付記する学位プログラム名: 霊長類学・ワイルドライフサイエンス・リーディング大学院 / 京都大学 / 新制・課程博士 / 博士(理学) / 甲第23406号 / 理博第4741号 / 新制||理||1680(附属図書館) / 京都大学大学院理学研究科生物科学専攻 / (主査)教授 平田 聡, 教授 村山 美穂, 教授 伊谷 原一 / 学位規則第4条第1項該当 / Doctor of Science / Kyoto University / DFAM

Few studies utilise modern species distribution data and modeling to make predictions for examining potential fossil localities. Instead, species distribution modeling is often used for palaeoenvironmental interpretations. Using palaeoclimate data to model potential past distributions for a species provides a prediction showing areas where its fossil remains may be found. In this study, the current, Last Glacial Maximum, and Last Interglacial potential distributions of the arid steppe-obligate saiga antelope (Artiodactyla: Bovidae, Saiga tatarica) were modeled using the species distribution model Maxent. Few fossil records exist, but available fossil locality records were used to validate both palaeo models, resulting in speculative predictions about where the saiga may have lived. Known fossil localities of saiga from the Last Glacial Maximum time period were located within predicted moderately suitable environments, while four of seven Last Interglacial fossil localities were located within predicted moderately suitable environments, suggesting that models can accurately identify areas where fossils for the saiga can be found. Specifically, these models suggest saiga fossils may be located in northwestern and northeastern China, the western and central regions of the Middle East, and southern Alaska. The predicted areas in northeastern China are of particular interest because saiga fossils have not been identified in this region, but some palaeontologists theorize that northeast China may have been suitable for saiga in the past. The models lend credence to this argument.

While range-wide population declines have prompted extensive research on greater sage-grouse (Centrocercus urophasianus), basic information about southern periphery populations, such as the Bald Hills population in southern Utah, has not been documented. The objective of this research was to determine habitat preferences and space use patterns of the Bald Hills sage-grouse population which occurs in an area of high potential for renewable energy development. I tracked 66 birds via VHF telemetry in 2011 and 2012 and surveyed vegetation plots throughout the study area. I found that the population was primarily one-stage migratory with seasonal distributions that did not correspond well with previously developed suitable habitat maps (based on local biologist knowledge and lek data) for all seasons; I also found that mean home range sizes ranged from 82 km2 to 157 km2. Nesting hens did not select for any measured vegetation characteristics within the study area, while brood-rearing hens selected for high forb cover. Birds at summer sites (non-reproductive bird locations during the summer season) selected for greater grass and forb cover and lower shrub cover compared with random sites. Overall, Bald Hills sage-grouse used areas with greater shrub canopy cover and lower grass and forb cover than recommended in habitat guidelines. Ten predictor variables were used to model suitable seasonal habitat using Maximum Entropy (maxent). All models were created for the Bald Hills population and projected to the Bureau of Land Management Cedar City Field Office management area and produced excellent model fit (AUC > 0.900). The Bald Hills population had similar nesting and winter habitat preferences as other populations but different brood-rearing and summer habitat preferences. I found local management techniques to be an important driver of seasonal habitat selection; birds selected for areas that had undergone habitat treatments (such as broadcast burn and crushing) within the previous 10 years. My results indicated the Bald Hills periphery population occupies marginal habitat and has adapted unique seasonal habitat preferences. Managers of isolated, fringe, and low-density populations should develop locally specific management guidelines to address the unique adaptations and ensure the persistence of these populations.

Several reports of North American bumble bee (Bombus Latreille) decline have been documented across the continent, but no study has fully assessed the geographic scope of decline. In this study I discuss the importance of Natural History Collections (NHC) in estimating historic bumble bee distributions and abundances, as well as in informing current surveys. To estimate changes in distribution and relative abundance I compare historic data assembled from a >73,000 specimen database with a contemporary 3-year survey of North American bumble bees across 382 locations in the contiguous U.S.A. Based on my results, four historically abundant bumble bees, B. affinis, B. occidentalis, B. pensylvanicus and B. terricola, have declined by 72 - 96% relative abundance across their native distribution, while B. bifarius, B. bimaculatus, B. impatiens, and B. vosnesenskii appear to be relatively stable. Finally, I provide some notes on the distribution, abundance, and frequency of Nosema bombi infections in Alaskan B. occidentalis.

No description available.

The developments of species distribution modeling techniques have brought new opportunities in the field of biological invasion management. In particular, statistical niche modeling for spatio-temporal predictions of species’ distribution is a widely spread tool that has proved its efficiency. The main purpose of this Master thesis is to study applicability of species distribution modeling to invasive alien species, with the aim of supporting efficient decision-making for their prevention. Some research questions are: how useful can species distribution modeling be for invasives’ prevention? Is distribution modeling technically feasible in the case of invasive species? What types of techniques are recommended to model distributions of IAS? What are the limits of such a tool? The methods employed to answer these questions are literature review and expert advice. I found that species distribution models can provide risk maps which are necessary to enable effective invasive alien species’ prevention. However intrinsic characteristics of invasives introduce uncertainties in the predictions made. Consequently several preliminary analyses should be conducted before applying the distribution model. Finally recommendations were made on the most appropriate distribution modeling technique to use depending on the urgency of the situation and the availability of data. / Utvecklingen av metoder för modelering av artdistribution har medfört nya möjligheter inom området hantering av biologiska invasioner. Statistisk nischmodelering för spatio-temporala förutsägelser av arters distribution är ett väl använt verktyg som har visat sig vara effektivt. Det övergripande målet med det här arbetet har varit att studera hur lämpad artmodelering är vid förebyggande av invasioner av främmande arter. Det har även undersökts huruvida metoden kan bidra till bättre och enklare beslutsfattande när det kommer till att förhindra sådana invasioner. Forskningsfrågorna lyder: hur användbart är fördelningsmodelering för förebyggande av spriding av invasiva arter? Är distributionsmodelering tekniskt genomförbar när det gäller invasiva arter? Vilka olika tekniker rekommenderas för att modelera spridningen av invasiva arter? Vilka begräsningar har modelerna? De metoder som används är litteraturöversikt och expertråd. Resultaten visar att artdistributionsmodelering kan bidra till att sammanställa riskkartor som är nödvändiga för att möjliggöra ett förebyggande arbete. Men speciella egenskaper hos de invasiva arterna som är svåra att förutse skapar osäkerheter i resultatet. Därför kan preliminära analyser med fördel genomföras innan modelering. I slutsatserna återfinns rekommendationer för vilken distributionsmodelteknik man bör använda, beroende av hur brådskande situationen är och om data finns tillgängligt.

Spatial information about invasive species abundance is critical for estimating impact and understanding risk to ecosystems and economies. Unfortunately, at landscape and regional scales, most distribution datasets provide limited information about abundance. However, national and regional invasive plant occurrence datasets are increasingly available and spatially extensive. We aim to test whether the frequency of these point occurrences can be used as a proxy for abundance of invasive plants. We compiled both occurrence and abundance data for nine regionally important invasive plants in the northeast US using a combination of herbarium records, surveys of expert knowledge, and various invasive species spatial databases. We integrated all available abundance information based on infested area, percent cover, or qualitative descriptions into abundance rankings ranging from 0 (absent) to 4 (highly abundant). Within equal area grid cells of 800 m, we counted numbers of occurrence points and used an ordinal regression to test whether higher numbers of occurrence points were positively correlated with abundance rankings. We compiled a total 49,341 occurrence points in 18,533 cells, of which 12,183 points (25%) within 4,278 cells (32%) had associated abundance information. In six of nine study species we found slight but significant positive overall relationships between abundance rank and occurrence frequency at high abundance ranks. However, at low abundance rankings the relationship tended to be negative and the magnitude of the overall difference in occurrence frequency was too small to be relevant to management. My results suggest that currently available occurrence datasets are unlikely to serve as effective proxies for abundance, and models derived from invasive plant occurrence datasets should not be interpreted as indicative of plant abundance and associated impact. Increased efforts to collect and report invasive species abundance information, and/or higher densities of occurrence points in heavily infested areas are strongly needed for regional scale assessments of potential abundance and associated impact.

A modelagem de distribuição de espécies tem como objetivo induzir um modelo para predizer a distribuição potencial de uma dada espécie. O modelo é projetado em um mapa de distribuição potencial que representa a probabilidade da presença da espécie em cada ponto. Esse processo de indução está relacionado com a estimativa do nicho fundamental da espécie, através da busca por relações entre dados georreferenciados de ocorrência da espécie e variáveis ambientais. Vários algoritmos de modelagem podem ser utilizados nessa tarefa. Oferecer diversos algoritmos pode tornar as ferramentas de modelagem mais completas. Porém, surge uma questão importante: qual algoritmo de modelagem escolher? Essa questão está relacionada com o desempenho preditivo das técnicas implementadas pelos algoritmos. Nesse contexto, o objetivo principal do trabalho foi organizar e especificar um método de análise de desempenho preditivo dos algoritmos de modelagem de distribuição de espécies. Através do método proposto é possível ter uma visão completa, estruturada e sistemática das etapas previstas em projetos de análise de desempenho preditivo dos algoritmos. O método pode ser utilizado como referência em estudos de validação de novos algoritmos, de comparação entre técnicas e na seleção de um ou mais algoritmos de modelagem. Como estudo de caso, o método proposto foi adotado nos testes de validação de um algoritmo baseado em Redes Neurais, desenvolvido e integrado ao framework openModeller, através da comparação com outros algoritmos já utilizados na modelagem. Além da própria validação, os testes tiveram como objetivo demonstrar a aplicabilidade do método. Os resultados mostraram que o algoritmo de Redes Neurais apresentou desempenho semelhante ao desempenho dos demais algoritmos, tendo sido, portanto, validado como adequado à tarefa de modelagem. Ainda no contexto da pesquisa, um algoritmo baseado na técnica de amostragem denominada Jackknife foi integrado ao openModeller, para aplicação na etapa de pré-análise. Testes relacionados com o tempo de execução foram realizados e uma versão paralela desse algoritmo foi desenvolvida. / The species distribution modeling aim is to induce a model to predict the potential distribution of a given species. The model is projected onto a potential distribution map that represents the presence probability of the species at each point. This induction process is related to the fundamental niche estimation of the species, through the search for relationships between georeferenced data of species occurrence and environmental variables. Several modeling algorithms can be used for this task. Providing different algorithms can make the modeling tools more complete. However, an important question arises: what modeling algorithm to choose? This issue is related to the predictive performance of the techniques implemented by the algorithms. In this context, the aim of this research was to organize and to specify a predictive performance analysis method of the species distribution modeling algorithms. Through the proposed method, it is possible to have a complete and structured vision of the steps in the planning of predictive performance analysis of the algorithms. The method may be used as a reference in validation studies of new algorithms, in comparison among techniques and in choosing one or more modeling algorithms. As a case study, the proposed method was adopted in the validation tests of an algorithm based on Neural Networks, developed and integrated into the openModeller framework, which was compared with other algorithms already used in modeling. Besides the validation itself, the tests intended to demonstrate the applicability of the method. The results showed that the Neural Networks algorithm presented similar performance to those of other algorithms and was validated as adequate to the modeling task. Still in the research context, an algorithm based on a sampling technique called the Jackknife was integrated to the openModeller, to be applied in the pre-analysis step. Tests related to the running time were carried out and a parallel version of this algorithm was developed.

O tamanduá-bandeira (Myrmecophaga tridactyla) se distribui amplamente ao longo da região Neotropical, porém é provável que esteja extinto da maior parte de sua extensão original, notadamente na América Central e nos limites austrais de sua distribuição. O táxon está ameaçado de extinção globalmente (IUCN) e também em âmbito nacional. Embora historicamente a espécie ocorra em todos os biomas brasileiros, hoje é considerada extinta nos Pampas, quase extinta na Mata Atlântica, sendo que na Caatinga sua presença necessita de confirmação e no Cerrado suas populações vem sofrendo drásticas reduções. Atualmente não há estudos de revisão da distribuição da espécie nos biomas brasileiros, tão pouco foi avaliado se as áreas mais adequadas à espécie estão sendo protegidas e o conhecimento existente é insuficiente para adotar estratégias de conservação adequadas. Dessa maneira, o presente estudo teve como principal objetivo modelar a distribuição potencial e atual do tamanduá-bandeira no Brasil e nos biomas brasileiros, a fim de identificar quais variáveis preditoras melhor explicam a ocorrência da espécie em diferentes escalas. Além disso, a partir dos modelos de distribuição atual, os biomas foram avaliados quanto à adequabilidade ambiental (i.e. probabilidade de presença) e foram realizadas uma análise de lacunas e a identificação de áreas prioritárias para a conservação. A distribuição potencial do tamanduá-bandeira foi melhor explicada em escala continental, por variáveis bioclimáticas (sazonalidade de temperatura e precipitação) e topográficas (altitude), enquanto que a distribuição atual foi bem explicada nas duas escalas, por variáveis de uso e cobertura da terra (porcentagens de cobertura arbórea, de silvicultura e de cana-de-açúcar). O Cerrado foi o bioma de maior adequabilidade ambiental à espécie, seguido da Amazônia, Pantanal, Mata Atlântica e Caatinga, sendo que não foram obtidos registros recentes para os Pampas. Menos de 10% da distribuição atual do tamanduá-bandeira no Cerrado e Pantanal encontra-se protegida por Unidades de Conservação, existindo uma lacuna parcial de conservação. Áreas prioritárias para a espécie incluem um corredor central no Cerrado, grande parte do Pantanal e áreas de transição (ecótonos) com outros biomas. Os resultados obtidos neste estudo permitiram preencher lacunas de conhecimento acerca da distribuição do tamanduá-bandeira, bem como dar suporte para o planejamento de sua conservação. / The giant anteater (Myrmecophaga tridactyla) is widely distributed throughout the Neotropical region, but is probably extinct from most of its range, notably in Central America and the southern limits of its distribution. The species is listed as Vulnerable on the IUCN and national Red Lists. Although historically present in all Brazilian biomes, there are no studies reviewing its distribution, nor has it been evaluated if the Brazilian federal conservation units are protecting the areas most suitable to the species. Thus, the aim of this study was to model the potential and current distribution of the giant anteater in Brazil and Brazilian biomes, to identify which predictor variables best explain the occurrence of the species at different scales. Current distribution models were used to evaluate the biomes environmental suitability (i.e. probability of presence) and a gap analyses were performed. Also, priority areas for conservation were identified. The potential distribution of the anteater was better explained on a continental scale by bioclimatic (seasonality of temperature and precipitation) and topographic (altitude) variables, while the current distribution was well predicted in both scales, by land cover variables (percentages of tree cover, silviculture, and sugarcane). The Cerrado was the biome of greater environmental suitability to the species, followed by the Amazon, the Pantanal, the Atlantic Forest and the Caatinga. No recent records were obtained for the Pampas. Conservation units protect less than 10% of the current distribution of the giant anteater in the Cerrado and Pantanal. Priority areas for the species include a central corridor in the Cerrado, much of the Pantanal and ecotones. The results obtained in this study helped to fill knowledge gaps on the distribution of the giant anteater in Brazil, supporting actions for its conservation.

A modelagem de distribuição de espécies tem como objetivo induzir um modelo para predizer a distribuição potencial de uma dada espécie. O modelo é projetado em um mapa de distribuição potencial que representa a probabilidade da presença da espécie em cada ponto. Esse processo de indução está relacionado com a estimativa do nicho fundamental da espécie, através da busca por relações entre dados georreferenciados de ocorrência da espécie e variáveis ambientais. Vários algoritmos de modelagem podem ser utilizados nessa tarefa. Oferecer diversos algoritmos pode tornar as ferramentas de modelagem mais completas. Porém, surge uma questão importante: qual algoritmo de modelagem escolher? Essa questão está relacionada com o desempenho preditivo das técnicas implementadas pelos algoritmos. Nesse contexto, o objetivo principal do trabalho foi organizar e especificar um método de análise de desempenho preditivo dos algoritmos de modelagem de distribuição de espécies. Através do método proposto é possível ter uma visão completa, estruturada e sistemática das etapas previstas em projetos de análise de desempenho preditivo dos algoritmos. O método pode ser utilizado como referência em estudos de validação de novos algoritmos, de comparação entre técnicas e na seleção de um ou mais algoritmos de modelagem. Como estudo de caso, o método proposto foi adotado nos testes de validação de um algoritmo baseado em Redes Neurais, desenvolvido e integrado ao framework openModeller, através da comparação com outros algoritmos já utilizados na modelagem. Além da própria validação, os testes tiveram como objetivo demonstrar a aplicabilidade do método. Os resultados mostraram que o algoritmo de Redes Neurais apresentou desempenho semelhante ao desempenho dos demais algoritmos, tendo sido, portanto, validado como adequado à tarefa de modelagem. Ainda no contexto da pesquisa, um algoritmo baseado na técnica de amostragem denominada Jackknife foi integrado ao openModeller, para aplicação na etapa de pré-análise. Testes relacionados com o tempo de execução foram realizados e uma versão paralela desse algoritmo foi desenvolvida. / The species distribution modeling aim is to induce a model to predict the potential distribution of a given species. The model is projected onto a potential distribution map that represents the presence probability of the species at each point. This induction process is related to the fundamental niche estimation of the species, through the search for relationships between georeferenced data of species occurrence and environmental variables. Several modeling algorithms can be used for this task. Providing different algorithms can make the modeling tools more complete. However, an important question arises: what modeling algorithm to choose? This issue is related to the predictive performance of the techniques implemented by the algorithms. In this context, the aim of this research was to organize and to specify a predictive performance analysis method of the species distribution modeling algorithms. Through the proposed method, it is possible to have a complete and structured vision of the steps in the planning of predictive performance analysis of the algorithms. The method may be used as a reference in validation studies of new algorithms, in comparison among techniques and in choosing one or more modeling algorithms. As a case study, the proposed method was adopted in the validation tests of an algorithm based on Neural Networks, developed and integrated into the openModeller framework, which was compared with other algorithms already used in modeling. Besides the validation itself, the tests intended to demonstrate the applicability of the method. The results showed that the Neural Networks algorithm presented similar performance to those of other algorithms and was validated as adequate to the modeling task. Still in the research context, an algorithm based on a sampling technique called the Jackknife was integrated to the openModeller, to be applied in the pre-analysis step. Tests related to the running time were carried out and a parallel version of this algorithm was developed.