Time Series Sampling and Data Assimilation in a Simple Marine Ecosystem Model
Document Type
Article
Publication Date
1-1-1996
Description
Simulated distributions of nutrients, phytoplankton and zooplankton were obtained from a simple marine ecosystem model that included nutrient inputs from episodic events. These distributions were then used in numerical identical twin experiments to test the ability of an adjoint data assimilation method to recover rate parameters, such as population growth and death rates, component initial conditions, and the amplitude of episodic events. Data were assimilated into the marine ecosystem model at monthly, bi-weekly and weekly intervals over a period of about 2 months. The ability to recover rate parameters and component initial conditions was determined primarily by the frequency and type of data that were assimilated. Assimilation of data at monthly intervals proved to be adequate for recovery of most of the rate parameters and some of the initial conditions. Bi-weekly data yielded better recoveries; however, increasing the data availability to weekly intervals did not significantly improve the results relative to the bi-weekly cases. The ability to recover biological rates with only monthly data suggests that these are fundamental aspects of marine ecosystems and can be resolved with only a few measurements. The availability of zooplankton information, even at a reduced frequency relative to phytoplankton or nutrient information, improved the ability to recover rate parameters with data more widely spaced in time. Recovery of component initial conditions was related to the timescales of the biological processes; faster processes required more frequent data. The recovery of the amplitude of the episodic events was related to the timing of the sampling relative to the event, rather than to the frequency at which data were available. The number of iterations needed for convergence when using data assimilation with the marine ecosystem model was dependent not only on the frequency and type of the input data series, but also on the structure of the marine ecosystem model. These results have implications for designing sampling strategies for measurement programs, such as the U.S. Joint Global Ocean Flux Study Hawaii Ocean Time-series and Bermuda Atlantic Time-series sites, so that these multidisciplinary data sets can be used with data-assimilative marine ecosystem models.
Citation Information
Lawson, Linda M.; Hofmann, Eileen E.; and Spitz, Yvette H.. 1996. Time Series Sampling and Data Assimilation in a Simple Marine Ecosystem Model. Deep-Sea Research Part II: Topical Studies in Oceanography. Vol.43(2-3). 625-651. https://doi.org/10.1016/0967-0645(95)00096-8 ISSN: 0967-0645