1. Habitat fragmentation, which involves habitat size reduction and isolation, is a major cause of biodiversity decline. However, interest in small isolated habitats has increased among ecosystem managers because these fragments can serve as remnant refuges for unique and/or endangered species in human-altered landscapes.

2. In a fragmented floodplain of northern Japan, we demonstrate how habitat fragmentation and habitat quality shape the refuges for a unique endangered minnow (Rhynchocypris percnurus sachalinensis) from biological invasions. We focused on two invasive minnows, topmouth gudgeon (Pseudorasbora parva) and rosy bitterling (Rhodeus ocellatus ocellatus), which are the dominant invasive species. By using a graph theoretical approach and structural equation modeling, we elucidated the relationships among habitat fragmentation, habitat quality and the abundance of invasive minnows with the endangered swamp minnow.

3. We found that the invasion of topmouth gudgeon, which has a high mobility and environmental tolerance, negatively affected swamp minnow populations. The invasive species may outcompete the native species in their overlapping trophic niche.

4.  Analyses indicated that habitat fragmentation (decreasing habitat size and connectivity) indirectly and positively affected the population abundance of swamp minnow by reducing the potential for invasion by topmouth gudgeon. We further found an interaction between the indirect effects of habitat fragmentation and local habitat quality: the indirect effects became more apparent in the relatively deeper habitats that were of better quality for natives. This result was likely attributed to the strong control effect of the abiotic stressor. Interspecific competition between the two minnows would be masked in shallow ponds because the native populations are primarily restricted by the abiotic factor, and its abundance is inherently limited regardless of pressure by invaders.

5.  In fragmented landscapes, evenly conserving all small fragments within a limited budget is difficult. The change in importance of small isolated ponds across a pond-depth gradient suggests that conservation priorities for small fragments should be established by considering local habitat quality. Our findings do not justify habitat fragmentation but suggest that both landscape-scale and local-scale perspectives synergistically aid in the management of biological invasions in modified habitat networks.

Definitions of the variable headings

Site：ID of study sites

River：River name for each study site

Pseudorasbora parva_CPUE：Relative abundance of topmouth gudgeon

Phoxinus percnurus_CPUE：Relative abundance of swamp minnow

Rhodeus ocellatus_CPUE：Relative abundance of rosy bitterling

dIIC_15：Delta Integral Index of Connectivity (threshold connectivity distance = 15 km)

Vegi：Vegetation cover ratio

Depth：Pond depth

Min DO：Minimum dissolved oxygen

EC：Electrical conductivity

PH：PH

*Note that all variables are log or arcsine transformed.