Excess tree deaths in 20 m cells at Barro Colorado Island were positively correlated across adjacent census intervals after species, size and 100 m block control, 1985–2015

We tested whether the same small forest cells had excess tree deaths in later census intervals. We used the Barro Colorado Island 50 ha plot in Panama. The primary data have 1,343,579 tree intervals and 154,989 deaths in six census intervals from 1985–1990 through 2010–2015. For each interval, we calculated cell deaths above the exact expected count with species, starting diameter class and 100 m block held fixed. We divided each excess count by its conditional standard deviation. The mean same-cell correlation across five adjacent map pairs was 0.254. A nominal 95% spatial-block range was 0.232 to 0.276. All five pair correlations were positive, from 0.228 to 0.295. Mean correlation was 0.113 for intervals with midpoints ten years apart. The adjacent-map result remained positive with other cell sizes, added follow-up duration groups, finer size classes and missing-diameter records. For trees at least 10 cm in diameter, the mean was 0.048; event noise differs between size selections. This is a descriptive result for recorded deaths in one forest. The expected counts use each interval’s observed death totals, so this is not a forecast. The block ranges assume independent exchangeable spatial blocks and omit other measurement error. The result does not identify a cause or establish persistent individual risk. The package has source provenance, selected data, code, all results, figures, source checks and repeat instructions.

forest dynamics · tree mortality · spatial persistence · Barro Colorado Island · open data · reproducibility