Back to the trend of the day: A strong cyclone is about to hit India's east coast (in Hindi)
Poly Rating · from the trend of the day

Cyclone Arnab Landfall Time and Location: 10 Past Misses

Nargis was expected to hit Bangladesh or south-east India, but it turned east into Myanmar's Irrawaddy delta, where about 138,000 people died or went missing.

  • No. 1: Cyclone Nargis (2008)
  • 10 places

01Final list

The leading AIs each built their own list of ten independently. Here is one result that is easy to check: every place has a reason, a source, and the list of AIs that included it.

Order: from most to least AI agreement.

A bright badge means this AI put the entry on its list. A dim badge means it did not.

  1. 1

    Cyclone Nargis (2008)

    Forecast to hit Bangladesh or south-east India, it turned east into Myanmar's Irrawaddy delta; 84,537 confirmed dead plus 53,836 missing.

    Included it: Claude, Gemini, MiMo, StepFun.

    In their lists: Claude: 1 · Gemini: 2 · MiMo: 1 · StepFun: 1

    What each AI noted 4
    • Claudeplace 1

      Forecast to hit Bangladesh or south-east India, it turned east and struck the Irrawaddy delta, a coast that had no recent major-cyclone history.

      Source ↗
    • Geminiplace 2

      Defied climatological expectations by hooking eastward straight into the low-lying Ayeyarwady Delta with sudden explosive strengthening.

      Source ↗
    • MiMoplace 1

      For days the maps aimed it at India–Bangladesh; the late eastward turn into the Irrawaddy Delta killed 138,373 people — after India had warned Myanmar 48 hours earlier.

      Source ↗
    • StepFunplace 1

      138,000 people died in a storm whose forecast existed - the warning that reached the public never did.

      Source ↗
    4 AIs put it here
    Source ↗
  2. 2

    Cyclone Mocha (2023)

    IMD's landfall point near Sittwe barely moved, but its landfall wind forecast rose from 130 to 180–190 km/h; ECMWF expected a later landfall than happened.

    Included it: Claude, Gemini, MiMo, StepFun.

    In their lists: Claude: 3 · Gemini: 10 · MiMo: 5 · StepFun: 6

    What each AI noted 4
    • Claudeplace 3

      The landfall point barely moved between bulletins, but the forecast landfall wind went from 130 km/h to 180–190 km/h in four days.

      Source ↗
    • Geminiplace 10

      Intensified into an extreme Category 5 storm while models wavered until 48 hours prior on whether Myanmar or Bangladesh would take the direct eye strike.

      Source ↗
    • MiMoplace 5

      The Sittwe landfall was nailed; the ceiling was not — it jumped to a Category-5-equivalent 270 km/h (JTWC) against IMD's 215, and the dead are counted as 145 or 463 depending on who counts.

      Source ↗
    • StepFunplace 6

      One cyclone, two countries: the global models split between Sittwe and Cox's Bazar until the final day or two.

      Source ↗
    4 AIs put it here
    Source ↗
  3. 3

    Cyclone Amphan (2020)

    Rapid strengthening took it to 240 km/h, but IMD's landfall miss was small. Reports give 5.5, 15 or 17 km; the global models missed by 50–80 km.

    Included it: Gemini, MiMo, StepFun. Did not include: Claude.

    In their lists: Gemini: 8 · MiMo: 4 · StepFun: 4

    What each AI noted 3
    • Geminiplace 8

      Exploded from Category 1 to Category 5 Super Cyclone in just 24 hours, shattering regional model intensity trend projections.

      Source ↗
    • MiMoplace 4

      Only the Bay's second super cyclone on record after 1999: it doubled to 240 km/h in 12 hours, then an eyewall cycle and dry air tore it down on the way in — landfall winds are still quoted as both 175 and 190 km/h.

      Source ↗
    • StepFunplace 4

      The Bay of Bengal's first super cyclone in 21 years - a 15-17 km landfall miss, but an intensity ramp-up that arrived faster than the models expected.

      Source ↗
    3 AIs put it here
    Source ↗
  4. 4

    Cyclone Michaung (2023)

    Early headlines put landfall at Chennai. It crawled north along the coast for three days, crossed near Bapatla in Andhra Pradesh, and Chennai got 530 mm of rain.

    Included it: Claude, MiMo, StepFun. Did not include: Gemini.

    In their lists: Claude: 4 · MiMo: 6 · StepFun: 8

    What each AI noted 3
    • Claudeplace 4

      The headlines said Chennai on 1 December and Andhra Pradesh by 3 December, and the storm then crawled up the coast and came ashore near Bapatla.

      Source ↗
    • MiMoplace 6

      It crawled north along the coast for three days with the landfall marker sliding from Chennai to Bapatla — and Chennai's 530 mm of rain fell before the centre arrived.

      Source ↗
    • StepFunplace 8

      Chennai braced for a landfall that came ashore about 300 km up the coast at Bapatla - the city drowned in rain instead.

      Source ↗
    3 AIs put it here
    Source ↗
  5. 5

    1999 Odisha super cyclone

    IMD expected it to move inland toward Bangladesh. Instead it stalled over Odisha for two days at 260 km/h, and mainland warnings came only two days before landfall.

    Included it: Claude, MiMo. Did not include: Gemini, StepFun.

    In their lists: Claude: 2 · MiMo: 3

    What each AI noted 2
    • Claudeplace 2

      IMD forecast it to move inland and turn toward Bangladesh. It stalled over Jajpur for about two days instead.

      Source ↗
    • MiMoplace 3

      It reached a basin-record 912 hPa, parked over Jagatsinghpur for two days, and the government counted 9,887 dead — other estimates run to 30,000.

      Source ↗
    2 AIs put it here
    Source ↗
  6. 6

    Cyclone Asani (2022)

    It took an erratic path with a sudden westward jog toward Andhra Pradesh. It peaked at 100 km/h but crossed the coast only as a deep depression.

    Included it: Gemini, MiMo. Did not include: Claude, StepFun.

    In their lists: Gemini: 3 · MiMo: 8

    What each AI noted 2
    • Geminiplace 3

      Stalled less than 50 km off Andhra Pradesh, executing an intricate coastal loop before disintegrating without crossing land at severe strength.

      Source ↗
    • MiMoplace 8

      The storm that made weathermen admit defeat: a sudden westward jog, a retreat toward the Bay, then a weak crossing of the same Andhra coast Arnab now targets.

      Source ↗
    2 AIs put it here
    Source ↗
  7. 7

    Cyclone Titli (2018)

    On 7 October JTWC rated its chance of forming as low because models disagreed. Three days later it crossed Palasa at peak strength: IMD 150, JTWC 195 km/h.

    Included it: Claude, StepFun. Did not include: Gemini, MiMo.

    In their lists: Claude: 7 · StepFun: 5

    What each AI noted 2
    • Claudeplace 7

      On 7 October JTWC gave it a low chance of forming because models disagreed, and three days later it hit Palasa with a 12 km eye.

      Source ↗
    • StepFunplace 5

      The eye came ashore at Palasa in Andhra Pradesh - and 85 people died in the Odisha districts that got the water instead of the wind.

      Source ↗
    2 AIs put it here
    Source ↗
  8. 8

    Cyclone Mora (2017)

    Forecasters had to watch for a northward turn. It turned and crossed near Chittagong/Cox's Bazar at about 120–130 km/h.

    Included it: Gemini, StepFun. Did not include: Claude, MiMo.

    In their lists: Gemini: 9 · StepFun: 9

    What each AI noted 2
    • Geminiplace 9

      Developed rapidly in the central Bay and sprinted northeastward at over 28 km/h, cutting warning preparation windows in half.

      Source ↗
    • StepFunplace 9

      A storm that had to be watched for a northward turn: it recurved and crossed at Cox's Bazar instead of running on into Odisha.

      Source ↗
    2 AIs put it here
    Source ↗
  9. 9

    Cyclone Mahasen (2013)

    JTWC's first forecast was above 155 km/h; it reached Bangladesh at about 85 km/h on a more westerly track. Only one source gives this.

    Included it: Claude. Did not include: Gemini, MiMo, StepFun.

    In their lists: Claude: 5

    What each AI noted 1
    • Claudeplace 5

      JTWC's first forecast had winds above 155 km/h, but it hit Bangladesh with about 85 km/h on a more westerly track than expected.

      Source ↗
    1 AIs put it here
    Source ↗
  10. 10

    Cyclone Sidr (2007)

    IMD's cyclone alert of 14 November named Odisha and West Bengal; Sidr struck Bangladesh the next day at IMD 215 km/h. Only one source gives this.

    Included it: Claude. Did not include: Gemini, MiMo, StepFun.

    In their lists: Claude: 10

    What each AI noted 1
    • Claudeplace 10

      IMD issued a formation alert on 11 November and a cyclone alert for Odisha and West Bengal on 14 November, but the storm hit Bangladesh on 15 November.

      Source ↗
    1 AIs put it here
    Source ↗

More about the top three ↓

02More about the top three

The ranking is above. Here is more about the three places at the top: what they are, why they are here, and what is worth knowing.

1Cyclone Nargis (2008)

No. 1 — Cyclone Nargis (2008). On the lists of 4 of 4 AIs; median place there: 1.

Nargis 2008-05-02 0645Z
NASA · Public domain · Wikimedia Commons

Did you know

  • Nargis reached Category 4 on 2 May. It was the only storm in its basin to do so three years in a row, after Mala (2006) and Sidr and Gonu (2007).

    Source ↗

Wikipedia · Wikidata

Did you know

  • The Irrawaddy Delta was once known as the "rice bowl" of the British Empire. The UN Food and Agriculture Organization estimated that Nargis affected 65% of Myanmar's paddies.

    Source ↗
History

  • In May 2008, more than a week after Cyclone Nargis, only about one in 10 people in need had received any aid. The military government refused visas to more foreign aid workers, and it warned people who hoarded or sold relief goods that they would be prosecuted.

    Source ↗
Where to go further

  • Open this NASA satellite picture of heavy rain from Nargis. Look at where the rain fell, then think about how that fits the storm's turn east into Myanmar.

    Source ↗

Video: Myanmar man makes guitars from Cyclone Nargis wood - BBC News · BBC News

Compare the top three ↓

2Cyclone Mocha (2023)

No. 2 — Cyclone Mocha (2023). On the lists of 4 of 4 AIs; median place there: 5.5.

Mocha 2023-05-14 0050Z
AHI imagery from JMA's Himawari 9 Satellite · CC BY 4.0 · Wikimedia Commons
2023 JTWC 01B IR satellite imagery
Joint Typhoon Warning Center · Public domain · Wikimedia Commons

Did you know

  • Mocha is named after Mokha, the Yemeni port city that also gave its name to mocha coffee. Yemen supplied the name. A Yale Climate Connections article called it the North Indian Ocean's strongest storm on record.

    Source ↗

Wikipedia · Wikidata

Did you know

  • Leftover moisture from Mocha met cold air far inland and fell as snow in northwestern Yunnan, China, where it piled up on roads. Roughly 1.27 million people were evacuated from Cox's Bazar, Bangladesh, by 14 May.

    Source ↗
History

  • In May 2023, experts feared Mocha could be Myanmar's worst storm since Nargis in 2008, which killed about 100,000 people. IMD's later bulletin put its top winds at 210 to 220 km per hour, barely below super-cyclone level.

    Source ↗
Keep in mind

  • Warnings did not reach everyone equally. Al Jazeera found most evacuations ran through existing community networks and left out the Rohingya. The vast majority were left behind in bamboo camp shelters. The military's claim of 85,000 evacuated could not be verified.

    Source ↗

  • Refugees International said failing to evacuate Rohingya from the camps, or even letting them leave, endangered countless lives. It also said the junta had already confined more than 100,000 Rohingya to displacement camps before the storm.

    Source ↗
Where to go further

  • Read this news story about the landfall near Sittwe. Start with the timing and the wind numbers around landfall, then compare them with the forecast numbers in the ranking.

    Source ↗

Video: Cyclone Mocha: How Are Cyclones Named? | When Will Cyclone Mocha Make Landfall? · moneycontrol

Compare the top three ↓

The top three side by side

Cyclone Nargis (2008)Cyclone Mocha (2023)Cyclone Amphan (2020)
In the AIs' lists4 / 44 / 43 / 4
Put first byClaude, MiMo, StepFun——

Places with the same number of AIs are ordered by their average place in the AIs' own lists.

More ratings

03Worth knowing

What the AIs noticed while building the list.

IMD's 24-hour landfall miss has shrunk: 157.5 km before modernisation, then 66.5 km, and an average of 19 km in 2021–25. A 2014 study quotes about 75 km; the figures cover different periods.

Noticed by:
Included it: MiMo, StepFun. Did not include: Claude, Gemini.
Source ↗

Storms are now missed less on where they hit than on how strong they get: Mocha's landfall point held while its wind forecast was raised again and again.

Noticed by:
Included it: Claude, StepFun. Did not include: Gemini, MiMo.
Source ↗

Two days before Dana (2024), GFS and ECMWF still disagreed on landfall time and place, and only about half of the Bay's monitoring buoys were working.

Noticed by:
Included it: MiMo. Did not include: Claude, Gemini, StepFun.
Source ↗

IMD reports 3-minute average winds and JTWC 1-minute averages, so the same storm gets different peaks: Titli 150 vs 195 km/h, Mocha 215 vs 270 km/h.

Noticed by:
Included it: Claude, Gemini, MiMo. Did not include: StepFun.
Source ↗

Arnab's forecast zone narrowed from Bapatla–Visakhapatnam to near Yanam. Its sustained winds are quoted as 115–125 km/h (ANI) and as 125–135 km/h (IMD bulletin).

Noticed by:
Included it: Claude, MiMo, StepFun. Did not include: Gemini.
Source ↗

04Questions people ask

Where will cyclone make landfall?

IMD expects Arnab to cross the Andhra Pradesh–Yanam coast near Yanam early on 28 October 2026, with the crossing window given as 2–4 am. The IMD bulletin gives sustained winds of 125–135 km/h, gusting to 150.

Source ↗
Where will hurricane make landfall?

A forecast landfall is a band, not a point. IMD's average 24-hour landfall miss was 19 km in 2021–25 and 77 km at 72 hours.

Source ↗
Where did cyclone make landfall?

Montha crossed Andhra Pradesh around midnight on 28–29 October 2025. Sources place it near Narasapuram or near Kakinada, which is why the dates and places differ.

Source ↗
Where will hurricane make landfall in Jamaica?

The most recent one, Hurricane Melissa, made landfall as a Category 5 near New Hope, Jamaica, on 28 October 2025 with 295 km/h winds.

Source ↗

05How we counted

In plain words, no formulas.

1The question

Top 10 Bay of Bengal Cyclones Whose Landfall Was Hardest to Forecast

2Why this topic

It sets the official forecast, two global models and the observed track side by side, so readers see a forecast landfall as a range, not a point, and understand why Arnab's warnings cover such a long stretch of the Andhra coast.

3Where it applies

Bay of Bengal cyclones that crossed the coasts of India, Bangladesh and Myanmar, from the satellite era to Arnab (2026)

4What we looked at

  • Distance between the landfall point forecast 24 hours ahead and the real landfall point
  • How far IMD's official forecast and the ECMWF and GFS global model runs were from each other
  • Error in the forecast peak wind speed, including rapid strengthening that was not foreseen
  • Whether the final warnings still gave people enough time to prepare
  • IMD / RSMC New Delhi, landfall forecast verification — How far, in km, and how late or early, in hours, IMD's official landfall forecast was for each North Indian Ocean cyclone, 2003–2025
  • IMD / RSMC New Delhi, Annual Verification Report 2024 — Average track forecast errors at 24, 48 and 72 h, and intensity errors, compared with the previous five-year average
  • PIB, answer to a Parliament question (2025) — IMD's average landfall-point error for 2021–25: 19.0, 34.4 and 77.3 km at 24, 48 and 72 h
  • NCMRWF, global-model verification for Amphan — Track and landfall errors of the global models, kept separate from IMD's official forecast

5Who ranked

The leading AIs, each building its own list of ten independently and naming a source for every place: Claude, Gemini, MiMo, StepFun.

6How the places were set

The same entry written differently by different AIs was counted as one. The more AIs put an entry on their list, and the higher they put it, the higher its place. "AI agreement" shows how many of the AIs had it on their own list. Sources are only those the AIs found themselves; nothing was added by hand.

06Each AI's own list

🇺🇸 Claude10 places · 2 notes

The same question, answered by Claude alone. The final list above combines all the AIs.

  1. 1
    Cyclone Nargis 2008 (Myanmar)

    Forecast to hit Bangladesh or south-east India, it turned east and struck the Irrawaddy delta, a coast that had no recent major-cyclone history.

    Source ↗
  2. 2
    1999 Odisha super cyclone

    IMD forecast it to move inland and turn toward Bangladesh. It stalled over Jajpur for about two days instead.

    Source ↗
  3. 3
    Cyclone Mocha 2023 (Myanmar/Bangladesh)

    The landfall point barely moved between bulletins, but the forecast landfall wind went from 130 km/h to 180–190 km/h in four days.

    Source ↗
  4. 4
    Cyclone Michaung 2023 (Andhra Pradesh)

    The headlines said Chennai on 1 December and Andhra Pradesh by 3 December, and the storm then crawled up the coast and came ashore near Bapatla.

    Source ↗
  5. 5
    Cyclone Mahasen 2013 (Bangladesh/Myanmar)

    JTWC's first forecast had winds above 155 km/h, but it hit Bangladesh with about 85 km/h on a more westerly track than expected.

    Source ↗
  6. 6
    Cyclone Hudhud 2014 (Visakhapatnam)

    It rapidly intensified on 11 October while stalled south-east of Visakhapatnam, and IMD's operational 960 hPa was later revised to 950 hPa.

    Source ↗
  7. 7
    Cyclone Titli 2018 (Andhra Pradesh/Odisha)

    On 7 October JTWC gave it a low chance of forming because models disagreed, and three days later it hit Palasa with a 12 km eye.

    Source ↗
  8. 8
    Cyclone Montha 2025 (Andhra Pradesh)

    One headline said the forecasts 'go south, so does landfall', and it came ashore near Narasapuram with only 95 km/h.

    Source ↗
  9. 9
    Cyclone Bulbul 2019 (West Bengal/Bangladesh)

    It peaked at 140 km/h on 8 November and weakened to a severe cyclonic storm just before landfall near Sagar Island.

    Source ↗
  10. 10
    Cyclone Sidr 2007 (Bangladesh)

    IMD issued a formation alert on 11 November and a cyclone alert for Odisha and West Bengal on 14 November, but the storm hit Bangladesh on 15 November.

    Source ↗

What Claude noticed

Mocha's forecast landfall point stayed put while its intensity forecast was raised repeatedly. Landfall location was well forecast, but the storm's strength was not.

Source ↗

For Arnab, the same landfall zone (Bapatla to Visakhapatnam, near Yanam) is forecast on 27 October. A miss of tens of km would matter because rainfall is expected over 16 districts.

Source ↗
🇺🇸 Gemini10 places · 2 notes

The same question, answered by Gemini alone. The final list above combines all the AIs.

  1. 1
    Cyclone Madi (2013)

    Executed an unprecedented 180-degree hairpin U-turn in the open sea, completely reversing from a northward track to a southwestward demise.

    Source ↗
  2. 2
    Cyclone Nargis (2008)

    Defied climatological expectations by hooking eastward straight into the low-lying Ayeyarwady Delta with sudden explosive strengthening.

    Source ↗
  3. 3
    Cyclone Asani (2022)

    Stalled less than 50 km off Andhra Pradesh, executing an intricate coastal loop before disintegrating without crossing land at severe strength.

    Source ↗
  4. 4
    Cyclone Arnab (2026)

    Underwent sudden intensification into a Very Severe Cyclonic Storm near Yanam while global models split sharply between Andhra and Odisha paths.

  5. 5
    Cyclone Jawad (2021)

    Prompted multi-state coastal evacuation alerts before shearing apart over the sea, never making an official severe landfall.

    Source ↗
  6. 6
    Cyclone Jal (2010)

    Maintained severe intensity across the central basin before unexpected high vertical wind shear tore its core apart hours before reaching Chennai.

    Source ↗
  7. 7
    Cyclone Roanu (2016)

    Skirted the entire eastern coastline of India for 1,500 km like a coastal train before accelerating violently into Chittagong.

    Source ↗
  8. 8
    Cyclone Amphan (2020)

    Exploded from Category 1 to Category 5 Super Cyclone in just 24 hours, shattering regional model intensity trend projections.

    Source ↗
  9. 9
    Cyclone Mora (2017)

    Developed rapidly in the central Bay and sprinted northeastward at over 28 km/h, cutting warning preparation windows in half.

    Source ↗
  10. 10
    Cyclone Mocha (2023)

    Intensified into an extreme Category 5 storm while models wavered until 48 hours prior on whether Myanmar or Bangladesh would take the direct eye strike.

    Source ↗

What Gemini noticed

Recurving and looping cyclones generate nearly double the 24-hour track errors (over 100-150 km) compared to straight west-northwest tracking storms in the Bay of Bengal.

Source ↗

Numerical Weather Prediction (NWP) models frequently diverge by over 400 km when a tropical cyclone decelerates below 5 km/h because subtle mid-latitude troughs dominate steering currents.

Source ↗
🇨🇳 MiMo10 places · 4 notes

The same question, answered by MiMo alone. The final list above combines all the AIs.

  1. 1
    Cyclone Nargis 2008

    For days the maps aimed it at India–Bangladesh; the late eastward turn into the Irrawaddy Delta killed 138,373 people — after India had warned Myanmar 48 hours earlier.

    Source ↗
  2. 2
    Cyclone Ockhi 2017

    Born unforecast beside Sri Lanka, it doubled its winds in two days while the projected crossing jumped from Tamil Nadu to Gujarat — 318 dead and 141 missing, most of them fishermen.

    Source ↗
  3. 3
    1999 Odisha super cyclone

    It reached a basin-record 912 hPa, parked over Jagatsinghpur for two days, and the government counted 9,887 dead — other estimates run to 30,000.

    Source ↗
  4. 4
    Cyclone Amphan 2020

    Only the Bay's second super cyclone on record after 1999: it doubled to 240 km/h in 12 hours, then an eyewall cycle and dry air tore it down on the way in — landfall winds are still quoted as both 175 and 190 km/h.

    Source ↗
  5. 5
    Cyclone Mocha 2023

    The Sittwe landfall was nailed; the ceiling was not — it jumped to a Category-5-equivalent 270 km/h (JTWC) against IMD's 215, and the dead are counted as 145 or 463 depending on who counts.

    Source ↗
  6. 6
    Cyclone Michaung 2023

    It crawled north along the coast for three days with the landfall marker sliding from Chennai to Bapatla — and Chennai's 530 mm of rain fell before the centre arrived.

    Source ↗
  7. 7
    Cyclone Gaja 2018

    It looped once in the Bay, crossed at Vedaranyam on an unusual west-southwest line at night — and NDMA's own study says rapid wind intensification made prediction difficult.

    Source ↗
  8. 8
    Cyclone Asani 2022

    The storm that made weathermen admit defeat: a sudden westward jog, a retreat toward the Bay, then a weak crossing of the same Andhra coast Arnab now targets.

    Source ↗
  9. 9
    Cyclone Thane 2011

    Winds nearly doubled in 48 hours offshore on Christmas week, and even at peak the agencies disagreed — IMD 140 km/h, JTWC 165 — before it broke over Cuddalore–Puducherry.

    Source ↗
  10. 10
    Cyclone Dana 2024

    Two days out, GFS and ECMWF runs still disagreed on both time and place — with only about half the Bay's monitoring buoys working.

    Source ↗

What MiMo noticed

IMD's average landfall-point error is now about 75 km at 24 h and 98 km at 48 h — yet for Fani (2019) it was just 11, 11 and 15 km at 24, 48 and 72 h, which is why a 215 km/h crossing killed 81 people instead of thousands.

Source ↗

Hard-to-forecast storms fail in three distinct ways — sharp turns (Nargis), sudden strengthening (Ockhi, Amphan, Mocha) and slow crawls along the coast (Michaung, Asani) — and no single error score captures all three.

Source ↗

Two days before a crossing, the world's best models can still disagree on where the line ends: for Dana (2024) — and again now for Arnab, 170 km from Visakhapatnam — GFS and ECMWF runs pointed different ways while only half the Bay's buoys were working.

Source ↗

Accuracy and catastrophe can coexist: WMO credited 'accurate forecasts' for limiting Mocha's casualties even as the death toll is recorded as 145 by ASEAN and 463 by Myanmar's NUG.

Source ↗
🇨🇳 StepFun10 places · 4 notes

The same question, answered by StepFun alone. The final list above combines all the AIs.

  1. 1
    Cyclone Nargis (2008, Myanmar)

    138,000 people died in a storm whose forecast existed - the warning that reached the public never did.

    Source ↗
  2. 2
    Cyclone Ockhi (2017)

    About 245 fishermen died at sea because the alert that went out was a fishing advisory, not a cyclone warning.

    Source ↗
  3. 3
    Cyclone Fani (2019)

    Model scenarios split between Tamil Nadu and Odisha for days, then the storm rapidly intensified - and IMD's final 24-hour landfall error was 11 km.

    Source ↗
  4. 4
    Cyclone Amphan (2020)

    The Bay of Bengal's first super cyclone in 21 years - a 15-17 km landfall miss, but an intensity ramp-up that arrived faster than the models expected.

    Source ↗
  5. 5
    Cyclone Titli (2018)

    The eye came ashore at Palasa in Andhra Pradesh - and 85 people died in the Odisha districts that got the water instead of the wind.

    Source ↗
  6. 6
    Cyclone Mocha (2023, Myanmar/Bangladesh)

    One cyclone, two countries: the global models split between Sittwe and Cox's Bazar until the final day or two.

    Source ↗
  7. 7
    Cyclone Yaas (2021)

    Its forecast landfall swung across the Odisha-West Bengal border - and both states moved 2.15 million people anyway.

    Source ↗
  8. 8
    Cyclone Michaung (2023)

    Chennai braced for a landfall that came ashore about 300 km up the coast at Bapatla - the city drowned in rain instead.

    Source ↗
  9. 9
    Cyclone Mora (2017, Bangladesh)

    A storm that had to be watched for a northward turn: it recurved and crossed at Cox's Bazar instead of running on into Odisha.

    Source ↗
  10. 10
    Cyclone Arnab (2026)

    Three days of forecasts pointing anywhere between Bapatla and Visakhapatnam, then a point forecast at Yanam - and the intensity was revised upward the whole way in.

    Source ↗

What StepFun noticed

IMD's own 24-hour landfall point error is now about 19 km on the 2021-25 average - smaller than the eye of most of these storms. The residual forecasting problem is no longer 'where' but 'how strong, how fast, and how much rain'.

Source ↗

Rapid intensification is the recurring failure mode: Ockhi produced no cyclone warning at all, Fani reached 215 km/h, Amphan became the Bay's first super cyclone in 21 years, and Titli was still intensifying within a day of landfall.

Source ↗

When the global models disagree, the honest forecast is a band rather than a point: for Yaas every model but bHWRF said north Odisha, and for Mocha the centres split between Sittwe and Cox's Bazar right up to landfall.

Source ↗

The deadliest Bay storms are not the ones with the biggest track error. Nargis killed about 138,000 with forecasts in hand, and Titli killed 85 after a landfall point that IMD got roughly right - the miss was in what people were told to do about the water, not the wind.

Source ↗