In this paper, we focus on the dust belt extending from North Africa through the Middle East and Central Asia to the deserts in western East Asia (Fig. 1). The Saharan dust belt (largely over 0–35∘ N, 16∘ W–25∘ E) is the world's largest source of aeolian desert dust aerosols, with annual emissions of 400–700 × 106 t of dust aerosols (D'Almeida, 1986; Schütz, 1980; Swap et al., 1996). There are two major dust sources within the Sahara – the Bodélé Depression in Chad and an area covering eastern Mauritania, western Mali, and southern Algeria (e.g., Middleton and Goudie, 2001; Schepanski et al., 2007; Ginoux et al., 2012; Yu et al., 2018).

The Middle East dust belt (roughly 13–38∘ N, 25–60∘ E) is the world's second largest dust source (Prospero et al., 2002; Goudie and Middleton, 2006; Huneeus et al., 2011; Kok et al., 2021). The Middle East and South Asian dust sources cover Sudan, the Arabian Peninsula, parts of Iran and Afghanistan, and Pakistan (Rezazadeh et al., 2013; Ginoux et al., 2012; Yu et al., 2018). The East Asian dust belt (25–50∘ N, 60–130∘ E) mainly includes the Taklamakan and the Gobi deserts (Prospero et al., 2002; Zhang et al., 2003; Ginoux et al., 2012) and is estimated to account for about 3 %–11 % of global dust emissions (Tanaka and Chiba, 2006). We did not consider the Gobi in our analysis due to a large area of missing data in IASI.

This study mainly uses the 10 µm DOD retrieved from IASI (LMD version) as the primary dataset to understand the day–night differences in DOD and dust plume layer height in the dust belt, along with ground-based observations. Results from IASI are compared with aerosol products from reanalyses. Meteorological variables from reanalysis and stations are used to examine their influences on the day–night differences in dust aerosols. All the datasets used in this study are summarized in Table 1.

IASI daytime 10 µm DOD (LMD version) has been validated against AERONET solar CAOD by Capelle et al. (2014, 2018) at some selected AERONET sites over the land and ocean for 2007–2016. In this work, we extend the previous daytime validation by including nighttime retrievals over the dust belt and to a longer period from 2007 to 2020. To compare IASI DOD with AERONET CAOD, we first sample AERONET solar and lunar CAOD within ±30 min of the IASI overpass time and IASI DOD pixels within a radius of 30 km from the AERONET sites. In total, 22 462 and 944 AERONET–IASI matchups for daytime and nighttime, respectively, are used in this study. Next, we convert IASI DOD at 10 µm in the IR band to 500 nm in the visible band (VIS) to be consistent with AERONET CAOD at 500 nm. An accurate conversion requires detailed information of the refractive index, size distribution, and the effective radius of dust particles (Capelle et al., 2014), which are usually not available over a large domain. Previously, Peyridieu et al. (2013) and Capelle et al. (2014, 2018) compared IASI DOD with AERONET station data by scaling AERONET AOD (550 nm) or CAOD (500 nm) to the equivalent 10 µm DOD using an empirically determined IR / VIS ratio at each AERONET site. Here, we follow a similar approach. At each AERONET station, the IR / VIS ratio is determined by regressing AERONET CAOD onto IASI DOD, with the slope of the regression being the IR / VIS ratio. However, the quality of such a linear fit depends on the sample size of IASI–AERONET collocations (Capelle et al., 2014). To prevent the ratios from being biased by the sample size, we exclude sites with fewer than 100 IASI–AERONET collocations for solar observations and 60 for lunar observations. Out of the 46 AERONET solar sites considered, only 5 sites (CATUC_Bamenda, Zinder_Airport, Banizoumbou, LAMTO-STATION, and NAM_CO) were excluded, whereas 4 out of the 11 lunar sites (Koforidua_ANUC, CATUC_Bamenda, Teide, and DEWA_ResearchCentre) were also excluded (see Tables 2 and 3). We found a mean IR / VIS ratio of ∼ 0.62 ranging from 0.31 to 2.06 for solar measurements and ∼ 0.57 ranging from 0.26 to 1.23 for lunar observations. A constant IR / VIS ratio of 0.60 (approximated by taking the mean of 0.62 for solar and 0.57 for lunar) is used to scale all IASI DOD from IR to VIS equivalent DOD at 500 nm for both data validation and the day–night difference analysis. Although the simple conversion method used here may lead to some uncertainties in the magnitude of the converted 500 nm IASI DOD, we found the calculated ratios to be largely within the range of empirically estimated IR / VIS ratios by Peyridieu et al. (2013) and Capelle et al. (2014, 2018) and largely consistent with the VIS / IR ratios used to convert IASI DOD (e.g., 1.54 by H. Yu et al., 2019, and 2.0 by Clarisse et al., 2019) for the conversion of IASI 10 µm DOD to 550 nm.

We evaluate IASI daytime and nighttime DOD against AERONET ground observations before using the product to understand the day–night differences in dust aerosols over the dust belt. Such evaluations can be achieved by Taylor diagrams (Taylor, 2001). A Taylor diagram compares datasets in terms of three statistics i.e., the Pearson correlation coefficient between two datasets, the standard deviations, and the centered root mean square error (RMSE). Figure 2 shows normalized Taylor diagrams that compare IASI DOD (scaled to 500 nm using the average IR / VIS ratio of 0.60) to AERONET CAOD (500 nm) for daytime (Fig. 2a) and nighttime (Fig. 2b) observations. The standard deviations and centered RMSEs of IASI DOD have been normalized by the standard deviation of AERONET CAOD (shown as REF). The results show IASI DOD is highly correlated with AERONET station observations with statistically significant (95 % confidence level) correlation coefficients ranging from 0.18–0.92 for solar sites and from 0.27–0.82 for lunar sites. The highest average correlation coefficient for solar data is observed in the Saharan and Sahelian dust belt with an average correlation coefficient of 0.77 ranging from as low as 0.52 in Tamanrasset (Tam) to as high as 0.87 in Oudjda (Ouj), followed by the Middle East sites with an average correlation coefficient of 0.75 ranging from 0.31 in Eilat (Eil) to 0.87 in Solar Village (Sol) and Kuwait University (Kuw). The performance of IASI over the Asian sites is highly variable with the lowest correlation coefficient of 0.18 at the NAM_CO (NAM) site to as high as 0.92 in Jaipur (Jai). The NAM site has the lowest sample size of IASI–AERONET collocations (N=15), and this may partly account for such a low correlation coefficient. These results are largely consistent with similar evaluations in past studies (Peyridieu et al., 2013; Capelle et al., 2014, 2018). However, we also notice underestimations of daytime DOD at a few sites, such as Eilat (Eil), with a weaker correlation coefficient of 0.36, a higher RMSE of 0.33, and a large negative bias of more than 100 % (see Table 2 and Fig. 2a). Similar large negative biases are also observed around other coastal sites over North Africa (e.g., Iza, Lag, and Tei), possibly due to the mixing of sea salt with dust aerosols and the complicated land surface conditions in the area leading to difficulties in DOD retrieval (Capelle et al., 2014, 2018). Similarly, nighttime DOD is also underestimated at Tei and Iza by more than 200 %

The correlations over lunar sites are generally lower than solar sites (Fig. 2b). While over sites like Teide (Tei), Dakar (Dak), and Santa Cruz Tenerife (SCT) correlations between IASI DOD and AERONET CAOD are higher than 0.7, correlations over other sites are around 0.27–0.69. Note that the smaller correlation coefficient at CAT is insignificant and may be due to the complex topography of the area that makes IASI retrieval difficult, resulting in a smaller IASI–AERONET collocated sample size (N=8). Some sites over the Middle East (e.g., DEWA_Research_Centre (DEW) and Migal (Mig)) are also characterized by slightly lower nighttime correlation coefficients. The discrepancies between IASI DOD and AERONET CAOD records at sites over complex topographic regions (e.g., Iza with an altitude of ∼ 2.4 km) are also observed by Capelle et al. (2018), who attributed such lower correlations partially to the heterogeneity of the land surface or to rapid varying near-surface dust plumes that may reduce the sensitivity of infrared sounders. Reasons for the lower correlation in lunar data could range from the smaller sample size of lunar data to the quality of data used in the evaluation, which are cloud-screened but not quality-assured. In general, IASI DOD at sites around dust source regions is better correlated with AERONET CAOD than sites around regions where dust is transported from source regions (e.g., the southern Sahel) and worsened in areas characterized by complex terrains and pollutants from either biomass burning, industrial emissions, or coastal sediments (e.g., Eilat).

In addition to these Taylor diagrams, we further examined the relationship between IASI DOD and AERONET CAOD by combining all data points for daytime and nighttime measurements as shown in Fig. 3a–b. Consistent with the Taylor diagrams, the density scatterplots reveal a good performance of IASI DOD with an overall correlation coefficient of 0.7 for solar observations (Fig. 3a) and 0.57 for lunar measurements (Fig. 3b). These values are quite close to the average correlations over all solar sites in Table 2 (0.75) and all lunar sites in Table 3 (0.55). We also notice some overestimations of IASI DOD for small CAOD values (< 0.5) in both daytime and nighttime records (mainly over coastal sites such as Mez, Eil, and Dak in daytime and Mig and Mez in nighttime), which warrant future investigation. Overall, Figs. 2–3 show LMD IASI captures the spatiotemporal distribution of dust aerosols over the dust belt in both daytime and nighttime well and can therefore be used to understand the day–night variations in dust aerosols.

In this section, we examine the characteristics and differences between daytime (09:30 local solar ECT) and nighttime (21:30 local solar ECT) DOD and dust layer height from LMD IASI, along with CAOD from nine selected AERONET stations and surface PM10 concentrations from three LISA sites. Here, a mean uniform scaling factor of 0.60 is used to convert both daytime and nighttime 10 µm DOD to 500 nm. Using individual ratios will slightly improve the consistency between IASI DOD and AERONET CAOD (not shown) but may lead to some biases in the day–night differences. Figure 4 shows the annual and seasonal mean climatology (2008 to 2020) of IASI 500 nm DOD, AERONET 500 nm CAOD, and surface PM10 concentration for daytime, nighttime, and day–night difference. Both daytime DOD and nighttime DOD show similar seasonal cycles. In winter (DJF), the dustiest regions occur in the southern parts of the Sahel to the Guinea Coast (Fig. 4b, g). By spring (MAM), the maximum DOD begins to transition northward to the central to northern parts of the Sahara (Fig. 4c, h) and maximizes around summer (JJA) in the central to northwestern Sahara (Fig. 4d, i). Similarly, a pronounced DOD maximum is observed in the central parts of the Arabian Peninsula, northwestern parts of the Indian subcontinent, around the Iraqi and Irani deserts, and the Taklamakan Desert in northwestern China in JJA. DOD is reduced in fall (SON), with a magnitude comparable to that in DJF over the Middle East and Asia, but is slightly stronger over the Sahara yet weaker over the Sahel (Fig. 4e, j). Such seasonal migration of dust maxima is largely driven by the meridional migration of the Intertropical Convergence Zone (ITCZ) and is generally consistent with previous studies about dust aerosols in this region via satellite retrievals (e.g., Ginoux et al., 2012; Pu and Ginoux, 2018; H. Yu et al., 2019; Chédin et al., 2020; Vandenbussche et al., 2020; Y. Yu et al., 2021; Li et al., 2021).

Figure 4 also demonstrates statistically significant (95 % confidence level) differences between daytime and nighttime DOD. The day–night differences in DOD, i.e., daytime minus nighttime, are positive over the major dust source regions (i.e., most parts of the Sahara; the central Arabian Peninsula; parts of South Asia around eastern Iran, southwest Afghanistan, and central Pakistan; and the Taklamakan Desert) yet negative over regions near dust sources (i.e., the southern Sahel to the Guinea Coast, the southeastern coast of the Arabian Peninsula, and central to southern India). It is also important to note that there is a seasonal variability in the magnitude of the day–night differences in DOD, with the largest magnitude of the day–night difference in DJF and MAM (Fig. 4l, m) and a weaker magnitude in JJA and SON (Fig. 4n, o). The spatial pattern of the day–night differences in DOD in JJA is generally consistent with the day–night difference in dust emissions over North African dust sources shown by Chédin et al. (2020; e.g., their Fig. 4) and Todd and Cavazos-Guerra (2016; e.g., their Fig. 8).

Surface observations of dust properties are also examined to compare with results from IASI products. AERONET CAODs overlaid as circles on the IASI DOD show the two datasets generally agree on the seasonal daytime and nighttime distributions of dust aerosols (Fig. 4a–j), but the day–night differences in most seasons are insignificant, partially due to the smaller sample size and lower quality (level 1.5) of lunar data. Among all the sites analyzed, only Iza and Mig sites show significant day–night differences in CAOD in MAM, with Mig largely consistent with IASI DOD. The inconsistency between IASI DOD and AERONET CAOD in some AERONET sites may also be partially due to the uncertainties resulting from using CAOD to approximate DOD and uncertainties in IASI retrievals (see discussion in Sect. 3.7) as well.

Surface observations of PM10 concentrations during both daytime and nighttime from LISA are also overlaid as stars on the IASI DOD in Fig. 4. For a consistent comparison with IASI DOD, LISA hourly data are averaged over time steps approximately within ±30 min of IASI pixels that fall within a 30 km radius from each LISA site. Although DOD and surface PM10 concentrations reveal different aspects of dust activities, i.e., IASI DOD shows vertically integrated column extinction due to coarse dust, while PM10 concentrations reveal near-surface concentrations of large particles including both dust and sea salt (usually dominated by dust in dust source regions), we found that these results share similarities in terms of the day–night variations. For instance, the day–night differences in PM10 over Cin in JJA and Ban in DJF (Fig. 4n, l; significant at the 90 % and 95 % confidence levels, respectively) are quite consistent with IASI DOD.

The IASI dust layer height is another variable that can be useful in characterizing the day–night difference in the distribution of dust aerosols. Figure 5 shows the annual and seasonal mean climatology of daytime, nighttime, and day–night differences in dust layer height. The dust layer height reaches about 2.4–3.6 km in dust source regions over the Sahara and the Sahel, the central Arabian Peninsula, and the deserts in Central and East Asia in the annual mean (Fig. 5a, f) and are generally higher in DJF and MAM seasons (Fig. 5b, c, g, h) and lower in JJA and SON (Fig. 5e, j). The lower summertime dust layer height is somewhat in contrast to previous studies using CALIOP (e.g., Yu et al., 2010; Clarisse et al., 2019; see Fig. S3 and more discussion below). Negative day–night differences in dust layer height, i.e., lower dust layer height in daytime than nighttime, are observed mainly in dust source regions (e.g., large parts of the Sahara, Arabian Peninsula, and Taklamakan Desert), while positive differences are found over the dust downwind regions (e.g., the southern Sahel to the Guinea Coast and large areas in the Indian subcontinent; Fig. 5k–o). The magnitude of the day–night differences in dust layer height shows relatively small seasonal variations. Overall, the spatial pattern of the day–night differences in dust layer height (Fig. 5k–o) is largely opposite to that of DOD (Fig. 4k–o), which is generally consistent with the dust emission index defined by Chédin et al. (2020) that shows a higher DOD and lower dust layer height in dust source regions.

The seasonal cycle of daytime (13:30 local solar ECT) and nighttime (01:30 local solar ECT) DOD and dust layer height from CALIOP is also investigated to compare with IASI data (Figs. S2 and S3, respectively). The seasonal cycle of CALIOP DOD is very similar to IASI, consistent with the findings of H. Yu et al. (2019), although IASI shows a larger area of higher DOD over the Guinea Coast in nighttime in DJF (Figs. 4g, S2g). The day–night differences in DOD from CALIOP are insignificant for most parts of the dust belt across all the seasons except for a narrow region over the northern Sahara and parts of Central Asia and western China in MAM, JJA, and SON and are in general opposite to IASI. Such an inconsistency in the day–night differences in DOD between IASI and CALIOP may partially be attributed to the low signal-to-noise ratio of CALIOP daytime data and differences in the overpass times of the two instruments (∼ 4 h apart). Moreover, because of the narrow swath width of CALIOP with a sampling rate of twice per month (a repeat cycle of 16 d), the afternoon and nighttime observations are not on the same day, which also influences the day–night differences in CALIOP DOD.

In contrast to IASI, dust layer height in CALIOP shows maximum altitude in JJA over major dust source regions and minimum in DJF (Fig. S3). Previous studies show IASI dust layer height is systematically biased low by about -0.4 to -0.8 km in comparison with CALIOP (Peyridieu et al., 2013; Capelle et al., 2014; Kylling et al., 2018); however, here we found that the maximum altitudes are comparable between the two datasets over North Africa. The afternoon (13:30 local solar ECT) and midnight (01:30 local solar ECT) dust layer heights from CALIOP are also not significantly different from each other in most parts of the dust belt, except around the southern Sahel to the Guinea Coast, the southern Arabian Peninsula, and parts of the western Taklamakan Desert in DJF; western China in MAM; northern and eastern Sahara and parts of Central Asia in JJA; and western China in SON (Fig. S4l–o), while sharing some similarities with IASI over the western Taklamakan Desert, Central Asia (JJA), and coastal northwestern Africa (JJA). The differences between the IASI and CALIOP dust layer height could be attributed to several factors (Peyridieu et al., 2013; Chédin et al., 2020), such as different definitions of dust layer height, e.g., arithmetic mean dust layer height in CALIOP versus cumulative extinction height in IASI, and different overpass times of the two instruments (CALIOP lags behind IASI by about 4 h). Kylling et al. (2018) found that the bias of the dust layer height in IASI (LMD version) would be lower if the CALIOP dust layer height was defined by cumulative extinction height instead of arithmetic mean and was shifted to the observation time of IASI. Their results (their Table 3) show a difference of -0.053±1.339 km between LMD IASI and CALIOP for the cumulative extinction and -0.607±1.187 km for the arithmetic mean.

We compare the seasonal cycle of daytime and nighttime IASI DOD with LISA and AERONET ground-based observations to better understand how day–night differences in dust aerosols propagate in seasons. Figure 6 shows monthly mean surface PM10 concentrations from three LISA sites (Ban, Cin, and Mbo; see locations in Fig. 1) and monthly mean DOD from IASI averaged over a 30 km radius around LISA sites. We average hourly PM10 concentrations around ±30 min of the IASI overpass time for a consistent comparison with IASI. Note that the seasonal cycle of LISA records is different from DOD, with a minimum in JJA associated with monsoon rainfall and a peak in DJF and MAM due to transported dust from the central Sahara (Marticorena et al., 2010). The three sites are along 13–14∘ N but aligned in an east–west trajectory of the Sahelian dust transect. Such observations reveal a clear spatial variability of dust with a higher dust concentration over Banizoumbou (Ban), which is close to the Saharan dust sources, but decreases westward in Cinzana (Cin) and M'Bour (Mbo), similarly to the findings of previous studies (Marticorena et al., 2010; Kaly et al., 2015). Daytime PM10 concentration is significantly higher than nighttime in DJF and early MAM at Ban and Cin, while nighttime dust concentration is higher than daytime from late MAM to early SON at Ban and Cin (Fig. 6a–b). Mbo shows a similar seasonal cycle as Ban and Cin but the day–night difference is largely insignificant (Fig. 6c). Like LISA PM10, daytime IASI DOD is higher than nighttime in most DJF and MAM months but lower in JJA at Ban and Cin, consistent with the results shown in Fig. 4. Note that different from PM10 concentrations, nighttime IASI DOD at Mbo is higher than daytime throughout the entire year. This disparity could partially be due to the fact that the Mbo site is located along the transport path of the boreal JJA dust plumes, but further from the major dust sources of North Africa; thus dust aerosols are likely mixed to higher altitudes, which may be sampled differently between the near-nadir-viewing IASI instrument and the surface measurements and differently between day and night.

A similar analysis is carried out over nine AERONET sites (blue dots in Fig. 1) for IASI DOD and AERONET CAOD as shown in Fig. 7. AERONET CAOD is collocated with IASI temporally and spatially for consistent comparison between the two datasets. CAOD and DOD show very similar seasonal cycles, with maxima in late MAM to JJA for stations in the Sahara (Dak) and off the west coast of North Africa (Iza and SCT) and the Middle East (Mig, Sha, Mez, and DEW), whereas the Guinea Coast stations (Ilo and Kof) showed maximum DOD or CAOD in late DJF to MAM. The largest biases between IASI and AERONET occur in Iza during JJA where both IASI daytime DOD and nighttime DOD are higher than AERONET solar and lunar CAOD. It is worth noting that the Iza site is located at a higher altitude (∼ 2.4 km) and may contain some uncertainties. In terms of the spatial variability of dust, both IASI and AERONET showed consistency over all sites with maximum DOD or CAOD in JJA over Dak, Iza, SCT, Mig, Mez, and DEW; maximum DOD or CAOD in DJF over Ilo and Kof; and maximum DOD or CAOD in late MAM over Sha. In terms of the day–night differences, AERONET is consistent with IASI at the Dak site during JJA, with higher CAOD during nighttime than daytime. Over the Guinea Coast (Ilo and Kof sites), nighttime CAOD or DOD is higher than daytime for most months from late JJA to DJF, which is consistent with IASI. While seasonal variations in day–night differences in CAOD are largely similar to IASI DOD at Dak, Ilo, and Kof in JJA and SON, discrepancies are found at SCT in JJA, Mez in JJA and SON, Mig in MAM, Sha from late MAM to SON, and DEW in MAM, probably in association with the relatively smaller sample size and relatively lower quality (level 1.5) of AERONET lunar data and impacts of sea salt on CAOD at the coastal stations.

IASI DOD and dust layer height are available only two times daily. To have a clear picture of the full diurnal cycle of dust aerosols, data with a higher temporal resolution are required. Here, we use station data, i.e., typically 5 to 15 min AERONET CAOD and hourly LISA PM10 concentrations, to further explore diurnal variations in dust over the dust belt. Figure 8 represents the diurnal cycle of surface PM10 concentration in the Sahel (first row) and CAOD at AERONET sites (last three rows) for each season, with vertical cyan lines highlighting the overpass time of IASI (09:30 and 21:30 local solar ECT). The results indicate that surface PM10 concentrations at the three LISA sites in the Sahelian dust belt peak around 09:00–11:00 LST in the late morning in all seasons, except in JJA and SON when PM10 concentrations are low due to precipitation scavenging. Both Cin and Mbo sites have evening peaks around 19:00–20:00 LST (except in JJA at Mbo; Fig. 8b, c), but it is not very evident at the Ban site, which shows a peak around midnight in MAM (Fig. 8a). The passing time of IASI is largely consistent with the timing of maxima in surface PM10 concentrations.

At AERONET sites, daytime records (06:00 to 17:00 for Dak, Ilo, and Kof sites; 05:00 to 18:00 for Iza, SCT, Mez, Mig, Sha, and DEW; light yellow shading in Fig. 8) are observed by the sun photometer, and nighttime data (18:00 to 05:00 for Dak, Ilo, and Kof; 17:00 to 06:00 for Iza, SCT, Mez, Sha, DEW; 17:00 to 05:00 for Mig; grey shading in Fig. 8) are from the lunar photometer; thus the discontinuity between daytime and nighttime records is likely due to the different instruments (Fig. 8d–l). Furthermore, level-1.5 lunar data also have a higher uncertainty compared to level-2.0 solar data. AERONET CAOD also peaks in the morning around 07:00–09:00 LST for sites in the Guinea Coast (Ilo and Kof) in DJF, and the peaks in the Sahel (Dak), at the North Atlantic sites (Iza and SCT), and in the Middle East (Mez) are also around 07:00–09:00 LST in JJA, which is consistent with previous work in this region (Schepanski et al., 2009; Marticorena et al., 2010; Kaly et al., 2015; Yu et al., 2021). A secondary peak of daytime CAOD occurs in the afternoon around 15:00 LST (e.g., at Dak, Kof, Ilo, and Mez sites; Fig. 8d, g, h, k). The nighttime peak of CAOD varies in different regions. In North Africa, CAOD maximizes around 03:00 in Dak, 20:00 and 04:00 in Iza, 22:00 in SCT, 04:00 in Ilo, and midnight in Kof, while in the Arabian Peninsula CAOD peaks around 05:00 in Mig, 20:00 in Sha, and around 04:00 in DEW but without a clear peak at the Mez site (slightly higher around 05:00 in MAM and 02:00 in DJF; Fig. 8d–l). Overall, the available AERONET data in the dust belt show that IASI daytime data largely capture the early morning peaks of CAOD, while the nighttime data partially capture the high CAOD either after its early evening peak or before its nighttime maxima. As revealed by the comparison with LISA and AERONET station data, the day–night variations in IASI DOD could be quite similar to ground-observed CAOD and surface dust concentrations at some sites but not others. Although IASI data contain only two time steps, its high spatial resolution and global coverage provide useful information that complements sparsely located ground observations to help understand the diurnal cycle of dust.

With global coverage and high temporal resolution, aerosol products from reanalysis would be great tools to study the diurnal cycle of dust if they largely capture the observed day–night dust variations shown in satellite retrievals. Here we examine daytime and nighttime DOD from MERRA-2 and EAC4 to examine whether they capture the day–night differences in DOD as revealed by LMD IASI over the dust belt. After the reanalysis datasets are sampled at IASI overpass times at each grid point for consistency with satellite observations as discussed in Sect. 2.2.5, the annual and seasonal climatology of daytime, nighttime, and day–night differences in DOD from MERRA-2 and EAC4 are presented in Figs. 9 and 10, respectively. Like IASI, the results of the seasonal mean climatology of MERRA-2 and EAC4 DOD from 2008 to 2020 also revealed a higher DOD in MAM and JJA in comparison with other seasons. The magnitude of the day–night difference in DOD is, however, weak in the reanalysis products and largely insignificant as compared to that of IASI (Figs. 9 and 10). The magnitude of the day–night difference in MERRA-2 DOD appears to be large only in the Bodélé Depression (centered around 17∘ N, 18∘ E), with a positive difference throughout the year and a negative difference to the southwest (Fig. 9l–o). Over northeastern Africa and the coastal area of the Arabian Peninsula and Central Asia, some areas also show significant negative differences, i.e., with greater nighttime DOD. The sign of the day–night differences in MERRA-2 DOD is largely consistent with IASI in some parts of the Bodélé Depression and southern Arabian Peninsula in DJF and MAM and Central Asia in JJA but not in other regions or seasons.

A slightly larger portion of the central to northern Sahara, the Middle East, Central Asia, and the eastern Taklamakan Desert is characterized by significant and negative day–night differences in DOD in EAC4 (Fig. 10l–o). In most of these areas, the day–night differences are opposite to that of IASI, except over the northeastern Sahara, the southern Arabian Peninsula, and northwestern Sudan in DJF and Central Asia in JJA. In short, aerosol reanalyses in general have difficulties in capturing the day–night differences in DOD shown by IASI. This may be partially because reanalyses do not assimilate nighttime observations (e.g., AERONET lunar data or infrared satellite products) to constrain AOD or DOD.

In this section we examine the impacts of meteorological conditions on the daytime and nighttime variations in dust aerosols, mainly DOD and dust layer height from IASI, over the dust belt using meteorological variables from MERRA-2, ERA5, IMERG, and LISA observational datasets.