Model sensitivity and robustness in the estimation of larval transport: A study of particle tracking parameters

• Sensitivity study of larval transport predictions to particle tracking parameters
• Predictions found to be sensitive to input parameters required for modeling
• Importance of parameter selection for biophysical models evaluated
• Methodology proposed to manage sensitivity and achieve model robustness
• Sensitivity found to be proportional to the strength of mixing in the system

Many marine organisms spend their early lives as planktonic larvae dispersed by ocean currents. Predictions of larval transport are important for a wide range of applications including the interpretation of population genetics, fisheries management, and the planning of no-take marine protected areas. A popular method for predicting larval transport is through the use of coupled ocean circulation and particle tracking models, termed “biophysical” models. Although much research has been done on the sensitivity and uncertainty of ocean circulation models, the sensitivity of particle tracking models for the assessment of larval transport has been largely overlooked. This study investigates the sensitivity of larval transport predictions to three input parameters universally required for particle tracking in biophysical models; namely the number of particles released, the particle release depth, and the particle tracking time. Using a three-dimensional biophysical model of the Southern California Bight, estimates of larval transport are quantified using a two-dimensional vertically-integrated particle density distribution (PDD) and the difference between PDDs is assessed using the fraction of unexplained variance (FUV). Overall, our study shows that larval transport predictions are sensitive to changes in all three input parameters and that the sensitivity is affected by the strength of mixing in the system. For the number of particles released, the FUV falls off rapidly as the number of particles increases. A minimum number of particles is identified that guarantees robustness of model predictions; this number increases as the complexity of the circulation patterns increases. For the particle release depth, the FUV between PDDs grew linearly as particles are released farther apart. The FUV is also inversely proportional to the strength of vertical mixing as the FUV is smaller in the winter when a deep mixed layer and weak stratification are present and larger in the summer when the system is strongly stratified. For the particle tracking time, the FUV between daily PDDs is much larger for short tracking times of 15 days or less than for longer tracking times of 20 days or more, showing a dependence on the length of time the particles take to be evenly mixed throughout the system. Our study quantifies the parameter sensitivity of larval transport predictions and presents a straightforward methodology to achieve robust predictions of larval transport from biophysical models.