Poisson point process simulation of spatial population dynamics

This repository contains source code (R/Rcpp) required to run spatial population dynamics simualation.

Installation

Windows

• Install R (R Installer) - latest version used is 3.6.2
• Install Rtools (Rtools Installer)
• Install required R packages from CRAN (devtools, Rcpp, BH)
• Install this source package using devtools

devtools::install_github("YegorGalkin/RcppSim")

Linux \ Ubuntu

TODO: like windows but without Rtools

MacOS

TODO: Like Linux

Simulator usage (1d)

1. Prepare named list of simulation options. Parameters:
   • area_length_x - area size
   • cell_count_x - used to split simulation interval into subintervals for filtering negliable interactions.
   • b - poisson birth rate
   • d - poisson death rate
   • dd - competitive death rate
   • seed - RNG seed
   • init_density - per unit starting density of speciments
   • death_kernel_r - max death kernel interaction radius
   • death_kernel_y - death kernel values on uniform grid on [0, death_kernel_r] interval
   • birth_kernel_r - max birth kernel interaction radius
   • birth_kernel_y - birth kernel values on uniform grid on [0, birth_kernel_r] interval
   • spline_precision - tries to keep spline approximation precise up to this value (for numerical integration and cdf inversal)
2. Run simulation for specified time or number of events

   sim$run_events(1e9)
   sim$run_for(100)

3. Extract and visualize results (using packages spatstat and tidyverse)

   sim$time
   sim$total_population
   sim$get_all_coordinates()
   
   points <- ppp(sim$get_all_coordinates(),
                 rep(0,sim$total_population),
                 c(0,sim$area_length_x),
                 c(-sim$area_length_x/2,sim$area_length_x/2)
                 )
                 
   pcf_grid = seq(0,1,length.out = 1001)    
   
   K_estimate <- Kest(points,r=pcf_grid,correction="Ripley")
   
   pcf_estimate <- data.frame(Kest=K_estimate$iso/2,x=pcf_grid)%>%
                   mutate(pfc=(Kest-lag(Kest))/(pcf_grid-lag(pcf_grid))/sim$area_length_x)%>%
                   pull(pfc)
                   
   ggplot(data=data.frame(x=pcf_grid,y=pcf_estimate),aes(x=x,y=y))+
   geom_point()

4. Check examples in this repository (tests, examples)