Published granulometric papers AND granulometric database proposal

In the frame of our research projects NKFIH K120620 'Paleoenvironmental reconstruction based on particle size and shape of aeolian dust deposits' and NKFIH KH130377 'Granulometric analysis of recent Saharan dust' we've published our new granulometric milestone papers! Critical questions of laser diffraction and automated static image analysis measurements, and problematic interpretations of grain size data were discussed in the following papers:

• Varga, Gy., Gresina, F., Újvári, G., Kovács, J., Szalai, Z. (2019). On the reliability and comparability of laser diffraction grain size measurements of paleosols in loess records. Sedimentary Geology 389, pp. 42-53.
• Varga, Gy., Kovács, J., Szalai, Z., Cserháti, Cs., Újvári, G., (2018). Granulometric characterization of paleosols in loess series by automated static image analysis. Sedimentary Geology 370, pp. 1-14.
• Varga, Gy., Újvári, G., Kovács, J. (2019). Interpretation of sedimentary (sub)populations extracted from grain size distributions of Central European loess-paleosol series. Quaternary International 502, Part A, pp. 60-70

On the reliability and comparability of laser diffraction grain size measurements of paleosols in loess records

Laser diffraction grain size data and size distributions of paleosols are widely used in paleoenvironmental reconstructions as physicochemical alteration-related proxies of past changes of the sedimentary environment. Different laser diffraction devices, optical theories, and optical settings are being applied nowadays, and ignorance of several uncertainty factors and drawbacks of this indirect grain size characterization approach has led to poorly comparable and reproducible granulometric datasets.

Here we present a detailed comparison of grain size results acquired by three state-of-the-art, widely used laser diffraction devices (Fritsch Analysette 22 Microtec Plus, Horiba Partica La-950 v2 and Malvern Mastersizer 3000). Grain size distributions were calculated using both the Fraunhofer and Mie scattering theories and a wide variety of optical settings (68 different complex refractive indices: 1.45–1.6–0.01i-1i) for 10 samples from loess-paleosol sequences of the Carpathian Basin.

Our findings demonstrate that optical settings have significant effects on grain size distributions, especially for the finest fractions (clay and fine silt populations). Interestingly, the selection of laser diffraction devices was an even more deterministic factor in grain size characterization as revealed by network analyses. Clustering of bulk and size-fractionated grain size records was primarily determined by the device used, and also by the applied optical settings. At the same time, the real physical differences of samples were found to be deterministic exclusively for the sand-sized fraction. As such, these findings emphasize the importance of the accurate description of applied methodological details in research papers, as comparability and reproducibility of granulometric datasets cannot be ensured with a lack of information about instrumental conditions and settings.

Granulometric characterization of paleosols in loess series by automated static image analysis

An automated image analysis method is proposed here to study the size and shape of siliciclastic sedimentary particles of paleosols of Central European loess sequences. Several direct and indirect measurement techniques are available for grain size measurements of sedimentary mineral particles. Indirect techniques involve the use of some kind of physical laws, however, all requirements for calculations are in many cases not known. Even so, the direct manual microscopic observation and measurement of large, representative number of grains is time-consuming and sometimes rather subjective. Therefore, automated image analyses techniques provide a new and perspective way to analyse grain size and shape sedimentary particles. Here we test these indirect (laser diffraction) and direct (automated static image analysis) techniques and provide new granulometric (size and shape) data of paleosols. Our results demonstrate that grain size data of the mineral dust samples are strongly dependent on shape parameters of particles, and shape heterogeneity was different between different size classes. Due to the irregular grain shape parameters, uncertainties have arisen also for determination of grain sizes. In this paper we present a possible correction procedure to reduce the differences among the results of the laser diffraction and image analysis methods. By applying new correction factors, results of the two approaches could become closer but the unknown thickness of particles remains a problem to solve.

The other presented correction procedure to assess the uncertain 3rd dimension of particles by their intensity-size relationships makes us able to reduce further the deviations of the two sizing methods.

Interpretation of sedimentary (sub)populations extracted from grain size distributions of Central European loess-paleosol series

Grain size proxies of aeolian dust deposits have widely been applied in environmental and sedimentary studies. However, large body of research papers are not taking into consideration that a complex grain size distribution curve cannot be an indicator of a single one environmental factor (e.g. wind speed/strength, transportation distance, aridity).

The aim of the present paper is to discuss the main differences of frequently used statistical methods and to provide possible interpretations of the results by applying these various approaches on the high-resolution loess-paleosol profile of Dunaszekcső, South Hungary (Central Europe). Beside single statistical descriptors (mean, median, mode) of grain size and simple indices of size-fraction ratios (U-ratio, Grain Size Index), some more complex algorithms were also used in our paper. The applied parametric curve-fitting, end-member modelling and hierarchical cluster analysis techniques are using the whole spectrum of the measured grain size distributions and provide a more reliable and more representative results even in case of small scale variations.

According to our findings, approaches which provide direct linkage among simple statistical descriptors and single atmospheric or other environmental elements are rather oversimplified as properties aeolian dust deposits are influenced by the integrated effects of several concurrent processes. Differences of more complex decomposition methods arise from the different approach and scope. End-members are determined from the unmixing based on the covariance structure of the whole grain size data-series of the section, while the parametric curve-fitting is based on the one-by-one deconvolution of the grain size distribution curves. End-members of loess-paleosol samples are regarded as representation of the average dust grain size distribution of various temporal sediment clusters of seasonal or other short-term intervals, while (sub)populations by parametric curve-fitting are proposed to illustrate process-related elements of background and dust storm depositional components for each sample. Results of cluster analysis represent similar grouping conditions as end-member modelling with a reduced sedimentary and genetically meaning.

Identification of Saharan dust particles in Pleistocene dune sand- paleosol sequences of Fuerteventura (Canary Islands)

New paper: 

Varga, Gy, Roettig, C.-B. (2018). Identification of Saharan dust particles in Pleistocene dune sand-paleosol sequences of Fuerteventura (Canary Islands). Hungarian Geographical Bulletin 67 (2), pp. 121-141.

DOI: https://doi.org/10.15201/hungeobull.67.2.2

Keywords: Saharan dust, Canary Islands, grain size, grain shape, automated image analysis

Abstract

Automated static image analysis and newly introduced evaluation techniques were applied in this paper to identify Saharan dust material in the unique sand-paleosol sequence of Fuerteventura (Canary Islands). Measurements of ~50,000 individual mineral particles per samples provided huge amount of granulometric data on the investigated sedimentary units. In contrast to simple grain size and shape parameters of bulk samples, (1) parametric curve-fitting allowed the separation of different sedimentary populations suggesting the presence of more than one key depositional mechanisms. Additional (2) Raman-spectroscopy of manually targeted individual particles revealed a general relationship among grain size, grayscale intensity and mineralogy. This observation was used to introduce the (3) intensity based assessment technique for identification of large number of quartz particles. The (4) cluster and (5) network analyses showed that only joint analysis of size, shape and grayscale intensity properties provided suitable results, there is no specific granulometric parameter to distinguish Saharan dust due to their irregular shape characteristics. The presented methods allowed the separation of Saharan dust-related quartz grains from local sedimentary deposits, but due to the lack of robust granulometric characterization of coarsest fractions and due to the diverse geochemical properties of North African sources, exact volumetric amount of deposited dust material and sedimentation rates could not be determined from these data.

Methodology

Samples were taken from 24 silty units considered as paleo-surfaces of stable geomorphic periods with reduced sand movements and relatively enhanced Saharan dust influence, additional dune sand and sand sheet samples were also investigated as references for intense sand transportation intervals. Detailed description of the units and stratigraphic analysis of selected sites can be found in the works of Faust, D. et al., (2015) and Roettig, C.-B. et al., (2017; under review). Air-dried and 2 mm sieved samples were measured by Malvern Morphologi G3-ID instrument in the Laboratory for Sediment and Soil Analysis (Geographical Institute, Research Centre for Astronomy and Earth Sciences, Hungarian Academy of Sciences).

The applied automated static image analysis technique is a new, innovative mode of grain size and shape analyses completed with chemical identity assessments of Raman spectrometry. In contrast to widely used laser diffraction measurements, image analysis provides direct observational data of particle size, and due to the automatic measurement technique large number of particles are characterized allowing us a more robust and objective granulometric description of particles compared to manual microscopic approaches.

7 mm³ of mineral particles per samples were dispersed by 4 bar compressed air onto a glass slide with 60 s settling time. The used 20× objective lens provide a 960× magnification, suitable for detailed characterization of particles in the size range between fine silt and fine sand fractions. Two-dimensional imaging was completed with the usage of additional vertical focal planes, two additional layers were applied above and two other ones below the focus, equivalent to a total of 27.5 µm.

The captured high-resolution grayscale images of ~50,000 individual mineral particles were automatically analysed by the device software to get a raw granulometric data-matrix. Each row of the table represents one sedimentary particle (with its own identity number), while the columns are various size and shape parameters, completed with light transmissivity data and Raman correlation scores.

Circle-equivalent (CE) diameter is the key size descriptor, calculated as the diameter of a circle with the same area as the projected two-dimensional image of a given mineral grain. Beside several other simple size properties (e.g. length, width, perimeter, sphere-equivalent volume), various shape parameters are derived from these sizes. Aspect ratio is the ratio of width and length, circularity describes the proportional relationship between circumference of a circle equal to the projected area of the particle and perimeter. Convexity (and solidity) parameters are measures of edge roughness by using the ratio of particle and convex hull perimeter (and area). Circularity and convexity values are also suitable to filter out stacked particles and aggregated particles, in this study particles with lower than 0.65 circularity and convexity values were excluded from further calculations.

Intensity mean and standard deviation parameters are determined from the grayscale images as a results of light transmissivity of particles. These values are dependent on mineralogy, particle thickness, chemical homogeneity and surface roughness (for detailed description of the method, see: Varga, Gy. et al., 2018). Intensity values together with chemical identity analyses of the build-in Raman spectrometer provide useful additional information for separation of granulometrically similar particles.

Identification of Saharan dust material

Based on the fact that the Saharan dust deposited at Fuerteventura is mainly (1) silt-sized and (2) contains a lot of quartz particles (regarded as exotic in the basaltic and carbonate-rich environment of the island), these two deterministic factors were evaluated separately to identify North African dust particles. Three different assessment methods were applied to determine the amount of Saharan dust material of the samples.

An indirect approach was applied to theoretically discriminate the silt-sized sedimentary subpopulations which were mathematically separated. The polymodal grain size distribution curves were partitioned into several unimodal Weibull-distributions by applying parametric curve-fitting technique (Sun, D. et al., 2002, 2004; Varga, Gy. et al., in press). According to the applied parametric curve fitting technique the polymodal particle size curves can be interpreted as sum of several, in this case three overlapping Weibull-functions which represent three sediment populations. According to published data on recent dust events from the area (Criado, C. et al., 2003; Menéndez, I. et al., 2007; von Suchodoletz, H. et al., 2009) and measurements of other far-travelled North African dust material (Varga, Gy. et al., 2016), the subpopulation with smallest particles are regarded as the product of long-ranged dust transport.

Raman-spectroscopy (at 785 nm wavelength with 3µm spot) was also applied to directly identify the quartz grains as an indicator of Saharan dust contribution. The acquired spectra of targeted particles were compared to Raman spectral reference libraries using KnowitAll® software from Bio-Rad to identify the minerals present.

The third applied technique was based on the grayscale intensity mean values of particles, the relatively high values were used as a proxy for quartz grains as it was found that there is a strong correlation between light transmissivity and chemical identity (especially in this special case of an environment characterized with the overwhelming majority of carbonate and quartz particles).

Cluster and networks analysis techniques were also applied to differentiate various mineral particle populations based on their general normalized shape (aspect ratio, circularity, convexity, solidity) and grayscale intensity (mean, standard deviation) values. Hierarchical cluster trees were created by using the Euclidean distance pairs of the selected parameters of separated quartz and carbonate size fractions (fine, medium, coarse silt and sand).

For network analysis 192×192 [(24 samples × 2 minerals × 4 size fractions) × (24×2×4)] matrix was compiled, where coefficient of determination was calculated for each pair of records based on the normalized shape and grayscale intensity parameters. This matrix was transformed into an adjacency matrix with values of 0 if r²<0.99 and 1 if r²≥0.99, in this way all of the similar mineral grains were coupled and the whole database can be handled as a network or a finite graph, where the similar records (nodes) are connected (edges) to each other. The Gephi network visualization software was used to analyse the compiled network by applying the ForceAtlas2 continuous graph layout algorithm (Jacomy, M. et al. 2014).

Static vs. dynamic image analysis

Granulometric characterization of paleosols in loess series by automated static image analysis

An automated image analysis method is proposed here to study the size and shape of siliciclastic sedimentary particles of paleosols of Central European loess sequences. 

New paper: Varga et al. (2018) Sedimentary Geology 370, pp. 1-14. DOI: j.sedgeo.2018.04.001

Several direct and indirect measurement techniques are available for grain size measurements of sedimentary mineral particles. Indirect techniques involve the use of some kind of physical laws, however, all requirements for calculations are in many cases not known. Even so, the direct manual microscopic observation and measurement of large, representative number of grains is time-consuming and sometimes rather subjective. Therefore, automated image analyses techniques provide a new and perspective way to analyse grain size and shape sedimentary particles. 

Here we test these indirect and direct techniques and provide new granulometric data of paleosols. Our results demonstrate that grain size data of the mineral dust samples are strongly dependent on shape parameters of particles, and shape heterogeneity was different of the different size classes. Due to the irregular grain shape parameters, uncertainties have arisen also for the sizes. 

In this paper we present a possible correction procedure to reduce the differences among the results of the laser diffraction and image analysis methods. By applying new correction factors, results of the two approaches could be get closer but the most definite factor, the unknown thickness of particles remained a problem to solve. The other presented method to assess the uncertain 3rd dimension of particles by their intensity-size relationships makes us able to reduce further the deviations of the two sizing methods.

Granulometry - 101

Interpretation of sedimentary (sub)populations extracted from grain size distributions of Central European loess-paleosol series

Varga, Gy., Újvári, G., Kovács, J. (in press). Interpretation of sedimentary (sub)populations extracted from grain size distributions of Central European loess-paleosol series. Quaternary International

Grain size proxies of aeolian dust deposits have widely been applied in environmental and sedimentary studies. However, large body of research papers are not taking into consideration that a complex grain size distribution curve cannot be an indicator of a single one environmental factor (e.g. wind speed/strength, transportation distance, aridity).

The aim of the present paper is to discuss the main differences of frequently used statistical methods and to provide possible interpretations of the results by applying these various approaches on the high-resolution loess-paleosol profile of Dunaszekcső, South Hungary (Central Europe). Beside single statistical descriptors (mean, median, mode) of grain size and simple indices of size-fraction ratios (U-ratio, Grain Size Index), some more complex algorithms were also used in our paper. The applied parametric curve-fitting, end-member modelling and hierarchical cluster analysis techniques are using the whole spectrum of the measured grain size distributions and provide a more reliable and more representative results even in case of small scale variations.

According to our findings, approaches which provide direct linkage among simple statistical descriptors and single atmospheric or other environmental elements are rather oversimplified as properties aeolian dust deposits are influenced by the integrated effects of several concurrent processes. Differences of more complex decomposition methods arise from the different approach and scope. End-members are determined from the unmixing based on the covariance structure of the whole grain size data-series of the section, while the parametric curve-fitting is based on the one-by-one deconvolution of the grain size distribution curves. End-members of loess-paleosol samples are regarded as representation of the average dust grain size distribution of various temporal sediment clusters of seasonal or other short-term intervals, while (sub)populations by parametric curve-fitting are proposed to illustrate process-related elements of background and dust storm depositional components for each sample. Results of cluster analysis represent similar grouping conditions as end-member modelling with a reduced sedimentary and genetically meaning.

How to fit Weibull functions?

You can separated the different sedimentary sub-populations by using a simple MS Excel software with SOLVER. The difference between the measured and modelled values (sum of two/three Weibull distributions) has to be minimized.