2016 Summer Internship Program
Forecasting the Effects of Environmental Changes on Dungeness Crab
This summer I had the incredible opportunity of being an intern for JISAO at the campus of the University of Washington. I was privileged to work with Dr. Samantha Siedlecki, who is a biogeochemical oceanographer and researcher at JISAO. With the guidance of Dr. Siedlecki and many others at JISAO, I was able to use a Generalized Linear Model to forecast the distribution of Dungeness crab along the U.S. west coast.
I began the research by analyzing a historical data set of Dungeness megalopae provided by C. A. Morgan from surveys funded by NOAA Fisheries and the Bonneville Power Administration. This data set contained the latitudes and longitudes of the documented megalopae distribution from 1998 to 2015. The next step in the research was figuring out yearly environmental variables at each point in space where the megalopae were documented in order to create a model trained by composition of the ocean at each point in space.
The environmental variable information was provided by the JISAO Seasonal Coastal Ocean Prediction of the Ecosystem (J-SCOPE) climatology, hind-cast and forecast models. Initially, the environmental variables extracted were salinity, temperature, dissolved oxygen concentration, nitrate concentration, and phytoplankton concentration. Learning how to extract these variables was a rewarding process as it involved thoughtful coding. The variables were extracted only over the dates what megalopae were documented, May and June. They also were extracted over depths of 30 meters as suggested by Jenifer Fisher from Oregon State University, an individual with expert knowledge in the biological distribution of Dungeness megalopae. We collaborated with many other scientists during the research and received lots of important help while setting up the data.
In addition to salinity, temperature, oxygen concentration, nitrate concentration, and phytoplankton concentration, two more variables were derived. A temperature gradient was computed as a variable from surface temperature in order to recognize how sensitive megalopae are to fronts. The other variable derived was the percent of the water column where the aragonite saturation was less than one because it is believed that megalopae are sensitive to waters where the aragonite saturation is less than one. This variable was calculated with an empirical relationship using temperature and oxygen fields (Alin et al. 2012).
After isolating the now seven variables to same points where megalopae were documented, I used a Principal Component Analysis (PCA) to see if any of the variables were significantly influencing each other. Using this approach, I found that each variable was individually important to model and had to compose combinations of variables to statistically test in the model. With help from my mentor and collaboration with Isaac Kaplan, I was able to figure out the best combination out of the ones we tested to fit in the Generalized Linear Model (GLM). The GLM was trained off of the environmental variables and the historical megalopae data. By some statistical magic, the GLM detects the relationship between the variables and the megalopae data and creates an equation that forecasts megalopae presence or absence based on only on the environmental variables. I was super excited to reach this point in the research and am so very grateful for my mentor, and everyone else who collaborated with this project. The approach we took in forecasting the megalopae distribution was fairly new and I’m excited to see how it is developed and used in the future.
Although I spent most of my time this summer working in Matlab, I definitely was able to enjoy the wonderments of Seattle during my stay. Some of my favorite activities were exploring the many small cafes around the area, visiting Pikes Place Market, and going to local concerts. Seattle is truly a unique city and has so much to offer in the summer. This summer I also made incredible friendships with the other interns and had a wonderful time getting to know everyone. I am so grateful to JISAO and my mentor Samantha Siedlecki for giving me such a valuable opportunity. This summer definitely reaffirmed my love for research and kindled a new interest in coding. It was an enormous honor to be an intern at JISAO and I would recommend anyone interested in applying to do so.
Click on image for full size PDF