Forecasting of air pollution episodes using machine learning in Kathmandu Valley

Air pollution, increasingly aggravated by climate change, presents a multifaceted threat encompassing environmental degradation, public health risks, and social disruptions. In Kathmandu, Nepal's capital, the city's distinctive bowl-shaped topography frequently results in the entrapment of pollutants, leading to prolonged periods of hazardous air quality. This study focuses on identifying the broad-scale synoptic conditions that intensify air pollution in the Kathmandu Valley across various seasons. Our analysis encompasses high pollution instances, scrutinizing alterations in local and external sources, transboundary transport, and meteorological conditions. Our prior research highlights the profound impact of transboundary smoke from agricultural burning in northwest India and forest fires in Nepal, particularly evident in early 2021 when severe pollution led to the closure of schools in Kathmandu. The objective of this study is to delineate the specific meteorological conditions that facilitate the ingress and retention of polluted air in the valley, driven by large-scale synoptic weather patterns. Subsequently, we plan to employ machine learning techniques for pattern recognition, aiming to precisely predict high pollution events. This approach is expected to serve as a vital instrument for timely intervention and strategic policy formulation, addressing the complex challenges of urban air pollution.