MODIS and GOES IFR Probabilities in the Pacific Northwest

IFR Probabilities (Left) from GOES-15 Imager (Upper Left) and MODIS (Lower Left), Brightness Temperature Difference (11 µm – 3.9 µm) (Right) from GOES-15 Imager (Upper Right) and MODIS (Lower Right). Terra data used at 0614 UTC, Aqua data used at 1027 UTC. All times as indicated. (Click to enlarge)

The Pacific Northwest is far from the sub-satellite point of GOES-West. Pixel size there is therefore greater than the nominal 4-km size at the sub-satellite point: Pixel size is more like 8 kilometers in the north-south by 5 kilometers in the east-west. The animation above shows GOES- and MODIS-based IFR Probabilities and Brightness Temperature Difference Products.

Both GOES and MODIS IFR Probabilities show an expansion (as observed) of reduced ceilings and visibilities as marine stratus penetrates inland over coastal Washington and Oregon. The visibility at Seattle drops as the high probabilities overspread the region. MODIS resolution is better able to depict the fingers of fog/stratus that penetrate up river valleys along the coast. GOES Temporal resolution, however, means that frequent updates are available. (Note the lack of MODIS data at 1300 UTC).

The brightness temperature difference fields from GOES and MODIS are different. The GOES field has a considerable false signal (as far as fog is concerned) related to changes in surface emissivity that occur in the arid intermountain west during summer. MODIS fields show less of this false signal because of differences in the spectral width of the channels.

The Day/Night band on Suomi NPP, below, also shows the extent of stratus over coastal Oregon and Washington. This visible image uses reflected lunar light for illumination.

Suomi NPP VIIRS Day Night Band, 0936 UTC, 15 July 2014

Dense Fog Advisories over Missouri

GOES-R IFR Probabilities (Upper Left), GOES-East Brightness Temperature Difference (10.7 µm – 3.9 µm) (Upper Right), Suomi/NPP Day/Night Band imagery (Lower Right), MODIS-based IFR Probabilitiy (Lower Left), times as indicated (Click to animate)

Moisture from departing late-day thunderstorms allowed for the development of dense fog over central Missouri overnight. The GOES-based IFR Probabilities, above, capture the low ceilings and reduced visibilities that developed. The traditional method of fog detection, the brightness temperature difference (BTD) between the 10.7 µm and 3.9 µm fields, was hampered by mid- and high-level clouds associated with the departing convection.

Polar-orbiting satellites such as Terra, Aqua and Suomi NPP can give high-resolution views of developing fog. In the present case, Terra overflew the region near 0400 UTC. The image below shows enhanced MODIS-based IFR Probabilities confined to central Missouri. An Aqua overpass at ~0800 UTC similarly gave a high spatial resolution view of the area. Of course, Terra and Aqua and Suomi NPP only give occasionaly snapshots. To see the ongoing development, temporal resolution as from GOES is key. But the polar orbiters can give an early alert if developing fog starts out at small scales that might be sub-pixel scale in GOES.

As above, but at 0400 UTC 14 July 2014 (Click to enlarge)

From the National Weather Service in St. Louis:

URGENT – WEATHER MESSAGE
NATIONAL WEATHER SERVICE ST LOUIS MO
527 AM CDT MON JUL 14 2014

MOZ041-047>051-059-141400-
/O.NEW.KLSX.FG.Y.0002.140714T1027Z-140714T1400Z/
BOONE MO-CALLAWAY MO-COLE MO-GASCONADE MO-MONITEAU MO-
MONTGOMERY MO-OSAGE MO-
INCLUDING THE CITIES OF…COLUMBIA…JEFFERSON CITY
527 AM CDT MON JUL 14 2014

…DENSE FOG ADVISORY IN EFFECT UNTIL 9 AM CDT THIS MORNING…

THE NATIONAL WEATHER SERVICE IN ST LOUIS HAS ISSUED A DENSE FOG
ADVISORY…WHICH IS IN EFFECT UNTIL 9 AM CDT THIS MORNING.

* TIMING…DENSE FOG HAS DEVELOPED AND WILL CONTINUE THROUGH 900
AM.

* VISIBILITIES…ONE QUARTER MILE OR LESS AT TIMES.

* IMPACTS…SIGNIFICANTLY REDUCED VISIBILITIES WILL LEAD TO
HAZARDOUS DRIVING CONDITIONS.

PRECAUTIONARY/PREPAREDNESS ACTIONS…

A DENSE FOG ADVISORY IS ISSUED WHEN DENSE FOG WILL SUBSTANTIALLY
REDUCE VISIBILITIES…TO ONE-QUARTER MILE OR LESS…RESULTING IN
HAZARDOUS DRIVING CONDITIONS IN SOME AREAS. MOTORISTS ARE ADVISED
TO USE CAUTION AND SLOW DOWN…AS OBJECTS ON AND NEAR ROADWAYS
WILL BE SEEN ONLY AT CLOSE RANGE.

The aviation portion of the AFD from St. Louis mentioned the probability of fog at 0800 UTC; the 1129 UTC update discussed the fog that was present over central Missouri:

.AVIATION: (For the 12z TAFs through 12z Tuesday Morning)
Issued at 609 AM CDT Mon Jul 14 2014

The first concern for this TAF package is that for low ceilings
and fog that have developed in the wake of precipitation that
exited the area overnight. In central MO and including KCOU, fairly
widespread dense fog has reduced visibilities to under 1/4SM for
much of the night. Further east and including metro area TAF sites,
trends indicate the potential for IFR cigs and MVFR visibility for
the first couple hours of the period. However, all sites affected
by fog should see an improvement through the morning hours as
ceilings lift and fog burns off. The second concern is that of a
second cold front, poised to move through the area today.
Currently, the cold front extends from roughly KDBQ southwestward
along the Missouri/Illinois border and just south of KAFK. While
showers may develop along the front as it moves through KUIN
during the late morning/early afternoon, greater instability
exists further south and east. Have currently continued VCSH
mention at KUIN and KCOU, and VCTS for metro TAF sites this
afternoon as the cold front moves through. Uncertainties regarding
coverage and exact timing preclude any TEMPO groups at this time.
The front should be south of all area TAF sites by 21Z, at which
time winds will have veered to the northwest and increased to
around 10-14KT. Winds will remain northwesterly through the end of
the period in the wake of the front, and while a mostly VFR
forecast is expected, reductions in ceilings/visibility may occur
with any storms that move over the terminals.

Distinguishing between stratus and fog over Pennsylvania

Suomi NPP 11.35 µm Infrared Imagery at 0609 and 0752 UTC, 10 July 2014 (Click to enlarge)

The orbital geometry of Suomi NPP allowed two high-resolution images of Pennsyvlania early in the morning of the 10th of July 2014. Can you tell from the imagery above if there is fog/stratus in the river valleys of Pennsyvlania? Are the relatively cool clouds from Pittsburgh northeastward towards Elmira, NY obstructing visibilities? Based on the IR (11.35 µm for Suomi NPP) imagery alone, above, that is a difficult question to answer. Historically, the brightness temperature difference between the longwave IR (11.35 µm) and the shortwave IR (3.74 µm) has been used to indentify water-based clouds. Imagery from Suomi NPP, below, highlights where water-based clouds (like stratus) exist. If the clouds are the same temperature as the surrounding land (likely the case for river fog), a single 11.35-µm image is of little help in identifying the clouds.

Suomi NPP 11.35 – 3.74 µm Brightness Temperature Difference at 0609 and 0752 UTC, 10 July 2014 (Click to enlarge)

The Day-Night band can also highlight where clouds exist, because lunar illumination reflects well off clouds. A 3/4-full moon ably illuminates the scene at 0609 UTC, but that moon has set at 0747 UTC and the Clouds are harder to see.

Suomi NPP Day/Night Band at 0609 and 0752 UTC, 10 July 2014 (Click to enlarge)

Both the Day/Night band and the Brightness Temperature Difference Fields (and any Infrared image) gives information about the top of the cloud. Fog existence is difficult to discern only from satellite data because the bottom of the cloud is not sampled. This is why a fused product (such as IFR Probability) that includes surface information (in the case of IFR Probability from the Rapid Refresh Model) is desirable. MODIS data can be used to compute IFR Probability, and a MODIS-carrying Aqua pass occurred in between the two Suomi NPP Passes shown above.

MODIS 11 – 3.9 µm Brightness Temperature Difference and IFR Probability at 0652 UTC, 10 July 2014 (Click to enlarge)

In the two images above, note how the IFR Probability Fields de-emphasize the low cloud areas that stretch northeastward from Pittsburgh towards Elmira. This is likely mid-level stratus. River Fog over northeast Pennsylvania is highlighted in the IFR Probability fields (and in the brightness temperature difference field). This image, which shows the GOES-based IFR Probability field at 0645 UTC, highlights the power of MODIS’ superior spatial resolution in the early detection of small-scale fog. The large region of reduced visibility around Elmira NY (meager surface observations suggest this large region of fog verified) appears in both MODIS- and GOES-based IFR Probability fields. Only the MODIS-based IFR Probability field, however, has a distinct river-valley signal over northeast Pennsylvania.

MODIS and GOES IFR Probability both suggest IFR conditions may be occurring over the Atlantic Ocean. The brightness temperature difference field shows no low cloud signal there because of a cirrus shield. IFR Probability gives a signal of fog here based on information from the Rapid Refresh.

MODIS-based IFR Probability over California

MODIS-based IFR Probabilities over the western US, 0932 UTC on 30 June 2014 (Click to enlarge)

MODIS data, although infrequent, can give a high-resolution estimate of whether of not fog/low clouds are forming in a region. This is particularly useful for cases with highly variable terrain (such as deep river valleys). In the image above, IFR Probabilities are very low over both the Central Valley of California and over the Salinas Valley closer to the coast. Fog/low Stratus are unlikely to be occurring.

Suomi/NPP Brightness Temperature Difference Fields (11.35µm – 3.74µm) at 0916 and 1057 UTC on 30 June 2014 (Click to enlarge)

The orbital geometry of Suomi/NPP on June 30th was such that it provided two close-up views of these two valleys around/after the time of the MODIS pass shown above. The animation above toggles between those two times. The brightness temperature difference field shows a general increase in return signal strength. But the MODIS IFR Probability field, top, and GOES-West IFR Probability fields (not shown) support the view that any clouds present are not having an impact on ceilings and visibility. A toggle that includes ceilings and visibilities is here.

Note that returns that suggest low clouds over Nevada are more likely due to emissivity differences in different soils, an effect that is more obvious in very dry conditions.

The Suomi/NPP Day/Night band can also provide imagery to help identify the tops of clouds. However, Day/Night Band imagery uses reflected moonlight. There was a new moon on June 27th, and that almost-new moon had set by the time of the images below. Thus, only Earthglow is illuminating any clouds that are present, and that feeble light is mostly overwhelmed by emitted city lights. It is therefore very difficult to identify any changes in cloudcover from the Day/Night band on this date.

Suomi/NPP Day/Night Band imagery at 0916 and 1057 UTC on 30 June 2014 (Click to enlarge)

Fog over Chicago

GOES-R IFR Probability and GOES-13 Brightness Temperature Difference, 0915 UTC on 27 June (Click to enlarge)

A cold winter and cool Spring have caused Lake Michigan to be much cooler than normal (Linked-to Figure is on this page). Colder-than-normal lake temperatures and warm summer dewpoints are a recipe for fog, and that fog has persisted over and near Lake Michigan this month. (Click here for a video of fog moving over Chicago on 26 June). The imagery above shows the GOES-R IFR Probability toggling with the GOES-13 Brightness Temperature Difference field (10.7µm – 3.9µm) at 0915 UTC. Abundant high clouds (cirrus from convection over the Plains) makes the brightness temperature difference method of detecting low clouds problematic. Because IFR Probability computation includes surface information, however, a useful signal near Lake Michigan that captures the extent of the fog is produced.

When high clouds prevent satellite predictors from being used, the IFR Probability field is typically fairly smooth. That is the case over most of Lake Michigan. Breaks in the high clouds at 0915 UTC over Chicago allow for satellite predictors to be used. Where that happens, IFR probabilities are larger, and the IFR Probability field is more pixelated.

Click here for a blog entry at the Washington Post on the fog.

Stratus and low Stratus over North Dakota

GOES-based GOES-R IFR Probabilities (Upper Left), GOES-13 Brightness Temperature Difference (10.7µm – 3.9µm) (Upper Right), GOES-R Cloud Thickness (Lower Left), Suomi/NPP Brightness Temperature Difference (11.35µm – 3.74µm) (Lower Right), all times as indicated (Click to enlarge)

A strength of the GOES-R IFR Probability field is that it highlights regions where IFR conditions are occurring and downplays regions where stratus is elevated off the surface, insignificant from an aviation point of view. In the image above, note how stations in the Missouri River valley of central North Dakota (Bismarck, Mercer County) are under a thick stratus deck that is highlighted in the brightness temperature difference fields, but surface ceilings and visibilities are good. IFR Probabilities there are small because the Rapid Refresh Model data does not suggest low-level saturation. Dickinson and Hettinger, over southwestern North Dakota, in contrast, show restricted visibilities (that worsen before sunrise, see below) in a region of higher IFR Probabilities. IFR Probabilities remain low over central North Dakota where visibilities are not restricted.

GOES-based GOES-R IFR Probabilities (Upper Left), GOES-13 Brightness Temperature Difference (10.7 – 3.9) (Upper Right), GOES-R Cloud Thickness (Lower Left), Suomi/NPP Brightness Temperature Difference (11.35 – 3.74) (Lower Right), all times as indicated (Click to enlarge)

Fog in the Valleys of Pennsylvania and New York

Suomi/NPP Day/Night Band and Brightness Temperature Difference (11.35 – 3.74) at 0658 UTC on 16 June 2014 (Click to enlarge)

Suomi/NPP Day/Night band “Visible Imagery at Night” from last night at 3 AM EDT over Pennsylvania and surrounding states shows cloud formation in the river valleys of Pennsylvania and New York. The brightness temperature difference also shows these cloud formations, but note over eastern New York how high cirrus prevents the detection of low clouds in river valleys using the brightness temperature difference field, but the cirrus is thin enough that the Day/Night band does show the low clouds. The brightness temperature difference field can show where clouds are present in regions where city lights in the day/night band might appear similar to clouds on a night when lunar illumination is strong (as was the case on 16 June 2014) — for example, along US Highway 220 from Lock Haven to Jersey Shore and Williamsport in southern Clinton and Lycoming Counties.

Of course, the Suomi/NPP satellite is seeing the top of the cloud, so it can be difficult to infer ceiling and visibility obstructions from these data. The GOES-based IFR Probability field from about the same time, below, shows hints of visibility restrictions, but the coarser resolution of GOES-13 (compared to Suomi/NPP) limits the ability of GOES to herald the development of fog. In addition, the 13-km resolution of the Rapid Refresh model data that are also used in the GOES-R IFR Probability fields is insufficient to resolve the small river valleys (such as Pine Creek that shows up very well in the S/NPP imagery in western Lycoming County). Portions of the Susquehanna River basins do show marginally enhanced probabilities, certainly something that would alert a forecaster to the possibilities of fog.

GOES-based IFR Probabilities, 0700 UTC on 16 June 2014 (Click to enlarge)

MODIS data can also be used to produce IFR Probabilities at infrequent intervals, but a timely overpass at 0744 UTC shows high probabilities in most of the river valleys of central Pennsylvania and upstate New York, with the highest probabilities near Elmira, NY.

MODIS-based IFR Probabilities, 0744 UTC on 16 June 2014 (Click to enlarge)

The GOES-R IFR Probabilities at 1000 UTC, below, show evidence that the fog/low stratus have become widespread enough in river valleys to be detected even by GOES. Elmira, NY, is reporting IFR conditions, and such conditions are also likely elsewhere in the valleys (although no observations are available to confirm that).

GOES-based IFR Probabilities, 1000 UTC on 16 June 2014 (Click to enlarge)

Use timely polar-orbiting satellite data — with high resolution — to confirm suspicions of developing fog in river valleys. Then monitor the situation with the good temporal resolution of GOES. During the GOES-R era, geostationary satellite spatial resolution will be increased and fog detection from GOES-R should occur with better lead time.

Colorbars

GOES-R IFR Probabilities (Upper Left), GOES-R Low IFR Probabilities (Upper Right), GOES-R MVFR Probabilities (Lower Left) and GOES-R Cloud Thickness (Lower Right), time as indicated (Click to enlarge)

The default colorbars for GOES-R IFR (Instrument Flight Rules) Probabilities, Low IFR Probabilities and MVFR (Mostly Visible Flight Rules) Probabilities change colors at different break points, as shown above (by where the arrows are, for example). IFR Probabilities switch from white to orange/yellow at around 40%, vs. 30% for LIFR and 55% for MVFR. A statistical analysis of the surface observations was used in the development of the colorbars. Highest skill at detecting the category was assigned the reddest colors. One could therefore infer from this that there is in general less skill in detecting MVFR conditions than LIFR conditions. For a given event, however, as shown above, MVFR probabilities will in general be larger than IFR or LIFR probabilities.

Let the color of the colorbar guide the interpretation. Black and white values mean the category (IFR conditions, for example) is unlikely, yellow means it’s possible, and red means it is highly likely. MVFR probabilities have higher thresholds than LIFR probabilities because of skill differences in predicting the different visibilities associated with the sky conditions.

Terrain and IFR Probabilities

GOES-R IFR Probabilities computed from GOES-West at 0930 UTC on 28 May, and Color-enhanced terrain (Click image to enlarge)

When IFR Probabilities are enhanced over high terrain, how confident can you be that IFR conditions are occurring? Surface observations are rare on mountain tops. It’s possible that clouds occurring are at one level as the terrain rises up into the clouds, and ceilings at adjacent stations can give an indication of the cloud base (in the present case, ceilings are about 9000 feet above sea level at Seattle, for example).

Suomi/NPP Day/Night band imagery can verify that clouds exist in the region where IFR Probabilities are elevated. The toggle below, of Day/Night band and Brightness Temperature Differences, shows compelling evidence (even in low light conditions) of clouds along the spine of the mountains in central Washington and Oregon.

Suomi/NPP VIIRS Day/Night Band and Brightness Temperature Difference (11.45 µm – 3.74 µm), 0935 UTC on 28 May (Click to enlarge)

High-resolution MODIS data are also used to produce IFR Probabilities, and they can be used to deduce the presence of low ceilings/reduced visibilities as well. The toggle below, from 1027 UTC, shows the brightness temperature difference field (11.0 – 3.9) from MODIS and the IFR Probability field. It is likely in this case that high clouds were shrouding the higher peaks of the Cascade Mountains.

Benefits of Resolution with a Polar-orbiting satellite

Toggle between GOES-13 Brightness Temperature Difference (10.7µm – 3.9µm) and GOES-R IFR Probability over Kentucky and surrounding states. Surface observations of ceiling heights and visibility are included, 0630 UTC 16 May 2014 (Click to enlarge)

The toggle above highlights a strength of the GOES-R IFR Probability fields compared to the GOES Brightness Temperature Difference when it comes to detecting low fog/stratus. The Brightness Temperature Difference field only sees the top of the cloud. In the toggle above, the region of elevated stratus, the stratus over western Virginia and western West Virginia is highlighted, but those clouds are unimportant for aviation/transportation, and IFR Probability fields ignore that region (save for the spine of the Appalachians where the mountains are rising up into the clouds, so ceilings are near the surface).

There are heightened IFR Probabilities in/around KEKQ (Monticello, Kentucky) at 0630 UTC: what is the character of that fog? It’s difficult to tell with the coarse GOES resolution (although someone familiar with the topography of eastern Kentucky might guess).

The imagery below toggle between the high-resolution (1-km) Suomi/NPP VIIRS Brightness Temperature Difference (11µm – 3.74µm) and the Day/Night Band at 0638 UTC. The higher resolution imagery allows the dendritic nature of the valley fog to appear in a way that is impossible with the coarser-resolution GOES data. Fog is initially developing in river valleys. Both the Brightness Temperature Difference and Day/Night imagery, however, are seeing only the top of the cloud and are not giving information about the likelihood of fog. But the cloud structure would alert a forecaster to the probability of developing fog (as does the time trend in the GOES-R IFR Probability fields).

Note how the cirrus shield east of the Appalachians shows up distinctly in both GOES and VIIRS brightness temperature difference fields. High clouds such as those prevent the satellite detection of fog/stratus at low levels. In those cases, only the IFR Probability field has a chance to detect fog if it is present.

Suomi/NPP Brightness Temperature Difference (11.0µm – 3.74µm) and Day/Night Band from VIIRS, 0638 UTC 16 MAy 2014 (Click to enlarge)

By 1000 UTC, the fog that was initially confined to river valleys over central Kentucky has expanded. In this case, Suomi/NPP data (or the trending of the GOES data) gives a forecaster a heads up on the development of overnight fog.

GOES-R IFR Probability over Kentucky and surrounding states. Surface observations of ceiling heights and visibility are included, 1000 UTC 16 May 2014 (Click to enlarge)

GOES-R Fog Product Examples

Fog detection fusing GOES, Terra/Aqua or Suomi/NPP Satellite data with Model output