Fig. 1: Spatio-temporal variations of observed dust storm days and simulated dust emission flux.

a Annual dust storm days (N_DS; days) averaged over 2001–2017 and c the linear trends of N_DS (days decade⁻¹; only stations with p < 0.1 are shown). The dust storm days are recorded at 2340 synoptic weather observations. The stations with N_DS < 0.1 days are not shown in (a). b Annual mean dust emission flux (F_emis; g m⁻² yr⁻¹) during 2001–2017 and d the linear trends of F_emis (g m⁻² yr⁻¹ decade⁻¹; p < 0.1 for regions marked with crosses) simulated by the dust emission model (DuEM v1; from the baseline experiment in Table 1) during 2001–2017. The black rectangle denotes the Eastern Sources (35-49 °N, 94-126.5 °E), from which the dust particles are mainly transported eastward and southward to affect the densely populated regions of China. e The temporal variations of observed dust storm days (black line marked with triangles) and simulated dust emission flux (magenta line marked with circles; from the baseline experiment) averaged at the synoptic stations over Eastern Sources during 2001–2017. The correlation between two curves is 0.75 (p < 0.001). Also shown are the linear trends (dash lines; p < 0.1). f Reduction of dust emission during 2010–2017 (2010s) relative to 2001, attributable to changes in surface wind (“Wind”), vegetation cover (represented using leaf area index “LAI”), soil moisture (“SOILM”) and all these three factors in northern China (from the baseline experiment: “Wind+LAI+SOILM”). These three factors account for 46%, 30%, and 24% of total dust emission reduction. The attribution is derived based on the experiments listed in Table 1.