The thirteen soil and sediment samples in this study (Table 1) were selected with the aim of representing the global scale variability of particle mineralogy. Samples, collected from different source areas worldwide, represent a depth of the first millimetres of the surface. Many of the samples had already been analysed in past laboratory studies, including simulation chamber experiments to investigate their physical, chemical and spectral optical properties at solar and MIR wavelengths (Baldo et al., 2020, 2023; Caponi et al., 2017; Di Biagio et al., 2014a, 2017, 2019, 2023).

For Northern Africa, we selected a soil from Morocco in the Northern Sahara and two samples from the Sahel, one in Niger and one in Chad (sediment from the Bodélé depression). These are important sources for medium and long-range dust transport towards the Mediterranean (Israelevich et al., 2002) and the Atlantic Ocean (Prospero et al., 2002; Reid et al., 2003). In particular, the Bodélé depression is one of the most active sources at the global scale (Goudie and Middleton, 2001; Washington et al., 2003). The two soils from the Middle East are from Saudi Arabia and Kuwait, which are important sources of dust to the Red and the Arabian seas (Prospero et al., 2002) and the North Indian Ocean (Leon and Legrand, 2003). For the second largest global source of dust, Eastern Asia, we studied one sample from the Gobi desert. For North and South America, soils were collected in the Sonoran Desert in Arizona and the Atacama Desert in Chile. The Sonoran Desert is a permanent source of dust in North America, the Atacama Desert is the most important source of dust in South America (Ginoux et al., 2012). For Southern Africa, we selected two soils from the Namib desert, one from the area between the Kuiseb and Ugab valleys (Namib-1, already studied in Di Biagio et al., 2017, 2019) and the other from the Huab ephemeral river bed (Namib-Huab), both of which are sources of dust transported towards the South-Eastern Atlantic (Vickery et al., 2013). A sample from Botswana was also selected, as an additional source from Southern Africa (Bhattachan et al., 2013, 2015). Surface sediment samples collected from two major dust hotspots in Iceland (H55, Hagavatn, and MIR45, Mýrdalssandur), which significantly contribute to the total dust emissions from this region (Arnalds et al., 2016), were also investigated. Iceland is the major documented emitting HLD source so far (Meinander et al., 2022).

Dust aerosols were re-suspended in the laboratory by mechanical shaking of the parent soil samples using the same protocol previously described by Di Biagio et al. (2017) and Baldo et al. (2023). To this aim, 5 g of soil sample (previously sieved at 1000 µm and heated at 100 °C for about 1 h) were placed in a Büchner flask and shaken at 100 Hz by means of a sieve shaker (Retsch AS200). The dust suspension in the flask was entrained by a pure N2 flow at 5 Lmin-1 to a glass vial where the dust aerosol was collected. The collection time lasted for about 1 h, whilst continuously shaking the soil, in order to collect around 1–2 mg of dust aerosols.

Infrared transmission spectroscopy was performed by means of the pellet technique (Di Biagio et al., 2014b; Volz, 1972) using Potassium Bromide (KBr for IR spectroscopy Uvasol^®, CAS No 7758-02-3, lot. B1978207 142) as transparent matrix in which dust grains collected from soil resuspension were dispersed. Pellet spectroscopy has various limitations and uncertainties to represent the optical behaviour of suspended aerosols, as discussed by Di Biagio et al. (2014b), however it represents a reference technique to explore the infrared optical properties of aerosols and their components, without the complication of resuspension processes. A quantity of 0.5 mg of dust was weighed and diluted in 300 mg of KBr (corresponding to 0.16 % of dust in KBr). Dust and KBr were weighed by means of a Mettler Toledo microbalance XPR26C whose maximum sensitivity is 1 µg. The dust-KBr mixture was mechanically shaken for few minutes to create a homogeneous mixing and slightly ground with an agate mortar. The obtained dust-KBr mixture was placed in the oven to dry at the temperature of 110 °C for at least 2 h and then pressed at ∼9.5 tcm-2 for 2 min to form a thin pellet. A quantity of 300.5 mg of powder (300 mg KBr plus 0.5 mg dust) was used to create a homogeneous pellet of 13 mm diameter (surface 1.33 cm2) and about 1 mm thickness. All laboratory manipulations were accomplished in clean conditions in an ISO7 room. A number of 2 to 3 pellets were produced per sample depending on the amount of the collected dust aerosols (with the exception of Arizona for which only 1 pellet could be produced), to test the repeatability of the procedure, for a total of 31 pellets. Additionally, six pure 300 mg KBr pellets were produced. All the pellets were kept in the oven at 110 °C until they were used for spectroscopy measurements in order to minimize water vapour absorption on their surface.

Absorbance spectra were recorded between 2.5–25 µm (4000–400 cm-1 wavenumber) at 0.5 cm-1 resolution by means of a PerkinElmer Spectrum Two FT-IR spectrometer. The instrument uses a Globar source, with a KBr beamsplitter and a deuterated triglycine sulphate (DTGS) detector. Pellets were placed in the spectrometer chamber. A background spectrum was acquired prior each pellet measurement and subtracted to remove the signatures of CO2 and H2O possibly present in the cell. A total of 10 scans were averaged to produce the dust-KBr and the pure KBr spectra, respectively. Within the 6 spectra of pure KBr, two showed slight contamination by dust grains in the pellet and were removed from the dataset. The other 4 pure KBr spectra did not show any indication of contamination, but showed a slightly different spectral variation particularly at low wavelengths (<5µm), which we attribute to some degree of water vapour absorption by the highly hygroscopic KBr occurring during the pellet production, which was performed at ambient pressure instead of standard vacuum conditions. To take this effect into account, each dust-KBr spectrum was corrected by subtracting the pure KBr spectrum that best fit the baseline of each dust-KBr pellet. All the collected raw spectra for KBr and dust are shown in Figs. S1–S5 in the Supplement. The uncertainty in the measured absorbance is less than 5 % and has been estimated as the 3σ variability of the signal in the regions of no dust absorption (A<0.01).

The pellet absorbance spectra acquired in this study are compared against in-situ absorbance spectra measured on suspended aerosols in a simulation chamber from the same LMLD source soils by Di Biagio et al. (2017) (Fig. S6, Text S1 in the Supplement). The comparison shows that the shape of the dust absorption in the 8–15 µm common spectral range is similar between pellets and suspended aerosols. Only the quartz band is in some cases overestimated in the pellet spectra. This is likely due to the presence of some grains of soils, much richer in quartz than the aerosols (Fig. 18 in Adebiyi et al., 2023), that were entrained together with the aerosol and therefore included in the pellet. This potential artefact, however, does not seem to influence the spectra in other regions within the 8–15 µm.

The identification and quantification of the main mineral phases composing the dust aerosol particles was performed by X-Ray Diffraction (XRD) and X-ray Absorption Near-Edge Structure (XANES) analyses (Formenti et al., 2014b, a). These were performed on dust samples collected on polycarbonate filters and re-suspended applying the same generation procedure as described in Sect. 2.2. Data acquisition and analysis was presented in previous studies (Baldo et al., 2020, 2023; Caponi et al., 2017; Di Biagio et al., 2017, 2019). More detailed information on measurement procedures is provided in Text S2 in the Supplement.

The mineralogy of the thirteen analysed samples is illustrated in Fig. 1 and reported in Table S1 in the Supplement. The eleven LMLD samples are composed of varying proportions of clays (illite, kaolinite, chlorite, palygorskite, muscovite; 45.5 %–75.6 % by mass), quartz (3.5 %–36.7 %), feldspars (orthoclase, albite, microcline; 2.2 %–26 %), calcite (2.0 %–21.7 %) and iron oxides (hematite, goethite; 0.7 %–5.8 %). Iron oxides are identified in different proportions of goethite and hematite in all LMLD, except for Botswana for which no iron oxides are detected (based only on XRD analysis for this sample). Calcite is found in all samples with the exception of Niger, Bodélé and Botswana. Dolomite is detected only in the Morocco dust.

As discussed by Baldo et al. (2020), a contrasting mineralogy is obtained for the Icelandic dust samples compared to LMLD. The Iceland-M dust is mostly composed of amorphous glass material (91.3 %), feldspars (anorthite; 3.5 %) and pyroxene (augite; 3.6 %), while Iceland-H is made of feldspars (anorthite, microcline; 53.0 %), pyroxene (augite; 29.3 %), olivine (forsterite; 7.2 %) and minor contribution of amorphous glass (8.0 %). Both Icelandic samples show the presence of iron oxides in the form of magnetite (1.4 %–2.0 %) with lower contributions of hematite and goethite (0.2 %–0.5 %) by XRD and Fe sequential extraction techniques. These observations are in line with analysis from same source areas in Iceland (González-Romero et al., 2024).

The absorbance spectra measured in the 2.5–25 µm range for the thirteen LMLD and the Icelandic dust samples are shown in Fig. 2. The comparison between dust absorbance and single mineral spectra is provided in Fig. 3 for Saudi Arabia, taken as an illustrative case for LMLD due to the presence of absorption signatures from multiple minerals in its spectrum, and for Iceland-H and Iceland-M for HLD.

As depicted in Figs. 2 and 3, the absorbance spectra show large sample-to-sample variability which reflects the diversity in terms of mineral content and speciation. A marked difference is observed for the Icelandic dust compared to the LMLD samples both in respect of the position and shape of the absorption bands. The largest absorbance peaks are found for both LMLD and Icelandic dust in the 6–12 µm and 15–25 µm regions due to the superposing contribution of multiple minerals, such as clays, quartz, carbonates and iron oxides (hematite, goethite) for LMLD and basaltic glass, feldspars, augite, forsterite and iron oxides (magnetite) for Icelandic dust. A full list of peak positions for single minerals are provided in Table S3 in the Supplement. Note that below 5 µm the absorbance signal shows an increase for decreasing wavelength for most of the samples, which is due to the combined effect of clay absorption around 2.8 µm and residual KBr signal not completely removed through baseline correction.

All LMLD samples display strong spectral signatures associated with clays between 9–11 µm, but the position and relative intensity of the peaks vary with clay speciation. The prevalence of kaolinite is reflected for instance in the presence of the double peaks at 9.7 and 9.9 µm and the secondary peak at 10.9 µm, as identified in the Niger and Bodélé samples. Conversely, a broader single peak located between 9.5–9.7 µm is identified for samples (i.e., Morocco, Arizona, Atacama, Namib-H) composed of different proportions of clay minerals (illite, kaolinite, montmorillonite, palygorskite, muscovite). The presence of quartz is associated in LMLD with a main absorption band at 9.3 µm, which induces the appearance of a more or less pronounced shoulder in the clay bands, and a secondary double peak at 12.5 and 12.8 µm, as clearly identified in the Niger, Bodélé, Gobi, Kuwait, Arizona and Botswana samples, where quartz content is the highest (18.9 %–36.9 % in mass). An additional region of intense absorption for LMLD is found in the 6.5–7.5 µm range associated with the specific features of carbonates (calcite, dolomite) also showing a secondary less intense peak at ∼11.4 µm. The intensity of the main calcite band at 7 µm band follows the trend in calcite content (see Table S1) but it is not always proportional to it. This could suggest that differences in size-dependent mineralogy between the different samples can affect the intensity of the absorption bands, as also discussed in Di Biagio et al. (2014b).

Three main broad peaks are identified for all LMLD samples above 15 µm. These are contributed by superimposing bands for the different clay species between 18–23 µm, together with those of quartz (two bands centred at ∼20 and 22 µm) and iron oxides (two large bands centred at ∼19 and 23 µm for hematite, and one at ∼18 µm for goethite).

For the Iceland-M sample dominantly composed of amorphous silicate (91.3 %), the spectrum is mainly shaped by the two broad absorption bands of amorphous glass centred at 9.5 and 22 µm. A third band at 14 µm is identified and potentially associated with anorthite (Salisbury et al., 1987, 1991, data not shown). For the Iceland-H sample several absorption bands of anorthite (representing about half of the mass for this sample) are identified, with the absorption peaks centred at 18.7, 21.4, and 26.5 µm. Microcline, augite, forsterite, and magnetite also contributed to the spectral absorbance above 17 µm and different absorption peaks of these minerals are identified, as shown in Fig. 3. The amorphous material contributes to absorption in the 8–12 µm region and above 15 µm. Small peaks are also detected in the 8–12 µm range but due to the scarcity of information on single mineral spectra, no specific attribution is possible.

The spectral diversity of Icelandic dust samples evidenced in this study is supported also by comparison with LMLD from literature data (Di Biagio et al., 2014b; Sadrian et al., 2023), confirming the different optical signatures between low, mid and high-latitude dust. This is shown in Fig. 4 where absorbance measurements in this study are combined with literature data on pellet dust samples from worldwide locations in the 2.5–25 µm spectral range. Data include the work by Di Biagio et al. (2014b) providing measurements for five natural dust aerosol samples originated from Niger and Algeria and collected at the ground-based sites of Banizoumbou (Niger) and Tamanrasset (Algeria) in 2006 during the AMMA campaign (African Multidisciplinary Monsoon Analysis) (Formenti et al., 2011; Rajot et al., 2008). Further spectroscopic data for 26 samples from diverse locations in USA (Arizona, California, Nevada, Utah), Africa (Chad, Botswana, Djibouti, Mali, Namibia), Spain (Las Canarias), Arabian Peninsula (Iraq, Kuwait, Qatar, Saudi Arabia), and Central and eastern Asia (Afghanistan, China) are provided by Sadrian et al. (2023). The Sadrian et al. (2023) dataset refers to surface soil samples sieved at 38 µm instead of aerosols. Figure 4 further illustrates the variability in the dust spectral signature across the 2.5–25 µm spectral range for dust of additional diverse geographic origin. The LMLD samples in the present work exhibit a spectral variability that is within the one identified in past studies and particularly in Sadrian et al. (2023).

The integral of the absorbance signal in the 2.5–25 µm range for the LMLD and Icelandic HLD investigated in this study and the contributions by the MIR (2.5–15 µm) and the FIR (15–25 µm) ranges are shown in Fig. 5. The integrated spectrum in the 8–12 µm region is also shown, as it corresponds to the MIR atmospheric window, where absorption by atmospheric gases is relatively low, therefore of interest for atmospheric radiative transfer as most of the dust MIR signal is in this region.

The integrated absorbance in the 2.5–25 µm range is within 0.33 and 3.29 for LMLD and 1.61–1.91 for Icelandic dust. The MIR contributes by 35 %–53 % (LMLD) and 45 %–47 % (HLD) to total absorbance, mostly within the 8–12 µm region, while the FIR contribution is within 47 %–65 % (LMLD) and 53 %–55 % (HLD), representing more than half of the total integrated absorbance.

Literature data on natural dust samples in Fig. 4 also confirm the relevance of the FIR absorbance for dust of varying origin. The integrated absorbance in the 2.5–25 µm is within 0.44 and 1.0 for the samples in Di Biagio et al. (2014b), and within 0.41 and 5.50 (with an average value of 1.7±1.2) in the 4–25 µm for the 26 samples in Sadrian et al. (2023). The FIR signal above 15 µm contributes 50 %–60 % of the integrated absorbance in Di Biagio et al. (2014b), and 28 %–63 % in Sadrian et al. (2023). Only one sample in Sadrian et al. (2023) displays a low contribution of the FIR (5 %) to the absorbance signal.

The few existing pellet spectroscopic measurements of natural dust aerosols beyond 25 µm (Fouquart et al., 1987; Volz, 1972, 1973), despite not allowing to explore the diversity of dust absorption from global sources, support anyhow the presence of a significant absorption signal continuously up to 40 µm (Fig. S8). These data were acquired at low spectral resolution and correspond to rainout dust samples collected in Germany (Volz, 1972), Saharan dust from Barbados (Volz, 1973) and Niger dust (Fouquart et al., 1987).

The relevance of dust absorption in the FIR for sources of varying mineralogy is further evidenced when looking at the spectra of single minerals composing LMLD and Icelandic HLD. Literature data of the imaginary part of the refractive index or absorbance spectra for single dust minerals extending through the FIR range up to 100 µm are shown in Fig. 6. Information on the datasets and the band centre wavelengths for all identified mineral absorption peaks are provided in Tables S2 and S3. As shown in Fig. 6, all minerals show multiple and often superposing absorption bands in the 25–100 µm wavelength range, therefore supporting the relevance of dust-radiation interaction in the FIR well beyond 25 and 40 µm, as measured for natural dust samples. The strongest peaks above 25 µm are identified for clays (kaolinite, illite; 28.8, 36.2, 51.3, 91.7 µm), iron oxides (hematite, goethite, magnetite; 22.6, 28.6, 29.1, 33.5, 37.3 µm), and carbonates (calcite, dolomite; 28.5, 33.0, 38.5, 45.9, 64.0 µm). Feldspars (albite, orthoclase, anorthite), pyroxene (augite), and olivine (forsterite) show several absorption peaks in the broad range 26–77 µm. The intensity of the absorption peaks is of comparable intensity, and in some cases higher, than those below 25 µm, particularly for hematite, calcite and dolomite.

On the other hand, for several minerals such as basaltic glass, quartz and microcline, information is lacking in the FIR. Figure 6, while showing the potential relevance of dust absorption throughout the FIR domain, also highlights the paucity of spectral data available for single dust mineral components, which sum up to the low available information for natural dust. For most of minerals, the data do not cover the full infrared spectral range, being limited either to the MIR or to part of the FIR. Information on the complex refractive index are available only for some minerals, while no data exist for feldspars, pyroxene and olivine. Almost all data available have been acquired for compressed pellets or often crystals, so they provide limited representativeness for suspended dust.

The results from this study show that the absorbance of dust in the FIR up to 25 µm is comparable in intensity to that in the MIR. These observations highlight the potential contribution of FIR interactions to the total dust direct radiative effect for low to high latitude dust sources. As a matter of fact, the FIR range is at present not specifically considered in model simulations due to the lack of knowledge of the dust optical properties at these wavelengths. However, it can be climatically relevant in different regions, notably dry areas and the upper troposphere, as well as high latitudes where infrared radiation control the radiative budget over large part of the year.

Another observation from the present analysis is the comparable intensity of the infrared absorbance, both in the MIR, FIR and the 8–12 µm window, for the Icelandic HLD compared to the LMLD samples. The integrated absorbance of HLD is indeed at the upper end of that measured for low and mid-latitude dust (Fig. 5). In a previous study, Baldo et al. (2023) estimated that Icelandic dust absorption in the solar range (370–590 nm) is at the upper end of typical values for LMLD as reported by Di Biagio et al. (2019), while in the near infrared (660–950 nm) the absorption by Icelandic dust is up to 2–8 times higher than most of dust aerosols from northern Africa and Asia. The results from Baldo et al. (2023) combined with observation from this work suggest the relevance of dust-radiation interaction for Icelandic dust across the whole atmospheric spectrum.

Data from this study also highlight the variability of dust spectral absorbance as measured up to 25 µm for dust of diverse origin and composition, particularly when contrasting LMLD versus Icelandic HLD types. Specific minerals are identified to modulate the intensity and the spectral variation of dust absorption bands, producing characteristic spectral signatures for LMLD (clays, carbonates, quartz, iron oxides) and Icelandic dust (amorphous silicate, anorthite, microcline, augite, forsterite). This suggests that the spectral signature of dust in the MIR and FIR could be used to identify mineralogical composition for different dusts, which can be used to differentiate the origin of airborne particles based on remote sensing infrared observations, such as those from IASI/IASI-NG and FORUM. Single mineral spectra also suggest the presence of additional specific mineral signatures beyond 25 µm, which supports further possibilities to detect and characterise global dust from ground-based and space-borne hyperspectral remote sensing observations in the FIR. Indeed, as demonstrated in Di Biagio et al. (2023), infrared spectral signatures can be used to derive quantitative information of dust mineralogy, which enables to distinguish dust sources and potentially follow dust plume evolution transport in the atmosphere. A specific analysis of the sensitivity of MIR and FIR remote sensing to dust aerosols is provided by Sellitto et al. (2025).