Linear date estimates showed no clear change in the coloring-to-fall interval for 34 Harvard Forest plants, 1991–2004 versus 2010–2023

We tested whether the interval from 50% leaf coloring to 50% leaf fall changed at Harvard Forest. We used HF003 revision 37 and 34 plant IDs from 13 taxa. Each ID had at least ten paired years in both 1991–2004 and 2010–2023 and no recorded replacement note. Linear interpolation between field visits gave 935 paired plant-years. The equal taxon mean interval rose from 10.32 to 10.54 days. The change was +0.22 days, with a 95% range of −0.91 to +1.19 days from 4,999 shared three-year calendar block draws. Planned checks also left the change uncertain. The result does not establish stability. Mean observation brackets became narrower in the later period. If dates can vary independently anywhere within those brackets, the mean change can range from −14.50 to +15.10 days. This extreme envelope has no probability level and shows the added assumptions in date interpolation. A source audit found three reused IDs with replacement notes, which were excluded after the first estimate. The initial 37-ID result is preserved. The source also has invalid values, duplicate keys, and conflicting documentation for 1992. An independent CSV calculation matched all selected dates and main estimates. A clean run matched all 28 outputs exactly. The package contains source files, provenance, code, results and three figures. The scope is one selected plant panel; the data do not identify a cause or measure carbon uptake.

autumn phenology · leaf coloring · leaf fall · Harvard Forest · observation intervals · null result · reproducibility