1. Introduction
Dust aerosols impact the global energy budget through interactions with radiation, clouds, atmospheric chemistry, biogeochemistry, and the cryosphere, with the total effect on the global energy budget being −0.2 ± 0.5 W m−2 (Kok et al. 2023, 2017). Northern China (NC), as the third largest source of dust, contributes ∼12% of global dust emissions (∼600 Tg) every year (Kok et al. 2021, 2023). Dust from NC is generated in the Gobi Desert (GD) and the Taklimakan Desert, a considerable part of which can be transported over long distances (Kai et al. 2019; Liang et al. 2022; Ridley et al. 2016; Song et al. 2021; Proestakis et al. 2018; Gkikas et al. 2022; Yao et al. 2020), affecting most parts of China (An et al. 2018; Chen et al. 2017a; Yao et al. 2021), Japan, South Korea (Zhang et al. 2003), the Pacific Ocean (Tan et al. 2017), and the United States (Fairlie et al. 2007). Therefore, radiative forcing attributed to dust is one of the largest sources of uncertainty in the study of global and regional climate change.
Due to the importance of dust over NC in climate change, it is crucial to continuously measure and investigate the climatological characteristics and radiative effects of dust. Systematic monitoring of dust events was established in China in the 1950s. Atmospheric visibility has been recorded to monitor dust events and investigate the sources, transportation, spatiotemporal variation, and regional characteristics of dust (Wang et al. 2008; Qian et al. 2004; Wang et al. 2005). However, only the types of dust events have been reported. To obtain data on dust concentrations, the China Meteorological Administration established a dust monitoring network in dust-source and downwind regions in NC, with PM10 observations, since 2003. However, these site-scale data have limitations in terms of their spatial continuity and vertical distribution (Wang et al. 2008). Satellite-based lidar is much more effective in monitoring, tracking, and analyzing dust than traditional ground-based observations, as it combines the unique advantages of lidar and the high orbit and large coverage of satellites (Wang et al. 2004; Chen et al. 2014). Thus far, different satellite platforms have been extensively used to monitor dust emissions and dust aerosol trajectories, such as the Advanced Very High Resolution Radiometer (AVHRR) (Iino et al. 2004), the Geostationary Operational Environmental Satellite (GOES) series, the Sea-Viewing Wide Field-of-View Sensor (SeaWiFS) (Yoon et al. 2011), the Total Ozone Mapping Spectrometer (TOMS) (Ren et al. 2017; Gao and Washington 2009), the Cloud–Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) (Gui et al. 2021; Huang et al. 2015; Proestakis et al. 2018), and the Moderate Resolution Imaging Spectroradiometer (MODIS) (Pu and Ginoux 2018). Among them, the dust optical depth (DOD), which is the column-integrated extinction due to dust particles and describes the overall dust loading, provided by CALIPSO or retrieved using MODIS aerosol products, is the basis for quantifying the effects of dust through aerosol–radiation interactions, and its spatial and temporal distribution characteristics are key to understanding the dust loading and its interactions with the Earth system (Liu et al. 2019; Pu and Ginoux 2018). For instance, Proestakis et al. (2018) presented the 3D climatology of the desert dust distribution over South and East Asia based on Cloud–Aerosol Lidar with Orthogonal Polarization (CALIOP). Gkikas et al. (2022) quantified the DOD with its monthly and year-to-year variability between 2003 and 2017 in both global and regional levels based on the MODIS Dust Aerosol (MIDAS) dataset, which combines MODIS-Aqua retrievals and Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2), reanalysis products. Additionally, Ginoux et al. (2012) used MODIS collection 5.1 level 2 aerosol products from the Aqua satellite to retrieve DOD, and MODIS DOD products have been widely used for subsequent dust studies and model validation (Yu and Ginoux 2022; Pu and Ginoux 2018; Gui et al. 2022; Song et al. 2021). More comprehensively, Song et al. (2021) analyzed regional and interannual variability based on decadal-scale global DOD derived from CALIOP and MODIS aerosol product and found that the two datasets compare reasonably well with the results reported in previous studies and the collocated Aerosol Robotic Network (AERONET) coarse-mode aerosol optical depth (AOD).
Although meteorological records and remote sensing techniques are widely used to study the climatological characteristics of dust, the climatological evolution mechanism and three-dimensional structure of dust are still poorly understood because of the lack of continuous and simultaneous dust observation data. Global climate models are indispensable for identifying the climatological and variation mechanisms of dust in NC (Chen et al. 2017b; Colarco et al. 2010; Shao and Dong 2006; Wang et al. 2008; Ridley et al. 2016). The state-of-the-art global climate models of the Coupled Model Intercomparison Project (CMIP) provide the basis for research on historical variations of DOD and its feedback (Eyring et al. 2016). Currently, phase 6 of CMIP (CMIP6). Given the contribution of CMIP6 to dust research, several studies have evaluated the performance of global climate models participating in CMIP6 in characterizing the spatial patterns of dust and their drivers over NC. For instance, Zhao et al. (2022) evaluated the global dust cycle in CMIP6 during 2005–14, utilizing the Aqua MIDAS dataset combined with the MERRA-2, and Copernicus Atmosphere Monitoring Service (CAMS) reanalysis data, in which the large uncertainties and intermodel diversity were found and the dominant drivers of dust emissions were revealed. Li et al. (2021) and Ali et al. (2022) addressed the errors in CMIP6 AOD from 2000 to 2014 in China and considered the influence of dust on them. However, owing to the lack of DOD observations, the deviations and improvements for CMIP6 DOD in historical simulations during the pre-2005 period have not been presented. Furthermore, the biases in trends of meteorological drivers which drive the variation of DOD remain to be revealed. Meanwhile, the CMIPs also provide a recognized basis for research on future projections of DOD and related impacts, but the errors and sources of errors in projections have not yet been quantified. Considering that CMIP6 future climate projections start from 2015, 7 years of CMIP6 future climate projections can currently be compared with observed data (Carvalho et al. 2022).
Thus, this study evaluated the performance of CMIP6 climate models and multimodel ensemble mean (MEM) in characterizing the spatial pattern and trend of DOD during March–August in historical simulations and under Shared Socioeconomic Pathway (SSP) future scenarios, with respect to ground-based observations, MODIS DOD, and MERRA-2 reanalysis data. Specifically, the aims of this study were to 1) reconstruct a DOD dataset for the pre-2002 period by utilizing empirical models of retrieved MODIS DOD and dust-event frequencies during 2002–21; 2) evaluate the CMIP6 DOD during March–August for the period 1980–2001 based on the reconstructed DOD; 3) evaluate the CMIP6 DOD during March–August for the period 2002–14 from historical simulations and during 2015–21 under different SSPs by utilizing MODIS DOD; and 4) detail the contributions of simulated trend of meteorological drivers to errors in trend of DOD.
2. Data and methods
a. CMIP6 model outputs
In this study, we obtained monthly simulations of monthly DOD at 550 nm (the variable named od550dust in CMIP6) and meteorological drivers [near-surface wind speed (WS, the variable named sfcWind in CMIP6), leaf area index (LAI), and total soil moisture content (MRSO)] during March–August based on the experimental outputs of the historical run of 10 global climate models participating in CMIP6 (shown in Table 1). The 10 models used an interactive dust scheme and have the output of DOD at 550 nm, in which dust emissions are parameterized as a function of surface wind speeds or wind stress, and account for the amount of bare soil, soil type, and aridity (Aryal and Evans 2021; Eyring et al. 2016; Thornhill et al. 2021; Petrie et al. 2021). In models in which dust emissions do not respond to variations in meteorology and climate, the connection between DOD and meteorological drivers is disjointed. The historical simulation of CMIP6 focuses on the period 1850–2014. Given that MODIS DOD is only available during 2002–21 and that the reconstructed DOD dataset is available during 1980–2021 with low resolution, we divided the years 1980–2014 into period 1: 1980–2001 (P1) and period 2: 2002–14 (P2) and evaluated the DOD of CMIP6 in P1 and P2 based on reconstructed DOD during P1 and MODIS DOD during P2. Meanwhile, for future scenarios, data from six models are available. The Scenario Model Intercomparison Project (ScenarioMIP) recommends four Tier one simulation protocols, reflecting different SSPs that result in different radiative forcing magnitudes by 2100: SSP1-2.6 (radiative forcing amounting to +2.6 W m−2 by 2100), SSP2-4.5 (radiative forcing amounting to +4.5 W m−2 by 2100), SSP3-7.0 (radiative forcing amounting to +7.0 W m−2 by 2100), and SSP5-8.5 (radiative forcing amounting to +8.5 W m−2 by 2100) (O’Neill et al. 2016; Tokarska et al. 2020). The variant label of the dataset was r1i1pif1 for the three periods of all experiments. The different variant label defines the realization number (r), initialization index (i), physics index (p), and forcing index (f), all being positive integers. For a given CMIP6 experiment, the “ripf” identifier is used to uniquely identify each simulation of an ensemble of runs contributed by a single model (for more details, visit https://github.com/WCRP-CMIP/CMIP6_CVs). Because CMIP6 future climate projections began in 2015, this study compared the CMIP6 future climate projections of DOD during the period 2015–21 (P3) with observational data. For comparison purposes, we interpolated the DOD from all model outputs in our work onto sites by averaging the data within 1° of the sites during P1 and onto a grid of the same resolution [2° × 1.5° (longitude × latitude)] by bilinear interpolation during P2 and P3.
Primary parameters of CMIP6 models used in this study.
b. DOD retrievals from MODIS observation
c. Surface dust records
The ground-based observation data on the frequencies of dust events (dust storms, blowing dust, and floating dust) in March–August used in this paper are from the ground-based fusion observation dataset (2019–May 2021) and the Dust (Storm) Dataset (V1.0) (1980–2018) (National Weather Bureau of China 1979), which were obtained from the National Meteorological Information Centre of the China Meteorological Administration (http://data.cma.cn/). Among them, the Dust (Storm) Dataset provides daily records of dust events occurrences from 2477 sites of the Chinese Meteorological Administration from 1980 to 2018. The ground-based fusion observation dataset, which records the daily ground-based observation elements of these 2477 sites, includes the records of dust events occurrences for the years 2019–May 2021. The definitions of dust events are the same in both datasets, and the dust event reports are generated by local observers. Dust events have been classified into three categories as dust storms, blowing dust, and floating dust by the National Weather Bureau of China (1979). Dust storms, which are the results of strong turbulent wind systems entraining dust particles into the air, reduce visibility to 1 km and below. Blowing dust is defined as a state in which dust is locally lifted off the ground by strong winds. Horizontal visibility may be reduced to 1–10 km. Suspended dust is defined as a weather phenomenon in which dust and fine dust float uniformly in the air, resulting in a horizontal visibility of less than 10 km (Wang et al. 2018, 2005; National Weather Bureau of China 1979). We calculated the monthly frequency of three types of dust events during 1980–2021 and used the calculated data to reconstruct the 1980–2001 DOD dataset. To ensure the reliability of the reconstructed DOD, we finally selected records from 479 sites and their corresponding empirical models of dust event frequencies and MODIS DOD for reconstructing the DOD dataset for P1. This was done according to data completeness, the magnitude of the frequency of dust events, and the study regions. The specific calculation steps are clarified in section 3. Furthermore, monthly DOD data (with a resolution of 0.625° × 0.5°) in March–August during 1980–2021 from MERRA-2 (Buchard et al. 2017; Gelaro et al. 2017) were compared with CMIP6 DOD.
d. Meteorological drivers
According to previous studies, dust emissions are parameterized as a function of surface wind speeds or wind stress and account for the amount of bare soil, soil type, and aridity in CMIP6 models we selected (Aryal and Evans 2021; Thornhill et al. 2021). To investigate the correlation between errors in simulated DOD and local drivers in March–August, we compared the regional average percentage trends in WS, LAI, and MRSO output from the CMIP6 models and observations (shown in Table S1). Depending on the available temporal coverage of data, the WS data for 1980–2014 were from the Historical Dataset of Surface Meteorological Observation in China from the National Meteorological Information Centre of the China Meteorological Administration (http://data.cma.cn/) and the WS data for 2015–21 were from the Global Land Data Assimilation System (GLDAS) (Rodell et al. 2004). Simultaneously, MOD13C2 version 6 products provide the normalized difference vegetation index (NDVI), with a resolution of 0.1° × 0.1°, which allows for consistent spatial and temporal comparisons of vegetation canopy greenness, a composite attribute of leaf area, chlorophyll, and canopy structure. Both NDVI and LAI products are effective in characterizing vegetation conditions on a global scale (Didan et al. 2015; Vermote et al. 2002). Furthermore, we used the monthly volumetric soil moisture (VSM) with a horizontal resolution of 0.25° × 0.25° from GLDAS for the period 2001–21 (Chen et al. 2013; Rodell et al. 2004). VSM from GLDAS exhibit similar spatial pattern and high consistency in drought trend with satellite-retrieved VSM (Liu et al. 2019). In addition, monthly LAI, VSM, WS, and in March–August during 1980–2021 from the MERRA-2 dataset were compared with the simulated meteorological drivers. Due to the different resolutions of the datasets, we harmonized the resolution of the data before averaging the regions. When two site datasets were compared, we chose the same site corresponding to both datasets; when the gridpoint dataset was compared to the site dataset, we interpolated the gridpoint data onto sites by averaging the data within 1° of the sites; and when two gridpoint datasets were compared, we interpolated the data onto a grid of the same resolution [2° × 1.5° (longitude × latitude)] by bilinear interpolation.
3. Reconstruction of site-scale DOD datasets pre-2002
Locations of 479 sites for the reconstruction of the DOD distribution over NC (33°–53°N, 73°–135°W) [44 sites in NWC (36°–47°N, 75°–96°W); 111 sites in GD (36°–47°N, 96°–112°W); 163 sites in NCP (34°–42°N, 112°–122°W); and 69 sites in NEC (42°–52°N, 112°–130°W)]. Color shading denotes the distribution of annual mean MODIS DOD.
Citation: Journal of Climate 37, 23; 10.1175/JCLI-D-23-0624.1
In this work, we used the leave N-out cross-validation method to build the multiple linear regression (MLR) empirical models for each site. A total of N consecutive years of data during 2002–21 were selected as the testing set, and the remaining 20 − N years of data were used to develop the MLR model and validate it with the testing set. Then, among all the sites with models, the sites with corresponding models passing the 90% significance test and with determination coefficients greater than 0.25 in both the training and testing sets were selected, thus obtaining empirical model sets with high consistency with the MODIS DOD distribution (Celisse and Robin 2008; Liu 2019). Continuously, the above model development process was repeated for all segments with 2, 3, and 4 consecutive years as the testing set (assigning N as 2, 3, and 4, respectively).
Consequently, after the above model development and validation process, we finally selected 497 sites (the distribution of sites is shown in Fig. 1) in the case of N = 2, with the R values and RMSEs for the reconstructed DOD and MODIS DOD being 0.886 and 0.058 in the training set and 0.856 and 0.066 in the testing set, respectively (Fig. 2). Overall, the reconstructed DOD was capable of accurately characterizing the trend in DOD over the four subregions, with the best performance in GD (Figs. S7 and S8). We applied the empirical models and the frequency of dust events during 1980–2001 at 497 sites to calculate the DOD over NC during 1980–2001.
Scatter diagrams of the reconstructed DOD against the MODIS DOD in the (a) training dataset and (b) testing dataset (N = 2) during 2002–21. The red line and the dotted black line are the linear regression and one-to-one line, respectively. The R, slope, matchups, mean absolute error (MAE), RMSE, and relative mean bias (RMB) of the linear regression are shown in the lower-right corner of the panels.
Citation: Journal of Climate 37, 23; 10.1175/JCLI-D-23-0624.1
4. Results and discussion
a. Spatial pattern of DOD in historical simulations
Figures S9 and S10 show the annual mean DOD during P1 and P2 outputs from 10 CMIP6 models and MEM, compared to MODIS and MERRA-2. The CMIP6 models are capable of capturing the spatial pattern of DOD, with a large magnitude difference from MODIS DOD. And the considerable uncertainty in the simulated DOD and the intermodel variability are found. Further, we investigated the relative errors of the CMIP6 DOD versus the reconstructed DOD during P1 and the CMIP6 DOD versus the MODIS DOD during P2, over NC and the four subregions (Fig. 3). The magnitudes of relative error are larger in the dustiest regions. During P1, the simulated DOD is generally lower than the reconstructed DOD. The regional-average relative error of MEM DOD is −29.54% over NWC and −79.49% over GD. During P2, DOD is overestimated over NWC and NCP (with relative errors in the MEM DOD of 91.05% and 23.53%, respectively), but underestimated over GD and NEC (with relative errors in the MEM DOD of −38.88% and −21.36%, respectively). The overestimation is mainly attributable to five models (CanESM5, CESM2, CESM2-WACCM-FV2, MRI-ESM2-0, and NorESM2-LM), among which CESM2-WACCM-FV2 also overestimate the DOD in NCP and NEC. The DOD in MERRA-2 is also lower than the reconstructed DOD and MODIS DOD, with average relative errors of 48.42%–21.94% over NC, respectively. Concurrently, and consistent with Zhao et al. (2022), the MEM estimates are 1.2–1.7 times larger than the MODIS DOD, with overestimates in individual models of up to 7 times larger. Additionally, the intermodel standard deviations are much greater than for the MEM during P2, which indicates large discrepancies between individual models.
Boxplots of relative error, which reveal the error magnitudes of the CMIP DOD vs the reconstructed DOD during P1 (1980–2001) (red), and the CMIP DOD vs the MODIS DOD during P2 (2002–14) (blue). The three gray dashed lines indicate the relative errors of 1, 0, and −1, respectively. The sample capacity, regional average, and intermodel standard deviation of relative errors are labeled in the upper-right corner of each panel.
Citation: Journal of Climate 37, 23; 10.1175/JCLI-D-23-0624.1
To further investigate the magnitude differences and the correlation of models, we used Taylor diagrams to evaluate the performance of MERRA-2 and the 10 CMIP6 models in simulating the seasonal mean DOD in typical regions during P1 and P2 (Fig. 4), for which the monthly mean of the regional average DOD series during March–August was used in the calculations. Generally, the spatial pattern of DOD is better captured in the MEM and MERRA-2, with R values of 0.65 and 0.74 in P1 and 0.18 and 0.65 in P2, over NC, respectively. Meanwhile, the efficiency of the CMIP6 models in simulating DOD shows regional differences. For correlation, over NC, NCP, and GD, CanESM5, INM-CM4-8, and INM-CM5-0 performed the best and CESM2-WACCM also had better correlation with MODIS DOD during P2. Over NWC, CanESM5 and INM-CM5-0 have better correlation with MODIS DOD during P1. Over NEC, most of the CMIP6 models are negatively correlated with observations in P1, and insignificantly positively correlated in P2, indicating that most models perform poorly in simulating DOD over this region. In terms of standard deviation, CanESM5 has the largest standard deviation [ratio of standard deviation (RSTD) = 2.28–6.26]. The RSTD over all subregions for CESM2-WACCM-FV2 and the RSTD over NC, NCP, and GD for MRI-ESM2-0 are also greater than 1.
Taylor diagrams for DOD over NC, NWC, GD, NCP, and NEC among CMIP6 models and MERRA-2 data compared with MODIS DOD during (a) P1 and (b) P2. Azimuthal position: R; radial distance: RSTD; distance from REF point: RMSE. (Only points corresponding to R > 0 and with corresponding RSTDs < 1.5 are shown in the diagrams.)
Citation: Journal of Climate 37, 23; 10.1175/JCLI-D-23-0624.1
b. Interannual trends of DOD in historical simulations
Figure 5 summarizes the percentage trends (% yr−1) in regional average reconstructed DOD, MODIS DOD, and DOD from CMIP6 models during P1 and P2 over NC, NWC, GD, NCP, and NEC. Over NC, the observations indicate that DOD decreases significantly during P1 (with trends of −0.88% yr−1 in spring and −0.29% yr−1 in summer), while during P2, DOD decreases weakly (−0.40% yr−1) in the spring and decreases significantly (−0.97% yr−1) during the summer. The MEM captures the trends in DOD effectively during spring in P2, but the simulated trends are weaker than the observations. Positive trends in MEM DOD during P1 are mainly attributed to the INM-CM5-0, MRI-ESM2, and NorESM2-LM during spring and are mainly attributed to CanESM5 and GFDL-ESM4 during summer. INM-CM5-0, MRI-ESM2, CanESM5, and GFDL-ESM4 also simulated positive trends during P2. CESM2, CESM2-FV2, CESM2-WACCM, and CESM2-WACCM-FV2 captured more of the negative trend in DOD.
Regional linear percentage trends (% yr−1) in DOD during [a(1)],[a(2)] spring and [b(2)],[b(2)] summer in [a(1)],[b(1)] P1 and [a(2)],[b(2)] P2. The colored squares denote the magnitudes of the trends; the numbers denote the corresponding trend values; the numbers with blue and red font signify that the trends are statistically significant at the 90% confidence level (p < 0.1, p represents the p value of the t-statistic for a two-sided hypothesis test).
Citation: Journal of Climate 37, 23; 10.1175/JCLI-D-23-0624.1
As the dominant driver of dust emissions (Zhao et al. 2022; Pu and Ginoux 2018), the WS over dust-source regions was first utilized to investigate the causes of errors in the dust trends (Fig. 6). During P1, the large decrease in DOD over the whole of NC is dominated by the decrease in WS over NWC and GD, by 1.22% and 0.65% in spring, and by 1.44% and 0.6% in summer, respectively. Compared to observations, over NWC, five models simulate weak negative trends (from −0.01% to −0.2%) in WS during spring and seven models simulate weak negative trends (from −0.06% to −0.26%) in WS during summer. Over GD, six models and eight models simulate weak negative trends during spring (from −0.03% to −0.33%) and summer (from −0.04% to −0.6%), respectively. The changes in MERRA-2 WS during spring are also weaker than observed (−0.19% and −2.64%). During P2, NWC and GD surface wind speeds decrease by 0.96% and 0.93% in spring and by 0.72 and 0.48 in summer. During spring, CESM2-FV2, CESM2-WACCM-FV2, CanESM5, GFDL-ESM4, INM-CM4-8, and INM-CM5-0 simulate weak negative trends in WS over NWC, and CESM2, CESM2-FV2, CESM2-WACCM, CESM2-WACCM-FV2, GFDL-ESM4, and INM-CM5-0 simulate weak negative trends in WS over GD. The changes in MERRA-2 WS are also weaker than observed. Overall, the models that simulate more reasonable decreases in DOD (e.g., CESM2-WACCM and the MEM for spring; CESM2-FV2, CESM2-WACCM, CESM2-WACCM-FV2, and the MEM) tend to better capture the changes in WS. Concurrently, NDVI increases significantly (by 0.68%) over GD in spring, and soil moisture also increases slightly (but not significantly) during P2. Additionally, the increases in average DOD during spring simulated by GFDL-ESM4, INM-CM5-0, and NorESM2-LM are mainly due to the decreased simulated LAI (Fig. S11).
Linear percentage trends (% yr−1) of surface WS over [a(1)],[b(1)] NWC and [a(2)],[b(2)] GD in spring (MAM) and summer (JJA) during [a(1)],[a(2)] P1 and [b(1)],[b(2)] P2 in CMIP6 models, MERRA-2, and observations. Note that stars signify trends that are statistically significant at the 90% confidence level (p < 0.1).
Citation: Journal of Climate 37, 23; 10.1175/JCLI-D-23-0624.1
c. Changes in DOD under different SSPs
ScenarioMIP provides projections of different future greenhouse gas and aerosol concentrations based on SSPs of future emissions and land-use change generated by the integrated assessment model, which will inform the projections of future dust trends under global warming. The warming projected by CMIP6 is stronger than in the IPCC AR5 report, thus quantifying the error in future dust projections by CMIP6 models based on 2015–21 observations, answering the question of whether CMIP6 projections of future dust are reasonable, and informing the constraint of future projections of dust by utilizing observations. Given that CMIP6’s future climate projections begin in 2015, we evaluated the performance of DOD outputs from six CMIP6 models under four SSPs over NC during P3.
Figures S12–S15 indicate that models are also capable of capturing the spatial pattern of DOD in the four SSPs, but with significant intermodel variability. Table 2 shows the R, RSTD, and RMSE from the comparison of DOD simulated by CMIP6 models and MERRA-2 with MODIS DOD based on P3 over NC. Generally, MERRA-2 and the MEM outperform most models. Over the whole of NC, the magnitudes of the MEM DOD are closer to those of the MODIS DOD than most of the individual models, with RSTDs of 0.85–1.03. The R between MERRA-2 and MODIS DOD is 0.52, and the RSTD is 0.47. Among the six models, the DOD from CanESM5 is significantly correlated with the MODIS DOD, but significantly overestimated, with RSTDs from 4.23 to 5.09. At the same time, the R, RSTD, and RMSE from the comparison of the DOD simulated by the CMIP6 models and MERRA-2 with MODIS based on future scenarios over NWC and GD are shown in Table S2. Among the different subregions, the spatial patterns over GD are well characterized in the CMIP6 models. Over GD, the R values between the MEM DOD output by the CMIP6 models and the MODIS DOD are 0.39–0.53, which is attributed to the significant correlation of the DOD from CanESM5, INM-CM4-8, INM-CM5-0, and NorESM2-LM with the MODIS DOD. The MERRA-2 DOD, meanwhile, has a correlation coefficient of 0.64 with MODIS DOD over GD. Among them, the standard deviation of the DOD of CanESM5, MRI-ESM2-0, and NorESM2-LM is larger than that of MODIS DOD, and the RSTDs of CanESM5 are larger than 6. Over NWC, the magnitudes of DOD are well represented in the CMIP6 models. However, some models show negative correlations, and only the DOD output from CanESM5 is significantly correlated with the MODIS DOD, with RSTDs of 5.01–6.52, respectively. Among the four SSPs, the spatial R values between the MEM DOD and MODIS DOD over NC are largest in SSP3-7.0 (R = 0.36, RSTD = 1) and the spatial R over GD is largest in SSP1-2.6 (R = 0.53, RSTD = 1.57), and over NWC, the MEM DOD is not significantly correlated with the MODIS DOD.
The R, RSTD, and RMSE from the comparison of the DOD simulated by CMIP6 models and MERRA-2 with MODIS DOD under four SSPs (based on P3: 2015–21) over NC. The numbers in bold represent R values that are statistically significant at the 90% confidence level.
Figure 7 shows the percentage trend of DOD from the CMIP6 models, MERRA-2, and MODIS DOD over the whole of NC. Observations show that DOD increases during spring and decreases during summer. Under the four SSPs, MEM DOD increases in SSP1-2.6 and SSP5-8.5 and decreases in SSP1-2.6 and SSP3-7.0. During spring, the positive trend is captured by MERRA-2 and by CanESM5 (in SSP1-2.6 and SSP5-8.5), GFDL-ESM4 (in SSP1-2.6 and SSP5-8.5), INM-CM4-8 (in SSP1-2.6 and SSP5-8.5), INM-CM5-0 (in SSP3-7.0), MRI-ESM2-0 (in SSP2-4.5 and SSP5-8.5), and NorESM2-LM (in SSP1-2.6, SSP3-7.0 and SSP5-8.5), but with different magnitudes. During summer, the negative trend is captured by MERRA-2 and by CanESM5 (in SSP2-4.5 and SSP3-7.0), GFDL-ESM4 (in SSP1-2.6, SSP2-4.5 and SSP3-7.0), INM-CM4-8 (in SSP3-7.0), INM-CM5-0 (in SSP1-2.6 and SSP2-4.5), MRI-ESM2-0 (in SSP3-7.0 and SSP5-8.5), and NorESM2-LM (in SSP2-4.5 and SSP3-7.0), but with different magnitudes. Meanwhile, the trends in DOD over the different subregions and their comparison with observations are shown in Figs. S16 and S17. Observations show that DOD increases over GD and NCP and decreases over NWC and NEC during spring. The positive trends are captured by more models and the MEM in SSP1-2.6 and SSP5-8.5. The negative trends over NWC are captured by more models in SSP2-4.5, and the negative trends over NEC are captured by more models and the MEM in SSP3-7.0. During summer, DOD increases over NCP and decreases in other subregions. More models simulate a positive DOD trend over NCP in SSP1-2.6 (CanESM5, INM-CM4-8, INM-CM5-0, MRI-ESM2-0, and MEM), with MEM also simulating a weak positive trend in SSP1-2.6, SSP2-4.5, and SSP3-7.0. The negative trends during summer are captured by more models and the MEM in SSP2-4.5.
(a) Linear percentage trends in annual-mean DOD over NC from MODIS, MERRA-2, and the MEM under 4 SSPs in [a(1)],[b(1)] spring (MAM) and [a(2)],[b(2)] summer (JJA) during P3. The shading denotes the intermodel standard deviation of six CMIP6 models. The trends that are statistically significant at the 90% confidence level (p < 0.1) are labeled alongside the trend values. (b) Projected DOD trends in P3 under different SSPs. The stars signify trends that are statistically significant at the 90% confidence level.
Citation: Journal of Climate 37, 23; 10.1175/JCLI-D-23-0624.1
Figure 8 shows the percentage trends of WS over NWC and GD in spring and summer during P3 in CMIP6 models, MERRA-2, and observations. Meanwhile, Figs. S18 and S19 show the same but for LAI and MRSO. According to GLDAS WS, the positive trend in DOD over GD is dominated by increases in WS of 0.29% yr−1 during spring and 0.56% yr−1 during summer. The trends of LAI and MRSO are not significant, but the positive trends in LAI slow down compared to those in P2 and the MRSO decreased during P3. The positive changes in GD wind speed are captured by more models in SSP1-2.6 (INM-CM4-8, MRI-ESM2-0, NorESM2-LM, and the MEM during spring; and INM-CM4-8, INM-CM5-0, MRI-ESM2-0, and the MEM during summer). The negative changes in NWC wind speed are captured by more models in SSP2-4.5 during spring and by more models in SSP5-8.5 during summer. Under the four SSPs, the MEM simulates negative trends in both LAI and MRSO over GD during spring and summer, which can promote dust emissions, except for LAI in spring under SSP3-7.0. Furthermore, the MEM simulates positive trends in WS under SSP1-2.6, SSP2-4.5, and SSP3-7.0 during spring and under SSP1-2.6 and SSP3-7.0 during summer. The increases in MEM DOD under SSP5-8.5 over GD are mainly attributable to the simulated negative trend in LAI and MRSO. The decreases in MEM DOD under SSP3-7.0 are due to the simulated positive trend in LAI.
Linear percentage trends (% yr−1) of WS over [a(1)],[a(2)] NWC and [b(1)],[b(2)] GD in [a(1)],[b(1)] spring (MAM) and [a(2)],[b(2)] summer (JJA) during P3 in CMIP6 models, MERRA-2 and GLDAS. Stars signify trends that are statistically significant at the 90% confidence level.
Citation: Journal of Climate 37, 23; 10.1175/JCLI-D-23-0624.1
5. Discussion and conclusions
CMIP6 provides an important basis for studying historical variations and future projections of DOD and its impacts with global climate models developed by the world’s leading climate research institutions, but the biases in simulating DOD remain to be detailed. In this study, a site-scale DOD dataset for the period before 2002 was reconstructed based on MODIS DOD and site-scale dust-event frequency, and then, the DOD under SSP1-2.6, SSP2-4.5, SSP3-7.0, SSP5-8.5, and historical simulations in CMIP6 were evaluated based on MODIS DOD and the reconstructed DOD.
By comparing the spatial patterns of DOD between the observed data and the CMIP6 models, we found that the models and their MEM are capable of capturing the spatial pattern of DOD, albeit with considerable uncertainty and intermodel variability in magnitude. Over NC, the regional average DOD is underestimated by 56.09% during 1980–2001 and overestimated by 30.97% during 2002–14 in the MEM. The intermodel standard deviations are much greater than those of the MEM during 2002–14, suggesting large discrepancies among individual models. Geographically, the magnitudes of relative error are larger in the dustiest regions (NWC and GD). Among the four SSPs, the simulated DOD under SSP1-2.6 has the best spatial correlation with MODIS DOD over GD, but is significantly overestimated.
By comparing the temporal changes of DOD between the observations and the CMIP6 models, we found that in the historical period, the MEM captures the trends in DOD effectively during spring in P2, but the simulated trends are weaker than the observations. CESM2, CESM2-FV2, CESM2-WACCM, and CESM2-WACCM-FV2 captured more of the negative trend in DOD. Very few models accurately capture the interannual trends in DOD, which can mainly be attributed to the different trends and their contributions to dust evolution of the simulated WS, MRSO, and LAI. Under the four SSPs, more models and the MEM under SSP1-2.6 capture the positive trend of dust in the different subregions, which is mainly attributable to the positive changes in simulated WS over GD.
Given that the global climate models participating in CMIP6 are incapable of accurately simulating the contributions of meteorological drivers to dust variations, resulting in significant uncertainties in simulations of historical and future dust trends, further work is needed to improve the quantification of dust–climate interactions in models, and the detailed outputs of meteorological drivers. Meanwhile, as variations in dust and meteorological drivers are associated with large-scale atmospheric circulation, further studies are also needed to investigate the contributions of variations in atmospheric circulation simulated by climate models. Furthermore, based on the future projections of global climate models participating in CMIP6, many studies have projected the future variation in dust, but there is no agreement on the future trends in dust over NC. Wang and Cheng (2022) and Zhao et al. (2023) suggested that, under the different SSPs, the dust emission flux will continue to decrease, due to the variations in meteorological factors unfavorable for dust emissions, such as an increase in vegetation and a decrease in wind speed. However, the distribution and trend of drought index based on the outputs of CMIP6 climate models indicate that, under the different SSPs, the continuous increase in temperature and decrease in precipitation over dust-source regions will promote dust emissions (Mao et al. 2021). Considering the significant bias in dust projections output from CMIP6 models, we advocate that the bias in future projections should be corrected utilizing emerging constraints or ensemble empirical mode decomposition methods before projecting the changes in dust loading by the end of the twenty-first century from the model outputs of CMIP6 under the four SSPs.
Acknowledgments.
This research was supported by grants from the National Natural Science Foundation of China project (42030608), National Science Fund for Distinguished Young Scholars (41825011), National Natural Science Foundation of China project (42175153), National Key Research and Development Program of China (2023YFC3706305), Youth Innovation Team of China Meteorological Administration (CMA2024QN13), Third Xinjiang Scientific Expedition Program (2022xjkk0903), the Joint Fund of the National Natural Science Foundation of China and the China Meteorological Administration (U2242209), and Basic Research Fund of CAMS (2023Z021). The authors declare no competing interests.
Data availability statement.
DOD and meteorological drivers (WS, LAI, and MRSO) outputs from the models participating in CMIP6 are obtained from monthly https://esgf-node.llnl.gov/search/cmip6/. The MODIS aerosol optical property data (MOD04_L2) and NDVI (MOD13C2) are downloaded from Earth Data Search (https://search.earthdata.nasa.gov/search, last access: 6 November 2021), a web application developed by NASA’s Earth Observing System Data and Information System (EOSDIS). The ground-based observation data on the frequencies of dust events and WS are available from the National Meteorological Information Centre of the China Meteorological Administration (http://data.cma.cn/). GLDAS products and MERRA-2 datasets are obtained from the Goddard Earth Sciences Data and Information Services Center (GES DISC) (https://earthdata.nasa.gov/eosdis/daacs/gesdisc).
REFERENCES
Albani, S., and Coauthors, 2014: Improved dust representation in the Community Atmosphere Model. J. Adv. Model. Earth Syst., 6, 541–570, https://doi.org/10.1002/2013MS000279.
Ali, M. A., and Coauthors, 2022: Evaluation and comparison of CMIP6 models and MERRA-2 reanalysis AOD against satellite observations from 2000 to 2014 over China. Geosci. Front., 13, 101325, https://doi.org/10.1016/j.gsf.2021.101325.
An, L., and Coauthors, 2018: Temporal and spatial variations in sand and dust storm events in East Asia from 2007 to 2016: Relationships with surface conditions and climate change. Sci. Total Environ., 633, 452–462, https://doi.org/10.1016/j.scitotenv.2018.03.068.
Aryal, Y. N., and S. Evans, 2021: Global dust variability explained by drought sensitivity in CMIP6 models. J. Geophys. Res. Earth Surf., 126, e2021JF006073, https://doi.org/10.1029/2021JF006073.
Buchard, V., and Coauthors, 2017: The MERRA-2 aerosol reanalysis, 1980 onward. Part II: Evaluation and case studies. J. Climate, 30, 6851–6872, https://doi.org/10.1175/JCLI-D-16-0613.1.
Carvalho, D., S. Rafael, A. Monteiro, V. Rodrigues, M. Lopes, and A. Rocha, 2022: How well have CMIP3, CMIP5 and CMIP6 future climate projections portrayed the recently observed warming. Sci. Rep., 12, 11983, https://doi.org/10.1038/s41598-022-16264-6.
Celisse, A., and S. Robin, 2008: Nonparametric density estimation by exact leave-p-out cross-validation. Comput. Stat. Data Anal., 52, 2350–2368, https://doi.org/10.1016/j.csda.2007.10.002.
Cesnulyte, V., A. V. Lindfors, M. R. A. Pitkänen, K. E. J. Lehtinen, J.-J. Morcrette, and A. Arola, 2014: Comparing ECMWF AOD with AERONET observations at visible and UV wavelengths. Atmos. Chem. Phys., 14, 593–608, https://doi.org/10.5194/acp-14-593-2014.
Chen, B., P. Zhang, B. Zhang, R. Jia, Z. Zhang, T. Wang, and T. Zhou, 2014: An overview of passive and active dust detection methods using satellite measurements. J. Meteor. Res., 28, 1029–1040, https://doi.org/10.1007/s13351-014-4032-4.
Chen, S., and Coauthors, 2017a: Emission, transport, and radiative effects of mineral dust from the Taklimakan and Gobi Deserts: Comparison of measurements and model results. Atmos. Chem. Phys., 17, 2401–2421, https://doi.org/10.5194/acp-17-2401-2017.
Chen, S., and Coauthors, 2017b: An overview of mineral dust modeling over East Asia. J. Meteor. Res., 31, 633–653, https://doi.org/10.1007/s13351-017-6142-2.
Chen, Y., K. Yang, J. Qin, L. Zhao, W. Tang, and M. Han, 2013: Evaluation of AMSR-E retrievals and GLDAS simulations against observations of a soil moisture network on the central Tibetan Plateau. J. Geophys. Res. Atmos., 118, 4466–4475, https://doi.org/10.1002/jgrd.50301.
Colarco, P., A. Da Silva, M. Chin, and T. Diehl, 2010: Online simulations of global aerosol distributions in the NASA GEOS-4 model and comparisons to satellite and ground-based aerosol optical depth. J. Geophys. Res., 115, D14207, https://doi.org/10.1029/2009JD012820.
Danabasoglu, G., and Coauthors, 2020: The Community Earth System Model Version 2 (CESM2). J. Adv. Model. Earth Syst., 12, e2019MS001916, https://doi.org/10.1029/2019MS001916.
Didan, K., A. Barreto Munoz, R. Solano, and A. Huete, 2015: MODIS Vegetation Index User’s Guide (MOD13 series), https://vip.arizona.edu.
Dunne, J. P., and Coauthors, 2020: The GFDL Earth System Model Version 4.1 (GFDL-ESM 4.1): Overall coupled model description and simulation characteristics. J. Adv. Model. Earth Syst., 12, e2019MS002015, https://doi.org/10.1029/2019MS002015.
Eck, T. F., B. N. Holben, J. S. Reid, O. Dubovik, A. Smirnov, N. T. O’Neill, I. Slutsker, and S. Kinne, 1999: Wavelength dependence of the optical depth of biomass burning, urban, and desert dust aerosols. J. Geophys. Res., 104, 31 333–31 349, https://doi.org/10.1029/1999JD900923.
Eyring, V., S. Bony, G. A. Meehl, C. A. Senior, B. Stevens, R. J. Stouffer, and K. E. Taylor, 2016: Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization. Geosci. Model Dev., 9, 1937–1958, https://doi.org/10.5194/gmd-9-1937-2016.
Fairlie, T., D. J. Jacob, and R. J. Park, 2007: The impact of transpacific transport of mineral dust in the United States. Atmos. Environ., 41, 1251–1266, https://doi.org/10.1016/j.atmosenv.2006.09.048.
Gao, H., and R. Washington, 2009: The spatial and temporal characteristics of TOMS AI over the Tarim Basin, China. Atmos. Environ., 43, 1106–1115, https://doi.org/10.1016/j.atmosenv.2008.11.013.
Gelaro, R., and Coauthors, 2017: The Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2). J. Climate, 30, 5419–5454, https://doi.org/10.1175/JCLI-D-16-0758.1.
Ginoux, P., J. M. Prospero, T. E. Gill, N. C. Hsu, and M. Zhao, 2012: Global-scale attribution of anthropogenic and natural dust sources and their emission rates based on MODIS Deep Blue aerosol products. Rev. Geophys., 50, RG3005, https://doi.org/10.1029/2012RG000388.
Gkikas, A., and Coauthors, 2022: Quantification of the dust optical depth across spatiotemporal scales with the MIDAS global dataset (2003–2017). Atmos. Chem. Phys., 22, 3553–3578, https://doi.org/10.5194/acp-22-3553-2022.
Gui, K., and Coauthors, 2021: Three-dimensional climatology, trends, and meteorological drivers of global and regional tropospheric type-dependent aerosols: Insights from 13 years (2007–2019) of CALIOP observations. Atmos. Chem. Phys., 21, 15 309–15 336, https://doi.org/10.5194/acp-21-15309-2021.
Gui, K., and Coauthors, 2022: Record-breaking dust loading during two mega dust storm events over northern China in March 2021: Aerosol optical and radiative properties and meteorological drivers. Atmos. Chem. Phys., 22, 7905–7932, https://doi.org/10.5194/acp-22-7905-2022.
Gui, K., and Coauthors, 2023: Quantifying the contribution of local drivers to observed weakening of spring dust storm frequency over northern China (1982–2017). Sci. Total Environ., 894, 164923, https://doi.org/10.1016/j.scitotenv.2023.164923.
Hsu, N. C., M.-J. Jeong, C. Bettenhausen, A. M. Sayer, R. Hansell, C. S. Seftor, J. Huang, and S.-C. Tsay, 2013: Enhanced Deep Blue aerosol retrieval algorithm: The second generation. J. Geophys. Res. Atmos., 118, 9296–9315, https://doi.org/10.1002/jgrd.50712.
Huang, J. P., J. J. Liu, B. Chen, and S. L. Nasiri, 2015: Detection of anthropogenic dust using CALIPSO lidar measurements. Atmos. Chem. Phys., 15, 11 653–11 665, https://doi.org/10.5194/acp-15-11653-2015.
Iino, N., K. Kinoshita, A. C. Tupper, and T. Yano, 2004: Detection of Asian dust aerosols using meteorological satellite data and suspended particulate matter concentrations. Atmos. Environ., 38, 6999–7008, https://doi.org/10.1016/j.atmosenv.2004.02.070.
Kai, K., and Coauthors, 2019: Large-scale dust event in East Asia, as revealed by the Himawari-8 DUST RGB, lidar network observations, and field survey. E3S Web Conf., 99, 8–11, https://doi.org/10.1051/e3sconf/20199901004.
Kok, J. F., and Coauthors, 2017: Smaller desert dust cooling effect estimated from analysis of dust size and abundance. Nat. Geosci., 10, 274–278, https://doi.org/10.1038/ngeo2912.
Kok, J. F., and Coauthors, 2021: Contribution of the world’s main dust source regions to the global cycle of desert dust. Atmos. Chem. Phys., 21, 8169–8193, https://doi.org/10.5194/acp-21-8169-2021.
Kok, J. F., T. Storelvmo, V. A. Karydis, A. A. Adebiyi, N. M. Mahowald, A. T. Evan, C. He, and D. M. Leung, 2023: Mineral dust aerosol impacts on global climate and climate change. Nat. Rev. Earth Environ., 4, 71–86, https://doi.org/10.1038/s43017-022-00379-5.
Li, X., Y. Liu, M. Wang, Y. Jiang, and X. Dong, 2021: Assessment of the Coupled Model Intercomparison Project phase 6 (CMIP6) model performance in simulating the spatial-temporal variation of aerosol optical depth over Eastern Central China. Atmos. Res., 261, 105747, https://doi.org/10.1016/j.atmosres.2021.105747.
Liang, P., B. Chen, X. Yang, Q. Liu, A. Li, L. Mackenzie, and D. Zhang, 2022: Revealing the dust transport processes of the 2021 mega dust storm event in northern China. Sci. Bull., 67, 21–24, https://doi.org/10.1016/j.scib.2021.08.014.
Liu, S., 2019: Leave-p-out cross-validation test for uncertain Verhulst-Pearl model with imprecise observations. IEEE Access, 7, 131 705–131 709, https://doi.org/10.1109/ACCESS.2019.2939386.
Liu, Y., Y. Liu, and W. Wang, 2019: Inter-comparison of satellite-retrieved and Global Land Data Assimilation System-simulated soil moisture datasets for global drought analysis. Remote Sens. Environ., 220, 1–18, https://doi.org/10.1016/j.rse.2018.10.026.
Mahowald, N. M., D. R. Muhs, S. Levis, P. J. Rasch, M. Yoshioka, C. S. Zender, and C. Luo, 2006: Change in atmospheric mineral aerosols in response to climate: Last glacial period, preindustrial, modern, and doubled carbon dioxide climates. J. Geophys. Res., 111, D10202, https://doi.org/10.1029/2005JD006653.
Mao, R., D.-Y. Gong, S.-J. Kim, Q. Zong, X. Feng, and X.-X. Zhang, 2021: Increasing spring dust storms in the future over the Taklimakan Desert, Northwest China: Implications from changes in circulation pattern frequency in CMIP6. Environ. Res. Commun., 3, 111002, https://doi.org/10.1088/2515-7620/ac37ee.
Marticorena, B., and G. Bergametti, 1995: Modeling the atmospheric dust cycle: 1. Design of a soil-derived dust emission scheme. J. Geophys. Res., 100, 16 415–16 430, https://doi.org/10.1029/95JD00690.
National Weather Bureau of China, 1979: Criterion of Surface Meteorological Observation (in Chinese). Meteorological Press, 7 pp.
O’Neill, B. C., and Coauthors, 2016: The Scenario Model Intercomparison Project (ScenarioMIP) for CMIP6. Geosci. Model Dev., 9, 3461–3482, https://doi.org/10.5194/gmd-9-3461-2016.
O’Neill, N. T., T. F. Eck, A. Smirnov, B. N. Holben, and S. Thulasiraman, 2003: Spectral discrimination of coarse and fine mode optical depth. J. Geophys. Res., 108, 4559, https://doi.org/10.1029/2002JD002975.
Petrie, R., and Coauthors, 2021: Coordinating an operational data distribution network for CMIP6 data. Geosci. Model Dev., 14, 629–644, https://doi.org/10.5194/gmd-14-629-2021.
Proestakis, E., and Coauthors, 2018: Nine-year spatial and temporal evolution of desert dust aerosols over South and East Asia as revealed by CALIOP. Atmos. Chem. Phys., 18, 1337–1362, https://doi.org/10.5194/acp-18-1337-2018.
Pu, B., and P. Ginoux, 2018: How reliable are CMIP5 models in simulating dust optical depth? Atmos. Chem. Phys., 18, 12 491–12 510, https://doi.org/10.5194/acp-18-12491-2018.
Pu, B., and Q. Jin, 2021: A record-breaking trans-Atlantic African dust plume associated with atmospheric circulation extremes in June 2020. Bull. Amer. Meteor. Soc., 102, E1340–E1356, https://doi.org/10.1175/BAMS-D-21-0014.1.
Pu, B., and Coauthors, 2020: Retrieving the global distribution of the threshold of wind erosion from satellite data and implementing it into the Geophysical Fluid Dynamics Laboratory land–atmosphere model (GFDL AM4.0/LM4.0). Atmos. Chem. Phys., 20, 55–81, https://doi.org/10.5194/acp-20-55-2020.
Qian, W., X. Tang, and L. Quan, 2004: Regional characteristics of dust storms in China. Atmos. Environ., 38, 4895–4907, https://doi.org/10.1016/j.atmosenv.2004.05.038.
Ren, X., Y. Xie, and G. Ren, 2017: Spatiotemporal characteristics of TOMS-based dust aerosol optical depth in northern China during 1978–2005. Earth Sci. Inform., 10, 41–53, https://doi.org/10.1007/s12145-016-0277-z.
Ridley, A. D., L. C. Heald, F. J. Kok, and C. Zhao, 2016: An observationally constrained estimate of global dust aerosol optical depth. Atmos. Chem. Phys., 16, 15 097–15 117, https://doi.org/10.5194/acp-16-15097-2016.
Rodell, M., and Coauthors, 2004: The Global Land Data Assimilation System. Bull. Amer. Meteor. Soc., 85, 381–394, https://doi.org/10.1175/BAMS-85-3-381.
Sayer, A. M., N. C. Hsu, C. Bettenhausen, and M.-J. Jeong, 2013: Validation and uncertainty estimates for MODIS Collection 6 “Deep Blue” aerosol data. J. Geophys. Res. Atmos., 118, 7864–7872, https://doi.org/10.1002/jgrd.50600.
Sayer, A. M., N. C. Hsu, J. Lee, W. V. Kim, and S. T. Dutcher, 2019: Validation, stability, and consistency of MODIS collection 6.1 and VIIRS version 1 Deep Blue aerosol data over land. J. Geophys. Res. Atmos., 124, 4658–4688, https://doi.org/10.1029/2018JD029598.
Seland, Ø., and Coauthors, 2020: Overview of the Norwegian Earth System Model (NorESM2) and key climate response of CMIP6 DECK, historical, and scenario simulations. Geosci. Model Dev., 13, 6165–6200, https://doi.org/10.5194/gmd-13-6165-2020.
Shao, Y., and C. H. Dong, 2006: A review on East Asian dust storm climate, modelling and monitoring. Global Planet. Change, 52 (1–4), 1–22, https://doi.org/10.1016/j.gloplacha.2006.02.011.
Skrivanek, S., and M. Com Llc, 2009: The use of dummy variables in regression analysis what is a dummy variable?, http://www.moresteam.com.
Song, Q., Z. Zhang, H. Yu, P. Ginoux, and J. Shen, 2021: Global dust optical depth climatology derived from CALIOP and MODIS aerosol retrievals on decadal timescales: Regional and interannual variability. Atmos. Chem. Phys., 21, 13 369–13 395, https://doi.org/10.5194/acp-21-13369-2021.
Swart, N. C., and Coauthors, 2019: The Canadian Earth System Model version 5 (CanESM5.0.3). Geosci. Model Dev., 12, 4823–4873, https://doi.org/10.5194/gmd-12-4823-2019.
Tan, S.-C., J. Li, H. Che, B. Chen, and H. Wang, 2017: Transport of East Asian dust storms to the marginal seas of China and the southern North Pacific in spring 2010. Atmos. Environ., 148, 316–328, https://doi.org/10.1016/j.atmosenv.2016.10.054.
Tanaka, T. Y., and M. Chiba, 2005: Global simulation of dust aerosol with a chemical transport model, MASINGAR. J. Meteor. Soc. Japan, 83A, 255–278, https://doi.org/10.2151/jmsj.83A.255.
Tanaka, T. Y., and M. Chiba, 2006: A numerical study of the contributions of dust source regions to the global dust budget. Global Planet. Change, 52, 88–104, https://doi.org/10.1016/j.gloplacha.2006.02.002.
Thornhill, G., and Coauthors, 2021: Climate-driven chemistry and aerosol feedbacks in CMIP6 Earth system models. Atmos. Chem. Phys., 21, 1105–1126, https://doi.org/10.5194/acp-21-1105-2021.
Tilmes, S., and Coauthors, 2019: Climate forcing and trends of organic aerosols in the Community Earth System Model (CESM2). J. Adv. Model. Earth Syst., 11, 4323–4351, https://doi.org/10.1029/2019MS001827.
Tokarska, K. B., M. B. Stolpe, S. Sippel, E. M. Fischer, C. J. Smith, F. Lehner, and R. Knutti, 2020: Past warming trend constrains future warming in CMIP6 models. Science, 6, eaaz9549, https://doi.org/10.1126/sciadv.aaz9549.
Tranmer, M., J. Murphy, M. Elliot, and M. Pampaka, 2020: Multiple Linear Regression. 2nd ed. Cathie Marsh Institute Working Paper 2020-01, 59 pp.
Vermote, E. F., N. Z. El Saleous, and C. O. Justice, 2002: Atmospheric correction of MODIS data in the visible to middle infrared: First results. Remote Sens. Environ., 83, 97–111, https://doi.org/10.1016/S0034-4257(02)00089-5.
Volodin, E., 2021: The mechanisms of cloudiness evolution responsible for equilibrium climate sensitivity in climate model INM-CM4-8. Geophys. Res. Lett., 48, e2021GL096204, https://doi.org/10.1029/2021GL096204.
Volodin, E., and S. V. Kostrykin, 2016: The aerosol module in the INM RAS climate model. Russ. Meteor. Hydrol., 41, 519–528, https://doi.org/10.3103/S106837391608001X.
Voss, K. K., and A. T. Evan, 2020: A new satellite-based global climatology of dust aerosol optical depth. J. Appl. Meteor. Climate, 83–102, https://doi.org/10.1175/JAMC-D-19-0194.1.
Wang, S., J. Wang, Z. Zhou, and K. Shang, 2005: Regional characteristics of three kinds of dust storm events in China. Atmos. Environ., 39, 509–520, https://doi.org/10.1016/j.atmosenv.2004.09.033.
Wang, W., and T. Cheng, 2022: Decrease trend of East Asia dust during the 21st century in CMIP6. ESS Open Arch., https://doi.org/10.1002/essoar.10512578.1.
Wang, X., Z. Dong, J. Zhang, and L. Liu, 2004: Modern dust storms in China: An overview. J. Arid Environ., 58, 559–574, https://doi.org/10.1016/j.jaridenv.2003.11.009.
Wang, X., J. Liu, H. Che, F. Ji, and J. Liu, 2018: Spatial and temporal evolution of natural and anthropogenic dust events over northern China. Sci. Rep., 8, 2141, https://doi.org/10.1038/s41598-018-20382-5.
Wang, X., and Coauthors, 2021: Spatio-temporal trends of dust emissions triggered by desertification in China. Catena, 200, 105160, https://doi.org/10.1016/j.catena.2021.105160.
Wang, Y. Q., X. Y. Zhang, S. L. Gong, C. H. Zhou, X. Q. Hu, H. L. Liu, T. Niu, and Y. Q. Yang, 2008: Surface observation of sand and dust storm in East Asia and its application in CUACE/dust. Atmos. Chem. Phys., 8, 545–553, https://doi.org/10.5194/acp-8-545-2008.
Wu, C., Z. Lin, Y. Shao, X. Liu, and Y. Li, 2022: Drivers of recent decline in dust activity over East Asia. Nat. Commun., 13, 7105, https://doi.org/10.1038/s41467-022-34823-3.
Yao, W., H. Che, K. Gui, Y. Wang, and X. Zhang, 2020: Can MERRA-2 reanalysis data reproduce the three-dimensional evolution characteristics of a typical dust process in East Asia? A case study of the dust event in May 2017. Remote Sens., 12, 902, https://doi.org/10.3390/rs12060902.
Yao, W., K. Gui, Y. Wang, H. Che, and X. Zhang, 2021: Identifying the dominant local factors of 2000–2019 changes in dust loading over East Asia. Sci. Total Environ., 777, 146064, https://doi.org/10.1016/j.scitotenv.2021.146064.
Yoon, J., W. Von Hoyningen-Huene, M. Vountas, and J. P. Burrows, 2011: Analysis of linear long-term trend of aerosol optical thickness derived from SeaWiFS using BAER over Europe and South China. Atmos. Chem. Phys., 11, 12 149–12 167, https://doi.org/10.5194/acp-11-12149-2011.
Yu, Y., and P. Ginoux, 2022: Enhanced dust emission following large wildfires due to vegetation disturbance. Nat. Geosci., 15, 878–884, https://doi.org/10.1038/s41561-022-01046-6.
Yukimoto, S., and Coauthors, 2019: The meteorological research institute Earth system model version 2.0, MRI-ESM2.0: Description and basic evaluation of the physical component. J. Meteor. Soc. Japan, 97, 931–965, https://doi.org/10.2151/jmsj.2019-051.
Zender, C. S., R. L. Miller, and I. Tegen, 2004: Quantifying mineral dust mass budgets: Terminology, constraints, and current estimates. Earth Space Sci., 85, 509–512, https://doi.org/10.1029/2004EO480002.
Zhang, X. Y., and Coauthors, 2003: Characterization of soil dust aerosol in China and its transport and distribution during 2001 ACE-Asia: 1. Network observations. J. Geophys. Res., 108, 4261, https://doi.org/10.1029/2002JD002632.
Zhao, A., C. L. Ryder, and L. J. Wilcox, 2022: How well do the CMIP6 models simulate dust aerosols? Atmos. Chem. Phys., 22, 2095–2119, https://doi.org/10.5194/acp-22-2095-2022.
Zhao, Y., and Coauthors, 2023: Multi-model ensemble projection of the global dust cycle by the end of 21st century using the Coupled Model Intercomparison Project version 6 data. Atmos. Chem. Phys., 23, 7823–7838, https://doi.org/10.5194/acp-23-7823-2023.
