You are here:
The impact of alternative nutrient kinetics and computational grid size on model predicted primary production and hypoxic area in the northern Gulf of Mexico
Citation:
Pauer, J., W. Melendez, T. Feist, J. Lehrter, B. Rashleigh, L. Lowe, AND Rick Greene. The impact of alternative nutrient kinetics and computational grid size on model predicted primary production and hypoxic area in the northern Gulf of Mexico. ENVIRONMENTAL MODELLING AND SOFTWARE. Elsevier Science Ltd, New York, NY, 26:1-13, (2020). https://doi.org/10.1016/j.envsoft.2020.104661
Impact/Purpose:
Mathematical models are increasingly being used to better understand coastal systems and to support policy and management decisions. However, with increases in model complexity (operating on increased number of grid cells and using multiple equations to describe chemistry, biology and transport), it is becoming more difficult to determine the accuracy and uncertainty associated with model predictions. Using a previously calibrated hypoxia model for the northern Gulf of Mexico we investigate model structure uncertainty, specifically how two phytoplankton nutrient limitation formulations (Monod versus Droop) and grid sizes (6km x 6km versus 2km x 2km) impact primary production and hypoxic area in the Gulf. Results show that both the selection of grid size and nutrient limitation formulation impact model prediction, especially the size of predicted hypoxic area which is a key indicator of riverine nutrient impact in the Gulf. For example, the same model on the 6km x 6km grid results in an average hypoxic area of almost 20% larger than running it on the 2km x 2km grid. This work highlights challenges of quantifying uncertainty of complex models in general, and recommends that improvements of such analyses should keep pace with the rate of model development.
Description:
An understanding of accuracy and uncertainty of model predictions is critical when models are used to support important policy and management decisions. The analyses behind these two concepts are challenging and time-consuming for complicated models containing multiple biogeochemical formulations and operating on high-resolution computational grids. For this reason, uncertainty analyses are seldom performed, especially methods to determine structural uncertainty. Here we use a previously-calibrated water quality model for the Gulf of Mexico to explore two features of model structural uncertainty: 1) how changes in the nutrient limitation formulation, Monod versus the Droop, and 2) effects of a change in grid size, 6km x 6km versus 2km x 2km, impact predicted production and hypoxic area in the Gulf of Mexico. Seasonal hypoxia in the northern Gulf of Mexico has been an environmental concern for many decades, and mathematical models have been used to better understand the issue. Running the same model on the 6km x 6km grid results in an average hypoxic area of almost 20% larger than running it on the 2km x 2km grid. However, predicted hypoxic area of the model running on the finer grid is more sensitive to changes (±25%) in river nutrient loading and settling rate and growth rate parameters than the courser grid. The model using the Monod nutrient formulation calculates higher primary production, especially over the summer months, but smaller summer average hypoxic area in comparison with the model with the Droop formulation. The Monod based model also predicts much higher primary production close to the river discharge locations (Mississippi and Atchafalaya rivers), while the Droop based model predicts higher primary production farther away from the rivers. The shifting of primary production when using the Droop formulation, probably explains the reason behind the larger predicted hypoxic area prediction when using Droop, although the Monod model generates higher primary production. This work highlights some of the challenges of quantifying uncertainty of complex models, and recommends that improvements of such analyses should keep pace with of the rate of model development.
URLs/Downloads:
DOI: The impact of alternative nutrient kinetics and computational grid size on model predicted primary production and hypoxic area in the northern Gulf of Mexico