Cyclone Arnab air quality: haze built up on the Andhra coast, then the storm cleaned it out

The air on the Andhra Pradesh coast was at its worst since at least mid-September in the days before Cyclone Arnab, and the storm then cleaned it out. At the GVM Corporation monitor in Visakhapatnam, PM2.5 averaged 98.9 µg/m³ on 24 October 2026, its highest daily mean since our archive began on 13 September; CPCB's bulletin put the city in the Poor band, at an AQI of 254, the next afternoon. By 2:30 pm on 27 October the same monitor read 4 µg/m³, and an hour later it went silent.

Arnab crossed the Andhra Pradesh and Yanam coast between 2 am and 8 am on 28 October as a very severe cyclonic storm, about 40 km south-west of Yanam and 60 km south-west of Kakinada, with sustained winds of 125 to 135 km/h gusting to 150 km/h (IMD press release, 28 October, 1:45 pm). Our data from the coast's monitors shows what happened to the air. Why the haze built up first is less clear.

Key findings

• The coast hazed up, inland did not. Visakhapatnam rose from 6.1 µg/m³ on 16 October to 98.9 on 24 October, against a 13 September to 14 October mean of 40.6. Rajamahendravaram rose from 6.8 to 90.1 (mean 28.3). Vijayawada stayed between about 31 and 44.
• The build-up started before the storm existed. Visakhapatnam passed 40 µg/m³ on 19 October. IMD first reported the low-pressure area that became Arnab on 23 October.
• The clean-out began before the heavy rain. Visakhapatnam's daily PM2.5 halved from 25 to 26 October; the fall to single digits came with the rain on 27 October.
• Odisha cleared too. Bhubaneswar, Cuttack and Paradeep fell below 10 µg/m³ in the hours before landfall.
• Monitors went dark before landfall. Visakhapatnam, Rajamahendravaram and Kakinada all stopped reporting by 12:30 am on 28 October.

Cyclone Arnab: what happened, according to IMD

In the 24 hours to 8:30 am on 28 October, IMD recorded extremely heavy rain (21 cm or more) over coastal Andhra Pradesh and Yanam and over Odisha. Garividi in Vizianagaram district had 26 cm, and Gajapati in Odisha 25 cm (IMD, 28 October, 1:45 pm). Reuters, reported by Open, said more than 180,000 people were evacuated across Andhra Pradesh and Odisha, and power lines were down in parts of Kakinada (Open, 28 October).

How bad did the air get before Cyclone Arnab?

The table compares each coastal monitor's usual level with its worst day before the storm and its last readings before landfall. PM2.5 is in µg/m³; days run 5:30 am to 5:30 am India time.

Visakhapatnam's 24 and 25 October were its two worst days in our archive, and Rajamahendravaram's 25 October its worst. The hourly peak came at 7:30 am on 25 October in Visakhapatnam (143 µg/m³) and at 5:30 pm the same day in Rajamahendravaram (128).

CPCB's official bulletin (India's NAQI, 24 hours to 4 pm) agrees. Visakhapatnam went from Good (44) on 19 October to Moderate (177) on 24 October, Poor (254) on 25 October, Moderate (147) on 26 October and Satisfactory (72) on 27 October. Rajamahendravaram went from Good (42) on 19 October to Poor (209) on 26 October, then Satisfactory (96) on 27 October. On 28 October Visakhapatnam, Rajamahendravaram, Kakinada and Eluru were all listed as absent "due to insufficient data" (CPCB, 19 October; 24 October; 25 October; 26 October; 27 October; 28 October).

Treat Kakinada's numbers as rough. Its hourly readings jump between about 12 and 178 µg/m³ within hours, and our quality checks held back several hours on 24 to 26 October as an erratic pattern.

Did the air get worse inland?

No. Away from the coast the numbers barely moved. Comparing 24 October with 27 October:

• Vijayawada (Rajiv Gandhi Park and Rajiv Nagar): 31.1 to 34.0 µg/m³.
• Kadapa (Yerramukkapalli): 37.8 to 36.3.
• Kurnool (Sita Rama Nagar): 41.2 to 45.6.
• Tirupati (Vaikuntapuram): 41.2 on 24 and 25 October; too few readings on 26 and 27 October.
• Hyderabad, nine TSPCB monitors: 37.7 to 30.8.

There is no monitor in Yanam, the Puducherry enclave closest to landfall. Puducherry's only monitor is in Puducherry town, nearly 600 km south.

South Odisha: the same pattern, and a cleaner finish

Odisha's monitors rose and fell in step with the Andhra coast.

A sensor at Berhampur University, part of the Aurassure University Clean Air Network rather than the official network, reached 59.3 µg/m³ on 24 October and fell to 1.5 at 10:30 am on 27 October. We have no official monitor in Berhampur (Brahmapur), Gopalpur or Rayagada. CPCB's bulletin moved Bhubaneswar from Moderate (111) on 24 October, with ozone the main pollutant, to Good (49) on 27 October.

Why did the air get worse before the storm?

We cannot say for sure. Researchers have described a pattern on the fringes of typhoons in East Asia: sinking air and light winds ahead of a storm trap pollution near the ground for a few days before landfall. A 2021 study of a typhoon over central Taiwan describes this (Lin et al., Atmospheric Chemistry and Physics, 19 November 2021). But a study of Typhoon Nuri in Hong Kong found the pre-landfall pollution there was mostly blown in from inland, not trapped by sinking air (Yang et al., Atmospheric Environment, July 2012).

We found no study of Arnab or of this coast that settles it, and our archive holds no wind data. Timing also cuts against a simple story: Visakhapatnam passed 70 µg/m³ on 20 October, three days before IMD reported the low-pressure area. Something other than the storm started the build-up. Whether Arnab's outer circulation made it worse on 23 to 25 October is not established.

Why did the air clean up?

Rain and wind, in two stages. Daily PM2.5 in Visakhapatnam fell from 96.6 on 25 October to 46.7 on 26 October, as Arnab closed from about 610 km to 220 km offshore (IMD, 26 and 27 October). IMD recorded no heavy rain (7 cm or more) on the Andhra coast in the 24 hours to 8:30 am on 27 October (IMD, 27 October), so this first drop came before the downpours. IMD had forecast squally winds along the Andhra Pradesh and Yanam coast from the morning of 27 October (IMD, 26 October).

The second stage came with the rain. Several places in Anakapalli and Visakhapatnam districts had more than 200 mm between 8:30 am on 27 October and 3 am on 28 October (Open, 28 October). Rajamahendravaram fell to 2 µg/m³ at 9:30 pm on 27 October. Raindrops collect particles as they fall and storm winds mix and carry them away; our explainer on rain and pollution covers how washout works.

Which monitors went offline?

We do not know why each monitor stopped; storms often cut power and data links. As of 4:30 pm on 28 October:

Srikakulam's monitor repeated one value for seven hours on the night of 27 October, a stuck sensor, then reported nothing until 11:30 am on 28 October. Vijayawada, Bhubaneswar, Cuttack, Paradeep and Hyderabad kept reporting.

Where does Arnab go next?

IMD expects the system to weaken to a deep depression late on 28 October and become a well-marked low by the evening of 29 October as it moves north-north-west, with heavy rain forecast for parts of Telangana, Chhattisgarh, Jharkhand and Bihar (IMD, 28 October, 2:30 pm). IMD also forecast rain for Delhi on 29 and 30 October but links it to western disturbances, not to Arnab. Our Delhi story covers what that rain may do to the capital's air before Diwali.

After the storm: indoor air and clean-up

Outdoor air on the coast is clean for now; the bigger risks are indoors. This is general guidance; follow your district administration.

• Run generators outside only. Generator exhaust contains carbon monoxide, which has no smell or colour. The US Centers for Disease Control and Prevention (CDC) says never to use generators, grills or other fuel-burning devices inside a home, garage or carport, and to keep generators at least 20 feet (about 6 m) from doors, windows and vents (CDC, reviewed 26 August 2026).
• Dry out fast to stop mould. CDC advises drying the home within 24 to 48 hours, throwing out anything soaked by flood water that cannot be cleaned and dried in that time, opening doors and windows while you work, and never mixing bleach with ammonia. Wear gloves, a mask and goggles (CDC, reviewed 26 August 2026).
• Why damp matters. The World Health Organization (WHO) links damp, mouldy homes to breathing symptoms, allergies and asthma (WHO guidelines on dampness and mould, 2009).
• Stay safe outside. India's National Disaster Management Authority (NDMA) advises drinking only boiled or chlorinated water, not entering damaged buildings and watching for broken electric poles and wires (NDMA cyclone do's and don'ts, circulated by the Puducherry government).

How we measured

• Data. Hourly PM2.5 from 33 monitors: APPCB monitors in Andhra Pradesh, OSPCB monitors in Odisha and TSPCB monitors in Hyderabad, all in CPCB's continuous network, plus one Aurassure sensor at Berhampur University, marked as such. No satellite estimates.
• Days. Midnight to midnight UTC, which is 5:30 am to 5:30 am India time. Times in the text are IST, for the hour that starts then.
• Valid days. A day counts only if a monitor had at least 18 real hourly readings. From 20 October we counted readings hour by hour; before that we used the archive's daily count. We dropped 44 of 462 monitor-days, and 31 had no data. 28 October is a partial day, pulled at about 6:30 pm IST.
• Stuck sensors. Five or more identical readings in a row count as a stuck sensor, and we dropped those hours (14 in the hourly file). We left out Vijayawada's HB Colony monitor entirely: it repeated values for hours at a time and spiked all month.
• "Last 12 hours." The mean of each monitor's last 12 readings before 5:30 am on 28 October, or before it went offline.
• Our numbers and CPCB's. CPCB's bulletin is a 24-hour average to 4 pm on the AQI scale, so its days differ from ours: Rajamahendravaram's Poor 209 on 26 October includes the evening of 25 October, when PM2.5 peaked. Use ours to compare monitors and hours, CPCB's for the official band.

Live readings: Visakhapatnam, Kakinada, Rajamahendravaram, Vijayawada and all of India. See also the northeast monsoon in Tamil Nadu.