Projected response of Arabian sea Oxygen minimum zone to climate change: Insights from a set of downscaled experiments

The land-locked northern boundary and seasonal high productivity in the Arabian sea (AS) leads to the formation and the maintenance of one of the most intense and thickest open ocean oxygen minimum zones (OMZ) there. Earlier studies based on both observation and model sensitivity experiments have reported that this perennial OMZ is highly sensitive to the strength of the monsoonal circulation and surface heating. Model simulations from the fifth phase of Coupled Model Intercomparison project (CMIP5) indicate significant changes in the Indian monsoonal circulation and the atmospheric heat fluxes under climate change. However, the future projection of AS OMZ under climate change remains largely uncertain and ill-understood. This is mainly due to a poor representation of the AS OMZ in the CMIP5 simulations and an important spread in their future oxygen projections for the region. Here we explore how downscaling CMIP5 global simulations with a high-resolution configuration of the Regional Ocean Modeling System (ROMS) model coupled to a nitrogen-based NPZD ecosystem model can help improving the representation of the AS OMZ and reduce the spread in CMIP5 projections. To this end, we performed a climatological “reference” simulation, i.e., the control simulation, where ROMS is forced with observed atmospheric and lateral boundary conditions, and a set of corresponding downscaled sensitivity experiment where ROMS is forced with atmospheric and lateral boundary conditions derived from global CMIP5 simulations. For the downscaling experiment, we chose two best performing models from the CMIP5 database based on their skill in simulating the present day (historical) climatology. The control simulation has been extensively validated against the observations for its skill in simulating the physical and biogeochemical variables. We explore the sensitivity of the downscaled oxygen distribution and OMZ to the regional model setup by varying the model resolution from 1/3deg to 1/12deg and expanding the model domain from a small AS-limited domain to one encompassing the full Indian Ocean. We show that the downscaled experiments improve the representation of different classes of oxygen (Oxic - O2 > 60mmol/l; Hypoxic - 60mmol/l >= O2 > 4mmol/l; and the Suboxic  - 4 mmol/l > O2 > 0 mmol/l) within the 0-1500m depth range. In particular, the downscaled experiments simulate a much smaller fraction of suboxic waters relative to hypoxic and oxic fractions, in agreement with observations.