Warm half-years did not make ammonia associations with EPT family richness stronger within 1,418 English river sites in 1989-2017; an autumn between-site term did not survive site controls

We tested if warmth makes the association between ammonia pollution and sensitive river macroinvertebrates stronger. We used the ChemPop data product: 37,484 spring and autumn samples from 1,418 sites in 7 English regions, 1989-2017, with 6-month chemistry and CHESS-met air temperature before each sample. The response was the number of Ephemeroptera, Plecoptera and Trichoptera (EPT) families. Models had site, region x year x month and method fixed effects. We fitted 28 region x season x period strata and pooled them. The primary within-site test was null. A higher-than-usual ammonia level did not have a steeper EPT deficit in warmer-than-usual periods: +0.035 families per log10(mg/l) per 1 C (95% CI -0.091 to +0.161). Between sites, autumn samples showed fewer EPT families at high-ammonia sites after regionally warm half-years: -0.27 (CI -0.49 to -0.05; 11 of 14 strata negative). Spring samples did not (+0.03). The autumn value was more negative than all 28 shifted-year placebos. But warmth x site-descriptor terms (altitude, northing, urban and arable cover) reduced it to -0.07 (CI -0.24 to +0.10), and BOD gave the same pattern as ammonia. Low-flow terms did not change it. At most, the data show a weak warm-summer sensitivity of degraded lowland or urban sites, about 1% of mean EPT richness. They do not isolate an ammonia effect. Limitations: air temperature is a proxy for water temperature, 6-month aggregates miss short peaks, ammonia censoring rises to 50-70% after 2004, and the robustness models were added after the first fit. The package includes code, source hashes, derived sample tables, all stratum estimates and placebo results.

freshwater macroinvertebrates · ammonia · temperature · pollution-climate interaction · EPT richness · English rivers · ChemPop · multiple stressors · fixed effects · placebo test · observation bias · censored water quality data