Wildlife biologists and conservationists are concerned about the effects of forest fragmentation and habitat loss on pond-breeding frog populations. Most studies have assessed the effects of habitat characteristics on the presence or absence of frogs bred in water ponds at multiple spatial scales. In this study, we focused on the effects of habitat composition on Dybowsky's frog population size. We used Dybowsky's frog egg-mass counts as a dependent variable and other habitat factors as independent variables to construct a generalized additive model (GAM), in order to describe the relationship between the frog population size and habitat factors. During the breeding periods of 2012 and 2014, we collected data from 105 water ponds in the forested landscapes of the Wupao Forest Farm, Eastern Wanda Mountains, Heilongjiang Province. Egg masses were detected at 98 breeding ponds (93.33%), 3 breeding ponds were dry during the breeding season of 2012, and egg mass counts were less than 29 at 67 breeding ponds. The size of breeding ponds ranged from 4 to 300 m² with an average size of 55.88 m². GAM response curves showed that the egg mass counts were positively associated with the size of breeding ponds from 0 to 150 m², but the egg mass counts declined with increasing breeding pond size beyond 150 m². The relationship between the egg mass counts and the pond size showed a contribution factor to GAM of 0.15. The egg mass counts were negatively associated with two habitat factors, namely forest canopy and sundries coverage ratio over the surface of pond water. The forest canopy and sundries coverage ratio contribution factors to GAM were -0.30 and -0.43, respectively. This study assumed that female frogs preferred ponds with higher water temperature. Water temperature was negatively correlated with forest canopy, and it changed from 7.5℃ to 13℃ with the change of forest canopy from 0% to 100%. We also assumed that shade over water surface was very useful for frog breeding. During breeding process, drainage area can be used to avoid predators; however, excess drainage may negatively affect spawning activity of frogs. Egg mass counts at landscape level were positively associated with forest area and the contribution factor to GAM was 17.99. Our results showed that the effect of forest size to egg mass counts was in accordance with many previous studies, since the forest around the breeding ponds was considered as the core habitat for frog living. Roads were another key habitat factor that was negatively associated with egg mass counts at landscape level. The egg mass counts declined near the roads and the contribution factor to GAM was 1.40. Multivariate models developed using within-pond variables explained more variation in egg mass counts than those developed using landscape-level factors; however, we suggested that both models were useful for predicting the breeding effort of frogs at individual sites. Since urbanization of eastern Wanda Mountains continues, forest regulations should effectively protect terrestrial habitats around the seasonal ponds and proactive techniques are required to assure the maintenance of pond-breeding forest populations.