Light pollution is increasing and artificial light sources have great impacts on animals. For migrating birds, collisions caused by artificial light pollution are a significant source of mortality. Laboratory studies have demonstrated that birds have different visual sensitivities to different colors of light, but few field experiments have compared birds’ responses to light of different wavelengths. We used three monochromatic lights (red, green, and blue) and polychromatic yellow light to study the impact of wavelength on phototaxis at two gathering sites of nocturnally migrating birds in Southwest China. For both sites, short-wavelength blue light caused the strongest phototactic response. In contrast, birds were rarely attracted to long-wavelength red light. The attractive effect of blue light was greatest during nights with fog and headwinds. As rapid urbanization and industrialization cause an increase in artificial light, we suggest that switching to longer wavelength lights is a convenient and economically effective way to reduce bird collisions.

Study Area

Our research sites were Zuomeiguo (ZMG) of Kaiyuan City (N23°37’21”, E103°30’24”, 2334 m above sea level) and Jinshanyakou (JSYK) of Zhenyuan County (N23°56’40”, E101°29’50”, 2467 m above sea level), both in Yunnan Province, China. Both sites are important concentration points of nocturnally migrating birds. These two sites are located at narrow passes in the Dahei Mountain and Ailao Mountain ranges that run from northwest to southeast. The distance between these two sites is approximately 200 km. The habitats of the two sites differ. JSYK has high forest coverage and is a middle-mountain moist evergreen broad-leaved forest with a tree height of about 10 m. In order to avoid having trees blocking the light, the test site was an open area where the trees had been cut. ZMG is a Karst landform, and the test site was located in an abandoned coalmine. The vegetation is mainly low shrubs and there is no forest (Figure 1; Figure 2). In the autumn, migratory birds must pass over these mountains during migration to the south. However, due to the obstruction posed by the mountains, birds gather and migrate along the valley on the eastern slopes and finally climb over some of the lower narrow mountain passes (Figure 1). The areas surrounding both gathering sites are often shrouded by heavy fog due to the southwest autumn monsoon. In these conditions, many migrating birds collide with lighted buildings near the narrow passes of the mountain ridges or with automobiles at night (Huang et al. 2006, Yang et al. 2009). Each autumn, the local forestry departments use the phototaxis of migratory birds to catch them with lights at the gathering sites to conduct ringing. In the past 20 years, the number of birds ringed during nocturnal migration in Yunnan Province has exceeded 140,000 (Zhao 2015).

Light Study Regimen

This research was simultaneously carried out at both sites from September 18 to October 27, 2017, and, additionally, at JSYK from September 10 to October 18, 2018. At each site one mist net (mesh size 36 mm; length 12 m; height 2.6 m; net with five pockets) was erected on the mountain slope that faced the direction of migration, from northeast to southwest, and a directional light source used to attract the migrating birds was placed 0.5m behind the net (Figure 2). We used LED lamps produced by Zhongshan Ruiyi Lighting Electric Appliance Co., LTD. Led lamps, model RY-ZWTGD-100. The light sources were red, yellow, green, and blue in color and the beam was of the scattering type. Red, green and blue lights were monochromatic while the yellow light was polychromatic and of the same color as the light source used by the banding station to trap birds. The power of each light source was 100W (Table 1). The lights were turned on at exactly 20:00 each night, and each color was shone continuously for one hour before being switched to a light source of another color. The illumination ended at 24:00. This illumination period corresponded to the peak activity period of nocturnally migrating birds (Han et al. 2007). The order in which the different-colored light sources were presented was rotated each night (Appendix Table 3). Captured birds were released after banding.

Data Analysis

The bird abundance data were tested for over dispersion by using “qcc” package in R. To analyze the effect of light color, light order, and weather on number of birds caught, we applied Generalized Linear Mixed Models (GLMM) with a log-link function and negative binomial distribution, based on the results of the overdispersion test. From the fully saturated model, significant fixed effects were selected based on a significance level of 0.05 in a likelihood-ratio test. The models were run in the statistical programming environment R, version 3.2.5 (R Development Core Team, 2015) with the function “glmer. nb” from package “lme4” (Bates et al., 2014).

Nocturnal migration intensity was highly variable among nights. We included migration night as a random factor to account for the fact that baseline migration levels vary among nights, and therefore all light treatments on a given night are likely to be related.

Light color was included as a categorical fixed effect. Other categorical fixed effects were weather (clear or foggy),  wind (head or tail), light order, study locations (ZMG and JSYK), and the interaction between light color and light order. For weather, we differentiated between clear or foggy skies. For wind direction, there were mainly two types of wind at JSYK and ZMG in fall, southwest and northeast. These correspond to headwinds and tailwinds, respectively, for migrating birds.

When analyzing the difference in power of different light colors for all birds put together, a nonparametric test (1-sample Kolmogorov-Smirnov (K-S) test) was first performed to determine if the data had a normal distribution. The survey data were found to not have a normal distribution, so we used a Friedman nonparametric test. In order to analyze the differences in the composition of bird species captured in the two sites, we used the Wilcoxon nonparametric test. We counted the number of individuals of each bird species captured. Due to the inconsistencies in survey intensity, the data we used for analysis were the capture rates of each species at each station. The data were statistically analyzed using SPSS version 22 software.