Demographic Noise

Crowley Model

  • Exploring the extent to which demographic stochasticity in a key class can reverberate through the entire population dynamics.
  • Starting with a simple test case of a modified Crowley model:
\begin{align} \dot x &= b_1 x (K - x - y) - d_1 x + c_1 x y \\ \dot y &= b_2 y (K - x - y) - d_1 y - c_2 x y \end{align}
\begin{align} \dot x &= b_1 x (K - x - y) - d_1 x + c_1 x y \\ \dot y &= b_2 y (K - x - y) - d_1 y - c_2 x y \end{align}
  • Model is implemented on a branch of the warningSignals package. Model in crowley.c, Depends on modified gillespie.c library in this package.
  • Branching necessary as implementation simplifies the gillespie function dependencies, but these changes would break compatibility with warning_signals.c and its R functions. Once warning_signals.c is updated to the new interface format I can remerge the branches.
  • compile with make crowley for the moment, will add R interface soon.
  • Branches need to be pushed to the github server separately, see the github directions, but essentially
 git push origin branchname

will do the trick.


#### Parameters for small class

  • Values: b_1 = 0.11, b_2 = 0.6, d_1 = d_2 = c_1 = 0.1, c_2 = 4, K = 10000.
  • Expected sizes: E(x) = 500, E(y) = 4500, (analytic)
  • Variation (from simulation) SD(x) ≈ 120, SD(y) ≈ 1000
  • For K = 1000, means are 1/10th and extinction very common within 500 time units, and possible in either group. Extinctions of smaller population not uncommon for K = 10000 in 5000 time units.
  • Freezing dynamics of x (i.e. b_1 = .00, d_1 = .00, c_1 = .00) greatly decreases fluctuations in Y (more than an order of magnitude, SD(y) ≈ 70).
  • Analytic description should be straight-forward linear noise approximation.