Fog over the Pacific Northwest: Where is it?

GOES-17 imagery in this blog post is made from GOES-17 Data that are preliminary and non-operational!

Night Time Fog Brightness Temperature Difference (10.3 µm – 3.9 µm) from GOES-16 and (preliminary, non-operational) GOES-17, 1202 UTC on 3 December 2018 (Click to enlarge)

The toggle above shows GOES-16 and GOES-17 Night Fog Brightness Temperature difference fields over the Pacific Northwest shortly after 1200 UTC on 3 December 2018.  The Pacific NW is a lot farther from the sub-satellite point (nadir) of GOES-16 (75.2 West Longitude) than from the sub-satellite point (nadir) of GOES-17 (at 137.2 W Longitude).  Thus, the GOES-16 view is has inferior spatial resolution.  There are also different parallax shifts for clouds between the two views.  The scene includes plenty of stratus, based on the observations, and isolated pockets of IFR conditions:  Spokane WA, Stampede Pass, WA, Pendleton OR, Medford OR, Salem OR.  It’s difficult to tell at a glance from the Brightness Temperature Difference field where the lowest ceilings and poorest visibilities are — because the satellite sees the top of the cloud, and it’s difficult to infer cloud base properties from infrared imagery of the cloud top.

The Advanced Nighttime Microphysics RGB can also be used to detect low ceilings, because its green component is the Night Fog Brightness Temperature Difference as shown above.  The toggle below compares the GOES-16 and GOES-17 RGBs.  Again, it is difficult to pinpoint at a glance where the IFR conditions are occurring in this product.

There are subtle differences in the colors of the RGB from the two satellites.  These are related to the distance from the sub-satellite point.  Limb cooling will cause the brightness temperatures from the GOES-16 Clean Window to be slightly cooler than the GOES-17 values, and that will affect the color:  The Clean Window is the Blue component of the RGB.  In addition, the Split Window Difference values will be slightly different because of the different amounts of limb cooling in 12.3 µm and 10.3 µm brightness temperatures.  GOES-17 data also includes striping that is being addressed with ongoing calibration work with the satellite.

Advanced Nighttime Microphysics RGB from GOES-16 and GOES-17 at 1202 UTC on 3 December 2018 (Click to enlarge)

The toggle below shows, (from GOES-16 data) the 10.3 µm – 3.9 µm Brightness Temperature Difference, the Nighttime Microphysics RGB, and the IFR Probability fields at 1202 UTC. IFR Probability fields include information about low-level saturation from the Rapid Refresh model, and that information allows the product to screen out regions of mid-level stratus; thus, the region of IFR conditions is better captured.  There are three main regions:  eastern WA southwestward to Pendleton OR, Seattle southward into Oregon, and the peaks of central Washington.

GOES-16 Night Fog Brightness Temperature Difference (10.3 µm – 3.9 µm), Nighttime Microphysics RGB and IFR Probability, 1202 UTC on 3 December 2018 (Click to enlarge)

Fog over the Dakotas and Minnesota: Where is it?

GOES-16 “Red Visible” Band 2 0.64 µm imagery at 1502 UTC on 21 November 2018 (Click to enlarge)

Consider the GOES-16 Visible (Band 2, 0.64 µm) Image above.  Where is the fog in this image?  Certainly you can tell where clouds exist, and if animated, you could identify snow on the ground (because it wouldn’t move like clouds do).  Alternatively, you could toggle between the visible and the Snow/Ice band (Band 5, 1.61 µm), below;  regions of snow/ice — such as in western North Dakota, or northeastern South Dakota, or Ontario appear bright in the 0.64 µm but dark in the 1.61 µm.

GOES-16 “Red Visible” Band 2 0.64 µm imagery and “Snow/Ice” Band 5 1.61 µm imagery at 1502 UTC on 21 November 2018 (Click to enlarge)

The Day Fog Brightness Temperature Difference product (3.9 µm – 10.3 µm) highlights low clouds. Stratus clouds with water droplets are scatterers of incoming solar radiation. The clouds over Minnesota and Iowa appear to be composed of much smaller water droplets, however, because they are so much warmer — the brightness temperature difference is much larger.  Smaller droplets are better scatterers of incoming solar radiation.  The image below shows the field, also at 1502 UTC on 21 November.  Are there any differences in this field that suggests fog might be present in one location, but not in the other?

GOES-16 Day Fog Brightness Temperature Difference (3.9 µm – 10.3 µm) imagery at 1502 UTC on 21 November 2018 (Click to enlarge)

The Day Snow Fog RGB composite, below, highlights regions of low clouds, snow/ice, and higher clouds.  Snow (and clouds made of ice) are shaded red, low clouds are shades of grey/blue.  Where is the fog?

GOES-16 Day Snow Fog RGB imagery at 1502 UTC on 21 November 2018 (Click to enlarge)

GOES-R IFR Probability Fields are a better predictor of where IFR conditions (that is, reduced visibility and lowered ceilings as might occur with fog) are occurring.  It combines satellite information and low-level informatio0n about saturation (from the Rapid Refresh model).  This fusing of data accentuates a satellite strength (detection of low clouds made up of water droplets — that is, stratus) and the model strength (namely, where are the low-level saturated?)   In the image below, fog and high probabilities of IFR conditions neatly overlap.  This toggle is between the visible, Day Snow-Fog RGB and IFR Probability field.

GOES-16 IFR Probability at 1502 UTC on 21 November 2018 along with surface observations of ceilings and visibility (Click to enlarge)

Fog under multiple cloud layers

GOES-16 Night Fog Brightness Temperature Difference (10.3 µm – 3.9 µm), 1137 – 1452 UTC on 31 October 2018 (Click to animate)

Consider the imagery above: The Night Fog Brightness Temperature Difference (10.3 µm – 3.9 µm) product can be used to highlight regions of low clouds (cloud made up of water droplets) because those water droplets do not emit 3.9 µm radiation as a blackbody (but do emit 10.3 µm radiation nearly as a blackbody), and the conversion of sensed radiance to brightness temperature assumes blackbody emissions. Thus, the 10.3 µm brightness temperature is warmer than at 3.9 µm. In the enhancement used, low stratus clouds are shades of cyan. Is there fog/low stratus along the coast of the Pacific Northwest? How far inland does it penetrate. It is impossible in this case (and many similar cases) to tell from satellite imagery alone because multiple cloud layers associated with a storm moving onshore prevent the satellite from seeing low clouds. The animation shows the Brightness Temperature Difference field along with surface reports of visibility and ceilings.  IFR and near-IFR conditions are widespread, but there is little correlation between their location and the satellite-only signal.

GOES-16 Night Fog Brightness Temperature Difference (10.3 µm – 3.9 µm), 1137 – 1452 UTC on 31 October 2018, along with surface reports of ceilings (AGL) and visibility (Click to animate)

GOES-R IFR Probability fields fuse together satellite information and model data to provide a better estimate of where IFR conditions might be occurring. The animation below, for the same times as above, shows a high likelihood of IFR conditions (as observed) over much of the eastern third of Washington State. The satellite doesn’t give information about the near-surface conditions in this case, but the Rapid Refresh data strongly suggests low-level saturation, so IFR probabilities are high. The field also correctly shows small likelihood of IFR probailities over coastal southern Oregon and northern California. The Rapid Refresh data used have 13-km resolution, however; fog at scales smaller than that may be present — in small valleys for example.

GOES-16 IFR Probabilities, 1137 – 1452 UTC on 31 October 2018 (Click to animate)

IFR Probability can distinguish between stratus and fog

Night Fog Brightness Temperature Difference (10.3 µm – 3.9 µm). 1152 UTC on 11 October (Click to enlarge)

Consider the Night Fog Brightness Temperature Difference product, above. The enhancement is such that blue regions show where 10.3 µm brightness temperatures are warmer than brightness temperatures at 3.9 µm. At night, this occurs because cloud water droplets do not emit as much 3.9 µm radiation as a blackbody; the conversion of the sensed radiation to a brightness temperature assumes blackbody emissions have occurred, however, and as a result, the 3.9 µm brightness temperature is cool. Cloud water droplets do emit as a blackbody at 10.3 µm, so the brightness temperature over cloud water droplets is warmer than the it is at 3.9 µm. Blue, then, signifies stratus clouds in this enhancement. Can you tell where the fog is?  Do you think IFR conditions are occurring wherever the Brightness Temperature Difference field is blue?  How about under the extensive cirrus shield, associated with former Hurricane Michael, that is over the southeastern fifth of the image?

The Advanced NIght time Microphysics RGB for the same time, below, uses as one of its inputs (the ‘green channel’ in the RGB) the Night Fog Brightness Temperature Difference (Click here for a toggle between the two to reinforce that statement).

Night Time Microphysics RGB at 1152 UTC on 11 October (Click to enlarge)

GOES-R IFR Probability fields, below, for the same time are useful because Rapid Refresh model data used in IFR Probability include low-level model estimates of saturation.  If saturation does not exist underneath the cloud deck, then IFR Probabilities are suppressed.  Don’t expect fog, then, underneath the strong signal in the Brightness Temperature Difference field over most of the Midwestern United States!  In addition, the Rapid Refresh Model data defines potential fog under the big cirrus shield over the Appalachians.

GOES-R IFR Probability fields at 1152 UTC on 11 October (Click to enlarge)

Click here for the Advanced Microphyics RGB with observations of ceilings and visibility.  Here are the observations over the Brightness Temperature Difference.  Observations plotted on top of IFR Probability are shown below.

GOES-R IFR Probability fields at 1152 UTC on 11 October along with surface observations of ceilings and visibilities (Click to enlarge)

IFR Conditions with cold-season extratropical cyclones

GOES-16 Night Fog Brightness Temperature Difference (10.3 µm – 3.9 µm), Advanced Night Time Microphysics RGB, and GOES-16 IFR Probability fieds at 1142 UTC on 8 October 2018 (Click to enlarge)

Fog and low stratus were widespread on 8 October over the Plains, in particular over Iowa and Minnesota.  What satellite tools exist to highlight such regions of lowered ceilings and reduced visibility?

When IFR conditions — fog and low stratus — occur with extratropical cyclones that generate multiple cloud layers, satellite detection of low clouds is difficult because higher clouds get in the way of the near-surface view. The animation above steps through the Night Fog Brightness Temperature Difference (10.3 µm – 3.9 µm; low clouds in the default enhancement are cyan), the Nighttime Microphysics Red-Green_blue (RGB) composite (low clouds in the RGB are cyan to yellow, depending on the temperature) and the GOES-R IFR Probability field (Probabilities for IFR conditions are highest in orange/red regions) for 1142 UTC on 8 October 2018, when low ceilings were widespread over the Plains and East Coast. Abundant high clouds rendered the Night Fog Brightness Temperature difference product (and, by extenstion, the Night Time Microphysics RGB, because the RGB uses the Night Fog Brightness Temperature Difference as its ‘Green’ Component) ineffective in outlining potential regions of low clouds. In contrast, the IFR Probability field was able to highlight low clouds under the high clouds because it fuses satellite data (ineffective at this time) with Rapid Refresh model estimates of low-level saturation.

There are regions — southern Lake Erie, for example — where the lack of high clouds allows the Brightness Temperature Difference field, and the Nighttime Microphysics RGB to operate with success in identifying low clouds.

The toggle below shows the Night Fog Brightness Temperature Difference and the IFR Probabiity fields over the eastern portion of the country.  Very small-scale (in the horizontal) features, such as river fog, are a challenge for IFR probability because the Rapid Refresh horizontal resolution of 13 km may not resolve river valleys.

GOES-16 Night Fog Brightness Temperature Difference (10.3 µm – 3.9 µm) and GOES-16 IFR Probability fieds at 1142 UTC on 8 October 2018 (Click to enlarge)

River Fog over Wisconsin

GOES-16 IFR Probability, 0747-1152 UTC on 30 August 2018 (Click to enlarge)

GOES-16 IFR Probability incorporates information about low-level saturation from the Rapid Refresh Model. The model data used changes hourly, and that can cause an hourly change, a pulsing, in the IFR Probability animation in cases when the forecast model evolution is changing.

In this case, visible imagery at sunrise (enhanced because of the low light) suggests the IFR Probability field is overpredicting the extent of River Fog in southwest Wisconsin. Note, however, that the highest IFR Probabilities do align with river valleys where fog is observed.

GOES-16 ABI Band 2 (0.64 µm) and GOES-R IFR Probabiilty, 1152 UTC on 30 August 2018 (Click to enlarge)

Dense Fog over the Midwestern United States

GOES-R IFR Probability Fields, 0337 – 1332 UTC on 8 August 2018 (Click to enlarge)

The longer August nights over the upper Midwest (for example, Madison Wisconsin’s night is about 70 minutes longer now than it was at the Summer Solstice) can allow for dense fog to form when light winds and clear skies follow a cloudy, damp day. The animation above shows the evolution of the GOES-R IFR Probability fields as the dense fog develops, a fog for which advisories were issued.  The horizontal extent of the widespread fog is captured well in the IFR Probability fields.  There are a couple of things worth noting.

There is a subtle — but noticeable — change in the IFR Probability Fields each hour in the animation. That change is related to model data in this fused product. Rapid Refresh Model output are used to identify regions of low-level saturation that must occur with fog. (The inclusion of this data helps IFR Probability better distinguish — compared to satellite imagery alone — between elevated stratus and fog). When output from newer model runs is incorporated (and that happens every hour), the IFR probability fields are affected. The amount of change is testimony to whether the sequential runs of the Rapid Refresh are consistent in capturing the developing fog. In this case, there were differences from one model run to the next.

GOES-R IFR Probability fields, 1202 UTC on 8 August 2018 (Click to enlarge)

Two boundaries are apparent in the animation above, captured in the 1202 UTC image above, and in the 1302 UTC image below. These boundaries are related to the Terminator, the dividing line between night and day. During an hour around sunrise, rapid changes in reflected 3.9 µm solar radiation make the detection of low clouds difficult. Temporal adjustments are incorporated into the IFR Probability fields to create a cleaner field. In the 1202 UTC image, above, the effects of sunrise are occurring along the NNW-SSE oriented boundary near the Mississippi River in southwestern WI (The boundary is parallel to the Terminator line, so it will be vertical — parallel to a Longitudinal Line — on the Equinoxes).   In the 1302 UTC image, below, the second boundary is over far northwestern Wisconsin, far southeastern Minnesota, and eastern Iowa.  The region between these two westward-propagating boundaries is where information from previous times is used in the IFR Probability fields.  Thus, when the second boundary passes, you may observe rapid changes in IFR Probability fields.

GOES-R IFR Probability fields, 1302 UTC on 8 August 2018 (Click to enlarge)

The Night Fog Brightness Temperature Difference (10.3 µm – 3.9 µm) can be used to detect stratus, and that field is also a constituent (the ‘green’ part) of the Advanced Nighttime Microphysics RGB. The animation below compares these two products used to detect stratus with IFR Probability at 1002 UTC on 8 August 2018.  The Night Fog Brightness Temperature Difference fails to highlight regions of fog over central WI (and elsewhere), so neither it nor the RGB give a consistent signal over the entire fog-shrouded region.

Night Fog Brightness Temperature Difference (10.3 µm – 3.9 µm), GOES-R IFR Probability, and Advanced Nighttime Microphysics RGB at 1002 UTC on 8 August 2018 (Click to enlarge)

Detecting Fog under a Pall of Smoke in the Pacific Northwest

HRRR forecast of Vertically-Integrated Smoke over the Pacific Northwest (for more information see text), forecast valid at 1200 UTC on 30 July 2018 (Click to enlarge)

When smoke covers a geographic region, visible detection of low-level fog is difficult because smoke can scatter or obscure the signal from the low-level clouds.  The image above shows a Vertically Integrated Smoke Forecast from a High-Resolution Rapid Refresh Model simulation in Real Earth (link). Thick smoke was predicted to occur over the coast of Oregon.

The animation below steps through the GOES-R IFR Probability, and then the ‘Blue Band’ (0.47 µm), the ‘Red Band’) (0.64 µm), the ABI channel with the highest spatial resolution, the ‘Veggie Band’ (0.86 µm) and the ‘Snow/Ice Band’ (1.61 µm). Two things of note: Because of the low sun angle, and the enhanced forward-scattering properties of smoke at low sun angle, it is very hard to detect fog through the smoke in the visible wavelengths near Sunrise. As the wavelength of the observation increases, scattering is less of an issue. In addition, smoke is more transparent to longer wavelength radiation. Thus, the cloud edges become more apparent under the smoke in the 0.86 µm and especially the 1.61 µm imagery compared to the visible.

GOES-16 IFR Probability, and GOES-16 Single Bands (Band 1, 0.47 µm, Band 2, 0.64 µm, Band 3, 0.86 µm and Band 5, 1.61 µm) at 1342 UTC on 30 July 2018, along with surface observations of Ceilings, Visibility, and Visibility Restrictions at 1400 UTC. (Click to enlarge)

GOES-16 IFR Probability can indicate the regions of fog and corresponding restrictions in visibility because it relies on longer wavelength observations from GOES-16 (3.9 µm and 10.3 µm, principally) and information about low-level saturation from the Rapid Refresh model.  Smoke is mostly transparent to radiation in the infrared unless it becomes extraordinarily thick  (indeed, that is one reason why smoke is difficult to detect at night);  thus, the brightness temperature difference between the shortwave (3.9 µm)  and longwave (10.3 µm) infrared channels on GOES-16’s ABI can highlight cloud tops made up of water droplets that occur underneath elevated smoke.

As the Sun gets higher in the sky, fog edges beneath the smoke become more apparent because forward scattering decreases.  The animation below of the visible confirms this. In contrast, the fog edge in the IFR Probability is well-represented during the entire animation (although the horizontal resolution of the infrared channels on GOES-16 (at the sub-satellite point) is only two kilometers vs. 1/2-km for the Red Visible).

Note that IFR conditions are also occurring during this animation due to smoke over southwest Oregon.  GOES-R IFR probability detects only cloud-forced IFR conditions, but not IFR conditions because of thick smoke only.  Again, this is because smoke detection in the infrared is a challenge for ABI Channels and Rapid Refresh Model output does not as yet predict visibility restrictions due to smoke.

GOES-16 ABI ‘Red Visible’ Imagery (0.64 µm), 1342 – 1557 UTC on 30 July 2018, along with surface observations of Ceilings, Visibility, and Visibility Restrictions (Click to enlarge)

GOES-16 IFR Probability, 1342 – 1557 UTC on 30 July 2018, along with surface observations of Ceilings, Visibility, and Visibility Restrictions (Click to enlarge)

IFR Probability indicating a Cold Front

GOES-16 IFR Probability Fields, 0607-1422 UTC on 1 June 2018 (Click to enlarge)

The animation above shows increasing amounts of IFR Probability moving southward and westward over eastern Wisconsin as a cold front with Lake-influenced air pushed inward. Ceilings lowered and visibilities reduced as the front passed, and the IFR probability field’s motion could be used to predict when the temperature change occurred. High clouds did not impede the satellite view of this event, so the brightness temperature difference field, below, also showed the stratus deck as it moved inland. Missing from the satellite-only view of this event, of course, is the information on whether the cloud is extending to the surface. Furthermore, the cloud signal vanishes from the brightness temperature difference product at sunrise, as it flips sign from positive at night to negative during the day as amounts of reflected solar 3.9 µm radiation increase.

GOES-16 Night Fog Brightness Temperature Difference (10.3 µm – 3.9 µm), 0607-1422 UTC on 1 June 2018 (Click to enlarge)

A Meteorogram for Milwaukee for the 24 hours ending at 1500 UTC on 1 June shows the dramatic change in ceilings, temperature and wind between 10 and 11 UTC on 1 June. A similar meteorogram for Madison (here), shows a less dramatic change at 1400 UTC on 1 June.

Meteorogram for KMKE (Mitchell International Airport, Milwaukee, WI) from 1500 UTC 31 May through 1500 UTC 1 June. Note the abrupt change in surface conditions at 1100 UTC on 1 June as the front moved through (Click to enlarge)

An animation of Low IFR Probabilities is shown below.

GOES-16 Low IFR Probability Fields, 0607-1422 UTC on 1 June 2018 (Click to enlarge)

Fog over Kansas

GOES-16 IFR Probability fields, 0557 – 1157 UTC on 21 May 2018 (Click to animate)

Dense Fog Advisories were issued over Kansas by the Goodland KS Forecast Office on Monday 21 May 2018. The IFR Probability field, above, computed using GOES-16 and Rapid Refresh Model output shows the development of fog as it developed westward across Kansas. (Surface winds in the region were light (source); Plot 2 (from here)).

High clouds around Kansas impeded the detection of low stratus/fog from satellite in some regions.  Those regions benefit from the fused data methods of IFR Probability:  where satellite data alone cannot be used, model data can give important information.  Consider the ‘Night Fog’ Brightness Temperature Difference field, below (10.3 µm – 3.9 µm) for the same period as shown above.  High clouds are apparent over southeast Kansas and Missouri, and also occasionally over Nebraska.  High clouds over McCook, Nebraska, just north of the Kansas-Nebraska state line in southwest Nebraska, prevent the satellite detection of fog even as the ceilings and visibilities decrease. (Here is a toggle of the three products at 1002 UTC).

Similarly, IFR or near IFR conditions develop over southwestern Missouri, but a cirrus shield there means that Brightness Temperature Difference (10.3 µm – 3.9 µm) fields, and the Nighttime Microphysics RGB (that relies on the Brightness Temperature Difference field) cannot observe the low clouds.

GOES-16 ‘Night Fog’ Brightness Temperature Difference (10.3 µm – 3.9 µm), 0557 – 1157 UTC on 21 May 2018 (Click to animate)

GOES-16 Nighttime Microphysics RGB, 0557 – 1157 UTC on 21 May 2018 (Click to animate)

When you use a product to ascertain the presence of fog/low stratus, be certain that you understand its limitations.