Insights on multivariate updates of physical and biogeochemical ocean variables using an Ensemble Kalman Filter and an idealized model of upwelling. (June 2018)

• Record Type: Journal Article
• Authors: Yu, Liuqian; Fennel, Katja; Bertino, Laurent; Gharamti, Mohamad El; Thompson, Keith R.
• Is Part Of: Ocean modelling. Volume 126(2018)
• Page Start: 13
• Page End: 28
• Publication Date: 2018-06
• Subjects: Ensemble Kalman Filter -- Multivariate updates -- Coupled physical–biogeochemical ocean model; Oceanography -- Periodicals; Océanographie -- Périodiques; Oceanography; Periodicals; 551.46
• DOI: 10.1016/j.ocemod.2018.04.005
• Languages: English
• ISSNs: 1463-5003

Abstract

Highlights: Multivariate updates of physical–biogeochemical fields outperform isolated updates. Success depends on close correlation between physical and biogeochemical properties. Multivariate updates can preserve correlation between density and nutrient properties. Abstract: Effective data assimilation methods for incorporating observations into marine biogeochemical models are required to improve hindcasts, nowcasts and forecasts of the ocean's biogeochemical state. Recent assimilation efforts have shown that updating model physics alone can degrade biogeochemical fields while only updating biogeochemical variables may not improve a model's predictive skill when the physical fields are inaccurate. Here we systematically investigate whether multivariate updates of physical and biogeochemical model states are superior to only updating either physical or biogeochemical variables. We conducted a series of twin experiments in an idealized ocean channel that experiences wind-driven upwelling. The forecast model was forced with biased wind stress and perturbed biogeochemical model parameters compared to the model run representing the "truth". Taking advantage of the multivariate nature of the deterministic Ensemble Kalman Filter (DEnKF), we assimilated different combinations of synthetic physical (sea surface height, sea surface temperature and temperature profiles) and biogeochemical (surface chlorophyll and nitrate profiles) observations. We show that when biogeochemical and physical properties are highly correlated (e.g., thermocline and nutricline), multivariate updates of both are essential for improving model skill and can be accomplished by assimilating either physical (e.g., temperature profiles) or biogeochemical (e.g., nutrient profiles) observations. In our idealized domain, the improvement is largely due to a better representation of nutrient upwelling, which results in a more accurate nutrient input into the euphotic zone. In contrast, assimilating surface chlorophyll improves the model state only slightly, because surface chlorophyll contains little information about the vertical density structure. We also show that a degradation of the correlation between observed subsurface temperature and nutrient fields, which has been an issue in several previous assimilation studies, can be reduced by multivariate updates of physical and biogeochemical fields.