● tracked Akvanfix 19.6N 84.9E · 20 Nov 11:30 IST60 km/h nowCyclonic Stormpeak 60 km/hWatchIMD bulletin 39page 20 Nov 17:52 IST● tracked Akvanfix 19.6N 84.9E · 20 Nov 11:30 IST60 km/h nowCyclonic Stormpeak 60 km/hWatchIMD bulletin 39page 20 Nov 17:52 IST
Storm tracked tracking stopped

Akvan

Akvan came ashore close to Gopalpur, Ganjam, between 07:30 and 08:30 IST on 20 Nov.

18 Nov 03:48 – 20 Nov 17:54IST, how long we followed it
173-hourly watches
bulletin 39IMD, updated 20 Nov 17:52 IST
IMD fix 19.6N 84.9E · 20 Nov 11:30 IST · 60 km/h
1 · What to expect

Wind and storm surge are the main dangers on the coast.

rivers, hillsfloods and landslides after landfall
2 · Odisha, on the ground

Akvan crossed the coast close to Gopalpur between 07:30 and 08:30 IST. The heaviest rain fell in Ganjam: 21 cm at Patrapur.

21 cmPatrapur (Ganjam), 24 h
21 cmHinjili (Ganjam), 24 h
19stations with 10 cm or more
15districts on orange alert, next 24 h

From IMD's Bhubaneswar centre, which covers Odisha the way AWARE covers Andhra Pradesh. Rain gauges measure far more than the satellite estimates: the satellite gives Ganjam about 5 cm on average.

3 · How the forecast moved

IMD's forecast landfall moved 550 km up the coast, from Machilipatnam to Gopalpur, before Akvan arrived.

16 Novfirst forecast landfall, near Machilipatnam
07:30–08:30IST on 20 Nov, when it crossed close to Gopalpur
~3 hearlier than the last day of forecasts said
70–80km/h at the coast; forecasts before 18 Nov said 107–130

Every IMD bulletin carries a forecast track. Followed one by one, they show the landfall moving north in two jumps on 17 and 18 Nov, then holding on Gopalpur for two days. The next screen asks whether two AI models, given only satellite pictures, could have called it sooner.

4 · Where the AI models said it would land, vs IMD

Reading satellite pictures alone, Inkling 2 Small named Gopalpur at 17 Nov 11:30, when IMD still said Kakinada. It did not stick with it.

204 / 318 / 396 kmaverage miss before 18 Nov 14:30: Inkling / Sonnet / IMD
16–38 kmIMD's miss once it settled; models 100–570 km
~1.2 daystoo early, typically, for the models; IMD 2–6 h late
95–140km/h Inkling expected; it came in at 70–80

At ten moments from 16 to 19 Nov, both models saw only satellite pictures, with no IMD position or forecast. Early on they sometimes beat IMD, but each good call was followed by a jump back south, they expected landfall a day or more too soon, and once IMD settled it was far better than either.

5 · Seen from space

Akvan's cold cloud mass at 11:30Z sits just north of the Odisha coast near 20.9N 86.6E.

4%of Palnadu under storm cloud
26 mmmost rain on any AP district in 24 h, by satellite
65 kmInkling 2 Small's storm-centre error, latest
154 kmClaude Sonnet 5.5's storm-centre error, latest

Every 3 hours, Claude Sonnet 5.5 and Inkling 2 Small read the same new satellite pictures, with IMD's position held back, and each writes a note. Inkling 2 Small places the storm's centre more precisely: on average 94 km from IMD's position, against 152 km for Sonnet, over 13 readings. Both read clouds and rain well; IMD's position is still the one to use.

6 · Who is in the way

No district is forecast to get gale-force wind.

A gale is wind of 62 km/h or more. Nobody can evacuate, drive a convoy or launch a boat in it, so everything has to happen before it arrives.

7 · What to do, and when

No evacuation is needed on IMD's current track; keep crews ready for power cuts.

0people the hedged plan moves to shelters
0 → 0expected deaths: doing nothing → hedged plan (wind and surge only)
0power crew and rescue team moves
21 Novnow
people in shelters (from the evacuation order until they head home)gale (62+ km/h): no evacuation, convoys or boats

Blue bars are people in shelters; orange hatching is the gale. Evacuations finish before each district's gale begins.

8 · The AI agents, as it happened

4 AI models ran the response live, every 6 hours, until tracking stopped. DeepSeek V4.1 Flash finished closest to the best plan, for now.

2.09harm locked in, DeepSeek V4.1 Flash (5 slots)
2.26harm locked in, Inkling 2 Small (1M) (5 slots)
4.67harm locked in, GLM 5.3 Flash (5 slots)
5.99harm locked in, Claude Sonnet 5.5 (12 slots)

Harm locked in: how much worse off the coast is because of an agent's orders so far, in life-equivalents (one equals one death). Lower is better; 0 means nothing lost yet. Sonnet evacuates GJM. DeepSeek, GLM and Inkling joined at 17:30 IST on 19 Nov, so they have fewer slots in which to lose ground; the rule-based policies are in the detail.

9 · How long we tracked Akvan

We followed Akvan from 18 Nov 03:48 to 20 Nov 17:54 IST, a few hours after it came ashore.

17watch runs, every 3 hours: bulletin, plan, scores
1–39IMD bulletins read, 15 Nov 05:30 to 20 Nov 11:30 IST
126-hour decision slots for Sonnet and the policies; 5 for the three newer models
13satellite readings, 19 Nov 14:07 to 20 Nov 17:52 IST

Akvan had crossed the coast and weakened into a deep depression over south Odisha, so the 3-hourly watch was stopped on 20 Nov. The final grading, which needs ERA5 rain about 5 days after the storm, has not been run, so every score on this page is provisional.

Remember

This is decision support from a model built on public data, not an official advisory. Follow IMD and the SDMA.

Tracking stopped at 20 Nov 19:15 IST; the page is no longer rebuilt every 3 hours. Last built 20 Nov 2026 13:46 UTC.

Eval boards

Who plansbest?

We give four AI models and simple rules the same job: run the response to two real cyclones. Then we replay their plans on what really happened.

1 · Knowing the whole storm in advance

With hindsight, the best AI models, DeepSeek V4.1 Flash and Inkling 2 Small, both score about 0.46 out of 1.

0.46DeepSeek V4.1 Flash
0.46Inkling 2 Small (1M)
0.30Claude Sonnet 5.5
0.25GLM 5.3 Flash

Averages over the 18 tasks every agent ran. The best possible plan scores 1.00 and the standard procedure 0.21: the score is the share of avoidable harm a plan avoids, so 0.21 means no better than the rulebook.

2 · Working from forecasts, as it happens

Working only from IMD's forecasts, the best AI model, Claude Sonnet 5.5, scores 0.57 and comes second, behind “hedged forecast policy” (0.68).

0.68Hedged forecast policy
0.57Claude Sonnet 5.5
0.55Standard procedure on forecasts
0.54DeepSeek V4.1 Flash

Every 6 hours an agent sees only the bulletins issued so far, gives its orders, and the real weather plays out. The rule that does well plans for IMD's usual forecast errors instead of trusting the latest track. Averages over the 6 tasks every agent ran.

3 · Where the models lose points, per task over 18 tasks

All four keep deaths at home near the best plan, then lose lives to floods: rescue teams wait in the wrong places.

Best planDeepSeek V4.1 FlashInkling 2 SmallClaude Sonnet 5.5GLM 5.3 Flash
Deaths at home9590959693
Deaths of people stranded by floods1623273641
People evacuated78,43376,30679,85876,366112,750
Still in shelters at the end0005,9911,467

DeepSeek and Inkling lose the fewest people to floods (23 and 27 a task; the best plan, 16). GLM moves 44% more people than needed, doubling shelter days. Left in shelters at the end: GLM about 1,500, Sonnet about 6,000. Use a model to read and draft; leave where boats and crews go to an optimiser and to officers.

Red: more than 15% worse than the best plan. A plan that broke a rule is left out (DeepSeek, one task).

About

Reading thenumbers

CycloneSim asks one question of every plan: how much harm did it cause, to people's lives and to their homes and daily life?

In twenty seconds

Two storms that happened, and one on its way.

1 · One unit for every harm

Every kind of harm is counted in one unit: the life-equivalent.

1 death= 1 life-equivalent
12,500days spent in a shelter = 1
62,500household-days without power = 1

Deaths count most, on purpose. But moving a whole district “just in case” is not free, and neither are days in a shelter after the storm has gone.

2 · One score

The score asks: how much of the harm that could have been avoided, did a plan avoid?

1.0the best possible plan
0.49avoids three quarters
0.21no better than the rulebook
< 0.2broke a rule
3 · Expected deaths

Expected deaths are an average from the model, not a forecast of the toll.

100,000people each facing a 1-in-10,000 risk
= 10expected deaths

People die in the model when they stay home in strong wind or storm surge, or when flood water cuts them off and no rescue team reaches them. People in shelters are safe in the model.

4 · How it is built

Built from IMD's bulletins, satellite rain and public maps, and checked against the exact best plan.

26coastal districts, AP and south Odisha
3 htime step
2 + 1cyclones replayed, and one tracked live