Sharp decline of dust events induces regional wetting over arid and semi-arid Northwest China in the NCAR Community atmosphere model

Dust event data are directly derived from strong sand-dust storm in China and the Supporting Dataset provided by the China Meteorological Administration (CMA). According to the observation manual prepared by the World Meteorological Organization, the present weather code (ww) related to dust weather are ww = 06, 07, 08, 09, 30–35, 98. These dust event observations are classified into three categories based on the intensities (Shao et al 2013): dust storm event (DS), blowing dust event (BD), and floating dust event (FD). According to observational results about dust mass concentration differences among these three categories, it was found that if a DS occurs at a station, the mass concentration is almost 3 times greater than that of BD, or 9 times greater than that of FD (Wang et al 2008). To consider the weight coefficients for three categories in describing dust activity at the site, the overall frequency of all dust activities can be measured by dust event index (DI) defined as (Wang et al 2008, Lee et al 2014),

$$\begin{equation}{\text{DI}} = 9 \times {\text{DS}} + 3 \times {\text{BD}} + {\text{FD}}.\end{equation} \tag{ 1 }$$

The surface-gridded precipitation data are from Monthly Dataset of surface climate data in China from CMA. The spatial resolution of precipitation dataset is 0.5° × 0.5° from 1961 to present based on 2472 station observations in China. According to Alexander (2016), for a model grid point or given station, the extreme precipitation events are defined when daily precipitation is larger than the 95th (or 99th) percentile value of all rainy days (daily precipitation >0.1 mm). Extreme precipitation (R95p or R99p) is defined as the sum of precipitation during all events exceeding the threshold.

We use annual precipitation data of historical simulations from 14 GCMs in table S1 from the Detection and Attribution Model Intercomparison Project (DAMIP, Gillett et al 2016) in CMIP6 (Eyring et al 2016). Here, we used the historical all forcing simulations including all natural and anthropogenic forcings (ALL) and three single forcing simulations in DAMIP. The single forcing experiments include the well-mixed GHG only and AER only, as well as natural forcing only simulations (NAT), designed to quantify the relative contributions of different drivers to regional and global climate change. The first member for individual models is chosen for fair comparisons, although CMIP6 GCMs provide large ensemble members. All the simulated data are interpolated to a 2.5° × 2.5° horizontal resolution by bilinear interpolation.

CAM5 is the atmospheric component of the Community Earth System Model version 1 (Neale et al 2010). The physical parameterizations of dust cycle processes in CAM5 are directly from dust entrainment and deposition model (Mahowald et al 2006). We use an improved size distribution for dust emissions based on brittle fragmentation theory (Kok 2011, Albani et al 2014). CAM5 adopts a two-moment stratiform cloud microphysics scheme, which can predict the mixing ratios and number concentrations of ice crystal and cloud droplet (Morrison and Gettelman 2008). The complete set of physics packages in CAM5 can be used to effectively investigate the aerosol direct and indirect effects (Ghan et al 2012). As suggested by Sagoo and Storelvmo (2017), we utilize an updated ice nucleation parameterization instead of the standard scheme (Meyers et al 1992), which enables the INP concentrations to be calculated based on the function of large dust particle concentration in addition to air temperature (DeMott et al 2015). This version reduces the efficiency of Wegner–Bergeron–Findeisen process to improve the cloud phase partitioning in mixed-phase clouds (Tan and Storelvmo 2016, Tan et al 2016). The deep convective cloud process in CAM5 is parameterized by the Zhang–McFarlane deep convection scheme in which the convective precipitation depends on properties that are closely related to local atmospheric instability (Zhang and McFarlane 1995).

Here, sensitivity experiments with different dust emission amounts in CAM5 are conducted to represent the effect of declining dust events. This GCM utilizes the finite-volume dynamical core with 0.9° × 1.25° in the horizontal resolution and 30 vertical levels. All the numerical experiments are performed with fixed monthly mean sea surface temperature and sea-ice concentrations during the whole simulated period. Two sets of numerical experiments including 25 year simulations are conducted: one with dust emissions (denoted dust experiment) and the other without dust emissions (nodust experiment). We further conduct an additional experiment with dusts but without dust-ice cloud interactions (DCI) in mixed-phase clouds (noice experiment). All the simulations are analyzed for the last 20 years of the 25 year numerical experiments. The difference between noice and dust experiments (noice-dust) represents DCI effect induced by decreasing dusts; the difference between nodust and noice experiments (nodust-noice) indicates dust direct effect (DDE); and the change between nodust and dust experiments (nodust–dust) represents the dust total effect (DTE = DCI + DDE). In addition to the extreme Nodust experiment, one sensitivity experiment is also conducted with half dust emissions (halfdust experiment). The dust emission factor, labeled by EF, can be used to change the amount of dust emissions. The default value of EF is 1/0.35. A value of EF = 1/0.7 is used to decrease the amount of dust emissions by half.

Figure 1 summarizes inter-annual variations of DI and precipitation over NWC during 1961–2013. It shows a significant decline of DI over NWC (figure 1(a)) with −3.88 yr⁻¹ (figure 1(b)). Furthermore, figure S1 shows a positive-to-negative transition of DI trend from 1977 to 1978, and a much steeper decreasing trend of DI since the late 1970s (−5.28 yr⁻¹), indicating a sharp decline of NWC dust events during this period, mainly due to reductions of surface wind speeds (Wu et al 2022). Similar findings were previously reported, and related to the Arctic climate changes such as the Arctic amplification and loss of Arctic sea ice associated with global warming (Gong et al 2006, Zhang et al 2006, Wu et al 2018, Liu et al 2020, Shang and Liu 2020). Correspondingly, the NWC annual precipitation exhibited a much larger increasing trend (0.93 mm yr⁻¹) since the late 1970s, compared to the trend of 1961–2013 (0.65 mm yr⁻¹). These concurrent observational changes in dust events and precipitation over NWC indicate that the sharp decline of dust events may be a possible driver of the regional wetting over recent decades. Additionally, historical simulations show that GCMs significantly underestimate the NWC wetting (Chen and Frauenfeld 2014, Wu et al 2018, Xin et al 2020). We also show the absolute (figure 1(e)) and relative trend of precipitation (figure 1(f)) for a comparison between the observations and GCMs in CMIP6 (table S1), indicating large underestimation of precipitation trend, especially for relative trend. The underestimation suggests other physical mechanisms to further drive this regional wetting. Observational evidence shows a south-flood-north-drought pattern over eastern China during recent decades (e.g. Wang 2001, Zhou et al 2009), which is also shown in figure 1(c). Many previous studies have mainly attributed this dipole pattern to weakening of East Asian summer monsoon induced by anthropogenic aerosols and internal forcings (Zhou et al 2009, Wang 2001, Song et al 2014, Xie et al 2016).

Zoom In Zoom Out Reset image size

To access the climate response to dusts, we firstly evaluate the spatial distribution and seasonal variations of climatology precipitation and dust burden over NWC against observations (or reanalysis) in figure S2. The CAM5 can generally capture the spatial distribution and seasonal variations of precipitation and dust burden. However, the CAM5 overestimates the regional precipitation and dust burden (figure S2), compared with GPCP precipitation (or MERRA-2 reanalysis of dust burden). Figure 2 further shows changes in annual precipitation induced by decreasing dusts over NWC, based on these sensitivity experiments with different dust emission amounts. The result indicates that decreasing dusts significantly increase annual precipitation over NWC in figure 2(a) with a regionally averaged value of 2.30 mm per month (almost 9% in table 1), with the largest percentage increase in the center of NWC in figure 2(b). The increased NWC precipitation is attributed mainly to DCI-induced change in figure 2(c), whereas the contribution of DDE is negligible (also see figure S3). Further inspection shows that the precipitation increase is dominated by DCI-induced change in convective precipitation with 2.65 mm per month (29.56% in table 1). Accordingly, extreme precipitation is also enhanced for R95p with 13.46% and R99p with 17.88% in table 1. Seasonal differences in figure S4 further substantiate that DCI-induced convective precipitation mainly occurs during March–May (MAM) and June–August (JJA), due to larger dust burden in these two seasons (figure S2(f)). These results support observational findings that significant NWC wetting is attributed mainly to the convective precipitation increase in boreal summer (Zhou et al 2010, Han et al 2016). In addition, the sensitivity experiment with half dust emissions also shows an increase in annual precipitation, especially convective precipitation (not shown). Therefore, decreasing dusts can significantly enhance the regional convective precipitation over NWC, and further decreases dust events by increasing regional soil moisture and vegetation cover (Liu et al 2004, Wu et al 2022), thus resulting in a positive feedback loop between reduced dusts and increased precipitation.

Zoom In Zoom Out Reset image size

To better understand physical mechanisms of regional precipitation change, we examine changes in cloud microphysical and macrophysical properties. Derived from Dust experiment, figure 3(a) presents the spatial pattern of dust burden with higher dust burden over the NWC region, owing to existence of the Taklamakan and Gobi deserts. Dust aerosols can effectively act as INPs in mixed-phased clouds to alter ice cloud microphysical properties (DeMott et al 2010, 2015, Kanji et al 2017, Tobo et al 2019). Decreasing dusts significantly reduce the ice number concentration (figure 3(b)), and increases ice effective radius (figure 3(c)) over East Asia, especially over NWC. Fewer but larger ice crystals induced by less INPs increase the solid hydrometeor fall velocity and precipitation efficiency, resulting in significant decrease in ice water path (figure 3(d)). However, fewer INPs lead to an increase in liquid water path (figure S5), mainly because of the weakened Wegner–Bergeron–Findeisen process in mixed-phase clouds (Lohmann and Feichter 2005, Shi and Liu 2019). Compared to the ice-phase hydrometeors, the changes in liquid droplet number concentration and effective radius are insignificant (figure S5) due to a limited ability of dust particles to serve as droplet nucleating particles (Kumar et al 1995). Dust-induced changes in the cloud microphysical properties further influence cloud macrophysical properties in middle-to-high attitudes through affecting the precipitation efficiency and cloud lifetime (Sagoo and Storelvmo 2017). There are significant decreases in cloud fraction (figure 4(a)) and cloud optical depth (figure 4(b)) in mid-high attitudes through DCI induced by decreasing dusts. Compared to DCI, the changes in cloud fraction and cloud optical depth through DDE are considerably smaller (figures 4(a) and (b)), mainly because DDE does not directly affect the cloud microphysical properties.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Furthermore, we show the vertical profile of changes in air temperature induced by decreasing dusts over NWC through DCI and DDE in figure 4(c). A decrease in clouds on mid-high altitudes induced by decreasing dusts through DCI warms the low-level atmosphere (figure 4(c)) due to enhanced shortwave heating rate (figure S6(a)). Furthermore, the cloud change cools the mid-level atmosphere (figure 4(c)) through decreasing longwave heating rate (figure S6(a)). The low-level warming and mid-level cooling through DCI significantly enhance the vertical temperature gradient (−0.51 °C) between 400 hPa and surface in figure 4(d). We also calculate the mean regional convective instability, defined as the difference of equivalent potential temperature between 400 hPa and surface. The negative (positive) value of regional convective instability indicates the atmospheric instability (stability) associated with convective precipitation. Decreasing dusts result in a significant increase in convective instability, or a more unstable atmosphere owing to enhanced dipole structure of low-level warming and mid-level cooling. The enhanced regional convective instability leads to an increase of 2.83 ± 3.07 J kg⁻¹ in convective available potential energy (CAPE) and convective mass fluxes of 0.21 ± 0.14 g m⁻² s⁻¹. These enhancements in CAPE and convective mass fluxes significantly increase the precipitation, especially convective precipitation and extreme precipitation in table 1. DDE shows insignificant and smaller changes in vertical temperature gradient, convective instability, CAPE and convective mass fluxes in figure 4(d), leading to insignificant change in precipitation (figure 2). Due to considerably larger precipitation increases in MAM and JJA, we also show the changes in these climate variables induced by dusts during the two seasons (figure S7), which is consistent with annual mean results in figure 4(d). Based on these results, we propose that the effect of decreasing dusts through enhancing atmospheric instability plays the non-ignored role in the convective invigoration and the increase of NWC precipitation.