Effects of roads on giant panda distribution: a mountain range scale evaluation

Abstract

Few studies have focused on the mountain ranges scale effects of roads on wildlife. This lack of data could lead to an underestimation of the negative impact of roads on animal populations. We analyzed a dataset that included 74.4% of the giant panda population and covered 78.7% of the global giant panda habitat to estimate road-effect zones for major roads, and to investigate how these major roads influenced the distribution of giant pandas on a mountain range spatial scale. We found that the density of giant panda signs was significantly decreased by proximity to major roads. The effect zone reached 5,000 m from national roads and 1,500 m from provincial roads. Structural equation model analysis revealed that the strongest negative impact of major roads on giant pandas was via the reduction of nearby forest cover. The results should provide a better understanding of the impact of anthropogenic infrastructure and regional economic development on wildlife, thus providing a basis for conservation policy decisions. We suggest that the environmental impact assessment of proposed roadways or further researches on road ecological effects should expand to a larger scale and consider the possible habitat degradation caused by road access.

Figure 1

Study area. Map was created with package “ggplot2” in R environment^,.

Figure 2

The difference in densities of giant pandas signs (densities of signs in buffers near roads minus those in random copies) within 20000 m from roads. The solid line with black circles represents national roads, and the dashed line with open circles represents provincial roads. The dotted line indicates no difference between the densities of signs near roads and random buffers.

Figure 3

Structural equation models (SEMs) for the difference in densities of giant panda signs in buffers around roads and those in random buffers. (a) Difference of densities of giant panda signs near national roads: the land cover was defined by the proportion of forest and construction land. (b) Difference of densities of signs near provincial roads: the land cover was defined by forest cover and water cover. The values associated with the paths are the standardized path coefficients, and the thickness of black (positive) and red (negative) paths is proportional to the standardized path coefficients. Solid arrows indicate significant relationships (P-value < 0.05), and dashed arrows refer to non-significant paths. Double-headed arrows indicate covariance estimates. *Indicates results significant at the 0.05 level or lower, **indicates results significant at the 0.01 level or lower, and the superscript “a” indicates coefficients modeled as fixed parameters with no measurement error.

Figure 4

Conceptual model describing the expected associations between environmental factors and difference in densities of giant panda signs between buffers around roads and random buffers. The environmental factors include differences in land cover, defined as a latent variable with three indicator variables (proportion of forest, construction land and bodies of water), distance from roads, difference in average elevation, and differences in human population densities. Arrows represent possible path directions, and double-headed arrows indicate covariance estimates.