One of the major tasks for conservation and fishery biologists is the
prediction of future population dynamics from relatively short time se
ries of past data. Inspired by a recent paper of Alan Hastings and Kev
in Higgins (Science, 263 (1994) 1133-1136) on long-term transients in
realistic model ecosystems for spatially distributed populations, we e
xplored the problem of predictability in analogous systems of coupled
logistic equations. For such systems, with parameters in the 'chaotic
range' as studied by A. Hastings and K. Higgins, we show that such pre
dictions are at best extremely difficult. Consequences for conservatio
n and fishery are explored.