2014—2023 年东亚地区沙尘气溶胶质量浓度再分析数据集

Abstract

Dust storms are among the most severe hazardous weather phenomena affecting northern China and adjacent regions.The primary dust source areas—including the Alxa-Hexi Corridor,the Tengger Desert,and the southern Mongolian Gobi Desert—emit more than 800 Mt of dust annually.During spring,the interaction between the Siberian high and Mongolian cyclones generates strong near-surface winds and enhanced vertical convection, forming a three-dimensional “uplift-suspension-transport”structure that promotes dust storm development.Under ongoing global warming,declining spring precipitation over the Mongolian Plateau and extensive desertification—currently affecting over 75% of Mongolia—are expected to further intensify transboundary dust transport into China,with severe consequences for public health,agriculture,and transportation.These challenges underscore the urgent need for long-term,high-quality dust datasets to improve understanding of dust emission mechanisms and forecasting capabilities. Atmospheric models are essential tools for simulating dust emission,transport,and deposition,as well as for assessing impacts on climate,ecosystems,and human health.However,large uncertainties in emission parameter-izations and long-range transport processes persist,often resulting in substantial biases in simulated dust concentrations,in some cases differing from observations by up to two orders of magnitude.Recent advances in atmospheric observation systems provide valuable constraints,including China's nationwide hourly PM₁₀ monitoring network and satellite remote sensing products with broad spatial coverage and multi-dimensional aerosol information,such as MODIS aerosol optical depth (AOD).In this context,data assimilation methods grounded in Bayesian theory offer an effective framework for integrating observational data with model simulations to generate spatially continuous and more accurate dust reanalysis datasets.Despite progress,existing studies have primarily focused on individual dust events,and long-term dust reanalysis efforts remain limited due to observation biases,sparse data coverage over source regions,transport errors,and the strong spatiotemporal variability of dust emissions. Building upon a self-developed dust storm assimilation system,this study integrates ground-based PM₁₀ observations,bias-corrected satellite AOD data,and an effective valid time shift ensemble Kalman filter (VTS-EnKF)designed to jointly correct dust intensity and transport position errors.Using this framework,we construct a high-resolution (0. 25°×0. 25°,3-hourly)three-dimensional dust aerosol mass concentration reanalysis dataset for East Asia during spring (March-May)over the period 2014—2023.This dataset provides a robust basis for investigating long-term dust variability,transboundary transport processes,and associated impacts on climate,the environment,and public health. Comparisons with MERRA-2 dust reanalysis demonstrate clear advantages of the newly developed dataset. While MERRA-2 exhibits reasonable agreement at low dust concentrations (<75 μg·m^-3),it substantially underestimates dust levels and exhibits larger uncertainties under moderate to severe dust conditions,particularly in dust-affected regions.Analysis of springtime dust variability from 2014 to 2023 reveals pronounced interannual and spatial heterogeneity,with dominant dust activity over the Tarim Basin and the Gobi Desert in China and episodic contributions from the Mongolian Gobi.Relative to observations,prior simulations tend to overestimate dust concentrations,whereas data assimilation introduces widespread negative analysis increments,reducing the regional mean concentration from 65. 24 to 39. 99 μg·m^-3.Notably,the reanalysis accurately captures both the intensity and timing of dust events in densely populated areas.Overall,the assimilation framework substantially improves dust representation,reducing RMSE by 76.9% and yielding a more reliable depiction of monthly and interannual dust variability.