A set of numerical simulations were conducted to determine the potential interannual variation of phytoplankton biomass in the Arabian Sea. A four-layer, reduced gravity model was forced with interannually varying, monthly mean winds for the years 1981 to 1990 and a seasonally varying, monthly mean climatological heat flux. Passive tracers, with fixed source and decay rates, were used as proxies for phytoplankton. The tracers were released in the second layer, and upwelling combined with interlayer mixing transported the tracers into the first (the top) layer of the model. In the top layer advection, diffusion and specified decay rate produced distributions of tracer concentration that we compare with coastal zone color scanner (CZCS) images. In addition to maps of tracer distribution, we constructed time series from areal averages of tracer concentration. The time series are particularly useful for analyzing the temporal variability of the tracer concentrations. There was substantial interannual variability in the results and a strong seasonal signal associated with the southwest monsoon. The years of maximal tracer concentration were 1990 and 1985. The first was associated with a decadal maximum in wind speed and the second with a large wind stress curl over the western Arabian Sea. The years with minimal tracer concentration were 1981 through 1983. We compare our results with published analyses of composited CZCS images during the southwest monsoon in 1981 and 1982. We find general agreement between our results and the CZCS images during the peak of the southwest monsoon; however, we find that during the early phase of the monsoon, when the winds are strengthening rapidly with time, the CZCS images are badly aliased by year to year changes in the time periods used to composite available satellite data. ¿ American Geophysical Union 1995