Abstract
This study compares the skills of two numerical models having the same horizontal resolution but based on different principles in representing meso- and submesoscale features of ocean dynamics in the Lakshadweep Sea (North Indian Ocean). The first model, titled LD20-NEMO, is based on solving primitive equations using the NEMO (Nucleus for European Modelling of the Ocean) modelling engine. The second one, titled LD20-SDD, uses a newer Stochastic-Deterministic Downscaling method. Both models have 1/20° resolution and use the outputs from a Global Ocean Physics Analysis and Forecast model at 1/12° resolution available from Copernicus Marine Service (CMEMS). The LD20-NEMO uses only a 2D set of data from CMEMS as lateral boundary conditions. The LD20-SDD consumes the full 3D set of data from CMEMS and exploits the stochastic properties of these data to generate the downscaled field variables at higher resolution than the parent model. The skills of the three models, CMEMS, LD20-NEMO and LD20-SDD are assessed against remotely sensed and in-situ observations for the four-year period 2015-2018. All models show similar skills in reproducing temperature and salinity, however the SDD version performs slightly better than the NEMO version. This difference in resolution is particularly significant in simulation of vorticity and computation of the share of the sea occupied by highly non-linear processes. While the NEMO and SDD model show similar skill, the SDD model is more computationally efficient than the NEMO model by a large margin.
Introduction
To understand coastal ecosystems, fisheries, and local currents, scientists require high-resolution ocean models capable of resolving "mesoscale" features like eddies and fronts. However, increasing a model's resolution is traditionally computationally expensive; doubling resolution typically requires a ten-fold increase in computing power. Most current research relies on "dynamical downscaling," where a fine-scale model is nested inside a coarser global model to solve complex fluid dynamics equations across the entire grid. This paper explores a more cost-effective alternative: Stochastic-Deterministic Downscaling (SDD).
Methodology
The study compares two models at 1/20° resolution, both nested within a 1/12° CMEMS global parent model:
- LD20-NEMO (Deterministic): The standard approach using the NEMO engine to solve primitive equations. This requires a 96-core High-Performance Computing (HPC) cluster.
- LD20-SDD (Stochastic-Deterministic): A novel approach using a modified version of Objective Analysis. It treats small-scale fluctuations as a random process and uses spatial correlation functions to reconstruct fine-scale details from the coarser parent model's 3D data.
The SDD method assumes that at very small scales, ocean dynamics behave somewhat like turbulence, which can be modeled statistically. By calculating correlation lengths (as seen in the figure above), the model can extract "embryonic" features already present in coarse data and sharpen them.
Key Results and Insights
Comparable Accuracy with Massive Efficiency
Both models showed similar skills in reproducing temperature and salinity. Validation against 325 Argo float profiles and satellite data (OSTIA, GHRSST MUR) confirmed that both models perform well down to 2000m. Notably, the SDD model showed a slightly better bias than the NEMO version compared to observations. The most striking result, however, was efficiency: the SDD model is approximately 100 times more computationally efficient, running on a single-core Windows PC in roughly the same time the NEMO model took on 96 cores.
Resolving Vorticity and Non-Linearity
Vorticity—the local spinning motion of water—is a critical test for high-resolution models. The SDD model resolved higher values of vorticity and enstrophy than the parent model, identifying complex structures around the Lakshadweep islands that the coarse model missed.
The models identified that highly non-linear processes (Kibel number > 0.5) occupy about 20–25% of the Lakshadweep Sea in early spring, a detail often underestimated by coarser models.
Conclusion and Future Work
The SDD method represents a significant step toward democratizing high-resolution ocean modeling, making it accessible to researchers without supercomputer access. While the results are promising, the authors note that further validation with high-resolution current velocity data is needed to definitively prove which model is more realistic at the smallest scales. Future work will also explore variable-specific correlation lengths and test the method in different oceanic regimes, such as the Arctic or high-energy boundary currents.
Summary updated