Accurate resource estimation is essential in mining projects, as it directly influences economic feasibility assessments and operational planning. A fundamental step in this process is the delineation of geological domains, traditionally based on sample data and geological interpretation. Within each domain, resource estimation is commonly performed using methods such as ordinary kriging (OK) or inverse distance weighting (IDW). These methods often simplify domain boundaries as hard, assuming abrupt changes in mineral grades at domain limits. However, in many geological settings, such as porphyry copper deposits, mineralisation transitions gradually between domains, forming soft boundaries. Relying solely on hard boundary assumptions may misrepresent these transitional zones, leading to biased estimations. This study investigates non-stationary estimation techniques that account for spatial correlations across soft boundaries, improving the accuracy and geological validity of resource models. We evaluate and compare four estimation methods: 1. Ordinary kriging (OK) within domains: Assumes hard boundaries and performs estimation separately within each domain. 2. Ordinary kriging with a non-stationary variogram (OKNS): Introduces a location-dependent variogram, where parameters (range, nugget, and sill) vary with distance to the boundary. 3. Simple kriging with a variant mean (SKVM): Incorporates a local mean that adjusts based on proximity to the boundary, capturing local grade variations. 4. Hybrid approach (OKSKVM): Combines OK in stable areas with SKVM in transition zones, ensuring a balance between numerical accuracy and geological continuity. The methodologies are applied to a porphyry copper data set, where experimental variograms and contact analysis confirm the presence of a soft boundary between two adjacent domains. Cross- validation results show that OKNS minimises estimation errors, while OKSKVM maintains geological continuity by ensuring a smooth grade transition. These findings highlight the importance of non- stationary models in accurately estimating resources within domains with soft boundaries. In particular, the study emphasises the critical role of precise grade estimation in the short-term stage of mining, where accurate resource modelling directly informs production scheduling and excavation strategies. Integrating refined estimations into short-term plans ensures optimal extraction sequencing and minimises operational risk, ultimately enhancing both resource recovery and economic performance. The proposed approaches provide a more robust framework for resource evaluation in mining projects, supporting more informed and adaptive short-term planning decisions.