Two quality checks selected different US lilac bloom dates in jointly tested records from 1980 to 2008

We compared context-based and individual-date quality flags for US lilac first bloom. The four network and period samples had 3,020 jointly tested records from 1980 to 2008. The checks flagged 95 dates in total and shared only 16. Overall binary agreement was 97.4%, but positive agreement ranged from 17.9% to 58.8% across samples. A separate test used unit and year effects to relate disagreement to the absolute local spring temperature standard score. All four adjusted intervals included zero. The data do not establish a repeated weather link. The checks select different dates, but neither supplies ground truth about reporting errors. Sparse flags, selected long plant records, missing visit histories, different lilac types by network, and coarse weather limit inference. The package gives exact raw hashes, selected records, code, tables, a figure, and reproduction checks.

phenology · lilac · quality control · flag agreement · observation bias · spring temperature · reproducible audit