Adjusted tree death maps had positive correlations between adjacent census intervals in 20 m cells at Barro Colorado Island, 1982–2015

We tested whether local tree death patterns persisted between adjacent census intervals in the Barro Colorado Island 50-hectare plot, Panama. Eight censuses labelled 1982 to 2015 give 1,578,430 selected tree intervals and 174,849 deaths. Starting diameter was at least 1 cm. We adjusted death counts in 20 m cells for species, starting diameter class and 100 m location. The mean correlation between six pairs of adjacent adjusted maps was 0.253. Its 95% spatial block bootstrap interval was 0.231 to 0.275. Every pair had a positive correlation, from 0.228 to 0.295. Added follow-up duration control gave a mean of 0.249. The original alignment had the largest correlation among all 1,250 whole-map translations in each pair. The mean correlation was lower for trees at least 10 cm in diameter, but lower counts can contribute to this difference. The package has selected outcomes, source checks, all results, figures and reproduction code. These results show partial persistence of local departures after the tested controls in one forest. They do not identify a cause or validate a forecast.

tree mortality · Barro Colorado Island · spatial ecology · temporal persistence · forest census · reproducible analysis