Fog detection when Cirrus is present: Southern Plains edition

GOES-16 Night Fog Brightness Temperature Difference (10.3 µm – 3.9 µm) at 0906, 1006, 1106, 1206, 1306 and 1401 UTC on 8 November 2019 (Click to enlarge)

Consider the stepped 1-hour images of the ‘Night Fog’ Brightness Temperature Difference (10.3 – 3.9) field, above. (Here’s one with full temporal resolution) Extensive jet stream cirrus is preventing GOES-16 from seeing evidence of low clouds, and the satellite alone is therefore failing to identify regions of low ceiling and fog over much of Texas.  (The Nighttime Microphysics RGB that depends on the Night Fog brightness temperature difference to detect low clouds suffers from a similar shortcoming here : link )

When Cirrus covers an area of low clouds, GOES-R IFR Probability does allow for low cloud/fog detection because it incorporates low-level saturation information from the Rapid Refresh to delineate regions of fog.  Similarly, it can screen out regions of mid-level stratus that can appear as fog in both the Night Fog brightness temperature difference and the Nighttime Microphysics RGB: Note the region around Childress TX, for example, where IFR Probability is low, but where the Night Fog Brightness Temperature Difference has a strong signal in the presence of elevated stratus.  Regions around Midland are highlighted in IFR Probability — Rapid Refresh data there are correctly diagnosing low ceilings and reduced visibilities.  (Here’s IFR Probability with an animation at full temporal resolution).  Because Rapid Refresh model data input into the IFR Probability computation changes every hour, you will sometimes see a pulsing in the imagery.

Use IFR Probability to detect regions of low ceilings and reduced visibility when cirrus has overspread your area.

GOES-R IFR Probability at 0906, 1006, 1106, 1206, 1306 and 1401 UTC on 8 November 2019 (Click to enlarge)

 

Fog Detection when Cirrus is present

GOES-16 NightTime Microphysics, Night Fog Brightness Temperature Difference (10.3 µm – 3.9 µm) and IFR Probability at 1206 UTC on 4 November 2019, along with 1200 UTC observations of ceilings and visibilities (Click to enlarge)

Fog can be a challenge to detect from satellite for two reasons: High clouds can block the satellite’s view of low clouds, or low clouds (stratus) can be detected but the satellite cannot accurately determine the cloud base. The toggle above cycles between the Night Fog Brightness Temperature difference (10.3 µm – 3.9 µm), the Nighttime Microphysics RGB (the Night Fog Brightness Temperature Difference is the green component of that RGB), and GOES-R IFR Probability.

The Night Fog Brightness Temperature Difference can be used to detect low stratus clouds because the small water droplets that make up those clouds have different emissivities at 10.3 µm and 3.9 µm: water droplets do not emit 3.9 µm radiation as a blackbody would, but they do emit 10.3 µm radiation more nearly like a blackbody. Thus the amount of energy at 3.9 µm collected by the satellite is smaller than it would be above a blackbody emitter. The computation of the blackbody temperature associated with the collected energy assumes a blackbody emission. Because that’s not the case at all wavelengths above clouds made of water droplets, 3.9 µm brightness temperatures are colder than those computed from observed 10.3 µm energy.

The Night Fog brightness temperature difference field allows an observer to infer the presence of low clouds over the Pacific Ocean, in and around Portland OR on the Columbia River to the ocean, and also northward towards Puget Sound, and in the Willamette Valley between Portland and Eugene OR. There is also a signal over north-central Oregon that is likely unrelated to clouds — variable surface soil emissivity in that region is affecting this brightness temperature difference (The GOES-R Cloud Phase product and Cloud Mask both show no cloud signal in that region).

Note that a cirrus shield is apparent over the northeast half of the image. The cirrus covers most of Puget Sound in Washington, and the Brightness Temperature Difference field can therefore not give information about low clouds (except where breaks in the cirrus occur — such as in the Strait of Juan de Fuca).

The Nighttime Microphysics RGB is a way to identify low clouds — but its information is coming from the Night Fog brightness temperature difference and it has limitations that are similar to those with the brightness temperature difference, namely: what is going on under cirrus clouds, and how close to the ground are the low clouds that are detected? In the RGB, low clouds will have a yellowish hue (red — the split window difference and green — the night fog brightness temperature difference — make yellow). The RGB shows different values over Portland than over the Willamette Valley because the clouds are warmer over Portland (as detected by the clean window 10.3 as the blue part of the RGB) and the split window difference field (the red part of the RGB) shows larger values over the Willamette Valley.

IFR Probability shows a a strong and consistent signal from Portland south to Eugene, and also over the Pacific, and north of Portland into the southern Puget Sound. In these regions, Rapid Refresh Model data are showing the presence of low-level saturation, a good indicator that IFR conditions may be present. IFR Probability shows small values under the region of low clouds detected by satellite over the Strait of Juan de Fuca; the Rapid Refresh model there is not showing low-level saturation. There is also a small signal in the clear skies over north-central Oregon where a brightness temperature difference is occurring because of soil-driven emissivity differences.

Use IFR Probability to ascertain the likelihood of fog underneath cirrus decks, or to determined the likelihood of stratus decks reaching the surface (that is, if stratus is actually fog).

Finding IFR Conditions during the day

GOES-16 Visible (0.64) imagery and GOES-16 IFR Probabilities, 1521 UTC on 7 October 2019 (Click to enlarge)

GOES-R IFR Probability fields from GOES-16, shown above in a toggle with visible imagery, can identify where IFR conditions are most likely under an extensive cloud shield. The image above shows an AWIPS combination of the visible imagery and the IFR Probability field at the same time (an intermediate step is also shown in the toggle). IFR Probability is aligning well with stations reporting IFR conditions. Stations reporting higher ceilings under the clouds are mostly outside the band of higher IFR Probabilities.

Placing the IFR Probability field underneath the Visible Image that is somewhat transparent allows a user to see where IFR conditions are most likely in the cloud field. This is a good situational awareness tool for daytime Fog.

Fog in the Mountains

GOES-16 Band 2 (0.64 µm) over Montana and northern Idaho, 1656 UTC on 23 September 2019. Major highways are also shown, including east-west I-90 (Click to enlarge)

Consider the image shown above, visible imagery (0.64 µm) from the GOES-16 Advanced Baseline Imager.    Can you tell at a glance where the IFR Conditions exist?  If you are familiar with northern Rockies topography, and know that the prevailing surface winds in this area were southwesterly, perhaps you can make a good intuitive guess:  low ceilings and reduced visibilities are likely on the upwind side of the topography.

METAR observations at 1700 UTC on 23 September 2019 plotted over topography (Click to enlarge)

GOES-R IFR Probability fields are computed from visible and infrared imagery as well as from Rapid Refresh model estimates of low-level saturation. IFR Probability fields, below, correctly show increased likelihood of IFR conditions over the mountains, and mostly on the windward side of the topography. Mullan Pass, north of the center of the image, at ~6000 feet above sea level, is observing IFR conditions.

GOES-16 IFR Probability fields, 1656 UTC on 23 September 2019 (Click to enlarge)

The animation below steps through the topography, visible imagery, ‘day fog’ brightness temperature difference (3.9 µm – 10.3 µm) and IFR Probability. The Brightness Temperature Difference field and the visible imagery show little relationship to the IFR Probability field.

Topography, GOES-16 ABI Band 2 (0.64 µm), ‘Day Fog’ Brightness temperature Difference (3.9 µm – 10.3 µm) and IFR Probability fields, all at 1656 UTC on 23 September 2019 (Click to enlarge)

An added benefit of IFR Probability fields: It can be used day or night.

Dense Fog over Kansas

Dense Fog Advisory issued by Goodland Kansas on 19 August 2019 (Click to enlarge)

Dense Fog developed over portions of the central Plains on Monday morning, 19 August 2019; The Goodland Kansas office of the National Weather Service issued Dense Fog Advisories for part of their County Warning Area, as shown above.  A variety of satellite-based products are useful for monitoring dense for from satellite.  The one used for the longest time is the so-called Night Fog Brightness Temperature Difference field, shown below, that detects (in blue/cyan, in the default AWIPS color enhancement used) clouds that contain water droplets.  Small water droplets do not emit 3.9 µm radiation as a blackbody would, but they do emit 10.3 µm radiation as a blackbody.  Thus the brightness temperature that is computed by detecting via satellite the amount of radiation emitted and converting those numbers of photons to an emitting temperature (a conversion that does assume blackbody emission!) will be warmer for 10.3 µm radiation than for at 3.9 µm radiation.  An animation of that field is shown below.

GOES-16 ABI ‘Night Fog’ brightness temperature difference field (10.3 µm – 3.9 µm), 1026 – 1421 UTC on 19 August 2019 (Click to animate)

Note in the animation above that dense fog over the higher Plains or western Kansas and Nebraska is well-articulated by this field; dense fog over Iowa and eastern Nebraska is not depicted with clarity. The animation also shows a well-known feature of the Night Fog Brightness Temperature difference field — it loses a signal during sunrise as the amount of reflected 3.9 µm radiation increases.

GOES-R IFR Probability fields, below for the same time, better capture the horizontal extent of low ceilings and reduced visibilities (i.e., IFR conditions) in this case by marrying the brightness temperature difference field with low-level saturation as predicted by the Rapid Refresh model. If satellite detection suggests clouds are present — or likely — and low-level saturation is predicted by the model, then IFR Probabilities will be large. The animation below highlights the IFR conditions over western Nebraska and Kansas — but it also highlight IFR conditions in eastern Nebraska and western Iowa, regions that the Night Fog brightness temperature difference does not capture dramatically. Note that during the animation, IFR probability fields can change quickly — as updated Rapid Refresh data (that is, a more recent model run that is often more accurate) are incorporated into the algorithm.

GOES-R IFR Probabilities, 1026 – 1431 UTC on 19 August 2019 (Click to enlarge)

Because the Night Fog brightness temperature difference does not outline fog features well in eastern Nebraska and Iowa, the Nighttime microphysics Red/Green/Blue composite, that uses the Night Fog Brightness temperature difference as its green component (Night Time Microphysics Quick Guide), similarly does not outline them, unlike IFR Probability. Be alert to the strengths and weaknesses of the product used for detecting fog.

Night time Microphysics RGB, Night Fog Brightness Temperature Difference, and GOES-16 IFR Probabililty, all at 1026 UTC on 19 August 2019 (Click to enlarge)

Fog in the Mid-South on August 8th

GOES-R IFR Probability, 1002 UTC on 8 August 2019, along with surface reports of ceilings and visibility (Click to enlarge)

GOES-R IFR Probability fields from 1002 UTC on 8 August 2019, above, show a region of high probabilities mostly colocated with surface observations of ceilings and visibilities.  IFR conditions are widespread from north of Memphis TN to southern Illinois.  IFR Probability is a fused product, using both satellite imagery to detect low clouds and Rapid Refresh model data to identify regions of low-level saturation.  Where both indicators are present (for example, over extreme western Kentucky and adjacent regions of southern Illinois and northwest Tennessee), IFR Probabilities are very high and IFR conditions are observed.  There are also regions where only model data can be used — because satellites are not detecting low clouds (because high clouds are blocking the view);  such a case is over northwest Arkansas, where IFR Probabilities are high (albeit not as high as over western Kentucky) and the field is not pixelated like a satellite image can be.  IFR Conditions are observed there as well.  (Click here for a screen shot from later in the morning of the Paducah National Weather Service office front page showing Dense Fog Advisories in Tennessee, issued by the Memphis Forecast Office).

The toggle below compares the IFR Probability field with the GOES-16 ‘Night Fog’ brightness temperature difference.  There is no consistent signal in the brightness temperature difference field to indicate fog on the ground.  The color enhancement for the brightness temperature difference is created so that teal to blue is positive.  Clouds made of liquid water droplets — such as stratus, or fog — will have a strong positive value because of the emissivity properties of small water droplets.  (That is, those droplets emit 10.3 µm radiation mostly as a blackbody would, but do not emit 3.9 µm radiation as a blackbody would).  Note in particular the very dark enhancement over northwest Arkansas (suggestive of cirrus there that will block any satellite view of low clouds) and the grey enhancement over west central Tennessee, north of Memphis, where mid-level clouds are similarly blocking a good view of the stratus at the surface.

GOES-16 IFR Probability and GOES-16 ‘Night Fog’ Brightness Temperature Difference field (10.3 µm – 3.9 µm) at 1002 UTC on 8 August 2019 (Click to enlarge)

The Nighttime Microphysics RGB can also be used to alert forecasters to fog — but a main component of that detection is the Night Fog brightness temperature difference field, and where the Night Fog brightness temperature difference fails to identify fog, as above, Nighttime Microphysics will as well, as shown in the toggle below.  Which color is useful for identifying fog in the RGB below?  Almost all of them!

Fog and Low Stratus under cirrus

Night Fog Brightness Temperature Difference (10.3 µm – 3.9 µm), 0801-1311 UTC on 7 May 2019 (Click to animate)

Consider the animation above, of the Brightness Temperature Difference product (10.3 µm – 3.9 µm) centered on Colorado on the morning of Tuesday 7 May 2019. The surface observations show widespread IFR conditions, but because of widespread high clouds over the region, the brightness temperature difference shows little signal that is consistent with low clouds (blue or cyan in this enhancement). There isn’t much horizontal spatial correlation between observations of low ceilings/reduced visibility and the Brightness Temperature Difference product. When low clouds are overlain by high clouds, don’t expect a satellite detection of low clouds to work. Note also how the Brightness Temperature Difference field loses its signal as the sun rises and a general increase in the amount of reflected solar shortwave (3.9 µm) infrared radiation increases.

GOES-R IFR Probability includes near-surface information that is useful, especially when mid-level or high clouds obscure the satellite view of low clouds. Rapid Refresh estimates of near-surface saturation are used to gauge the probability of IFR conditions. In the case on 7 May 2019, that information allowed GOES-R IFR Probability to approximate exactly where the lowest ceilings and more reduced visibilities were. There is therefore a much better spatial correlation between the location of surface observations showing IFR conditions, and the IFR Probability field. This has a tacit implication on how well the Rapid Refresh Model is simulating the evolution of the atmosphere. Note also that in contrast to the Brightness Temperature Difference field, a consistent signal is maintained through sunrise.

GOES-R IFR Probability, and surface observations of ceilings/visibility, 0801-1311 UTC on 7 May 2019 (Click to animate)

Because the NightTime Microphysics RGB relies on the Night Fog Brightness Temperature difference field, it rises or falls on fog detection on the shoulders of the Brightness Temperature Difference field. On this day, it fell. Click here for the Nighttime Microphysics RGB animation; the animation below toggles between the three fields — Night Fog Brightness Temperature Difference, Nighttime Microphysics RGB, and IFR Probability — at 1002 UTC on 7 May 2019.

Night Fog Brightness Temperature Difference (10.3 µm – 3.9 µm), Nighttime Microphysics RGB and GOES-R IFR Probability fields, 1001 UTC on 7 May 2019 (Click to enlarge)

Fog surrounding convection in the Midwestern United States

GOES-16 IFR Probability and GOES-16 ‘Night Fog’ Brightness Temperature Difference (10.3 µm – 3.9 µm) at 1001 UTC on 2 May 2019, along with surface observations of ceilings and visibility (Click to enlarge)

Widespread Fog occurred over the midwestern United States on Thursday morning, 2 May 2019, with some Dense Fog Advisories issued. Some of the fog developed under an extensive cirrus canopy associated with strong convection over north-central Illinois, and those cloud layers make satellite-based detection of stratus near the surface (i.e., fog) a challenge.  The toggle above between the Night Fog Brightness Temperature Difference (10.3 µm – 3.9 µm) and GOES-16 IFR Probability highlights the difficulty in using the brightness temperature difference alone to outline regions of fog. In contrast, the Rapid Refresh model information on low-level saturation allows IFR Probability to alert a forecaster to the likelihood of fog in regions underneath cirrus, regions such as southeastern Illinois and southwestern Indiana.

There is an excellent spatial correlation between regions with high IFR probability and low ceilings/reduced visibilities, and also with regions with low IFR probability and non-IFR conditions (central Iowa, for example, and central Illinois southwestward into central Missouri).

Because the Night Fog Brightness Temperature difference is a central feature of the Nighttime Microphysics RGB product — Night Fog Brightness Temperature difference is the green part of that RGB, as apparent in the toggle below — Nighttime Microphysics also is challenged to identify fog in regions under cirrus

Advection Fog and Multiple Cloud Layers

GOES-16 Night Fog Brightness Temperature Difference (10.3 µm – 3.9 µm), Low IFR Probability and Advanced Nighttime Microphysics RGB at 0837 UTC on 13 March 2019 (Click to enlarge)

An intense early Spring storm produced blizzard conditions over the western Great Plains on 13 March 2019 (See this blog post for example). The southerly flow in advance of the storm moved moist air over a dense snowpack over the upper Midwest, resulting in dense Advection Fog. The animation above cycles between the Night Fog Brightness Temperature Difference (10.3 µm – 3.9 µm), the Advanced Nighttime Microphysics RGB that uses the Night Fog Brightness Temperature Difference as the ‘green’ and GOES-16 Low IFR Probabilities.

At 0837 UTC, high clouds had not yet covered all of the upper Midwest, and the satellite still viewed stratus clouds over Wisconsin and Minnesota (bright cyan in the Night Fog Brightness Temperature difference enhancement). Note the difference in the Low IFR Probability field in regions where high clouds are present (Iowa and states to the south and west) and where they are not. Low IFR Probability has a pixelated look over Wisconsin because modest satellite pixel-level variability is included in the IFR probability field. Over Iowa, in contrast, satellite data gives no information about reductions to visibility under the thick cloud cover; low-level data from the Rapid Refresh model drives the Low IFR Probability field there. Probabilities are high in regions where observations suggest IFR/Low IFR conditions are present. This suggest the Rapid Refresh is simulating the evolution of the atmosphere well over the Plains.

The Advanced Nighttime Microphysics RGB can detect low clouds — but only when high clouds are not in the way. For this case, the RGB signal over Wisconsin is consistent with stratus and fog, but over Iowa the signal is consistent with cirrus that is occurring there. Surface restrictions in visibility and lowered ceilings are similar in the two states, and the IFR Probability field has similar values in the two states.

By 1502 UTC on 13 March (below), high clouds had overspread the entire Midwest; Low IFR Probability relies almost entirely on Rapid Refresh Model estimates of low-level saturation.  That is why the field over northern Minnesota and northern Wisconsin is flat.  Some holes in the high clouds are suggested by the pixelated look of the field in Iowa and southern Wisconsin.

GOES-16 Low IFR Probability, 1502 UTC on 13 March 2019 (Click to enlarge)

Cloud Layers and Detection of IFR Conditions

GOES-16 Night Fog Brightness Temperature Difference (10.3 µm – 3.9 µm), Nighttime Microphysics RGB and GOES-16 IFR Probability at 1116 UTC on 22 February 2019; Surface observations of ceilings and visibilities at 1100 UTC are also plotted (Click to enlarge).

A strong storm embedded within a subtropical jet stream over the southern United States was associated with widespread fog on the morning of 22 February 2019. This screen-capture from this site shows Dense Fog Advisories over much of Georgia, and over regions near Dallas. Which products allowed an accurate depiction of the low ceilings and reduced visibilities?

The toggle above cycles between the Night Fog Brightness Temperature Difference (10.3 µm – 3.9 µm), which product identifies low clouds (cyan blue in the default AWIPS enhancement shown) because of differences in emissivity at 3.9 µm and 10.3 µm from small water droplets that make up stratus clouds, the Nighttime Microphysics RGB, which RGB uses the Night Fog Brightness Temperature Difference as it green component, and the GOES-16 IFR Probability product.  IFR conditions are defined as surface visibilities between 1 and 3 miles, and ceiling heights between 500 and 1000 feet above ground level.  The plotted observations help define where that is occurring.  Multiple cloud layers from Arkansas east-northeastward make a satellite-only detection of IFR conditions challenging.  IFR Probability gives useful information below cloud decks because model-based saturation information from the Rapid Refresh Model fill in regions below multiple cloud decks where satellite information about low clouds is unavailable.

The toggle below shows the same three satellite-based fields (Night Fog Brightness Temperature Difference, Nighttime Microphysics RGB and IFR Probability)  at the same time, but centered over Oklahoma.  In this case, the Rapid Refresh Data are used to screen out a region of elevated stratus over northeast Oklahoma. Note that these is little in the Night Fog Brightness Temperature Difference field to distinguish between the IFR and non-IFR locations.

GOES-16 Night Fog Brightness Temperature Difference (10.3 µm – 3.9 µm), Nighttime Microphysics RGB and GOES-16 IFR Probability at 1116 UTC on 22 February 2019; Surface observations of ceilings and visibilities at 1100 UTC are also plotted (Click to enlarge).

GOES-R IFR Probability over the southeast United States in this case is identifying regions of IFR conditions underneath multiple cloud decks (and also where only the low clouds are present) by incorporating low-level saturation information from the Rapid Refresh model. Over Oklahoma, non-IFR conditions under an elevated stratus deck are identified (and screened out in IFR Probability fields) by the lack of low-level saturation information in the Rapid Refresh.