非表示:
キーワード:
-
要旨:
Tropospheric ozone (O3) is a key greenhouse gas and pollutant that is receiving increasing attention globally. While there are many sources of tropospheric O3, precursors from human activity (Anthro) and open biomass burning (BB) are the only ones that can be controlled. As such, it is crucial for policymakers to understand the relative contributions of the two. However, determining the contribution of O3 can be challenging as it cannot be directly observed. It must be calculated by chemical transportation model (CTM) simulation which could be biased for unreal emission inventory, or estimated by real observations that assumes too simple chemical and transportation processes.
In this paper, we propose a solution by developing a deep learning (DL) model that combines both CTM simulations and observations. The DL model is able to learn a generalized relationship between unobservable O3 contribution from Anthro or BB sectors and observable mixing ratio of tracers simulated by CTM with full chemistry and transportation processes. The DL model then, when applied to observed tracers, could avoid the bias from model to provide an accurate estimation of the contributions in reality.
Our results indicate the contribution from BB to tropospheric remote ozone mixing ratio is no larger than that from Anthro emission from a global perspective, even when uncertainties are deliberately tuned to bias BB. Therefore, the reduction of anthropogenic emissions should be the top priority for controlling global background O3 levels, at least for the time period of 2016-2018 studied.
