Predicting bloom initiation on the Texas (USA) coast: Combining satellite imagery with an individual-based model Chapter uri icon


  • A key question when modeling harmful algal blooms is: where do blooms start? Here we combined a previously developed, spatially explicit individual-based model (IBM) for Karenia brevis with a satellite imagery ensemble model to identify patches of K. brevis that may serve as seed populations in the southern Gulf of Mexico. We focused on the region north of the Yucatan peninsula, Yucatan, Mexico to identify potential K. brevis populations, which were then used to seed the IBM and track the cells through time. Individuals were tracked from ~1 Jul through 31 Dec (~180 days) and the arrival of a sustained presence of cells (i.e. consecutive days with cells present) at the coast of Texas was compared with a 10-year time series of abundance captured with an Imaging FlowCytobot (IFCB) at Port Aransas, Texas, USA. The IFCB captures images of all phytoplankton cells (10-150 m) and has provided early warning for 8 harmful algal blooms, specifically Dinophysis ovum and K. brevis. Modeled cells were allowed to vertically migrate in the water column to maximize their access to light and nutrients, but there was no cell division and no cell death. Two different, highly simplified, nutrient fields were tested, one based on salinity (nutrient concentration increases with decreasing salinity) and one based on mixed layer depth (nutrient concentration increases with depth; MLD). To account for randomness associated with each individual cell, each year was tested with multiple model runs and the results averaged. In 2015, a bloom of K. brevis was identified at Port Aransas on 14 Sept. Retrospective analysis of the satellite imagery indicated the potential presence of a patch of K. brevis north of the Yucatan Peninsula on 12 July. Cells were seeded in this region of Mexico and tracked forward in time. The first sustained presence of modeled cells in the region near Port Aransas occurred between 7 - 18 Sept, dependent upon how the nutrient field was derived. Model runs utilizing the MLD-based nutrient field were better at capturing the timing of arrival while those utilizing the salinity based nutrient field appeared to better capture the bloom dynamics once the bloom had arrived. Results varied by year, with the starkest difference being 2010, when no large bloom of K. brevis was observed along the coast of Texas. Results from the model utilizing a salinity based nutrient field indicated a large bloom, with cells arriving in mid-September. In contrast, the results utilizing a MLD-based nutrient field indicated no cells until mid-November, well beyond the typical start period for a bloom in Texas (September-October). As the true nutrient field is much more complex than those used here, the incorporation of better nutrient fields could greatly improve future results and the potential for forecasting of blooms.

author list (cited authors)

  • Henrichs, D., Tomlinson, M., & Campbell, L.

complete list of authors

  • Henrichs, Darren||Tomlinson, Michelle||Campbell, Lisa

editor list (cited editors)

  • Hess, P.

Book Title

  • Proceedings of the 18th International Conference on Harmful Algae

publication date

  • January 2018