Supplemental information 2: Candidate negative binomial models considered in the analysis of beaver dam abundance in six ecoregions of Quebec. All models included the log of stream length as an offset and random intercept for each ecoregion.

Models / Fixed effects / Random slopes
Null / Intercept
Dam building / Alder cover + Gradient+ Clay
Dam building – Clay random slopes / Alder cover + Gradient+ Clay / Clay
Dam building – Alder cover random slope / Alder cover + Gradient+ Clay / Alder cover
Dam building – Gradient random slope / Alder cover + Gradient+ Clay / Gradient
Food availability / Hardwood cover + Non-forested cover
Food availability – Non-forested cover random slopes / Hardwood cover + Non-forested cover / Non-forested cover
Food availability – Hardwood cover random slopes / Hardwood cover + Non-forested cover / Hardwood cover
Human disturbance / Road intersection
Human disturbance – Road intersection random slopes / Road intersection / Road intersection
Global / Alder cover + Gradient+ Clay + Hardwood cover + Non-Forested cover + Road intersection
Global – Alder cover random slopes / Alder cover + Gradient+ Clay + Hardwood cover + Non-Forested cover + Road intersection / Alder cover
Global – Gradient random slopes / Alder cover + Gradient+ Clay + Hardwood cover + Non-Forested cover + Road intersection / Gradient
Global – Hardwood cover random slopes / Alder cover + Gradient+ Clay + Hardwood cover + Non-Forested cover + Road intersection / Hardwood cover
Global – Non-forested cover random slopes / Alder cover + Gradient+ Clay + Hardwood cover + Non-forested cover + Road intersection / Non-forested cover
Global – Clay random slopes / Alder cover + Gradient+ Clay + Hardwood cover + Non-forested cover + Road intersection / Clay
Global – Road intersection random slopes / Alder cover + Gradient+ Clay + Hardwood cover + Non-forested cover + Road intersection / Ecoregion + Road intersection