Earlier tree deaths had a positive link to later small tree entry in 20 m cells at Barro Colorado Island, 1985–2015

Does local tree death help explain where new small trees enter the measured forest population in the next census interval? We used eight open tree censuses from one 50 ha plot at Barro Colorado Island, Panama. The main test has 6,250 records from 1,250 fixed 20 m cells and five interval pairs in 1985–2015. We related later entry of trees 10 to less than 20 mm in diameter to earlier death basal area of trees at least 100 mm in diameter. Controls included past recruitment, live small trees, live large-tree basal area, and block and period means. The death coefficient was 0.140 on log-transformed rates, with a nominal 95% spatial resample interval of 0.080 to 0.197. The positive link remained under all planned checks. Coefficients fitted on three earlier pairs reduced centered-log prediction error by 0.70% in two later pairs. The spatial interval was 0.15% to 1.37% when both models were refitted in each draw. The gain was small. Test block means were supplied, so this is a test of local spatial differences rather than total recruitment forecasts. The data show a conditional link in one forest. They do not identify a causal effect of tree death or measured canopy gaps. The package has source extracts, hashes, code, predictions, resample draws, figures and repeat instructions.

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