Particle morphomics by high-throughput dynamic image analysis

A novel omics-like method referred to as “particle morphomics” has been proposed in the present study. The dynamic images of >2,000,000 particles per sample in sediments, soils and dusts were collected by a Sympatec GmbH QICPIC particle size and shape analyzer, and the morphological descriptors of each particle including equivalent diameter, sphericity, aspect ratio and convexity were extracted as the “particle morphome”. Various multivariate analyses were adopted to process the high-throughput data of particle morphome including analyses of alpha and beta diversities, similarity, correlation, network, redundancy, discretion and principal coordinate. The outcome of particle morphomics could estimate the morphological diversity and sketch the profile of morphological structure, which aided to develop a morphological fingerprint for specific particle samples. The distribution and properties of particle assemblages of specific morphology could also be evaluated by selecting particles with respect to filter criteria. More importantly, the particle morphomics may be extended to investigate and explain the biogeochemical and environmental processes involved with particle morphology if linked with external variables.


Index for Supporting Information
Supporting Texts: Text S1. Measurement principle of QICPIC system …………………………………. 3 Text S2. Workflow of extracting data by WINDOX 5 …………………………….. 4-6 Text S3. Definitions of morphological descriptors …………………………………... 7 Text S4. Calculations of morphology distribution parameters ……………………….. 8 Supporting Tables: Table S1. Sampling information for particle samples ……………………………….. Text S1. Measurement principle of QICPIC system The QICPIC particle size and shape analyzer is an instrument, which can directly measure and analyze the particle size and shape of a large number of fast moving particles. Its measurement is based on the principle of optical imaging, and the following figure present the diagram of particle image measurement. When measuring, the pulsed light emitted from the high-frequency pulsed light source passes through the beam expander to obtain parallel pulsed light, which irradiates on the dispersed single particle in the test area; through a special optical imaging system, the clear image of each particle at the orthogonal orientation of the projection direction is obtained. After a large number of image data are processed by computer, the single particle close-up image, or the morphological information and distribution of all the statistical particles in a sample can be obtained. The particle size range of QICPIC measurement is from 1 to 20000 μm. The stroboscopic technology uses pulsating light source. The time of each pulse is less than 1×10 -9 s, and the stroboscopic rate of the lens is about 109 times/s. The high-speed camera system has a resolution of 1024×1024 pixels and a gray scale of 256. Even particles with a moving speed of up to 100 m/s can be clearly imaged. The QICPIC measurement employs the high-speed camera system CMOS, which is capable of shooting 500 images per second, a single-winding double-stranded cable with 1.25 GByte/s of transmission rate, a specially designed PCI interface, and a workstation with dual-core processing technology. Therefore, it can realize the real-time transmission of a large number of test data and images, and obtain tens of thousands to millions of particle images in a very short time, which ensures the density and total amount of test particles, and achieves the well representative results.

Text S2. Workflow of extracting data by WINDOX 5
WINDOX 5 the special software for QICPIC measurement and analysis, which comprises of multi-purpose program groups (packages) for sensor control, real-time storage and processing of high-throughput images, database management, and data output. During the QICPIC measurement, the high-throughput images will be stored real-timely in standard Database as an FDB file. The image data of every single particle including X-coordination, Y-coordination and number of set (black) pixels are encoded as 16 bit values in hexadecimal coding, as shown in the example below. After the QICPIC measurement, we could choose the calculation properties such as equivalent diameter, sphericity, aspect ratio and convexity (Text S3) in the Work Page, and perform the calculation. The calculation results including morphological descriptors and morphology distribution parameters (Text S4) are integrated in the FDB file. The Application Program provides a programming language, whose grammar is similar to BASIC language, to retrieve and export the morphological data. In Application Program, we can retrieve the morphological data from FDB file by using the statement "JOURNAL" with the function of returning a report of the current measurement data as a string of characters, and a series of Template Commands (see table below). The statement "DATAEXPORT" is used to export the image analysis data into the CSV file. Particle size at the maximum of the linear distribution density; n: quantity measure μm @dmxg.(n) Particle size at the maximum of the logarithmic distribution density; n: quantity measure μm @xstdd.(n) Standard deviation of the particle size distribution n: quantity measure μm @SKEW.(n) Skewness of a size distribution; n: quantity measure μm @RRSB.D(n) RRSB Fineness parameter d'; n: quantity measure @RRSB.N(n) RRSB Fineness parameter n; n: quantity measure @RRSB.R2(n) Regression quality of an RRSB fit n: quantity measure @RRSB.S(n) Standard deviation of the RRSB fit n: quantity measure @xexp.(n) Statistical expectation of the particle size distribution n: quantity measure μm @NX Number of classes of a size distribution @x.(k) Upper limit of particle size class k k: particle size class, k out of {1,...,@NX} μm @Q.(k,n) Cumulative distribution value at the upper particle size limit of class k k: particle size class, k out of {1,...,@NX} n: quantity measure % @R.(k,n) Residue distribution (1-Qn) k: particle size class, k out of {1,...,@NX} n: quantity measure % @dQ.(k,n) Fraction (percentage in size class ) k: particle size class, k out of {1,...,@NX} n: quantity measure % @q.(k,n) Linear distribution density 1/mm k: particle size class, k out of {1,...,@NX} n: quantity measure @xgm.(k) Geometric mean value of the particle size class k k: particle size class , k out of {1,...,@NX} μm @qx.(k,n) Logarithmic distribution density k: particle size class, k out of {1,...,@NX} n: quantity measure @SD.NS Number of classes of the shape factor distribution @SC.S(k,s) Mean shape factor of the particles in size class k k: (1 ≤ k ≤ 1000); s: calculation mode @SD.dmS (s,n) Shape factor at the maximum of the shape factor distribution density s: calculation mode; n: quantity measure @SD.EXC(s,n) Kurtosis Excess of a size distribution s: calculation mode; n: quantity measure @SD.Q(k,s,n) Cumulative distribution value for shape factor class k k: (1 ≤ k ≤ 1000); s: calculation mode n: quantity measure % @SD.q(k,s,n) Linear distribution density value for shape factor class k (1 ≤ k ≤ 1000) s: calculation mode; n: quantity measure @SD.R(k,s,n) Residue distribution value for shape factor class k k: (1 ≤ k ≤ 1000); s: calculation mode n: quantity measure % @SD.S(k,s) Upper limit of shape factor class k used for the calculation of the distribution (1 ≤ k ≤ 1000) s : calculation mode @SD. Sexp(s,n) Statistically expected value of the shape factor distribution s: calculation mode; n: quantity measure @SD.SKEW(s,n) Skewness of a size distribution s: calculation mode; n: quantity measure @SD.Sm(k,s) Mean value of the shape factor for class k, (1 ≤ k ≤ 1000); s: calculation mode @SD.Sstdd(s,n) Standard deviation of the shape factor distribution s: calculation mode; n: quantity measure Wherever in this list, the parameter "n" is mentioned, and describes the quantity measure (0=number, 1=length, 2=area, 3=volume) of the distribution; the parameter "s" is calculation mode for the shape factor (1=sphericity, 2=aspect ratio, 3=convexity).

Text S3. Definitions of morphological descriptors
The equivalent diameter is defined as the diameter of a circle with the same area as the projection area of the particle. The sphericity is the ratio of the perimeter of the equivalent circle to the real perimeter of the particle. The aspect ratio is the ratio of minimal to the maximal Feret diameter of the particle. The convexity is defined as the ratio of the projection area and the area of the convex hull of the particle. The length of a fiber is defined as the longest direct path from one end to another within the particle contour without loops or deviations. The diameter of a fiber is calculated by dividing the projection area by the sum of all lengths of the branches of the fiber. The elongation is the ratio of diameter and length of a fiber.

Text S4. Calculations of morphology distribution parameters
Expectation: where x is the morphological descriptor value (i.e., equivalent dimeter, sphericity, aspect ratio or convexity) Standard deviation: where n is the number of size or shape factor class, andx is the mean value of a specific descriptor. Kurtosis: where m4 is the fourth central moment, and m2 is the variance. The sample was collected from a small harbor in the estuary of Yangtze River. It was comprised of spongy fine particles with a lighter color. Some reeds were grown on it.

Skewness:
6 Sediment E119°58'59", N32°53'6" The sample was collected at the bank of a small river with a gentle flow in Xinghua, Jiangsu. The sample was comprised of particles of various sizes with a black color. Some weeds were grown on it. 7 Sediment E119°58'58", N32°53'6" The sample was collected at the central riverway of the same river of collecting No. 6 sample. The sample had a similar character with No. 6 sample but without plants growing on it.
8 Soil E121°20'44", N31°29'11" The sample was collected at a paddy field in the suburban area of Baoshan Dist., Shanghai. The field has been reclaimed for a long time with a high fertility. It was about to spring plant when sampling. 9 Soil E121°45'47", N31°12'54" The sample was collected from a rape field at Sanjia Port in Pudong Dist. of Shanghai, and comprised of silt and clay particles with a darker color.
10 Soil E121°29'50", N31°20'25" The sample was an urban soil and collected at a construction site to be developed in Yangpu Dist. of Shanghai. The sample was withdrawn from top layers, and comprised of varying sizes of particles with a moderate color. The sample was potential backfill.
11 Soil E121°29'50", N31°20'25" The sample was the subsoil of No. 10 sample with a higher density and a light yellow color. The sample was an old and compacted urban soil.
12 Dust E121°29'29", N31°18'7" The sample was collected at the windowsill of a student dormitory in the Handan Campus of Fudan University with a light grey color.
13 Dust E121°17'10", N31°8'1" The sample was collected at the ground close to the residential area. A dust point source was a nearby building materials factory.
14 Dust E121°17'10", N31°8'1" The sample was collected at a nearby site close to the sampling location of No. 13 sample but more close to the building materials factory. The sample has a similar character with No. 13 sample.  .987(**) 1 * correlation is significant at the 0.05 level (2-tailed); ** correlation is significant at the 0.01 level (2-tailed); particle sample number (including duplicates) = 28. Abbreviation: VMD, volume mean diameter; VMS, volume mean sphericity; VMA, volume mean aspect ratio; VMC, volume mean convexity; SD, standard deviation.