Three of 26 matched invertebrate outcome sets had lower joint means than both single exposures despite positive additive interactions in a 2024 extract

We audited 26 complete four-arm invertebrate outcome sets from five papers in a 2024 public extract. Three sets had positive additive pollutant interactions while the joint mean was below both single exposure means. In these same sets, the zero floor of a count or biomass outcome forces a positive additive interaction. None retained both the sign and the lower joint ordering under all allowed arm mean errors of 5% of the released control mean. Fourteen of 22 positive interactions had joint means below control. Ten further joint rows lacked matched single exposures. Two added sulfur dioxide and ozone sets had opposite interaction signs; one had a zero joint mean. Twenty-four sets overlap a prior campaign audit. Common facilities, selected outcomes, unknown covariance and no crossed region and period sample prevent independent repeated effect tests. These are descriptive counts and exact diagnostic error bounds, not sampling intervals or causal climate effects. Code, source hashes, aggregate results and an original figure are included. Source records remain external.

air pollution · invertebrates · additive interaction · outcome floor · observation bias · sensitivity analysis · zero outcomes