By

JPSS Volcanic Hazards Initiative: VOLCAT SO2 Properties

This post is a companion to the previous post on VOLCAT SO2 Alerts. Please see that post for background on the CrIS instrument and definitions of the acronyms used here. In particular, it deals with the SO2 cloud properties derived in VOLCAT by the CrIS SO2 algorithm.

The CrIS algorithm estimates the height and column concentration of SO2 present across a wide range of background conditions including many cloudy scenes. The algorithm works primarily by analyzing the spectral residuals between the measured infrared spectrum and some spectrum considered to represent the SO2-free background atmosphere. As discussed in the previous post, the algorithm does trigger some false alerts in the presence of large quantities of water vapor, such as in the deepest convective clouds in tropical storms. The algorithm is designed to estimate a probability density function for the likely altitude of the SO2 layer and to estimate the partial column concentration for many height intervals, that is the column loading of SO2 in a given height interval. This is accomplished by determining the layer height and column loading in the case of many possible background (SO2-free) atmospheres. Although figures relating to these probabilistic indicators of SO2 hazard are not currently generated in VOLCAT alerts, they will be added in the near future, directly addressing the changing IAVW roadmap for quantitative (with uncertainty) SO2 hazards which will take effect in the coming years.

This post will detail considerations of the SO2 cloud properties currently given in the VOLCAT alerts. Currently, VOLCAT alerts provide maps of SO2 properties similar to those already generated for ash with a few differences. Many of the SO2 cloud property figures are identical to those for ash (two false color images, a longwave infrared image, a visible image (reflectance), and a 12-11 µm split window image). The SO2-specific images currently generated by VOLCAT are:

  1. VIIRS-only SO2 probability based on the 8.6-10.8 µm (online-offline) VIIRS brightness temperature difference. This probability correlates well with whether or not SO2 is visible in the false color (R: 12-10.8µm, G: 10.8-8.6 µm, B: 10.8 µm) image, typically visible as either a bright green cloud or more yellow if ash is present. 
  2. CrIS-only SO2 heights. Currently, VOLCAT does not interpolate SO2 heights to the VIIRS swath since this would generate unrealistic heights on the cloud edges or between two nearby directions. Consequently, the heights are shown only on the CrIS field of view (FOV) footprints. 
  3. A CrIS-VIIRS fusion product for the total column concentration (mass loading). Because this is a continuous field in the satellite viewing geometry (unlike SO2 layer height), it can be interpolated to the VIIRS swath, generating additional detail over a CrIS-only image. As is common for SO2 measured by other sensors, the SO2 loading is measured here in Dobson Units (DU), where for SO2, 1 DU = 0.0285 g m-2. 

Between these three images for SO2 properties, many nuances of the presence and hazards associated with SO2 can be determined. The same example from the previous post (Nishinoshima on 31 July, 2020) is shown in Fig. 1. In this example a clear multispectral ash signal is present (pink colors), with some indication that SO2 is present within the ash clouds (yellowish) in the multispectral image. The VIIRS SO2 probability is near-zero in the region of most concentrated ash and jumps to near-certainty just outside of this region where the ash is no longer opaque.  Otherwise, the VIIRS-only SO2 probability is very small in most of the scene. This necessitates the use of the CrIS-integrated products to assess the SO2 cloud hazards.  

Figure 1: VOLCAT SO2 cloud properties for the 31 July, 2020 eruption cloud of Nishinoshima, Japan. 

As the CrIS SO2 heights suggest, this cloud is mostly between 5-9 km asl, though it is somewhat lower-altitude near the source where it is generally between 2-5 km asl though there is some scatter in the heights there. Becasure the CrIS heights are statistical representations of the true layer height subject to many possible background atmospheres, it is typical that there is some local scatter in the height estimates. This should be interpreted as statistical noise and the individual height retrievals should mostly be considered in aggregate. This does not preclude spatial variations in height, but these should be considered as local averages.  As described in a previous post, the difference between the 90th percentile height and maximum retrieved height reported in the alert information gives a sense of the overall uncertainty, that is, if this is a large difference, then the distribution of heights is very fat tailed and the overall set of heights is very uncertain.  This consideration only applies to clouds large enough to have a meaningful set of percentiles. As described in the previous post on alerts, high scatter among detected CrIS FOVs and little spatial correlation or clustering can be a good indication of false detections. Lastly, as described in the previous post, the ground pattern of CrIS FOVs can have as much as 30% gaps. Consequently, apparent “holes” in the SO2 cloud do not necessarily mean there is no SO2 there (e.g. Fig 1). As these “holes” are relatively small, they should simply be ignored. 

The last SO2 cloud property currently generated by VOLCAT is the fused CrIS-VIIRS total column loading. The CrIS algorithm takes into account the estimated height of SO2 when calculating the loading, so in some cases, random errors that assign a very low altitude will cause an unrealistically large loading. As this is a local effect, it is most likely only in false alarms. In the example in Fig. 1, the fusion of VIIRs with CrIS not only fills almost all of the gaps in the CrIS swath, but adds additional details that would be difficult to distinguish in a CrIS-only approach such as the sharp gradients in loading near the source region. Note that in VOLCAT SO2 loading images, the color scale is logarithmic with the values 0.32, 3.16, and 31.63 representing 10-1.5, 100.5, and 101.5 respectively.  This scaling has been chosen as most appropriate since testing indicates that the loading in these clouds can span many orders of magnitude even within the same cloud.  In general, SO2 loading less than 1 DU is considered low concentration, whereas loading greater than about 10 DU is considered moderate concentration, and loading higher than about 50 DU are considered high concentration. Because the CrIS SO2 retrieval is based on retrieving the height and column loading in a 1 km-thick layer of SO2, SO2 column loading can be viewed crudely in reference to some threshold concentration. In particular, the World Health Organization (WHO) sets 500 µg m-3 as a 10-minute exposure threshold for respiratory hazards. In the common units of SO2 remote sensing, this corresponds to a point concentration of 17.5 DU km-1, that is, a column loading of 17.5 DU for an assumed 1 km-thick layer or 8.75 DU for an 0.5 km-thick layer. On the color scale given in VOLCAT alerts, 17.5 DU is right about at the boundary between green and yellow colors (approx. 101.25 DU).

This product suite generally detects SO2 from four sources: 

  1. Anthropogenic sources such as power plants (low concentration, low altitude, small area) 
  2. Biomass burning sources such as wildfires (very low concentration, any altitude, any area)
  3. Passive (non-explosive) volcanic degassing (low-moderate concentration, altitude similar to source volcano, small-moderate area)
  4. Active (explosive) volcanic degassing (any concentration, altitude, and area). 

Only this last category is of significant interest for aviation support of volcanic hazards avoidance. The first three sources, though interesting in their respective applications, can generally be distinguished from active volcanic degassing by the SO2 cloud concentration and height, but also proximity to a currently restless volcano. Because the CrIS SO2 algorithm operates day and night aboard both Suomi-NPP and NOAA-20, each CrIS instrument has nearly complete global coverage twice per day, approximately 50 minutes apart. Consequently, new SO2 emissions are theoretically detectable within about 11 hours of emission. As a result, significant new SO2 emissions cannot travel more than a few hundred km without being detected. Thus any new SO2 detections more than several hundred km away from a restless volcano are unlikely to have originated from an erupting volcano and can instead be assumed either one of the non-volcanic sources or a false alarm.