During the evening of April 21, 2020, several supercells developed in the Texas Panhandle into western Oklahoma producing hail resulting in numerous reports. Not only was the size of hail large, the amount of hail that fell was significant as noted by reports of accumulated hail of several inches deep. As storms snailed away from their deposited hail swaths, GOES-16 was there to view these swaths. This post will showcase how forecasters can view freshly deposited hail swaths using multispectral imagery at night.
Figure 1 depicts the Nighttime Microphysics RGB with the “clean” longwave infrared band (Band 13, 10.3 um) overlaid while making any temperatures warmer than -25 C transparent. This allows for cloud layer and phase information to be provided within the Nighttime Microphysics RGB while also monitoring very cold cloud top temperatures associated with severe storms as well as overshooting tops denoting an updraft penetrating into the tropopause. As the convective clouds and overshooting tops pull away from the border of the Texas Panhandle and western Oklahoma, notice one to three subtle, magenta-colored lines oriented northwest to southeast as skies clear (see annotated images here and here). These lines correlate well with highest reflectivities that had passed over previously in time as well as with storm reports of hail. These lines are a result of altered environmental and ground temperatures (in this case cooling) from the copious amounts of hail left behind their parent storms.
While the default composition ranges of the Nighttime Microphysics RGB within AWIPS-2 allow for just subtle signatures within the imagery, there is room for adjustment to enhance this signal. A good place to start is by looking at a 4 panel display of the RGB (Figure 2) being discussed and each of the its components set to a black to white scale with the RGB’s default ranges.
Sample the signal in question within each of the RGB’s components and adjust the range within the components until there is good contrast between the signal and background. Once the component’s ranges are adjusted to your liking, adjust them within the RGB itself. In this case, there was room for improvement mainly within the red component (the Split Window Difference, 10.3-12.3 um) and the blue component (“Clean” IR, Band 13, 10.3 um). However when adjusting the red component, the RGB itself became too saturated in red coloring. Therefore we can tone down the saturation of red by lowering the red component’s gamma. Figure 3 is what the new composition looks like now (click here for actual recipe values).
Do you think the hail swaths show up better in the adjustments as shown in Figure 4? Making an adjustment on the fly works in this case, but might not for all cases. Thus adjustments to RGBs are advised to only be made on a case by case basis, especially with those having a large dependency on infrared temperatures. It is also worth noting that this approach of using multispectral satellite imagery to observe accumulated hail swaths is dependent on cloud-free skies as well as enough hail to actually make an alteration to the ambient environmental and ground temperatures.
So how does this apply to IDSS? Well besides increasing overall situational awareness, noting hail swaths in near real time lends confidence in derived products like MRMS MESH tracks through observations. This could also allow forecasters to help core partners like DOT maintenance crews target areas for hail removal should they request it. Additionally, this could also provide an opportunity for target LSR intel gathering, or even safety messaging as accumulated hail can produce hazardous travel conditions through slick roadways and reduced visibility should hail fog form (which by the way is also detectable in this approach through the use of the 3.9-10.3 um “Fog/Low Stratus” Brightness Temperature Difference as the Nighttime Microphysics RGB’s green component).
Flooding within the Red River of the North basin is practically a yearly rite of passage for those living there signaling the exit of winter and coming of summer. This area straddling the North Dakota and Minnesota border is well prepared for river and overland flooding because of its frequency. Spring of 2020 is proving to be no different with major flooding occurring over much of the Red River and its tributaries due to seasonal snow melt.
It can be difficult to keep a pulse on how impactful flooding really is with a lack of comprehensive reports in an ever-evolving and fluid situation. Plus with extensive flood mitigation systems in place, the majority of floods that occur in this area are nothing more than just an inconvenience for most people. Still for the minority that have property impacted by flooding, it can be quite notable and our services can be tailored to them, or at least towards the emergency services aiding them. Additionally, a threat to life can come to fruition if one finds themselves in very cold floodwaters (an example could be a car sliding off the road into floodwaters). This is ultimately our mission as National Weather Service meteorologists: to protect life and property.
This post will explore the satellite imagery that aided NWS Grand Forks in flood operations in mid-April 2020. In this case, flooding was due to springtime snow melt, but the following procedures and workflows may be applicable to warm season convection as well.
On April 10, 2020, moderate to major flooding was occurring along the Red River and its tributaries within the central and northern basin. Wondering if there were any areas experiencing impactful flooding outside of current flooding warnings, forecasters initially turned to GOES-R ABI River Flood Products for help in highlighting areas of observed floodwater coverage. Overlaying flood warning polygons, forecasters then looked for areas highlighted by the ABI River Flood Extent product not encompassed by polygons. This was the case in western Polk and Marshall counties, Minnesota, as noted in Figure 1. Values higher than 60% (orange and red coloring) were of particular interest as lower values into the 30-50% range were believed to be that of non-impactful standing meltwater in agricultural fields. While the ABI River Flood Products are updated every hour, the ABI’s spatial resolution of 1 km can smooth out the spatial extent of potentially impactful flood waters. VIIRS offers this same imagery at a finer resolution of 375 m, but at the expense of one or two images per day which require a daytime, cloud-free sky to provide useful information.
Luckily this area of interest was mostly cloud-free during one of the VIIRS passes. It confirmed higher percentage values in the same areas of interest, particularly in northwest Polk County, Minnesota (Fig 2).
There is additional satellite imagery available to further provide details on floodwaters and its potential impacts. Higher resolution satellite imagery down to 10 m from Sentinel-2 and Landsat 8 has recently become available on the web and already processed for quick viewing at sites like Sentinel-Hub and Remote Pixel. A timely, cloud-free pass from the Sentinel-2 satellite over the area of interest was available to forecasters for interrogation (Fig 3). It revealed extensive break out water from the Red River north of Grand Forks, North Dakota, and surrounding Oslo, Minnesota, although these areas were well within the flood warning polygons. What about the other areas of expansive meltwater in agricultural plots between Alvarado, East Grand Forks, and Tabor, Minnesota? Sentinel-2 imagery hinted that some of these floodwaters might be over roads and surrounded farmsteads. These areas of adjacent flooded plots of land correlated nicely with higher percentages within the ABI and to an extent VIIRS River Flood Extent products, thus confirming impactful flooding would possible here.
Forecasters took this approach of using satellite imagery to hone in on targeted areas for intel gathering of flood impacts. Looking for road closures on state Department of Transportation and county websites, data mining on social media, and calling emergency managers confirmed impactful flooding in these areas. Thus, these areas could warrant a flood headline. Ultimately the decision was made to not issue an Aerial Flood Warning as the majority of these areas fell just within current flood warning polygons.
So does the application of this imagery stop here? Not quite. ABI and VIIRS River Flood Extent products highlight nicely the extent of observed floodwater. In this case it was used as a source for the graphic in Figure 4. The imagery answers the “where” and “when” of flooding within the graphic while the photo depicts the impact, capped off with safety messaging. This imagery also yielded high confidence in overland flooding impacts which was messaged via the Hydro section within the Area Forecast Discussion from NWS Grand Forks. Additional imagery analysis was conveyed to the North-Central River Forecast Center which was useful in assessing the current state of snowpack, river and ditch ice, and floodwater expanse.
In summation, satellite imagery from ABI and VIIRS River Flood Products as well as from Sentinel-2 provided excellent details in gauging flooding impacts from river and overland flooding. Starting with the most coarse spatial resolution, yet highest temporal resolution imagery like ABI was a good starting point in searching for floodwaters over a broad area. Then, incrementally honing in on highest flood extent signals at higher spatial resolution imagery proved to be a good workflow in pinpointing areas of interest to target intel gathering of current impacts from flooding. Furthermore, this imagery was useful in messaging through graphics, discussions, and collaboration.
Did you know Nighttime Microphysics RGB can by used to diagnose floodwaters (Fig 5)? The highest contribution of sensed floodwater comes from the Split Window Difference product. In this case, there was high contrast between lower values of liquid bodies of water and higher values of land. However, due to dependence on sensed infrared energy, seasonal and air mass differences can change the appearance of floodwaters within the RGB. You can adjust the RGB to help draw out this floodwater signal from the Split Window Difference as well.
Besides high resolution imagery from Sentinel-2 and Landsat 7/8, cloud penetrating Sentinel-1 offers similar resolution imagery while capable of detecting liquid water versus snow and bare ground (Fig 6). Although, it does take a more careful eye in interpretation of the imagery, especially when looking at “waterlogged” snowpack which can give a false signal for liquid water. This type of imagery can be extremely useful when pesky clouds prevent imaging radiometers from sensing the ground.
Well, we can turn to the ABI on board GOES-East to help fulfill this answer, specifically the 1.61 um channel known as the Snow/Ice Band. At this wavelength, solar radiation energy is strongly absorbed by snow and ice with little reflected energy travelling back towards the sensor. This is why snow and ice show up relatively dark in the 1.61 um channel compared to snow-free ground and liquid phased clouds. While both are efficient at absorbing in this wavelength, ice still more strongly absorbs radiation than snow, allowing ice to appear even darker than snow. Additionally, wide areas of liquid water are even more strongly absorbed in the 1.61 um. This makes flood waters including major flooding along the Red River of the North and other ice-free lakes very dark. It is possible that darker swaths in southern Minnesota into Iowa may be enhanced from higher soil moisture from rain prior to a wintry mix producing ice.
With a mostly clear sky in place on April 4, we can use these properties to get an idea of where ice accretion occurred. While we are at it, let’s go down the multispectral imagery path and look at the Day Snow-Fog RGB which includes the 1.61 um channel.
Notice a dark red swath extending down the spine of Minnesota, into eastern South Dakota and into Iowa. There is also another area in southeastern Minnesota, again into Iowa, and western Wisconsin. These are areas where ice from freezing rain and sleet accrued. The streak-like nature of this signature within Iowa and southern Minnesota points to convective elements and showery activity leading to widely varied accretions and likely associated impacts telling us this was a very difficult forecast to pinpoint. Overall, guidance from HREF and NBM did well highlighting the general area, although they likely smoothed out these high spatially varied accretions within their ensemble systems.
The Day Snow-Fog RGB utilizes the 1.61 um channel as it’s green component. Less reflectance of snow and ice in this channel leads to lesser green values added to the overall combination. Within the RGB’s red component, the visible-like 0.87 um channel, snow is much more reflective with little or no reflectance from ice accrued areas. These differences make for easily noticeable contrasts between snow and ice, making it easier for forecasters to diagnose areas that experienced overall more ice than snow.
While the darker signature of icing helped forecasters see where icing may have been more prevalent than snowfall, it does not mean this was the only area of icing. Just to the west of the dark red strip in western Minnesota significant icing still occurred as depicted in this USGS photo of ice accrual on a river gage near Fargo disrupting data transmission. However, a transition from wintry mix later to accumulating snowfall lead to accumulated snow hiding the icing signature. The same could be said for patch around southeastern Minnesota.
As the power of an April sun warmed ground temperatures in the Upper Midwest, ice and thinner snowpack quickly disappears on the Day Snow/Fog RGB and CIRA’s Snow/Cloud Layers RGB.
So how is this information useful after the storm? Not only does it increase a forecaster’s situational awareness of potentially highest impacted areas, it can be used as an Impact Decision Support Service tool for illustrating to partners where exactly these conditions occurred. Forecasters at the National Weather Service in Grand Forks, North Dakota, and hydrologists at the North Central River Forecast Center used this imagery to gauge important spatiotemporal characteristics of the associated liquid water equivalent leading to conclusions of which locations would first see this water move into area rivers already undergoing flooding from winter snowpack melt. Additionally, satellite imagery including the Day Snow Fog RGB image above was used in a hydrologic briefing to core partners within the Red River of the North basin, giving them an idea of these important characteristics while using imagery to justify some aspect of updated river forecasts.
March 2020 started off rather windy and unstable in the Northern Plains leading to convection during a month that still sees predominately wintertime stratiform precipitation. There were several instances of showers producing lightning, graupel/snow, and convectively forced severe wind gusts in addition to synoptically driven high winds. Designating convective cells that may be more impactful than others in the cool season (late fall, winter, and early spring) can be difficult using remote sensing instruments like satellite and radar due to their shallow heights and relatively warm cloud tops. Satellite appearance of shallow convection tends to be unimpressive in the infrared using default color schemes made for taller, colder cloud topped warm season convection. Also, due to the shallow depth of the cloud layer, radar can easily overshoot this layer composed of what tends to be marginal returns. Lastly, lightning activity tends to be much lower than that in the warm season, if present at all.
But there are some ways to make subtle features of shallow convection more easily tracked by adjusting the satellite display. Let’s explore..
First let’s take a look at an event on March 4, 2020, that produced several convectively induced high wind gusts at times exceeding 75 mph. The Day Snow-Fog RGB has proven useful in monitoring wintertime precipitation, mainly because of its use of near-infrared bands to help delineate snow cover, bare ground, and cloud top microphysical composition. It is also well known that using RGB’s capable of tracking cloud features while color sorting aforementioned properties (looking at you Day Cloud Phase Distinction RGB) can be quite useful for tracking forcing mechanisms, areas of potential convective initiation, and stage in convective life cycle.
Figure 1 shows convective clouds moving through eastern Montana into the Dakotas and northern Wyoming. Convection is already ongoing in the start of the loop with a line of showers attached to a cold front extending from near Great Falls to Glasgow, Montana. The cold front can be tracked on satellite propagating south and east undercutting additional convection in the “warm sector” ahead of the front. Ahead of the front are additional clusters of growing cumulus clouds as the convective temperature was reached, aided by ample insolation over a snow free surface beneath a relatively cool column of air (focus of cloud growth probably had some influence from differential heating and convergence augmented by variations in terrain). These cumulus clouds initially exhibit a highly textured, bubble-like appearance with whiter coloring indicating non-glaciated clouds. They very quickly become bright pink while maintaining their texture signaling glaciation of clouds through ascent. Finally they turn from cumulonimbus to stratocirrus losing texture, acquire more purple coloring, and move west to east in broader upper flow as parcels detrain from their initial paths of low level ascent, thus indicating the majority of additional ascent is likely over. The ability to track these mesoscale forcing mechanisms and stages in storm growth can be important to forecasters as this may give lead time to potential initiation or advection of impactful convection, as well as convective decay.
Now let’s focus on the tristate region of the Dakotas and Montana, an area where several severe wind gusts were measured. Using Rapid City’s observed sounding as representative of the environment, this area was primed to transfer not only cloud layer winds of around 50 kt towards the surface, but any evaporation of descending precipitation cores may help accelerate these winds from aloft towards the surface (DCAPE ~700 J/kg, inverted V sounding).
We can utilize procedures meant for tracking severe convection in the warm season while making some minor adjustments to help gear towards a cool season environment. Figure 2 is an example of one such procedure. The range of the Day Cloud Phase Distinction RGB’s red component (10.3 um, Band 13, “Clean IR”) was edited from [Min: 7.5, Max: -53.5] to [Min: -10, Max -70] in an effort to desaturate red coloring by shifting this thermal sensing band towards the colder, better representing the temperatures typical in the cool season. Convective evolution is represented by a transition of colors from cyan to green to yellow/red as clouds glaciate and then cool. The inclusion of the visible spectrum within the Day Cloud Phase Distinction RGB also gave clues on storm development featuring more turbulent cloud texture near the upshear side of continually developing anvils, indicative of sustained updrafts reaching their equilibrium level. A separate, standalone look at Band 13 offered a way to track coldest cloud tops, although adjustments were also made here shortening the range from [Min: 55.0, Max: -109.0] to [Min: 0, Max -70] while making all values warmer than -25 C transparent. This allowed robust convection to be more easily tracked in an area of poor radar coverage with more red colored tops corresponding to relatively taller, stronger updrafts. One particular cluster of showers moved through Glendive, MT, with its AWOS measuring a 47 kt (54 mph) wind gust 18:56 UTC followed by Dickinson, ND, reporting 66 kt (76 mph) wind gust 19:56 UTC. This cluster exhibited relatively rapid cloud top cooling around 40 minutes prior to reaching Glendive and another prior to reaching Dickinson. Lastly, lightning activity can be analogous to updraft strength. Adjustments to the GLM FED appearance were made to shorten the default range of 1-260 to 1-20 helping draw out meager lightning rates. While overall lightning activity was low (as can be typical in cool season convection in the Northern Plains), this adjustment did help bring attention to cells that had relatively high activity compared to neighboring cells. One cell that stands out emanates from the same cluster that moved through Glendive and Dickinson but on its southern flank. Looking at MRMS isothermal reflectivity at -20C, the cell holding highest lightning activity has reflectivity exceeding 40 dBZ. All of these signals, upstream reported severe wind gusts, and our knowledge of the thermodynamic and kinematic environment can steer the forecaster into thinking this particular cell is strong and chances of producing severe downdraft wind gusts are high. This prompted NWS Bismarck’s to issue its earliest severe thunderstorm warning with an eventual 81 mph wind gust being reported at a mesonet site in Hettinger County, North Dakota.
A lightning-producing cell precipitating rain, snow, and graupel formed west of Bismarck, North Dakota, moving southeast into northeast South Dakota, during the evening of March 3, 2020. Using a similar procedure as displayed before, Figure 3 shows relatively rapid cooling of an expanding cloud top/anvil as noted in Band 13 (default color scheme shown). Values exceeding 35 dBZ in MRMS isothermal reflectivity at -20 C correlated nicely with the appearance of sensed lightning by GLM and ground based networks.
A graupel/snow depositing shower moved through Fargo, North Dakota, near sunrise March 3, 2020. This along with other showers left a trail of accumulated graupel/snow apparent on the Day Snow-Fog RGB (Figure 4). This is seen as more orange shaded streaks atop the background darker red coloring of the older snowpack. The shower moved through the Fargo-Moorhead metro right as sunlight was becoming available. Thus, it was beneficial to adjust the near infrared components of the RGB to account for little visible light to work with. This helped brighten the image revealing the metro was right on the edge of new accumulation (which lined up nicely with radar).
Later in the day, clouds peeled away revealing additional showers had deposited new frozen precip accumulation throughout eastern North Dakota into northwest Minnesota (Figure 5).
The main takeaway of this post is to encourage forecasters to become more comfortable with adjusting satellite products to help draw out convective features. Such adjustments may be especially necessary during the cool season. That’s not to say just blindly make adjustments, rather forecasters need to understand what is being adjusted and how that would affect the appearance and meaning of edited products. After determining an environment is possible for impactful convection, forecasters should think about adjusting satellite display to make subtle convective features typical in the cool season more easily tracked.
On February 12, 2020, a very strong arctic cold front swept through the upper tier of the central CONUS bringing blizzard conditions, dangerous wind chills colder than -50 F, wind gusts exceeding 60 mph, and crashing temperatures to portions of the Dakotas, Minnesota, Iowa, and Nebraska. Despite little snowfall expected with this front (generally between 1-2 inches), blowing snow and significant visibility reductions appeared likely should the existing snow pack be susceptible to being lofted and blown around. But prior to this impactful frontal passage, forecasters were left with a difficult decision: just how susceptible is the existing snow pack to become lofted into blowing snow?
One tool that could help forecasters answer this question is close examination of snow pack appearance on the 1.61 um Snow/Ice band offered on GOES-R ABI. It’s has been shown that the Snow/Ice band is sensitive to the amount of liquid water in a volume of snow and ice, i.e. its “water to ice crystal ratio” (CIMSS Blog example). We can apply this in operations by looking for “old, crusted over snow” from “new, fresh and fluffy snow.” In Figure 1, a swath of newly felled snowfall across eastern SD, southern MN, northern IA, into WI and MI can be seen as a lighter shade of grey compared to darker northern neighbors (you need to wait for bright white clouds to pull away to reveal the dark snow pack below). This lighter shaded area is where forecasters can more confidently delineate a fresher snow pack that may be more susceptible to being lofted and blown around. Lastly, this loop applies adjustment to the display range of the Snow/Ice band in order to more easily draw out this area, changing from the AWIPS default Min-Max of 0-100 to an adjusted Min-Max of 3-30.
As the cold front encroached upon Canadian border states of the Northern Plains overnight, close examination of the GOES-East Nighttime Microphysics RGB (Figure 2) revealed the exact location of the arctic cold front by way of rapidly advancing, arching area of low stratus towards the south embedded or underneath mid-upper level clouds moving northeastward. Higher clouds were associated with a weak mid level and surface wave moving east along the strong baroclinic zone charging southward.
Figure 3 annotates the location of the arctic cold front as well as METAR observations with the Nighttime Microphysics RGB. Closely scrutinizing satellite imagery for any subtle details can help forecasters latch onto synoptic and mesoscale features, particularly those that bring hazardous weather like this arctic front whose impacts started immediately after frontal passage. Rapid temporal tracking of this feature offered by GOES-East could give forecasters details like timing of onset to impacts, something very important to IDSS.
As the sun rose on February 12, 2020, forecasters at NWS Grand Forks were anxious to see if blowing snow could be viewed on satellite imagery as area observations, reports, webcams, and radar suggested (as it turned out the crust on snow pack might have been broken north of the aforementioned area of newly felled snow due to very gusty winds exceeding 60 mph aiding to the production of blowing snow). Luckily a GOES-East mesoscale sector was over the FGF area (thanks DMX!) during this time allowing 1 minute imagery to provide the most up to date satellite view available. A look at the Day Snow-Fog RGB (proven to be useful in monitoring blowing snow during the day) gave indications of horizontal convective rolls associated with blowing snow, evident moving out of southern Manitoba into northern North Dakota, but perhaps not as quite obvious for operational usefulness (Figure 4).
While the blowing snow plumes were somewhat noticeable in Figure 4, the default Day Snow-Fog RGB composite ranges within AWIPS doesn’t show features well in times of low light, i.e. near sunrise/sunset and near poleward locations in winter. Since all three of the components (0.86 um, 1.61 um, 3.9-10.7um) that make up the Day Snow-Fog RGB are sensitive to solar reflectance, we can adjust the composite ranges of this RGB to become more representative of the little solar reflectance available just after sunrise in this high latitude location. Shrinking the RGB composite ranges from the default (R: 0-100, G: 0-70, B: 0-30) to (R: 0-20, G: 0-15, B: 0-25) allows the forecaster to more clearly see swaths of horizontal convective rolls that make up blowing snow (Figures 5 & 6). Forecasters can adjust RGB ranges on the fly like this to make features easier to track, making it more operationally useful. Just don’t forget to readjust the range as more sunlight becomes available!
Figure 7 displays how blowing snow could be tracked on satellite throughout the day as dry arctic air scoured away clouds moving south and east near the front (stratus colored white/lavender as well as some higher cirrus as orange/red). Blowing snow extended all along the Dakotas and Minnesota border with blowing snow seemingly influenced by not only the Red River Valley in eastern North Dakota and northwest Minnesota (Figure 8), but also by the Coteau De Prairie/Sisseton Hills in eastern South Dakota and southwestern Minnesota (Figure 9). Significantly reduced visibility as noted by METARs can be matched with the blowing snow plumes seen funneling down the lower elevation of the Red River Valley (Figure 8) and Buffalo River Valley (Figure 9). Also, cessation of blowing snow can be noted within the Red River Valley as the horizontal convective rolls dissipated north to south in Figure 7. Lastly, blowing snow plumes could be noted advecting southward out of southeastern South Dakota into the relatively snow-free far northeast Nebraska towards the end of the loop in Figure 9.
Analysis of GOES-East imagery during this event provided information on antecedent conditions as well as precise tracking of the arctic cold front and blowing snow causing blizzard conditions behind the front. This information was used directly in operations helping refine the boundary of blizzard warnings and winter weather advisories, social media messaging, as well as IDSS support to core partners of the National Weather Service.