Atlantic Finally Showing Signs of Life? AI Models Discussion 8-24-2026

  • Andy Hazelton
    Andy Hazelton

Share

Atlantic Finally Showing Signs of Life? AI Models Discussion 8-24-2026

Today’s AI models blog will focus on the Atlantic, where a wave has just emerged off the coast of Africa and has been tagged as Invest 96L by the National Hurricane Center. The wave has a 60% chance of development over the next 7 days.

There are some interesting discrepancies in the model forecasts for 96L. Dynamical models are generally fairly aggressive with development, but AI based models less so for now, at least in the short term. For example, GFS develops 96L into a hurricane and recurves it well north of the Caribbean (Figure 1), while AIGFS keeps the system weaker and brings it into the Caribbean Islands (Figure 2). The environment looks similar between the two models, with shear from a TUTT near the Caribbean. The difference seems to be mostly in the track and early intensification: GFS develops quickly and gains latitude, while AIGFS is slow to develop and stays on a more westerly heading. There are similar differences between the ECWMF and AIFS, although the operational ECMWF is not as aggressive as GFS.

Figure 1: Operational GFS forecast of 850-hPa vorticity, 500-hPa height, and 200-hPa wind initialized at 12 UTC August 24, 2026 and valid at 12 UTC August 31, 2026. Figure from www.tropicaltidbits.com.
Figure 2: AIGFS forecast of 850-hPa vorticity, 500-hPa height, and 200-hPa wind initialized at 12 UTC August 24, 2026 and valid at 12 UTC August 31, 2026. Figure from www.tropicaltidbits.com.

Taking a look at some other AI models, most of them generally seem to agree with the AIFS and AIGFS. Both FourCastNet-v3 (Figure 3) and Panguweather (Figure 4) show a TD or strong tropical wave moving through the Virgin Islands in 6-7 days.

Figure 3: FourCastNet-v3 forecast of 10-meter wind and MSLP initialized at 12 UTC August 24, 2026 and valid at 12 UTC August 30, 2026.
Figure 4: PanguWeather forecast of 10-meter wind and MSLP initialized at 12 UTC August 24, 2026 and valid at 12 UTC August 30, 2026.

None of the models show any significant intensification in the MDR or near the islands. The MDR is likely too dry/stable, and near the islands a TUTT (tropical upper tropospheric trough) will likely bring some southwesterly shear over the system (Figure 5). The westerly shear will likely be less of an issue if it lifts north earlier like GFS shows.

Figure 5: AIFS forecast of 250-hPa wind initialized at 12 UTC August 24, 2026 and valid at 12 UTC August 30, 2026.

Taking a look at some ensemble forecasts will also help us to assess the possible forecast outcomes for 96L. The AIFS ensemble (Figure 6) shows most members tracking just north of the islands, with a large spread in possible tracks after.

Figure 6: EC AIFS ensemble initialized at 12 UTC August 24, 2026 and valid out to 240 hours. Data from www.weathernerds.org.

Google DeepMind’s 50-member ensemble (Figure 7) shows a few members that develop early, but more showing development and intensification later when the system reaches the Southwest Atlantic. This seems plausible if the system is able to survive the marginal MDR and hostile shear near the Caribbean.

Figure 7: Google DeepMind 50-member ensemble forecasts of TC tracks and intensity over the Atlantic initialized at 06 UTC August 20, 2026, valid out to 360h. Data from Google.

The 1000-member ensemble from Google DeepMind offers some interesting insights. One cluster of the ensemble (Figure 8) shows early development in the next 2-3 days, with most members recurving in the vicinity of Bermuda.

Figure 8: Google DeepMind 1000-member ensemble tracks for one cluster for Invest 96L, initialized at 12 UTC August 24, 2026 and valid out to 360h. Figure from www.tcatlas.org.

Another cluster (Figure 9) has TC genesis much later after the system moves past the TUTT and reaches the vicinity of the Bahamas. Many of these members also recurve in the vicinity of Bermuda, but some show a potential land threat due to the later development.

Figure 9: Google DeepMind 1000-member ensemble tracks for a second cluster for Invest 96L, initialized at 12 UTC August 24, 2026 and valid out to 360h. Figure from www.tcatlas.org.

Both clusters show a large range in intensity for 96L, which makes sense. There are definitely scenarios where it could find a favorable environment in the Southwest Atlantic and intensify. But there are also scenarios where a combination of dry air, shear, and land interaction cause the system to dissipate before intensifying much or even developing into a TC.

The general steering flow suggests this hopefully won’t be a threat to the Continental United States, but it’ll be worth watching, particularly if it waits to develop and doesn’t gain much latitude over the next few days.

We’ll keep an eye on things and be back with another update soon!