To simplify soil landscapes into meaningful management units, soil individuals can be grouped into homogeneous classes that can be presented and interpreted through maps. This approach promotes the dimensions of soil security by facilitating effective management of soil with similar characteristics. Numerical soil classification models have been proposed to reduce the uncertainties and subjectivity in expert-based soil groupings. This study aimed to numerically cluster soil profiles using multiple algorithms, following a "pedogenon"-inspired methodology applied at the field scale. The study was performed on a 58,000-ha low-relief landscape in an arid to semi-arid region in Iran. Common clustering models were applied, including hierarchical clustering using Euclidean distance (HC) and Mahalanobis distance (HM), k-means (KM), Partitioning Around Medoids (PAM), and fuzzy c-means (FCM). Additionally, a biclustering (BiC) algorithm was evaluated for the first time in soil classification. Following the pedogenon methodology, environmental soil-forming factors were used as inputs for the numerical classifications. These were evaluated on their Akaike information criterion (AIC) derived from a generalized least squares model of ten physio-chemical soil properties at harmonized depth intervals. The HM algorithm achieved the lowest AIC, indicating the best fit, followed by FCM, KM, HC, PAM, and BiC. The HM clusters aligned well with the conceptual understanding of soil scientists as expressed through the USDA Soil Taxonomy, reflecting coherent changes in soil properties along the landscape. Additionally, the HM clusters provided a clearer separation in the soil properties with depth than a legacy Soil Taxonomy subgroup map. The clustering algorithms that were easiest to interpret were also the most accurate, thereby enhancing communication across scientific disciplines, policymakers and farmers. Despite its relatively weaker performance, the BiC model showed moderate alignment with conventional clustering algorithms, demonstrating a strong similarity with KM cluster patterns and therefore warrants further investigation.