Advanced descriptive statistics of random reflectance measurements on plant-based biochars-do they even matter?

Numerous recent publications have demonstrated the relationship between random reflectance and the proportion of the fully carbonized fraction (equivalent to the fusinite maceral) contained in a biochar sample. These findings have motivated international frameworks and independent carbon registries to consider random reflectance among the core analytical proxies required to assess biochar permanence in soil. However, skepticism for the application of the proxy still persists, with main challenges revolving around the aspects of data acquisition and data interpretation. This is mostly attributed to the fact that the methods applied in the microscopic study of biochar were originally developed and standardized for the study of coal, where the calculation of the average and standard deviation of a 100 measurements on collotelinite, accompanied by a histogram showing the frequency distribution of the measured values, is often enough to describe and report this optical property.

Even though biochar samples are petrographically much simpler than coal and other sedimentary rocks, they have peculiarities that require a more careful consideration when applying standard petrographic techniques for their study. Biochar manufacturing conditions play a major role in the extent of the carbonization degree of the feedstock and this in turn has an impact on the heterogeneity of the formed biochar grains often resulting in complex distributions of the reflectance values, not always accurately captured in the basic descriptive statistics (mean and standard deviation, etc.), particularly in the case of polymodal distributions. For this reason a higher number of measurements, ranging between 300-500, on fields of view selected along a regular grid, is required to acquire meaningful average and standard deviation values, as opposed to coal samples, where a “run of the sample” on parallel traverses where collotelinite occurs is a common practice.  

Advanced descriptive statistics have long been successfully used for the evaluation of grain size analysis of clastic sedimentary rocks for the assessment of reservoir properties and depositional environment. This study attempts to investigate the frequency and probability distributions and derived advanced descriptive statistics of random reflectance measurements acquired from 50 plant-based biochar samples, in order to characterize their heterogeneity with regards to the proportion of the fully carbonized fraction. The calculated advanced descriptive statistics include the coefficient of variation and confidence intervals, measures of central tendency (median and mode), measures of dispersion (variance and interquartile range), shape parameters (skewness and kurtosis), and probability-related measures (probability density function, cumulative distribution function, and percentiles, particularly those associated with the established “inertinite benchmark”-IBR_o2). In addition to those, the study attempts a comparison between the IBR_o fractions determined by the measurement of 3-4 points per field of view, against those determined by just measuring the point located at the crosshair of each field of view, together with the convergence of the acquired set of measurements to the mean and median of each sample. Findings are expected to contribute to a mathematically more robust characterization of the acquired datasets, providing greater rigor in how this data can be utilized with regards to the assessment of biochar carbon permanence.