In engineering geology, a reasonable assessment of the spatial distribution of uncertainty in a region is vital in guiding research, saving money, and shortening the period. However, the traditional modeling process requires a lot of manual interaction, and the uncertainty of the geological model cannot be accurately quantified and utilized. This paper proposes a novel implicit geological modeling and uncertainty analysis approach based on the triangular prism blocks, which is divided into data point acquisition, ensemble model with divide-and-conquer tactic (EMDCT), uncertainty analysis, and post-processing. By employing machine learning algorithms, the EMDCT gives superior results for implicit modeling. The sensitivity analysis of the prediction results is further evaluated via information entropy. According to the distribution of uncertainty, supplementary boreholes are selected as additional knowledge to retrain the local components of the model to enhance their performances. The implicit modeling method is applied to real hydraulic engineering problems by employing the EMDCT, and the proposed model has obvious advantages in the implicit geological characterization. The overall accuracy in the working area with sparse boreholes reaches 0.922, which is 0.013 higher than the traditional method. By evaluating the distribution of uncertainty, an accuracy of 0.962 can be achieved, which is equivalent to reducing 10 boreholes.