利用生態棲位模擬探討鄰域鳥種間之棲位分化

全文

(1)國立台灣師範大學生命科學系碩士論文. 利用生態棲位模擬探討鄰域鳥種間之棲位分化 Niche differentiation between parapatric parrotbills (Paradoxornis webbianus and P. alphonsianus)? A test using ecological niche modeling. 研究生：曹子軒研究生：Tzu-Hsuan Tsao 指導教授：李佩珍博士指導教授：Dr. Pei-Jen Shaner 指導教授：李壽先博士指導教授：Dr. Shou-Hsien Li. 中華民國 101 年 8 月.

(2) 致謝經過了兩年多的學習與努力，才有今日的成果。這篇論文的完成，首先要感謝我的指導老師李佩珍，李老師是和我同時進入師大生科系的新老師，身負沉重的研究及教學壓力，然而卻非常有耐心的指導我這個沒有任何研究經驗的非本科生，即使在我放棄第一個論文題目時，也願意讓我做另一個與實驗室完全不同類型的題目，並且花相當多的時間和我一起研究，以及教導我許多統計及科學知識，讓我能獨力完成研究。我很幸運能夠進入李佩珍老師的研究室，而這也要感謝林登秋老師的推薦，才讓我有機會和李老師學習。林老師是我在大學時期就認識的老師，上課相當的認真且生動有趣，很高興能在師大生科系再次遇到林老師，讓剛來的我不會感到如此陌生。要完成碩士學位我想並不是想像中的那樣順利。在碩一即將結束時，我因為在決定論文題目遇到困難而曾經想休學，我的共同指導老師李壽先，也許是因緣際會，剛好對於我想研究的東西頗感興趣，因此在那時提供了一個機會讓我研究我想研究的東西。雖然我的研究內容本身也與李壽先老師的研究主題沒有太大關係，但李老師卻非常熱心的協助我，盡可能的提供我需要的資料，讓我完成一個難得的跨領域研究。除此之外，李壽先老師也花相當多的時間在教導我論文寫作以及報告技巧，讓我了解的不只是科學寫作，也是做任何事都應該要.

(3) 有的態度和邏輯判斷。在兩位李老師的督促及教導之下，我的碩士生涯不單單只是修課、研究、寫論文然後畢業。在這兩年半之間，我有非常多上台報告我研究內容的經驗，甚至是在美國的國際生態會議上分享。若沒有兩位老師的督促及耐心，我想我是不可能上台分享的，然而，不論以後是不是從事學術研究，上台報告或分享應該是必要的過程。此外，在我畢業口試之後，我也有機會到美國當短期研究生，看看國外學者的研究以及體驗美國生活，算是完成我小時候的小夢想。真的非常感謝兩位李老師在這兩年多來的指導及學習的機會。由於我所待的兩個研究室皆不是以我的研究內容為主題，要直接得到專業的建議是相當不容易，因此能完成這篇論文也要感謝兩位口試委員：台大森林系的丁宗蘇博士以及北大公衛系的 Bruno Walther 博士。兩位具有生態棲位模擬專業的學者花了很多時間修改我的論文。其中 Bruno Walther 博士在我剛接觸這個題目時就給了我相當多的專業知識，特別感謝他的熱心協助。兩位口試委員除了給了我很多重要的建議及修改方向外，也讓整個口試過程像是一場有趣的科學討論，而不單單只是一個考試的過程，倍感收穫。此外，我的研究內容是以統計為基礎，必要有足夠的統計知識才能夠了解，因此特別感謝許鈺鸚老師能讓我旁聽她的統計課程，經過許老師的講解我更能融會貫通.

(4) 統計的知識。也很感謝台大的李瑋珠老師讓我這外校生旁聽她的多變量課程，讓我稍微了解一個非常難懂但我必須知道的統計知識。我的研究必須得依靠相當龐大的資料量，以得到較可靠的分析結果，因此很感謝郭玉民、雷富民、楊曉君、梁偉、王龍武、楊燦朝、 Jin-Won Lee 等學者以及中華鳥會所提供的資料，才能讓我完成此篇論文。當然，最感謝的還是李壽先老師，李老師擔心我的資料量不夠，一直非常熱心的幫我向各學者們蒐集資料。能順利完成我的碩士學位也得感謝我親愛的學長姐及同學們。感謝和我一起進這個研究室的吳佩真、郭儆寰，能認真聽我報告、給予建議。感謝江伊韓每次都幫我製作精緻的投影片及耐心聽我無聊的報告。感謝陳宣汶學長特別指導我統計相關的知識。感謝褚瑞華、洪心怡、林容仟、葉佳芬、張伊鈞學姐們，給我很多報告、寫作的經驗以及研究建議。感謝柯伶化學姐、羅諠憶、周琮焜、陳冠廷、李苡柔、陳偉欽、馮騰輝等實驗室的夥伴們，能在每次 meeting 耐心聽完我這實驗室異類的報告。特別感謝柯伶樺學姊、吳佩真、江伊韓、羅諠憶、林苡柔、陳偉欽能在我口試時全程幫忙，讓我順利通過。最後，要感謝我的家人，感謝我的父母願意讓我在花三年的時間攻讀碩士，花錢養我、包容我住在家裡當米蟲，也感謝我的哥哥提供我一些就讀碩士的經驗，讓我不會對研究感到害怕。沒有家人的鼓勵.

(5) 及支持，我也不會有機會去研究科學，更不可能攻讀碩士學位。. 謝謝你們!.

(6) Contents Abstract …………………………………………………………...……. i 摘要 …………………………………………………………………… iii Introduction ………………………………………………………… 1 Method and Materials …………………………………………… 7 Study area ………………………………………………………… 7 Species presence records ………………………………………… 7 Environmental data layers ………………………………………… 9 Modeling algorithm ……………………………………………… 10 Test of random spatial distribution ……………………………… 12 Test of niche equivalency and niche partitioning …………...…… 13 Test of secondary contact ………………………………………… 15 Results ……………………………………………………………… 16 Current-day sympatric zone ……………………………………… 16 Non-random distribution and niche partitioning in the sympatric zone ……………………………………………………………… 16 Last glacial maximum distributions and secondary contact ……… 18 Discussion …………………………………………………………… 19 Niche partitioning and competition ……….……………………… 19 Ecological parapatry ……………………………………………… 21 Secondary contact and non-equilibrium distribution …………… 22 Implication and future work ……………………………………… 23 Literature cited …………………………………………………… 25.

(7) Table …………………………………………………………………34 Table 1. The 19 bioclimate variables in WorldClim, and their current-day and last glacial maximum ranges for the study area ……………………………………………………………… 34 Table 2. The structure matrix of the first canonical discriminant function, and the discriminant loadings of original bioclimatic variables ………………………………………………………… 35 Figure ……………………………………………………………… 36 Figure 1. The outcomes of competitive exclusion between two species …………………………………………………………… 36 Figure 2. The study area and species presence records ………… 37 Figure 3. The binary distributions of P. webbianus and P. alphonsianus ……………………………………………………… 38 Figure 4. The potential sympatric zone of P. webbianus and P. alphonsianus ……………………………………………………… 39 Figure 5. The presence records of P. webbianus and P. alphonsianus within current-day sympatric zone ……………………………… 40 Figure6. The actual and null average nearest neighbor distance (ANND) of P. webbianus and P. alphonsianus …………………… 41 Figure 7. The actual and null D and I values of P. webbianus and P. alphonsianus ……………………………………………………… 42 Figure 8. The ranges of temperature seasonality and isothermality occupied by P. webbianus and P. alphonsianus within the sympatric zone ……………………………………………………………… 43 Figure 9. The binary distributions of P. webbianus and P..

(8) alphonsianus during last glacial maximum (LGM) …………….… 44 Appendix …………………………………………………………… 45 Appendix I. The presence records of P. webbianus and P. alphonsianus ………………………...…………………………… 45 Appendix II. Map of the study area and contact region between P. webbianus and P. alphonsianus …………………………… 46 Appendix III. Preliminary tests on the changes in test AUC values w i t h i n c r e a s i n g s t u d y a r e a s f o r P. w e b b i a n u s a n d P. alphonsianus ……………………………………………………… 47 Appendix IV. Predicted distributions of P. webbianus and P. alphonsianus with and without low resolution data points …… 48 Appendix V. Predicted distributions of P. webbianus and P. alphonsianus with original and trimmed data sets ………… 49 Appendix VI. The effects of thresholds on predicted sympatric zone of P. webbianus and P. alphonsianus, and on statistical test results ………………………………………………..…………… 50 Appendix VII. Predicted distributions of P. webbianus and P. alphonsianus based on five and 19 bioclimatic variables ……………………………………………………………… 51 Appendix VIII. The first canonical discriminant function scores of P. webbianus and P. alphonsianus …………………..……………… 52 Appendix IX. The probability distributions of P. webbianus and P. alphonsianus during last glacial maximum ………………….…… 53.

(9) Abstract Competition is one of the key mechanisms determining species range limits, community structures, and ecological speciation. At geographic scale, competitive exclusion is often proposed to be a cause of parapatric distribution between species that have similar niche requirements. However, it is exceedingly difficult to demonstrate competitive exclusion in natural settings. In this study, I used ecological niche modeling to predict potential distributions of two closely-related avian species in Asia, Paradoxornis webbianus, and P. alphonsianus. The current-day distributions of the two species indicate that they share an area of potential sympatric zone in Southwestern China, within which both species exhibit non-random spatial distributions (P. webbianus in the northeastern region, and P. alphonsianus in the southwestern region). The niche identity test shows that the two species occupy different niches within the sympatric zone. Furthermore, the discriminant analysis points to temperature variability as the most likely niche dimensions along which P. webbianus and P. alphonsianus differentiate. Paradoxornis webbianus appears to be more tolerant of temperature fluctuations than P. alphonsianus. The potential distributions of the two species during the last glacial maximum (21,000 years ago) suggest that they have maintained a similar area of potential sympatry since their divergence approximately 30,000 to 25,000 years ago. Therefore, their spatial segregation within the current-day sympatric zone is more likely a result of competition than secondary contact. In conclusion, this study provided strong support for the role of competition in maintaining parapatric i.

(10) distribution between two recently-diverged species. Ecological speciation through competition and niche partitioning, therefore, might play a key role in the rapid divergence among closely-related species such as P. webbianus and P. alphonsianus.. Keywords: competition, ecological niche modeling, Paradoxornis, niche partitioning, parapatric distribution, sympatric speciation. ii.

(11) 摘要物種競爭是決定物種分布範圍、群聚結構以及生態種化的關鍵因素之一。而競爭排除被認為是棲位相近的二物種，在地理上鄰域分布的主要原因。然而，要在自然的情況下證明競爭排除則相當困難。本研究以分布於亞洲的二近緣鳥種：棕頭鴉雀(Paradoxornis webbianus)以及灰喉鴉雀(Paradoxornis alphonsianus)，利用生態棲位模擬來預測其潛在分布。由現今的潛在分布顯示二物種在中國四川具有一潛在共域區，而二物種在此區域內的實際分布則為非隨機分布(宗頭鴉雀分布於東北方，而灰喉鴉雀位於西南)。棲位相同檢測的結果顯示二物種在此潛在共域區內佔有不同的氣候棲位。此外，判別分析的結果顯示，溫度變異最可能是導致二物種分化的棲位面向：棕頭鴉雀可能比灰喉鴉雀更能忍受氣溫的變動。模擬二物種在末次冰河最盛期(約二萬一千年) 的潛在分布顯示，二物種自從二萬五千年至三萬年分化以來，皆維持一至的共域區。因此，二物種在現今潛在共域區內的空間劃分很可能是導因於競爭排除而不是近期的次級接觸。本研究的結果支持物種間的競爭能夠維持二近緣物種的鄰域分布。因此，由競爭以及棲位分化所導致的生態種化，很可能是造成近緣物種間(如棕頭鴉雀及灰喉鴉雀)快速分化的重要因素。. iii.

(12) 關鍵字：競爭、生態棲位模擬、鴉雀屬、棲位分化、鄰域分布、同域種化. iv.

(13) Introduction Competition is arguably one of the most important mechanisms determining species range limits, community structures, and ecological speciation. The principle of competitive exclusion states that when different species share the same resource that is limiting, the species that can utilize the resource more efficiently will eventually outcompete, and exclude other species (Gause 1934). As a result, two species with the same niche requirements cannot stay in sympatry. Instead, they are likely to form a parapatric distribution at their respective range boundaries. Past studies have demonstrated that competitive exclusion can drive parapatric distributions (Connor & Bowers 1987 review; Bull 1991 review; Sexton et al. 2009 review). However, most of these studies were manipulative experiments conducted in closed systems, using plants or other species with little or no mobility as model species (e.g. Berendse 1983; Wethey 2002; Cunningham et al. 2009). For highly mobile organisms with a wide geographic distribution, such as large mammals or birds, it is extremely difficult to test competitive exclusion. In order to demonstrate competitive exclusion between two parapatrically-distributed species, they must have an overlapping fundamental niche (Hutchinson 1957) within which they compete with each other. The outcome of the competition is that one or both of the species can only utilize portions of their fundamental niches (i.e. the realized niche; Hutchinson 1957). Therefore, it is a prerequisite to establish species’ fundamental niches before competition-related hypotheses about niche partitioning can be tested. However, the 1.

(14) multidimensionality nature of the niche concept makes it extremely difficult to measure fundamental niches (e.g. Pulliam 2000). As an alternative, ecological niche modeling is often used to quantify species’ fundamental niches (Soberón & Peterson 2005). Ecological niche modeling generates potential distribution of a species based on its occurrence or abundance data, and relevant environmental data (Guisan & Zimmermann 2000 review; e.g. Nielsen et al. 2005; Seoane et al. 2005; Murray et al. 2009). There are different approaches to model building such as correlative or mechanistic, unsupervised or supervised with expert opinions. At regional or continental scale, using correlative and unsupervised models can be a more efficient and flexible approach (Seoane et al. 2005; Murray et al. 2009; Kearney & Porter 2009 review). The environmental data layers are used to encapsulate various niche dimensions. The potential distribution of a species is a spatial representation of its fundamental niche. The actual distribution of a species, on the other hand, represents its realized niche. By comparing the potential and actual distributions of parapatrically-distributed species, one can discern competitive exclusion between species (Anderson et al. 2002; Costa et al. 2008). There are two types of possible outcomes when using this approach to test competition between two species. The first outcome is that the actual presence records of the two species are randomly distributed within the potential sympatric zone, which is the overlapping area of the potential distributions of the two species, or their shared fundamental niches (Figure 1a). This outcome indicates that competitive exclusion. 2.

(15) does not occur between the two species, and either species can freely fulfill their respective fundamental niche within the sympatric zone. In contrast, the second outcome is that the actual presence records of the two species are not randomly distributed within the potential sympatric zone, which includes total absence of either one of the two species (Figure 1b), or spatial segregation of the two species (Figure 1c). These patterns suggest possible competition between the two species. As a result, one or both of the species cannot fulfill their respective fundamental niche within the sympatric zone. Many studies have successfully applied ecological niche modeling to investigate competition (Anderson et al. 2002; Cadena & Loiselle 2007; Costa et al. 2008; Ritchie et al. 2009; Acevedo et al. 2010; Pellissier et al. 2010). Specifically, Anderson et al. (2002) predicted potential distributions of two closely-related, and parapatrically-distributed species, Heteromys australis, and H. anomalus, in Central and South America. Their results showed that one of the two species dominated within their potential sympatric zone, suggesting that competitive exclusion occurred between them. Costa et al. (2008) examined six pairs of amphibians and reptiles in Oklohoma, U.S.A., and found that one pair exhibited a similar pattern with one species dominating within the sympatric zone, indicating competitive exclusion. In this study, I used Paradoxornis webbianus and P. alphonsianus to test competitive exclusion between two parapatrically-distributed species. Paradoxornis webbianus is widely distributed from northern Vietnam to southeastern Siberia, including the Korean Peninsula, and the island of Taiwan. Paradoxornis alphonsianus, on the contrary, is restricted to. 3.

(16) high-altitude areas in Sichuan province, Guizhou province, and Yunnan province in China, and northern Vietnam (Appendices I & II). They have a narrow contact zone (sympatric zone based on actual presence records) along the western edges of the Chengdu Plain, Yunnan province, and Guizhou province (Appendix II). According to a phylogeny based on mitochondrial DNA, P. alphonsianus alphnosianus, P. alphonsianus yunnanensis, and some individuals of P. webbianus suffusus belong to the same branch (Yeung et al. 2011). In addition, data on nuclear loci indicated that the two species diverged approximately 30,000 to 25,000 years ago (Lin & Li, unpublished data). Therefore, P. alphonsianus has recently diverged from P. webbianus suffuses. This study includes P. webbianus suffusus, and two subspecies of P. alphonsianus (P. alphonsianus alphnosianus and P. alphonsianus yunnanensis), as the study species. The theory of niche conservatism predicts that recently diverged species should occupy a similar niche because species tend to retain their ancerstor’s fundamental niche through time (Wiens & Graham 2005, Wiens et al. 2010). A recent review by Peterson (2011) also pointed out that most species do not alter their coarse-resolution niche, such as temperature or precipitation, at the scale of hundreds of thousands of years. Because P. webbianus suffusus and P. alphonsianus were recently diverged, they are likely to share similar niches, and therefore are subject to current competition (Violle et al. 2011). In fact, field observations suggest that both species inhabit similar habitats (e.g. shrub lands, grasslands, and woodlots), and overlap in diets (e.g. insects, seeds and. 4.

(17) fruits; Wu & Chen 1986; Yang 2004; Robson 2007). Other observations also suggest that P. alphonsianus replace P. webbianus suffusus above 1,000-meter altitude in Chengdu Plain, Yunnan province, and Guizhou province (Robson 2007). Therefore, I hypothesize that the parapatric distribution of P. webbianus suffusus and P. alphonsianus is a result of competitive exclusion. An alternative explanation for the parapatric distribution is secondary contact. Secondary contact refers to the situation where two species that have undergone allopatric speciation come into contact as a result of range expansion. In this scenario, their parapatric distribution merely reflects the fact that they have not had enough time to fill their fundamental niches (disequilibrium distribution; Davis 1986). Assuming niche consevratisim is strong, one can use ecological niche modeling to simulate species’ paleo-distribution (Kozak et al. 2008), and assess how likely that the parapatric distribution is caused by secondary contact. Specifically, a lack of potential sympatric zone at time of species divergence would suggest allopatric speciation, and provide support for the alternative explanation of parapatric distribution driven by secondary contact. In order to test competitive exclusion between P. webbianus suffusus and P. alphonsianus, I used ecological niche modeling to generate their potential distributions. My predictions are: (1) the actual distributions of these two species within the area of potential sympatry are non-random, and (2) the two species differentiate in niche uses within the area of potential sympatry. In addition, I compared the current-day sympatric. 5.

(18) zone between P. webbianus suffusus and P. alphonsianus to that during the last glacial maximum, and evaluated the possibility that secondary contact is the reason for their parapatric distribution.. 6.

(19) Method and materials Study area The study area was set based on the minimum convex polygon of all presence records of P. webbianus suffusus and P. alphonsianus, buffered by four decimal degrees (one decimal degree equals approximately 111 kilometers at equator). My study area thus extends from Northern China to Northern Vietnam range from 97.595°E to 126.055°E, and 17.726°N to 40.685°N (Appendix II). Preliminary tests suggested that smaller study areas produced models with lower test AUC values (area under the receiver operating characteristic curve; Appendix III). Although AUC has been questioned as a measure of relative performance of models with different study areas, there are currently no alternative methods (Lobo et al. 2007). On the other hand, extremely large study areas may lead to model overfitting and reduced biological realism (Anderson & Raza 2010). Therefore, I limited the study area to a four-decimal-degree buffer, retaining the higher AUC values while avoiding the drawbacks of large study areas. The resulting study area was 5,668,290 km2.. Species presence records The presence records of P. webbianus suffusus and P. alphonsianus were obtained from museum archives, birding records, literature (Wu & Chen 1986; Yang 2004; Lu et al. 2006), and personal observations (Y. M. Kuo, F. M. Lei, W. Liang, L. W. Wang, T. C. Yang, and X. J. Yang personal communications). Because some P. alphonsianus individuals were identified as a subspecies of P. webbianus suffusus in early studies 7.

(20) (e.g. Wu & Chen 1986; Yang 2004; Liu et al. 2010), I used photographs and specimen to confirm suspicious records. If the evidence was inconclusive, I excluded the records from the data sets. This confirmation process removed 20 inconclusive records of P. webbianus suffuses, and 6 inconclusive records of P. alphonsianus. Through these processes, I gathered a total of 801 records of P. webbianus suffusus, and 177 records of P. alphonsianus (Figure 2a). For records that did not include geographical coordinates, I georeferenced the locations in Google Earth manually based on their names and descriptions. In order to evaluate if spatial resolutions of the original records affect predicted distributions, I excluded data points with a spatial resolution of more than 200 km2 (e.g. the location description is a county). This method resulted in 124 data points being excluded from the 801 records of P. webbianus suffusus, and 38 data points being excluded from the 177 records of P. alphonsianus. The predicted distributions of these two species with or without low resolution data points were qualitatively similar (Appendix IV). Therefore, I retained all data points in subsequent analyses. Species presence records are often clustered due to sampling bias towards locations with easy access, such as established transects, parks and campuses (Reddy & Davalos 2003; Hortal et al. 2008). This could lead to locations of clustered presence data being predicted as of high suitibility (Appendix V). To overcome this issue, I randomly chose data points that are at least 0.5 decimal degrees apart from one another, and obtained trimmed data sets of 228 presence records for P. webbianus. 8.

(21) suffusus, and 48 presence records for P. alphonsianus (Figure 2b). Preliminary tests suggested that data trimming could reduce the predicted suitability of large urban areas with highly clustered presence records (e.g. WuHan metropolitan area; Appendices II & V), and increase the predicted suitability of rural areas with few records (e.g. Longyan city; Appendices II & V). In addition, previous studies suggest that a more even distribution of presence records could reduce the impact of spatial autocorrelation between environmental variables on model predictions (Segurado et al. 2006; Dormann 2007; Veloz 2009; Anderson & Raza 2010). For these reasons, I used the trimmed data sets for all subsequent niche modeling.. Environmental data layers I used 19 bioclimatic variables from WorldClim (Hijmans et al. 2005) to model the potential distributions of P. webbianus suffusus and P. alphonsianus. These 19 variables can be broadly categorized into temperature-related and precipitation-related factors (Table 1). The WorldClim dataset was generated using interpolation from monthly climatic data between 1950 and 2000 collected from weather stations around the world, and has been widely used in ecological niche modeling. The spatial resolution of WorldClim data set used in this study is 2’30”, which is approximately 4.65*4.65 kilometers or 21.6 square kilometers at the equator. Although WorldClim data sets are available at higher resolution, the movement ability of P. webbianus could range from 100 m to 2 km (Lee et al. 2010; S. H. Li, personal communication). Therefore,. 9.

(22) the 2’30” resolution of WorldClim should be appropriate for current study.. Modeling algorithm I used MaxEnt (Phillips et al. 2006; Phillips & Dudik 2008) to predict potential distributions of the two species. MaxEnt is based on the algorithm of maximum entropy and assumes that a species is evenly distributed within the study area (maximum entropy), and does not show any preference for specific environmental conditions. By gradually adding environmental variables (limiting factors) into the model, the even distribution will approach the actual distribution of the species. In comparison with other algorithm, MaxEnt produces relatively high AUC (Phillips et al. 2004; Elith et al. 2006; Hernandez et al. 2006; Wisz et al. 2008), and has therefore been widely applied (Elith et al. 2011). In this study, I randomly divided the presence records into 50% training data and 50% testing data. The training data set is used to build the model, and the testing data set is used to test the model’s performance. I repeated this process 50 times, each time using a random seed to divide the presence records. I set regularization at 1, maximum number of background points at 10000, iteration at 500, convergence threshold at 0.00001, and left all other settings as default values in MaxEnt. Using the average from the 50 runs with the randomly-split training and testing data sets, I obtained test AUC values that are greater than 0.7, indicating sufficient discrimination ability (Pearce & Ferrier 2000). I converted the continuous probability output from MaxEnt to a. 10.

(23) binary format using the lowest presence threshold (Pearson et al. 2007). And by overlaying the binary potential distributions of the two species, I was able to determine the sympatric zone. The lowest presence threshold method converts the estimated continuous probabilities that are greater than or equal to the lowest estimated probability among all cells with actual presence records in the training dataset into “suitable”(or probability = 1), and the rest to “unsuitable” (or probability = 0). This is also referred to as the minimum training presence threshold in MaxEnt. The process of converting continuous probabilities to a binary format can be very sensitive to the choice of the thresholds. However, the lowest presence threshold has been suggested to be particularly suitable for non-migratory species, because for these species, it is likely to represent the worst environmental conditions in which they can sustain (Pearson et al. 2007). Preliminary runs using eight different thresholds indicated that five of them produced consistent results on the spatial distribution and niche partitioning tests. (Appendix VI). The remaining three thresholds,. however, produced extremely small sympatric zones that prevented any further analyses (Appendix VI). Environmental variables are often spatially-correlated. In addition, a model with too many variables can be overfitted (Peterson 2007). However, when information on key environmental variables that limit a species’ distribution is not available (e.g. physiological tolerance) (Elith et al. 2011), it is a conservative approach to include all possible variables. Therefore, I included all 19 bioclimatic variables for the niche modeling. In order to evaluate the potential issues with overfitting, I compared the. 11.

(24) predicted distributions between the saturated models with all 19 variables, and the less saturated models with only five variables, and both produced similar predicted distributions (Appendix VII).. Test of random spatial distribution In order to test whether or not P. webbianus suffusus and P. alphonsianus are randomly distributed within the sympatric zone, I used the average nearest neighbor distance (Clark & Evans 1954; Goreaud & Pelissier, 2003) implemented in Geospatial Modelling Environment (Beyer 2011) as the measure of the degree of spatial randomness in species presence records within the sympatric zone (Figure 5). First, I generated 200 sets of pseudo-presence records that are randomly distributed within the sympatric zone for each of the two species using the Monte Carlo method. The number of the pseudo-presence records generated for each data set is equal to the number of actual presence records for each of the two species, and the same data trimming rule was applied (the pseudo-presence records are randomly chosen from points that are at least 0.5 decimal degrees apart from one another). I then calculated the average nearest neighbor distance for each of the 200 pseudo-presence data sets, and used them to draw a null distribution under the assumption of random distribution for each of the two species. Finally, I compared actual average nearest neighbor distance to the null distribution. If the actual distance is lower than the 5th percentile of the null distribution, I concluded that the spatial distribution of the species within the sympatric zone is non-random.. 12.

(25) Tests of niche equivalency and niche partitioning In order to test if there is niche partitioning between P. webbianus suffusus and P. alphonsianus within the sympatric zone, I performed a niche identity test (Warren et al. 2008, 2010), followed by a discriminant analysis of all environmental variables to investigate the key niche dimensions that separate the two species. For these two analyses, I used the trimmed presence records of P. webbianus suffusus (n = 57) and P. alphonsianus (n = 41) within the sympatric zone (Figure 5). The niche identity test uses the continuous probability outputs from MaxEnt to calculate D and I values (Schoener 1968; Vaart 2000):. D p X , pY   1 . 1  pX ,i  pY ,i 2 i 1 I  p X , pY   1  i p X ,i  pY ,i 2. . equation 1. . 2. equation 2. p X ,i and pY ,i are the normalized suitability score of species X and Y in cell i. Therefore, for the entire study area,. i p X ,i  1. and. i pY ,i  1. A. value of zero for either the D or I value indicates complete niche partitioning ( p X ,i  pY ,i  0 for any given cell). A value of 1 for either D or I value indicates total niche overlap ( p X ,i  pY ,i for any given cell). If the actual D and I values based on the presence records are statistically smaller than the mean D and I values from the null distributions assuming an identical niche between the two species, I conclude that the species have different niches. The actual D and I values for P. webbianus suffusus and P. alphonsianus were generated by running. 13.

(26) ecological niche modeling for the sympatric zone only (because this is the geographic area where competition between the two species is most likely to have occurred), using the entire trimmed dataset. This procedure yields a suitability map within the sympatric zone for each of the two species in the form of continuous probability, based on which actual D and I values can be computed following equations 1 and 2. The null distributions of the D and I values were created by pooling the presence records of the two species, and then randomly splitting all records into two sets with the same number of the original presence records for each of the two species. The two sets of presence records were used to simulate two new hypothetical species that have an identical niche. Finally, by repeating the randomization of presence records, and running ecological niche modeling for each randomized data set for 200 times, I generated the null distributions of D and I values under the assumption of niche equivalency. The discriminant analysis (e.g. Tinker et al. 2008; Rodder & Engler 2011) was performed using the trimmed data set of P. webbianus suffusus and P. alphonsianus within the sympatric zone. The values of the 19 bioclimatic variables for all of the cells with actual presence records were entered into the discriminant analysis. The resulting canonical functions that effectively discriminate between the two species reflect the potential niche dimensions along which P. webbianus suffusus and P. alphonsianus differentiated (SPSS Version 19).. 14.

(27) Test of secondary contact In order to test the secondary contact hypothesis, I modeled the distributions of these two species during the last glacial maximum, which is approximately 21,000 years ago, immediately following the divergence of the two species. By projecting the models of current distributions onto the bioclimatic data layers of the last glacial maximum (Collins et al. 2006) (Table 1), I generated paleo- distributions of the two species. A substantial sympatry during the last glacial maximum that overlaps with current-day sympatry would indicate a low likelihood of secondary contact. On the other hand, a lack of sympatry during the last glacial masimum would suggest allopatric speciation and support the secondary contact hypothesis.. 15.

(28) Results Current-day sympatric zone The predicted distribution of P. webbianus suffusus extended from northern Vietnam to northern China (the predicted area is about 3,602,198 km2), largely reflecting the actual presence records except for the western Guizhou province which was predicted to be suitable but contained no presence records of P. webbianus suffusus (Figure 3a). The predicted distribution of P. alphonsianus was much more restricted, covering eastern Sichuan province, Guizhou province, Yunnan province, western Guangxi province, and northern Vietnam (the predicted area is about 1,273,098 km2) (Figure 3b). The actual presence records of P. alphonsianus were even more restricted than the predicted distribution. Specifically, eastern Sichuan and Guizhou provinces, and western Yunnan province were predicted to be suitable habitats but did not have any presence records of P. alphonsianus. The AUC values of these current distribution models were acceptable for both species, with an AUC of 0.74 for P. webbianus suffusus and 0.92 for P. alphonsianus. The lowest presence threshold was 0.09 for P. webbianus suffusus, and 0.17 for P. alphonsianus. By overlaying the distributions of P. webbianus suffusus and P. alphonsianus, I produced a potential sympatric zone in eastern Sichuan province, Guizhou provinces, and western Yunnan province (area of potential sympatry is about 852,583 km2) (Figure 4a).. Non-random distribution and niche partitioning in the sympatric zone The distributions of P. webbianus suffusus and P. alphonsianus were 16.

(29) not random within the sympatric zone based on the Monte Carlo test (Figure 6). The actual average nearest neighbor distance was 0.65 decimal degrees for P. webbianus suffusus (n=57), and 0.67 decimal degrees for P. alphonsianus (n=41). The range of average nearest neighbor distance of the null distribution was 0.66 to 0.82 decimal degrees for P. webbianus suffusus, and 0.68 to 0.93 decimal degrees for P. alphonsianus. The niche identity test showed that P. webbianus suffusus and P. alphonsianus had substantial niche differentiation within the sympatric zone. The actual D and I values were 0.64 (n=57) and 0.89 (n=41), lower than the 5th percentile of D and I values of the null distributions (Figure 7). The range of null D values was 0.75 to 0.92, and that of null I values was 0.94 to 0.99. The first canonical discriminant function had 100% discriminating ability, and was the only significant function in the model (χ2 = 33.35, df = 15, P = 0.004), which produced an overall 76% correctly-classified cases (the proportions of cases correctly classified are: P. webbianus suffusus, 47/57 or 0.82; P. alphonsianus, 27/41 or 0.66; both combined, 74/98 or 0.76). There are 13 bioclimatic variables that passed the tolerance test (Table 2). The discriminant loadings of the first canonical function indicate that a higher function 1 score was correlated with less temperature seasonality and high isothermality (Table 2). Function 1 score was higher for P. alphonsianus (Appendix VIII), suggesting P. alphonsianus was more suited in areas with less seasonal fluctuations in temperatures, and relatively constant diurnal temperatures throughout the. 17.

(30) year (Figure 8).. Last glacial maximum distributions and secondary contact Last glacial maximum distributions of P. webbianus suffusus and P. alphonsianus were extensively overlapped, suggesting they were not allopatrically distributed (Figure 9). Instead, they shared a sympatric zone similar to the current day situation (Figure 4b). Therefore, it is unlikely that the non-random distributions of these two species within the sympatric zone are the result of secondary contact. The last glacial maximum distribution of P. webbianus suffusus was approximately 28% smaller than its current day distribution, restricted to the middle portion of China (Figure 9a). The last glacial maximum distribution of P. alphonsianus was 16% smaller than its current day distribution, restricted to eastern Sichuan province, Yunnan province, eastern Guangxi province, and northern Vietnam (Figure 9b) (See Appendix IX for probability paleo-distribution of P. webbianus suffusus and P. alphonsianus). The potential sympatric zone during the last glacial maximum extends from eastern Sichuan province and western Guangxi province to northern Vietnam (Figure 4b), with 89% of it overlapping with current-day sympatric zone.. 18.

(31) Discussion This study demonstrated that P. webbianus suffusus and P. alphonsianus occupy a potential sympatric zone (Figure 4), indicating that their fundamental niches are overlapped. Within this sympatric zone, the two species exhibit a non-random, segregated distribution, suggesting neither of them can fulfill their fundamental niches within the sympatric zone. The last glacial maximum distributions of P. webbianus suffusus and P. alphonsianus produced an area of potential sympatry similar to the current-day situation (Figure 4). Therefore, it is unlikely that the two species are in secondary contact. Furthermore, there is evidence for niche partitioning along temperature-related niches between P. webbianus suffusus and P. alphonsianus. Specifically, P. alphonsianus is associated with areas of lower temperature variability, and P. webbianus suffusus with areas of higher temperature variability (Figure 8). These results provide strong support for the role of competition in shaping the current parapatric distribution of P. webbianus suffusus and P. alphonsianus.. Niche partitioning and competition Competition is considered a major driving force for niche partitioning (Hutchinson 1957; Schoener 1974). In this study, within the area of potential sympatry, P. webbianus suffusus and P. alphonsianus differentiated in temperature-related niches, suggesting the possibility of condition-specific competition (Taniguchi & Nakano 2000) in driving their niche partitioning. Condition-specific competition refers to the situations where the competitive abilities of species, and their competitive 19.

(32) outcomes, are influenced by environmental conditions. For example, under a set of fixed conditions, species A might be more efficient in resource use than species B. However, when the conditions change, as often the case in natural environments, species B might become more efficient in resource use than species A (e.g. Gause & Witt 1935; Finstad et al. 2011). In my study, P. webbianus suffusus is more suitable in areas with higher temperature variability, whereas P. alphonsianus is more suitable in areas with lower temperature variability. This could be a result of increased competitive ability of P. webbianus suffusus in an environment with large temperature fluctuations, and/or increased competitive ability of P. alphonsianus in a relatively more stable environment. Other forms of competition, such as direct competition for limiting resources, or apparent competition between species with a shared natural enemy, could also drive patterns of species distributions (e.g. Lubchenco 1980) Although this study does not have data to exclude either of these possibilities, there are generally very few studies that provide support for the role of apparent competition in driving species parapatric distribution (Bull 1991 review; Sexton et al. 2009 review). A recent analysis of 100 bird species in France by Julliard et al. (2006) revealed that birds that are habitat specialists tend to aggregate with other specialist birds, and similarly, birds that are habitat generalists tend to aggregate with other generalist birds. In my study, P. alphonsianus is relatively more specialized in temperature-related niches (i.e. P. alphonsianus has a narrower niche breath in terms of less temperature. 20.

(33) seasonality and high isothermality) than P. webbianus suffusus. Therefore, their spatial segregation could also reflect this “specialist-generalist” gradient in bird communities. Theory predicts that ecological specialization should be favored by natural selection under stable environments (Wilson & Yoshimura 1994; Futuyma & Moreno 1998). In such environments, specialists are expected to be more competitive than generalists. Therefore, as a generalist ancestor species with wide geographic distribution continues to expand its range into locally stable environments, specialist species could rapidly evolve from always being more competitive than their generalist ancestor. It is interesting to note that western Guizhou province is predicted to be a suitable region for P. webbianus suffususin yet no actual presence records exist. On the other hand, this region is predicted to be unsuitable for P. alphonsianus during last glacial maximum, but they do have presence records in this region today (Figures 3b & 9b). This suggests that P. alphonsianus might have expanded their range into this region since last glacial maximum and potentially excluded P. webbianus suffususin from this region.. Ecological parapatry It is reasonable to speculate that the models might have over-predicted the fundamental niches of P. webbianus suffusus and P. alphonsianus, and that the two species do not have overlapping fundamental niches. In this case, the parapatric distribution of P. webbianus suffusus and P. alphonsianus is a result of each being limited by their own separate sets of niche requirements (ecological parapatry;. 21.

(34) Bull 1991). One important reason for over-predicting a species distribution in ecological niche modeling is the omission of key niche variables. However, previous studies have shown that species distributions of birds were strongly influenced by temperature-related conditions (Root 1988; Pigot et al. 2010), for which many variables are included in the current study. Therefore, it is not likely that key niche variables were missing from the models. In addition, considering that their divergence time is relatively short, under the assumption of niche conservatism, these two species are expected to share some overlapping fundamental niches. In fact, previous studies reported that the two species are similar in resource use (Wu & Chen 1986; Yang 2004; Robson 2007). Therefore, the hypothesis of ecological parapatry is a weaker explanation compared to the competitive exclusion hypothesis for the observed parapatric distribution of P. webbianus suffusus and P. alphonsianus.. Secondary contact and non-equilibrium distribution For recently-diverged species such as P. webbianus suffusus and P. alphonsianus, it is possible that they have not had sufficient time to reach equilibrium distribution. Therefore, even if they have a parapatric distribution today, it does not mean their equilibrium distribution would remain parapatric. Consequently, one cannot truly differentiate between parapatric distribution driven by competitive exclusion and by secondary contact. For studies using species that had been diverged a long time ago, equilibrium distribution can often be assumed. However, in order to understand the role of competition in driving niche divergence and. 22.

(35) promoting ecological speciation, it is necessary to study recently-diverged, closely-related species. In this study, I investigated the possibility of secondary contact between P. webbianus suffusus and P. alphonsianus by modeling their last glacial maximum distributions. Although ecological niche modeling has its limitations when used to predict paleo-distributions (e.g. the assumption of niche conservatism, the quality of paleo-climate data), it remains one of the most effective tools in establishing species paleo-distributions. Luckily, because the divergence of P. webbianus suffusus and P. alphonsianus is fairly recent, their niches are expected to closely resemble their common ancestor, which suggests that the assumption of niche conservatism is likely met. The results on last glacial maximum distributions suggest that P. webbianus suffusus and P. alphonsianus had already shared a potential sympatric zone that covers a similar area to their current-day sympatric zone soon after their divergence. In fact, Araujo and Pearson (2005) demonstrated that most bird species in Europe are in their equilibrium distributions, possibly as a result of the stronger dispersal ability of birds. Therefore, P. webbianus suffusus and P. alphonsianus should have had enough time to reach their equilibrium distribution since their divergence.. Implication and future work Competition can facilitate ecological speciation by promoting niche differentiation and expansion (e.g. Rosenzweig 1978; Schluter 1994; Bolnick 2001; Agashe & Bolnick 2010). To my knowledge, this is one of the few studies that demonstrated rapid niche differentiation along. 23.

(36) specific niche dimensions (i.e. temperature) between two closely-related species at a continental scale. This niche differentiation driven by competition may explain why P. webbianus suffusus and P. alphonsianus can exhibit a high level of morphological dissimilarity with such a short divergence time. Future studies that focus on testing the role of competition in rapid ecological speciation should yield fruitful results. In a correlational ecological niche modeling, the predicted species distribution is based on the correlations between actual species presence records and environmental data. Therefore, in theory, it should reflect the realized niche of the species (Kearney 2006). However, with an increased geographic coverage of species presence records, the predicted distribution will approach the fundamental niche of the species (Soberón & Peterson 2005; Phillips & Dudik 2008). This also means that a correlational ecological niche modeling almost always underestimates a species’ fundamental niche, which guarantees a conservative test on the role of competition in niche differentiation. It is interesting to note that the predicted distribution of P. alphonsianus covers a large area in west Yunnan province that does not have many presence records (Figure 3b). A possible explanation is that another species (P. brunneus) in this area (MacKinnon et al. 2000), which is closely related to both P. webbianus suffusus and P. alphonsianus (Yeung et al. 2011), is competitively excluding P. alphonsianus from this part of its fundamental niche. This study illustrates the potential of using correlational ecological niche modeling to test competition-related hypotheses with multiple pairs of species from a community or across a phylogenetic tree.. 24.

(37) Literature cited Acevedo P., Ward A.I., Real R. & Smith G.C. (2010). Assessing biogeographical relationships of ecologically related species using favourability functions: a case study on British deer. Divers. Distrib., 16, 515-528. Agashe D. & Bolnick D.I. (2010). Intraspecific genetic variation and competition interact to influence niche expansion. Proc. R. Soc. B, 277, 2915-2924. Anderson R.P., Peterson A.T. & Gomez-Laverde M. (2002). Using niche-based GIS modeling to test geographic predictions of competitive exclusion and competitive release in South American pocket mice. Oikos, 98, 3-16. Anderson R.P. & Raza A. (2010). The effect of the extent of the study region on GIS models of species geographic distributions and estimates of niche evolution: preliminary tests with montane rodents (genus Nephelomys) in Venezuela. J. Biogeogr., 37, 1378-1393. Araujo M.B. & Pearson R.G. (2005). Equilibrium of species’ distributions with climate. Ecography, 28, 693-695. Berendse F. (1983). Interspecific competition and niche differentiation between Plantago lanceolata and Anthoxanthum odoratum in a natural hayfield. J. Ecol., 71, 379-390. Beyer H. (2011). Geospatial modelling environment, version 0.5.5 Beta. www.spatialecology.com Bolnick D.I. (2001). Intraspecific competition favours niche width expansion in Drosophila melanogaster. Nature, 410, 463-466. Bull C.M. (1991). Ecology of parapatric distributions. Annu. Rev. Ecol. Syst., 22, 19-36. Cadena C.D. & Loiselle B.A. (2007). Limits to elevational distributions in two 25.

(38) species of emberizine finches: disentangling the role of interspecific competition, autoecology, and geographic variation in the environment. Ecography, 30, 491-504. Clark P.J. & Evans F.C. (1954). Distance to nearest neighbor as a measure of spatial relationships in populations. Ecology, 35, 445-453. Collins W.D., Bitz C.M., Blackmon M.L., Bonan G.B., Bretherton C.S., Carton J.A., Chang P., Doney S.C., Hack J.J., Henderson T.B., Kiehl J.T., Large W.G., McKenna D.S., Santer B.D. & Smith R.D. (2006). The community climate system model version 3 (CCSM3). J. Climate., 19, 2122-2143. Connor E.F. & Bowers M.A. (1987). The spatial consequences of interspecific competition. Ann. Zool. Fenn., 24, 213-226. Costa G.C., Wolfe C., Shepard D.B., Caldwell J.P. & Vitt L.J. (2008). Detecting the influence of climatic variables on species distributions: a test using GIS niche-based models along a steep longitudinal environmental gradient. J. Biogeogr., 35, 637-646. Cunningham H.R., Rissler L.J. & Apodaca J.J. (2009). Competition at the range boundary in the slimy salamander: using reciprocal transplants for studies on the role of biotic interactions in spatial distributions. J. Anim. Ecol., 78, 52-62. Davis, M.B. (1986). Climatic instability, time lags, and community disequilibrium. In: Community Ecology (eds Diamond, J. & Case, T.J.). Harper & Row, New York, pp. 269–284. Dormann C.F. (2007). Effects of incorporating spatial autocorrelation into the analysis of species distribution data. Globa.l Ecol. Biogeogr., 16, 129-138. Elith J., Graham C.H., Anderson R.P., Dudik M., Ferrier S., Guisan A., Hijmans R.J., Huettmann F., Leathwick J.R., Lehmann A., Li J., Lohmann L.G., Loiselle. 26.

(39) B.A., Manion G., Moritz C., Nakamura M., Nakazawa Y., Overton J.M., Peterson A.T., Phillips S.J., Richardson K., Scachetti-Pereira R., Schapire R.E., Soberón J., Williams S., Wisz M.S. & Zimmermann N.E. (2006). Novel methods improve prediction of species' distributions from occurrence data. Ecography, 29, 129-151. Elith J., Phillips S.J., Hastie T., Dudik M., Chee Y.E. & Yates C.J. (2011). A statistical explanation of MaxEnt for ecologists. Divers. Distrib., 17, 43-57. Finstad A.G., Forseth T., Jonsson B., Bellier E., Hesthagen T., Jensen A.J., Hessen D.O. & Foldvik A. (2011). Competitive exclusion along climate gradients: energy efficiency influences the distribution of two salmonid fishes. Global. Change. Biol., 17, 1703-1711. Futuyma, D.J. & Moreno, G. (1988). The evolution of ecological specialisation. Annu. Rev. Ecol. Syst., 19, 207-233. Gause G.F. (1934). The struggle for existence. The Williams & Wilkins company, Baltimore. Gause G.F. & Witt A.A. (1935). Behavior of mixed populations and the problem of natural selection. Am. Nat., 69, 596-609. Goreaud F. & Pelissier R. (2003). Avoiding misinterpretation of biotic interactions with the intertype K-12-function: population independence vs. random labelling hypotheses. J. Veg. Sci., 14, 681-692. Guisan A. & Zimmermann N.E. (2000). Predictive habitat distribution models in ecology. Ecol. Model., 135, 147-186. Hernandez P.A., Graham C.H., Master L.L. & Albert D.L. (2006). The effect of sample size and species characteristics on performance of different species distribution modeling methods. Ecography, 29, 773-785.. 27.

(40) Hijmans R.J., Cameron S.E., Parra J.L., Jones P.G. & Jarvis A. (2005). Very high resolution interpolated climate surfaces for global land areas. Int. J. Climatol., 25, 1965-1978. Hortal J., Jimenez-Valverde A., Gomez J.F., Lobo J.M. & Baselga A. (2008). Historical bias in biodiversity inventories affects the observed environmental niche of the species. Oikos, 117, 847-858. Hutchinson G.E. (1957). Concluding remarks. Cold Spring Harbor Symp. Quant. Biol., 22, 415-427. Julliard R., Clavel J., Devictor V., Jiguet F. & Couvet D. (2006). Spatial segregation of specialists and generalists in bird communities. Ecol. Lett., 9, 1237-1244. Kearney M. (2006). Habitat, environment and niche: what are we modelling? Oikos, 115, 186-191. Kearney M. & Porter W. (2009). Mechanistic niche modelling: combining physiological and spatial data to predict species' ranges. Ecol. Lett., 12, 334-350. Kozak K.H., Graham C.H. & Wiens J.J. (2008). Integrating GIS-based environmental data into evolutionary biology. Trends. Ecol. Evol., 23, 141-148. Lee J.W., Kim H.Y. & Hatchwell B.J. (2010). Parental provisioning behaviour in a flock-living passerine, the Vinous-throated Parrotbill Paradoxornis webbianus. J. Ornithol., 151, 483-490. Liu H.J., Tin T.S., Fang W.H., Lin W.H., Tsai M.C. & Yen C.W. (2010). The Avifauna of Taiwan. Forestry Bureau, Taipei. Lobo J.M., Jiménez-Valverde A. & Real R. (2007). AUC: a misleading measure of the performance of predictive distribution models. Global. Ecol. Biogeogr., 17, 145-151.. 28.

(41) Lu Z., Zhou F., Pan H.P. & Xu Y.L. (2006). New record of bird species in Guangxi—Paradoxornis alphonsianus. Journal of Guangxi Agricultural and Biological Science, 25, 1. Lubchenco J. (1980). Algal zonation in the New England rocky intertidal community: an experimental analysis. Ecology, 61, 333-344. MacKinnon J.R., Phillipps K. & He F. (2000). A field guide to the birds of China. Oxford University Press. Murray J.V., Goldizen A.W., O'Leary R.A., McAlpine C.A., Possingham H.P. & Choy S.L. (2009). How useful is expert opinion for predicting the distribution of a species within and beyond the region of expertise? A case study using brush-tailed rock-wallabies Petrogale penicillata. J. Appl. Ecol., 46, 842-851. Nielsen S.E., Johnson C.J., Heard D.C. & Boyce M.S. (2005). Can models of presence-absence be used to scale abundance? - Two case studies considering extremes in life history. Ecography, 28, 197-208. Pearce J. & Ferrier S. (2000). Evaluating the predictive performance of habitat models developed using logistic regression. Ecol. Model., 133, 225-245. Pearson R.G., Raxworthy C.J., Nakamura M. & Peterson A.T. (2007). Predicting species distributions from small numbers of occurrence records: a test case using cryptic geckos in Madagascar. J. Biogeogr., 34, 102-117. Pellissier L., Brathen K.A., Pottier J., Randin C.F., Vittoz P., Dubuis A., Yoccoz N.G., Alm T., Zimmermann N.E. & Guisan A. (2010). Species distribution models reveal apparent competitive and facilitative effects of a dominant species on the distribution of tundra plants. Ecography, 33, 1004-1014. Peterson A.T. (2007). Why not WhyWhere: The need for more complex models of simpler environmental spaces. Ecol. Model., 203, 527-530.. 29.

(42) Peterson A.T. (2011). Ecological niche conservatism: a time-structured review of evidence. J. Biogeogr., 38, 817-827. Phillips S.J., Anderson R.P. & Schapire R.E. (2006). Maximum entropy modeling of species geographic distributions. Ecol. Model., 190, 231-259. Phillips S.J., Dudik M. & Schapire R.E. (2004). A maximum entropy approach to species distribution modeling. In: Proceedings of the twenty-first international conference on machine learning. ACM Banff, Alberta, Canada, p. 83. Phillips S.J. & Dudik M. (2008). Modeling of species distributions with Maxent: new extensions and a comprehensive evaluation. Ecography, 31, 161-175. Pigot A.L., Owens I.P.F. & Orme C.D.L. (2010). The environmental limits to geographical range expansion in birds. Ecol. Lett., 13, 705-715. Pulliam H.R. (2000). On the relationship between niche and distribution. Ecol. Lett., 3, 349-361. Reddy S. & Davalos L.M. (2003). Geographical sampling bias and its implications for conservation priorities in Africa. J. Biogeogr., 30, 1719-1727. Ritchie E.G., Martin J.K., Johnson C.N. & Fox B.J. (2009). Separating the influences of environment and species interactions on patterns of distribution and abundance: competition between large herbivores. J. Anim. Ecol., 78, 724-731. Robson C. (2007). Handbook of the birds of the world, Vol 12. Lynx Edicions, Barcelona. Rodder D. & Engler J.O. (2011). Quantitative metrics of overlaps in Grinnellian niches: advances and possible drawbacks. Global. Ecol. Biogeogr., 20, 915-927. Root T. (1988). Energy constraints on avian distributions and abundances. Ecology, 69, 330-339.. 30.

(43) Rosenzweig M.L. (1978). Competitive speciation. Biol. J. Linn. Soc., 10, 275-289. Schoener T.W. (1968). Anolis lizards of bimini - resource partitioning in a complex fauna. Ecology, 49, 704-726. Schoener T.W. (1974). Resource partitioning in ecological communities. Science, 185, 27-39. Schluter D. (1994). Experimental evidence that competition promotes divergence in sdaptive radiation. Science, 266, 798-801. Segurado P., Araujo M.B. & Kunin W.E. (2006). Consequences of spatial autocorrelation for niche-based models. J. Appl. Ecol., 43, 433-444. Sexton J.P., McIntyre P.J., Angert A.L. & Rice K.J. (2009). Evolution and ecology of species range limits. Annu. Rev. Ecol. Evol. S., 40, 415-436. Seoane J., Bustamante J. & Diaz-Delgado R. (2005). Effect of expert opinion on the predictive ability of environmental models of bird distribution. Conserv. Biol., 19, 512-522. Soberón J. & Peterson A.T. (2005). Interpretation of models of fundamental ecological niches and species’ distributional areas. Biodivers. Inform., 2, 1-10. Tinker M.T., Bentall G. & Estes J.A. (2008). Food limitation leads to behavioral diversification and dietary specialization in sea otters. Proc. Natl. Acad. Sci. USA, 105, 560-565. Taniguch Y. & Nakano S. (2000). Condition-specific competition: Implications for the altitudinal distribution of stream fishes. Ecology, 81, 2027-2039. Vaart A.W. (2000). Asymptotic statistics. Cambridge University Press. Veloz S.D. (2009). Spatially autocorrelated sampling falsely inflates measures of accuracy for presence-only niche models. J. Biogeogr., 36, 2290-2299. Violle C., Nemergut D.R., Pu Z.C. & Jiang L. (2011). Phylogenetic limiting similarity. 31.

(44) and competitive exclusion. Ecol. Lett., 14, 782-787. Warren D.L., Glor R.E. & Turelli M. (2008). Environmental niche equivalency versus conservatism: quantitative approaches to niche evolution. Evolution, 62, 2868-2883. Warren D.L., Glor R.E. & Turelli M. (2010). ENMTools: a toolbox for comparative studies of environmental niche models. Ecography, 33, 607-611. Wethey D.S. (2002). Biogeography, competition, and microclimate: The barnacle Chthamalus fragilis in New England. Integr. Comp. Biol., 42, 872-880. Wiens J.J., Ackerly D.D., Allen A.P., Anacker B.L., Buckley L.B., Cornell H.V., Damschen E.I., Davies T.J., Grytnes J.A., Harrison S.P., Hawkins B.A., Holt R.D., McCain C.M. & Stephens P.R. (2010). Niche conservatism as an emerging principle in ecology and conservation biology. Ecol. Lett., 13, 1310-1324. Wiens J.J. & Graham C.H. (2005). Niche conservatism: Integrating evolution, ecology, and conservation biology. Annu. Rev. Ecol. Evol. S., 36, 519-539. Wilson, D.S. & Yoshimura, J. (1994). On the coexistence of specialists and generalists. Am. Nat., 144, 692-707. Wisz M.S., Hijmans R.J., Li J., Peterson A.T., Graham C.H., Guisan A. & Distribut N.P.S. (2008). Effects of sample size on the performance of species distribution models. Divers. Distrib., 14, 763-773. Wu Z. & Chen K. (1986). The Avifauna of Guizhou. Guizhou People's Publishing House. Yang L. (2004). The avifauna of Yunnan China: Passeriformes. Yunnan Science and Technology Press. Yeung C.K.L., Lin R.C., Lei F.M., Robson C., Hung L.M., Liang W., Zhou F.S., Han. 32.

(45) L.X., Li S.H. & Yang X.J. (2011). Beyond a morphological paradox: Complicated phylogenetic relationships of the parrotbills (Paradoxornithidae, Aves). Mol. Phylogenet. Evol., 61, 192-202.. 33.

(46) Tables Table 1. The 19 bioclimate variables in WorldClim, and their current-day and last glacial maximum ranges for the study area. Variable. Present-day range. LGMa range. Bio1: Annual mean temperature. -7.4~26.7 (°C). -11.1~25.1 (°C). Bio 2: Mean diurnal range Bio 3: Isothermalityb Bio 4: Temperature seasonalityc Bio 5: Maximum temperature of warmest month Bio 6: Minimum temperature of coldest month. 4.2~16.4 (°C) 17~58 (°C *100) 2.17~11.77 (°C) 3.9~37.4 (°C). 4.9~39.2 (°C) 18~77 (°C *100) 2.16~11.75 (°C) 0.2~43.1 (°C). -26.2~18.8 (°C). -42.7~13.8 (°C). Bio 7: Temperature annual range Bio 8: Mean temperature of wettest quarter. 13.1~48.2 (°C) -0.7~29.5 (°C). 17.1~63.8 (°C) -6.6~27.4 (°C). Bio 9: Mean temperature of driest quarter Bio 10: Mean temperature of warmest quarter Bio 11: Mean temperature of coldest quarter Bio 12: Annual precipitation Bio 13: Precipitation of wettest month Bio 14: Precipitation of driest month Bio 15: Precipitation seasonalityd Bio 16: Precipitation of wettest quarter Bio 17: Precipitation of driest quarter. -16.2~27.9 (°C) -0.7~30.2 (°C) -16.2~23.3 (°C) 167~4589 (mm) 49~1108 (mm) 0~216 (mm) 11~150 104~2774 (mm) 2~773 (mm). -21.6~24.0 (°C) -4.3~28.8 (°C) -21.6~21.3 (°C) 142~5531 (mm) 43~1353 (mm) 0~267 (mm) 11~152 91~3386 (mm) 0~975 (mm). Bio 18: Precipitation of warmest quarter Bio 19: Precipitation of coldest quarter. 97~2774 (mm) 3~995 (mm). 85~3386 (mm) 1~1243 (mm). a. LGM = last glacial maximum Isothermality = (Bio2/Bio7)*100 c Temperature seasonality is the standard deviation of the monthly temperatures among the 12 months d Precipitation seasonality is the coefficient of variation (CV) of the monthly precipitations among the 12 months. b. 34.

(47) Table 2. The structure matrix of the first canonical discriminant function and the discriminant loadings of original bioclimatic variables. Variable name. Function1 loading. Isothermality Temperature seasonality. .548 -.536. Mean diurnal range Temperature annual rangea Precipitation seasonality Precipitation of warmest quarter Precipitation of coldest quartera Precipitation of driest quarter Maximum temperature of warmest month. .460 -.396 .323 -.258 -.224 -.215 -.190. Mean temperature of warmest quartera Mean temperature of coldest quartera. -.175 .167. Precipitation of warmest quarter Mean temperature of driest quarter Annual precipitation Precipitation of wettest quarter Precipitation of wettest month Minimum temperature of coldest month Annual mean temperature Mean temperature of wettest quarter. .151 .151 -.146 .095 .078 .078 .040 -.024. a. These variables failed minimum tolerance criteria (0.001) and were excluded in the final model.. 35.

(48) Figures. Figure 1. The outcomes of competitive exclusion between two species. The circles denoted the fundamental niches of the two species, with the green and red dots representing their respective realized niches. The area of potential sympatry is where the fundamental niches overlap. The first outcome is that the two species are randomly distributed within the sympatric zone (a), indicating a lack of competitive exclusion. The second and third outcomes are that, one species is completely absent from the sympatric zone (b), or the two species are segregated within the sympatric zone (c), both indicating competitive exclusion (modified from Costa et al. 2008). 36.

(49) Figure 2. The raw (a) and trimmed (b) data sets of species presence records. The study area was shaded. The green dots denoted P. webbianus (abbreviate as “Pws”), and the red dots denoted P. alphonsianus (abbreviate as “Pa”). The blue dots represent locations where both species were recorded.. 37.

(50) Figure 3. The binary distributions of P. webbianus and P. alphonsianus. The binary distributions of P. webbianus (a) and P. alphonsianus (b) were converted from the probability distribution using the lowest presence threshold of 0.09 and 0.17, respectively. Paradoxornis webbianus was abbreviated as “Pws” and P. alphonsianus as “Pa.” 38.

(51) Figure 4. The potential sympatric zone of P. webbianus and P. alphonsianus. The current-day sympatric zone (a) covered Chengdu Plain, Guizhou province, and western Guangxi province. The last glacial maximum (LGM) sympatric zone (b) covered Chengdu Plain, western Guangxi province, and northern Vietnam. P. webbianus was abbreviated as “Pws” and P. alphonsianus as “Pa.” 39.

(52) Figure 5. The presence records of P. webbianus and P. alphonsianus within current-day sympatric zone. There were 57 actual presence records for P. webbianus, and 41 for P. alphonsianus within current-day sympatry. P. webbianus was abbreviated as “Pws” and P. alphonsianus as “Pa.”. 40.

(53) Figure 6. The actual and null average nearest neighbor distance (ANND) of P. alphonsianus and P. webbianus. The actual ANND values (blue arrows) of P. alphonsianus (a) and P. webbianus (b) were significantly lower than the null distributions.. 41.

(54) Figure 7. The actual and null D and I values of P. webbianus and P. alphonsianus. The actual D value of 0.64 was lower than the 5th percentile of the null D distribution based on 200 replicates (a). The actual I value of 0.89 was lower than the 5th percentile of the null I distribution based on 200 replicates (b).. 42.

(55) Figure 8. The ranges of temperature seasonality and isothermality occupied by P. webbianus and P. alphonsianus within the sympatric zone. Temperature seasonality and isothermality were the two bioclimatic variables with the highest loadings in the first canonical discriminant function. Paradoxornis webbianus (green circles, abbreviated as “Pws”) occupied areas with a higher temperature seasonality, and P. alphonsianus (red circles, abbreviated as “Pa”) occupied areas with a higher isothermality.. 43.

(56) Figure 9. The binary distributions of P. webbianus and P. alphonsianus during last glacial maximum (LGM). The binary distributions of P. webbianus (a) and P. alphonsianus (b) were converted from the probability distribution using the lowest presence threshold of 0.09 and 0.17, respectively.. 44.

(57) Appendix Appendix I. The presence records of P. webbianus and P. alphonsianus. The presence records included all subspecies. The green dots denoted P. webbianus (abbreviated as “Pws”), and the red dots denoted P. alphonsianus (abbreviated as “Pa”). The blue dots represented locations where both species were recorded.. 45.

(58) Appendix II. Map of the study area and contact region between P. webbianus and P. alphonsianus. The contact region included Sichuan, Guizhou, Yunnan, and Guangxi.. 46.

(59) Appendix III. Preliminary tests on the changes in test AUC values with increasing study areas for P. webbianus and P. alphonsianus. Raw presence records and default parameter settings in MaxEnt were used to generate the AUC values. A total of eight different study areas were tested, beginning with the smallest study area based on the minimum convex polygon (MCP), and increasing gradually by the amounts of buffers added to the MCP.. Study area. AUC value P. webbianus. P. alphonsianus. MCP + 0 decimal degree. 0.746. 0.934. MCP + 1 decimal degree. 0.776. 0.94. MCP + 2 decimal degrees. 0.802. 0.943. MCP + 3 decimal degrees. 0.816. 0.947. MCP + 4 decimal degrees. 0.838. 0.953. MCP + 5 decimal degrees. 0.854. 0.957. MCP + 6 decimal degrees. 0.867. 0.962. MCP + 7 decimal degrees. 0.883. 0.963. 47.

(60) Appendix IV. Predicted distributions of P. webbianus and P. alphonsianus with and without low resolution data points. A spatial resolution of more than 200 km2 was categorized as low resolution data points. The predicted distributions of P. webbianus with all data points (a) and with low resolution data points excluded (b) were qualitatively similar. The predicted distributions of P. alphonsianus with all data points (c) and with low resolution data points excluded (d) were also qualitatively similar.. 48.