Meanwhile, the Spinning Enhanced Visible and InfraRed Imager (SEVIRI) on EUMETSAT’s Meteosat-8 satellite also collects full-disk imagery. At the time of this event, it was in geostationary orbit at 41.5 degrees East, over the Indian Ocean. Its perspective provides coverage of an area of the Earth that is minimally visible to GOES-16 and GOES-17. Band 6 of this detector has a wavelength band centered at 7.35 microns, with a band pass that overlaps that of the GOES ABI instruments.

Because the antipode of the Hunga Tonga volcano (20.545 S, 175.394 W) is located approximately over the shared borders of Mali, Algeria, and Niger, Meteosat-8’s spatial coverage provides the opportunity to view the interference pattern of the pressure wave at the midpoint of its first traversal of the globe.

The EUMETSAT website provides access to the Meteosat-8 datasets. The option to download historical data requires registration, but you may request access via a free online account. There are multiple mechanisms provided to download data once you are logged into your account. I chose the API endpoint and the product “EO:EUM:DAT:MSG:MSG15-RSS - Rapid Scan High Rate SEVIRI Level 1.5 Image Data – MSG”, then browsed by date. A typical image file download webpage looks like the image to the left.

The image files are in EUMETSAT’s https://api.eumetsat.int/data/browse/collections “”MSG” format, with a “.nat” extension, as shown above. (Support for Meteosat MSG format file support was recently introduced in ENVI® 5.6.2). These files are quite large, so I wrote some ENVI code to extract only the single band of interest for subsequent analysis.

The pressure wave is not readily apparent in any individual image from a single time. However, this feature can be enhanced by making a difference image at two adjacent time steps. The radiance difference image is literally the radiance at one time versus the radiance at the previous time step.

pro meteosat_delta_images
compile_opt strictarr
on_error, 2
e = envi(/current)
if (e eq !null) then e = envi(/headless)
 dir = dialog_pickfile($
    Title = 'Select directory of Meteosat band files for time differencing', $
    /dir, /must_exist)
if (dir eq '') then return
files = file_search(filepath(root = dir, 'MSG1-*-Band-*.dat'), count = c)
files = files.sort() ; ensure they're in time order
if (c eq 0) then return
r0 = e.openraster(files[0])
d0 = r0[0].getdata()
gain = (r0[0].metadata)['DATA GAIN VALUES']
offset = (r0[0].metadata)['DATA OFFSET VALUES']
mask = where(d0 eq 0)
d0 = temporary(d0)*gain + offset
foreach r, r0 do r.close
for i = 1, c - 1 do begin
    print, (i + 1).tostring() + '/' + c.tostring(), ' ', files[i]
    r1 = e.openraster(files[i])
    d1 = r1[0].getdata()
    mask = where(d1 eq 0)
    d1 = temporary(d1)*gain + offset
    delta = d1 - d0
    delta[mask] = 0
    fn = filepath(root = file_dirname(files[i]), $
        file_basename(files[i], '.dat') + '_diff.dat')
    if ~file_test(fn) then begin
        nr = ENVIRaster(delta, URI = fn, INHERITS_FROM = r1[0])
        nr.save
        nr.close
    endif
    foreach r, r1 do r.close
    d0 = d1
endfor
end

	  

This was repeated for the GOES ABI datasets, as well. The difference images were visually inspected to remove any “noisy” images from the sequence. Additionally, I had downloaded image sets that overlapped in time over two or more sensors, so another step involved finding the “best” image sequence to make a coherent animation, across all three sensors. I chose a sequence of GOES-17 – GOES-16 – Meteosat-8 – GOES-17 for the final video.

There are many ways to approach the task of generating an image animation for these datasets. For this example, I went old-school IDL® to generate the image overlays and annotations, using a combination of Direct Graphics and Object Graphics. Although the original imagery can be over 5,000 pixels per side, for GOES ABI data, the animation input will be normalized to only 1,024 pixels per side for all three datasets.

The following snippet grabs the pixel locations of the continental outlines for a geostationary satellite view at 41.5 degrees East for the Meteosat-8 data, in a 1024 x 1024 buffer.

window, /free, /pixmap, xsize = 1024, ysize = 1024
map_set, sat_p=[35786023.0,0,0], /satellite, 0., 41.5, /noborder ; East
map_continents, /hi
mask = tvrd(channel = 0)
wdelete, !d.window
xy = array_indices(mask, where(mask ne 0))

	  

(For the GOES data, simply modify the longitude appropriately for each satellite’s location.)

These steps were repeated for the two GOES datasets, an exercise left for the reader. The thresholded and scaled radiance difference images, with continental overlays applied, were then sequenced by time stamp over all three datasets, adding an appropriate text annotation to each.

deltas = ['10.', '1.']
foreach delta, deltas do begin
  restore, 'GOES16_ABI_Band_9_Delta_' + delta + '_Tonga.sav'
  allframes = frames
  allsources = hash()
  foreach k, frames.keys() do begin
    allsources[k] = ‘16’
  endforeach
  restore, 'GOES17 + '_ABI_Band_9_Delta_' + delta + ‘ _Tonga.sav'
  foreach k, frames.keys() do begin
    allsources[k] = ‘17’
    AllFrames[k] = frames[k]
  endforeach

  restore, 'Meteosat-8 SEVIRI Band 6_Delta_' + delta + '_degree_Tonga.sav'
  foreach k, frames.keys() do begin
    allsources[k] = ‘M’
    AllFrames[k] = frames[k]
  endforeach
  outframe = 1

  ks = (AllFrames.keys()).sort()
  time0 = 15*24+4+10/60. ; approximate eruption time
  foreach k, ks, index do begin
    source = allsources[k]
    day = strmid(k, 6, 2)
    hr = strmid(k, 8, 2)
    mi = strmid(k, 10, 2)
    t1 = day*24+hr+mi/60.
    deltat = t1 - time0
    hh = long(deltat)
    mm = abs(round((deltat - hh)*60.))
    time = 'Time from eruption (HH:MM) = ' + $
      (deltat lt 0 ? '-' : '') + hh.tostring() + ':' + string(mm, format = '(I2.2)')
    if (index eq 0) then begin
      i = image(AllFrames[k], dim=[1024, 1024], margin = [0, 0, 0, 0])
      t0  = text(.01, .95, ["Hunga Tonga-Hunga Ha'apai eruption", $
        'Sources: NOAA-NASA GOES-16 & -17 ABI L2 Band 9', $
        'Meteosat-8 SEVIRI Band ' + SEVIBand, $
        'Radiance Difference (+/- ' + delta + ')'], $
        /normal, color = 'Yellow')
      t  = text(.01, .92, time, /normal, color = 'Yellow')
    endif else begin
      i.setdata, AllFrames[k]
      t.string = time
    endelse

    write_png, filepath('frame' + string(outframe++, format = '(I3.3)') + '.png'), $
      i.copywindow()
  endforeach
  i.close
endforeach
	  

In this example, I wrote the frames individually to PNG output files in separate directories first in order to inspect them. However, the following video creation step could be integrated into that logic.

pro tongamovie
compile_opt strictarr
on_error, 2
d1 = dialog_pickfile(/must_exist, /dir, $
    title = 'Select directory of 1-delta PNG images')
if (d1 eq '') then return
d2 = dialog_pickfile(/must_exist, /dir, $
    title = 'Select directory of 10-delta PNG images')
if (d2 eq '') then return
n = file_search(d1 + '*.png', count = c)
if (c eq 0) then return
; write the MP4 video to the same directory as where this routine lives
oVid = IDLffVideoWrite(filepath(root=routine_dir(), 'tonga.mp4'), FORMAT='mp4')
fps = 3
; put the 10 and 1 radiance difference images side-by-side
newx = Round(1920/2*.75)
stream = oVid.AddVideoStream(newx*2, newx, fps)
fin = bytarr(3, 2*newx, newx)
for i = 1, c do begin
    f = 'frame' + string(i, format = '(i3.3)') + '.png'
    i1 = congrid(read_png(filepath(root = d1, f)), 3, newx, newx)
    i2 = congrid(read_png(filepath(root = d2, f)), 3, newx, newx)
    fin[0, newx, 0] = i1
    fin[0, 0, 0] = i2
    r = ovid.put(stream, fin)
endfor
end

	  

Notice that the pressure wave is still clearly visible after a full traversal back to the origin in the animation, a feature that might be quite difficult to pick out by eye in any individual, static image. I have not followed up further to see how much longer the feature is visible in the IR. I would suspect that it would require setting the radiance clamping threshold (currently set to 1.0 in this example) to lower and lower values before the signal disappears completely into the background noise as the wave’s energy slowly dissipates.

It is safe to assume that over the coming years many dissertations for advanced degrees will be generated from this singular event.