Qualitative and quantitative detection of iron oxides using reflectance spectroscopy is an important task in soil mapping practice [1
]. Soil iron oxide content reflects the duration and intensity of pedogenic processes [7
]. Accordingly, they appear as either a standalone crystallized mineral (e.g., hematite (Fe2
) and goethite (FeOOH)), or a pseudo amorphous entity (e.g., ferryhydrates) that coats the soil particle surface and colors the soil, even if their content is low [5
]. The aim of the new NASA mission ‘Earth Surface Mineral Dust Source Investigation’ (EMIT) [9
] is to map the surface mineralogy of the soil in order to investigate dust sources with a special emphasis on iron oxides using a hyperspectral remote-sensing sensor onboard the International Space Station (ISS) EMIT will help answer essential scientific questions on whether iron oxide type of aerosol warms or cools the atmosphere, and will enable better modeling of dust transmission. Soil organic matter (SOM) and iron oxides are spectrally active across the visible–near infrared (VIS–NIR) spectral region (400–1000 nm), but their spectral signatures overlap, potentially hindering their identification. Therefore, it is important to study the impact of different types of organic matter (OM) on the spectral signal of iron oxides. Soil is a complex system that is extremely variable in physical and chemical composition both temporally and spatially. Soil spectroscopy, although complex as well, can cluster several soil properties based on their spectral responses and can reduce the dimension of uncertainty in the soil system using multivariate algorithms [10
In the last few decades, extensive studies have been performed to investigate the role of OM and iron oxides in the soil reflectance spectrum (e.g., [11
]). OM is spectrally active across the entire VIS–NIR region and different spectral ranges across this region (as well the Short Wave Infra-Red, SWIR; 1000–2500 nm) have been found to correlate with OM content. For example, Mathews et al. [11
] found that SOM correlates with reflectance values in the 500–1200 nm range, whereas Beck et al. [13
] suggested that the 900–1220 nm region is more suitable for SOM detection. Tian et al. [18
] deduced that the entire VIS–NIR spectral range (400–1000 nm) is sensitive to SOM detection. Iron oxides, as either crystallized or amorphous minerals, are also spectrally active across the VIS–NIR region based on the electronic transition of iron cations Fe3+
, as summarized by Hunt et al. [6
]. The most common iron oxides in Israeli soils are hematite and goethite minerals, whereas their reflectance spectra generally display specific features linked to Fe3+
-induced absorption [19
]. Along with other non-crystalline iron-oxide fractions, these minerals impact the common reddish color found in these soils. The main absorption of hematite is at around 550 nm and 880 nm (red color) and at 480 nm and near 920 nm (yellowish-brown color) for goethite. Thus, Henderson et al.’s [16
] claim that the spectral regions of OM and iron (and Mn) oxides overlap and might spectrally affect each other is not surprising.
Although Sukhdev et al. [1
] created an extended review of hyperspectral analysis for iron oxides in soils, the researches mostly considered iron oxides as a standalone parameter in soil. Li et al. [20
] suspected that the high content of iron oxides masks the spectral features of SOM in the 600–1000 nm range. Peng et al. [21
] showed that the SOM spectrum across the 622–851 nm range is obscured when iron-oxide content is high. Liu et al. [22
] investigated the effects of iron oxides on predictions of SOM content using spectral-based models. They used 267 soil samples from South China which were replete with iron oxides to demonstrate that this overlap affects the spectral-based models. Adding the external parameter orthogonalization (EPO) algorithm provided a more accurate model. Richter et al. [23
] observed that strong iron absorption makes it difficult to extract SOM content directly from the spectrum. As SOM content varies by up to 10% in mineral soils [24
], the opposite effect, i.e., the SOM’s quantitative effects on iron-oxide spectral features is important. Baumgardner et al. [12
] suggested that SOM plays a dominant role in spectral properties when its content exceeds 2.0%, but no actual measurements were provided. Galvao and Vitorello [25
] demonstrated that SOM obliterates the effect of iron reflectance and color when its content is higher than 1.7%. Whereas there have been many studies on SOM, there is hardly any mention in the literature of differences between OM types (sources) at either low or high SOM contents in mineral soils. This is a crucial factor, as OM is not a homogeneous material and its composition can vary based on the chemical characteristics of its source and the effects of aging [26
]. In this study, we systematically examined how the various amounts of different types of OM affect the iron oxides’ spectral signal.
To that end, we physically mixed two different pure decomposed OM types from different origins with a selected red soil from Israel (Rhodoxeralf) and compared the authentic result to a synthetic spectral linear-mixing model. The study reports the spectral effects of increasing levels of these OM types in the soil and provides spectral thresholds for the critical spectral changes, where quantitative assessment might be significantly biased.
A significant effect of OM on the spectral signal of iron oxide in red soil was observed at around 800–1000 nm. Although the spectral changes were visible at low OM content (0.85–1.5% SOM), we set a threshold of 25% spectral change, above which the effect was significant. The SOM values under this threshold were found to be around 1–2%. It is interesting to note that Baumgardner et al. [12
] concluded that a quantitative assessment of SOM can only be achieved when it is above 2%. This stands in good agreement with Galvao and Vitorello [25
], who found that the threshold value for masking out the iron-oxide signal is 1.7% SOM. It is then possible that the OM spectral assessment is strongly controlled by the iron-oxide signal. It is also interesting to note that at around 2% SOM, the OM effect began to depend on its type. Here, A2 and A5 had different spectral effects on the iron-oxide signal, especially above 2% SOM. This suggests that in OM-rich soils, SOM exerts a significant effect on the iron-oxide spectral response.
Many studies have reported that the spectral slope across the VIS–NIR region is a sufficient parameter to assess OM content e.g., [18
], while others have provided specific spectral features across the VIS–NIR region [16
]. Nonetheless, in the literature, there is no concrete explanation for these OM spectral assignments in the VIS–NIR spectral region. The slope in the visible region may be related to the irregularity of OM particles relative to the shape of the mineral, characterized by different scattering effects that then control the slope. The specific wavelengths reported in this region call for a more concrete explanation because OM itself has no absorption features in the VIS–NIR region except, most likely, chlorophyll residue at 670 nm [26
]. Accordingly, and based on the results of this paper, we propose the following mechanism:
The iron oxides in mineral soils act as a stable background for the OM spectral response. As the slope of the OM in the VIS–NIR region increases with increasing OM, it actually, and practically, masks the stable iron-oxide spectral response (where iron oxides are present). Accordingly, when a specific wavelength is extracted in the spectral-based SOM models, it measures the changes in the iron-oxide absorption features due to the OM masking effect, and not only the change in the OM slope. As crystallized iron oxides have a specific wavelength (depending on environmental conditions), the spectral-based models that report specific wavelengths are based on the spectral features of the iron (or other) oxides in the soils. Accordingly, the SOM estimation by reflectance across the VIS–NIR region is subject to the soil-formation process, where iron oxides and OM are formed according to the five major soil-forming factors [48
]. As the iron-oxide absorption features and the SOM spectral response can differ from one climate zone to another, caution must be exercised in establishing a generic model to assess OM in the presence of iron-oxide or vice versa. It should be noted that several studies have reported a saturation phenomenon with increasing SOM percentage e.g., [12
]. The proposed mechanism can explain the nonlinearity of the OM’s masking effect on the iron oxide absorption signal. The results of this paper showed that below 1–1.5% SOM caused less than 25% change in the iron-oxide absorption band, the masking effect was robust, and no difference was observed between the two OM types (A2 and A5). Above this threshold, each OM species had a distinct spectral effect. This observation should be considered when different climatogenic soils are used to assess SOM spectrally. In an arid region where SOM content is low, no distinction between OM sources is necessary and a generic spectral-based model to assess SOM may be possible. In more humid regions, however, where the OM may change and the SOM content may be high, such a model should be generated with caution.
Based on this discussion, we can conclude that if spectral modeling is used to predict SOM content from soil spectroscopy information, both the origin of the SOM and the soil’s mineralogy need to be taken into account to improve SOM-detection accuracy. This may pose a problem if a robust SOM spectral analysis is conducted using continental or global soil spectral libraries in which both the iron oxides and the OM species might vary. As the VIS–NIR spectral region has some limitations in predicting SOM, as already noted, we strongly suggest that the entire spectral VIS–NIR–SWIR region be used for such predictions to obtain more spectral information as examined here.
It goes without saying that the opposite effect may also be significant, i.e., the analysis of iron oxides in the soil might be significantly affected by both OM content (up to 25% change at 1–2% SOM) and OM species (above 25% change at 2–10% SOM). This observation should be taken into account in any attempt to quantitatively estimate soil iron-oxide content using reflectance spectroscopy if the SOM content is above 2%.
The significant difference between authentic and synthetic mixtures occurs because of the complex situation in the authentic soil system relative to the simple linear-mixing effects (Hapke 1980 [49
]). Both organic and mineral particles have different shapes and irregularities that affect the soil’s reflectivity differently across the VIS–NIR region. In the real world, OM is not a standalone endmember. Instead, it interacts with the soil minerals in many nonlinear processes. This makes it highly risky to try to mimic a real-life soil spectrum using the linear mixing of pure spectra taken from a spectral library or simple measurements. As soil is a complex system, we attempted to minimize other effects that might hinder accurate iron-oxide analysis, such as soil moisture and particle-size distribution. The carefully prepared authentic mixtures used in this study, although far from real natural conditions, nevertheless better represent “natural” conditions than the synthetic mixtures, and highlight the importance of the OM species in the spectral masking effect. Thus, a synthetic mixture of soil components such as OM and iron oxides (and perhaps also other soil minerals) should be used to form soil spectral analogs with caution. Any materials in such a mixture will have chemical and physical interactions amongst themselves, such as electrostatic adsorption, van der Waals bonding effects, and other physical adsorption processes that may bias the linear relationships. The authentic mixture as used here, with a preparation stage that renders it sufficiently homogeneous, is a more reliable approach to investigating such effects by simulating a near-real natural environment, thereby providing insight into the effects of two significant soil chromophores in the VIS–NIR region.
The origin of the OM has an effect on soil reflectance in the VIS–NIR region. At low SOM content, up to 1.5%, the two OM types behaved more similarly. However, as the OM content increased beyond this level, their effect on the iron-oxide signal was greater and different: A5 OM obliterated the iron-oxide signal at 4% SOM and A2 at 8% SOM. The overlapping spectral responses of OM and iron oxides in the VIS–NIR spectral region should be taken into account when both iron-oxide and OM contents in the soil are to be spectrally estimated due to this eventual obliteration ("saturation") effect of OM on the iron-oxide spectral signal. We set 25% spectral change as the threshold above which precaution should be taken in the quantitative assessment of iron oxides; this threshold corresponded to 1.9% and 1.6% SOM content for soil–A2 and soil–A5 authentic mixtures, respectively. The iron-oxide spectral feature acts as a secondary chromophore in the assessment of OM content where the saturation plot between OM content and the soil’s spectral response is due to the complete masking effect of the OM spectral response on the iron-oxide signal. The use of synthetic mixing models in soil research is problematic, especially if OM and mineral endmembers are mixed. The problem stems from the inability to mimic the nonlinear processes occurring in the complex soil system. Although this paper provides a case study of only two OM species and one soil sample, it demonstrates the problem posed by mixtures in soils and calls for an investigation of more mixtures that may be used in quantitative and qualitative analyses of soil spectral information. It is important to note that this paper presents an individual and artificial case study that highlights the effect of OM on iron-oxide spectral features. More studies investigating other soils and different types of OM, which better represent the natural interaction between soil minerals and OM, are strongly warranted.