Effects of Sample Handling on Visible and Near-Infrared Soil Reflectance

Abstract

Soil reflectance is a cumulative attribute determined by interactions between light (photons) and the physical, chemical, and biological properties of soil. Soil properties such as organic matter, moisture, mineral oxide contents, soil texture, and surface roughness all influence soil reflectance at unique wavelengths. Many standard sample preparation techniques are designed to alter soil properties, so as to homogenize samples to improve the consistency of reflectance data collected. This study aims to quantify the effects of one standard sample preparation activity, drying, and repeated wetting–drying cycles on visible–near-infrared soil reflectance, by collecting reflectance data from nine samples which were dried then wetted for a total of three times. This study demonstrates the major, permanent effects of a drying and wetting cycle on soil reflectance, and then presents a model which can be used to correct for these effects. These results have direct implications for remote sensing activities, soil libraries, and soil spectral libraries. These results show that without correction, data from soil spectral libraries, spectral data collected from stored samples, and spectral data transferred between studies, have limited utility for characterizing soils as they exist in the field. Soil spectral data collected in the lab requires correction before it may be used to predict soil properties in the field, as standard sample handling procedures change intrinsic soil properties and introduce systematic error into these data.

Graphical abstract

This is a visual representation of the abstract.

Keywords

soil spectroscopy spectral libraries remote sensing visible wavelengths near-infrared NIR wavelengths

Introduction

Detailed scientific soil characterization is typically an expensive, time-consuming, and resource-intensive activity,¹ which requires numerous specialized instruments, chemicals, lab facilities, and appropriately trained technical staff. By contrast, the use of visible (Vis) and near-infrared (NIR) spectroscopy to predict soil properties is a relatively rapid and cost-effective soil characterization option.^2–4 Apparent relationships between soil reflectance spectra and soil properties such as clay content,^5,6 organic matter/carbon content,^7,8 soil moisture,^9–11 cation exchange capacity,^2,12 and various minerals¹³ indicate that soil spectroscopy may be a viable alternative to laboratory measurements for soil characterization in many circumstances. It is reported that the prediction of soil properties with spectroscopic modeling costs <10% of conventional, laboratory-standard measurement techniques.⁴

The use of visible–near-infrared (Vis-NIR) soil reflectance data to generate models that predict soil properties is a tried-and-tested practice. Various regression,^14,15 Bayesian,¹⁶ machine learning,¹⁷ and other statistical approaches have been applied to generate prediction models. The accuracy of the resulting prediction models is highly variable, with examples of model R² ranging between 0.01 for the prediction of percent sand¹⁸ to 0.99 for the prediction of g/kg of carbon.¹⁴

Prediction models based on soil spectral data present a significant opportunity for characterizing soil attributes as soil reflectance is a cumulative property determined by the physical, chemical, and biological characteristics of soil.^19,20 For example, the wavelength at which incoming energy is absorbed, is determined by the electronic transitions (visible wavelengths) and vibrational energies (NIR wavelengths) of chemical bonds in the molecule.²¹ The amount of energy absorbed is determined by the penetration depth of light through the sample, the amount of material present, and its temperature.

Soil biology influences the chemical and physical characteristics of the soil²² by facilitating nutrient cycling, altering soil structure through the creation of aggregates, and many other interactions.²³ Physical characteristics of the soil surface also dictate the amount and direction of light scattered from the soil surface.²⁴ For example, the influence of particle size on soil reflectance is demonstrated in Figure 1,²⁵ with smaller particle sizes typically corresponding to greater reflectance in the wavelength domain shown.

Figure 1.

Demonstration by Wu et al.²⁵ that soil reflectance increases with decreasing particle size. All but the <1 mm line are samples with sizes less than the indicated range. The spectrum of a sample with mixed particle sizes is indicated by the label “<1 mm.”

Standard sample preparation procedures include grinding,^26,27 sieving,^28,29 and drying of soil samples prior to data collection.³⁰ By their nature, these standard preparation procedures alter the physical and chemical soil properties; however, numerous other unintended alterations also occur and are well documented. For example, the intentional disintegration of aggregates during sample grinding, crushing, and sieving intentionally reduces the mean particle size of a sample, but also decreases the soil surface roughness. Lower soil surface roughness (a smoother soil surface) increases soil reflectance,^31–33 also shown in Figure 1.²⁵ Losses of organic carbon and nutrients previously protected within aggregates, due to oxidization, will alter the chemical (and somewhat the physical) properties of soil.³⁴ These physical and chemical alterations resulting from sample preparation procedures may cause major changes in the shape of the soil reflectance curve, as well as the position and depth of absorption features.²⁰ Hence, in order for soil reflectance data to accurately represent the properties of soil, it is critical to maintain the physical, chemical, and biological integrity of soil samples during sampling and all other activities occurring prior to reflectance data collection.

Although numerous studies^35–37 document the effects of heating, wetting, and drying on soil properties such as particle size (Figure 1), soil surface chemistry, organic matter content, and soil structure (aggregation, cohesion, mechanical strength, and cementation), few specifically relate these effects to changes in Vis-NIR soil reflectance.

In one known study,³⁸ 227 soil samples from the CSIRO soil library were purposefully ground to five different particle sizes. Mid-infrared (MIR) soil reflectance data collected from each of the samples demonstrated that grinding reduced the variability (standard deviation) in reflectance measurements (when replicate measurements were undertaken) and increased the depth and precision (sharper, deeper, and narrower features) of absorption features associated with organic carbon and soil minerals such as kaolinite.

One study considering the parameters of the SOILSPECT radiative transfer model,²⁴ measured the specular reflectance of three soils (clay, peat, and a fine sandy soil) as a function of soil humidity (moisture content [MC]) and surface roughness (related to particle size of the soils). It was found that soil moisture affected soil reflectance across the entire Vis-NIR range measured (five bands between 450 and 2450 nm), while the spectral effects of soil surface roughness were wavelength dependent and varied by soil type.¹ Breure et al.¹ considered the influence of common sample preparation techniques on the accuracy of a partial least square regression prediction model for available phosphorus (P), potassium (K), soil organic carbon (SOC), clay, and pH, based on Vis, NIR, and MIR soil reflectance data. After averaging results from the Vis-NIR and MIR prediction models, the root mean square error (RMSE) of models trained with reflectance measurements from milled samples was lower (0.42%) than that for models trained with data from in situ (1.7%), unprocessed (2.7%), or air-dried (1.4%) samples. However, closely inspecting the results of Breure et al.,¹ there is a slight difference in the best (lowest RMSE %) result presented when only models incorporating Vis-NIR data are considered. In that case, the RMSE was 0.88% for milled samples, 0.82% for air-dried samples, 1.5% for unprocessed samples, and 1.8% for in-situ samples. In all cases (RMSE ± 0.1%) except for milled samples, models incorporating the MIR data actually had a greater RMSE than comparable models generated with Vis-NIR. One interpretation of this result is that milling improves the performance of soil property prediction models compiled with MIR spectral data, but less so for models generated with Vis-NIR data. More generally, the results of Breure et al.¹ were significantly different once the samples were processed in any way.

Soil spectral measurements (collected in the lab) are occasionally included as a data source in remote sensing soil prediction models. Examples of studies incorporating soil spectra from soil spectral libraries,^39,40 soil libraries,⁴¹ and samples collected in the field⁴² exist. For instance, Biney et al.⁴² compared the performance of soil organic carbon (SOC) predictive models prepared only with soil spectral data collected in the field or in the lab (prior to and after sample drying) with the performance of a model prepared only with Sentinel-2 remote sensing data. They found that SOC prediction accuracy (R² and RMSE) was higher in the models incorporating soil spectral data from dry soil samples (with the model R² in the range of 0.7–0.92 and model RMSE in the range of 1.00–1.20), compared to wet soil samples, and between data collected in the lab (R²= 0.54, RMSE = 2.40) and in the field (R²= 0.5, RMSE = 2.50). However, Biney et al.⁴² did not consider the effects of additional sample handling (e.g., rewetting the samples) on their reflectance measurements, or subsequent effects on model performance.

In each of the published studies described above, the spectral effects of drying, wetting, and other handling procedures of soil samples were not reported. Hence, the present paper establishes the significance of soil drying and repeated cycles of drying and wetting on soil reflectance. First, new data demonstrating the major spectral effects of drying, then subsequent wetting and drying cycles on Vis and NIR (400–2500 nm) soil reflectance are presented. Secondly, explanations for the alterations seen in soil reflectance are explored. Finally, the implications of these results for remote sensing, soil libraries, soil spectral libraries, the transferability of soil spectral data, and predictive models based on soil spectroscopic data are discussed. It is hoped that these results and the associated discussion will inspire within relevant fields: firstly, an appreciation of the magnitude of the spectral effects of drying on soil reflectance; secondly, a critical reflection on the suitability, applicability, and appropriate use of spectral data from soil spectral databases, and thirdly, further consideration of the ideal spectral data collection methods for the prediction of soil properties in the field.

By applying these insights, it may be possible to avoid or correct systematic bias introduced by standard soil preparation procedures including drying. Indeed, a correction model for the samples included in this study is presented. In turn, the use of accurate and representative spectral data is expected to improve experimental and modeling results, improving the reliability of models used to estimate soil properties with Vis-NIR reflectance data.

Experimental

Materials and Methods

In 2019, nine soil samples were collected from a grazing paddock near Muttama, New South Wales (NSW), Australia, by Bishop and colleagues from the University of Sydney. The sampled paddock is situated 290–340 m above sea level on a southwest-facing slope. The relative position of each sampling site is indicated in Figure 2.

Figure 2.

Sampling locations for nine samples used in this experiment were in a grazing paddock near Muttama, NSW. A road is visible at the top of the image as well as multiple paddock trees (green dots) and a cultivated field to the right (east).

After collection in 2019, the samples were dried at 40 °C then stored in plastic, screw-top jars. Clay, silt, sand, and organic matter content, soil texture, and bulk density data were collected for each sample (Table I). The soil texture classifications made were clay-loam (CL), loam (L), silty-loam (ZL), and sandy-CL (SCL).

Table I.

Particle size distribution (clay, silt, and sand %), soil texture, organic matter content, and bulk density of Muttama soil samples near the time of collection.

Sample	Clay (%)	Silt (%)	Sand (%)	Soil texture	Organic matter (%)	Bulk density (g/cm³)
ISS10A	25.41	18.65	55.93	CL	3.86	1.30
ISS14A	22.12	17.05	60.84	L	3.19	1.36
ISS17A	25.06	20.89	54.05	CL	4.38	1.27
ISS20A	21.10	25.83	53.07	ZL	2.24	1.41
ISS25A	27.19	15.86	56.95	CL	3.48	1.33
ISS26A	18.28	28.56	53.16	ZL	4.26	1.27
ISS27A	20.24	6.84	72.93	SCL	3.71	1.34
ISS28A	18.54	26.02	55.43	ZL	2.83	1.37
ISS36A	14.47	27.58	57.94	ZL	3.91	1.30

The reflectance data collection started by placing 50 g of each stored soil sample in a Petri dish and reducing the MC to 0% by drying for 48 h at 105 °C in a sample drying oven. This drying procedure is recommended by Australian Department of Sustainable Natural Resources⁴³ to fully dry samples. Water was then incrementally added to each sample with a pipette, intermittent to reflectance data collection at each MC. The MC of each sample was incrementally increased by 1% or 5% MC intervals, with the amount of water added to reach the desired soil MC calculated using the following equation:⁴³

Gravimetric soil water content (%) = [\frac{(mass of moist soil (g) - mass of oven - driedsoil (g))}{mass of oven - dried soil (g)}] \times 100

(1)

For instance, 2.5 mL of water was added to 50 g of the oven-dry soil in a Petri dish and mixed with a clean spoon to increase the soil MC by 5%, or 0.5 mL of water was added and mixed in to increase the soil MC by 1%. Reasonable efforts were made not to crush soil aggregates or apply excessive force during mixing.

Once the soil became fully saturated (with water pooling in the Petri dish), reflectance data collection was halted, and the sample returned to the oven for further drying. The sample was left in the soil drying oven at 105 °C for 48 h, as per the initial drying activity. Reflectance data were then collected in two additional rounds at all moisture levels measured in the first round. Results from the initial data collection are referred to as Round 1 (R1) data, while data from the two latter data collection activities are referred to as Round 2 (R2) and Round 3 (R3) data, respectively.

Soil reflectance data were collected using an ASD FieldSpec Spectroradiometer (Malvern Panalytical) in the range of 400–2500 nm with a handheld contact probe according to the standardized soil reflectance data collection procedures recommended by Viscarra Rossel et al.⁴⁴ and Shepherd et al.⁴⁵ Five reflectance measurements were collected at each MC, with the average of the five measurements shown below. Between each sample and each MC, the ASD was calibrated with a white Spectralon tile.⁴⁶ Data cleaning, visualization, and plotting were undertaken in Python using the Pandas, NumPy, Matplotlib, and Seaborn libraries.

Results and Discussion

Soil reflectance data from this study demonstrate a significant, wavelength-dependent, increase in the dry (0% moisture) soil reflectance following the first drying and wetting cycle (between the R1 and R2 data), both when averaged over the nine soil samples (Figure 3) and for each soil individually (Figures 4 and A1). In every case (for each sample), the reflectance in R2 is greater than the reflectance in R1 at all wavelengths. This data show a clear effect of drying and wetting on soil spectra, with soil reflectance increasing by 25–43% (Figure 5) for wavelengths between 650 and 2200 nm after the first round of drying and wetting. The increase in reflectance between R1 and R2 increases with wavelength to ∼1800 nm (Figures 3, 4, and A1), but the increases are proportional to the initial reflectance (Figure 5). For more than half of the samples (Figures 4 and A1), the soil reflectance in R2 is more than one standard deviation above the mean of the reflectance measurements recorded in R1 at all wavelengths considered. We consider differences <2% (the average standard deviation across all wavelengths) not to be significant.

Figure 3.

The mean reflectance from the dry (0% MC) Muttama samples for all three rounds (R1–R3) with the standard deviations as functions of wavelength, color-coded for each round as shading around the mean curves.

Figure 4.

(a–d) The dry (0% MC) soil reflectance for samples (a) ISS17 and (c) ISS25 in R1–R3, and (b and d) the changes in reflectance between subsequent rounds of drying and wetting.

Figure 5.

Proportional change in reflectance as a function of wavelength between R1 and R2 for each of the nine Muttama samples, and the average of all samples. Linear models demonstrating the suggested correction factors for three wavelength domains, 400–650 nm, 650–2200 nm, and 2200–2500 nm, are also shown.

On average (Figure 3), there is not a significant difference between the reflectance in the R2 and R3 of measurement. However, when soils are considered individually (Figures 4 and A1), the soil reflectance in R3 is more than one standard deviation above the mean reflectance for R2 for some samples (e.g., ISS25 and ISS26). Note that the differences are wavelength-dependent and most pronounced in the range of 1000–1800 nm.

The qualitative shape of the reflectance curves and the presence of small absorption features around 1400, 1900, and 2200 nm were maintained for all samples in all rounds, as shown in Figures 3, 4, and A1.

In all three rounds (R1–R3) of data collection, dry (0% MC) reflectance increased (Figures 4 and A1) between 400 and 1350 nm for all nine samples, with the greatest spectral angle (steepest increase)^47,48 between 650 and 750 nm. Around 1400 nm, an absorption feature associated with the vibration of oxygen–hydrogen (O–H) bonds is apparent. This absorption feature has a typical depth of ∼0.02 reflectance units (2% reflectance). Beyond this feature, reflectance continues to increase until ∼1900 nm where the second O–H absorption feature is apparent. This feature is deeper than the feature near 1400 nm and there is a high spectral angle on the far side of this absorption feature before a peak in the spectral curve is reached at ∼2130 nm. Another absorption feature is centered at 2210 nm. The spectral curve then oscillates gently before declining toward 2500 nm.

In R2 and R3, the reflectance is higher than in R1 at all wavelengths considered, with larger spectral angles in the 650–750 nm range than in R1 (Figures A1a, A1c, A1e, A1g, A1i, A1k, and A1m). Note that the dry spectrum for each round is measured at the beginning of each wetting and drying cycle.

The standard deviation (Figures 4a, 4c, and A1) of the five reflectance measurements collected at each wavelength in each round is variable between samples and between rounds, but typically decreases in later rounds. This result is consistent with reports in the literature of wetting and drying cycles altering soil and soil surface properties such as aggregate concentration and surface roughness, thereby increasing the homogeneity of a sample.^2,20

The proportional change in soil reflectance identified following the drying and wetting cycles is calculated with:

Δ S_{1, 2} = \frac{(R_{2} (λ) – R_{1} (λ))}{R_{1} (λ)}

(2)

where R_i(λ) is the measured soil reflectance at each wavelength with i = 1 and 2 for R1 and R2, respectively. Although the absolute reflectance consistently increases across the Vis and NIR range with increased sample handling (Figures 3 and 4), it is important to note that the proportional change in soil reflectance between R1 and R2 (given by Eq. 2) is close to constant from 650 to 2200 nm (Figure 5). This result is found for the average over all samples and for each sample individually. Outside of the range 650–2200 nm, the proportional change in reflectance between R1 and R2 rises sharply between 400 and 650 nm, and more gently above 2000 nm. The 1900 nm absorption feature stands out in Figures 3 to 5, with the size of the absorption feature (Figures 4 and A1) increasing in some cases (ISS17 and ISS25), remaining similar (ISS20 and ISS25), or decreasing (ISS10, ISS14, ISS26, ISS28, and ISS36) in other cases.

In Figure 5, between 400 and 650 nm, the proportional change in soil reflectance is in the range of 1–26% and the change in reflectance between R1 and R2 increases approximately linearly with wavelength. In the range of 650–2200 nm, the proportional change is in the range of 10–39% but is effectively constant with wavelength. In the range of 2200–2500 nm, the proportional changes are even greater and increase with wavelength up to a maximum of 43% for sample ISS20 (Figure 5). The existence of these three wavelength domains in which ΔS_1,2 varies approximately linearly suggests a method of correcting spectra from processed samples with linear correction models in the three wavelength domains. Figure 5 presents the combined model for the sample averaged quantity ΔS_1,2 for this data. The linear model in each domain is detailed further in Table II, with coefficients and intercepts defined.

Table II.

Correction factors for these wetted and dried samples for three wavelength domains.

Range of λ	400–650 nm	650–2200 nm	2200–2500 nm
Gradient m	0.08	0	0.01
Intercept b	–24.5	27.38	5.5

Based on the results presented in Figure 5, the model appears reasonable. It may be possible to recalibrate soil reflectance and the model defined by Eq. 2 and Table II with data collected from samples that have been dried and wetted, through the development of wavelength-specific correction factors described next.

The suggested correction approach from Eq. 2 and Figure 5 is to write

Δ S_{1, 2} = \frac{(R_{2} (λ) - R_{1} (λ))}{R_{1} (λ)} = m l + b

(3)

which can be rearranged to yield

R_{1} (λ) = \frac{R_{2} (λ)}{(m l + b + 1)}

(4)

where R₁(λ) is the original, corrected, spectra, R₂(λ) is the measured spectra from a processed sample, m is the gradient of the line segment and b is the intercept. Thus, the combination of a measured reflectance spectra R₂(λ) and the linear model coefficients in Table II allows estimation of the reflectance spectrum R₁(λ) for the once-dried and otherwise non-handled soil sample. It is relevant that the model represented by Eqs. 2 and 4 and Table II does not reconstruct the reflectance spectrum of the field sample. However, they illustrate one possible approach for this and demonstrate the importance of such reconstructions and correction procedures for multiple dried and wetted samples. Of course, correction factors relevant to other wavelengths or bands (such as satellite remote sensing bands) would need to consider an appropriate adjustment for each parameter (e.g., NIR ×0.3, Blue ×0.2).

The reflectance spectra of nine soil samples from a grazing paddock in NSW, Australia, show significant alteration as a result of drying and wetting. This alteration is clearly apparent as a substantial increase in soil reflectance between R1 and R2 of reflectance data collection. By comparison, reflectance typically changed little between R2 and R3. Changes in Vis-NIR soil reflectance resulting from sample preparation activities are clearly apparent. These are due to the drying and wetting cycles and correspond to the alteration of intrinsic soil properties. In principle, based on other analyses, these alterations include changes to the organic matter content, particle size distribution, soil surface chemistry, bulk density (pore size), and soil structure.

For instance, the physical and chemical effects of soil wetting and drying are summarized succinctly by Bartlett and James⁴⁹ in a study comparing the chemistry and behavior of a soil stored and then rewetted in the lab versus a soil that was collected moist from the field. They identified immediate changes in soil surface chemistry and soil organic matter as a result of drying. During storage, samples were found to be metastable with their characteristics varying through time, especially if a rewetting/moistening event occurred during the storage period.⁴⁹ Also found that the behavior of a rewetted soil differed from that of a soil that maintained its MC during storage. Hence, they recommended that soil properties are best maintained for short-term storage and transport in a sealed polyurethane bag near field capacity (a measure of MC), and for longer periods at 4 °C, while still moist. They also noted that most of their results had been previously discovered but were too often forgotten or ignored in the development and application of standard soil handling procedures.

The recommendation of Bartlett and James⁴⁹ is consistent with the results of numerous other studies which have recorded a drying event causing irreversible changes in a soil.^50–52 For example, Karube and Abe⁵⁰ found the maximum water-holding capacity of aliphatic (volcanic) soils from Japan reduced by 23–29% after air-drying. For other soils, changes in pore size (the space between soil particles),^37,53,54 soil aggregation,^55,56 mechanical strength,⁵⁷ biological properties, and soil structure⁵⁸ have been recorded after a drying event. Mineral or microbial-induced cementation^36,59 or hydrophobicity⁶⁰ may also occur as a result of drying soil samples.

In a study concerning the abundance and type of soil microbial communities in Cambisol soils on a sheep-grazing property in North Lancashire, England,⁶¹ it was found that a soil wetting and drying cycle caused a significant reduction in soil fungi, the ratio of fungi to bacteria, and the total weight of soil bacteria. The study also identified a loss of nitrogen and carbon from the soils during rewetting, caused by nutrient leaching. Similar results were obtained by Miller et al.⁶² for chaparral (thin, rocky, and nutrient-poor) soils from California; however, that study found that losses in carbon were associated with soil respiration. These studies are mentioned as they highlight the potential loss of dissolved and other labile carbon fractions during a drying and wetting cycle.^61,62

Soil grinding and aggregate disintegration are other common practices applied in soil laboratories and are recommended as a standard procedure prior to the collection of soil reflectance data.⁴⁴ Crushing and sieving soil to particle sizes ≤2 mm is also typical;^20,44 however, the exact sieve mesh size and method vary according to the project research aims. For example, studies concerned with aggregates may include particles anywhere in the range of 1–9 mm in diameter.^35,63–65 Crushing, grinding, and sieving procedures are designed to homogenize soil samples to allow more representative subsampling²⁰ and reduce the effects of surface roughness on soil reflectance.^2,33

Stenberg et al.²⁰ claim that there is “no doubt” that soil grinding has a “substantial effect” on soil reflectance. The increase in reflectance is reportedly particularly pronounced for clay-rich soils with a high proportion of aggregates. The disintegration of aggregates during grinding increases soil reflectance due to a decrease in mean particle size and soil surface roughness,²⁰ and so changing the amount of Rayleigh scattering from the soil surface (among other effects). The link between soil reflectance and particle size is well established,^25,31,66,67 as in Figure 1. The effective composition of the soil surface will also change upon crushing or grinding, as soil components such as organic matter break down,^68,69 and components previously hidden inside aggregates become exposed.

Data from this study demonstrates that the drying and wetting of soil samples changes their reflectance across the Vis and NIR wavelengths. Though the impact on soil reflectance was greatest following the first round of drying and wetting, each round resulted in a distinct change (usually an increase) in soil reflectance across all wavelengths measured. Between measurement rounds, soil reflectance curves (Figures 4 and A1) generally maintained their form and qualitative shape, suggesting that increased reflectance in later rounds was a product of particle fining and/or decreases in surface roughness^19,70,71 instead of the alteration of the soil mineral composition. Specifically, the maintenance of the absorption feature centered at 2210 nm, which is associated with Fe-oxides, chloride salts, and other hydrated mineral complexes^72,73 remains throughout all three rounds (R1–R3) of data collection.

From the small differences between the R2 and R3 spectra, it is plausible that the spectral effects of multiple drying–wetting cycles would converge to a stable reflectance spectrum once all particle and biochemical processes have occurred. However, crucially, as demonstrated here, the soil reflectance is permanently altered and this stable, altered, reflectance spectrum is significantly different from the spectrum after minimal processing. This is because the sample no longer represents the original sample, but instead an intrinsically modified sample.

It is thus clear that the handling and processing of soil samples, including standard drying and homogenization processes, alter the soil reflectance spectra and should be avoided if possible and should be corrected for in real-world comparisons. Thus, since they routinely use these processes, soil libraries and soil spectral libraries have very limited utility for characterizing soils as they would occur in the field, if the spectral data cannot be appropriately corrected. Also, soil samples or soil data shared between research groups should be accompanied by a full, detailed disclosure of sample handling, preparation, and data collection procedures applied. This will allow other potential users to consider the comparability and usefulness of such samples or data, apply appropriate corrections, and identify any likely sources of error.

The models described in Eqs. 3 and 4 and Table II are one means of approximating the correction factors required to convert between the spectrum of a dried sample and the spectrum of the same sample after an additional drying and wetting cycle, for these samples. Other correction models may be required to account for other sample preparation processes.

Also, instead of drying soil samples, it may be possible to remove the effect of soil moisture on soil spectra without drying soil samples when soil reflectance data is available at two known MCs. This is because strong (usually linear) relationships exist between soil MC and soil reflectance at multiple wavelength combinations.⁷⁴ Hence, these known relationships, calibrated for each soil as per the recommendations of McGuirk and Cairns,⁷⁴ could be used to model the dry reflectance of a sample.

Despite reports that soil spectroscopists are fully capable of predicting soil properties from (processed) soil spectroscopic measurements and avoiding problems such as model over-fitting,⁴ it is clear from at least our analyses and the cited literature that the effects of sample handling remain a source of systematic error in soil spectral data. Instead, spectral modeling of both in-field and dry soil reflectance may be an option for avoiding the incorporation of error from sample handling activities, and its potential should be investigated further. The importance of validating spectroscopic models is well recognized. However, it remains unclear how model validation could eliminate the effects of sample handling when all samples are subject to the same sample preparation activities known to alter soil spectral, physical, and chemical characteristics. This is true even when techniques such as “explainable” machine learning are applied.

Thus, despite being first recognized more than 40 years ago, the insights and recommendations regarding the effects of sample drying and storage on soil properties detailed by Bartlett and James⁴⁹ remain highly relevant. All remote sensing and soil spectroscopic models include an underlying assumption that spectra and other data collected from soil samples are representative of soil properties in the field at the time of sample collection. Though drying and milling samples prior to data collection are known to improve the statistical performance and precision of spectroscopic prediction models in some instances,¹ presumably by homogenizing the samples, the present analysis and numerous other studies document the intrinsic changes to reflectance spectra and soil samples caused by drying, wetting, grinding, and crushing.^{20,37,54,55,58,69} Ignoring these results has caused and will cause systematic error to be incorporated into spectroscopic (remote sensing and ground-based) model validation and calibration data sets. Clearly, this is an undesirable and unacceptable practice.

Instead of limiting spectral data collection to the laboratory environment, it is suggested that when feasible, spectral data is collected three times: in situ in the field, once the sample is collected, then again in the laboratory once dried or otherwise processed prior to storage or further testing. Analysis of these three spectral data sets will eventually allow correction and calibration factors to be determined. In turn, this will facilitate the correction of current and previous spectral data collected in the laboratory to ensure such data is representative of the soil in its field condition.

Conclusion

Standard soil sample preparation procedures applied in soil laboratories significantly alter soil reflectance spectra. In particular, by studying the effects of sample wetting and drying for nine soil samples from south-eastern Australia, significant changes in soil reflectance are demonstrated following a single cycle of sample drying and wetting. Thus, this paper's analysis and associated discussion demonstrate that soil libraries and other soil repositories where samples have undergone multiple drying–wetting cycles, destructive homogenization procedures, or biochemical processing, are of very limited utility for comparisons with field spectra, predicting soil properties in the field, or remote sensing spectral analyses. Arguably, they are only useful if the effects of these alterations can be avoided or corrected. One simple linear correction model for three wavelength domains, developed for the nine soil samples analyzed, is presented, acknowledging that this model does not account for soil-to-soil variability.

Footnotes

Acknowledgments

We thank T. Bishop and colleagues who collected the samples and provided access to the laboratory equipment. We also thank CUAVA, the ARC Centre for CubeSats, UAVs and their Applications for the provision of a scholarship, in part, to support this work.

Data Availability Statement

The data which support this study are available at https://unisyd-my.sharepoint.com/:x:/r/personal/savannah_mcguirk_sydney_edu_au/_layouts/15/Doc.aspx?sourcedoc=%7B64E96E44-7DB7-4B83-B488-ED00B2705220%7D&file=dryspectra.csv&wdLOR=cBA43AC1F-2E4C-AE49-822E-B577229627E6&action=default&mobileredirect=true.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors would like to thank the Australian Research Council (ARC) for funding (grant ID: IC170100023) the Australian Research Council Centre for CubeSats, UAVs, and their Applications (CUAVA), based at the University of Sydney, which in turn, supported this work.

ORCID iD

Savannah L. McGuirk

Appendix

References

Breure

T.S.

Prout

J.M.

Haefele

S.M.

Milne

A.E.

, et al. “Comparing the Effect of Different Sample Conditions and Spectral Libraries on the Prediction Accuracy of Soil Properties From Near- and Mid-Infrared Spectra at the Field-Scale”. Soil Tillage Res. 2022. 215: 105196. https://doi.org/10.1016/j.still.2021.105196

Ben-Dor

Banin

. “Near-Infrared Analysis as a Rapid Method to Simultaneously Evaluate Several Soil Properties”. Soil Sci. Soc. Am. J. 1995. 59(2): 364–372. https://doi.org/10.2136/sssaj1995.03615995005900020014x

Gomez

Lagacherie

. “Mapping of Primary Soil Properties Using Optical Visible and Near Infrared (Vis-NIR) Remote Sensing“. In: N. Baghdadi, M. Zribi, editors. Land Surface Remote Sensing in Agriculture and Forest. London, Oxford: ISTE Press Ltd; Elsevier, 2016. Pp. 1–35.

Viscarra Rossel

R.A.

Behrens

Ben-Dor

Chabrillat

, et al. “Diffuse Reflectance Spectroscopy for Estimating Soil Properties: A Technology for the 21st Century”. Eur. J. Soil Sci. 2022. 73(4): E13271. https://doi.org/10.1111/ejss.13271

Yaron

Faigenbaum-Golovin

Granot

Shkolnisky

, et al. “Removing Moisture Effect on Soil Reflectance Properties: A Case Study of Clay Content Prediction”. Pedosphere. 2019. 29(4): 421–431. https://doi.org/10.1016/S1002-0160(19)60811-8

Post

J.L.

Noble

P.N.

. “The Near-Infrared Combination Band Frequencies of Dioctahedral Smectites, Micas, and Illites”. Clays Clay Miner. 1993. 41(6): 639–644. https://doi.org/10.1346/CCMN.1993.0410601

Hong

Guo

Chen

Linderman

, et al. “Exploring the Potential of Airborne Hyperspectral Image for Estimating Topsoil Organic Carbon: Effects of Fractional-Order Derivative and Optimal Band Combination Algorithm”. Geoderma. 2020. 365: 114228. https://doi.org/10.1016/j.geoderma.2020.114228

Žížala

Minařík

Zádorová

. “Soil Organic Carbon Mapping Using Multispectral Remote Sensing Data: Prediction Ability of Data with Different Spatial and Spectral Resolutions”. Remote Sens. 2019. 11(24): 2947. https://doi.org/10.3390/rs11242947

Wang

Zhang

Johnson

V.C.

. “New Methods for Improving the Remote Sensing Estimation of Soil Organic Matter Content (SOMC) in the Ebinur Lake Wetland National Nature Reserve (ELWNNR) in Northwest China”. Remote Sens. Environ. 2018. 218: 104–118. https://doi.org/10.1016/j.rse.2018.09.020

10.

Lobell

D.B.

Asner

G.P.

. “Moisture Effects on Soil Reflectance”. Soil Sci. Soc. Am. J. 2002. 66(3): 722–727. https://doi.org/10.2136/sssaj2002.7220

11.

McGuirk

S.L.

Cairns

I.H.

. “Soil Moisture Prediction with Multispectral Visible and NIR Remote Sensing”. ISPRS Ann. Photogram. Remote Sens. Spatial Inform. Sci. 2022. 3: 447–453. https://doi.org/10.5194/isprs-annals-V-3-2022-447-2022

12.

Nduwamungu

Ziadi

Tremblay

G.F.

Parent

L-É

. “Near-Infrared Reflectance Spectroscopy Prediction of Soil Properties: Effects of Sample Cups and Preparation”. Soil Sci. Soc. Am. J. 2009. 73(6): 1896–1903. https://doi.org/10.2136/sssaj2008.0213

13.

Omran

E.S.E.

. “Rapid Prediction of Soil Mineralogy Using Imaging Spectroscopy”. Eurasian Soil Sci. 2017. 50(5): 597–612. https://doi.org/10.1134/S106422931705012X

14.

Madari

B.E.

Reeves

J.B.

Machado

P.L.

Guimarães

C.M.

, et al. “Mid- and Near-Infrared Spectroscopic Assessment of Soil Compositional Parameters and Structural Indices in Two Ferralsols”. Geoderma. 2006. 136(1–2): 245–259. https://doi.org/10.1016/j.geoderma.2006.03.026

15.

Cozzolino

Moron

. “The Potential of Near-Infrared Reflectance Spectroscopy to Analyse Soil Chemical and Physical Characteristics”. J. Agric. Sci. 2003. 140(1): 65–71. https://doi.org/10.1017/S0021859602002836

16.

Guerrero

Javadi

S.H.

Mouazen

A.M.

. “Automatic Detection of Quality Soil Spectra in an Online Vis-NIR Soil Sensor”. Comput. Electron. Agric. 2022. 196: 106857. https://doi.org/10.1016/j.compag.2022.106857

17.

Bai

Z.J.

Xie

M.D.

B.F.

Luo

D.F.

, et al. “Estimation of Soil Organic Carbon Using Vis-NIR Spectral Data and Spectral Feature Bands Selection in Southern Xinjiang, China”. Sensors. 2022. 22(16): 6124. https://doi.org/10.3390/s22166124

18.

Chang

C.W.

Laird

D.A.

Hurburgh

C.R.

. “Influence of Soil Moisture on NIR Reflectance Spectroscopic Measurement of Soil Properties”. Soil Sci. 2005. 170(4): 244–255.

19.

Ben-Dor

. “Quantitative Remote Sensing of Soil Properties”. Adv. Agron. 2002. 75: 173–243. https://doi.org/10.1016/s0065-2113(02)75005-0

20.

Stenberg

Viscarra Rossel

R.A.

Mouazen

A.M.

Wetterlind

. “Visible and Near Infrared Spectroscopy in Soil Science”. In: Sparks

D.L.

, editor. Advances in Agronomy. Amsterdam: Elsevier, 2010. Pp. 163–215.

21.

Condit

. “Application of Characteristic Vector Analysis to the Spectral Energy Distribution of Daylight and the Spectral Reflectance of American Soils”. Appl. Opt. 1972. 11(1): 74–86. https://doi.org/10.1364/AO.11.000074

22.

Abbott

L.K.

Murphy

D.V.

. Soil Biological Fertility: A Key to Sustainable Land Use in Agriculture. Dordrecht, The Netherlands: Kluwer Academic Publishing, 2003.

23.

Abbott

L.K.

Murphy

D.V.

. “What is Soil Biological Fertility?”. In: L.K. Abbott, D. V. Murphy, editors. Soil Biological Fertility: A Key to Sustainable Land Use in Agriculture. Dordrecht, The Netherlands: Springer, 2007. Pp. 1–15.

24.

Jacquemoud

Baret

Hanocq

. “Modeling Spectral and Bidirectional Soil Reflectance”. Remote Sens Environ. 1992. 41(2–3): 123–132. https://doi.org/10.1016/0034-4257(92)90072-R

25.

Zheng

Yang

. “Effects of Different Particle Sizes on the Spectral Prediction of Soil Organic Matter”. Catena. 2021. 196: 104933. https://doi.org/10.1016/j.catena.2020.104933

26.

Dalal

R.C.

Henry

R.J.

. “Simultaneous Determination of Moisture, Organic Carbon, and Total Nitrogen by Near Infrared Reflectance Spectrophotometry”. Soil Sci. Soc. Am. J. 1986. 50(1): 120–123. https://doi.org/10.2136/sssaj1986.03615995005000010023x

27.

Reeves

McCarty

Mimmo

. “The Potential of Diffuse Reflectance Spectroscopy for the Determination of Carbon Inventories in Soils”. Environ. Pollut. 2002. 116(S1): S277–S284. https://doi.org/10.1016/s0269-7491(01)00259-7

28.

Russell

C.A.

. “Sample Preparation and Prediction of Soil Organic Matter Properties by Near Infra-Red Reflectance Spectroscopy”. Commun. Soil Sci. Plant Anal. 2003. 34(11–12): 1557–1572. https://doi.org/10.1081/CSS-120021297

29.

Brunet

Barthès

B.G.

Chotte

J.-L.

Feller

. “Determination of Carbon and Nitrogen Contents in Alfisols, Oxisols, and Ultisols from Africa and Brazil Using NIRS Analysis: Effects of Sample Grinding and Set Heterogeneity”. Geoderma. 2007. 139(1–2): 106–117. https://doi.org/10.1016/j.geoderma.2007.01.007

30.

Morgan

C.L.

Grunwald

Brown

D.J.

Sarkhot

D.V.

. “Comparison of Soil Reflectance Spectra and Calibration Models Obtained Using Multiple Spectrometers”. Geoderma. 2011. 161(3–4): 202–211. https://doi.org/10.1016/j.geoderma.2010.12.020

31.

Baumgardner

M.F.

Silva

L.F.

Biehl

L.L.

Stoner

E.R.

. “Reflectance Properties of Soils”. Adv. Agron. 1986. 38: 1–44. https://doi.org/10.1016/S0065-2113(08)60672-0

32.

Orlov

D.S.

. “Quantitative Patterns of Light Reflection by Soils. 1. Influence of Particle (Aggregate) Size on Reflectivity”. Dokl. Soil Sci. 1966. 13S: 1495.

33.

Dunn

B.W.

Beecher

H.G.

Batten

G.D.

Ciavarella

. “The Potential of Near-Infrared Reflectance Spectroscopy for Soil Analysis: A Case Study from the Riverine Plain of South-Eastern Australia”. Aust. J. Exp. Agric. 2002. 42(5): 607–614. https://doi.org/10.1071/ea01172

34.

Kravchenko

A.N.

Negassa

W.C.

Guber

A.K.

Rivers

M.L.

. “Protection of Soil Carbon Within Macro-Aggregates Depends on Intra-Aggregate Pore Characteristics”. Sci. Rep. 2015. 5(1): 16261. https://doi.org/10.1038/srep16261

35.

Pojasok

Kay

B.D.

. “Assessment of a Combination of Wet Sieving and Turbidimetry to Characterize the Structural Stability of Moist Aggregates”. Can. J. Soil Sci. 1990. 70(1): 33–42. https://doi.org/10.4141/cjss90-004

36.

Gowthaman

Nakashima

Kawasaki

. “Effect of Wetting and Drying Cycles on the Durability of Bio-Cemented Soil of Expressway Slope”. Int. J. Environ. Sci. Technol. 2021. 19: 2309–2322. https://doi.org/10.1007/s13762-021-03306-1

37.

Kodikara

Barbour

S.L.

Fredlund

D.D.

. “Changes in Clay Structure and Behaviour Due to Wetting and Drying”. Paper presented at: 8th Australian–New Zealand Conference on Geomechanics Hobart, Tasmania. 1999; Feb 14 - Feb 19. Pp. 179–186.

38.

Le Guillou

Wetterlind

Viscarra Rossel

R.A.

Hicks

, et al. “How Does Grinding Affect the Mid-Infrared Spectra of Soil and Their Multivariate Calibrations to Texture and Organic Carbon?”. Soil Res. 2015. 53(8): 913–921. https://doi.org/10.1071/SR15019

39.

Castaldi

Chabrillat

Don

Van Wesemael

. “Soil Organic Carbon Mapping Using LUCAS Topsoil Database and Sentinel-2 Data: An Approach to Reduce Soil Moisture and Crop Residue Effects”. Remote Sens. 2019. 11(18): 2121. https://doi.org/10.3390/rs11182121

40.

Wang

Guan

K.Y.

Zhang

C.H.

Lee

, et al. “Using Soil Library Hyperspectral Reflectance and Machine Learning to Predict Soil Organic Carbon: Assessing Potential of Airborne and Spaceborne Optical Soil Sensing”. Remote Sens. Environ. 2022. 271: 112914. https://doi.org/10.1016/j.rse.2022.112914

41.

Mzid

Castaldi

Tolomio

Pascucci

, et al. “Evaluation of Agricultural Bare Soil Properties Retrieval from Landsat 8, Sentinel-2 and PRISMA Satellite Data”. Remote Sens. 2022. 14(3): 714. https://doi.org/10.3390/rs14030714

42.

Biney

J.K.M.

Vasat

Bell

S.M.

Kebonye

N.M.

, et al. “Prediction of Topsoil Organic Carbon Content with Sentinel-2 Imagery and Spectroscopic Measurements Under Different Conditions Using An Ensemble Model Approach with Multiple Pre-Treatment Combinations”. Soil Tillage Res. 2022. 220: 105379. https://doi.org/10.1016/j.still.2022.105379

43.

Australian Department of Sustainable Natural Resources. “Soil Survey Standard Test Method Soil Moisture Content”. 1990. https://www.environment.nsw.gov.au/resources/soils/testmethods/mc.pdf [accessed Jun 21 2024].

44.

Viscarra Rossel

R.A.

Behrens

Ben-Dor

Brown

D.J.

, et al. “A Global Spectral Library to Characterise the World's Soil”. Earth-Sci. Rev. 2016. 1(155): 198–230. https://doi.org/10.1016/j.earscirev.2016.01.012

45.

Shepherd

K.D.

Ferguson

Hoover

van Egmond

, et al. “A Global Soil Spectral Calibration Library and Estimation Service”. Soil Secur. 2022. 7: 100061. https://doi.org/10.1016/j.soisec.2022.100061

46.

Labsphere. “Spectralon Diffuse Reflectance Material”. https://www.labsphere.com/product/spectralon-diffuse-reflectance-material/ [accessed Oct 10 2023].

47.

Ward

K.J.

Chabrillat

Neumann

Foerster

. “A Remote Sensing Adapted Approach for Soil Organic Carbon Prediction Based on the Spectrally Clustered LUCAS Soil Database”. Geoderma. 2019. 353: 297–307. https://doi.org/10.1016/j.geoderma.2019.07.010

48.

Chang

C.-I.

. “Hyperspectral Target Detection: Hypothesis Testing, Signal-to-Noise Ratio, and Spectral Angle Theories”. IEEE Trans. Geosci. Remote Sens. 2022. 60: 1–23. https://doi.org/10.1109/TGRS.2021.3069716

49.

Bartlett

James

. “Studying Dried, Stored Soil Samples: Some Pitfalls”. Soil Sci. Soc. Am. J. 1980. 44(4): 721–724. https://doi.org/10.2136/sssaj1980.03615995004400040011x

50.

Karube

Abe

. “Water Retention by Colloidal Allophane and Imogolite With Different Charges”. Clays Clay Miner. 1998. 46(3): 322–329. https://doi.org/10.1346/CCMN.1998.0460311

51.

Kubota

. “Aggregate-Formation of Allophanic Soils: Effect of Drying on the Dispersion of the Soils”. Soil Sci. Plant Nutr. 1972. 18(2): 79–97. https://doi.org/10.1080/00380768.1972.10433277

52.

Maeda

Takenaka

Warkentin

. “Physical Properties of Allophane Soils”. In: Advances in Agron. Academic Press. Amsterdam, The Netherlands: Elsevier, 1977. Pp. 229–264.

53.

Zhao

Wang

. “Soil-Water Characteristic Curve for Unsaturated Loess Soil Considering Density and Wetting-Drying Cycle Effects”. Zhongnan Daxue Xuebao (Ziran Kexue Ban) [J. Cent. South Univ. (Sci. Technol.)]. 2012. 43(6): 2445–2453.

54.

Monroy

Zdravkovic

Ridley

. “Evolution of Microstructure in Compacted London Clay During Wetting and Loading”. Géotechnique. 2010. 60(2): 105–119. https://doi.org/10.1680/geot.8.P.125

55.

Lado

Ben-Hur

Shainberg

. “Soil Wetting and Texture Effects on Aggregate Stability, Seal Formation, and Erosion”. Soil Sci. Soc. Am. J. 2004. 68(6): 1992–1999. https://doi.org/10.2136/sssaj2004.1992

56.

Layton

J.B.

Skidmore

E.L.

Thompson

C.A.

. “Winter-Associated Changes in Dry-Soil Aggregation as Influenced by Management”. Soil Sci. Soc. Am. J. 1993. 57(6): 1568–1572. https://doi.org/10.2136/sssaj1993.03615995005700060029x

57.

Gao

Q.-F.

Zeng

Shi

Z.-N.

Zhang

. “Evolution of Unsaturated Shear Strength and Microstructure of a Compacted Silty Clay on Wetting Paths”. Int. J. Geomech. 2021. 21(12): 04021233. https://doi.org/10.1061/(ASCE)GM.1943-5622.0002207

58.

Hochman

Dor

Mishael

. “Diverse Effects of Wetting and Drying Cycles on Soil Aggregation: Implications on Pesticide Leaching”. Chemosphere. 2021. 263: 127910. https://doi.org/10.1016/j.chemosphere.2020.127910

59.

Dexter

A.R.

Horn

Kemper

W.D.

. “Two Mechanisms for Age-Hardening of Soil”. J. Soil Sci. 1988. 39(2): 163–175. https://doi.org/10.1111/j.1365-2389.1988.tb01203.x

60.

Kaiser

Kleber

Berhe

A.A.

. “How Air-Drying and Rewetting Modify Soil Organic Matter Characteristics: An Assessment to Improve Data Interpretation and Inference”. Soil Biol. Biochem. 2015. 80: 324–340. https://doi.org/10.1016/j.soilbio.2014.10.018

61.

Gordon

Haygarth

P.M.

Bardgett

R.D.

. “Drying and Rewetting Effects on Soil Microbial Community Composition and Nutrient Leaching”. Soil Biol. Biochem. 2008. 40(2): 302–311. https://doi.org/10.1016/j.soilbio.2007.08.008

62.

Miller

A.E.

Schimel

J.P.

Meixner

Sickman

J.O.

Melack

J.M.

. “Episodic Rewetting Enhances Carbon and Nitrogen Release from Chaparral Soils”. Soil Biol. Biochem. 2005. 37(12): 2195–2204. https://doi.org/10.1016/j.soilbio.2005.03.021

63.

Young

. “A Method of Measuring Aggregate Stability Under Waterdrop Impact”. Trans. ASAE. 1984. 27(5): 1351–1354. https://doi.org/10.13031/2013.32970

64.

Henin

Monnier

Combeau

. “Méthode pour l'étude de la stabilité structurale des sols”. Ann. Agron. 1958. 1: 73–92.

65.

Kemper

W.D.

Chepil

W.S.

. “Size Distribution of Aggregates”. In: Black

C.A.

, editor. Methods of Soil Analysis: Part 1, Physical and Mineralogical Properties, Including Statistics of Measurement and Sampling. Madison, Wisconsin: American Society of Agronomy, 1965. Pp. 499–510.

66.

Hunt

G.R.

. “Visible and Near-Infrared Spectra of Minerals and Rocks: I. Silicate Minerals”. Mod. Geol. 1969. 1: 283–300.

67.

Pieters

C.M.

. “Strength of Mineral Absorption Features in the Transmitted Component of Near-Infrared Reflected Light: First Results from RELAB”. J Geophys. Res.: Solid Earth. 1983. 88(B11): 9534–9544. https://doi.org/10.1029/JB088iB11p09534

68.

Croft

Kuhn

N.J.

Anderson

. “On the Use of Remote Sensing Techniques for Monitoring Spatio-Temporal Soil Organic Carbon Dynamics in Agricultural Systems”. Catena. 2012. 94: 64–74. https://doi.org/10.1016/j.catena.2012.01.001

69.

Powlson

D.S.

. “The Effects of Grinding on Microbial and Non-Microbial Organic Matter in Soil”. J Soil Sci. 1980. 31(1): 77–85. https://doi.org/10.1111/j.1365-2389.1980.tb02066.x

70.

Cierniewski

. “A Model for Soil Surface Roughness Influence on the Spectral Response of Bare Soils in the Visible and Near-Infrared Range”. Remote Sens. Environ. 1987. 23(1): 97–115. https://doi.org/10.1016/0034-4257(87)90073-3

71.

Cierniewski

. “The Influence of the Viewing Geometry of Bare Rough Soil Surfaces on Their Spectral Response in the Visible and Near-Infrared Range”. Remote Sens. Environ. 1989. 27(2): 135–142. https://doi.org/10.1016/0034-4257(89)90013-8

72.

Bishop

J.L.

. “Visible and Near-Infrared Reflectance Spectroscopy”. In: Bishop

J.L.

Bell

Moersch

J.E.

, editors. Remote Compositional Analysis: Techniques for Understanding Spectroscopy, Mineralogy, and Geochemistry of Planetary Surfaces. Cambridge, UK: Cambridge University Press, 2019. Pp. 68–101.

73.

Bishop

J.L.

Davila

Hanley

Roush

. “Dehydration–Rehydration Experiments with Cl Salts Mixed Into Mars Analog Materials and the Effects on Their VNIR Spectral Properties”. Paper presented at: 47th Lunar and Planetary Conference, The Woodlands, Texas. 2016. 1645. https://www.hou.usra.edu/meetings/lpsc2016/pdf/1645.pdf [accessed Jun 21 2024].

74.

McGuirk

S.L.

Cairns

I.H.

. “Relationships Between Soil Moisture and Visible–NIR Soil Reflectance: A Review Presenting New Analyses and Data to Fill the Gaps”. Geotechnics. 2024. 4(1): 78–108. https://doi.org/10.3390/geotechnics4010005