The western area of Abu Dhabi's Late Jurassic Arab Formation is a huge ultra-sour gas reservoir with an areal gradient in composition. Early on, the data suggested a trend in H2S concentrations along the axis of the field with a sour gas entering from the southwest and migrating to the northeast. This sour gas contaminated the existing reservoir fluid and created the current areal gradient. Gas properties also varied, and in particular, dew point pressure. This paper describes the innovative methodology used by ADNOC Sour Gas to input the varying compositions into the simulation model. The methodology consists of two major steps. In the first, PVT data was analyzed and correlations between H2S and other components were established. The second step involved using PETREL to create compositional maps. Ultimately each grid block was assigned a unique composition based on the H2S concentration at that location resulting in a continuously varying compositional gradient. Concentrations of other components were assigned based on the H2S concentration. The result was a dynamic model which duplicates the areal distribution in composition and accurately predicts the varying dew point pressures using a single Equation of State. Simulation predictions of condensate and sulfur production has been verified by actual plant yields. Four years of production has shown the veracity of the initialization of the composition in the model as no modifications to the original compositional distribution was required.