Islands in warm tropical seas can create convection through a combination of factors that work together to foster deep convection. A recent example from the island of Guadalcanal in the Solomon Islands during the day of Monday 20 April 2026 illustrates this. Guadalcanal is famous as the site of a pivotal campaign of land and naval battles during World War II but today is known as a lush island of tropical rainforests, beaches, and the home of the nation’s capital city Honiara.

There are multiple ways that an island can foster convection. First and foremost, during the day, the island heats up faster than the surrounding ocean. Hot air rising, of course, is at the heart of deep moist convection. But this is enhanced by two additional factors: coastal breezes and orographic lift. Coastal breezes are a frequent byproduct of daytime island heating. Intro to Meteorology textbooks and popular media often simplify this as the air over the land gets warm causing it to rise while cooler ocean-based-air rushes in to fill the gap. The truth is more complex but more interesting; if you really want to know more it’s at the end of this post (be warned, though, it’s pretty technical). For now, though, we’ll note that as the island heats up it causes an offshore breeze to form. For a large continental landmass, this breeze can penetrate inland dozens of kilometers. For a small island, the breeze can’t penetrate that far because it will soon collide with the breeze coming from the other side of the island. This creates a low level convergence zone which enhances the upward lift caused by the daytime heating. Finally, the terrain of the island also has an impact. Many of these tropical islands are quite rugged, which means that as the wind rushes onshore it undergoes orographic lifting by the many mountains. This enhances the upward motion even more. This map of the terrain of Guadalcanal (courtesy of Wikipedia) shows that the island is dominated by an east-west ridge. Any flow from the north or the south is going to be enhanced by this vertical lift.

We’ll take a look at Guadalcanal on this day, starting with the true color view from the Himawari-9 Advanced Himawari Imager (AHI). Guadalcanal is the island at the center of this animation. The loop starts at 2130 UTC, which is 8:30 AM local time. The island starts clear, and is a beautiful deep green. For those of you who are mostly familiar with looking at GOES satellite imagery, you might think it is green because of the tropical vegetation that isn’t normally seen over the continental United States. However, it’s important to remember that the AHI has a true green channel unlike the US geostationary satellites and so its colors are going to appear a little more saturated. Shortly after the start of this loop localized convection forms across the island. However, if you watch the animation to the end, howver, you see that the convection starts to get squeezed into the middle of the island by the sea breezes on either side, forming a band of clouds that runs alongside the island’s mountainous spine.

It may be good to evaluate the kind of environment in which this convective development is taking place. Here’s a plot of the NOAA Physical Sciences Laboratory satellite-observed sea surface temperatures (SSTs). The Solomon Islands are circled in yellow, and it’s clear from this map that SSTs are a toasty 29-30 C (84-86 F). As such, there’s a lot of latent heat available to support convection.

Furthermore, the vertical structure of the atmosphere is also supportive of convection. Here is a skew-T plot of a nearby NUCAPS-retrieved vertical profile at 0338 UTC (2:38 PM local time), provided courtesy of the NUCAPS Savvy tool maintained by NASA SPoRT. The profile is generally well-mixed, with a near dry-adiabatic low level and a largely moist-adiabatic environment aloft. The instability is large, however, with CAPE values ranging from over 3200 to nearly 4300 J/kg depending on the values used. If low-level air parcels can ascend to around 1400-1600 m, they’ll breach the level of free convection and ascend on their own.

However, remember that terrain map above? That central mountainous spine has an elevation of around 1000-1200 m, so much of the work is already being done by orographic lift. The convergence and daytime heating doesn’t have to be very strong at all to unlock deep moist convection. Let’s see what happens over the next few hours. This loop is the same True Color RGB product as earlier, but now runs from 0100 to 0500 UTC (noon to 4:00 PM local) and covers the initiation of the deep convection.

Looking at the same period but through the Band 13 (10 micron) infrared channel helps to show how deep this convection was in a way that’s not possible with the true color alone. Here we see a burst of cold brightness temperatures as the convection ascends to the tropopause.

Additional RGB products can provide useful insight as well. The Day Convection RGB shows bright yellow plumes where the vertical growth is the most extensive while the red clouds are where ice particles can be found.

The Day Microphysics RGB also highlights some useful information. The salmon and golden areas indicate thick clouds with small water droplets (salmon) or ice particles (golden) which are commonly associated with deep convection. The green clouds are high-level ice clouds, which show that the cloud top is glaciating and thus the storm is transitioning rapidly to the mature stage.

Despite the strong forcing and favorable thermodynamics, this storm didn’t last very long. One possibility is that the deep convection to the south created an outflow boundary, which can be seen in the satellite loops as lines of clouds propagating northward. This cold, dense air would be kryptonite to any deep convection and cut it off just as it was getting going. At the same time, the day was getting late and the natural solar heating was dying out. Here’s the Band 13 view of the end of this event, from 0500 to 0900 UTC (4:00 PM to 8:00 PM local time). Sunset is around 0715 UTC (6:15 PM local time).

As promised, here is a more detailed description of coastal breeze formation. This post’s author has a special interest in coastal breeze formation (as depicted in this paper published in the Journal of Atmospheric Sciences) so he’s happy to take the time to offer a more precise explanation. As is well-known, the land heats up more readily than the water does due to differences in thermal heat capacity. This means that the air that is adjacent to the land also gets warmer than the air over the water. This creates a density gradient from land to sea, which in turn results in localized baroclinicity (we weren’t kidding when we said this was going to be a technical explanation) as the isobars remain largely static while the isopycnals tilt from parallel to the isobars before the heating to crossing them afterward. According to Kelvin’s Circulation Theorem, circulation in a baroclinic fluid is constant with time, but a barotropic fluid is going to cause a change in the circulation. Therefore, as the environment transitions from baroclinic to barotropic, it goes from zero circulation to a measurable one. That circulation is the coastal breeze.

View only this post Read Less

Altocumulus castellanus clouds (ACCAS) unexpectedly developed in southern Wisconsin yesterday morning, producing beautiful cloud features and eventually a severe thunderstorm. ACCAS form due to mid-level instability and manifest in towering billows that can often produce pileus clouds as they burst through a more stable layer.

The rooftop cameras of the UW-AOSS building caught this development nicely.

A model sounding from the High-Resolution Rapid Refresh (HRRR) did hint at mid-level instability due to an elevated mixed layer atop a temperature inversion. The model shows a mid-level lapse rate of 8.7 C/km and 155 J/kg of most-unstable CAPE (MUCAPE). However, the model did not forecast convection or precipitation.

There is some evidence that frontogenesis at 700mb may have served as a forcing mechanism for the convection, but it remains uncertain.

The NOAA/CIMSS LightningCast model caught this ACCAS development, and predicted elevated probabilities of lightning 20-45 minutes prior to first flashes (depending on position in the line).

Several cells produced 1″ hail in far southeastern Wisconsin.

The ProbSevere v3 models (PSv3) never exceeded 19% on these cells. However, an experimental image-based AI model produced > 50% probability of severe for several time steps (it predicts the probability of any severe hazards within the next 45 minutes and 20 km). The image-based approach may be able to better exploit spatial context and multi-modal signals from meteorological observations than the tree-based PSv3.

As a meteorologist, it is not always pleasant to be surprised by unexpected weather, but it is always a delight to behold beauty and wonder in the sky.

View only this post Read Less

1-minute Mesoscale Domain Sector GOES-19 (GOES-East) True Color RGB images from the CSPP GeoSphere site (above) highlighted notable smoke plumes from two large wildfires in southeast Georgia (the Pineland Road Fire near the Florida border, and the Brantley Highway 82 Fire farther to the northeast), in addition to another wildfire in northern Florida on 21 April 2026. Those particular wildfires were burning in parts of Georgia and Florida that were experiencing Exceptional Drought conditions.

1-minute GOES-19 GeoColor RGB images with an overlay of Next Generation Fire System (NGFS) Fire Detection polygons (below) provided a close-up view of the smoke plume and thermal signature associated with the Brantley Highway 82 wildfire in southeast Georgia — which was nearly under control at 700 acres in size during the early morning hours on 21 April, but then quickly grew out of control as surface winds abruptly shifted to the northeast then to the east (occasionally gusting to 22-23 mph), with the fast-moving wildfire burning an area of 5000 acres by the end of the day. Numerous evacuation orders were issued, about 87 homes were destroyed and schools throughout Brantley County were closed due to air quality concerns.

At 1758 UTC, the wildfire first exhibited a 3.9 µm shortwave infrared brightness temperature of 138ºC (below) — which is the saturation temperature of GOES-19 ABI Band 7 detectors. Also of note, at that time the Fire Radiative Power of the thermal signature was a rather high 6594 MW.

Periodic bursts of brighter-white pyrocumulus clouds rose above the hazy gray smoke plume, such as was seen just west of the fire’s thermal anomaly at 2019 UTC (below).

View only this post Read Less

5-minute CONUS Sector GOES-19 (GOES-East) imagery (above) revealed the brief development of what appeared to be a standing wave cloud — caused by a vertically-propagating gravity wave — that was anchored near the north coast of Saginaw Bay, Michigan on 20 April 2026. At first glance, this author was reminded of similar-appearing standing wave clouds that form along the Minnesota coast of Lake Superior.

A toggle between GOES-19 images and Topography (below) seemed to show that the northwestern edge of the cloud feature lined up with the final (subtle) drop in topography near the coast — however, since there was not the strong NW offshore surface wind flow common to the aforementioned Minnesota example, perhaps the onshore (and slightly upslope) lake breeze played a role in vertical gravity wave initiation? Until a sound explanation rooted in science is stumbled upon, this event will fall into the coveted “What the heck is this?” blog post category.

View only this post Read Less