Note: GOES-R IFR Probabilities are computed using Legacy GOES (GOES-13 and GOES-15) and Rapid Refresh model information; GOES-16 data will be incorporated into the IFR Probability algorithm in late 2017.
The weather.gov website on Wednesday morning 10 May 2017 showed two dense fog advisories, one near Cincinnati, OH and one near Greensboro, NC. The aviation weather website showed an IFR Sigmet in between the two regions of dense fog. The fog formed along a stationary front that sat over the region.
How well did GOES-R IFR Probabilities and GOES-13 Brightness Temperature Difference fields capture this event? The animation of GOES-R IFR Probability, below, computed using data from GOES-13 and the Rapid Refresh Model, shows enhanced probabilities early in the evening that increased with time. The orientation of the field — from west-northwest to east-southeast — aligns well with the regions of developing fog.
GOES-R IFR Probability fields, 0100, 0400 and 0700 – 1100 UTC on 10 May 2017 (Click to enlarge)
The brightness temperature difference field, below, did not perform as well in outlining the region of low ceilings/reduced visibilities because of the presence of high clouds that interfered with the ability to detect low clouds. Consequently, the highest brightness temperature differences (3.9 µm – 10.7 µm) do not align so well with the regions of developing fog. Note also that at the end of the animation — 1100 UTC — increasing amounts of reflected solar 3.9 µm radiation is changing the character of the field from negative to positive. In contrast, the IFR Probability fields (above) maintain a consistent signal through sunrise.
GOES-13 Brightness Temperature Difference fields (3.9 µm – 10.7 µm), 0700-1100 UTC on 10 May 2017 (Click to enlarge)
GOES-R IFR Probability fields, hourly from 0215 UTC through 1115 UTC on 28 September 2016 (Click to enlarge)
IFR and Low IFR Conditions developed over parts of Virginia and the Carolinas Piedmont Region during the morning of 28 September 2016. The screengrab below, from the Aviation Weather Center, shows the areal extent of the reduced visibilities and/or low ceilings. (The text of the IFR Sigmet is here.) Fog over southeastern Virginia is developing under multiple cloud decks associated with the convection near a front. IFR Probabiities in this region are determined by Rapid Refresh data that shows low-level saturation; the flat-looking field over that region is characteristic of model-only IFR Probability fields. Farther to the southwest, over western North Carolina, IFR Probabilities are determined by both satellite and model data; notice how pixelated the data are in that region.
Screengrab from the Aviation Weather Center at 1215 UTC on 28 September 2016 (Click to enlarge)
Suomi NPP overflew the eastern United States shortly after 0730 UTC, and the toggle below shows the Day Night Visible Band and the Brightness Temperature Difference field (11.45 – 3.74 ). Water-based clouds (yellow and orange in the enhancement used) are detected just to the west of cirrus and mixed-phase clouds (black in the enhancement used). The 0737 UTC IFR Probability field, at bottom, had model-data only as predictors in regions where Suomi NPP shows multiple cloud layers. Note also that the 1-km resolution of Suomi NPP is resolving the developing valley fogs in the Appalachian mountains of Ohio, West Virginia and Kentucky. There are only a few pixels in the IFR Probability field that are suggesting valley fog development — but note in the end of the animation at the top of this post that more valley pixels show IFR Probability signals. When GOES-R is flying, its superior (to GOES-13) 2-km resolution should mitigate this too-slow identification of valley fogs.
Suomi NPP Day Night Band Visible (0.70) and Brightness Temperature Difference (11.45 – 3.74) fields at 0736 UTC on 28 September 2016 (Click to enlarge)
GOES-R IFR Probability fields at 0737 UTC, and surface observations of ceilings and visibilities at 0800 UTC (Click to enlarge)
GOES-R IFR Probabilities, Hourly from 0215-1415 UTC 17 March 2016 (Click to enlarge)
GOES-R IFR Probabilities captured the development of coastal fog over coast of the Atlantic Ocean from Long Island south to North Carolina on the morning of March 17 2016 behind a weak cold front. IFR Conditions penetrated into the Delaware and Lehigh River Valleys over Pennsylvania. In general, GOES-R IFR Probability fields captured the region of IFR conditions as it developed and expanded. Note that the IFR Probabilities remained elevated through 1400 UTC along the eastern shore of Chesapeake Bay. The 1413 UTC webcam image from Tilghman Island, below, (source), shows an offshore fogbank.
GOES-R IFR Probabilities, hourly from 0400 through 1515 UTC, 5 February 2016 (Click to enlarge)
IFR Conditions frequently occur with storms along the East Coast. Satellite detection of such conditions is very difficult because of the multiple cloud layers that accompany cyclogenesis. The IFR Probabilities, above, have a character that reflects their determination solely from Rapid Refresh Data. That is, Satellite Predictors were not considered over much of New England because of the presence of multiple cloud layers, as suggested in the Water Vapor animation below.
IFR Probability fields are initially entirely offshore in the animation above, and IFR conditions are not observed over southern New England. Note how IFR probabilities initially increase over land over southern New Jersey and then quickly move northeastward into southern New England as IFR Conditions develop. Because satellite predictors are unavailable in these regions (on account of the many clouds layers), the simultaneous development of high IFR Probabilities with observed IFR Conditions argues for a good simulation of the observed weather by the Rapid Refresh. Fused data products such as IFR Probability fields join the strengths of different systems to provide a statistically more robust field than is possible from the individual pieces.
When daytime arrives — at around 1215 UTC in the animation above — a distinct transition is apparent in the GOES-R IFR Probability fields. This occurs because Satellite Data — visible satellite data — can be used during daytime to articulate the regions of cloudiness with more precision. Because cloudiness in general is better defined, IFR Probability fields (that require the presence of clouds) increase somewhat, and the color table used emphasizes that change.
GOES-13 Water Vapor 3-hourly Animation, 0400-1300 UTC, 5 February 2016 (Click to enlarge)
GOES-R IFR Probabilities, hourly from 0000 UTC through 1315 UTC on 22 October 2015 (Click to enlarge)
Dense fog developed over South and North Carolina on the morning of 22 October 2015, just inland from the coast (Click here for screenshot from the National Weather Service homepage from Wilmington NC). The animation, above, shows the hourly evolution of the GOES-R IFR Probability fields from just after sunset on the 21st through sunrise on the 22nd. Highest probabilities of IFR Conditions overlap the stations where IFR Conditions occur. GOES-R Cloud Thickness also gives information on where the thickest clouds are; the 1115 UTC Cloud Thickness (the final field before twilight conditions change the 3.9 µm emissivity used to diagnose Cloud Thickness) shows the thickest clouds removed from the coast; compare that region to the regions with fog remaining at 1315 UTC in the animation above. There is good overlap. The maximum cloud thickness in the just-before-sunrise image below is a bit over 1000 feet, in northeastern North Carolina; according to this scatterplot that relates pre-sunrise cloud thickness to dissipation time, fog dissipation should occur within 3 hours, that is by 1415 UTC.
GOES-R Cloud Thickness, 1115 UTC on 22 October (Click to enlarge)
IFR Probabilities have better statistics in outlining regions of dense fog. This is because fog that reduces visibility and low stratus that does not reduce visibility can look very similar to a satellite. IFR Probability fields incorporate near-surface information in the guise of Rapid Refresh model predictions of low-level saturation that better refine regions where low stratus extends down to the ground. There are regions where brightness temperature difference has a fog-like signal with high ceilings/good visibility (central South Carolina, for example). These regions have low IFR Probability values because the Rapid Refresh model does not predict low-level saturation. Fusing satellite and model data yields a better product.
GOES-13 Brightness Temperature Difference (10.7µm – 3.9µm) and IFR Probability fields, 0615 UTC on 22 October 2015 (Click to enlarge)
The visible animation, below, shows that fog dissipated completely shortly after 1430 UTC, in accordance with expectations based on the Cloud Thickness.
GOES-13 Visible Imagery, 1315 through 1445 UTC on 22 October (Click to enlarge)
Suomi NPP Day Night Visible (0.70 µm) Image, 0744 UTC on 1 October 2015 (Click to enlarge)
When thick clouds are present, as along the East Coast of the United States early on 1 October 2015, as depicted by the Suomi NPP Day Night band (0.70 µm) image above, satellite detection of low clouds is problematic. This is why GOES-R IFR Probability fields incorporate information from model data so that useful guidance can be produced on whether IFR Conditions exist.
IFR Probabilities increase slowly over coastal South and North Carolina after midnight on 1 October — and the fields do a good job of outlining where IFR Conditions are occurring. In most locations, at most times, the fields are not pixelated. The smooth nature arises when model fields (which are relatively smooth) are used (and satellite data that are more pixelated are not used) to generate the IFR Probability Fields. Some holes in the extensive cloud cover occur over North Carolina during the animation: the IFR Probability field takes on a more pixelated appearance when that happens — and the Probability value increases when satellite data can also be used as a Predictor.
GOES-R IFR Probability fields created with GOES-13 Imager and Rapid Refresh model Data, 0500-1215 UTC on 1 October 2015 (Click to enlarge)
The GOES-R IFR Probability field at 1145 UTC includes a north-south oriented artifact. To the east of the obvious line, day-time predictors are used in the GOES-R IFR Probability computation; to the west, night-time predictors are used. One of the daytime predictors is Visible Imagery that is used to cloud-clear more accurately. The IFR Probability where daytime predictors are used is larger because there is more confidence that a cloud does exist.
GOES-R IFR Probability Fields computed from GOES-East and Rapid Refresh Data, hourly from 2300 7 May through 1200 8 May, and surface observations of ceilings/visibility (Click to enlarge)
GOES-R IFR Probability Fields showed large values at sunset over the Atlantic Ocean east of New Jersey and the Delmarva Peninsula. As night progressed, that fog penetrated inland. The IFR Probability field accurately depicts the region where visibilities due to fog were reduced. The 0400 UTC image in the animation above (reproduced below), has qualities that highlight the benefit of a fused product. The Ocean to the east of the southern Delmarva peninsula is overlain with multiple cloud layers that make satellite detection of low clouds/fog problematic. In this region, satellite data cannot be used as a predictor and the GOES-R IFR Probability field is a flat field. Because the GOES-R IFR Probability product includes information from the Rapid Refresh model (2-3 hour model forecasts, typically) and because saturation is indicated in the lowest 1000 feet of the model, IFR Probabilities over the ocean are high in a region where satellite data cannot be used as a predictor.
As above, but for 0400 UTC 8 May 2015 only (Click to enlarge)
GOES-R IFR Probabilities can also be computed using MODIS data, which data has better spatial resolution than GOES (1-km vs. nominal 4-km). The toggle below of the MODIS brightness temperature difference and the GOES-R IFR Probability shows a very sharp edge to the expanding fog field over New Jersey.
MODIS Brightness Temperature Difference (11µm – 3.9µm) and MODIS-based GOES-R IFR Probabilities, ~0250 UTC on 8 May, 2015 (Click to enlarge)
The Gulf Stream is apparent in the Brightness Temperature Difference field above and IFR Probabilities are high over the ocean between the coast and the Gulf Stream. In the absence of observations, how much should those high IFR Probabilities be believed. There is high dewpoint air (mid-50s Fahrenheit) along the East Coast at this time, and advection fog could be occurring, for example. Suomi NPP also overflew the region shortly after midnight. The toggle below, of brightness temperature difference and the Day Night Band confirms the presence of (presumably) low clouds over the cold Shelf Water of the mid-Atlantic bight.
Suomi NPP Brightness Temperature Difference (11.35 µm – 3.74 µm) and Suomi NPP Day Night Band Visible Imagery (0.70 µm) at night, 0643 UTC on 8 May, 2015
GOES-R IFR Probabilities, 0800 through 1400 UTC on 27 April 2015 (Click to enlarge)
Fog developed along Lake Erie after sunrise on Monday 27 April 2015. The fog and low ceilings were associated with a line of light showers, so multiple cloud layers were present. These layers inhibit satellite detection of fog/low stratus. GOES-R IFR Probabilities, above, computed using GOES-13 Satellite data and Rapid Refresh Model Output show very low probabilities at 0800 and 1015 UTC (stratus clouds are observed); at 1215 UTC, IFR Probabilities increase in the counties adjacent to Lake Erie in Pennsylvania and New York; at 1315 and 1400 UTC, IFR Probabilities are high, and IFR conditions are observed in both Erie PA and Dunkirk NY.
GOES-13 Brightness Temperature Difference Fields, 0400 through 1000 UTC on 27 April 2015 (Click to enlarge)
Brightness Temperature Difference fields overnight, above, showed evidence of water-based clouds over much of the area. The fields are moving south, however, leaving the lakeshore behind. The toggle below is of Brightness Temperature Difference and IFR Probability from 1215 UTC. It is far more difficult to relate features in the brightness temperature difference field with reductions in observations at the surface than it is to relate IFR Probability fields with surface observations. Note also that the character of the brightness temperature difference field below has changed because reflected solar radiation at 3.9 µm has become important.
Toggle between GOES-R IFR Probability fields and GOES East Brightness Temperature Difference Fields at 1215 UTC on 27 April 2015 (Click to enlarge)
Brightness Temperature Difference Fields (10.7µm – 3.9µm) from GOES-East, hourly from 0400-1300 UTC on 2 March 2015 (Click to enlarge)
A series of frontal systems along the east coast caused multiple cloud layers and IFR conditions over much of the deep south and piedmont from Mississippi to Georgia and up through Virginia overnight on 1-2 March 2015. The animation of Brightness Temperature Difference (10.7µm – 3.9µm), above, is testimony to the difficulty in using that product as a fog detection device when multiple cloud layers are present: Many stations underneath cirrus show IFR Conditions. (Note also how the signal changes at 1300 UTC — the end of the animation — as the sun rises and increasing amounts of 3.9µm solar radiation is reflected off the clouds).
The GOES-R IFR Probability fields, below, better identify regions of reduced visibility underneath mixed cloud layers. It does this by incorporating model data (from the Rapid Refresh) into its suite of predictors. Thus, where low-level saturation is indicated under multiple cloud layers, IFR Conditions can be assumed to be occurring, and computed IFR Probabilities are large. The hourly animation of GOES-R IFR Probability, below, shows a good overlap between large IFR Probabilities and IFR (or near-IFR conditions). The flat IFR Probability field that is widespread over the Piedmont region of the Carolinas is typical of an IFR Probability field determined mostly by model output. Where high clouds break, pixelated regions develop (and IFR Probabilities increase) in the field.
GOES-R IFR Probability fields, 0400-1300 UTC on 2 March (Click to enlarge)
The 1215 UTC and 1300 UTC imagery in the IFR Probability animation above include a discernible nearly north-south line. (The 1215 UTC image is below). This is the terminator. To the right of that line, where IFR Probabilities are slightly larger (dark orange), daytime predictors are being used; to the left of that line, IFR Probabilities are slightly smaller (lighter orange) and nighttime predictors are being used. Why is the probability a bit larger in the daytime predictors? In part this is because visible imagery can be used to ascertain whether clouds are present. You can be a bit more confident that IFR conditions are present because clouds are present in the visible imagery.
GOES-R IFR Probability fields, 1215 UTC on 2 March (Click to enlarge)
GOES-East Water Vapor imagery, every 2 hours, 0130 – 1330 UTC 23 December 2014 (Click to enlarge)
A slow-moving storm system is producing widespread fog and low clouds on the east coast (and in the middle of the country as well). The water vapor animation above shows the cloud cover associated with the system. Water vapor imagery such as this suggests many different cloud layers, and in such cases the IFR Probability fields (below) rely on Rapid Refresh Data to provide information because Satellite signals of low clouds cannot occur in the presence of cirrus contamination. A simple Brightness Temperature Difference product would give little information about near-surface clouds over the Southeast.
GOES-East-based GOES-R IFR Probabilities and surface-based observations of ceilings and visibilities, hourly from 0145 through 1245 UTC 23 December 2013 (Click to enlarge)
IFR Probability fields show a flat nature that occurs when satellite data cannot be used as a Predictor because of the presence of high clouds/multiple cloud layers. The Probability values are suppressed; interpretation of those values should be colored by the knowledge of the presence or absence of high clouds. In the example above, when high clouds briefly separate over central South Carolina around 0600 UTC, a region of higher IFR probability is shown. The algorithm is more confident that fog/low stratus exists because Satellite Predictors can also be used in that region. What changes is the ability of the GOES-R IFR Probability algorithm to assess the probability of IFR conditions because more predictors can be included; in the region where the high clouds part, satellite information about the low clouds can be included, and IFR Probabilities increase as a result.