Adjacent spring-stage intervals had negative residual covariance in four US shrub samples in 1965-2008

We tested whether the leaf-to-first-bloom interval and the first-to-full-bloom interval had opposite annual deviations. Four separate East and West samples in 1965-1986 and 1987-2008 contained 13,320 plant-years. After plant, year, and March-April temperature adjustment, residual correlations ranged from -0.251 to -0.179. Each exploratory 95% resampling interval for covariance was below zero. A first-bloom date occurs with opposite signs in the two intervals. Under an independent date-error model, an error standard deviation of 2.42 to 3.19 days would remove the negative covariance at the point estimates. These values fit the residual variance limits. They do not measure actual error. Missing visit records, selected complete triples, mixed taxa, coarse weather, and model assumptions prevent a claim of biological compensation. The package gives a data audit, code, source hashes, an error-sensitivity method, and figures.

phenology · observation bias · shared date error · mathematical coupling · lilac · honeysuckle · regional replication