Soil sodicity is a major limitation to wheat production in Australia, yet estimating its yield impact remains challenging due to interactions with climate, management, co-occurring soil constraints, and heterogeneous data support. This study compares three approaches for quantifying and mapping yield penalties attributable to sodicity across the Northern Grains Region (NGR) during 2000-2016. Two empirical approaches were evaluated alongside simulations based on a process-based cropping system model, the Agricultural Production Systems sIMulator. Climatic conditions were stratified into wet, moderate, and dry years based on August vapour pressure deficit, representing atmospheric demand during a critical growth period around anthesis in the NGR. The approaches produced systematically different estimates in both magnitude and spatial extent. Empirical Approach 1 and APSIM, both anchored to spatially distributed soil-profile data, generated coherent regional patterns and higher mean penalties, with APSIM consistently producing the largest values. In contrast, Empirical Approach 2, trained on spatially clustered yield-monitor data, produced lower regional averages but greater within-field variability and more pronounced high-end penalties, particularly under dry conditions. These differences reflect contrasts in data support, model structure, and representation of sodicity effects. The empirical approaches estimate realised sodicity-related yield penalties within the observed data domain. In contrast, APSIM estimates the yield gain from removing sodicity relative to sodicity-constrained water-limited yield potential, resulting in larger absolute penalties in environments with higher attainable yield. The strongly scale-and method-dependent estimates of yield penalties therefore provides complementary insights for regional to paddock-scale decision-support for managing soil constraints in dryland cropping systems.