It has long been recognized that the Arab-D reservoir in Ghawar field has been significantly dolomitized and that the distribution of dolomites is highly heterogeneous across this reservoir. Previous studies indicated that dolomite occurs with either a stratigraphic or non-stratigraphic distribution; when mapped, dolomite tends to form several parallel linear trends across the field. Although stratigraphic dolomite was suggested to be formed early from highly evaporated pore fluids sourced from overlying evaporite deposits, non-stratigraphic dolomite was thought to be generated primarily from hydrothermal fluids sourced from below. This study focuses primarily on these non-stratigraphic dolomites, and proposes that: (i) these dolomites initially formed via seepage reflux, but were reinforced by late stage hydrothermal dolomitization; and (ii) reflux is also responsible for the formation of parallel, linear trends of dolomite. The reflux model hypothesizes that an evaporative lagoon (which is the source of dolomitizing fluids) formed during the falling stage systems tract of a depositional sequence, and that with continuing sea-level fall this lagoon migrated progressively towards deeper parts of an intrashelf basin adjacent to the Ghawar field, leaving behind lines of dolomite bodies along a series of temporary coastlines. Two-dimensional reactive transport models have been built to test this hypothesis, and have resulted in a predicted pattern of dolomite bodies that agrees with both the observed vertical distribution of non-stratigraphic dolomite, as well as the mapped lateral distribution of the dolomite trends. In addition, the major ion compositions of Late Jurassic seawater are calculated based on fluid inclusion data in the literature. Using Jurassic seawater in current models leads to the absence of anhydrite cements and less potential of over-dolomitization than using modern seawater.