WSPS1 Reproducibility Package: Climate, not invasion, is the dominant driver of native wasp community change

Reproducibility package for manuscript "Climate, not invasion, is the dominant driver of native wasp community change: a multi-method framework for managing honeybee predation under global warming" submitted to Journal of Applied Ecology, 2026. Authors: Sung-Hyun Min, Kyeong-Yong Lee, Su-Bae Kim, Sang-Mi Han The package contains a single-file standalone Python script (reproduce.py) that loads three site-week capture CSVs (Forest reference site, Apis mellifera apiary, Apis cerana apiary; 635 weekly observations across 19 site-year clusters, 10,441 total Vespa captures, 2019-2025) and reproduces all primary statistical results: species capture totals, Kruskal-Wallis test among sites, GAM thermal niche parameters (T_opt, deviance explained), and GEE Poisson regression with AR(1) and exchangeable working correlation structures, including a late-summer (Aug-Sep) seasonal sensitivity analysis. The frozen reference values JSON enables automated PASS/FAIL verification (11/11 checks pass). Version 3.0.0 changelog: realigned the GEE specification in reproduce.py to the original analysis used in the manuscript. Earlier releases (v1.0.0, v2.0.0) included a numeric site covariate that produced output values differing from the manuscript by approximately ±0.01 IRR. The original specification is the no-site formulation: log(E[Y]) = beta0 + beta1*temp_z + beta2*humid_z + beta3*wind_z + beta4*V_velutina_z, with site-level heterogeneity absorbed by cluster-robust sandwich standard errors via groups = site-by-year. ICC(temperature) = 0.012-0.017 (98.3-98.8% temporal variance) supports the pooled approach. Under v3.0.0 the script output reproduces every value in frozen_reference_values.json exactly. Data and code released under CC BY 4.0. Funding: Rural Development Administration, Republic of Korea, Project No. RS-2023-00232224.