Abstract: Soil parameters exhibit inherent spatial variability and uncertainty, and their rational and accurate characterization is fundamental to improving the precision of slope reliability and risk assessments. Random field theory provides an effective framework for quantitatively capturing the spatial variability of soil properties. However, traditional unconditional random field (URF) models fail to fully incorporate the spatial location information of site investigation data, leading to significant discrepancies between simulated and observed values at measurement points, which further reduces the accuracy of slope reliability and risk evaluation. To address this issue, this study proposes a Constraint Seed Method(CSM) for constructing conditional random fields (CRFs) of soil parameters using available site investigation data. By introducing observational data at the random field generation stage, CSM imposes spatial constraints on independent random seeds, thereby achieving precise control of the spatial distribution of soil parameters. The CRFs generated by CSM possess two major advantages: (i) the simulated values at observation points strictly match the measured data, ensuring high consistency between the model and field measurements; and (ii) in unobserved regions, the simulated results exhibit statistical consistency with measured data within the 95% confidence interval, effectively reflecting the inherent spatial correlation of soil properties. Two representative slope cases are presented to validate the proposed method. Results demonstrate that CSM achieves significantly higher computational accuracy and efficiency than other methods. Incorporating borehole data causes the probability density function of the posterior factor of safety to converge, with a substantial reduction in variance. The estimated failure probability and risk differ markedly from those derived solely from prior information, providing a much closer approximation to actual engineering conditions and highlighting the corrective effect of site data on risk perception. Furthermore, as the number of boreholes increases, the estimated failure probability and risk values progressively converge toward their true levels. Owing to its simplicity, robustness, and high computational efficiency, CSM can offer a practical and effective approach for optimizing site investigation programs and supporting risk management of slope engineering projects.