: https://volcano.ssec.wisc.edu/alert/individual/148688  

The CrIS algorithm can successfully detect SO2 in a variety of background atmospheres including when clouds are present; however, it is sensitive to large quantities of water vapor and thus can trigger false alarms in the presence of very deep convective clouds. This is a particular problem in the tropics where large, very moist cyclones occasionally trigger false alarms.  In order to quickly identify possible false alerts, other images and the SO2 properties must be consulted. In particular, the spread of SO2 heights can give a good indication of a true or false alert.  In a true alert, multiple adjacent CrIS FOVs with SO2 will generally have been detected and their heights will not usually vary by more than a few kilometers.  In a false alert, several CrIS FOVs may have been determined to have SO2, but they may not be adjacent or closely clustered and their heights will typically vary by a lot, perhaps as much as 10 km (e.g., Fig. 4). S-NPP CrIS is more prone to these false alerts due to one of its detectors suffering from high radiometric noise; however, a fix for this is currently being tested and will soon become live, greatly reducing the false SO2 alerts from S-NPP CrIS. Despite this fix, VOLCAT still detects many possible SO2 clouds which may either be non-volcanic (and thus unrelated to volcanic ash) or minor false alerts. In all of these cases, these clouds are either low concentration relative to typical volcanic sources or are very small, containing only one or a few FOVs. The next post on SO2 cloud properties will discuss identifying these in greater depth.    

Figure 4: Example VOLCAT SO2 height for a true (left) and likely false (right) detection. Left: SO2 heights from Nishinoshima on 31 July, 2020. Right: SO2 heights for a false detection due to a tropical cyclone. Note the good clustering in horizontal and vertical directions for the true detection and the poor spatial correlation for the false detection (https://volcano.ssec.wisc.edu/alert/individual/164901).

References:

Wang,  L.,  Tremblay,  D. A.,  Han,  Y.,  Esplin,  M.,  Hagan,  D. E.,  Predina,  J.,  Suwinski, L., Jin, X., and Chen, Y. (2013).  Geolocation assessment for CrIS sensor data records. Journal of Geophysical Research:  Atmospheres, 118(22): 12,690–12,704. DOI: 10.1002/2013JD020376