Abstract
Accurately predicting the stability of a biochar sample placed in the environment is important for guiding climate policies and the emerging voluntary market for carbon removal. The stability of biochar in soils varies with feedstock type, pyrolysis parameters, and environmental conditions. Previous assessments have correlated biochar stability estimates to single features – like pyrolysis temperature, elemental molar ratios, or incubation duration – but these assessments used different datasets and methodologies and reached different conclusions. Therefore, our aim was to develop an open dataset of biochar decomposition experiments, and to develop a transparent data preparation and processing toolchain, enabling reproducibility of scientific results. We first made an inventory of all published biochar incubation experiments, and then collected the incubation data and an extensive set of associated metadata (i.e., biomass and biochar properties, pyrolysis and incubation conditions). In a second step, we developed a data analysis toolchain, including functions for extrapolation of the incubation data to longer times and models for correlation between metadata and estimated biochar stability. In the extrapolation step, the incubation data was fitted to decay functions. Care was taken to explore the effect of using different fitting algorithms and constraints, and to apply a recalibration of the incubation temperature. For the correlation step, several strategies were applied, including both single-feature linear regressions to reproduce previous results and multi-feature regressions based on decision trees. So far, a dataset of 135 observations with more than 8000 data elements was collected making it one of the largest biochar stability datasets available. For the first time, raw biochar decomposition data is also compiled for 111 observations (mostly laboratory incubations). The initial data exploration revealed that although pyrolysis temperatures in the range 350 to 700°C are well represented, there is a data gap at higher temperatures with only a few data points at 1200°C. Likewise, only two observations are available with a molar H/C ratio below 0.2. These gaps can guide design of future incubation studies, as these parameters are often seen as indicators of stability. During curve fitting with single, double, or triple exponential models, we noted that the choice of initial conditions was important for finding a good fit, but we also noted that in many cases fitting uncertainties were high, residuals were not necessarily randomly distributed, and that some observations did not fit well to any type of exponential model (likely due to experimental conditions). Nevertheless, we were able to approximately reproduce the fitting results reported in Woolf et al. (2021). Finally, linear correlations were established between predicted stability and pyrolysis temperature, molar H/C ratio, but also other features available in the dataset, yielding similar correlation coefficients as previously reported (0.1 to 0.4). Attempts to understand the variability in predicted stability (using principal component analysis) and to develop multi-variate and non-linear models have so far not significantly improved model performance without overfitting. Opportunities remain to use the compiled data for other types of modelling, e.g. in soil carbon models.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.