We studied the population dynamics of the plant at four sites at annual censuses, 1982-1987, during the rainy season (June-October) to determine spatial and temporal variability in vital rates and in sensitivity of population growth. Results of general interest included that among the 16 population projection matrices constructed for the 4 year × 4 site combinations, the ones with the highest growth rates had higher sensitivity to transitions of early stages. Matrix elements did not vary independently, but instead elements were correlated; for example environmental states that were good for growth were also good for reproduction. There was a negative correlation between elasticity and variability of matrix elements. We also found that el Niño years were favorable for recruitment of seedlings from a dormant seed bank; with 10-fold higher seedling emergence than other years, a surprising inland impact of global oceanic climate patterns. Of most general importance for understanding the long term demographic impacts of events that affect single stages in particular years, our study showed that elasticity structure at a site and year typical of high population growth was quite distinct from elasticity structure of other site/year combinations. This suggested that events (i.e., plant-animal interactions) in one year (or one state of the habitat) could be distinct in their contributions to long term population growth from similar events in other years (or other states of the habitat).