In 1953–2020, JMA stations permanently lost 16% of animal and 6% of plant phenology series (frog 64%), unrelated to nearby population; animal dates also became later than temperature predicts

Question: did animal records of the Japan Meteorological Agency (JMA) phenology network lose detection relative to plant records at the same stations, and does this loss explain the non-temperature delay in animal first dates? Data: JMA records (65 items, 102 stations, 1953-2020), GHCN-Daily and GHCN-M v4 station temperature, GHSL population within 5 km. We used 2,808 station-items with at least 3 records in 1953-1962. A station-year was active if the station recorded a mandatory plant item. Results: (1) The share of active station-years with a record rose for the 16 mandatory plant items (0.84 in the 1950s to 0.92 in the 2010s; +0.018 per decade, 95% station-bootstrap CI 0.012 to 0.025) and fell for the 11 mandatory animal items (0.87 to 0.76; -0.010, CI -0.018 to -0.002). (2) Permanent loss (no record in the last 5 or more active years of the station) affected 15.8% (CI 13.0-18.8) of mandatory animal series and 5.6% (3.0-8.9) of mandatory plant series; the station-paired difference was +0.12 (0.08-0.16). Tonosama frog first sighting was lost at 37 of 58 stations, firefly at 24 of 72 and Shiokara dragonfly at 17 of 71. (3) Optional items lost 60-62% of series for both plants and animals, mostly in the 1960s, so their loss is a programme change and not an animal signal. (4) Loss was not associated with log population within 5 km (log-odds 0.02 per 10-fold population, CI -0.66 to 0.73) or with station urban warming. (5) After control for 60-day preseason station temperature, mandatory animal first dates became later by 1.43 days/decade (1.16-1.70); mandatory plants -0.17 (-0.56 to 0.18). The animal-minus-plant difference was 1.1 to 1.8 days/decade in all 7 checks, including a cherry-flowering 'biological thermometer' in place of instrument temperature and 1988-2020 only. (6) Animal series that lost detection drifted more (+2.93 days/decade) than series with stable detection (+0.82, CI 0.50-1.13), but stable series still drifted later than plants (-0.23). Dates did not become later in the last 3-5 records before a permanent loss. Thus detection loss explains part, but not all, of the animal delay, and the loss is not predicted by population near the station. This reproduces the animal delay of SCI 7a2ca78c with a different temperature method. Limits: records do not separate true absence from no observation; most animal records moved to 'nearby' sites after 1987; the population measure is for 2020 at 1-km resolution.

phenology · Japan Meteorological Agency · detection · first sighting · frog · firefly · monitoring network · observation bias · climate change · urbanization · GHCN-Daily · GHSL