Philadelphia Refinery Explosion in CONUS and Mesoscale Sectors

June 21st, 2019 |

GOES-16 Shortwave Infrared imagery (3.9 µm) from the CONUS sector (orange label) and Mesoscale sector (white label) scans over Philadelphia at the nominal time of 0821 UTC on 21 June 2019 (Click to enlarge)

An earth-shaking refinery explosion (News story 1, 2) occurred in Philadelphia, PA under variably cloudy skies on Friday morning 21 June 2019. The toggle above shows GOES-16 3.9 µm Shortwave Infrared imagery from the mesoscale sector (with 1-minute imagery) and from the CONUS sector (with 5-minute imagery) at similar nominal times, 0821 UTC. There is a noticeable difference between the character of the signal in the mesoscale sector and the CONUS sector. This chart shows that a CONUS sector scan (shown in Blue at that link for Mode 6A that applies to GOES-16) occurs over a span of time from either 76 seconds to 233 seconds, or from 376 seconds to 533 seconds every 10 minutes (600 seconds). The nominal time of the image will be when the scan starts. Thus, the CONUS time at 08:21 UTC — that shows a very hot spot — is scanning over Philadelphia at some time in the following 2+ minutes after the scan start time (08:21:54). The Meso scale time of 08:21 UTC is scanning Philadelphia at very close to 08:21:05 UTC.

The mesoscale sector from 0820 to 0826 UTC shows no scene quite so hot as the CONUS sector. The warmest brightness temperature from the mesoscale sector, 51.5 C, occurs in the 0823 UTC image. This contrasts with the warmest brightness temperature of 118.2 C in the CONUS sector! This suggests an ephemeral explosion or rapidly-changing cloud cover that frequently masked the view (or both!).  The GOES-16 CONUS sector scanned just at the right time; the Mesoscale sector, even with higher temporal resolution, did not see the worst of the explosion.

GOES-16 Shortwave Infrared imagery (3.9 µm) from the a Mesoscale sector (white label) over Philadelphia from 0821 to 0826 UTC on 21 June 2019 (Click to enlarge)

Toggles between the CONUS image at 0821 and the mesoscale sectors at 0822 UTC and at 0823 UTC suggest that GOES-16 CONUS sector scanned Philadelphia shortly after 0822 UTC. The shift between the CONUS and mesoscale sectors is in opposite directions at the two times. (Note that Philadelphia County is outlined in both of those linked-to toggles.)

As has been observed with other similar events (for example, here, here and here), a closer look at the area showed that a nighttime thermal signature of the fire was also evident in GOES-16 Near-Infrared 1.61 µm and 2.24 µm imagery (below).

GOES-16 Near-Infrared (1.61 µm and 2.24 µm) and Shortwave Infrared (3.9 µm) images [click to play animation | MP4]

GOES-16 Near-Infrared (1.61 µm and 2.24 µm) and Shortwave Infrared (3.9 µm) images [click to play animation | MP4]

Transport of Canadian wildfire smoke across the Northeast US

May 22nd, 2019 |

GOES-16 CIMSS Natural Color images, with an overlay of the Smoke Detection Product ;click to play animation | MP4]

GOES-16 CIMSS Natural Color RGB images, with an overlay of the Smoke Detection Product [click to play animation | MP4]

GOES-16 (GOES-East) CIMSS Natural Color Red-Green-Blue (RGB) images with an overlay of the Smoke Detection Product (above) revealed curved filaments of wildfire smoke moving southeastward across the Northeast US and the adjacent offshore waters of the Atlantic Ocean on 22 May 2019. The smoke filaments were classified as Medium/High confidence by the algorithm — additional information on GOES-R Aerosol Detection Products in AWIPS is available here and here.

During the preceding overnight hours, with ample illumination from the Moon (in the Waning Gibbous phase, at 92% of Full) smoke filaments were evident over the Atlantic Ocean on Suomi NPP VIIRS Day/Night Band (0.7 µm) imagery at 0722 UTC or 3:22 AM Eastern Time (below). Note that the smoke did not exhibit a signature in the corresponding VIIRS Infrared Window (11.45 µm) image, since thin smoke layers are effectively transparent to infrared radiation.

Suomi NPP VIIRS Day/Night Band (0.7 µm) and Infrared Window (11.45 µm) images at 0722 UTC [click to enlarge]

Suomi NPP VIIRS Day/Night Band (0.7 µm) and Infrared Window (11.45 µm) images at 0722 UTC [click to enlarge]

Daily composites of Suomi NPP VIIRS True Color RGB images with VIIRS Fire Detections viewed using RealEarth (below) showed that thick smoke from wildfires in northern Alberta — primarily the Chuckegg Creek Fire that forced evacuations in the town of High Level — was initially lofted above the meteorological clouds over the Northwest Territories and Nunavut on 19 May and 20 May, before eventually moving southeastward across central/eastern Canada.

Daily composites of Suomi NPP VIIRS True Color RGB images with Fire Detections, 18-22 May [click to play animation]

Daily composites of Suomi NPP VIIRS True Color RGB images with Fire Detections, 18-22 May [click to play animation]

HYSPLIT model 72-hour back trajectories from 3 points corresponding to the smoke filaments seen in the GOES-16 imagery off the Northeast US coast (below) confirmed an initial anticyclonic transport from the region of the Alberta wildfires, with a subsequent southeastward transport across Canada and eventually the Northeast US.

HYSPLIT model 72-hour back trajectories from 3 points off the Northeast US coast [click to enlarge]

HYSPLIT model 72-hour back trajectories from 3 points off the Northeast US coast

6-hourly GFS 500 hPa analyses (source) shown below help to explain the smoke transport as seen in both the VIIRS imagery and the HYSPLIT trajectories — a ridge of high pressure was present over western Canada early in the period, with a transition to a deepening longwave trough over eastern Canada with a shortwave trough digging across Quebec and the Maritimes on 21-22 May. Strong descent of the trajectories occurred during the final 12 hours of transport, on the back side of the digging shortwave trough.

6-hourly GFS 500 hPa analyses [click to enlarge]

6-hourly GFS 500 hPa analyses, from 12 UTC on 18 May to 12 UTC on 22 May [click to enlarge]

Wildfire in Alaska

May 1st, 2019 |

GOES-17 Shortwave Infrared (3.9 µm) and "Red" Visible (0.64 µm) images [click to play animation | MP4]

GOES-17 Shortwave Infrared (3.9 µm) and “Red” Visible (0.64 µm) images [click to play animation | MP4]

On 01 May, GOES-17 (GOES-West) Shortwave Infrared (3.9 µm) and “Red” Visible (0.64 µm) images (above) showed the thermal anomaly (or fire “hot spot”) and dispersion of smoke from the first moderate-size wildfire of 2019 in the Interior of Alaska — the Oregon Lakes Impact Area Fire about 7 miles southwest of Fort Greely. This fire grew from 30 acres to 4000 acres in a 24-hour period, aided by warm daytime temperatures with low relative humidity values and southwest winds late in the day on 30 April (surface data). The Oregon Lakes Impact Area Fire was burning in a remote area just west of the Delta River which was previously burned by the 2013 Mississippi Fire; that area also contained unexploded ordnance dropped by military aircraft during training exercises.

A toggle between Suomi NPP VIIRS Day/Night Band (0.7 µm) and Shortwave Infrared (3.74 µm) images at 1216 UTC or 4:16 am local time (below) revealed the nighttime glow of the fire, along with a more accurate depiction of the size and location of the thermal anomaly.

Suomi NPP VIIRS Day/Night Band (0.7 µm) and Shortwave Infrared (3.74 µm) images [click to enlarge]

Suomi NPP VIIRS Day/Night Band (0.7 µm) and Shortwave Infrared (3.74 µm) images [click to enlarge]

Although the color enhancements were different, a comparison of Shortwave Infrared images from Suomi NPP (3.74 µm) at 1216 UTC and GOES-17 (3.9 µm) at 1220 UTC (below) demonstrated the advantage of imagery from polar-orbiting satellites at high latitudes. In this example, the 375-meter resolution VIIRS image showed 2 distinct fire hot spots that were not apparent in the lower spatial resolution — 2 km at nadir, decreasing to about 4 km over Alaska — GOES-17 image.

Shortwave Infrared images from Suomi NPP (3.74 µm) and GOES-17 (3.9 µm) [click to enlarge]

Shortwave Infrared images from Suomi NPP (3.74 µm) and GOES-17 (3.9 µm) [click to enlarge]

A larger-scale view of GOES-17 Shortwave Infrared and Visible images from 02-04 UTC on 02 May (below) showed the fire as it exhibited its peak 3.9 µm infrared brightness temperature (51.3ºC or 324.5 K at 0210 UTC) and the smoke plume had drifted over 100 miles to the southeast, moving over Beaver Creek, Yukon (CYXQ). While most of the smoke was apparently lofted above the boundary layer, the surface visibility at Fort Greely PABI was briefly reduced to 6 miles at 09 UTC or 1am local time on 02 May. Note the lack of “false cold pixels” adjacent to the warmest 3.9 µm pixels — this is due to a recent change to the GOES-R ABI Band 7 resampler, as detailed in this blog post.

GOES-17 Shortwave Infrared (3.9 µm) and "Red" Visible (0.64 µm) images [click to play animation | MP4]

GOES-17 Shortwave Infrared (3.9 µm) and “Red” Visible (0.64 µm) images [click to play animation | MP4]

A comparison of Visible and Shortwave Infrared images from GOES-17 and GOES-15 (below) highlighted the improved fire detection and monitoring capability of the new GOES-R series. The higher spatial resolution (0.5 km vs 1.0 km at nadir for Visible, and 2 km vs 4 km at nadir for Shortwave Infrared) and more frequent image scans (10 minutes for GOES-17 Full Disk vs 15-30 minutes for GOES-15 CONUS sector) along with better Image Navigation and Registration (INR) were especially valuable at the higher latitudes of Alaska. For example, the subtle behavior of the fire’s smoke column vertical jump at 2350 UTC was only apparent in the GOES-17 Visible imagery.

GOES-17 Visible (0.64 µm, top left), GOES-15 Visible (0.63 µm, top right), GOES-17 Shortwave Infrared (3.9 µm, bottom left) and GOES-15 Shortwave Infrared (3.9 µm, bottom right) images [click to play animation | MP4]

GOES-17 Visible (0.64 µm, top left), GOES-15 Visible (0.63 µm, top right), GOES-17 Shortwave Infrared (3.9 µm, bottom left) and GOES-15 Shortwave Infrared (3.9 µm, bottom right) images [click to play animation | MP4]

Since the fire was also located within the GOES-17 Mesoscale Domain Sector #2, 1-minute imagery provided an even better depiction of the fire’s smoke column vertical jump and downstream smoke transport (below).

GOES-17 Visible (0.64 µm, top left), GOES-15 Visible (0.63 µm, top right), GOES-17 Shortwave Infrared (3.9 µm, bottom left) and GOES-15 Shortwave Infrared (3.9 µm, bottom right) images [click to play animation | MP4]

GOES-17 Visible (0.64 µm, top left), GOES-15 Visible (0.63 µm, top right), GOES-17 Shortwave Infrared (3.9 µm, bottom left) and GOES-15 Shortwave Infrared (3.9 µm, bottom right) images [click to play animation | MP4]

Change to the GOES-R ABI Band 7 (3.9 µm) Resampler

May 1st, 2019 |

GOES-17 3.9 µm imagery around a fire at 23:30 UTC on 17 February 2019 with the former interpolation scheme (left), the updated interpolation scheme (center) and the difference field between the two (right). The yellow box shows the approximate fire location over Mexico. (Image courtesy Chris Schmidt, CIMSS)

GOES-R Advanced Baseline Imagery (ABI) detections must be interpolated from the detector grid on the satellite to a grid that is fixed and geographically referenced. This is accomplished by applying a truncated sinc function in both north-south and east-west directions to the data on the detector grid. Sinc functions include small negative tails adjacent to the large central maximum; for fifteen out of sixteen ABI bands, those subtractions are not detectable. For Band 7, however, the shortwave infrared band at 3.9 µm, the ABI band with the largest dynamic range (and 14 bits of information), the interpolation from detector space to the fixed grid pixel can introduce negative values of radiances and careful observers have seen Cold Pixels Around Fires, the so-called CPAF effect.

An improved interpolation for Band 7 only has been implemented (on 23 April for GOES-16 and on 18 April for GOES-17) in the GOES-R Ground System that reduces the negative tail in the Truncated Sinc function. In the single image above, from GOES-17 at 23:30 UTC on 17 February, the “old” truncated sinc function (denoted ‘Original’ in the image) has generated a falsely cold pixel — white in the greyscale enhancement — off the southeast corner of the warm pixels shown in black.  The cold pixels are not present when the new, improved interpolation scheme is used. Note, however, that the Data Max annotated in the image has cooled by 2K with the improved interpolation;  a fire is nevertheless obvious.

Consider the animation below, for example, (from this blog post on the Cranston fire), that used the ‘old’ interpolation scheme.  Cold pixels (in white) occasionally appear around the periphery the fire (in red) in the center of the image. The new interpolation means that such cold pixels will no longer appear in the data.

GOES-16 ABI visible imagery (0.64 µm) and shortwave infrared imagery (3.9 µm) over the Cranston fire, 1842 UTC on 25 July 2018 to 0227 UTC on 26 July 2018  (Click to enlarge)

The image below shows a fire at 1641 UTC on 29 April 2019, after the CPAF change was implemented into the GOES-R Ground System (two different enhancements are shown). No artificial cold pixels are present. The hottest pixel is 405 K, which would have produced a CPAF under the original truncated sinc kernel.

GOES-16 3.9 µm Imagery at 16:41 UTC on 29 April 2019 (Image courtesy Chris Schmidt, CIMSS)(Click to enlarge)