Trend tests
Concern At Site 1 the share of pollution-sensitive groups is declining (p = 0.042) while the total score shows no clear trend.
A seasonal Kendall test compares each survey only with surveys from the same season in other years, so the spring-summer gap cannot pose as a trend. One value per season and year is used, and a trend is reported only with at least 10 same-season comparisons.
| Site | Measure | Surveys | Result | Evidence | Change per decade |
|---|---|---|---|---|---|
| Site 1 | Survey score | 27 | no clear trend | p = 0.24 | -1.7 points |
| Site 1 | Share of groups found that are pollution-sensitive | 27 | declining | p = 0.042 | -8 pts |
| Site 1 | Mayfly, stonefly and caddisfly groups found | 27 | declining | p = 0.092 | under 0.1 groups |
| Site 2 | Survey score | 16 | no clear trend | p = 0.63 | +0.6 points |
| Site 2 | Share of groups found that are pollution-sensitive | 16 | no clear trend | p = 0.29 | -7 pts |
| Site 2 | Mayfly, stonefly and caddisfly groups found | 16 | no clear trend | p = 1.00 | under 0.1 groups |
Mayfly, stonefly and caddisfly groups at Site 1 are also declining (p = 0.092). Site 2 has fewer surveys and shows no clear trend. No sudden shift stands out at either site: the strongest candidate at Site 1, around 2021, would arise by chance 39% of the time.
Results with p between 0.05 and 0.1 are a signal to watch rather than proof.
How this was worked out
Seasonal Kendall test (Hirsch, Slack and Smith 1982) on season-year medians; seasonal Sen slope; turning points by a permutation test of the strongest mean shift (999 shuffles).