Importance of the Landlocked Pond Smelt Hypomesus nipponensis as a food resource of the Little Tern Sternula albifrons on inland Andong Lake of Korea: A video image analysis

We carried out the diet study of the little tern on the sandy islet in inland Andong Lake, Korea, during the beeding season (April to July 2018). To identify its diet and examine the importance of the main prey species as a food resource, we set two remote-control video cameras with 4K-resolution on the islet. One thousand two hundred seventy-five still images that the tern had prey in its bill were identified at the species level and measured on a monitor. Then, they were classified to five length-categories and compared among months and breeding stages. Freshwater fishes dominated the diet (100%; eleven species overall), where the landlocked pond smelt Hypomesus nipponensis (80.8%) and largemouth bass Micropterus salmoides (13.7%) were the primary and next essential prey species, respectively. The average prey item length was 51.04 ± 20.89 mm and significantly differed among months and breeding stages (P < 0.001, respectively). 50–75 mm prey length category was the most frequent in the diet (42.2%). In April and May, larger fish >50 mm constituted the greater part of their diet (93.1%, 66.3%, respectively), whereas the diet in June and July consisted of smaller fishes <50 mm (56.2%, 68.8% respectively). The occurrence frequency of prey length categories also varied significantly among the breeding stages (P < 0.001): 1–25 mm and 50–75 mm were overrepresented and underrepresented, respectively, at the chicks in the nest stage. On the other hand, 50–75 mm was preferred for the pre-laying and incubation stages. In terms of the survival condition of pond smelts, the before- and after water surface temperatures of the day when terns flew away showed a significant difference (P = 0.004), where a threshold looks like between 29.11°C and 30.04°C. These results support the prey abundance hypothesis that, when cold-water pond smelts might wholly swim down into the deeper lake in the hot summer, the terns might also leave their colony for another foraging place with higher prey availability.


Introduction 44
The little tern is a seabird predator that breeds on spits of sand, shingle, and shell 45 fragments on seashores or in estuaries as well as rivers, lakes, and reservoirs [1,2]. It 46 9 / 32 second (30), and image compression (H.265). Except for special days, i.e., checking a 141 bad WiFi signal and temporarily withdrawing camera due to an increase in water level, 142 we usually recorded tern's prey delivery events 17 h a day from 05:00 to 22:00. After 143 chicks left the islet, we shot parent birds' feeding behaviors by a portable 4K-video 144 camera (Raven, RED). identified prey at the level of species and determined its numerical abundance [42]. 151 Then, we classified undistinguishable long shots or blurred pictures into "unidentified". 152 The frequency of prey occurrence was compared among months (April-July) and 153 breeding stages: (1) the pre-laying (Male brings prey to female for mating), (2) the 154 incubation (Male delivers prey to incubating female), (3) the chicks in the nest (Parents 155 deliver prey to chicks in the nest), and (4) the mobile chicks on the ground period 156 (Parents deliver prey to frisking chicks). 157 158

Calculation of prey length 159
To obtain prey length data, we also measured little tern's culmen length (LCL) and 160 prey's total length (PTL) on a monitor (resolution: 1920 × 1080, length: 15.6 inches) by 161 a vernier caliper (nearest 0.01 mm). Then, we selected only side shots or frontal shots 162 with a vertical bill because those with a diagonal bill to the camera had become 163 logarithmically shorter in LCL than their actual values [43]. In addition, we excluded 164 those in which even an end of bill or prey was covered with other subjects in 165 measurement for minimizing bias. The little tern carries fish crosswise in its bill [44], 166 being hung vertically, where there may be less distorted in length. Since a head of the 167 fish, however, tends to tilt to some extent, we measured the PTL in consideration of the 168 unfolded length of the bent fish after a preliminary measurement. Although the male 169 tern is greater in body size than the female, it is not easy to visually distinguish between 170 the two sexes [45]. Therefore, we determined the actual PTL by comparing the 171 measured LCL and PTL and average LCL (30.56 mm) measured in Korea [11]. 172 With a division of 25 mm, we complementarily classified PTLs to 5 length categories: 173 (1) 1-25 mm, (2) 25-50 mm, (3) 50-75 mm, (4) 75-100 mm, (5) ≥100 mm. Based on 174 this classification information, we determined the preferred prey size of little tern adults 175 and chicks as follows, respectively: (1) very small, (2) small, (3) medium, (4) large, and 176 (5) very large. To verify temporal variation in prey size, we also analyzed the frequency 177 of these categories in 4 months and breeding stages, respectively. 178 179

Aquatic factors 180
The investigation of the water surface temperature and dissolved oxygen (hereafter 181 WST and DO, respectively) was conducted to assess their effects on tern's primary prey. 182 Since the tern opts to prey mainly on fish nearer the surface or in water less than 15 183 cm deep [46, 28, 47], we surveyed the water factors between the surface and 1 m deep 184 during the breeding season. We obtained these data from automatic measurement 185 system (Troll 9500, Insitu) set by dam management authority, which automatically 186 measured them four times a day (06:00, 12:00, 18:00, and 23:00). We used the average 187 value of both depth categories and 4-hour categories as a day value. The WST and DO 188 of the day at which all little terns migrated into different foraging place in the late 189 breeding season were defined as potential survival thresholds, under which the main 190 prey fish could no longer survive and thus moved into deeper water. By comparing both 191 "before-" and "after-" moving values (1st to 15th and 16th to 31th, 15th and 16th of July, 192 respectively) the correlations between aquatic factors, little terns, and pond smelts were 193 examined. 194 195

Data analysis and Statistics 196
Chi-squared test and Kruskal-Willis test were used to assess differences in diet 197 composition and occurrence frequency of prey length categories and ones in average 198 PTL between months and breeding stages, respectively. However, a comparison of two 199 length variables or "before-" and "after-" moving" values was conducted by Mann-200 Whitney U-test or Student t-test. We expressed data as mean ± SD and set the 201 significance level at P < 0.05 for all statistical tests. All computations were carried out 202 in PASW Statistics 18. 203 204

205
The date when little terns arrived for the first time at the sand islet on Andong Lake was 206 the 9th of April, and the one when they left for the last time was the 16th of July. In total, 207 31 breeding nests were made there, and 78 eggs were laid and incubated. We gained 208 video clips of ca. 2,000 h 89 days long and succeeded to capture 1,275 best still images 209 that terns had preys in their bills. 210 Most of the images (97.2%) were identified at the species level. Freshwater fishes 211 dominated the diet, accounting for 100% of prey by number and a total of 11 species 212 (Table 1). On the contrary, insects and crustaceans did not occur in the current study. 213 For identified prey items, the pond smelt (80.8%) was the primary prey species, and the 214 largemouth bass (Micropterus salmoides; 13.7%) was the next essencial one (Table 1). 215 The latter appeared suddenly on the prey identification list in late May, which consisted 216 mostly of very small and small fries (for bass 47.4%, 33.7%, respectively; Table 1 and 217 Fig 4). The other nine fish species constituted a minor part of the diet (2.8%; Table 1  The length of pond smelt used as prey also differed significantly among 4 breeding 273 stages and months (Kruskal-Willis test: n = 781, χ 2 = 118.250, P < 0.001; n = 981, χ 2 = 274 104.624, p < 0.001 ; Fig 7A and 7B). The medium and larger pond smelts >50 mm and 275 the small ones <5 mm were delivered for females on the mating period and young in the 276 nest, respectively. In terms of the survival condition of pond smelts, both environmental factors were 287 estimated within a normal range during the entire breeding season from 9th of April to 288 15th of July [Fig 8 and Fig 9]. The average WST and DO were 22.40 ± 4.35℃ (range = 289 8.27-30.01℃) and 9.49 ± 0.98 mg/L (range = 7.22-12.28 mg/L), respectively. However, 290 they showed a contrary result in July when terns flew away. Between DO between the 291 1st to 15th and 16th to 31th of July, there was not found a significant difference, 292 whereas those in WST did (Mann-Whitney U-test: P < 3.313, P < 0.001, respectively; 293 Table 2). In addition, DO between 15th and 16th of July did not differ, whereas WST 294 did (Mann-Whitney U-test: P = 0.186, Student t-test: P = 0.004, respectively; Table 2). 295 Therefore, a threshold WST for little terns looks like between 29.11℃ and 30.04℃, the 296 average WST of their presence (15th of July) and non-presence days (16th of July) 297 [ Table 2 and Fig 8], respectively.  RCVCs that we set on the islet to capture the images that the little tern had prey in its 322 bill was much effective for identifying their prey species. Non-interference in the 323 breeding activity on colony and revealment of the greater part of prey species were the 324 substantial merit of this method. Above all things, we could record foraging activities at 325 dawn and after sunset attributed to IR mode, which helped for taking all aspects of prey. 326 Besides, the estimation of PTL and LCL on a captured image reflected the actual length 327 approximately. The maximum value of PTLs was 112.93 mm in the current study, which 328 was similar to or did not exceed that of the previous study in Andong Lake (74.0-105.1 329 mm in standard length) [48]. 330 Eleven fish species including pond smelt and largemouth bass dominated tern's diet, 331 and insects and crustaceans, also rare in other habitats [3, 5, 27], were not detected, 332 which probably seems to result from the year effect. It was observed in the previous 333 year that a chick fed on a butterfly brought to itself (unpubl. data). 334 The length of prey delivered to little tern chicks in this study (mean = 45.86 ± 20.55 335 mm) was similar to that of the UK (30-70 mm) [3]. Generally, the prey type of terns is 336 affected by the fish abundance of foraging place [49,50]. The ichthyofauna in Andong 337