inst/examples/fishcollapse_knit_.md

In cboettig/warningsignals: Detection of Early Warning Signals for Catastrophic Bifurcations

Warning Signals in Fish Collapse

• Author Carl Boettiger [e-mail]
• Date: 10 Nov 2011
• Description: an example trying to detect early warning signals in data from fisheries collapses

Set up markdown format and image uploads.

Load required libraries

Load the data

Visualize data

Compute some indicators

Define some indicators <!--begin.rcode window_var <- function(X, windowsize=(length(X)/2)){ out <- sapply(0:(length(X)-windowsize), function(i){ var(X[(i+1):(i+windowsize)]) }) c(rep(NA, length(X)-length(out)), out) }

window_autocorr <- function(X, windowsize=(length(X)/2)){ out <-sapply(0:(length(X)-windowsize), function(i) acf(X[(i+1):(i+windowsize)], lag.max=1, plot=F)$acf[2]) c(rep(NA, length(X)-length(out)), out) } end.rcode-->

Reformat the data, uses data.table to perform computations over species <!--begin.rcode require(data.table) fish <- data.table(subset(dat_scotia, Year < 1992)) tmp <- data.frame(species = fish$variable, Year = fish$Year, Stock = fish$value, variance = fish[, window_var(value), by="variable"]$V1, acor = fish[, window_autocorr(value), by="variable"]$V1)

dat <- melt(tmp, id=c("Year", "species")) end.rcode-->

Cod are approaching a crash, but lobster are going strong, but both seem to show the same pattern.

Note the indicator patterns vary widly and rather arbitrarily among species

cboettig/warningsignals documentation built on May 13, 2019, 2:12 p.m.