Talking points for “Utility of GOES Imagery in Forecasting / Nowcasting Severe Weather”.

  1. Title
  2. Learning Objectives –The key to this training session is gaining an understanding of the utility of GOES satellite imagery in combination with other datasets (i.e. NWP output, surface and upper air data, radar etc.) in severe weather forecasting / nowcasting. The learning objectives are listed in a roughly chronological order going from assessing model output in the 1-2 day period leading up to the event, to the day of the event itself, to hours before the event, to while the event is unfolding. The role of observational data becomes increasingly important as we get closer to the time of the event.
  3. The first learning objective will be to assess model performance. This pertains to comparing observational data to NWP output to gain confidence in one model solution versus another.
  4. In the situation where the position error of a feature of interest (i.e. 500 mb trough) is carried along from the previous run, a comparison with observational data can be made to determine which model solution is more correct in terms of location of the feature of interest.
  5. We will focus in on a case where various models had different solutions the day before a potential severe weather event. Comparison with observational data can be used to show which of the model solution(s) are outliers and therefore be treated with less confidence relative to other model solutions. In this case, the potential severe weather event is forecast to occur on April 23 over the Great Plains. The trough responsible for the potential severe weather episode is investigated as it moves through the western US. We will be looking at various model forecasts of 500 mb height from the 0000 UTC 22 April model runs, water vapor imagery from 0045 – 1215 UTC 23 April 2007 is overlaid. GFS, NAM and UKMET 500 mb height forecasts show model solution divergence beginning over Nevada by 0600 UTC and becoming more evident by 1200 UTC. The UKMET is fastest (furthest southeast), the NAM is the slowest (furthest northwest) and the GFS is in between. The low center circulation does appear to be further southeast in the water vapor imagery by 1200 UTC, however this is subtle so we look for more evidence of that in other data. Overlay the GOES winds in the vicinity of 500 mb at 0915 UTC to compare the GOES winds with the winds from the 3 model runs. Determine which of the model(s) fits the data better and which model(s) do not. Comparison between GOES winds and model output is particularly useful far from the 00 and 12 Z upper air times. The frame at 1215 UTC displays the observed 500 mb plots. Compare the plotted data with the height/wind data from the 3 models. The NAM appears too slow with the trough, it should be further southeast. The GFS is better but still some disagreement with the Flagstaff observation which suggests the trough is slightly further southeast. The UKMET has the best agreement with the Flagstaff and Grand Junction observations. Keep in mind, this trend in position of the 500 mb trough has been present in previous model runs as well (not shown for April 21 runs) with the NAM being slower, the UKMET faster and the GFS in between. Noticing this trend would yield greater confidence in the UKMET in terms of position of the 500 mb trough, with slightly less confidence in the GFS and less confidence in the NAM forecast for the late afternoon/evening of April 23 in the Plains.
  6. The second learning objective is air mass identification. Sometimes there are signatures in the visible satellite imagery that can be used in tandem with surface and upper air data to delineate different air masses / boundaries.
  7. In this example, GOES visible imagery shows a boundary between an unstable air mass to the south and a stable air mass to the north. The boundary is a warm front that was reinforced by an earlier MCS outflow boundary. The cloud streets to the south of the boundary characterize an unstable air mass. The orientation of the cloud streets are parallel to the surface winds. The stable wave clouds to the north of the boundary characterize a stable air mass. The orientation of the stable wave clouds are perpendicular to the wind at inversion top level.
  8. GOES visible imagery 1302 – 1932 UTC 11 May 2000. METARs may be toggled on/off with the “11May00” check box in the controls frame. For this case we will be focusing on Iowa. In the animated imagery we observe the development of a region of cloud streets moving northeast, with a region of stable wave clouds to the northeast. If we toggle on the METARs we can detect a low pressure area associated with the cloud fields discussed above moving to the northeast. A warm front exists northeast of the low, and a cold front extends southwest of the low. Note the movement of the unstable cloud streets towards the northeast as the southwest flow moves in.
  9. GOES visible imagery 1932 UTC 11 May – 0045 12 May 2000. METARs may be toggled on/off with the “11May00” check box in the controls frame. The stable wave clouds go away, likely due to a combination of daytime heating and advection of the boundary between the cloud streets and wave clouds towards the northeast. Convective initiation occurs around 2245 UTC just ahead of the low in the region of the stable wave cloud and cloud streets boundary. The convergence in this region is maximized. The conditions were favorable for tornadoes this day so this storm produced a tornado at 2345 UTC, soon after convective initiation.
  10. GOES visible imagery 1300 – 1900 UTC 31 May 1985. METARs may be toggled on/off with the “31May85” check box in the controls frame. Identify the various air masses / boundaries across Ohio, Pennsylvania and western New York. A north-south oriented line of convection is moving through the area during the morning hours, this moves eastward during the day. The highest dewpoints are immediately east of the cold front, with south/southwest winds advecting moisture northward. Some regions in northern Ohio transitioned from stable wave clouds to unstable cloud streets with daytime heating. Clearing is taking place across much of northern Ohio and into northwest Pennsylvania. In Ohio, there are indications of a convergence line (a line of enhanced Cu) ahead of the cold front. Note the transition from southwest winds over much of Ohio to more backed, southerly to southeasterly winds across northwest PA, western NY extending into Ontario. There is a region of subsidence in southeast Ohio.
  11. GOES visible imagery 1900 – 2354 UTC 31 May 1985. METARs may be toggled on/off with the “31May85PM” check box in the controls frame. The first thunderstorms in the region of interest initiate along a pre-frontal trough in northeast Ohio. These thunderstorms quickly become severe and move eastward towards Pennsylvania. Note that thunderstorms activity is inhibited south of this line of storms in east central / southeast Ohio (recall the subsidence observed in the previous slide). Later on, thunderstorms develop on the cold front. The environment was favorable for tornadoes on this day, there were numerous F3 and F4 tornado events as well as a F5 tornado from northeast Ohio into northwest Pennsylvania.
  12. The third learning objective will deal with identification of changes in the pre-storm environment.
  13. One of the major reasons we see changes in the pre-storm environment are due to Mesoscale Convective System (MCS) activity. Monitoring MCS activity for potential effects on later convection is particularly important because NWP output can struggle with this, meaning that monitoring observational data is that much more important. Whenever we observe an MCS we should be thinking of its consequences on potential future convective activity. The first step is to identify the region that has been stabilized by the outflow from the MCS. The next step is to identify the boundary between the stable air mass you just identified and the potential unstable air mass, thunderstorms often develop along MCS outflow boundaries. Finally, the outflow boundary should be monitored for new convective development, or the interaction between a storm that develops in the warm sector and the MCS outflow boundary. The interaction can be favorable for a period of time or unfavorable depending on storm motion, instability, moisture depth, shear etc.
  14. Example of MCS and its effects on later convective development. GOES visible imagery from 1415 – 2215 5 May 2008. METARs may be toggled on/off with the “5May08” check box in the controls frame. Early in the loop we see a MCS in Kansas that leads to a well defined outflow boundary that moves southwest. Around 1900 UTC we can trace the outflow boundary as the line of cloud streets to the southwest of the boundary, and stable wave clouds to the northeast of the boundary. Around 2030 UTC we see convective initiation occurring along the MCS outflow boundary. There were numerous reports of large hail with these thunderstorms with the largest report being 4.25” in diameter.
  15. GFS forecast CAPE from 1200 UTC 24 May 2008, overlays are surface wind (GFS24May_Wind),dewpoint (GFS24May08_Td), 500 mb height and wind (GFS24May08_500), temperature (GFS24May_T), QPF (GFS_24May08_QPF), CIN (GFS24May08_CIN). At 500 mb we see a low over Wyoming that is forecast to move north during the day, a shortwave trough associated with this system is forecast to move northeastward across the plains during the day. At the surface we see a cold front / dryline forecast to move eastward, in response to the shortwave trough. Note the dewpoints and CAPE in the warm sector with CAPE values ranging from 4000 J/Kg in OK/KS to 3000 J/Kg in NE and lesser values further north in South Dakota. CIN values by 0000 UTC suggest convection initiation likely along the dryline / front from SD to NE to KS with more CIN further south in OK and TX.
  16. As in previous slide, except this is the NAM forecast 1200 UTC 24 May 2008. We show the NAM forecast here as well, which depicts a similar scenario to the GFS. Of particular interest is the magnitude of the CAPE in the warm sector in NE and KS – a large region in the 3000 to 4000+ J/Kg range.
  17. GOES 10.7 um IR imagery during the overnight and morning hours of 24 May 2008. Note the MCS over Kansas and Nebraska during the overnight hours. The MCS stabilizes a large area in its wake. This will have a profound effect on the CAPE forecast, particularly over Nebraska and Kansas. Overlays (24MAYGFS00Z_QPF and 24MAYNAM00Z_QPF) are the GFS and NAM QPF forecasts respectively, from the 00Z run. The GFS had the majority of the QPF too far northwest while the NAM also had this problem as well as a lack of QPF with the MCS over Kansas. The models typically do not have a good handle on MCS’s, and in particular the effects on future convective development. This is where monitoring observational data is critical for anticipating effects on convection later in the day. Refer back to slides 16 and 17 (the 12Z NAM and GFS forecasts) and show the 24May_METAR overlay to see where the models were overestimating CAPE.
  18. SPC Mesoanalysis of SBCAPE at 0000 UTC 25 May 2008 and (toggle) MLCAPE. CAPE values are significantly lower than forecast in the warm sector across Kansas and Nebraska. South of the MCS in Oklahoma, CAPE values are in the 3000 – 5000 J/Kg range.
  19. GOES visible imagery from 1545 – 2310 UTC 24 May 2008. MCS outflow boundaries can be seen across Kansas and Oklahoma. A storm initiates along the dryline / MCS outflow boundary intersection in Oklahoma, we will look at this in more detail in the next slide. The convection in Nebraska is weak (just one severe weather report – 0.75” hail) and storms do not develop in Kansas. Supercells do develop in South Dakota close to the 500 mb low, where 500 mb temperatures are much colder than further south. The MCS had a significant influence on the warm sector and subsequent distribution of severe weather reports which were primarily confined to South Dakota and Oklahoma.
  20. GOES visible imagery from 1445 – 2202 UTC 24 May 2008 centered over northern Oklahoma. Outflow from the MCS stabilized the area of north central to northeast Oklahoma, northward into Kansas. The outflow boundary intersects the dryline in northwest Oklahoma where convective initiation occurs. CAPE values in this region were in the 3000-4000 J/Kg range. The initial storm moves into the stable air (note the stable wave clouds) and dissipates, however, an outflow boundary from this storm intersects the MCS outflow boundary to cause additional convective initiation. The later storm propagates along the MCS boundary and produces a EF2 tornado.
  21. Now we will look at a short case study to analyze changes in the pre-storm environment. The case is from 5 July 2000. We begin by looking at NWP output. The image shows CAPE from the 1200 UTC run of the Eta model. Advance the frames to inspect the various forecast times at 3 hour increments. Also make use of the overlays in the controls frame, these include the 1200 UTC run Eta model fields of 500 mb wind, surface dewpoint, MSLP, surface temperature and surface wind. Our forecast area is northeast Colorado. 500 mb winds show southwest flow in the 25-30 knot range over the forecast area. MSLP and surface winds show pressure falls taking place to the west during the day with a broad low pressure area across Colorado and Wyoming by 0000 UTC. Winds across northeast Colorado are forecast to be easterly at 1800 UTC and become more southeast, then south/southeast by 0000 UTC in the forecast area. Forecast temperature and dewpoint fields show relatively high dewpoints in the morning to mid-day hours (in the northeast/easterly flow regime). Forecast temperatures reach the mid-90’s by 2100 UTC and the corresponding dewpoints drop dramatically at this time, likely in response to mixing out the moisture. This leads to the most unstable air mass in place in the late morning to early afternoon hours across the forecast area.
  22. GOES visible imagery during the morning hours and METARs. During the morning hours, we see the MCS in Nebraska and the outflow associated with it moving to the southwest. Winds in northeast Colorado are east/northeasterly by 1800 UTC with dewpoints in the upper 50s to 60s, consistent with the Eta forecast up to that time.
  23. Now we look later in the day, this is the GOES 10.7 um imagery from 1625 – 2325 UTC 5 July 2000. The MCS in Nebraska can be seen early in the loop, moving east out of the picture. The outflow boundary can be easily identified in the IR channel since the air mass is relatively moist compared to its surroundings. The moist air mass will show up as cooler in the IR imagery, which is a lighter color in this color table. Speed up the loop to watch the MCS outflow boundary moving from southwest Nebraska into northeast Colorado. Recall the Eta model had temperatures rising in this region and dewpoints dropping due to mixing. In reality, we see the moist air mass moving westward, providing for an unstable air mass and potential enhancement for convection along the outflow boundary. Also, the elevation rises appreciably across this region - North Platte (2800 feet) to northeast Colorado on the moist side of the boundary (~4000 feet) – so that the effects of elevated heating will contribute to higher surface theta-e.
  24. GOES visible imagery 1645 5 July – 0202 6 July 2000. We see indications of the MCS outflow boundary moving westward across northeast Colorado around 2200 UTC as indicated by a line of cumulus. This continues to move westward, but we lose track of it in the visible imagery as the cumulus along it dissipates. The early convective development occurs in southwest Nebraska, along the southern edge of the MCS outflow boundary, as well as in Wyoming (and later the Nebraska panhandle) as storms initiate there. Keep in mind, the storms in Wyoming and the Nebraska panhandle have high LCL’s as the temperature / dewpoint spread was very high in the environment where these storms initiated. However, we do see an interesting development with a thunderstorm in the south central portions of the Nebraska panhandle, starting around 0030 UTC. This thunderstorm has a more impressive overshooting top with it, and begins turning to the right around 0045, these are indications of a more intense thunderstorm (we’ll look at radar reflectivity in the next slide). A bit later, around 0115 UTC this storm is looking very intense and has storm scale features around it, there are cumulus above an invigorated RFD, we will discuss this in more detail later in this session. For now, we’ll say this signature is associated with an intensifying thunderstorm. The thunderstorm intensified dramatically as it encountered the more unstable air mass from the MCS outflow boundary. Also, it may have moved to the right along the boundary, helping to intensify the storm. The later is hard to say for sure, but the storm motion is at least a hybrid of right movement due to dynamical effects as well as propagation along the MCS outflow boundary.
  25. KGLD 0.5 degree base reflectivity along with METARs 0036 – 0206 6 July 2000. The MCS outflow boundary shows up nicely in this loop.