elkhunter505
WKR
- Joined
- Jul 4, 2018
- Messages
- 570
Sorry, let me be more specific. The science surrounding how it impacts deer herds, how it spreads, and how to kill it is clear. We are still uncertain about the risk to humans because there hasn't been any evidence of spread yet.How exactly is the science on cwd very clear? My understanding is we are still are still trying to prove/disproves the theories on how tse’s work. Sounds like you are pretty up to date - do you have any good references to share?
When we think about how CWD moves across the landscape, they use the ecological diffusion model a lot of the time. That model basically quantifies how CWD spreads across different habitat types. In that sense, in heavily forested areas, it is harder for deer to move and congregate than it is in agricultural fields or other habitat types. So based on GPS telemetry data about the natural movement process, they can use habitat type data to estimate how fast CWD will spread and where it is most likely to spread in the future based on where we know that CWD is currently and how prevalent it is in those areas. The ecological diffusion models that have been published have been shown to fit the true CWD spread process very well; however, because of the scale of CWD, it is very computationally intensive to fit that model to a large dataset so they rely on simulation modeling to take models like that and make predictions into the future. In this sense, they rely on papers from the past that show true data in how CWD spreads across the landscape and then build simulation models off of those dynamics. They then add some uncertainty surrounding each parameter in those models and run the simulations 100s or 1000s of times to get a bunch of different scenarios of how CWD spreads under different scenarios. The hard part for the public and for the modelers is the models get so complicated when we are trying to make them as close to the true process as possible that they become a black box to many people including me sometimes. The models use differential equations and all sorts of other calculus, algebra, and statistics which in this sense is likely not understandable to the normal person that isn't trained in these methods.
For example, this paper, is unbelievably dense but showed that their model accurately predicted and forecasted the true process using data from over 100,000 whitetails that were harvested and tested (2562 positives) in Wisconsin between 2002 and 2014. Figure 6 is a bit weird to look at but it gives you a sense of the accuracy of their predictions and forecasts in comparison to the data that they have available. In this case, they built a mechanistic model, which basically means constraining certain portions of the model based on the ecological knowledge that we have, to fit their data. They proved that their model forecasted, aka predicted prevalence rates into the future, better than other models that did not contain mechanisms and just fit based on data alone. They created forecasts by simulating that model into the future based on how it fit previous data and found it close to the true process.
However, for the average person to look at this type of model and see that they are constraining the process through a mechanism can sometimes feel like they are making the model "do what they want" when what they are truly doing is using the knowledge we've gained in the past from previous studies to help the model find realistic solutions. An example of how data could give unrealistic solutions would be just random chance of the data that you get from collared animals on their survival. Many mammals experience the lowest survival rates when they're young and then that survival rate becomes better and better as they age. If you don't have a high enough sample size, the distribution of the ages of the animals you have collared may not be truly accurate of the population. In this sense, if you have few young animals and more old animals and more of the old animals die, a model with no mechanism is going to estimate a higher survival probability for the younger animals than the older animals even though we know that is likely not true. If you put a mechanism in place, it can allow your model to give you more realistic estimates with the data that you have because in general, we know that neonates are at the highest risk of death and older animals are at the lowest.
Edit for typo on the old vs young animal survival.