Examining the effect of salinity on dolphin mortality using Lagrangian particle tracking in a hydrodynamic model

Abstract

Every year, many dolphins are found dead on the beaches and waterways of the Gulf of Mexico and Mississippi Sound. In 2019, an unusual mortality event (UME) happened, where 337 bottlenose dolphins were stranded between Louisiana, Mississippi, Alabama, and Florida. The National Oceanic and Atmospheric Administration (NOAA) determined that the cause of this UME was protracted exposure to low salinity waters, based on observations of skin lesions and the environmental conditions during that period of time. It is often unclear where dolphins initially died, as their carcasses are found stranded on beaches days after they die. To investigate this further, we used a hydrodynamic model (EFDC+) of the Mississippi Sound. Our goal was to track dolphin carcass movement and simulate salinity at the time and location of each predicted dolphin's death. I represented the movement of 19 dolphin carcasses using particles within the lagrangian particle tracking (LPT) module. The results enabled us to predict the original location of death for each dolphin. The average simulated salinity of all the dolphins' most probable original place of death was below five, except for two cases. I compared these results to the salinity of the Mississippi Sound during Bonnet Carré spillway opening and non-opening dates. Our findings highlight the significant impact of the spillway's opening on the reduction of salinity and its association with dolphin mortality. Furthermore, I calibrated our model simulating the movement of dead dolphins using data from GPS-tagged turtle carcasses and wooden effigies in April 2017 collected by NOAA in the same area as our study domain. Firstly, this data was used to conduct a sensitivity analysis on the previously configured LPT model, which involved altering certain parameters. The results from sensitivity analysis were then used to perform a calibration by adjusting the most sensitive parameters to get the best match between the modeled and observed trajectories. The analysis found that adding wind drag to the LPT model significantly improved its predictive capabilities. The study also investigated the effect of transitioning from a two-dimensional (2D) to a three-dimensional (3D) model. The results revealed that in our study domain with smooth bathymetry a 2D model is good enough and more efficient for modeling purposes.