One Polish dataset causes the within-series rise of masting variability in MASTREE+ v2; 101 other studies show no detectable CVp trend and no consistent link to summer warming

Question: Pearse et al. (2017) found that the interannual variability of seed production (CVp) increased in 1900-2014. Their main model compares different records across 20-year periods. Studies of European beech found that CVp decreased where summers became warmer. We tested two points in MASTREE+ v2 (2024 release). Does CVp change inside the same series? Does that change follow local summer warming? Approach: We used 689 continuous seed, fruit, cone and flower series (102 studies, 264 taxa). We calculated CVp in 10-year windows (8 or more years) and fitted series fixed-effect models. CIs come from a bootstrap that resamples studies. Checks: study weighting, leave-one-study-out, random-effects meta-analysis, windows shifted by 5 years, and two temperature products (GHCN-CAMS and CRU TS 4.10). Results: The pooled within-series trend is +4.1% per decade (95% CI -1.8 to +5.5). One dataset causes most of it: D00013, Polish State Forests visual scores of the percentage of trees reproducing (94 series, 28% of windows; +6.3% per decade). In D00013, mean reproduction also increased, and with log mean as a covariate its CVp trend is about zero. Without D00013, the trend is +1.2% per decade (CI -2.7 to +4.7). With equal study weights, it is -0.5% (CI -4.2 to +3.5). A random-effects meta-analysis of 48 studies gives -1.0% (CI -4.8 to +3.1), with high heterogeneity (I2 = 0.80). A Pearse-style species-centred period model gives +7.2% per 20 years (p = 0.047) with D00013 and +3.1% (p = 0.08) without it. In models with series and decade fixed effects, local summer temperature does not predict lower CVp. Of 16 estimates (2 products, 2 window alignments, 4 samples), 14 are positive and 2 are negative. None has a CI below zero, and the three with a lower bound at or above zero depend on weighting or window alignment. Subgroup trends, for example seed traps and Fagaceae, change with the window alignment. Limitations: CVp from short windows is noisy. We did not correct for count sampling noise, trap-number changes or stand age. The temperature data are 0.5-degree grid cells. D00013 is not in the 2017 compilation, so this study does not re-test the original estimate. It shows that in MASTREE+ v2 a general CVp trend over time is not robust.

mast seeding · masting · seed production · coefficient of variation · MASTREE+ · climate change · temporal trend · fixed effects · meta-analysis · data influence · European beech · GHCN-CAMS · CRU TS