Global weather datasets may be underestimating Himalayan snowfall, says international study

Some of the snowiest parts of the Himalayas may be receiving significantly more snowfall than existing global weather datasets indicate, according to a new international study. Scientists involved in the research concluded that “Existing weather models are failing to accurately capture snowfall across the region’s rugged mountain terrain," according to Gaurav Talwar's Times of India report.

The study, published this month in the peer-reviewed Monthly Weather Review journal of the American Meteorological Society, found that widely used global atmospheric datasets consistently underestimate snowfall across the western-central Himalayas, particularly during major winter storms. Researchers said the findings could help improve snowfall forecasts, water resource planning and avalanche hazard assessments across the region.

The research was led by Siddharth Gumber of the British Antarctic Survey, UK Research and Innovation, Cambridge, in collaboration with scientists from the UK Met Office, the University of Leeds, IIT Kharagpur, the University of Cambridge, the University of Sheffield and the University of Birmingham.


Local snowfall hotspots missed by global datasets

According to the study, snowfall patterns in the Himalayas vary sharply over short distances because of the region's steep and complex terrain, making them difficult for coarse-resolution global weather datasets to capture accurately.

Using a high-resolution atmospheric model calibrated with newly collected snowfall observations, the researchers identified several localised snowfall hotspots that are either significantly underestimated or entirely absent from existing global weather products.

One of the most striking findings came from the western Himalayas, particularly around Manali in Himachal Pradesh. The study found that the best available conventional snowfall analyses underestimated seasonal snowfall by as much as 37% during a single winter. Incorporating the new observations substantially reduced this error, with the improved model estimating seasonal snowfall of more than 800 kg per square metre in some locations—well above estimates generated by conventional global datasets.

Gap widens during heavy snowstorms

The researchers found that the underestimation was not limited to the Himalayas. Global reanalysis datasets underestimated snowfall across all three mountain regions examined in the study—the western-central Himalayas, the European Alps and the Rocky Mountains in the United States.

The differences became even more pronounced during intense snowfall events. The high-resolution model was able to reproduce both the timing and intensity of major snowstorms much more accurately than existing global datasets.

To validate their approach, the researchers tested the atmospheric model across the three mountain systems using observations from frozen lakes. Snow accumulating on the lake surface changes the pressure beneath the ice, allowing scientists to derive an independent estimate of snowfall in remote mountain regions.

Explaining the methodology, lead author Gumber said, “The research team deployed novel snowfall sensors in high mountain regions to measure the timing and intensity of snowfall. Based on Archimedes’ principle, the instruments measured changes in water pressure beneath frozen lakes to directly estimate the mass of snow accumulating on the lake surface, providing highly accurate snowfall measurements.”

Why accurate snowfall estimates matter

The researchers said more precise snowfall estimates have implications far beyond weather forecasting. Snowfall in the Himalayas feeds major river systems, affects glacier mass balance and plays a critical role in agriculture, hydropower generation and water security across northern India.

Improved snowfall mapping could also strengthen avalanche forecasting and provide scientists with better tools to assess how climate change may reshape snow-dependent ecosystems across the Himalayan region.

(With TOI inputs)