Amphibians are one of the taxa that are particularly vulnerable to roads and traffic. To mitigate amphibian roadkill, previous studies have used dead animal data to identify road segments with high risk of roadkill. However, the estimation using dead animal data may be spatially biased by traffic and scavenging. This study aimed to evaluate and compare the factors affecting occurrence of both live and dead amphibians on roads. We conducted a comprehensive route census and counted the occurrence of live and dead amphibians on road between July and October 2022 and May and October 2023. Additionally, we investigated the change of traffic and removal of dead animals along landscape gradient. We constructed joint species distribution models to explain the occurrence of live and dead animals by landscape factors. From these analyses, we compared the landscape factor affecting amphibian occurrence between live and dead animals. We observed 755 live and 483 dead animals of 11 species during all surveys. Effects of proportion of surrounding paddy fields differed between live and dead animals for many of target species. Dead animals were more likely to be detected in the intermediate landscape than live animals. Traffic volume was the highest in the intermediate landscape, which have caused many dead animals there. Differences in occurrence between live and dead animals for some species could not be sufficiently explained by traffic and removal of dead animals. We emphasized the importance to consider obtaining information not only from dead animals, but live animals to mitigate amphibian roadkill.