Missing Data Can Be A Pain

I spent most of this week balancing two major tasks. The first, which occupied less of my time, was to make a nice visual for a conference that the lab is attending next week. I won't be sharing that figure, since the information in that figure will likely change over time as we add new information to the model. However, it is certainly ready to go for the conference, and provides a nice understanding of the complexity of this Mississippi Sound and Bight model. 

The other task, which I haven't finished yet, is to work through some current problems with the model. As Kim explained, when you balance a model--work through the inputs and calculations at the start of the model--the program will alert you to errors. These errors may be a combination of factors, including more animals being removed from the model than actually exist in the model, animals having unreasonable ecotrophic efficiencies (the proportion of their biomass consumed by higher predators), or problems with animals' diets. The first error--more organisms removed than exist in the model--comes from a mismatch between the starting biomass and the removals from fishing that we simulate in the model. Since the starting biomasses we input to the model are based on landings (the fish catches that are brought onto the boats and tallied), the biomasses are not reflective of all the organisms in the model area, also known as the stock biomasses. 

For our model, we have a few mismatches that I am working through right now, but one of the challenges is that stock assessment data are harder to come by than landings data. Landings data are pretty well consolidated by the appropriate state or national agency, while stock assessments are usually gray literature--information published outside of traditional journal publishing. Some of these reports are 300-400 pages long, and I only need a few sections of information from these stock assessments for our work. Luckily, Excel can extract data from PDF tables without having to type each entry by hand. For some situations, though, the data are not published as tables, but instead as figures. Here, I can use some visual tools that help extract data from figures. While these tools are not perfect, they can provide good estimates of the actual data, useful in this situation when perfection isn't necessarily the ultimate goal. Here, you can see the open circles that I have identified as data points I'd like to extract, and the program uses visual cues and exact data points (the x- and y-axis marks you can see I've added in blue and red) to determine the x- and y-coordinates of the data points. 

So what's next? Well, now that I have the data, I need to determine the units that the authors of each report used and how to convert these values to usable data points for our model. This process will likely take time as each species may require unique calculations. Don't expect this to be the focus of next week's blog, because next week will be all about the conference we are attending here in Biloxi. Until next time!

---