Effect of the spatial arrangement of habitat patches on the development of fish habitat models in the littoral zone of a Canadian Shield lake

Abstract

We developed fish habitat models in a Canadian Shield lake using (i) a sampling-site approach based on analytical units having a surface area equal to that of sampling sites (S ~ 200 m2), (ii) a constant-multiple approach in which the analytical units constituted grouping of adjacent sampling sites in units of increasing sizes (e.g., 2S or 3S), and (iii) a habitat-patch approach in which only contiguous sampling sites with similar environmental characteristics were merged. The best models explaining within-lake variations in fish density, biomass, and community structure on the littoral zone were obtained using the constant-multiple approach, but the predictive power of these models was highly variable (0 < R2 < 0.9) compared with the habitat-patch approach (0.27 < R2 < 0.49). For these approaches, intrinsic variables (estimated inside the analytical units) explained on average 16%–27% of the variations of fish descriptors compared with 6%–32% for extrinsic variables (observed outside analytical units or related to the spatial arrangement of habitat characteristics). Our study suggests that habitat patches are reliable analytical units with which to develop fish-habitat models. Our study also indicates that inclusion of variables that refer to landscape characteristics may significantly improve the predictive power of fish habitat models.