The urban heat island effect is closely related to cicada density in metropolitan Seoul

Background. Cryptotympana atrata and Hyalessa fuscata are the most abundant cicada species in the Korean Peninsula, where their population densities are higher in urban areas than in rural ones. The urban heat island (UHI) effect, wherein human activities cause urban areas to be significantly warmer than surrounding rural areas, may underlie this difference. We predicted a positive relationship between the degrees of UHI in urban areas and population densities of C. atrata and H. fuscata.

significantly higher in urban areas with high UHI than in other groups. Specifically, densities of 23 C. atrata in high UHI areas were approximately seven and four times higher compared to those 24 in low UHI and in suburban groups, respectively. The order of magnitude was greater in H. 25 fuscata, where densities in high UHI group were respectively 22 and six times higher than those 26 in low UHI and in suburban groups.

27
Discussion. These results suggest that the UHI effect may be closely linked to high cicada 28 densities in metropolitan Seoul, although the underlying mechanism for this remains unclear.

INTRODUCTION
Cicadas are likely to benefit from urban warming, since their development and life history events 53 require high thermal supplies. In western Japan, Cryptotympana facialis, a closely related species 54 of C atrata, has sharply increased in urbanized areas, owing to better thermal adaptation to 55 warmer urban areas than to rural areas (Moriyama & Numata, 2008). 56 Cryptotympana atrata Fabricius (Tribe Cryptotympanini) and Hyalessa fuscata Distant 57 (Tribe Sonatini) are two popular cicada species inhabited in the Korean peninsula (Lee, 2008). 58 These two species are widely distributed in major cities, and their noisy calling songs in summer 59 are nuisance to city dwellers. Measurements of exuviae densities reveal significantly higher 60 densities of both species in urban areas than in countryside areas (Kim et al., 2014). Several 61 hypotheses have been proposed to explain the high cicada density in an urban environment, such 62 as host plant availability (Kim et al., 2014), predator avoidance strategy, habitat fragmentation 63 (Takakura & Yamazaki, 2007), and urban soil compaction (Moriyama & Numata, 2015). To our 64 knowledge, however, none of these hypotheses, have been tested to explain the abundance of 65 two cicada species in urban areas. 66 In this study, we aimed to elucidate the relationship between UHI and the population 67 densities of C. atrata and H. fuscata. We examined the densities of two species in three groups: 68 areas of high and low UHI intensity in metropolitan Seoul, and suburban areas. If urban warming 69 was beneficial for their development, the population density of each species was expected to be 70 higher in warmer urban areas than in cooler urban areas. The location, total geographic area, and total number of trees in each sampling locality are 104 reported in table 1. Four random sites were determined in each group; and three replicates were 105 randomly picked in each site. The overall geographic area of each replicate was approximately 106 10,000 m 2 , and the distance between radii of two replicates was more than 100 m, to avoid 107 overlapping cicada dispersal areas (Karban, 1981). Replicates were standardized as residential   Sidak post hoc tests were carried out to investigate pairwise difference among UHI groups. We 146 also evaluated differences among groups by a non-parametric, Kruskall-Wallis one-way 147 ANOVA test, in comparison with GLM.

149
Species compositions 150 C. atrata and H. fuscata constituted most of cicada species in all groups, in which C. atrata 151 comprised approximately 30%, and H. fuscata almost 70% (Fig. 1). Nevertheless, one-way 152 ANOVA showed no difference in species composition among three groups in both C. atrata (F2, 153 33 = 0.083, P > 0.05) and H. fuscata (F2, 33 = 1.136, P > 0.05) ( In total, resource-weighted densities of the two species were highest in the high UHI group, 159 followed by densities in suburban and low UHI groups (Fig. 2, 3). Regarding C. atrata, the 160 difference between the high and low UHI groups was 6.82 times in area-weighted density, and 161 7.16 times in tree-weighted density, meanwhile, between the high UHI and suburban groups, it 162 was 4.64 times in area-and 4.81 times in tree-weighted densities. The density difference between 163 the high UHI and other groups was more remarkable in H. fuscata. The difference between high 164 and low UHI groups was 22.12 times in area-weighted densities and 22.77 times in tree-weighted 165 densities, whereas between high UHI and suburban groups, it was 6.5 times in area-and 2.27 166 times in tree-weighted densities. Population densities of C. atrata were consistently highest in high UHI group, followed by 170 densities in suburban group, and lowest in low UHI group (Fig. 2). The results of GLM showed that whereas the UHI, sampling period, and interaction between them were significant (P < 0.05), 172 the site and replicate were not (P > 0.05) ( Table 3). The amount of variance explained by the 173 UHI group was highest in the models (h = 0.72 in both area-and tree-weighted densities), 174 compared to other explanatory variables (h < 0.3). Sidak post hoc tests showed highest 175 resource-weighted densities in the high UHI group (P < 0.001, Table 4), but no difference was 176 found between low UHI and suburban groups (P = 0.972). In comparison to the GLM model, 177 non-parametric tests also yielded a significant difference among groups (Kruskal-Wallis test; for 178 area-weighted density  2 (2, N = 72) = 18.35, P < 0.001, for tree-weighted density  2 (2, N = 72) 179 = 18.65, P < 0.001). Pairwise comparison revealed cicada population densities in the high UHI 180 group were significantly higher than those in the low UHI group (Mann-Whitney U test; for 181 area-weighted density U = -24.42, P < 0.001, for tree-weighted density U = -24.67, P < 0.001) 182 and suburban group (for area-weighted density U = -17.83, P = 0.007, for tree-weighted density 183 U = -17.83, P = 0.007).

184
Besides the UHI group, the sampling period and interaction between them were also 185 significant factors affecting resource-weighted densities (P < 0.05) (Fig. 2). Cicada densities of 186 each group were more pronounced in the second sampling period compared to the first.

187
Specifically, the area-weighted density and tree-weighted densities increased 2.9 and 3.27 times 188 in the second sampling period. Similar to C. atrata, log(x + 1) transformed densities of H. fuscata were greatest in high UHI 192 group, closely followed by densities in suburban and low UHI group (Fig. 3). The GLM model 193 on transformed densities showed that both the UHI and the sampling period were significant (P < 194 0.05), whereas the interaction between them was not (P > 0.05) ( Table 5). The site was 195 significant for both densities (P < 0.05); meanwhile, the replicate was significant for tree-196 weighted density (P < 0.05) but not area-weighted density (P > 0.05) (Table 5)

201
Sidak tests on multiple comparisons indicated that the resource-weighted densities of the 202 high UHI group were significantly greater than those of the low UHI group or the suburban 203 group (P < 0.001), whereas there was no difference between the low UHI and suburban groups 204 (P > 0.05) ( Table 4). Results of the Kruskal-Wallis test identified significant differences across 205 UHI groups (for area-weighted density  2 (2, N = 72) = 18.75, P < 0.001, for tree-weighted 206 density  2 (2, N = 72) = 18.51, P < 0.001). Likewise, Mann-Whitney U tests indicated densities 207 in the high UHI group to be greatest in comparison to other groups (P < 0.05); no difference was 208 observed between low and suburban groups (P > 0.05).

209
The sampling period was significant to densities of H. fuscata across three groups (Fig. 3), 210 which coincided with the pattern observed in C. atrata. However, no significant interaction 211 between the UHI group and the sampling period was found, although both high UHI and low 212 UHI groups exhibited noticeable variation between two samplings, compared to suburban groups.

214
Results of exuviae enumeration surveys showed that UHI variation was a significant factor in 215 densities of both C. atrata and H. fuscata in metropolitan Seoul. Cicada densities were 216 significantly higher in urban areas with high UHI than in those with low UHI or in suburban  Separate research to investigate exactly how these two cicada species actually respond to water-261 stress conditions of their host plants will be helpful to elucidate such a causal relationship. 262 We also consider the adaptation of individual species to local habitat conditions as a       Figure 2. Comparison on resource-weighted densities of C. atrata among three groups.

424
Multiple Sidak post hoc tests show that densities in urban areas with high UHI group are 425 significantly higher than in urban areas with low UHI and suburban groups (P < 0.05), 426 whereas no significant difference is found between low UHI and suburban groups (P > 0.05). high UHI group are significantly higher than urban areas with low UHI and suburban groups 432 (P < 0.05), whereas no significant difference is found between low UHI and suburban groups 433 (P > 0.05).