Computing the value of information in management of a stochastic resource

pdg-control

After reading Costello et. al. (2010), taking a look at value of information in a dynamic context. Consider the model of Sethi et. al. (2005) (Costello is part of et al); a dynamic fisheries model with stochastic growth (a la Reed, (1979)), stochastic measurement assessing the stock size, and stochastic implementation of the harvest. I first imagine the case where the real system is quite deterministic, say, with a tiny amount of growth stochasticity only. I solve this optimal harvest problem using SDP and compare it to three other scenarios (simulating 100 realizations of each).

  1. where our estimate of the growth stochasticity is much higher, reflecting our lack of knowledge about growth, (in our the computation of the optimal solution)
  2. where our estimate of both growth and stock measurement is higher
  3. where our estimate of all three is higher.

Compare this to solving the optimal solution under increasing uncertainty (that is, increasing the sources of stochasticity in the real system, and having the manager solve for the optimal solution given that uncertainty). Still feels like uncertainty should impact profits, but not much:

Even underestimating rather than overestimating the noise isn’t so costly, only impacts cases where the population crashes (i.e. the manager assumes a nearly deterministic system, and scenarios add additional stochasticity into the reality).

Similarly for an Allee effect model, overestimating uncertainty doesn’t result in very suboptimal (conservative) behavior…

Costello C, Rassweiler A, Siegel D, De Leo G, Micheli F and Rosenberg A (2010 2010). “Marine Reserves Special Feature: The value of spatial information in MPA network design.” Proceedings of the National Academy of Sciences, 107. ISSN 0027-8424, https://dx.doi.org/10.1073/pnas.0908057107

Sethi G, Costello C, Fisher A, Hanemann M and Karp L (2005). “Fishery management under multiple uncertainty.” Journal of Environmental Economics and Management, 50. ISSN 00950696, https://dx.doi.org/10.1016/j.jeem.2004.11.005

Reed W (1979). “Optimal escapement levels in stochastic and deterministic harvesting models.” Journal of Environmental Economics and Management, 6. ISSN 00950696, https://dx.doi.org/10.1016/0095-0696(79)90014-7

Evolution stuff

  • More writing on Wrightscape manuscript. Reply to Peter.
  • rfishbase new version to CRAN?
  • Create table 1 for treebase
  • Touch up pacakge documentation in treebase

Early warning

Large replicate sims with fine sampling – creating memory issues on nodes on farm. Too hot for SOCK, switching to MPI…

Package Documentation

Just as a test, I ran staticdocs build_package function on rfishbase, sticking the output into inst/doc/html (from the package root directory).

library(staticdocs)
build_package("rfishbase", "inst/doc/html")

I then created the gh-pages branch with just the html docs and pushed to github.

git branch gh-pages
git checkout gh-pages
rm -rf data/ man/ R/ README.md NEWS NAMESPACE DESCRIPTION demo/
mv inst/doc/html/* .
rm -rf inst/
git add *
git commit -a -m "gh-pages documentation"
git push origin gh-pages
git checkout master

Which means that the documentation now appears at: https://ropensci.github.com/rfishbase/

We can think about customizing the css to some ropensci standard if we like. Should motivate me to get some more pretty examples in the documentation too. I think there’s a way to get vignettes to show up as well, will dig around a bit.

Also created an index for all packages at https://ropensci.github.com