A 1998-2020 farmland-versus-forest heat contrast in US NestWatch nest success did not repeat in 2021-2025 nests and depended on the habitat code and heat measure

Lauck et al. (2023) reported that hot weather during nesting decreased NestWatch nest success in agricultural land and increased it in forest in 1998-2020. We rebuilt their variables from NestWatch Open Dataset version 6 and gridMET: success means at least one fledgling, land use comes from the habitat code at the nest, and heat is the 46-day maximum Tmax z-score against 1980-2000. We fitted logistic models with species, year and region fixed effects. We then applied the same method to 76,398 nests from 2021-2025 that the original study did not use. In 160,946 nests from 1998-2020, we found the published pattern: the agriculture slope was -0.070 logit per SD (SE 0.032), the forest slope +0.046 (SE 0.021) and the contrast -0.116 (SE 0.039). The contrast stayed negative with region-by-year, site and observer controls, nest-discovery filters, outcome-code filters and location-precision filters. In 2021-2025, the contrast was +0.085 (SE 0.080). The difference between periods was +0.201 (SE 0.089, p = 0.023). The contrast was negative in all five earlier year blocks and positive in both later blocks. Within 1998-2020, the contrast was absent with the 100-m habitat code (-0.014, SE 0.031) or with the 46-day mean temperature (-0.026, SE 0.039). It was mostly western (-0.295 west of 100 W, -0.062 east). Developed sites had negative heat slopes in both periods. Limitations: the forward period is short and had 31% power for a significant negative contrast of the earlier size. We used fixed-effect models, not the authors' GLMM, and did not add land-cover proportion covariates. NestWatch lay dates are often estimated, and the outcome does not use exposure days. NestWatch data are CC BY-NC and are not bundled; code, aggregate results, hashes and a retrieval script are included.