Kalmaegi Moves Through the Philippines, 90W on the Horizon: AI Models Discussion 11-3-2025

The Atlantic and East Pacific look to be quiet in the next week or so as we start the final month of hurricane season, so we will keep our focus on the West Pacific, where a couple of systems near the Philippines have our focus.

Typhoon Kalmaegi

Typhoon Kalmaegi is currently moving through the Philippines as a Category 2. The inner core got more organized last night just before it approached the Philippines, but thankfully it hasn’t had quite enough time to rapidly intensify past Category 2 (85 knots). Some easterly shear helped keep the storm in check and slow its organization yesterday. Hurricane models (HAFS and HWRF) showed this shear slowing the intensification at times. In contrast, several runs of Google DeepMind (FNV3) were way too aggressive with development, with many members making the storm a Category-5 equivalent and the ensemble mean down in the 940s-950s for pressure across several cycles (Figure 1).

There have been many cases this year (including Melissa) where DeepMind/FNV3 was a top performer, and in the case of Melissa it was able to pick up on the extreme RI. However, there are clearly other cases (such as this) where the TC training dataset or synoptic forecast is not correct, in this case leading to excessive deepening. This will be something to look at as this model is improved and others are developed as well.

Going forward, models do show some intensification as Kalmaegi moves out into the South China Sea and moves west towards Vietnam under an Upper-level ridge (Figure 2).

While conditions do look generally favorable for Kalmaegi to intensify after passing through the Philippines, the persistent easterly upper-level winds (Figure 3) could continue to cause some shear and asymmetry in the core.

This uncertainty in just how symmetric the storm will be able to become with this persistent easterly shear is probably why some ensemble members are only showing modest deepening, while others are still intensifying the storm to a supertyphoon (Figure 4). The mean of the DeepMind ensemble has been fairly persistent in showing a strong typhoon (pressure in the 940s-950s) at landfall in Vietnam (Figure 1). It will be worth watching to see how intact the structure is after crossing through the Philippines, and how it responds to the environment when it is back over water.

Invest 90W

The other system of considerable interest is Invest 90W, located to the southeast of Guam. As Figure 4 shows, the Google DeepMind ensemble has been showing a considerable signal for this to turn into a powerful typhoon, with many members showing a supertyphoon. There has been some uncertainty as to whether the storm would recurve ahead of a trough to the north of it later this week (Figure 5), but recent runs of most models have trended towards a solution where it misses this first trough and gets trapped under a ridge that moves it towards Luzon.

In general, the environment looks favorable for intensification, perhaps rapid intensification. The storm will be embedded in upper level easterly flow under a ridge (Figure 6), so we’ll have to watch and see if there’s any easterly shear issues like with Kalmaegi. Overall, though, the environment looks extremely favorable for intensification, and dynamical hurricane models generally agree with DeepMind in showing rapid intensification into a supertyphoon.

This will definitely be a storm to watch very closely in the northern Philippines in the coming days. The ridge to the north may eventually break down and allow the storm to recurve to the NE, but at this point most models are suggesting that may not happen until after the storm gets close to or passes over Luzon (Figure 7).