Climate change is one of the major environmental problems that humanity is facing. Studies suggest the use of negative emission technologies (NETs) to substantially decrease atmospheric CO2 concentration. Biochar can be applied to the soil for long-term carbon sequestration and simultaneous increase in soil fertility. However, the change in biochar properties after field application and exposure has not yet been thoroughly investigated, necessitating the need to determine if biochar can maintain its desirable properties with aging. Rough set-based machine learning (RSML) can be used to generate a data-driven model to predict the effect of different attributes on % change in C content of biochar. Four rules were accepted that relate the effect of feedstock type, pyrolysis temperature, aging method, and aging duration to % change in C content of biochar. Rules 1 and 2 cover 27 % and 14 % of the training set with 100 % accuracy, while Rules 3 and 4 cover 14 % and 10 % of the training set with 66.67 % accuracy. The performance of the rules was assessed via a k-fold (k=10) cross validation method and checked for mechanistic plausibility. The findings of the study can help maximize the potential benefit of biochar in climate mitigation.