Altitudinal range-size distribution of breeding birds and environmental factors for the determination of species richness: An empirical test of altitudinal Rapoport’s rule and rescue effect on a local scale

Range-size distributions are important for understanding species richness patterns and led to the development of the controversial Rapoport’s rule and Rapoport-rescue effect. This study aimed to understand the relationship between species richness and range-size distribution in relation to environmental factors. The present study tested the following: (1) altitudinal Rapoport’s rule, (2) climatic and ambient energy hypotheses, (3) non-directional rescue effect, and (4) effect of environmental factors on range-size group. Altitudinal species range-size distribution increased with increasing altitude and showed a negative relationship with climatic variables and habitat heterogeneity, and a positive relationship with primary productivity. These results support the altitudinal Rapoport’s rule and climatic hypothesis; however, they do not fully support the ambient energy hypothesis. Results from testing the non-directional rescue effect showed that the inflow intensity of species from both directions (high and low elevations) affected species richness. And we found that the 2nd and 3rd quartile species distribution were the main cause of a mid-peak of species richness and the non-directional rescue effect. Additionally, the 2nd quartile species richness was highly related to minimum temperature and possessed thermal specialist species features, and the 3rd quartile species richness was highly related to habitat heterogeneity and primary productivity. Although altitudinal range-size distribution results were similar to the altitudinal Rapoport’s rule, the mid-peak pattern of species richness could not be explained by the Rapoport’s-rescue effect; however, the non-directional rescue effect could explain a mid-peak pattern of species richness.


41
To identify species richness patterns, analyses of geographical patterns in species richness 42 have been traditionally performed [1,2]. However, a few decades ago, great attention was given to the 3 43 necessity of studying species range-size distributions [2]. This concern regarding range-size 44 distributions led to the development of the controversial Rapoport's rule [3]. Rapoport's rule states 45 that higher latitudinal species have wider latitudinal ranges than that of lower latitudinal species and 46 was developed by Stevens [3]. The effect of this phenomenon on species richness is explained by the 47 rescue effect that suggests there is a source-sink dynamic in populations based on the influence of 48 immigration on extinction [4], which is called the Rapoport-rescue hypothesis [3]. This phenomenon 49 has been extended to an altitudinal gradient [5,6]. However, ever since this phenomenon was defined 50 as a rule, there have been many associated controversies related to different results obtained for 51 different taxa, sampling effort, geographical scale, and mechanism used [7-10].

52
One of the main underlying mechanisms of Rapoport's rule is that species range-size 53 distributions are determined by climatic conditions [3,6]. Organisms living at high latitudes or 54 altitudinal areas, where climatic conditions are highly variable, have broader physiological thermal 55 tolerances [11]. Therefore, this hypothesis proposes that organisms living in these areas will have a 56 wider distribution range. Other ecological determinants of range-size distribution are associated with 57 topographical complexity and habitat heterogeneity [12,13], which can be described by the ambient 58 energy hypothesis. This hypothesis states that there exists a fine subdivision regarding limited food 59 sources by the topographical habitat structure that promotes greater specialist species [14,15]. Thus, 60 more frequent interaction among species could occur at higher latitudes, thereby species range-size 61 distributions are wider with increasing latitude [15]. Although the importance of habitat heterogeneity 62 is constantly mentioned together with climatic conditions, testing of the ambient energy hypothesis 63 has not frequently occurred and there are many cases where habitat and topographical heterogeneity 64 have not been distinguished from each other [13]. In addition, the altitudinal approach for this 65 hypothesis has not yet been applied in advanced studies. 66 The predicted consequence of the altitudinal Rapoport's rule is an increase in species 67 richness from higher to lower elevations, which is termed the Rapoport-rescue hypothesis [6].

68
According to a study by Stevens [6], this phenomenon occurred in cases that showed decreasing 69 patterns of species richness with increasing altitudes. The results from this study showed that low 70 altitude localities had relatively more species near the edge of their range than high altitude sites [6].

71
Based on the hypothesis by Stevens, wide-range species are the main contributors to geographical 72 patterns in species richness. However,  stated the directions of three putative 73 rescue effects in altitudinal gradients. Based on these, two hypotheses were formed as follows: when 74 species richness exhibits a decreasing pattern with increasing altitude, an influx of species occurs at 75 high or low altitudes, and when species richness indicates a mid-peak, the influx of species occurs via 76 a non-directional rescue effect [16]. In the latter case, the increase in species richness might have 77 occurred because of the other range species rather than because of the wide-range species. Therefore, 78 different recue hypothesis should be applied based on different species richness patterns along 79 altitudinal gradients. Thus, it is important to identify which range species increased the species 80 richness and identify what environmental factors affect the distribution of each range species.

81
The present study aimed to understand the range-size distribution patterns and underlying 82 mechanism along altitudinal gradients, and the relationship between species richness and range-size in 83 relation to environmental factors. To understand the range-size distribution patterns and underlying 84 mechanism along altitudinal gradients, we tested the altitudinal Rapoport's rule and climatic and 85 ambient energy hypotheses. Meanwhile, in a previous study, a hump-shaped pattern of species 86 richness along an altitudinal gradient in the study site was identified [17]. Therefore, to determine the 87 reason for such mid-peak patterns in species richness, we also tested the non-directional rescue effect 88 and identified environmental factors associated with the range-size distribution group. The present study was conducted in mixed or deciduous forested areas in Jirisan National 92 Park (South Korea), with altitudinal range from 200 to 1,400 m above sea level (asl). The altitudinal 93 range in the park was from 110 to 1,915 m asl; however, we excluded the subalpine forest (up to 94 1,400 m asl) from the survey area to minimize the differences in bird communities among forest types 95 [17].

115
The horizontal habitat diversity was calculated by the Shannon-Wiener diversity index (H′) using the 116 area of that particular habitat type (abundance) and the number of different habitat types (richness)

117
[17], which was used as an indicator of habitat heterogeneity.

118
Analysis was conducted by model selection and multimodel inference using a generalized 119 linear model. Before adding variables to the model, we identified the correlation between variables 120 and eliminated maximum temperature correlated (r ≥ |0.7|) with minimum temperature (r = 0.991; 121 S1 Table). We developed a set of seven candidate models and calculated Akaike's information 122 criterion adjusted for small sample sizes (AICc) and Akaike weights (w i ) [18]. The high-confidence 123 set of candidate models consisted of models with Akaike weights within 10% of the highest, which 124 were used to compute model-averaged parameter estimates [17,[19][20][21]. All statistical analyses were 125 performed using the R Studio 1.1.383 software program (packages bbmle, AICcmodavg, and 126 MuMin).

128
To demonstrate the non-directional rescue effect, the midpoint of each species was 129 calculated using the median of each species range [22,23]. Then, the distance between the two groups 130 was calculated, which was divided by the center altitude (800 m asl) that showed the highest species 131 richness [17] and averaged the midpoints of all the species detected at each group by the left and right 132 group (> 800 m asl, < 800 m asl). In the case of a plot where species only existed on the left or right 133 side, it was assumed that there was no influence of non-directional rescue effect and these species 134 were given a value of 0. Using this method, we identified the intensity of species inflow along the 135 distance from the center altitude. The relationship between the number of species and distance 136 between the mean mid-point was analyzed using the best-fit curve (linear, quadratic, and exponential) 7 137 estimation function in SPSS 20.

Effect of environmental factors on the range-size distribution
139 group (quartile method) 140 All bird species were divided using the quartile method based on their identified range-size 141 distribution, i.e., less than 25% species (1st quartile species), between 25% and median number of 142 species (2nd quartile species), between median number and 75% species (3rd quartile species), and 143 more than 75% species (4th quartile species) [2]. To identify which quartile group increased the 144 species richness, present/absent data of each quartile group were used and analyzed using the 145 independent samples t-test in SPSS 20. The effects of environmental factors (climate, primary 146 productivity, and habitat heterogeneity) were analyzed in the quartile groups that were determined to 147 be affecting species richness. Analyses were conducted using model selection and multimodel 148 inference in the R Studio 1.1.383 software program (packages bbmle, AICcmodavg, and MuMin) and 149 the number of species for each quartile species.

151
Species range-size distribution (Altitudinal Rapoport's rule) 152 We tested the altitudinal Rapoport's rule in 53 breeding bird species from 142 plots. The 153 patterns in mean altitudinal range-size distributions showed a tendency of broader range-size 154 distribution with increasing elevation (Fig 1). All curves (linear, quadratic, and exponential) 155 represented by a significant relationship (P < 0.001, P < 0.001, P < 0.001, respectively; Fig 1). The 156 highest value of R 2 was a quadratic curve (R 2 = 0.41 ; Fig 1.b); however, linear and quadratic curves 157 showed a slight difference (Fig 1.a and b). The lowest value of R 2 was an exponential curve (R 2 = 158 0.38 ; Fig 1.c). showing two supported models that had Akaike weights within 10% of the highest weight (Table 1).

167
The best model of mean altitudinal range included minimum temperature and habitat diversity (w i = 168 0.870; Table 1). The second ranked model was the full model, which contained the added variable of 169 vertical coverage of vegetation, in which the Akaike weights were 6.7 times lower than that of the 170 best model (w i = 0.870 vs w i = 0.130; Table 1 Multimodel averaged parameter estimates including the two supported models over the mean 178 altitudinal range showed negative correlation with minimum temperature and habitat diversity, and 179 positive correlation with overstory vegetation (P < 0.001, P < 0.001, P = 0.043, respectively; Table   180 2).

184
Testing of the non-directional rescue effect 185 We demonstrated the non-directional rescue effect using the intensity of species inflow. We 186 found that species richness showed a tendency of increasing with increasing of distance between mean mid-points (Fig 2). A quadratic curve represented that the intensity of species inflow increased 188 up to 300 m; however, it slightly decreased after 300 m (R 2 = 0.19, P < 0.001 ; Fig 2.b). An 189 exponential curve showed a same value of R 2 with a quadratic curve. The lowest value of R 2 was a 190 linear curve (R 2 = 0.16, P < 0.001 ; Fig 2.a). We utilized more detailed methodology to identify which range-size distribution group 197 increased species richness. From analysis of independent samples t-test, we found that the 2nd and 3rd 198 quartile species contributed to increasing species richness (Fig 3). The 2nd and 3rd quartile species 199 showed a significant differences in species richness between present and absent of each range-size 200 quartile species (P = 0.002 and P = 0.009, respectively; Fig 3). Whereas, the 1st and 4th quartile 201 species did not show a significant differences in species richness between present and absent (P = 202 0.447 and P = 0.195, respectively; Fig 3). The 4th quartile species showed substantial difference in 203 the value of mean between present and absent (mean = 8.196 ± 2.443 and mean = 5.000 ± 0, 204 respectively; Fig 3); however, the 4th quartile species did not show a significant difference because 205 the most of areas were contributed by the 4th quartile species (present, n = 141 and absent, n = 1).   We identified two supported models that showed Akaike weights within 10% of the highest 229 value ( Table 5). The best model included habitat diversity and vertical coverage of vegetation (w i = 230 0.625; Table 5). The second ranked model was the full model that also contained the minimum 231 temperature ( Table 5). The model containing vertical coverage of vegetation was 16.9 times more 232 likely to be the best explanation for the 3rd quartile species richness (w i = 0.625 vs w i = 0.037; Table   233 5). When habitat diversity was included in the 3rd quartile species richness model, Akaike weights 234 were 8.5 times higher than those eliminated in the model (w i = 0.289 vs w i = 0.034; Table 5).

235
Multimodel averaged parameter estimates including the two supported models in the 3rd quartile 236 species richness represented a positive relationship with habitat diversity, and understory and 237 overstory vegetation (P = 0.001, P = 0.005, and P = 0.032, respectively; Table 6). geographical scale, and mechanism used [7][8][9][10]24,25]. Stevens [6] stated that compared to sampled 253 point studies, regional surveys are more likely to be biased owing to unequal sampling. If an intensive 254 survey is undertaken only in one elevation band, then species richness will be biased upward and 255 altitudinal range will be biased downward [6]. Thus, in the present study, we conducted a field survey 256 that utilized identical sampling intensity (S2 Table) and sampled using a point count survey method.

257
Additionally, we performed the field survey restrictively in a mixed or deciduous forested area [17].

258
Food sources are the most influential factor on the distribution of birds during the breeding season. significantly stronger support for the altitudinal Rapoport's rule. Our study site was located at above 263 23° N latitude (S2 Table), thus the above-mentioned geographical features were influential. Although 264 the present study was conducted in a relatively small mountainous area having a low elevation range, 265 our results are similar to those found for the altitudinal Rapoport's rule.

267
To understand the phenomenon that higher altitudinal species have wider altitudinal ranges, 268 we tested the underlying mechanisms, i.e., the climatic and ambient energy hypotheses. According to 269 the climatic hypothesis, we assumed that species that have a broader physiological thermal tolerance  (Table 1 and 2), thus 274 supporting the climatic hypothesis.

275
According to the ambient energy hypothesis, greater habitat heterogeneity and primary  increasing altitude according to the altitudinal Rapoport's rule (Fig 1); however, species richness 290 showed a mid-peak pattern [17]. Therefore, these two results did not show proper logical flow. To (high and low elevations). Thus, we identified the intensity of species inflow using this new method.

302
From testing of the non-directional rescue effect, the species richness showed a tendency of increasing 303 with increasing of species inflow (Fig 2). According to our prediction, the reason for higher species 304 richness at mid-elevation was owing to species inflow from other areas apart from the mid-altitude 305 area. Thus, our results supported the non-directional rescue effect. To demonstrate the reason behind 306 higher species richness at mid-elevation in relation to species range-size distribution, we used the 307 quartile method.

308
Effect of environmental factors on the range-size distribution 309 group 310 We found that the 2nd and 3rd quartile species contributed to the increased species richness 311 at mid-elevation (Fig 3). From the results of distribution patterns of each quartile species, we found 312 that the distributions of the 4th quartile species were skewed toward high altitudes, the 2nd quartile 313 species were skewed toward low altitudes, and the 3rd and 1st quartile species were distributed over 314 the entire altitudinal range (Fig 4). As shown in Fig 4,  species. Because the most of areas were equally contributed by the 4th quartile (Fig 3 and 4). The 1st 318 quartile species did not contribute to species richness either, because these species showed only a 319 small number of detections and possessed specialist species features (Fig 3) Fig 4), and possessed thermal specialist species 330 features. Whereas, the 1st quartile species was influenced by only habitat diversity and distributed 331 over the entire altitudinal range (S3 Table; Fig 4), was identified having a features of habitat specialist 332 species. Meanwhile, the 3rd and 4th quartile species are composed of species having a wide altitudinal 333 range-size. Thus, we assumed that the 3rd and 4th quartile species are not influenced by habitat and 334 temperature, and possess generalist species features [34,35]. However, we found that the 3rd quartile 335 species was influenced by habitat heterogeneity and primary productivity. Whereas, the 4th quartile 336 species was influenced by primary productivity and minimum temperature (Table 6; S3 Table). A 337 previous study conducted on latitudinal differences, the 3rd and 4th quartile species were strongly 338 influenced by primary productivity compared to other quartile groups [2], showed a coincidence with 339 our results. To achieve a better understanding of these patterns, competition among species related to 340 niche are required. From these results, we determined that the cause of mid-peak pattern of species 341 richness was not inflow of habitat specialist species [14,15], but owing to the influence of minimum 342 temperature, habitat heterogeneity, and primary productivity on the distribution of the 2nd and 3rd 343 quartile species. they do not fully support the ambient energy hypothesis. There was some logical error between the 352 Rapoport's rule and mid-peak pattern of species richness. Thus, we tested the non-directional rescue 353 effect, and the results supported this effect. Using the quartile method, we found that the 2nd and 3rd 354 quartile species richness were the main contributors to the mid-peak of species richness and the non-