Primary production and summer hypoxia in Lynch Cove

Neil Banas, 2011

This diagnostic model relates marine circulation and terrestrial nutrient inputs to plankton biology and oxygen levels in Lynch Cove, at the tip of Hood Canal in Puget Sound, WA. This model is intermediate in complexity between, on the one hand, the simple box models and back-of-the-envelope calculations often used to diagnose hypoxia dynamics, and, on the other hand, detailed three-dimensional simulations. This model is one of several tools used by the Hood Canal Dissolved Oxygen Program Integrated Assessment and Modeling team to diagnose the contribution of terrestrial nutrient loading to summer hypoxia in Lynch Cove. Please see chapter 3.4 of the HCDOP final report or contact a member of the HCDOP IAM team before drawing conclusions from this model.

Further technical information is below. You can change model parameters with the sliders. The model grid cells and the dots in the inset stock-flux diagram are also clickable.



The model

This model is a simplified implementation of a new, tidally averaged model of estuarine circulation and nutrient cycling called tonicella. The model domain is two-dimensional, representing a slice down the main axis of Lynch Cove, horizontally coarse (8 boxes) but vertically well-resolved (40 layers). The seaward (left-hand) boundary condition is based on Twanoh ORCA data from summer 2005. In this application, the total volume flux associated with the two-layer estuarine exchange flow is treated as a free, user-specified parameter. In other words, in contrast to many oceanographic models, this model does not predict the strength of the circulation that supplies marine nutrients: instead it addresses the question, "Given a certain level of marine loading and terrestrial N input, what biochemical patterns would we expect, under conditions like those in summer 2005?"

A five compartment nutrient cycling model ("NPZDO": nitrate+ammonium, phytoplankton, zooplankton, detritus, and dissolved oxygen) is run in each cell of the model, with each compartment also subject to advection by the exchange flow and vertical mixing (with a constant vertical diffusivity). The NPZDO model is an extension of the NPZD model used by Banas et al. (2009) in a study of the Washington-Oregon coast: see that study for details. Default model parameters are the same as those used by Banas et al. (2009), with the following exceptions: the initial slope of the phytoplankton growth-light curve was set to 0.125 based on historical C14 uptake data from Puget Sound; the light attenuation coefficients for seawater and phytoplankton were set to 0.2 m^-1 and 0.03 m^-1 (mmol N/m3)^-1, based on a fit to Lynch Cove CTD data; and the half-saturation for nutrient uptake was set to a more standard value of 1 mmol/m3. This ecosystem model is currently being adapted for use throughout Puget Sound and the Salish Sea as part of the MoSSea project. It is also under continuing development based on a wealth of data from the Washington-Oregon outer coast as part of the PNWTOX project.


Comparison with observations

See the HCDOP final report, chapter 3.4, for a description of a tuning and validation experiment in which this model was systematically compared with summer 2005 observations in order to constrain our estimates of the relative contributions of terrestrial and marine nutrients to Lynch Cove. The data-model comparison, for a family of "best guess" model cases including the default parameter values in the visualization above, is summarized in this figure. Red lines = model cases with varying values of exchange-flow strength, detrital remineralization rate, and detrital sinking rate. Black dots = observations.

lynch cove model validation


The visualization

This online visualization is an adaptation of NPZmachine 3.4 by Neil Banas, a general, open-source tool for simulating and visualizing marine ecosystem dynamics. If you want to tinker and adapt this visualization for other purposes, start there. Here is the source code for this Lynch Cove application. It's built using Processing. It's actually more than a visualization: the model is actually running in your web broswer!