A Week For Model Upgrades

I spent most of my time this week making updates or upgrades - if you will - to two different ecosystem models. To prepare our MissDelta model for the first model scenario, Kim and I have decided to make some small adjustments to better reflect data availability. For instance, in our previous work, we started our model in 2000 even though the Mississippi Department of Marine Resources' oyster monitoring dataset wasn't robust until 2006. We back-calculated 2000 data from the 2006 data for our previous work and that model still showed strong replicability of trends in oyster biomass through time. Now we have decided to shift our model to start in 2006 to better align with these oyster data, but also to start after Hurricane Katrina, since we haven't seen any hurricanes as extreme in recent years, and we have 20 years of data to use for our model. For readers who are not ecosystem modelers: this model does not attempt to recreate the past. We provide a start value for the biomass of all of our organisms and then a time series for as many groups as possible, and we calibrate our model to the time series. This means that we make small adjustments to other values which help align the model's predictions with the observed values. A helpful way to think of the modeling process is that our model performs best when it is recreating past behaviors. For instance, if we have years of data collected during a flood event to calibrate the model, the model is likely to be able to replicate changes in biomasses if we give it a future flood scenario. Alternatively, if we calibrate the model on data where there was never a very hot summer, then the model will likely struggle to replicate behaviors of model groups during intense heat waves in future scenarios. Hence, why having 20 years of data provides a strong baseline on which to calibrate our model. 

At the same time, I have been working to upgrade my microplastics model, which still needs the same calibration process, but I am not trying to replicate trends caused by microplastics pollution. Instead, I am still working to replicate trends in model group biomasses, which will help explain how future microplastics inputs to the model area will be transferred from one model group to another through predation. Refining this model is slightly more challenging, since I am working with a subset of model groups from the larger model and since I am working with a number of model groups that do not have validated time series for their biomasses. With fewer time series, I have less data on which to calibrate my model, though I have decided to include a few extra model groups with time series to make things work. Maybe if someone regularly collected nematode data and published it, we could make better models that include our tiny animals near the base of food webs, but alas.

Stay tuned, because over the next few weeks I will provide more updates on how these models are getting on, since I have some important (and exciting) upcoming deadlines for these tasks.