Category: Synthetic NSSL WRF-ARW Imagery

This blog entry will look at an example of the synthetic fog product (from the 4-km NSSL WRF-ARW model) for an event that took place during the overnight hours of November 19 to 20, 2011.

Here is the synthetic fog product during the overnight hours:

http://rammb.cira.colostate.edu/templates/loop_directory.asp?data_folder=training/visit/loops/19nov11_syn_fog

There is southerly flow advecting moisture from the Gulf of Mexico northward.  The increasing moisture, in combination with nighttime cooling leads to the development of a large area of stratus clouds across the Plains from Texas to Nebraska and eastward across the southeast states.  This enhancement will depict regions of low-level clouds in the blue colors.  Note that the imagery from 0000-08000 UTC is based on the 0000 UTC November 18 model run, and the imagery between 0900-1200 UTC is based on the 0000 UTC November 19 model run.  There is a discontinuity between 0800 and 0900 UTC for this reason, but the trend in the low-level clouds as discussed earlier is still valid.

Let’s assess how well the model forecast was.  We’ll look at the GOES Low Cloud / Fog product (also known as the shortwave albedo product):

http://rammb.cira.colostate.edu/templates/loop_directory.asp?data_folder=training/visit/loops/19nov11_sw_albedo

In this color table, low-level clouds / fog are depicted by the lighter / white colors.  High-level clouds are darker / black and color enhanced above a temperature threshold.  The model did a pretty good job with the development of low-level stratus across the Plains and into the southeast.  Fog can be assessed by looking at surface observations (not shown).

The key to remember here is that this product aids in visualization.  Model fields of relative humidity could have yielded the same conclusions, however by looking at a forecast model field of something you’re already familiar with for diagnosing stratus / fog (i.e., the GOES fog product), it offers a different perspective on assessing the possibility of low-level stratus / fog during the forecast period of interest.

Upon inspection of the synthetic infrared (10.35 micron) imagery from the NSSL WRF-ARW model:

http://rammb.cira.colostate.edu/templates/loop_directory.asp?data_folder=training/visit/loops/13oct11_syn_ir

your attention may be drawn to the region of southwest Colorado since we see a region of cold brightness temperatures  that does not move and persists for the duration of the loop (1600-0000 UTC).  Could these be low clouds / fog? Let’s look at the GOES visible imagery for 1630 UTC:

Skies are clear across southwest Colorado, so we are not looking at low clouds / fog.  However, note that there is snow cover over the mountains.

What we’re looking at in the synthetic imagery is a representation of  snow cover.  The snow cover data that goes into the NSSL WRF-ARW model is relatively coarse, therefore the cold ground signature is spread out over relatively large areas when there is snow cover in the model.  This signature shows up quite easily early in the cold season with the large temperature difference between snow cover and no snow cover, and will increase in coverage moving into winter as snowfall coverage increases.

Let’s analyze the following loop of the synthetic fog product, generated from the 4-km NSSL WRF-ARW model:

http://rammb.cira.colostate.edu/templates/loop_directory.asp?data_folder=training/visit/loops/10oct11_syn_fog/

In this color table, grey into light blue represents increasing confidence in liquid water clouds.  Our example from the 0000 UTC 10 October 2011 NSSL WRF-ARW model run shows a large area of liquid water clouds (most likely stratus) across Texas extending northward through the central US.  The darker shades of grey and black correspond to ice clouds (most likely cirrus) forecast by the model.

The region of blue in Arizona and Utah extending southward into northwest Mexico that does not move catches your attention.  A quick look at the visible satellite imagery:

http://rammb.cira.colostate.edu/templates/loop_directory.asp?data_folder=training/visit/loops/10oct11_goes_vis

shows that the region is mostly cloud free.  However, the lighter shade of grey extending northwest to southeast through the four corners of AZ/CO/UT/NM corresponds to the blue in the synthetic fog product; this false cloud signature is a consequence of surface emissivities at the two channels.  This is not a model error, rather, an observed feature in GOES-11.  Similarly, the other blue region in the northeast quadrant of AZ and southwest AZ into northwest Mexico are also consequences of the surface emissivity.

The easiest way to identify a false signature is to look at the loop, and the areas that don’t move at all throughout the duration of the loop are likely false emissivity signatures.

Synthetic satellite imagery can be useful in forecasting temperature.  This example from September 20-21, 2011 demonstrates the utility of synthetic imagery from the 4-km NSSL WRF-ARW model in forecasting the overnight low temperature.

Focusing on southeast Wyoming, examine the synthetic infrared imagery from late afternoon (2000 UTC) through the late night hours (0800 UTC):

http://rammb.cira.colostate.edu/templates/loop_directory.asp?data_folder=training/visit/loops/20sept11_syn&image_width=800&image_height=600

Early in the loop, there is afternoon convective clouds in southeast Wyoming, which diminishes by 0300 UTC.  However, notice the region of clouds (indicated by the colder brightness temperatures) developing across southeast Wyoming in the 0300-0800 UTC time range.  If the forecast is correct, the cloud cover would keep temperatures from cooling down as quickly.

Now let’s analyze what actually happened by looking at the GOES IR imagery:

http://rammb.cira.colostate.edu/templates/loop_directory.asp?data_folder=training/visit/loops/20sept11_goes&image_width=800&image_height=600

Note the clouds developing into southeast Wyoming in the 0500-0745 UTC time range, similar to what was forecast by the WRF-ARW model.  This caused locations under the cloud cover to not cool off as quickly, for example, look at the temperature trace (red line) for Cheyenne, WY:

Temperatures at Cheyenne cooled off gradually to about 1000 UTC, then the cloud cover dissipated allowing temperatures to cool more rapidly to an overnight low of 36 F.

Similarly, at locations further east in the southern Nebraska panhandle, cloud cover kept Kimball, NE from cooling off too rapidly for an overnight low of  42:

And an overnight low of 40 at Sidney, NE:

Meanwhile, at Douglas (about 150 miles north of Cheyenne) in east central Wyoming:

Temperatures cooled off more rapidly as skies remained clear, and the overnight low was 25 F.

Real-time synthetic imagery from the 4-km NSSL WRF-ARW model may be viewed here:

http://rammb.cira.colostate.edu/ramsdis/online/goes-r_proving_ground.asp#Synthetic_GOES-R_Imagery_from_Real-Time_NSSL_4_km_WRF-ARW

Contributor:  Becca Mazur, NWS forecast office, Cheyenne, WY

A new synthetic difference product is now being produced from the NSSL WRF-ARW output: the fog product, or 10.35 – 3.9 μm.  In order to generate synthetic 3.9 μm imagery in a timely manner, it’s necessary to assume that it’s always night because the solar reflected calculations are too expensive.  This allows the traditional fog product to be displayed for all forecast hours of the WRF.  In the example above, the color scheme for the synthetic fog product (left) is designed such that grey into light blue represents increasing confidence in liquid water clouds.  Note the forecast stratus clouds in the San Francisco Bay area, and compare that to the observed GOES-11 visible imagery on the right.  The WRF-generated fog product forecast does a good job with the location and dissipation time of the low clouds.  Such a forecast might be quite useful for the aviation industry given how often low clouds and fog cause delays at the West Coast airports.

Synthetic imagery from the 4-km NSSL-WRF ARW model is produced at CIRA and is available in real-time here:

http://rammb.cira.colostate.edu/ramsdis/online/goes-r_proving_ground.asp#Synthetic%20GOES-R%20Imagery%20from%20Real-Time%20NSSL%204%20km%20WRF-ARW

The following loop shows Hurricane Irene at the time it was east of Florida:

http://rammb.cira.colostate.edu/templates/loop_directory.asp?data_folder=dev/lindsey/loops/nssl_wrf_g13_irene&image_width=1024&image_height=384

The right side shows the synthetic infrared imagery from the 0000 UTC 25 August model run, and the corresponding GOES infrared imagery is on the left.  Keep in mind some of the known biases with the model such as brightness temperatures typically warmer than GOES, and anvil cirrus from convection that is typically underdone.  Synthetic imagery can be compared with GOES imagery to look for forecast trends in the Hurricane.

For this blog entry, we’ll consider applications of the NSSL 4-km WRF-ARW model synthetic imagery towards a severe weather event that occurred on June 22, 2010.  The synthetic imagery is one of the GOES-R proving ground real-time products.   Synthetic imagery is model output that is displayed as though it is satellite imagery.  Analyzing synthetic imagery has an advantage over model output fields in that the feature of interest appears similar to the way it would appear in satellite imagery.   The primary motivation for looking at synthetic imagery is that you can see many processes in an integrated way compared with looking at numerous model fields and integrating them mentally.

Figure 1 shows the WRF-ARW synthetic imagery for the 6.95 um (water vapor) band.  The forecast times are indicated at the bottom middle portion of the image, they are from 1200 UTC to 0300 UTC so we are looking at the 12 to 27 hour forecast from the 0000 June 22, 2010 model run.  The model shows an upper-level low over Montana and Wyoming moving eastward.  South of this feature, we can see a region of relatively fast moving warmer brightness temperatures (the red colors moving from Arizona and Utah towards Colorado).  This appears to be associated with an upper-level jet streak.  Another example of a region of warmer brightness temperatures would be across Michigan moving towards Ohio, Pennsylvania and New York.  With both features in the east and the west, there appears to be convection developing during the late afternoon hours.  Remember, we’re looking at mid to upper level features in the water vapor imagery.  The main role of the synthetic water vapor imagery is identifying shortwaves and jet streaks that may play a role in the initiation, maintenance and intensity of convection.  It’s important to understand what you’re looking at in the water vapor imagery when you see a region of warmer brightness temperatures, we’ll discuss this more in future blog entries and in VISIT training sessions that address this topic.  Next, let’s look at lower levels, so we turn to the synthetic IR imagery.

Figure 2 shows the WRF-ARW synthetic imagery for the 10.35 um IR band.  We’re looking at the same time period as the water vapor loop we just looked at.  The advantage to this channel is that low-level features will show up.  This is useful when analyzing cloud cover, to see if clouds will dissipate and allow for sufficient insolation to warm up the surface.  At 1400 UTC we can see low-level clouds showing up as the colder brightness temperatures across western Nebraska and northeast Colorado, these are forecast to dissipate by afternoon hours, however note the higher level clouds forecast across eastern Colorado.  A morning MCS exists across eastern Nebraska moving eastward towards Iowa.   We see the early afternoon convection just ahead of the upper low in Wyoming and Montana by 2000 UTC, while isolated storms develop in eastern Colorado and the Nebraska panhandle shortly thereafter, in the region of strong southwest flow aloft.  Additional convection develops further south in Texas later.  Upscale growth occurs during the late afternoon and evening hours, particularly over Nebraska where there is stronger flow aloft then further south in Texas.  It appears to be an MCS over Nebraska by 0300 UTC.  The best use of the synthetic imagery is to look at the forecast in the morning hours, follow the trends in GOES and other observational data during the day to gauge how much confidence you should have in the model forecast.

Figure 3 shows the GOES 10.7 um IR band over the same time period as the forecast imagery we just looked at.  Notice the low-level clouds in western Nebraska and northeast Colorado dissipated, as was forecast.  The high level clouds in Colorado were well forecast, and looked to be covering a greater area than forecast.  The early thunderstorm activity in Wyoming near the upper low is forecast well.  Notice the later storms in western Nebraska and Kansas, they have much large anvil cirrus canopies than forecast by the synthetic imagery, this is a known bias in the model so that storms in the model will typically appear smaller than observed in GOES.  Upscale growth into an MCS late in the loop in Nebraska seems to be handled well also, and keep in mind that the anvil cirrus canopy will always appear larger in GOES than in the synthetic imagery.

Figure 4 shows the GOES 6.5 um water vapor imagery over the same time period.  The brightness temperatures will generally appear warmer in the synthetic imagery compared to GOES imagery.  The main role of the synthetic water vapor imagery is identifying shortwaves and jet streaks that may play a role in the initiation, maintenance and intensity of convection.  Keep this in mind as you examine the synthetic water vapor imagery, then look at GOES visible imagery and surface observations to see where the key low-level convergence boundaries exist.

The synthetic imagery has exciting potential as an additional tool in forecasting severe thunderstorms, just keep in mind we are looking at model output with its familiar limitations.

For more information on severe weather applications of the synthetic imagery from the NSSL 4-km WRF-ARW model, you may take to this VISIT training session:

http://rammb.cira.colostate.edu/training/visit/training_sessions/synthetic_imagery_in_forecasting_severe_weather/

On the morning of 19 Jan. 2011, a nice example of using the simulated WRF imagery to forecast low clouds presented itself.  The image on the right is the 17-hour forecast from the 00Z WRF run, and low clouds are evident in the Arkansas River Valley in southeast Colorado, as denoted with the arrow.  The left image is the corresponding GOES-13 IR image at 17 UTC, showing that these clouds were indeed observed.  A forecaster might then look at the forecast images over the next several hours to see whether the model dissipates the clouds.