Why 1-minute satellite data matters: Monitoring Fires

February 18th, 2016 |

GOES-14 0.63 µm Visible (top) and 3.9 µm Shortwave Infrared (bottom) images [click to play MP4 animation]

GOES-14 0.63 µm Visible (top) and 3.9 µm Shortwave Infrared (bottom) images [click to play MP4 animation]

Extensive wildfires (well-forecast by the Storm Prediction Center) occurred over the southern Plains on Thursday 18 February 2016, while GOES-14 was operating in SRSO-R mode. A comparison of 1-minute GOES-14 Visible (0.63 µm) and Shortwave Infrared (3.9 µm) images (above; also available as a large 112 Mbyte animated GIF) showed the broad areal coverage of smoke plumes and fire hot spots (dark black to yellow to red pixels) during the day over eastern Oklahoma.

GOES-14 0.63 µm Visible (left) and 3.9 µm Shortwave Infrared (right) images [click to play MP4 animation]

GOES-14 0.63 µm Visible (left) and 3.9 µm Shortwave Infrared (right) images [click to play MP4 animation]

Of particular interest was a rapidly-intensifying and fast-moving grass fire over northwestern Oklahoma, in Harper County just west-northwest of the town of Buffalo, which burned 17,280 acres (media report). Note the warm air temperatures as seen in the surface plots — the high of 90º F at Gage OK (KGAG, south of the fire) tied for the warmest February temperature on record at that site. A closer view of the Buffalo fire is shown above — county outlines are shown as dashed white lines, while US and State highways are plotted in violet (also available as a large 63 Mbyte animated GIF). The shortwave infrared images revealed the initial appearance of a color-enhanced fire hot spot (exhibiting an IR brightness temperature of 327.5 K) at 2045 UTC; three minutes later (at 2048 UTC), the IR brightness temperature had already increased to 341.2 K (red enhancement) which is the saturation temperature of the GOES-14 shortwave IR detectors. The hot spots could also be seen racing northeastward toward the Oklahoma/Kansas border, with the fire eventually crossing US Highway 183 (which runs south-to-north through Buffalo and across the Kansas border). The early detection and subsequent accurate tracking of such rapid fire intensification and propagation could only have been possible using 1-minute imagery.

The two plots below show GOES-14 pixel values of 3.9 µm IR brightness temperature at the initial Buffalo fire site (top plot, at 36:51º N, 99:48º W) and at a site just to the northeast (bottom plot, at 36:54º N, 99:43º W) through which the moving fire propagated. The blue line shows every value, nominally at 1-minute intervals. The red dots show points sampled every five minutes. Very small temporal scale changes in the fire cannot be captured with a 5-minute sampling interval.

GOES-14 Shortwave Infrared (3.9 µm) Brightness Temperatures at 36:51:36º N, 99:48:27º W, 2040-2230 UTC on 18 February 2016 [click to enlarge]

GOES-14 Shortwave Infrared (3.9 µm) Brightness Temperatures at 36:51:36º N, 99:48:27º W, 2040-2230 UTC on 18 February 2016 [click to enlarge]

GOES-14 Shortwave Infrared (3.9 µm) Brightness Temperatures at 36:54:44º N, 99:43:22º W, 2115-2200 UTC on 18 February 2016 [click to enlarge]

GOES-14 Shortwave Infrared (3.9 µm) Brightness Temperatures at 36:54:44º N, 99:43:22º W, 2115-2200 UTC on 18 February 2016 [click to enlarge]

 

GOES-15 (left), GOES-14 (center), and GOES-13 (right) 3.9 µm Shortwave Infrared images covering the initial period 2030-2100 UTC [click to play animation]

GOES-15 (left), GOES-14 (center), and GOES-13 (right) 3.9 µm Shortwave Infrared images covering the initial period 2030-2100 UTC [click to play animation]

For the Buffalo fire, a three-satellite comparison of Shortwave Infrared (3.9 µm) images from GOES-15 (operational GOES-West), GOES-14, and GOES-13 (operational GOES-East) is shown for the initial 30-minute time period 2030-2100 UTC (above). The images are displayed in the native projection of each satellite. In terms of the first unambiguous fire hot spot detection (via a hot color-enhanced image pixel) during that initial period, it would appear from the image time stamps that both GOES-14 and GOES-13 detected the fire at 2045 UTC — however, because GOES-14 was scanning a much smaller sector, it did indeed scan the fire at 20:45 UTC (while GOES-13 scanned the fire at 2049 UTC, 4 minutes after its larger scan sector began in southern Canada). Also note that there were no GOES-15 images during that 30-minue period between 2030 and 2100 UTC, due to the satellite having to perform various “housekeeping” activities — so if a NWS forecast office AWIPS were localized to use GOES-15, initial fire detection would not have been posible until reception of the 2100 UTC image (which actually scanned the fire at 2104 UTC).

A faster animation covering a longer 2.5-hour period from 2030-2300 UTC is shown below. Again, a true sense of the fast northeastward speed of fire propagation could only be gained using 1-minute imagery.

GOES-15 (left), GOES-14 (center), and GOES-13 (right) 3.9 µm Shortwave Infrared images covering the 2.5-hour period 2030-2300 UTC [click to play animation]

GOES-15 (left), GOES-14 (center), and GOES-13 (right) 3.9 µm Shortwave Infrared images covering the 2.5 hour period 2030-2300 UTC [click to play animation]

GOES-14 Shortwave Infrared (3.9 µm) images [click to play animation]

GOES-14 Shortwave Infrared (3.9 µm) 9mages [click to play animation]

The animation above shows another view of 1-minute GOES-14 Shortwave Infrared (3.9 µm) imagery, centered over northeastern Oklahoma — in these images, the hottest fire pixels are darkest black. Time series of infrared brightness temperature values at two individual fire pixels (shown here) are plotted below. The Blue lines show the 1-minute data; Red dots show how 5-minute monitoring would have adequately captured the events. Pixel Brightness Temperature changes that occur on the order of 1 or 2 minutes are common, and peak values can be missed with 5-minute granularity.

GOES-14 Shortwave Infrared (3.9 µm) Brightness Temperatures at, 2138-2301 UTC on 18 February 2016 at 35:31:17 N, 96:05:55 W [click to enlarge]

GOES-14 Shortwave Infrared (3.9 µm) Brightness Temperatures from 2138-2301 UTC on 18 February 2016, at 35:31:17º N, 96:05:55º W [click to enlarge]

GOES-14 Shortwave Infrared (3.9 µm) Brightness Temperatures at, 2138-2301 UTC on 18 February 2016 at 35:23:51 N, 95:20:52 W [click to enlarge]

GOES-14 Shortwave Infrared (3.9 µm) Brightness Temperatures from 2138-2301 UTC on 18 February 2016, at 35:23:51º N, 95:20:52º W [click to enlarge]

In the GOES-R era, Fire Products will be produced every 5 minutes. Individual NWS Forecast Offices will be able to request Rapid-Scan Imagery (1-minute intervals) over a 1000 km x 1000 km mesoscale sector.

===== 19 February Update =====

Seen below are RealEarth comparisons of Aqua MODIS and Suomi NPP VIIRS true-color Red/Green/Blue (RGB) images from the early afternoon of 18 February (before the Buffalo OK fire) and 19 February (after the Buffalo OK fire), which revealed the long southwest-to-northeast oriented burn scar. As seen on the GOES-14 animation above, the fire crossed US Highway 183 just to the north of Buffalo (that portion of the highway was closed for several hours).

Aqua MODIS true-color images on 18 February and 19 February [click to enlarge]

Aqua MODIS true-color images on 18 February and 19 February [click to enlarge]

Suomi NPP VIIRS true-color images on 18 February and 19 February [click to enlarge]

Suomi NPP VIIRS true-color images on 18 February and 19 February [click to enlarge]

In addition, a comparison of Suomi NPP VIIRS true-color and false-color images (below) helps to discriminate between the darker burn scar and the cloud shadows seen on the true-color image — the Buffalo fire burn scar appears as varying shades of brown in both the true-color and the false-color images.

Suomi NPP VIIRS true-color and false-color images [click to enlarge]

Suomi NPP VIIRS true-color and false-color images [click to enlarge]

===== 27 February Update =====

Landsat-8 false-color RGB images on 18 February (a few hours prior to the start of the fire) and 27 February (several says after the fire) [click to enlarge]

Landsat-8 false-color RGB images on 18 February (a few hours prior to the start of the fire) and 27 February (several says after the fire) [click to enlarge]

A comparison of 30-meter resolution Landsat-8 false-color (created using OLI bands 6/5/4) RGB images from 18 February (about 3.5 hours prior to the start of the Buffalo OK fire) and 27 February (several days after the fire) provided a very detailed view of the burn scar. Note that a few green fields remained within the burn scar, and also appeared to prevent the spread of the fire along portions of its perimeter — this is a result of the vast difference between the very low moisture content of the dry grassland (which burned quickly and easily) and the high moisture content of the well-irrigated fields of winter wheat, alfalfa, and canola crops.