[H-CG26] 農業残渣焼却のもたらす大気汚染と健康影響および解決への道筋

[HCG26-06] Impact of autumn crop residue burning on PM_2.5 over North India using a meteorology-chemistry model, a high-density surface observation network, and inverse modeling

*Mizuo Kajino^1,4、Kentaro Ishijima¹、Joseph Ching²、Kazuyo Yamaji³、Rio Ishikawa⁴、Tomoki Kajikawa⁴、Tomoki Nakayama⁵、Yutaka Matsumi⁶、Prabir Patra^7,8、Sachiko Hayashida⁷ (1.Meteorological Research Institute, Japan Meteorological Agency、2.The Education University of Hong Kong、3.Kobe University、4.University of Tsukuba、5.Nagasaki University、6.Nagoya University、7.Research Institute for Humanity and Nature、8.Japan Agency for Marine-Earth Science and Technology)

キーワード：Crop Residue Burning, Air Polltion, North India, NHM-Chem

In recent years, increase in PM_2.5 surface air concentrations in Delhi during the intensive burning period of rice crop residues in the state of Punjab is a cause of concern. However, quantifying the impacts of crop residue burning on air quality is not straightforward, mainly because fire emissions detected by satellites may be underestimated: Fire detection is sometimes affected by clouds or haze over the area. Another reason is the lack of surface observation data, as the monitoring stations in Punjab are located in urban areas and not in agricultural areas. Against this background, a campaign of a dense in-situ surface observation network with low-cost sensors called CUPI-G (Compact and Useful PM_2.5 Instruments with Gas sensors) was conducted in autumn 2022 (Singh et al., 2023). We performed the numerical simulations using Japan Meteorological Agency’s regional meteorology-chemistry model (NHM-Chem) with the daily fire emissions (GFAS) and other emission sources such as anthropogenic and natural sources. Our simulations showed significant underestimation of PM_2.5 concentrations using GFAS emissions during the episodic periods November 1-3 (plume 1) and November 8-12 (plume 2), identified by Singh et al. (2023). We made an inverse estimation of the additional emissions associated with crop residue burning using the tagged NHM-Chem simulations and the surface observation datasets of PM_2.5 and CO by minimizing a cost function with the L-BFGS-B algorithm (Byrd et al., 1995; Zhu et al., 1997). We set five area-based tags in burning areas and two temporal tags (0:00-12:00 and 12:00-24:00 UTC; corresponding to local daytime (5:30-17:30) and nighttime (17:30-5:30)). The new simulation successfully reproduced the increase in PM_2.5 concentrations both in the source (Punjab) and in the downwind states (Haryana and Deli NCR) for the plume 1, while the simulation for the plume 2 was unsuccessful. The reason for the failure of the simulation for plume 2 may be mainly due to the substantial deviations between the simulated and observed wind speed. Certainly, the inverse modeling does not work consistently if the meteorological simulation fails. It is also due possibly to the coarseness of the spatial and temporal resolutions of tags: more tags would have been required to minimize the deviations between simulated and observed PM_2.5 concentrations.