Current and Past Research
The Value of MPAs for Stock Assessments
The overall aim of this thesis is to assess the degree to which having an No-take Marine Protection Area (NTMPA) increases the ability to estimate life-history parameters relevant to stock assessment models (i.e. growth, fecundity and natural mortality). MPAs are implemented on the basis of protecting habitats, increasing species diversity, conservation of biodiversity, improving fisheries yields and as a control of experiments. An argued research-related benefit is that because there is no fishing inside an MPA it may be be possible to directly estimate the previous mentioned parameters (Growth and fecundity are usually estimated from auxillary information and data typically collected for stock assessment purposes usually provide no information about natural mortality). The maximum improvement in estimability will be quantified in addition to the extent that this improvement is reduced or lost when there is spillover from the MPA. Draft of proposal is available: here.
Sub-kilometer length scales in coastal waters
Research focusing on patchiness (spatial variability) in biological (bioluminescence and fluorescence) and several physical parameters in coastal waters. Coastal systems are disproportionately large contributors to ocean productivity, thus identifying the scales of this patchiness is of critical importance to understanding coastal ecosystem dynamics. With fine scale data measured from an autonomous underwater vehicle (AUV) in the Mid-Atlantic Bight, Monterey Bay and San Luis Obispo Bay, my contribution and work focused on modeling and identifying sub-kilometer scales of variability from this data with geostatistical models, in particular variogram analyses. Results indicate the presence of scales ranging from 10’s to 100’s of meters and provide valuable insight into ecosystem dynamics, sampling design, and resource allocation for future studies. This work has been accepted for publication in Continental Shelf Research.
Discrete-time surival analysis with applications to educational data
Research focusing on formalizing and testing a semi-parametric discrete-time survival model. In such a model, the effects of time are fit with standard logistic regression techniques, but the effect of the covariates on the logit hazard (hazard probability) are assumed to be unknown and fit with non-parametric regression techniques (e.g. smoothing splines). Performance of semi-parametric models to fully parametric models was compared through rigorous simulations and model comparison metrics. In addition to discovering favorable results for the semi-parametric model, interesting connections were made between lifetable quantities and the fully parametric logistic regression model. This work has been submitted to the Journal of Educational Statistics.
Geostatistical models for assessing spatial and temporal rockfish abundance
Senior thesis investigating and learning statistical approaches to data with apparent spatial autocorrelation, with motivation to define a meaningful control area for the proposed (now implemented) Point Buchon marine protection area. (MPA) With several years of catch data, different Kriging models were explored and fit to assess potential control areas with similar abundance patterns as the protected area, by looking at seasonal patterns in the protected and surrounding areas. Although no decisive evidence was found for a particular control area, some very basic baseline assessments for the protected area were obtained.
Publications
Blackwell, S. M., M. A. Moline, A. Schaffner, T. Garrison and G. Chang. (2008) Sub-kilometer length scales in coastal waters. Continental Shelf Research, 28, 215-226