Realistic ocean models require significant computing power, especially as resolution increases from scales of tens of kilometers to a single kilometer. These models are used for short-term and seasonal weather predictions, as well as in long-term climate simulations, all of which are used routinely for decision making. The better the resolution, the better potential dangers can be understood and mitigated, resulting in a better situation for all… right?
That's the question Eric P. Chassignet and Xiaobiao Xu, of Florida State University's Center for Ocean-Atmospheric Prediction Studies, asked in an overview paper published on July 31 in
Advances in Atmospheric Sciences.
"Increasing the resolution allows you to resolve more and more small-scale ocean features, and the question that then arises is as to whether there is corresponding improvement in the overall representation of the ocean circulation and at what cost," said Chassignet, who directs the center and serves as distinguished professor of oceanography. "In other words, what is the optimal ratio of resolution and computational resource that truly leads to a significantly better understanding of ocean physics and the Earth’s climate?"
Coarse resolution models, with a horizontal resolution on the order of 100 kilometers, are mostly used for climate applications and the ocean currents in this class of models tend to be broad and steady. When the model’s resolution is increased to approximately 10 kilometers, the currents become unstable, forming swirling ocean mesoscale eddies, somewhat like storms in the atmosphere. Just like storms, they have an impact on other components of the earth system.
Modeled surface relative vorticity in the Gulf Stream region depicting the flow rotation and shear as a function of horizontal grid spacing (~50, 6, and 2 km, respectively). In addition to a well defined Gulf Stream and associated mesoscale eddies, one can observe an increased number of small-scale features (submesoscale) that arise when the model horizontal resolution is increased to ~2 km. (Image by Advances in Atmospheric Sciences)
However, resolving the mesoscale eddies is not enough to accurately model the ocean circulation, according to Chassignet. His team determined that increasing resolution to approximately a kilometer, which makes the model able to simulate smaller, sub-mesoscale eddies, shifted their model of the Gulf Stream to a realistic rendition that more closely resembled actual observations.
“We argue that resolving sub-mesoscale features is as significant a regime shift as resolving the mesoscale eddies was,” Chassignet said.
Yet the resolution comes at a price and with a concern, since each time the model’s resolution is increased by a factor of two, it requires increasing computational power by a factor of 10. According to Chassignet, more work is needed to better understand if increasing resolution improves the overall representation of ocean water masses. For example, an increase in the resolution does not always deliver a clear bias improvement in temperature or salinity drift.
"The next step is to have routine, sub-mesoscale-resolving global ocean models so we can fully assess their ability to model the ocean and quantify their impact in climate models," Chassignet said, noting that close collaborations with computer scientists are essential moving forward to ensure computer systems that can more efficiently handle the modeling needs for Earth systems.
The Chinese Academy of Sciences President's International Fellowship Initiative supported this work in part.
