Next Article in Journal
Behavior of Mono Helical Pile Foundation in Clays under Combined Uplift and Lateral Loading Conditions
Next Article in Special Issue
Morphology of Dome- and Tepee-Like Landforms Generated by Expansive Hydration of Weathering Anhydrite: A Case Study at Dingwall, Nova Scotia, Canada
Previous Article in Journal
Zinc(II) Sulfanyltribenzoporphyrazines with Bulky Peripheral Substituents—Synthesis, Photophysical Characterization, and Potential Photocytotoxicity
Previous Article in Special Issue
Novel Vision Transformer–Based Bi-LSTM Model for LU/LC Prediction—Javadi Hills, India
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Applied Sciences
Volume 12
Issue 13
10.3390/app12136826
applsci-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Characterisation of Coastal Sediment Properties from Spectral Reflectance Data

by
Jasper Knight
1,* and
Mohamed A. M. Abd Elbasit
2
1
School of Geography, Archaeology & Environmental Studies, University of the Witwatersrand, Johannesburg 2050, South Africa
2
School of Natural and Applied Sciences, Sol Plaatje University, Kimberley 8301, South Africa
*
Author to whom correspondence should be addressed.
Appl. Sci. 2022, 12(13), 6826; https://doi.org/10.3390/app12136826
Submission received: 9 May 2022 / Revised: 31 May 2022 / Accepted: 3 June 2022 / Published: 5 July 2022
(This article belongs to the Special Issue Geomorphology in the Digital Era)
Browse Figures
Versions Notes

Abstract

:
Remote sensing of coastal sediments for the purpose of automated mapping of their physical properties (grain size, mineralogy and carbonate content) across space has not been widely applied globally or in South Africa. This paper describes a baseline study towards achieving this aim by examining the spectral reflectance signatures of field sediment samples from a beach–dune system at Oyster Bay, Eastern Cape, South Africa. Laboratory measurements of grain size and carbonate content of field samples (n = 134) were compared to laboratory measurements of the spectral signature of these samples using an analytical spectral device (ASD), and the results interrogated using different statistical methods. These results show that the proportion of fine sand, CaCO3 content and the distributional range of sediment grain sizes within a sample (here termed span) are the parameters with greatest statistical significance—and thus greatest potential interpretive value—with respect to their spectral signatures measured by the ASD. These parameters are also statistically associated with specific wavebands in the visible and near infrared, and the shortwave infrared parts of the spectrum. These results show the potential of spectral reflectance data for discriminating elements of grain size properties of coastal sediments, and thus can provide the baseline towards achieving automated spatial mapping of sediment properties across coastal beach–dune environments using hyperspectral remote sensing techniques.

1. Introduction

Sediment properties of sandy beaches and sand dunes, including grain size, carbonate content, moisture content, organic content, magnetic susceptibility and grain mineralogy, are most commonly measured and quantified based on field observations or field sampling, and then laboratory analysis of these samples using different analytical equipment [1,2,3,4]. Following this, a range of statistical techniques (e.g., calculation of moment measures, multivariate analyses) can be used on the grain size data in particular, in order to characterise sediment properties and to interpret depositional processes and environments and their changes over time and space, e.g., [5,6,7,8,9,10,11]. This standard methodology has been undertaken on many beaches and dunes worldwide, resulting in an understanding of spatial patterns of different sediment properties (based mainly on grain size) across different coastal depositional environments, e.g., [12,13,14,15]. The main problem of such a field-based approach is that it provides only a limited view of local-scale coastal sediment properties and dynamics, which is often strongly affected by the specific spatial and temporal context of field sampling at individual sites. In addition, studies also use different sampling strategies and methods of data analysis, which means that results from these individual studies may not be comparable. By contrast, remote sensing methods using a variety of platforms have potential to consistently map and quantify spatial patterns of sediments and landforms across beach–dune systems, and this has been undertaken in several studies e.g., [16,17,18,19,20,21]. There are fewer studies, however, that have examined spectral data on sediment properties and stratigraphy. Sediment cores have been examined using different hyperspectral imaging techniques, mainly in the shortwave infrared (SWIR) wavebands, in order to identify stratigraphic variations in sediment grain size and mineralogy [22,23,24]. These studies have been used to produce spectral time series maps that represent variations in sediment properties through the cores, rather than identify individual spectra that represent certain sediment endmembers. There are only a few studies that have examined the spectral properties of sediments in coastal environments, and these have considered the role of variations in water content and mineralogy as key factors influencing their spectral signatures [18,25]. Mineral compositions can then be used to derive endmembers for spatial modelling.
Most work on spectral signatures of sediment has been done on river depositional environments [26,27], and work on coastal sediments can be informed by these previous studies. For example, river and coastal sediment samples in NE Italy were evaluated by Ciampalini et al. [28] using an analytical spectral device (ASD) in order to derive a spectral library representing sample grain size and mineralogy, which was then compared to laboratory results. Principal component analysis was then used to identify sediment provenance endmembers. The same research approach was used by Ibrahim et al. [19] along the Belgian coast. Grain size properties along beaches were examined using Landsat visible, near infrared (VNIR) and thermal infrared bands in SE India [29], but these bands may have been influenced by a high concentration of heavy minerals (50–80%) at this site. Using IKONOS imagery, Park et al. [30] showed that all spectral bands have a good correlation (>0.8) with grain size, and Williams and Greeley [31] showed that different spectral bands from synthetic aperture radar imagery are affected by surface moisture. Thus, there are several studies that have analysed the spectral properties of beach sand but these have tended to focus on the role of local environmental factors rather than the application of different techniques or methodologies. A key question is how location-specific measurements can be applied to similar depositional settings elsewhere [27,32] or how patterns of (for example) grain size, calcium carbonate (CaCO3), organic carbon or biomass content can be mapped across space using automated remote sensing techniques [21,33,34,35].
Although field and laboratory hyperspectral devices have been used to derive data on coastal sediment properties [16,28,36], there have been hitherto no published studies using the spectral properties of sediments from coastal settings in South Africa. This study uses laboratory hyperspectral measurements of sediment samples collected from a beach–dune system on the coast of South Africa, focusing on relationships between selected properties of the field samples (including grain size and carbonate content) and their associated spectral signatures. The aims of this study are to describe the nature of beach–dune samples in terms of their spectral signatures and to examine these relationships using statistical methods. This can be considered as a first step towards developing a robust methodology for automated mapping of sediment properties across beach–dune environments applicable globally.

2. Study Area and Methods

The study area examined, from which surface sediments were sampled, is at Oyster Bay, Eastern Cape Province, South Africa (Figure 1). Prevailing winds in this region are towards the northeast (in summer) and the west/northwest (in winter). Tidal range is high microtidal/low mesotidal and swell waves from the Southern Ocean have a significant wave height of >5 m [37]. Oyster Bay is an asymmetrical zeta-shape embayment [38] with an extensive sandy beach that is 6.1 km in total length and with a variable beach width of 30 to 290 m at low tide. Bedrock headlands to the east and west define the overall shape of the bay. An extensive supratidal zone is present, containing parallel-aligned transverse dunes with crests that are 40–50 m apart, similar to those found elsewhere along the South African coast [39]. Dune migration periodically blocks off the mouth of the incoming Klipdrift River. The landward boundary of the supratidal dunes at the back of the beach is marked by dune migration into a zone of highly vegetated and variably cemented linear palaeodunes that extend for ~40 km along this coastline. These ridges broadly correspond to the Nahoon Formation of the late Pleistocene Algoa Group, covering the period of marine isotope stages 5 to 2 inclusively [40,41,42]. Holocene-age dunes in this region, fronting the eroded older dunes, correspond to largely unvegetated foredunes of the Schelm Hoek Formation and are composed of unconsolidated calcareous aeolian sand [43].
Surficial (top 5 cm) sediment samples (~400 g each, n = 134, labelled 6–139) were collected in the field across the beach–supratidal dune system in the centre of the Oyster Bay embayment (Figure 1c). A random sampling approach was used but covering the full width of the beach including the intertidal zone. These samples were bagged, labelled, and sampling locations and their geomorphic settings marked using a Garmin Etrex 20 handheld GPS (x y accuracy ± 3 m). In addition, shells of different species (that were broken and did not contain organisms) were also collected from the intertidal zone. Samples were of the Cape brooding oyster (Ostrea atherstoni, sample 1), Brown mussel (Perna perna, sample 2), Agulhas ridged nut clam (Lembulus belcheri, sample 3), Southern cuttlefish (Sepia australis, sample 4), and a mixed shell sample combining these and other shell species found within the intertidal zone (sample 5) (Figure 2). In the laboratory, shell samples (n = 5) were dried and crushed using a pestle and mortar to generate broken fragments >2 mm diameter. Sediment samples were dried, sieved to remove the >2 mm fraction, and a subsample (~50 g) evaluated for CaCO3 content using the loss on ignition method. In this method, the subsample was weighed, combusted in a muffle furnace for 5 h at 950 °C, and reweighed. Combustible CaCO3 content (% of sample mass) was then calculated. Three replicates were undertaken for each sample, and the results averaged. Variation between the replicates was commonly <0.1%. The grain size distribution for each sample was measured using a Mastersizer 3000 Hydro EV for the size range 0.01–2000 μm with a subsample size of ~5 g. Each subsample was sonificated for 20 s prior to measurement, and five individual grain size distribution patterns were measured using the Mastersizer, and the average taken. The key grain size distribution parameters (D10, D50, D90, kurtosis, skewness, standard deviation and mean) generated by the Mastersizer software were used for analysis. Additionally, a derived parameter herein called span, which describes the width of the particle size distribution, was calculated as
S p a n = ( D 90 − D 10 ) D 50
where D90 and D10 are the 90th and 10th percentile values of the grain size distribution, and D50 is the median grain size.
The spectral signatures of sediment and shell samples were acquired under controlled environmental conditions in the laboratory using an Analytical Spectral Device (ASD) FieldSpec®3spectrometer (ASD Inc., Boulder, CO, USA) and a light mug (Figure 3). The instrument measures wavelengths from 350 to 2500 nm and compares samples to a white reference panel. The 2 mm-sieved sediment and shell samples were placed in 100 mL glass bottles (Figure 3) and then placed on top of the light mug to measure the reflectance from each sample. Five spectral scans were captured for each sample to ensure spectral stability and an average reflectance was considered for further analysis. The spectrometer was calibrated using the white reference Spectralon® (Figure 3). The spectrometer was recalibrated after every 20 sample scans. The spectral measurement were stored in a notebook computer connected to the device. Figure 3 shows the laboratory setup used in this study.
Following ASD data collection, the spectral data were first combined and converted from digital numbers to reflectance values using the ViewSpecPro® software. The spectral resolution was 1 nm which causes several data redundancy difficulties and affects the processing time. The spectral data also went through several pre-processing steps including removing noise-affected spectra located at the edge of the scans. First, the reflectance data less than 375 nm and greater than 2460 nm were removed from further analysis in order to disregard edge effects. Second, the moisture absorption spectral bands (at ~2500, 1950 and 1450 nm; [44]) were therefore eliminated from the final pre-processing stage. A correlation analysis was first performed on the data to identify the spectral wavelengths that show a significant association with different sediment properties. Based on the correlation analysis, specific wavelength regions were then selected. Linear regression analysis and partial least squares regression analysis were then performed on the selected spectral wavelengths based on the magnitude of the correlation coefficient. The data were partitioned into training and testing datasets. Approximately 70% (94 samples) of the dataset was used for the statistical model training, and 30% (40 samples) was used for model testing and validation.

3. Results

3.1. Site Geomorphology and Sediment Dynamics

Oyster Bay contains transgressive transverse dunes within the supratidal part of the beach (Figure 1b), and in the field these are observed to be asymmetric in profile and actively migrating towards the northeast, in the direction of the regional prevailing wind [45] and reflecting the relatively high sediment availability in Oyster Bay. The transgressive dunes show steep slipfaces (Figure 4b) and migrating free dunes over the beach surface (Figure 4c). Older vegetated dunes are left as residual eroded hummocks (Figure 4d).

3.2. Sediment Properties

Detailed laboratory analysis of sediment sample grain size data (Table 1) shows that the samples (n = 134) are remarkably uniform. In terms of texture, samples are dominantly (98%) medium sand with only one sample fine grained and two samples coarse grained. In terms of sorting, most samples (66%) are well sorted, 31% are moderately well sorted and 3% moderately sorted. For skewness, 98% are near symmetrical and 2% are coarse skewed. CaCO3 values vary from 8.22% to 27.29%. For kurtosis, >99% are mesokurtic and only one sample is leptokurtic. The sediment samples were collected from backshore, beach, dune crest, ramp, slipface and interdune positions (Figure 1c). There are some statistically significant differences between grain size end-members (fine and coarse/very coarse sand) and CaCO3 values between some of these sampling positions (Table 2), in particular in beach samples where wave action can contribute to sediment sorting and from supratidal dunes where wind transport leads to effective sediment sorting. As a result, there are some statistical differences between D10, D90 and D50 values.
Analysis of the covariation between different sediment properties shows that there are statistically significant relationships between several property types (Table 3). The properties that refer specifically to dimensional values of the grain size distribution (D10, D50 and D90) show evidence for very high correlation coefficients (<0.95) which is indicative of autocorrelation. Dimensionless parameters of skewness and kurtosis show more variable relationships but are also relatively strongly correlated (both positively and negatively) with grain size variables. The nondimensional parameter span broadly expresses the distributional range of particle sizes within the sample (Equation (1)) and thus has a high correlation coefficient with distributional parameters (Table 3). The independent parameter of CaCO3 content shows a strong positive (negative) relationship with coarse (fine) sand because of the mechanical break up of marine shells over time, forming relatively large shell fragments mixed in with coarse mineral sand [46].
Averaged spectral characterisation of sediment samples from different geomorphic positions at Oyster Bay are presented in Figure 5. There are generally similar patterns seen at all positions, consistent with their generally similar sediment grain size compositions (Table 2), with some consistent variability in the water absorption bands. There is greatest variability in particular within the SWIR at ~1850–2400 nm. It is also notable that beach samples show somewhat more variability than samples from other positions, with higher reflectance values (compared to other positions) in the VNIR and lower values in the SWIR (Figure 5a). Based on the high correlation coefficients of beach samples with fine sand and CaCO3 values (Table 2), we therefore speculate that this spectral variability of beach samples reflects the disproportionate influence of fine sand and CaCO3 from shell fragments within these samples. The nature of these samples are now explored in more detail.

3.3. Spectral Analysis of Sediment Samples

The spectral variation at the VNIR and SWIR bands can be examined in detail using the correlation matrices between selected sample particle size characteristics and CaCO3 content. Here, we systematically calculate the correlation coefficient of D10, CaCO3, D90, fine sand and kurtosis at 1 nm wavelength increments through the NVIR and SWIR wavebands (Figure 6). This shows that certain parts of the spectrum are associated with greater (positive or negative) correlation coefficient values and thus are more useful in terms of discriminating between different sediment properties at those wavelengths. For example, in the range ~700–1350 nm there is a clear discrimination between high positive correlations for D10 and D90 and high negative correlation for fine sand (Figure 6). Here, kurtosis and CaCO3 shows no correlation. Likewise, in the range ~1850–2450 nm there is greater statistical discrimination between fine sand (highest positive correlation) and CaCO3 (highest negative correlation) values. By contrast, the region ~1450–1700 nm is not useful for discriminating any sediment properties, because there are very low correlation coefficients throughout (i.e., all correlation coefficients are around zero).
The statistical relationships of CaCO3 values, fine sand and grain size span to different wavelengths are described in Table 4, which shows the outputs of a linear regression model for each variable. The results highlight that certain wavelengths have a statistically significant relationship to some sediment properties. For example, CaCO3 shows greatest significance in the wavelength range ~1052–1252 nm, which falls within the VNIR part of the spectrum. Fine sand has the greatest significance at shorter VNIR wavelengths (~852–952 nm), and span shows significance in isolated parts of the spectrum (2300, 2400 and 2447 nm) at the end of the SWIR range, which may be an artefact of sediment composition within the sample as a whole, e.g., [47]. Water absorption at the 1352 nm waveband has a strong signal and therefore this waveband is removed from the analysis (Table 4) in order to avoid erroneous overfitting.
CaCO3 values can be estimated using single wavelength relationships as:
C a C O 3 = 54.1 − 98.37 R 2335   ( R 2 = 0.718 ,   n = 94 )
where R2335 is reflectance at the 2335 nm wavelength. A similar model performance can be achieved at the wavelength between 2038 and 2435 nm. A multilinear relationship can improve the model estimation as follows:
C a C O 3 = 38.99 + 460.23 R 2200 − 1272.89 R 2300 + 1158.99 R 2335 − 409.99 R 2370 ( R 2 = 0.872 ,   n = 94 )
where R2200, R2300, R2335, R2335, and R2335 are reflectances at the 2200, 2300, 2335 and 2370 nm wavelengths, respectively. Fine sand can be estimated using the following linear relationship:
F   S a n d = 179.06 R 2450 − 41.01   ( R 2 = 0.463 ,   n = 94 )
where F Sand is the fine sand percentage, and R2450 is reflectance at the 2450 nm wavelength. Fine sand values can also be estimated using the reflectance from the wavelengths ranging between 552 and 1789 nm. This can also be used to develop an improved model to estimate the fine sand percentage as follows:
F   S a n d = 21.55 − 364.86 R 1462 + 369.07 R 2082   ( R 2 = 0.721 ,   n = 94 )
where F Sand is the fine sand percentage and R1462 and R2082 are reflectances at the 1462 and 2082 nm wavelengths, respectively. Span had very poor performing models when a single band was utilised. For example, the following model was the best performing single-wavelength model:
S p a n = 2.15 − 2.43 R 2038   ( R 2 = 0.183 ,   n = 94 )
where R2038 is the reflectance at the 2038 nm wavelength. A multilinear model can be developed using highly correlated wavelengths ranging between 852 and 2450 nm, as shown below:
S p a n = 1.68 − 24.66 R 852 + 29.62 R 952 + 59.44 R 2252 − 131.36 R 2300 + 79.47 R 2350 − 73.27 R 2400 + 62.52 R 2447   ( R 2 = 0.516 ,   n = 94 )
where R852, R952, R2252, R2300, R2350, R2400, and R2447 are reflectances at the 852, 952, 2252, 2300, 2350, 2400 and 2447 nm wavelengths, respectively.
Fine sand and span are not well predicted using single wavelength models (see Table 4), whereas CaCO3 shows a much stronger relationship. The models were then validated using 40 independent samples that were not used in the model development (Figure 7). It is notable that a multilinear model leads to a better fit between measured and estimated values.
Cross-validation of the output of the linear regression model (Table 4) through comparison between predicted and measured samples is shown in Figure 8. R2 values, adjusted for the number of variables considered in each model, are higher for CaCO3 with decreasing values for fine sand and span. An increased degree of scatter reflects the inability of the model to describe all of the sample points, and this is particularly the case for span (see Figure 7). Thus, CaCO3 and fine sand values show the most robust statistical relationships to the spectral measurement data.
The spectral characteristics of shell samples 1–5 are presented in Figure 9. Throughout, this shows higher absorption values in the SWIR with consistent dips between all samples at the wavelengths ~1100, 1600 and 2000 nm. The latter may correspond to the water absorption wavelength at ~1950 nm. There is also a very slight jump in reflectance at the water absorption wavelength at ~1450 nm. In addition, the individual shell samples show some variability in the VNIR bands in particular, because of the different shell colours present (Figure 2).
In order to consider whether there are any spectral differences between individual sediment samples with different values of CaCO3, fine sand and span, the samples with the highest and lowest values of these parameters (Table 1) are compared to each other (Figure 10). All these samples reflect the aggregated patterns shown in Figure 5a, in which there are decreases in reflectance in the water absorption bands. The samples also show that, irrespective of individual values of CaCO3, fine sand and span, there are similar reflectance values in the range 1400–1950 nm (see Figure 6). In more detail, comparison of the CaCO3 values within individual samples shows that shorter wavelengths have a higher reflectance where higher CaCO3 values are present, but that the sample with the lowest CaCO3 values has a higher reflectance at longer wavelengths (Figure 10a). This is mirrored by the results for fine sand (Figure 10b), where the signature for the sample with the lowest amount of fine sand (i.e., the coarsest sample) is very similar to the sample with the highest amount of CaCO3. The reason for this is that broken marine shells (as the source of CaCO3 in the sample) give rise to coarse rather than fine sediment [46]. The parameter span, as a reflection of sediment sorting, tends to reflect the presence of coarser outliers in the sample (see the potential autocorrelation with coarse sediment in Table 3) and is therefore of less interpretive significance than either CaCO3 or fine sand (see Figure 8).

4. Discussion

The presence of a wave-dominated shoreface and wide supratidal zone with migrating transverse dunes (Figure 1 and Figure 4) is typical of the south- and southeast-facing South African coast, in which the dunes have a net eastward migration rate of some 3–12 m yr−1 [39,48,49]. Sediment grain size analysis of field samples from Oyster Bay shows that overall they are fairly texturally uniform but with some significant differences in properties between different beach–dune sub-environments (Table 2) and a wide range of carbonate contents (Table 1). The values obtained for grain size properties are similar to other beach–dune systems in the region, e.g., [50,51]. The relatively limited textural and compositional differences mean that it is sometimes difficult to distinguish between such coastal samples, especially in wave-dominated shoreface environments and wind-affected supratidal environments, where sediments are relatively well sorted e.g., [2,4,11,52]. This is certainly the case with beach–dune sediments along the South African coast, e.g., [39,45,49,51,53]. However, detailed statistical analysis shows that different landforms and beach–dune settings at Oyster Bay have different sediment properties (Table 2). This is particularly the case with grain size endmembers (fine and coarse sand) and CaCO3 values that reflect the presence of broken marine shells (not land snail shells) and thus a shoreface source. It is also notable that there are similar overall spectral signatures of all samples (Figure 5), irrespective of their depositional environment (Table 2), which means that the depositional environment of any one sediment sample cannot be resolved by their spectral signature alone. One potential reason for this is that all of the field sediment samples examined here are quartz-dominated (data not shown) and thus, variations in mineralogy cannot be considered as a significant control on their spectral signatures, unlike in previous studies, e.g., [29,47,54]. Although the spectral signature of interstitial water is a dominant feature in other previous studies of coastal sediments e.g., [25,26,32], we deliberately excluded this by drying the samples prior to analysis. This enabled the spectral data of the Oyster Bay samples to be a better representation of grain size and CaCO3 properties (Figure 6), which is the primary aim of this study.
Spectral characteristics of sand systems (beaches, dunes and deserts) have been examined in several studies, e.g., [54,55], and these highlight the potential application of spectral analysis techniques to inform on, in particular, mineralogy and depositional environments [29,32,34,35,36,47]. Similar to this study, the VNIR part of the spectrum has been previously identified as the most useful in terms of sediment discrimination [55]. There are fewer studies, however, that have looked at grain size data and CaCO3 content. In detail, the spectral reflectance of these samples, however, revealed some fundamental differences between CaCO3 content, fine sand proportion and span (Figure 6). These properties also show statistically significant correlations with certain wavelengths (Table 4). However, despite the evidence for some differences in spectral reflectance at different wavelengths for samples with different values of CaCO3 and fine sand (Figure 10), this does not mean that spectral reflectance can be used to predict values of these sediment properties in unknown samples. This is because measured reflectance values at any wavelength are the net result of all grains within the entire sample and not one single component such as shell fragments. In addition, detrital sediment samples of different provenance or found in different depositional environments could have a range of lithologies, water, organic content or other materials, such as microplastics, that may affect spectral reflectance, e.g., [56]. Previous field studies also show the spectral dominance of water absorption signals, e.g., [18,25], and these tend to drown out any signals related to sediment grain size or CaCO3, hence the methodology applied in this study.
These results and their caveats highlight that the potential for spatial mapping of sediment properties across beach–dune environments using hyperspectral imaging techniques may be challenging because of (1) the uncertainties associated with the interpretation of spectral signatures, even under laboratory conditions, and (2) the multiple environmental factors that may be present in a natural beach–dune environment and that may also affect spectral reflectance signatures, including microtopography, vegetation/algae, and salt and water content. Studies on tidal flats also highlight specific problems related to silt and mud particles and chlorophyll content [25,30,35], and these also have to be considered along coastlines that may contain many different types of depositional settings, as well as the spatial transitions between them.

5. Conclusions

This study, based on field samples from a South African beach–dune system, shows both the complexity and potential of hyperspectral techniques to analyse the properties of these samples (with respect to grain size and CaCO3 content), and their limitations. The major conclusions from this study are:
Statistically, CaCO3, fine sand and span are the most important sediment properties in terms of their ability to distinguish between coastal depositional environments (Table 3);
These properties in particular have distinctive spectral signatures in different parts of the VNIR and SWIR wavebands (Table 4);
Fine sand and CaCO3 in particular are clearly distinguishable at ~1850–2450 nm in the SWIR waveband (Figure 6);
Shell content (giving rise to CaCO3 values) and different shell types show somewhat different spectral signatures (Figure 9).
It is notable that previous studies have not described these sediment properties using such analytical techniques and in such a level of detail. The results from this study provide the basis for working towards the automated mapping of a beach–dune environment using hyperspectral satellite data, which must be seen as a long-term goal vital for ongoing monitoring of climate change-sensitive environments.

Author Contributions

Conceptualization, J.K.; Formal analysis, M.A.M.A.E.; Investigation, J.K.; Methodology, J.K.; Software, M.A.M.A.E.; Visualization, M.A.M.A.E.; Writing—original draft, J.K.; Writing—review & editing, M.A.M.A.E. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

Data are available from the corresponding author.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. De Lange, W.P.; Healy, T.R.; Darlan, Y. Reproducibility of sieve and settling tube textural determinations for sand-sized beach sediment. J. Coast. Res. 1997, 13, 73–80. [Google Scholar]
  2. Kasper-Zubillaga, J.J.; Ortiz-Zamora, G.; Dickinson, W.W.; Urrutia-Fucugauchi, J.; Soler-Arechalde, A.M. Textural and compositional controls on modern beach and dune sands, New Zealand. Earth Surf. Proc. Landf. 2007, 32, 366–389. [Google Scholar] [CrossRef]
  3. Kwarteng, A.Y.; Al-Hatrushi, S.M.; Illenberger, W.K.; McLachlan, A. Grain size and mineralogy of Al Batinah beach sediments, Sultanate of Oman. Arab. J. Geosci. 2016, 9, 557. [Google Scholar] [CrossRef]
  4. Hallin, C.; Almström, B.; Larson, M.; Hanson, H. Longshore transport variability of beach face grain size: Implications for dune evolution. J. Coastal Res. 2019, 35, 751–764. [Google Scholar] [CrossRef] [Green Version]
  5. Pyökäri, M. Beach Sediments of Crete: Texture, Composition, Roundness, Source and Transport. J. Coast. Res. 1999, 15, 537–553. [Google Scholar]
  6. Smith, D.A.; Cheung, K.F. Empirical relationships for grain size parameters of calcareous sand on Oahu, Hawaii. J. Coast. Res. 2002, 18, 82–93. [Google Scholar]
  7. Alsharhan, A.S.; El-Sammak, A.A. Grain size analysis and characterization of sedimentary environments of the United Arab Emirates coastal area. J. Coast. Res. 2004, 20, 464–477. [Google Scholar] [CrossRef]
  8. Deidun, A.; Gauci, R.; Schembri, J.A.; Šegina, E.; Gauci, A.; Gianni, J.; Gutierrez, J.A.; Sciberras, A.; Sciberras, J. Comparative median grain size assessment through three different techniques for sandy beach deposits on the Maltese Islands (Central Mediterranean). J. Coast. Res. 2013, 65, 1757–1761. [Google Scholar] [CrossRef]
  9. Poizot, E.; Anfuso, G.; Méar, Y.; Bellido, C. Confirmation of beach accretion by grain-size trend analysis: Camposoto beach, Cádiz, SW Spain. Geo-Mar. Lett. 2013, 33, 263–272. [Google Scholar] [CrossRef]
  10. Choi, T.J.; Choi, J.Y.; Park, J.Y.; Um, H.Y.; Choi, J.H. The Effects of Nourishments Using the Grain-Size Trend Analysis on the Intertidal Zone at a Sandy Macrotidal Beach. J. Coast. Res. 2018, 85, 426–430. [Google Scholar] [CrossRef]
  11. Gunaratna, T.; Suzuki, T.; Yanagishima, S. Cross-shore grain size and sorting patterns for the bed profile variation at a dissipative beach: Hasaki Coast, Japan. Mar. Geol. 2019, 407, 111–120. [Google Scholar] [CrossRef]
  12. Pedreros, R.; Howa, H.L.; Michel, D. Application of grain size trend analysis for the determination of sediment transport pathways in intertidal areas. Mar. Geol. 1996, 135, 35–49. [Google Scholar] [CrossRef] [Green Version]
  13. Edwards, A.C. Grain Size and Sorting in Modern Beach Sands. J. Coast. Res. 2001, 17, 38–52. [Google Scholar]
  14. Gallagher, E.L.; MacMahan, J.; Reniers, A.J.H.M.; Brown, J.; Thornton, E.B. Grain size variability on a rip-channeled beach. Mar. Geol. 2011, 287, 43–53. [Google Scholar] [CrossRef]
  15. Pradhan, U.K.; Sahoo, R.K.; Pradhan, S.; Mohany, P.K.; Mishra, P. Textural Analysis of Coastal Sediments along East Coast of India. J. Geol. Soc. India 2020, 95, 67–74. [Google Scholar] [CrossRef]
  16. Deronde, B.; Kempeneers, P.; Forster, R.M. Imaging spectroscopy as a tool to study sediment characteristics on a tidal sandbank in the Westerschelde. Estuar. Coast. Shelf Sci. 2006, 69, 580–590. [Google Scholar] [CrossRef]
  17. Castillo, E.; Pereda, R.; de Luis, J.M.; Medina, R.; Viguri, J. Sediment grain size estimation using airborne remote sensing, field sampling, and robust statistic. Environ. Monit. Assess. 2011, 181, 431–444. [Google Scholar] [CrossRef]
  18. Manzo, C.; Valentini, E.; Taramelli, A.; Filipponi, F.; Disperati, L. Spectral characterization of coastal sediments using Field Spectral Libraries, Airborne Hyperspectral Images and Topographic LiDAR Data (FHyL). Int. J. Appl. Earth Obs. Geoinf. 2015, 36, 54–68. [Google Scholar] [CrossRef]
  19. Ibrahim, E.; Kim, W.; Crawford, M.; Monbaliu, J. A regression approach to the mapping of bio-physical characteristics of surface sediment using in situ and airborne hyperspectral acquisitions. Ocean. Dyn. 2017, 67, 299–316. [Google Scholar] [CrossRef]
  20. Smit, Y.; Ruessink, G.; Brakenhoff, L.B.; Donker, J.J.A. Measuring spatial and temporal variation in surface moisture on a coastal beach with a near-infrared terrestrial laser scanner. Aeolian Res. 2018, 31, 19–27. [Google Scholar] [CrossRef]
  21. Kim, K.-L.; Kim, B.-J.; Lee, Y.-K.; Ryu, J.-H. Generation of a Large-Scale Surface Sediment Classification Map Using Unmanned Aerial Vehicle (UAV) Data: A Case Study at the Hwang-do Tidal Flat, Korea. Remote Sens. 2019, 11, 29. [Google Scholar] [CrossRef] [Green Version]
  22. Jacq, K.; Giguet-Covex, C.; Sabatier, P.; Perrette, Y.; Fanget, B.; Coquin, D.; Debret, M.; Arnaud, F. High-resolution grain size distribution of sediment core with hyperspectral imaging. Sediment. Geol. 2019, 393–394, 105536. [Google Scholar] [CrossRef]
  23. Ghanbari, H.; Jacques, O.; Adaïmé, M.É.; Gregory-Eaves, I.; Antoniades, D. Remote Sensing of Lake Sediment Core Particle Size Using Hyperspectral Image Analysis. Remote Sens. 2020, 12, 3850. [Google Scholar] [CrossRef]
  24. Zander, P.D.; Wienhues, G.; Grosjean, M. Scanning Hyperspectral Imaging for In Situ Biogeochemical Analysis of Lake Sediment Cores: Review of Recent Developments. J. Imaging 2022, 8, 58. [Google Scholar] [CrossRef]
  25. Ryu, J.-H.; Na, Y.-H.; Won, J.-S.; Doerffer, R. A critical grain size for Landsat ETM+ investigations into intertidal sediments: A case study of the Gomso tidal flats, Korea. Estuar. Coast. Shelf Sci. 2004, 60, 491–502. [Google Scholar] [CrossRef]
  26. Small, C.; Steckler, M.; Seeber, L.; Akhter, S.H.; Goodbred, S., Jr.; Mia, B.; Imam, B. Spectroscopy of sediments in the Ganges–Brahmaputra delta: Spectral effects of moisture, grain size and lithology. Remote Sens. Environ. 2009, 113, 342–361. [Google Scholar] [CrossRef]
  27. Legleiter, C.J.; Stegman, T.K.; Overstreet, B.T. Spectrally based mapping of riverbed composition. Geomorphology 2016, 264, 61–79. [Google Scholar] [CrossRef] [Green Version]
  28. Ciampalini, A.; Consoloni, I.; Salvatici, T.; Di Traglia, F.; Fidolini, F.; Sarti, G.; Moretti, S. Characterization of coastal environment by means of hyper- and multispectral techniques. Appl. Geogr. 2015, 57, 120–132. [Google Scholar] [CrossRef]
  29. Rejith, R.G.; Sundararajan, M.; Gnanappazham, L.; Loveson, V.J. Satellite-based spectral mapping (ASTER and Landsat data) of mineralogical signatures of beach sediments: A precursor insight. Geocarto Int. 2022, 1–24, in press. [Google Scholar] [CrossRef]
  30. Park, N.-W.; Jang, D.-H.; Chi, K.-H. Integration of IKONOS imagery for geostatistical mapping of sediment grain size at Baramarae beach, Korea. Int. J. Remote Sens. 2009, 30, 5703–5724. [Google Scholar] [CrossRef]
  31. Williams, K.K.; Greeley, R. Laboratory and field measurements of the modification of radar backscatter by sand. Remote Sens. Environ. 2004, 89, 29–40. [Google Scholar] [CrossRef]
  32. Van der Wal, D.; Herman, P.M.J. Regression-based synergy of optical, shortwave infrared and microwave remote sensing for monitoring the grain-size of intertidal sediments. Remote Sens. Environ. 2007, 111, 89–106. [Google Scholar] [CrossRef]
  33. Rainey, M.P.; Tyler, A.N.; Gilvear, D.J.; Bryant, R.G.; McDonald, P. Mapping intertidal estuarine sediment grain size distributions through airborne remote sensing. Remote Sens. Environ. 2003, 86, 480–490. [Google Scholar] [CrossRef]
  34. Choi, J.K.; Ryu, J.H.; Eom, J. Integration of spatial variables derived from remotely sensed data for the mapping of the tidal surface sediment distribution. J. Coast. Res. 2011, 64, 1653–1657. [Google Scholar]
  35. Ibrahim, E.; Monbaliu, J. Suitability of spaceborne multispectral data for inter-tidal sediment characterization: A case study. Estuar. Coast. Shelf Sci. 2011, 92, 437–445. [Google Scholar] [CrossRef]
  36. Park, N.-W. Geostatistical integration of field measurements and multi-sensor remote sensing images for spatial prediction of grain size of intertidal surface sediments. J. Coast. Res. 2019, SI90, 190–196. [Google Scholar] [CrossRef]
  37. Wepener, V.; Degger, N. South Africa. In World Seas: An Environmental Evaluation: Volume II: The Indian Ocean to the Pacific, 2nd ed.; Sheppard, C., Ed.; Elsevier: Amsterdam, The Netherlands, 2019; pp. 101–119. [Google Scholar]
  38. Meeuwis, J.; van Rensburg, P.A.J. Logarithmic spiral coastlines: The northern Zululand coastline. S. Afr. Geogr. J. 1986, 68, 18–44. [Google Scholar] [CrossRef]
  39. Knight, J.; Burningham, H. The morphodynamics of transverse dunes on the coast of South Africa. Geo-Mar. Lett. 2021, 41, 47. [Google Scholar] [CrossRef]
  40. Roberts, D.L.; Botha, G.A.; Maud, R.R.; Pether, J. Coastal Cenozoic Deposits. In The Geology of South Africa; Johnson, M.R., Anhaeusser, C.R., Thomas, R.J., Eds.; Geological Society of South Africa/Council for Geoscience: Johannesburg/Pretoria, South Africa, 2006; pp. 605–628. [Google Scholar]
  41. Claassen, D. Geographical controls on sediment accretion of the Cenozoic Algoa Group between Oyster Bay and St. Francis, Eastern Cape coastline, South Africa. S. Afr. J. Geol. 2014, 117, 109–128. [Google Scholar] [CrossRef]
  42. Knight, J. The late Quaternary stratigraphy of coastal dunes and associated deposits in South Africa. S. Afr. J. Geol. 2021, 124, 995–1006. [Google Scholar] [CrossRef]
  43. Butzer, K.W.; Helgren, D.M. Late Cenozoic evolution of the Cape coast between Knysna and Cape St. Francis, South Africa. Quat. Res. 1972, 2, 143–169. [Google Scholar] [CrossRef]
  44. Rossel, R.A.V.; McBratney, A.B. Laboratory evaluation of a proximal sensing technique for simultaneous measurement of soil clay and water content. Geoderma 1998, 85, 19–39. [Google Scholar] [CrossRef]
  45. Schumann, E.H.; Illenberger, W.K.; Goschen, W.S. Surface winds over Algoa Bay, South Africa. S. Afr. J. Sci. 1991, 87, 202–207. [Google Scholar]
  46. Carter, R.W.G. Some problems associated with the analysis and interpretation of mixed carbonate and quartz beach sands, illustrated by examples from north-west Ireland. Sediment. Geol. 1982, 33, 35–56. [Google Scholar] [CrossRef]
  47. Fang, Q.; Hong, H.; Zhao, L.; Kukolich, S.; Yin, K.; Wang, C. Visible and Near-Infrared Reflectance Spectroscopy for Investigating Soil Mineralogy: A Review. J. Spectroscop. 2018, 2018, 3168974. [Google Scholar] [CrossRef]
  48. Lubke, R.A.; Sugden, J. Short-term change in mobile dunes at Port Alfred, South Africa. Environ. Manag. 1990, 14, 209–220. [Google Scholar] [CrossRef]
  49. Knight, J.; Burningham, H. Sand dunes and ventifacts on the coast of South Africa. Aeolian Res. 2019, 37, 44–58. [Google Scholar] [CrossRef]
  50. Harmse, J.T. Trend surface analysis of aeolian sand movement on the southwest African coast. S. Afr. Geogr. J. 1985, 67, 31–39. [Google Scholar] [CrossRef]
  51. Illenberger, W.K. Variations of sediment dynamics in Algoa Bay during the Holocene. S. Afr. J. Sci. 1993, 89, 187–196. [Google Scholar]
  52. Çelikoğlu, Y.; Yüksel, Y.; Kabdaşlı, M.S. Cross-shore sorting on a beach under wave action. J. Coast. Res. 2006, 22, 487–501. [Google Scholar] [CrossRef]
  53. Olivier, M.J.; Garland, G.G. Short-term monitoring of foredune formation on the east coast of South Africa. Earth Surf. Proc. Landf. 2003, 28, 1143–1155. [Google Scholar] [CrossRef]
  54. Bandfield, J.L.; Edgett, K.S.; Christensen, P.R. Spectroscopic study of the Moses Lake dune field, Washington: Determination of compositional distributions and source lithologies. J. Geophys. Res. 2002, 107, 5092. [Google Scholar] [CrossRef] [Green Version]
  55. Sadiq, A.; Howari, F. Remote Sensing and Spectral Characteristics of Desert Sand from Qatar Peninsula, Arabian/Persian Gulf. Remote Sens. 2009, 1, 915–933. [Google Scholar] [CrossRef] [Green Version]
  56. Moshtaghi, M.; Knaeps, E.; Sterckx, S.; Garaba, S.; Meire, D. Spectral reflectance of marine macroplastics in the VNIR and SWIR measured in a controlled environment. Sci. Rep. 2021, 11, 5436. [Google Scholar] [CrossRef] [PubMed]
Applsci 12 06826 g001 550
Figure 1. (a) Location of the study area at Oyster Bay, Eastern Cape, South Africa; (b) large-scale geomorphic setting of Oyster Bay with the sampling region (panel c) shown in the red box; (c) distribution of sediment sampling points 006–139 (background image in (c) from Google Earth, image date 25 August 2013, which is the latest available image before the sampling period).
Figure 1. (a) Location of the study area at Oyster Bay, Eastern Cape, South Africa; (b) large-scale geomorphic setting of Oyster Bay with the sampling region (panel c) shown in the red box; (c) distribution of sediment sampling points 006–139 (background image in (c) from Google Earth, image date 25 August 2013, which is the latest available image before the sampling period).
Applsci 12 06826 g001
Applsci 12 06826 g002 550
Figure 2. Shell samples collected from Oyster Bay. (a) Cape brooding oyster (Ostrea atherstoni, sample 1); (b) Brown mussel (Perna perna, sample 2); (c) Agulhas ridged nut clam (Lembulus belcheri, sample 3); (d) Southern cuttlefish (Sepia australis, sample 4).
Figure 2. Shell samples collected from Oyster Bay. (a) Cape brooding oyster (Ostrea atherstoni, sample 1); (b) Brown mussel (Perna perna, sample 2); (c) Agulhas ridged nut clam (Lembulus belcheri, sample 3); (d) Southern cuttlefish (Sepia australis, sample 4).
Applsci 12 06826 g002
Applsci 12 06826 g003 550
Figure 3. Laboratory spectral measurement system. The system involves an ASD spectroradiometer and a light mug device.
Figure 3. Laboratory spectral measurement system. The system involves an ASD spectroradiometer and a light mug device.
Applsci 12 06826 g003
Applsci 12 06826 g004 550
Figure 4. Dune and beach morphology at Oyster Bay. (a) Dissipative beachface within the lower part of the beach system; (b,c) migrating transverse dune ridges within the supratidal zone; (d) erosional hummock of an older vegetated dune system, now isolated within the upper part of the beach.
Figure 4. Dune and beach morphology at Oyster Bay. (a) Dissipative beachface within the lower part of the beach system; (b,c) migrating transverse dune ridges within the supratidal zone; (d) erosional hummock of an older vegetated dune system, now isolated within the upper part of the beach.
Applsci 12 06826 g004
Applsci 12 06826 g005 550
Figure 5. Results of spectral analysis of samples from different geomorphic positions at Oyster Bay. (a) Full average spectrum for samples from the different position; detailed results at (b) the VNIR (552–1352 nm) and (c) SWIR (1852–2450 nm) parts of the spectrum.
Figure 5. Results of spectral analysis of samples from different geomorphic positions at Oyster Bay. (a) Full average spectrum for samples from the different position; detailed results at (b) the VNIR (552–1352 nm) and (c) SWIR (1852–2450 nm) parts of the spectrum.
Applsci 12 06826 g005
Applsci 12 06826 g006 550
Figure 6. Correlation coefficient between reflectance values at different spectral wavelengths and sediment CaCO3, and particle size characteristics.
Figure 6. Correlation coefficient between reflectance values at different spectral wavelengths and sediment CaCO3, and particle size characteristics.
Applsci 12 06826 g006
Applsci 12 06826 g007 550
Figure 7. Relationship between estimated and measured (a) CaCO3, (b) fine sand %, and (c) span using a linear single wavelength model (model 1) and multilinear model (model 2).
Figure 7. Relationship between estimated and measured (a) CaCO3, (b) fine sand %, and (c) span using a linear single wavelength model (model 1) and multilinear model (model 2).
Applsci 12 06826 g007
Applsci 12 06826 g008 550
Figure 8. Cross validation results of (a) CaCO3 values, (b) fine sand and (c) span. Vertical axes are measured values. Small symbols show cross-validation prediction values.
Figure 8. Cross validation results of (a) CaCO3 values, (b) fine sand and (c) span. Vertical axes are measured values. Small symbols show cross-validation prediction values.
Applsci 12 06826 g008
Applsci 12 06826 g009 550
Figure 9. Spectral reflectance results for shell samples (SH) 1–5.
Figure 9. Spectral reflectance results for shell samples (SH) 1–5.
Applsci 12 06826 g009
Applsci 12 06826 g010 550
Figure 10. Examples of the spectral signature of individual sediment samples with the greatest range of (a) CaCO3 values (highest: #133, lowest: #98); (b) span values (highest: #61, lowest #139); and (c) fine sand percentage (highest #52, lowest #90).
Figure 10. Examples of the spectral signature of individual sediment samples with the greatest range of (a) CaCO3 values (highest: #133, lowest: #98); (b) span values (highest: #61, lowest #139); and (c) fine sand percentage (highest #52, lowest #90).
Applsci 12 06826 g010
Table
Table 1. Details of sediment samples examined in this study. Sample location codes are: BAB: Backshore, beach; BAC: Backshore, crest; BAI: Backshore, interdune; BAR: Backshore, ridge; BAS: Backshore, slipface; BAT: Backshore, trough; BEC: Beach, crest; BEI: Beach, interdune; BER: Beach, ridge; BES: Beach, slipface; BET: Beach, trough; CRI: Crest, interdune; CRR: Crest, ridge; CRS: Crest, slipface; CRT: Crest, trough; INT: Interdune, ridge; INS: Interdune, slipface; INT: Interdune, trough; RAS: Ramp, slipface; RAT: Ramp, trough; SLT: Slipface, trough. Sample numbers are indicated in Figure 1c.
Table 1. Details of sediment samples examined in this study. Sample location codes are: BAB: Backshore, beach; BAC: Backshore, crest; BAI: Backshore, interdune; BAR: Backshore, ridge; BAS: Backshore, slipface; BAT: Backshore, trough; BEC: Beach, crest; BEI: Beach, interdune; BER: Beach, ridge; BES: Beach, slipface; BET: Beach, trough; CRI: Crest, interdune; CRR: Crest, ridge; CRS: Crest, slipface; CRT: Crest, trough; INT: Interdune, ridge; INS: Interdune, slipface; INT: Interdune, trough; RAS: Ramp, slipface; RAT: Ramp, trough; SLT: Slipface, trough. Sample numbers are indicated in Figure 1c.
Sample #Location CodeLatLongElevation (m asl)D10 (Microns)D50 (Microns)D90 (Microns)KurtosisSkewnessCaCO3 (%)Fine sand (%)Medium Sand (%)Coarse Sand (%)Very Coarse Sand (%)
006BEC−34.17193824.6429234.2112093154780.950.0018.8324.3468.067.590.00
007BET−34.17198424.6426682.0482203405280.95−0.0113.7818.5768.0713.350.00
008SLT−34.17205224.6424894.9321992894210.970.0010.4031.4765.922.610.00
009BES−34.17201624.6423898.7773024888040.95−0.0221.403.1649.3644.732.75
010BET−34.17201124.6421287.8162263505470.96−0.0113.6916.5068.1215.380.00
011BET−34.17167924.6418795.1722013235380.98−0.0416.2025.1061.4111.890.88
012BEC−34.17151724.64227611.6612043074650.940.0017.0426.5766.996.440.00
013BEI−34.17170224.64236311.6611982884200.970.009.8931.7965.612.600.00
014CRS−34.17169024.6424769.7392794427150.96−0.0221.985.3057.3636.331.01
015SLT−34.17166624.6425768.2962033034530.95−0.0112.3027.3967.185.430.00
016BET−34.17193324.6433846.1342203405290.95−0.0114.5818.6967.8913.420.00
017BEC−34.17213724.6439029.7392543926090.96−0.0119.099.0166.6324.290.06
018BEC−34.17192924.6438057.8162073114700.95−0.0116.4425.2867.906.820.00
019CRS−34.17183324.64402912.3822483825890.95−0.0116.6010.4567.6921.830.03
020BET−34.17137424.6438915.8931892874400.96−0.0112.6634.0961.504.410.00
021BEC−34.17112024.6436439.0172143214850.960.0019.3522.2869.538.190.00
022INR−34.17136324.6436497.0952283746310.96−0.0318.7615.1560.8823.400.50
023INR−34.17149924.6432899.4982173515830.97−0.0416.9718.8562.4117.850.56
024RAS−34.17160024.6429899.9792183435430.96−0.0116.9019.0666.1414.790.00
025RAS−34.17186924.6430106.1342043185030.95−0.0212.9325.1064.5410.360.00
026BAB−34.17220324.6428618.5371993024590.95−0.0116.2128.5365.406.060.00
027BAT−34.17195124.6420911.0872193395280.95−0.0116.7819.1967.5013.300.00
028BAT−34.17198024.6416966.1342263726310.97−0.0321.1315.6460.7522.840.70
029BAC−34.17196124.6414417.0952113144720.950.0014.7323.8669.236.910.00
030BAT−34.17194224.6412645.8932003114900.96−0.0213.6126.9264.198.890.00
031BEC−34.17195524.6411515.6532213445300.950.0119.2218.1468.1913.670.00
032BES−34.17156524.6410336.1342614166690.95−0.0120.087.8860.6430.940.54
033BEC−34.17127724.6408093.2501872744020.950.0022.4138.3059.871.830.00
034BAT−34.17112524.6408658.2962023074660.950.0017.5827.0566.486.470.00
035RAS−34.17103824.6408417.8162193365140.950.0012.1319.3568.7111.940.00
036BAC−34.17112324.6409995.6532103164810.960.0020.1623.9268.257.830.00
037BAT−34.17119324.6412810.12525446610201.06−0.1326.589.3146.0634.298.86
038BAC−34.17122324.6417884.6921912844250.960.0019.2334.6062.353.040.00
039BAR−34.17137624.6416566.3741802643880.96−0.0112.3842.8155.961.220.00
040BAT−34.17162124.6417684.4512133275080.95−0.0114.7221.9767.0411.000.00
041BAC−34.17171324.6416327.3352393665650.95−0.0118.4413.1268.7118.160.00
042BAB−34.17200424.6399711.8082253635890.95−0.0114.9216.2663.4520.200.09
043BAC−34.17177524.6399736.8542353404880.950.0218.4214.6377.038.340.00
044BAC−34.17164824.6398909.2582193284940.96−0.0115.8419.7871.019.200.00
045BAS−34.17138824.6399738.2962533936110.96−0.0119.069.3066.0924.550.06
046BAC−34.17102024.63977310.7002133224900.960.0019.0022.2568.968.790.00
047BAS−34.17124724.6400235.4132033104810.96−0.0210.2726.4065.667.940.00
048RAT−34.17133024.6402026.3742063184950.95−0.0115.9624.5566.099.360.00
049RAS−34.17175624.6402006.1342393916710.97−0.0517.2012.7759.3126.551.37
050BAC−34.17188724.6407095.6532223505560.96−0.0117.0617.6066.0716.320.01
051BAC−34.17160224.6404813.7302173596250.99−0.0620.6318.4059.7420.101.31
052RAT−34.17123224.6405643.9711762483470.96−0.0115.2751.2148.620.170.00
053RAT−34.17126324.6406427.8161922743910.960.0016.6137.2761.611.120.00
054INT−34.17111724.6405366.3742013004480.960.0015.5728.3366.844.840.00
055BAT−34.17147324.6412315.8932073194980.95−0.0114.8224.1966.069.750.00
056BAT−34.17181424.6412657.3352053265300.95−0.0314.8923.9262.9813.000.07
057BAB−34.17215724.6418783.7301972894270.970.0013.4831.6565.293.060.00
058BAB−34.17193024.6390006.3742263505490.96−0.0115.0216.6167.7115.670.00
059CRI−34.17167924.6390457.5762073064540.95−0.0117.6726.1468.325.540.00
060BAC−34.17124824.6390708.5372143445680.96−0.0320.1220.3462.4917.060.11
061BAC−34.17112424.6392306.37424344312401.09−0.2524.6511.3246.8727.5811.11
062BAT−34.17124224.6393140.6061952914380.96−0.019.9731.8164.084.110.00
063MAI−34.17138124.6395565.8932193455510.96−0.0217.3218.7165.6415.630.01
064RAT−34.17168524.6393764.4512173546201.00−0.0719.6318.9060.1418.941.53
065BAT−34.17178624.6396847.8162163295040.960.0016.7620.7068.8310.470.00
066INT−34.17146524.6397313.7302143516160.99−0.0715.9219.7859.6918.781.41
067INR−34.17099624.63943413.1032083245100.95−0.0118.7723.5265.2111.270.00
068INS−34.17115424.6404068.7772193274880.960.0018.7519.9571.528.530.00
069CRI−34.17159424.629845−5.6412384067220.98−0.0423.6712.5855.5629.392.34
070CRI−34.17114724.63039212.8632463966390.960.0018.0511.0261.8826.900.21
071CRI−34.17061624.63082712.6222523885940.960.0019.256.4567.5922.950.01
072CRS−34.17029024.63110110.7002623925830.950.0019.477.5370.2022.270.01
073BAR−34.17139424.6316638.2961953024740.96−0.0218.6729.596.977.430.00
074BER−34.17243824.631051−0.3542233344980.960.0014.1118.1572.149.720.00
075BAR−34.17093824.6359536.1342203585810.950.0022.1017.6562.8519.470.03
076BAT−34.17161624.63574310.4591962874230.970.0011.4132.6064.592.810.00
077BAC−34.17193524.6357844.6922153315120.95−0.0117.5420.8567.5911.560.00
078BER−34.17250024.6365025.6532193375210.95−0.0114.7119.2368.1212.650.00
079BAT−34.17183424.6369198.5372413615450.96−0.0113.3412.6871.6415.680.00
080BAR−34.17111024.6370488.7772133435620.96−0.0220.4720.4962.9216.540.05
081CRS−34.17084724.63818913.5842534328080.99−0.0827.259.3853.0733.004.49
082BAB−34.17169024.6381494.9322343545300.940.0115.9214.3571.7313.920.00
083BAB−34.17233624.6384287.3352394298650.99−0.1025.3312.2849.4331.896.09
084BAI−34.17159524.6385242.0481972944420.96−0.0120.4530.7164.884.410.00
085INR−34.17102924.63880710.4592363715840.94−0.0120.3413.4965.9420.540.04
086CRI−34.17073324.63872213.5841902753960.960.0015.5037.3461.221.440.00
087BET−34.17248524.6401665.4131962814000.960.0012.2934.0964.391.510.00
088BET−34.17254124.6413055.6532343856420.95−0.0119.4813.6060.3025.740.36
089BET−34.17257024.6423991.5672754366990.95−0.0119.735.9658.2335.160.65
090BET−34.17285124.6437805.8933365178010.96−0.0124.001.0245.1651.742.09
091BAB−34.17252024.6442984.4512042964300.960.0012.5428.4868.453.070.00
092BAS−34.17200224.6442816.1341932723850.960.008.2637.8161.320.870.00
093BAS−34.17198124.6443187.3352002974430.96−0.018.8629.3066.254.450.00
094BAC−34.17137524.6444959.7392564036430.96−0.0117.998.8263.0627.870.25
095BAR−34.17076124.6444258.2962243385110.96−0.0117.3317.6470.8011.570.00
096BAT−34.17118824.6445816.6142153385530.97−0.0413.6720.5364.1414.330.54
097BAI−34.17182824.6449377.8162133305180.95−0.0113.8621.6566.1412.210.00
098BAT−34.17226124.6448225.4132012894160.960.008.2230.9966.752.260.00
099BAB−34.17260824.6452146.1342173295040.96−0.0114.4820.4069.1510.460.00
100BER−34.17289924.6454624.4512493755650.960.0013.0210.2371.1018.670.00
101BAC−34.17224424.6454328.0562353535350.95−0.0118.5114.3771.2314.400.00
102BAC−34.17191924.6455699.2582253495410.960.0021.1516.7868.3514.880.00
103BAS−34.17144724.64530510.7002944797960.96−0.0222.233.7550.5942.962.71
104BAT−34.17130624.6454697.5762273555640.96−0.0214.5616.2266.4317.320.02
105RAS-34.29699724.64551710.7002193405320.95−0.0118.1018.9567.4013.650.00
106CRS−34.17092724.64592610.45925044410301.11−0.1924.179.9348.9730.528.51
107SLT−34.17093924.6459579.4982143144600.950.0010.6022.9471.265.800.00
108CRR−34.17089024.64628210.4592293615710.960.0017.5815.2466.1718.570.02
109CRR−34.17111024.64657111.9012183515730.96−0.0120.5318.7063.1918.050.05
110CRT−34.17085824.6467117.3351952934420.960.0012.4331.1564.414.410.00
111CRS−34.17082024.6469279.2582503926250.97−0.0220.5110.0464.5425.210.21
112CRT−34.17081924.6470344.4512093225030.95−0.0111.8923.5766.0510.370.00
113CRR−34.17135024.64720913.1032133164690.950.0016.2222.9870.416.610.00
114RAS−34.17190824.64745219.8321792653910.950.0012.6742.7755.891.340.00
115CRR−34.17172424.64727321.9951902723880.960.0016.9538.5760.421.000.00
116BAC−34.17178224.64693520.3132043084670.95−0.0120.5126.5766.856.580.00
117BAS−34.17191024.64659912.6222002954370.960.0012.4229.6666.463.880.00
118BAC−34.17193024.64654417.9102203274890.960.0018.0319.7471.698.570.00
119BAC−34.17209524.64643712.6222443725710.96−0.0119.9511.5069.2419.250.01
120CRS−34.17134224.6482714.4511972803930.960.008.9934.2564.661.090.00
121CRS−34.17168224.64852111.4212023094760.95−0.0113.0926.6865.857.470.00
122CRS−34.17169124.64833413.8242032884070.960.0016.3630.6967.611.700.00
123CRS−34.17121824.64804813.1032664226780.94−0.0119.957.1960.0432.250.53
124CRS−34.17178124.64796922.7161892643720.960.0119.5741.9057.630.470.00
125CRS−34.17203224.64772116.9481882633710.960.0112.0742.4057.150.450.00
126INR−34.17232024.64758712.8632012954320.960.0011.0129.4967.173.340.00
127CRS−34.17230924.64735816.4683014667330.96−0.0220.232.8354.9841.191.09
128BAT−34.17238024.6471119.9792143305120.95−0.0113.4621.3467.1111.560.00
129BAC−34.17229324.64694812.3822133134580.950.0017.7423.5070.805.700.00
130BAT−34.17228624.64659110.9402173254880.960.0014.4120.8370.668.520.00
131BAC−34.17236724.64638410.4592123174780.960.0014.0523.1769.297.540.00
132BAB−34.17259524.6462534.9322203304970.960.0012.5319.3471.009.660.00
133BAB−34.17295524.6461326.8542855009410.96−0.0627.295.1544.9041.997.86
134BAB−34.17267924.6469234.6922083124710.95−0.0114.7125.0068.106.900.00
135BER−34.17306624.6474866.6142503735580.960.0014.1710.1072.0417.870.00
136BAB−34.17278724.6479205.1722163164630.960.0013.7022.1771.796.040.00
137BAT−34.17266724.6484873.4902213314940.960.0014.5018.9471.839.240.00
138CRS−34.17243124.6489886.6142293334850.960.0016.8616.6975.258.060.00
139CRT−34.17216024.6490224.4512102934080.970.0113.1627.6670.871.480.00
Table
Table 2. Table of p-values obtained for different sediment properties from different sampling locations (Table 1), analysed using Fisher’s partial least squares discriminant analysis. Significance levels are: ^ 0.1, * 0.05, ** 0.01 and *** 0.001.
Table 2. Table of p-values obtained for different sediment properties from different sampling locations (Table 1), analysed using Fisher’s partial least squares discriminant analysis. Significance levels are: ^ 0.1, * 0.05, ** 0.01 and *** 0.001.
LocationD10D50D90CaCO3Fine SandMedium SandCoarse SandVery Coarse Sand
Backshore, beach0.0002 ***0.0004 ***0.0041 **0.0455 *0.0120 *0.23130.0003 **0.0148 *
Backshore, crest0.0265 *0.0318 *0.0475 *0.0005 **0.14330.91290.0221 *0.1460
Backshore, interdune0.8668 ^0.58230.51540.95090.72690.81880.39360.8481
Backshore, ramp0.19030.35540.63080.95660.0472 *0.53590.71080.9965
Backshore, slip face0.42710.40770.54170.0813 ^0.26340.73300.42440.9953
Backshore, trough0.45340.82410.72960.0782 ^0.36780.88720.93390.4140
Beach, crest0.0113 *0.0144 *0.0817 ^0.55470.0814 ^0.19140.0192 *0.0967 ^
Beach, interdune0.0038 **0.0141 *0.0757 ^0.11130.0803 ^0.23430.0283 *0.0674 ^
Beach, ramp<0.0001 ***<0.0001 ***0.0022 *0.0705 ^<0.0001 ***0.58110.0003 **0.0241 *
Beach, slip face0.0069 **0.0076 **0.0365 *0.0067 **0.0172 *0.34040.0081 *0.1989
Beach, trough<0.0001 *<0.0001 ***0.0032 **0.0002 ***0.0004 **0.1425<0.0001 ***0.0379 *
Crest, interdune0.16260.35880.49350.0148 *0.51580.73720.49680.4123
Crest, ramp0.0007 ***0.0036 **0.0235 *0.0031 **0.0004 **0.52450.0163 *0.2081
Crest, slip face0.0879 ^0.0868 ^0.15490.0015 **0.0819 ^0.69030.0820 ^0.5586
Crest, trough0.0001 ***0.0019 **0.0328 *<0.0001 ***0.0018 **0.73330.0039 *0.4294
Interdune, ramp0.22320.20700.31840.98910.0532 ^0.47400.27200.8547
Interdune, slip face0.40230.27830.34840.0957 ^0.21410.84960.21970.9095
Interdune, trough0.46580.44450.64380.17650.29970.88220.38130.7004
Ramp, slip face0.91540.73600.72570.0839 ^0.94050.51690.55450.9972
Ramp, trough0.40200.39120.39700.10980.13860.40830.62160.4729
Slip face, trough0.60990.44540.43970.28220.43340.76910.39040.7306
Table
Table 3. Pearson Product Moment Correlation of different sediment properties across all samples (n = 134). F = fine, M = medium, C = coarse and VC = very coarse. Variable span is defined in the text. Significance levels are: ^ 0.1, * 0.05, ** 0.01 and *** 0.001.
Table 3. Pearson Product Moment Correlation of different sediment properties across all samples (n = 134). F = fine, M = medium, C = coarse and VC = very coarse. Variable span is defined in the text. Significance levels are: ^ 0.1, * 0.05, ** 0.01 and *** 0.001.
D10D50D90KurtosisSkewnessCaCO3F SandM SandC SandVC Sand
D101.000
D500.951 ***1.000
D900.754 ***0.903 ***1.000 ***
Kurtosis0.1390.318 ***0.634 ***1.000
Skewness−0.252 **−0.477 ***−0.793 ***−0.890 ***1.000
CaCO30.571 ***0.679 ***0.698 ***0.337 ***−0.463 ***1.000
F sand−0.914 ***−0.918 ***−0.761 ***−0.156 ^0.300 ***−0.557 ***1.000
M sand−0.201 *−0.332 ***−0.451 ***−0.383 ***0.478 ***−0.411 ***0.0301.000
C sand0.928 ***0.984 ***0.872 ***0.266 **−0.443 ***0.679 ***−0.880 ***−0.393 ***1.000
VC sand0.414 ***0.603 ***0.853 ***0.813 ***−0.895 ***0.566 ***−0.367 ***−0.529 ***0.545 ***1.000
Span0.403 ***0.642 ***0.902 ***0.782 ***−0.941 ***0.593 ***−0.493 ***−0.469 ***0.616 ***0.885 ***
Table
Table 4. Analysis of p-values of linear regression of CaCO3, fine sand and grain size span with selected wavelengths (see Figure 5). Significance levels are: ^ 0.1, * 0.05, ** 0.01 and *** 0.001.
Table 4. Analysis of p-values of linear regression of CaCO3, fine sand and grain size span with selected wavelengths (see Figure 5). Significance levels are: ^ 0.1, * 0.05, ** 0.01 and *** 0.001.
Wavelength (nm)CaCO3Fine SandSpan
5520.0928 ^0.62640.9885
6520.25070.98830.4891
7520.72110.0928 ^0.0734 ^
8520.73430.0096 **0.1257
9520.94100.0026 **0.2164
10520.0077 **0.79440.3640
11520.0100 *0.94150.2201
12520.0087 **0.49230.1494
14620.0834 ^0.21630.3150
15520.31620.82460.2073
16520.26050.98010.1364
17520.15250.89570.1300
17890.10640.0480 *0.0158 *
19620.0264 *0.79770.0528 ^
20280.44130.40950.2356
20820.96180.52960.6334
21520.42120.42930.7909
22000.58970.56750.6655
22520.30460.0789 ^0.3075
23000.0098 **0.47950.0085 **
23350.10320.86120.8394
23380.15300.80690.7031
23500.71370.0696 ^0.4356
23520.24430.0253 *0.9282
23700.32900.25000.7780
24000.35960.31500.0041 **
24200.32370.18920.2231
24350.28550.0127 *0.7258
24470.65000.0760 ^0.0024 **
24500.22810.0470 *0.5175
Adjusted R20.89780.75100.5510
Overall p-value2.2 × 10−16 ***2 × 10−16 ***4.88 × 10−16 ***
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

Knight, J.; Abd Elbasit, M.A.M. Characterisation of Coastal Sediment Properties from Spectral Reflectance Data. Appl. Sci. 2022, 12, 6826. https://doi.org/10.3390/app12136826

AMA Style

Knight J, Abd Elbasit MAM. Characterisation of Coastal Sediment Properties from Spectral Reflectance Data. Applied Sciences. 2022; 12(13):6826. https://doi.org/10.3390/app12136826

Chicago/Turabian Style

Knight, Jasper, and Mohamed A. M. Abd Elbasit. 2022. "Characterisation of Coastal Sediment Properties from Spectral Reflectance Data" Applied Sciences 12, no. 13: 6826. https://doi.org/10.3390/app12136826

APA Style

Knight, J., & Abd Elbasit, M. A. M. (2022). Characterisation of Coastal Sediment Properties from Spectral Reflectance Data. Applied Sciences, 12(13), 6826. https://doi.org/10.3390/app12136826

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop