How far cameras and thermal imagers can see along the border, hour by hour for 72 hours.

Cameras, thermal imagers and drone detectors along the Punjab and Rajasthan border work through two hard seasons a year: crop-residue smoke in October and November, and dense fog from December. By 5 October 2026 Punjab had recorded 84 stubble fires, 67 of them in the border districts of Amritsar and Tarn Taran (IANS via NewKerala). The Border Security Force says smugglers fly drones under cover of fog (The Week, November 2025).

Weather forecasts give visibility for people and aircraft, not range for a sensor, and they leave out today's smoke. Operators find out how far their cameras can see when the picture goes grey.

• Gives every hour of the next 72 a range for a day camera and a thermal imager, with a band from a dull target to a high-contrast one, and a class: good 5 km or more, degraded 1 to 5 km, poor under 1 km.
• Flags fog risk from dew point, wind, night and temperature, a heuristic checked against airport fog reports over two winters.
• Grades smoke and haze from modelled PM2.5 and aerosol optical depth on India's National AQI breakpoints, and shows which way the wind carries it.
• Counts satellite fire detections by 0.25° cell and by district on both sides of the border, the same way on each side, and tracks the season against last year.
• Checks itself against airport reports every run and publishes the result, with the formulas, on a methods page.

1. Forecast. Two batched requests to Open-Meteo bring hourly visibility, humidity, dew point, wind, rain and daylight from ECMWF's model, and PM2.5, PM10, dust and aerosol optical depth from the Copernicus atmosphere model, for 45 reference points and 12 airports.
2. Air. The model's visibility gives one extinction coefficient; the forecast smoke gives another, scaled to airport reports. The larger one wins each hour.
3. Range. Contrast fades with distance as exp(−σR) (Koschmieder). Thermal range adds water-vapour absorption, rain attenuation and a wavelength scaling that lets smoke pass and fog block.
4. Sectors. Each sector shows the worst of its reference points, hour by hour.
5. Fires. NASA FIRMS VIIRS detections are filtered for confidence and static heat sources, then counted by cell and district. Points never leave the pipeline.

Camera, thermal and counter-drone operators deciding which sensors to rely on tonight; commanders planning patrols around the hours when cameras see least; Shyena and Sangam planners sizing sensor coverage; district administrations watching smoke over border villages.

• Counts, not locations: fires are published only as cell and district totals, counted the same way on both sides of the border. No field-level maps that could be used against individual farmers.
• Reference points are public towns, crossings and heritage sites, not security positions.
• Every number carries its source and time, and the method, checks and failure modes are published.
• Advisory only. Follow IMD fog and weather warnings, CPCB and state pollution control board advisories, NDMA and state disaster management authorities.

• Drishti estimates what the air allows. Lens, resolution, mounting height, terrain and vegetation can only shorten the range.
• ECMWF calls its visibility diagnostic experimental, and the smoke model runs on a grid of about 40 km, so local plumes and patchy fog are missed.
• The fog score is a heuristic, not a probability. On the 2024-25 winter, which it was not tuned on, its high class caught 42% of fog hours and 43% of its warnings were false alarms; in October, with little fog, it had no hits.
• Satellites miss small, short or cloud-covered fires. Official crop-residue counts from CAQM and the Punjab Remote Sensing Centre use other methods.
• The free Open-Meteo service is for non-commercial use; a deployment needs a subscription or a self-hosted copy.

Pilot Drishti for your sectors, refreshed every hour and run on your own servers. contact@trinetra.in