Spectral nudging in the Tropics

Abstract. Spectral nudging allows forcing a selected part of the spectrum of a model's solution with the equivalent part in a reference dataset, such as an analysis, reanalysis or another model. This constrains the evolution in certain scales, typically the synoptic ones, while allowing the others to evolve freely. In a limited area model (LAM) setting, spectral nudging is commonly used to impose the large-scale circulation in the interior of the domain, so that the high resolution features in the LAM's forecast are consistent with the global circulation patterns. In a previous study developed over a Mid-Latitude domain, we investigated two parameters of spectral nudging that are often overlooked despite having a significant impact on the model solution. First, the cut-off wave number, which is the parameter determining the scales that are nudged and has a critical impact on the spatial structure of the model solution. Second, the spin-up time, which is the time required to balance the nudging force with the model internal climate and roughly indicates the starting point from when the results of the simulation contain useful information. The question remains if our conclusions for Mid-Latitudes are applicable to other areas of the planet. Tropical Latitudes offer an interesting testbed as its atmospheric dynamics has unique characteristics with respect to that further North and yet it is the result of the same underlying physical principles. We study the impact of these two parameters in a domain centred in the Gulf of Mexico, with a particular aim to evaluate their performance related to hurricane modelling. We perform 4-day simulations along 6 monthly periods between 2010 and 2015, testing several spectral nudging configurations. Our results indicate that the optimal cut off wavenumber lies between 1000 Km and 1500 Km depending on the studied variable and that the spin-up time required is at least 72 h to 96 h, which is consistent with our previous work. We evaluate our findings in four hurricane cases, allowing for at least 96 h of spin-up time before the system becomes a tropical storm. Results confirm that the experiments with cut-off wavenumbers near the Rossby Radius of Deformation perform best. We also propose a novel approach in which a different cut-off wavenumber is used for each variable. Our tests in the hurricane cases show that the latter set up is able to outperform all of the other spectral nudging experiments when compared to observations.