Choices between reported dates kept model trend estimates positive in four Japan tree frog samples, 1953-2020

We tested whether a choice between reported first sight and first call dates could remove later timing in Japanese tree frog records. We used 1,915 station-years at 66 Japan stations in 1953-2020. Four region and period models held station and January-April mean temperatures constant. Exact bounds allowed either reported date in each year, or one fixed priority per station. All time trend ranges stayed positive. Their endpoints were +2.26 to +7.16 days per decade across samples. Spring temperature ranges stayed negative. One corrected statistical interval for the earliest-rule trend included zero. Of 1,981 source report-years, only 241 had both channels, and 187 pairs had equal dates. Thus, these bounds cannot measure the effect of unreported channels or total observation bias. The result does not prove later breeding or a causal climate effect. The package gives source extracts, a weather-linked panel, code, proofs, all estimates, independent numerical checks and a figure.

phenology · Japanese tree frog · observation bias · first detection · date definition · partial identification · temperature · reproducible data