JMA animal first-detection dates in 1953–2020 were later next to station-wide gaps in animal records, unlike plant dates, but these gap terms did not explain the long-term animal delay

Question: do Japan Meteorological Agency (JMA) animal first-detection dates show a year-scale signature of detection effort, and does it explain their known delay beyond temperature? Data: 140,306 JMA records in 2,965 station series (24 animal first-song or first-sighting items, 41 plant items, 101 stations, 1953-2020) parsed from the JMA PDF tables, and GHCNm v4 monthly temperature at the same stations. Method: within-series models with series and item-year fixed effects and item-specific effects of 60-day preseason temperature, plus two detection terms: the same series had no record in the year before or after (own gap), and the share of other same-group series at the station with no record in both adjacent years (station-wide lapse). Standard errors are clustered by station. Results: a full station-wide lapse came with animal dates 5.85 days later (95% CI 4.14 to 7.56; 5.45 to 7.30 in five specifications) but not with later plant dates (-0.46, CI -2.36 to 1.44; all plant CIs include zero). The CI was above zero in 13 of 22 animal items and 2 of 33 plant items. An own gap came with dates 1.78 days later for animals (1.37 to 2.19) and 0.50 days for plants. These terms did not explain the long-term delay: the animal residual trend was +1.51 days/decade (SE 0.11) without them and +1.59 (0.10) with them, because gaps became less frequent after the 1970s. Plants: +0.17 and +0.18. Items with more non-detection had a larger residual delay for animals (Spearman rho 0.73, 22 items) but not plants (0.13). Dates did not become later in the 5 years before 43 station closures (-0.38 days, SE 0.50) or before 134 animal series ended. Station urban warming did not predict the residual trend. Conclusion: observer effort affects JMA animal first-detection dates at the scale of single years, so analyses should include an effort term. But record-gap measures of effort do not explain the delay of about 1.5 days/decade. Limits: gaps are proxies; records do not separate absence from no search; linear models with one monthly temperature window; per-item counts are not corrected for multiple tests.

phenology · first detection date · detection bias · observer effort · Japan Meteorological Agency · animal phenology · climate change · fixed effects · GHCNm · monitoring networks