Biochar has unique function to improve soil chemo-physical and biological properties for crop growth. Because changes of biochar in soil may affect its long-term effectiveness as an amendment, it is important to quantify and monitor biochar after application. In this paper, we propose a solution for this problem via hyperspectral image analysis. We treat the soil image as a mixture of soil and biochar signals, and then apply hyperspectral unmixing methods to predict the biochar abundance at each pixel. The final percentage of biochar can be calculated by taking the mean of the abundance of hyperspectral pixels. We have compared several hyperspectral unmixing methods based on least squares estimation and nonnegative matrix factorization with various sparsity constraints. Experimental results are evaluated by polynomial regression and root mean square errors against the ground truth data collected in the environmental labs. The results show that hyperspectral unmixing is a promising method to measure the percentage of biochar in the soil.