Kathmandu's pre-monsoon air pollution: Propelled by regional emissions, exacerbated by local meteorological conditions

Surging forest fire incidents in Nepal

The Forest Fire Detection and Monitoring System (FFDMS) of the Nepal Government’s Ministry of Forest and Environment, detected around 1300 forest fire cases in the country just in a month-long span between 5 March and 5April. However, the number of daily forest fire outbreaks saw sudden spikes since 21March, albeit with some day-to-day fluctuations (Figure 1). 

Plummeting air quality in Kathmandu valley

Simultaneously, a rapid assessment of the air quality data obtained from the Khumaltar air quality monitoring station in Lalitpur during this one-month period, revealed severe deterioration in air quality between March 21 and April 5 with levels of fine particulate matter (PM_2.5)  altering from 35–170 μgm^-3 and carbon monoxide (CO) concentrations ranging from 0.3–1.5 ppm; this is vis-à-vis the situation between 5-20 March  (both PM_2.5  and CO levels showing lower variability , ranging between 40 - 85 μgm^-3 and 0.4 - 0.6 ppm, respectively) (Fig 2, left panel).  

The mean PM_2.5 concentration for 21 March - 5 April was measured at 97.2 µgm-³, almost 1.5 times higher than the measured value of 65.9 µgm-³ for 5 March - 20 March. The mean CO concentration, on the other hand, was measured at 0.8 ppm for 21 March - 5 April , almost 60 percent higher than the mean value (0.5 ppm) for the preceding fifteen days. (Fig 2, right panel). Such high levels of concentration of pollutants, PM_2.5 in particular, increase the risk of cardiopulmonary diseases as well as all-cause mortality.

Fire hotspots: spatially clustered, positionally distant

While the US Government’s National Aeronautics and Space Administration’s (NASA) MODIS satellite images corroborated with the drastic change in the forest fire situation between 21 March  and 5 April  with fire incidents being more prevalent over time and space, it also revealed higher spatial concentration of the fire hotspots in the western/southwestern side of the Kathmandu valley (Figure 3)).

To be noted in this context, that nearly 26% outbreaks were detected in the Madesh province, 25% in Bagmati, 16% in Koshi, 14% in Lumbini and 18% cumulatively in the Gandaki, Karnali, and Sudurpashchim provinces, respectively.

The FFDMS data, on the other hand, shows only one localised case of forest fire in the valley (at Lalitpur) itself during this month-long phase, while several major fire hotspots – for instance, Chitawan (147 events), Makawanpur (110 events), Sindhuli (49 events) in the Bagmati province itself, or Parsa (174), Udayapur (77), Dang (54), in the adjacent Madhesh, Koshi, and Lumbini provinces, respectively -  were detected at over 100 kilometres away from the valley.

With smoke plumes often rising to 4000-5000 metres, way beyond the typical planetary boundary layer in the region, forest fires are known for increasing the chances of long-range transport of emissions like carbon monoxide (CO), fine particulate matter (PM_2.5) and ozone precursors.

But how does the transport and dispersion of the pollutants occur? Empirical analysis of the process is relatively scare. Thus, to get an empirical perspective of it, especially during this year’s coinciding phase of severe air pollution in the Kathmandu valley and raging forest fire outbreaks across Nepal, we used the Khumaltar station as our receptor location for monitoring the source, transport and dispersion of pollutants in the local air¹.

Meteorological conditions and pollutant dispersion

Forest fires from dried vegetations usually intensify during the pre-monsoon months in the southern Hindu Kush Himalaya (HKH) foothills (of India, Nepal and Bhutan) as well as at higher altitudes close to the cryosphere, due to dry weather conditions extending through the winter season.  With longer spells of dry weather conditions becoming common in the region under the exacerbating effects of climate change, forest fires also have amplified in frequency, scale and intensity over the past ten years or so.

This years’ fierce pre-monsoon forest fires in Nepal, for instance, came on the heels of a drier-than-normal winter season - Nepal received only nine per cent of the average winter rainfall by 23 January , 2025, and the HKH region, in general, saw record-low winter snowpack this winter – and an advanced drought warning from the National Agricultural Drought Watch.  

But it is the local meteorological condition, such as wind direction and velocity, atmospheric pressure and humidity etc. in the Kathmandu valley, that appears to play a crucial role in the dispersal of the pollutants.

Using pollution rose plots, we visualised CO and PM_2.5 concentrations together with wind direction between 5-20 March and 21 March 21 – 5 April, respectively. Our analysis revealed that the frequency of winds from the southern direction increased during the fortnight of escalating fire outbreaks, coupled with a rise in the percentage of calm conditions from 10% to 15%, respectively. In tandem, both average CO and PM_2.5 concentrations in the valley’s air increased by 64% (from 0.49 ppm between 5-20 March 5 to 0.82 ppm between 21 March and 5 April) and 47% (from 65.94 µgm^-³ between 5-20 March to 97.2 µgm^-³ between 21 March and 5 April), respectively (see Figs. 4 and 5).

Emissions source and transport

Simultaneously, we used the hybrid single-particle Lagrangian integrated trajectory (HYSPLIT) model - a standard tool for simulating transport and dispersion of air pollutants – to track the source and movement of emissions to the valley during the severe pollution days, during the forest fire period.

A 48-hour backward trajectory analysis for identifying the sources of the emissions influencing daily air quality at the receptor location indicated that during the most polluted days pollutant sources showed higher likelihood of association with locations in the west/southwest of the valley where the forest fires were spatially concentrated.

The 48-hour backward trajectory for March 8, one of the days of relatively lower PM_2.5 concentration in our month-long monitoring phase, traced incoming winds mainly from south/southeast sides of the valley. In contrast, the trajectory for 2 April, when high PM_2.5 concentration was detected in the receptor location air, traced incoming winds from the forest fire prone western / southwest sides of the valley (Fig 6 top panels).

The findings from a 120-hour backward trajectory analysis for the frequency of airmass trajectories from long-distance emission sources are consistent with the surface wind patterns traced by the 48-hour back trajectories, thereby corroborating with long-range transport of emissions to the Kathmandu valley from distant sources, during severe forest fire episodes.

The 120-hour trajectory analysis, during the coinciding peak phases of pollution and forest fires, revealed higher frequency of incoming air masses from fire-affected regions outside of the valley, and hence higher likelihood of long-range transport of pollutants in the valley. In contrast, the pre-peak-fire period analysis showed higher frequency of surface winds that are more likely to bring in localised pollutants (Fig 6, bottom panels).

1. Long-range transport of air pollutants refers to the atmospheric transport of air pollutants for a distance greater than 100 km. So, by local air we mean a moving air mass over a distance ≤ 100 km. ↩︎