A large brush fire developed and spread quickly during the afternoon of 11 July 2019 on the Hawaiian island of Maui, and continued to spread through the evening. By 05Z, the fire was reported to have burned 3,000 acres. The wildfire was captured by GOES-West, which includes the Hawaiian islands in the southwest corner of its 5-min CONUS sector.
The fire first becomes apparent in the 3.9 um shortwave IR imagery over the west-central portion of the island at 2051 UTC, and quickly heats to a brightness temperature of 127C in its hottest pixel (Figure 1). Compare this to a temperature of 41C over the same pixel just prior to wildfire initiation. The fire quickly expands and spreads to the southeast during the afternoon. The fire slows its movement and cools slightly during the evening.
Photos on social media showed a large smoke plume associated with this wildfire, which was also captured by GOES-West 0.64 um visible imagery (Figure 2). The smoke was observed pooling in the lower elevations within the center part of the island, while also being advected to the southwest in the mid levels. An intense updraft is apparent within the smoke field over the fire between 0221 UTC and 0331 UTC.
The natural true color RGB imagery available in AWIPS provides an image similar to what one would see from outer space. In lieu of a green band on ABI, the green component in this RGB is approximated by combining the 0.47 um, 0.64 um, and 0.86 um bands. Of course, the 0.47 um band is used for the blue component of the RGB, and 0.64 um for the red component. In this case, the brown smoke stands out against the white clouds, green land, and blue ocean.
GOES-West GLM data recently became available to several NWS offices. This, of course, benefits western US NWS WFOs and offices with Pacific forecasting duties. Recall that GLM gridded products are reformatted to the 2 km ABI fixed grid, so share the same parallax as the ABI imagery (shifted away from nadir). With GOES-East at 75.2W, and GOES-West at 137W, the “cutoff” for “better” GLM data (less parallax, better detection efficiency) is at 106W (equal distance from both satellite subpoints). Offices west of 106W should use GOES-West GLM, and offices east of 106W should use GOES-East GLM (see Fig 1).
A comparison between GOES-East and -West GLM FED and ABI VIS data over eastern Montana (near 106W) using ENI point data as a constant (with assumed very little parallax) shows a similar degree of parallax between the two satellites (Fig 2). Of course, from the East point of view, the GLM and ABI data are displaced to the north and west, while that from the West point of view are displaced to the north and East.
Another example from northeast California (120.5W), well west of 106W and closer to GOES-West subpoint, shows similar parallax between the two satellites meridionally (Fig 3). Zonally, however, there is significantly less parallax displacement in the GOES-West data (minor shift to the east) than in the GOES-East data (bigger shift to the west). Additionally, the nearer GOES-West is able to detect more flashes, and the GOES-East pixels are stretched further away from the satellite sub-point.
Analyzing GOES-East GLM data over the northeast US (near 75W and 38N), we observe no parallax in the east-west direction, but significant displacement to the north (Fig 4).
Now looking further south (~15N) but still near 75W, northern displacement is much less, with ENI and GLM detentions very similar in location (Fig 5). The west-east parallax remains negligible.
One final example, this time from GOES-West, shows parallax with lightning associated with Hurricane Barbara remnants just south of Hawaii. Being south of 20N but between 150W and 160W (well west of the 135W subpoint), the parallax appears negligible to the north, but a shift to the west is apparent. (Fig 6). Both ENI and GLD lightning data are used in this case to ensure data quality in this remote area.
Currently, only GOES-West GLM data within the GOES-West CONUS ABI sector are sent to WFO AWIPS (covers the western 1/3 of CONUS). Similarly, only GOES-East GLM data within the GOES-East CONUS sector is sent to WFO AWIPS (covers the full CONUS). Figures 7 and 8 show GLM coverage over the CONUS from both satellites, respectively for the same time period on 3 July 2019.
Therefore, while all CONUS offices are covered by GOES-East GLM data, only offices in the western 1/3 of the CONUS receive GOES-West GLM over their area of responsibility. If GOES-East GLM suffers a data outage, the eastern 2/3 of the US will be left without any GLM data. If GOES-West GLM suffers an outage, all of the CONUS will still have access to GLM data from GOES-East.
Fortunately, there are plans to extend GOES-West GLM coverage in AWIPS eastward to near Kentucky in the coming weeks.
The Storm Prediction Center included parts of southern Colorado in a Moderate Risk (four on a scale of five) for Severe Weather in the 1300 UTC Day 1 Convective Outlook on 26 May 2019. This was the first instance since 18 May 2010 that a portion of the Pueblo, CO CWA was included in a 1300 UTC or later SPC Day 1 Moderate Risk for Severe. All severe risks were possible, including significant tornadoes, significant hail, and severe wind gusts. Given the risk, 1-min imagery from GOES-16 was available over the region.
Analysis of GOES-17 water vapor imagery from the 26th revealed an impressive shortwave trough accelerating across the Baja California Peninsula and Gulf of California during the early morning hours, and lift across New Mexico and into west Texas during the afternoon into the evening while developing a negative tilt (Fig 1). A jet streak is analyzed rounding the southeast portion of the shortwave into E Co/W KS by the end of the loop. Convective initiation in an associated region of large scale ascent ahead of the shortwave and under the increasing mid-level flow is apparent during the afternoon near the end of the animation across the high plains. Meanwhile, a broad upper low shifts south across central California.
RAP analysis confirms the above mentioned features with the yellow contour representing 50+ knot 500 mb wind, and white contours represent relatively high values of 500 mb positive vorticity (Fig 2).
Morning Day Cloud Phase Distinction RGB imagery from GOES-16 implied abundant low-level moisture per widespread low stratus clouds (cyan), confirmed by dew points in the mid 50s across most of the eastern Colorado plains (Fig 3). By early afternoon under strong heating, much of the low stratus had eroded across the southeast plains, and mid-level clouds (green-yellow) were spreading across the area in response to the approaching shortwave. The Denver Cyclone mesoscale feature is diagnosed in the low stratus over northeast Colorado. Snow cover (green) is still readily apparent over the Colorado Rockies, a testament to the impressive snow totals seen this season.
Convection developed quickly during the early afternoon over southeast Colorado under strong large scale forcing and weak capping. One of the initial storms produced hail as large as 2.5″ in diameter in Springfield, CO (Baca County). A three-body scatter spike was detected in radar imagery, indicating large hail potential (Fig 4). Baca County in southeast Colorado is in an area of relatively poor radar coverage, with lowest radar tilts shooting 10,000 ft AGL. While this level upward generally provides enough information to make an informed (severe thunderstorm) warning decision, additional info such as that from satellite and lightning is useful in confirming warning decisions, particularly in marginal cases.
One-minute VIS and IR GOES-16 imagery in this case showed an above anvil cirrus plume (AACP) emanating from a persistent overshooting top (OT) early on in the life of the storm (Fig 5-8), confirming severe potential. The storm weakened quickly after 1920 UTC as evidenced by cloud top warming and a sudden loss of the OT and AACP in satellite imagery.
Meanwhile, a new storm developed to the northwest in Bent County, CO. At 2033 UTC, radar imagery showed a tightening velocity couplet and potential trend toward tornado development. Unfortunately after this time, the KPUX radar went down unexpectedly. Radar imagery capturing the storm during the 30 minutes leading up to the outage at 2033 UTC is shown in Fig 9.
With this new storm quickly developing a mature updraft and having its anvil sheared to the northeast, the updraft was exposed to the view of the satellite (from the southeast). As has been observed with a few other cases, the 500 m, 1-min visible imagery captured rotation within this exposed updraft, confirming a mesocyclone/supercell thunderstorm (Fig 10).
After a few minutes of no new radar data (2040 UTC), the warning forecaster decided to issue a tornado warning given: 1) the environment supported tornado development, 2) the last radar scan showed meager rotation just starting to develop, 3) a consistent rotating updraft apparent in visible satellite imagery, 4) cloud tops remained cold in IR satellite imagery, and 5) the storm had quickly developed an AACP apparent in VIS and IR satellite imagery. One-minute VIS and IR satellite imagery during the 30-min period leading up the warning decision at 2040 UTC are shown in Figs 10-11.
KPUX radar would return at 2045 UTC, and the storm went on to produce a confirmed tornado at 2054 UTC and again at 2210 UTC, in addition to numerous instances of large hail. A 2-hr long animation (2010 – 2210 UTC) of the storm is shown in Figs 12-14. The rotating updraft, OT, and AACP are all readily apparent throughout the life of the storm.
AWIPS provides an image combination option that automatically transitions imagery from IR to VIS and VIS to IR during the sunrise and sunset (Fig 1). This prevents the user from needing to change imagery/panels during the day/night transition, and provides pleasing imagery to share with the public.
Unfortunately, AWIPS does not make it easy to apply this option to RGB imagery. After some tinkering, I was able to get it to work.
The VIS/IR Sandwich RGB recently become available in AWIPS-II. Since this RGB is only usable during the day and primarily for monitoring convection, it makes sense to transition to IR alone at night. The example below shows mature convection driving south across central Texas during the afternoon into the evening of 9 June 2019 (Fig 2). The imagery automatically transitions from the VIS/IR Sandwich RGB to IR single-band imagery during sunset. Fig 3 provides a wider view.
The VIS/IR Sandwich RGB to IR transition provided an impressive view of the development and evolution of a prolific hail producing storm over Kansas during the early evening of 11 June (Fig 4).
Another useful transition is the cloud monitoring RGBs. Many blog posts have been written on the utility of the Nighttime Microphysics RGB for monitoring clouds at night, and the Day Cloud Phase Distinction RGB for doing the same during the day. This feature in AWIPS allows for a transition between the two RGBs within a single loop (Fig 5).
It should be noted that the transition feature can slow AWIPS, particularly if looking over a large area. Finally, feel free to reach out if you’d like the code to do this.
A vigorous trough and favorable thermodynamic environment led to the development of numerous severe storms and flash flooding during the day on 20 May through the evening and into the next morning. Storms produced over 25 tornadoes and many large hail and strong wind reports, in addition to flash flooding.
Water vapor imagery from the event showed the trough become negatively tilted as it swept across the Rockies into the plains from the 20th into the 21st (Fig 1). The negative tilt implies differential temperature/moisture advection in the vertical (increasing instability) and increasing wind shear. Combining the water vapor imagery with NWP analyses (in this case, hourly RAP) helps to improve analysis of the overall synoptic picture as features from the model can be connected to features in the imagery. Over time, this can improve ones ability to diagnose features in the imagery alone. Additionally, variances between the imagery and model can be noted and extrapolated into the model forecast. The NWP overlay in this case quantifies the strength of the trough and shows the strong jet rounding its base and advancing over the southern plains.
One-minute satellite imagery from GOES-16 was available to forecasters during this event. Visible imagery from west Texas/Texas Panhandle into western Oklahoma showed rapid thunderstorm evolution, including storm initiation and development of overshooting tops and above anvil cirrus plumes (Fig 2). The quick development of these storm top features was no surprise given the favorable setup, and indicated very strong updrafts and significant severe potential.
One-min IR imagery confirmed the trends and features, with rapid cooling rates implying swift updraft growth, the small cold regions indicating overshooting tops, and downstream warm regions surrounded by colder tops (enhanced V) suggesting the above anvil cirrus plumes (Fig 3).
A 19 hr long IR loop (17Z – 12Z) revealed persistent strong thunderstorm activity from west Texas across much of Oklahoma (Fig 4). Thunderstorms training over the same areas led to numerous reports of flash flooding.
A Mesoscale Convective System (MCS) containing a line of severe thunderstorms rolled across south Texas early in the morning on 03 May 2019. GOES-16 1-min imagery was available over the region to support forecast and warning efforts. One-minute IR imagery indicated a broad region of overshooting tops with gravity waves emanating out away from the updrafts (Fig 1). Gravity waves across the top of a convective system are formed when the updraft interacts with the stable tropopause. A strong updraft may “overshoot” the tropopause into the lower stratosphere, appearing as an overshooting top. In an effort to return to equilibrium, these air parcels go on to oscillate (sink and rise) past that equilibrium level. This air reaching and overshooting the tropopause is forced to oscillate outward and downstream by the mean flow, appearing as “waves” at the storm top.
Transverse banding (buzzsaw looking cloud area) was apparent in the imagery north of the main updraft region. These mid-upper level clouds are an indication of potential aircraft turbulence. There were indeed several aircraft reports of moderate turbulence through these cloud features.
GLM flash extent density overlaid on gray-scale IR imagery from the same period highlights the storm cores (highest lightning rates), as well as lightning flashes extending well away from the updrafts through the anvil region (Fig 2). The average flash area confirms smaller flashes associated with the active/newer updrafts, with longer flashes in the anvil away from the main updrafts.
One-minute visible imagery at sunrise revealed the storm top features in much greater detail, including the very active updraft/overshooting top region, and series of gravity waves across the anvil.
The day cloud phase distinction RB has been discussed as a useful tool for monitoring the cumulus cloud field leading up to and including convective initiation. Pure liquid cumulus clouds appear as cyan in the RGB because they have relatively high reflectance in the 0.64 um (high green) and 1.6 um (high blue) components, but are warm in the IR (low red). As convection begins to initiate and ice develops in the cloud top, the blue component decreases since ice does not reflect as well in the 1.6 um band (low blue), but the green component (0.64 um) stays the same or increases (high green). The red component begins to increase as the temperature of the clouds decreases (mid red). Therefore, for convective initiation, you are left with a transition from cyan to green.
In the 01 May case, you can see a transition from cyan to green in areas along the boundaries, indicating that convective initiation is imminent or occurring. There are quite a few orphan anvils in this case, indicating failed initiation but that the CAP is likely close to breaking. While these features are all apparent in the VIS alone as well, the color detail added to the imagery as a result of the combination of multiple channels makes these features and trends easier to diagnose. The high, 0.5 km, resolution available from the 0.64 um VIS is maintained in the RGB in AWIPS.
Once the convection matures, the color transitions from green to close to yellow since the convection is still highly reflective in the 0.64 um channel (high green), and is also now cold (high red), but still not as reflective in the 1.6 um channel (low green) due to the presence of ice.
Just as convection begins to initiate, attached are the three components to the RGB, plus the RGB, at 1721 UTC
It should be noted that with this case, the RGB was modified slightly. The 0.64 um (green) component had its max increased to 100% (from 78%) to account for the higher reflectance of cumulus clouds and mature convection during the middle of the day. This change is often necessary when using this RGB to monitor for convective initiation, and can be made easily in the “composite options” setting of the product. Without the change, the green component saturates quickly, leading to a loss of storm top detail in the imagery provided by the high resolution 0.64 um channel.