Memoire Online - A study of the birds in central Bangkok (Thailand) in order to implement the basis of a long term monitoring

TRAVAIL DE FIN D'ETUDES PRESENTÉ EN VUE DE L'OBTENTION DU DIPLOME DE MASTER BIOINGENIEUR EN GESTION DES FORETS ET DES ESPACES NATURELS

(c) Toute reproduction du présent document, par quelque procédé que ce soit, ne peut être réalisée qu'avec l'autorisation de l'auteur et de l'autorité académique de l'Université de Liège/Gembloux Agro-Bio Tech

TRAVAIL DE FIN D'ETUDES PRESENTÉ EN VUE DE L'OBTENTION DU DIPLOME DE
MASTER BIOINGENIEUR EN GESTION DES FORETS ET DES ESPACES NATURELS

Upon completion of this topic, I wish to sincerely thank all those who were in any way involved in its realization.

I want first of all to thank my two co-promoters, Pr. Marie-Claude Huynen and Pr. Tommaso Savini who allowed me to discover the world of ornithology in the incredible city of Bangkok and I sincerely thank them for their advices all along the redaction of the present work. Then, I would like to thank JuanMa for his warm welcome in Bangkok, for his help concerning the accommodation and the ways to go around the city. I don't know how I would have done without your little red bike. Thanks also to the wonderful folks I met in Bangkok: Mart', Lek et al., Jess, Barry and the amazing group «it's a bad idea», particularly Gwen and Franck. Thanks to all for the sharing, the support, the laugh...

Un tout grand merci à Marc Dufrêne et Anaïs Gorel, car sans eux pas de stats et sans stats... pas de TFE ! Vous vous êtes toujours rendus disponible quand j'avais des questions et je vous en suis très reconnaissante. Merci également au professeur Jan Bogaert pour ses pistes de discussion et à José Flahaux pour sa relecture consciencieuse.

Ensuite, au terme de ce master en Gestion des Forêts et des Espaces Naturels, j'aimerais remercier l'ensemble du cadre enseignant pour les différents cours prodigués lors de ces deux années de master. J'ai toujours apprécié la qualité des cours et le bon équilibre entre cours magistraux et visites de terrain. Merci de m'avoir accompagnée et transmis votre savoir !

Impossible de ne pas citer ensuite mes cokotteurs de ces dernières années aux Déchets et à l'Auberge, avec qui j'ai passé des moments inoubliables et sans qui la vie à Gembloux n'aurait surement pas été la même : Romy, Alex, Sosso, Chavroux, Pauline, Porco, Valou, Baz, Ana, Flo, Eme, Olivia, Boedts, Sophie, Renard, Arthur ainsi que François, Justine, Angeline, BM, Roxane, Constant, Mumu, Lewis, Val, Jey, Lio, Zara, Manon et Hélène. Et puisque bien entendu cette vie à Gembloux ne s'arrête pas au kot, j'ai une pensée pour mes complices du conseil, Fanfan, Sophie, Vic', Stritsky, Baz, Schreder, Const et Francky, je n'oublierai jamais ces merveilleux moments partagés ensemble. Un immense merci à Charles, Clément, Camille et Olivier, mes yolo d'acolytes de la rédaction, sans qui ce dernier mois n'aurait pas été si gay. Et pour finir, merci à Baptiste, Amandine, Henri, Kity, Scott et toutes les autres merveilleuses rencontres faites en ces bons vieux murs de Gembloux !

Pour terminer je ne remercierai jamais assez mes parents et mes trois petites soeurs, Claire, Chloë et Coraline, qui m'ont toujours soutenue à tous les niveaux, et sans qui je ne serais pas ce que je suis aujourd'hui ; ainsi que Thomas qui me supporte depuis presque deux ans.

Le voyage réalisé dans le cadre du présent travail a été rendu possible grâce au soutien financier de l'Académie de Recherche et d'Enseignement supérieur de la Fédération Wallonie-Bruxelles, Belgique (Commission de la Coopération au Développement)

ABSTRACT

Worldwide, urban sprawl, induced by current increasingly demographic pressure, has become a prominent concern in conservation ecology. Urban green patches are essential biodiversity hotspots in cities. Bangkok, capital of Thailand, is among the larger cities in Asia and did not escape the global growth of urbanization, fragmenting the green areas of its metropolis and seeing its biodiversity collapsing like elsewhere. As part of that thesis, we collected ornithological and environmental data into various green patches of Central Bangkok. Indeed, the goal of this study is to investigate the ornithological characteristics, together with the environmental factors affecting them in order to implement the basis of a long-term monitoring of the urban avifauna. Various indices were calculated to permit the description of the chosen green patches' ornithological and environmental characteristics. Then, different statistical methods were used in order to explain the previous calculated indices' and bird communities' distribution and how the environmental features affected them. We demonstrated that the green patch size and water cover rate influenced the most the ornithological characteristics indices in our study area. Several issues related to bird conservation in Bangkok are then discussed through the main findings of this thesis. Finally, perspective are set focusing on the fact that long-term data about birds collected across a city can help filling the gaps caused by our lack of understanding of the metropolitan landscapes design needs in order to better sustain the avian fauna in the cities.

Keywords: conservation ecology, urban green patches, birds, Bangkok, fragmentation, urban avifauna, monitoring, Southeast Asia

De par le monde, la croissance urbaine, conséquence d'une pression démographique exponentielle, est devenue une préoccupation capitale en écologie de la conservation. Les espaces verts urbains sont des importants centres névralgiques de biodiversité au sein d'une ville. Bangkok, capitale de la Thaïlande, fait partie des plus grandes villes d'Asie du Sud-Est et n'a pas échappé à la croissance urbaine généralisée. Les espaces verts de la métropole ont été intensivement fragmentés et la biodiversité s'est effondrée comme partout ailleurs. Dans le cadre de ce mémoire, nous avons recueilli des données ornithologiques et environnementales au sein de divers espaces verts dans le centre de Bangkok. En effet, l'objectif de cette étude est d'analyser les caractéristiques ornithologiques, ainsi que les facteurs environnementaux qui les affectent afin de mettre en place les bases d'un monitoring à long terme de l'avifaune urbaine de Bangkok. Divers indices ont été calculés pour permettre la description des caractéristiques ornithologiques et environnementales des espaces verts choisis. Ensuite, différentes méthodes statistiques ont été utilisées afin d'expliquer la distribution des indices précédemment estimés et de définir des communautés d'oiseaux. L'influence des caractéristiques environnementales sur ces distributions a ensuite été développée. Nous avons ainsi démontré que la taille et le taux de recouvrement en eau des espaces verts sont les deux variables environnementales qui influencent le plus la diversité ornithologique dans notre zone d'étude. Plusieurs suggestions pour la conservation des oiseaux à Bangkok ont ensuite été discutées à l'aide des principaux résultats apportés par ce mémoire. Finalement, les perspectives mettent l'accent sur l'importance d'un monitoring à long terme de l'avifaune au sein d'une métropole comme Bangkok.

Mots-clefs : écologie de la conservation, espaces verts urbains, oiseaux, Bangkok, fragmentation, avifaune urbaine, monitoring, Asie du Sud-Est

IV.4.4. Environmental explicatory factors of the ornithological distribution analysis 34

FIGURE 3: SPECIES RICHNESS AND ENDEMISM IN THE FOUR BIODIVERSITY HOTSPOTS OF SOUTHEAST ASIA.) 5

FIGURE 12: AMOUNT OF SPECIES CHARACTERIZED BY DIFFERENT DISTRIBUTION (NUMBER OF RECORDS) IN THE STUDY AREA. 40

FIGURE 14: AMOUNT OF SPECIES INDIVIDUALS CHARACTERIZED BY DIFFERENT RELATIVE DENSITIES IN THE STUDY AREA. 42

FIGURE 15: COMMUNITY SPECIALIZATION INDEX (CSI) DISTRIBUTION IN THE STUDY AREA 43

FIGURE 16: COMPARISON OF THE DESCRIBING PARAMETERS OF THE ORNITHOLOGICAL DATA CALCULATED IN THE 25 PATCHES STUDIED 44

FIGURE 18: FACTORIAL DESIGN CREATED WITH THE TWO FIRST AXIS OF THE PCOA CONCERNING THE ABUNDANCE DATA 46

FIGURE 19: REPRESENTATION OF THE ENVIRONMENTAL VARIABLES IN THE PEARSON AND SPEARMAN CORRELATION CIRCLES FORMED BY

FIGURE 20: FACTORIAL DESIGN CREATED WITH THE TWO FIRST AXIS OF THE PCA CONCERNING THE ENVIRONMENTAL DATA 50

FIGURE 21 REPRESENTATION OF THE ENVIRONMENTAL VARIABLES TOGETHER WITH THE ORNITHOLOGICAL DESCRIPTIVE PARAMETERS IN THE

PEARSON AND SPEARMAN CORRELATION CIRCLES FORMED BY THE TWO FIRST AXES OF THE PCA. 51

FIGURE 24: REPRESENTATION OF THE SPECIES ABUNDANCE AND ENVIRONMENTAL VARIABLES IN THE PLOT FORMED BY THE TWO FIRST AXES

TABLE 1: SYNTHESIS TABLE BRINGING CONSERVATION KEYS IN ORDER TO ALLEVIATE THE EFFECTS OF URBANIZATION ON BIRDS. 13

TABLE 4: CRUDE ORDINAL SCALE OF ABUNDANCE DEDUCTED FROM THE ENCOUNTER RATE DATA 25

TABLE 8: SPECIES SELECTED VIA THE INDVAL METHOD AS BEING SIGNIFICANTLY ASSOCIATED TO A GROUP OF SITES 47

TABLE 9: PEARSON AND SPEARMAN CORRELATION COEFFICIENTS BETWEEN THE ENVIRONMENTAL VARIABLES AND THE TWO AXES OF THE

PCA. 49
TABLE 10: PEARSON AND SPEARMAN CORRELATION COEFFICIENTS BETWEEN THE ORNITHOLOGICAL VARIABLES AND THE TWO AXIS OF THE

TABLE 11: GENERAL LINEAR MODELS AND SUMMARY STATISTICS FOR ORNITHOLOGICAL VARIABLES 52

TABLE 12: PEARSON AND SPEARMAN CORRELATION COEFFICIENTS BETWEEN THE ENVIRONMENTAL VARIABLES AND THE TWO AXES OF THE

YOU COULD THINK THAT THERE ARE ONLY PIGEONS
THAT EVERYONE TRIES TO GET RID OF...

...BUT THIS THESIS WILL SHOW YOU THAT BIRD IS AN
INCREDIBLE TAXA IN WHICH A LOT OF SPECIES
MANAGE TO ADAPT TO EVEN THE NASTIEST HABITAT

Bangkok, capital of Thailand, is among the larger cities in Asia with an estimated unofficial population of more than 10 million people (THAIUTSA et al., 2008) and did not escape the global growth of urbanization, fragmenting the green areas of its metropolis and seeing its biodiversity collapsing like elsewhere in Southeast Asia (SODHI et al., 2004; SODHI and BROOK, 2006).

Urban ecology actions are more urgent now than they have ever been, especially in developing countries that contain some of the world's largest metropolitan areas. According to the World Urbanization Prospects (UN, 2012), Asian cities host about half of the urban population of the world, with this number expected to increase by 1.7 times over the next four decades.

The Southeast Asian region is characterized by four biodiversity hotspots. When coupling that high biodiversity with the high human population density, the region comes to be one of the most endangered biodiversity hotspots where demographic and economic pressures have led to extensive conversion of forests and overexploitation of coastal resources (WILLIAMS, 2012).

Studies of the avian fauna in metropolitan areas show that cities generally remain hostile places to most native bird species. However, these areas in which people live, work and play could take on an increasingly vital role in sustaining biological diversity (TURNER, 2003). Wildlife diminution rates can only be arrested by reconciling activities in production landscapes (agriculture and urban) with the conservation of nature (ROSENZWEIG, 2003). Long-term data about birds collected across a city can help filling the gaps caused by our lack of understanding of the metropolitan landscapes design needs and allow to better sustain the avian fauna in the cities (TURNER, 2003).

Two features of importance will be especially highlighted throughout this master thesis. First, in a general context of decline and homogenization of populations of urban birds (DEVICTOR et al., 2008; MCKINNEY, 2006; SAX and GAINES, 2003), it is a key applied issue to understand and to predict their distribution and persistence in the modern, fragmented landscapes humans created. Second, urban green spaces are an essential foundation for a healthy population, a healthy economy and for ecological balance in any city (BOLUND and HUNHAMMAR, 1999; WHO, 2008) and it is thus essential to predict how their environmental composition affects birds to better understand the value of those urban green spaces (KOSKIMIES, 1989).

The impact of the intensive environmental changes on the avian fauna of Bangkok haven't been studied yet and long-term data on multi-species distribution are inexistent (ROUND, 2008). Thereby, a first step would be to study the distribution of existing avian fauna in Bangkok to establish a long-term monitoring and set priorities for its long term conservation.

This thesis reports on surveys of the avifauna within various vegetation patches of Bangkok with the aim of providing basis for long term monitoring actions.

In order to best achieve the stated goal, we centered our work on two principal research questions and the ensuing objectives (Figure 1).

After having achieved the present objectives, the basis to implement a long term monitoring will be set and a discussion will be oriented to bring preferences for Bangkok avifauna conservation.

This master thesis is divided into 7 sections. To put things into context, the first two sections consist in a brief introduction, followed by a literature review supporting the general framework of the study. Then, the third section will describe the area in which the study was realized. Section 4 will present the methodology and the analyses that we used in order to reach the objectives previously described. Section 5 will then show the results of the analyses and sections 6 and 7 will discuss and conclude the results obtained, finalizing with the perspectives regarding the long-term monitoring.

II. LITERATURE REVIEW

This section aims at setting the general context of this study, using what the literature can offer so far. It will start by a brief recall of the actual worldwide biodiversity crisis, focusing then more on the description of the situation in Southeast Asia. Afterwards, we will provide an overview of the bird taxa situation, starting with the general trend of the evolution of their populations in Bangkok. We will also show the bird's importance as environment indicators as well as the need of monitoring them to explain environmental changes. We will next have a look at the effect of urbanization on birds and the importance of urban green areas, finally giving conservation clues from studies made in other cities.

«La caractéristique la plus merveilleuse de notre planète est la présence de la vie et la
caractéristique la plus incroyable de la vie est sa diversité¹!» (BEUDELS, 2013)

II.1.1. Biodiversity in decline

The term «biodiversity» is a contraction of «biological diversity» and was defined by the Convention on Biological Diversity (UNCED, 5^th of June 1992) as: «The variability among living organisms from all sources including, inter alia, terrestrial, marine and other aquatic ecosystems and the ecological complexes of which they are part: this includes diversity within species, between species and of ecosystems.» First, this term was closely related to nature conservation but has been then associated with more functional and utilitarian notions, especially after the publication of the MILLENNIUM ECOSYSTEM ASSESSMENT (MEA; 2005) which connected interactions between people, biodiversity and ecosystems.

Indeed, despite the fact that most humans consider themselves above everything, the biodiversity that surrounds them is indispensable to their survival. In fact, the changes in human condition leads to changes in biodiversity and in ecosystems and thus, have an ultimate effect on the services provided by the ecosystems which make biodiversity and human well-being entirely linked together (MEA, 2005). Species conservation is not only giving the species the right to exist, it also adds value to human's life by providing supporting, provisioning, regulating and cultural functions as shown in Figure 2.

¹ «The most wonderful characteristic of our planet is the presence of life and the most incredible characteristic of life is its diversity»

Currently, biodiversity is unequivocally declining and some authors even speak about a «6^th extinction crisis» (LEAKEY and LEWIN, 1999; MEA, 2005). If the extinction of a species is indeed a natural process (75 to 95% of all species that have ever existed are now extinct), today's biodiversity is disappearing at a 100 to 1000 times higher rate than the mean natural extinction rate that occurred during the fifth previous extinctions. So, the concern is not about the occurrence of extinctions, but rather about the acceleration of the extinction process: if nothing is done, 50% of the actual species will have disappeared before the end of the XXI^st century (LEAKEY and LEWIN, 1999) and a world of pests and weeds will remain.

Nowadays, the challenge is to implement a sustainable development ensuring the social and economic viability of human societies while respecting the ecosystems (BROWN, 2001). Protecting those ecosystems requires a good basic knowledge of them, which can be improved by scientific research. As shown before, biodiversity is a quite nebulous and extremely large concept. Nonetheless, it must unconditionally be quantified in order to reach political decisions or to implement management measures or to reach a priority for actions. An important difficulty in quantifying biodiversity is that it is a multifaceted concept (PURVIS and HECTOR, 2000) and it must be done at a defined scale and with a defined and refocused objective (HOSTETLER, 1999).

Although Southeast Asia (Brunei, Cambodia, Indonesia, Laos, Malaysia, Myanmar, the Philippines, Singapore, Timor-Leste, Thailand, and Vietnam) incorporates four biodiversity hotspots (BRIGGS, 1996; SODHI et al., 2004; WILLIAMS, 2012) (Figure 3), the region faces several key social, scientific and logistical conservation challenges.

Figure 3: Species Richness and endemism in the four biodiversity hotspots of Southeast Asia. The red bars represent the percentage of species endemic to the respective hotspot. Numbers in parentheses represent total and endemic species known to science, respectively (SODHI et al., 2004)

Among all the world's tropical regions, Southeast Asia has the highest rate of habitat lost with a deforestation rate four times higher than elsewhere in the world (FAO, 2012). This fact is alarming knowing that compared to the other tropical regions, Southeast Asia has the highest mean proportion of country-endemic bird (9%) and mammal species (11%) as well as the second highest rate of country-endemic vascular plant species (25%) (SODHI et al., 2009).

The current major threats to biodiversity in Southeast Asia are predominantly from socioeconomic origin; including population growth, poverty, chronic shortage of conservation resources and corrupt national institutions. Hence, as the regional societies of Southeast Asia attempt to match

the living standards of developed nations, environmental issues are inexorably marginalized (SODHI et al., 2004; SODHI and BROOK, 2006). This is supported by the fact that the economic constraints are much larger in the developing countries like Thailand, than in North America or Europe and therefore it is difficult to find money for environmental management when basic needs and poverty are an immediate bigger concern (FRASER, 2002).

Lastly, research on Southeast Asian biodiversity over the past 20 years has been neglected in comparison to other tropical regions (SODHI and BROOK, 2006). Indeed, it appears that there is a shortage of local scientists conducting rigorous conservation biology research in Southeast Asia, with the current work dominated by descriptive work, mostly inventories (SODHI and LIOW, 2000). These trends are disturbing and the consequence can be even more severe for that region because of the habitat destruction that occurred during the last century. Some solutions were brought by SODHI and LIOW (2000) to improve the quality of conservation biology research of Southeast Asia. They range from an increased accessibility to the international conservation biology journals to the start of more multinational collaborative projects, more rigorous funds for long-term research, education of local scientists in research design to reach the standards in order to be published in international journals ...

«Birds are among the best known parts of the Earth's biodiversity. But nevertheless soundly quantified knowledge is far from complete for most species and regions.» (BIBBY et al., 1998a)

The bird taxa did not avoid the general biodiversity decline, and indeed, according to the IUCN Red List of Threatened Species (IUCN, 2013), 12% of the world avian species are now threatened. For a long time, the major threats for the birds were the variability of climatic events and their effect on vegetation. Those have been lately supplanted by the human impacts on the environment. During the last centuries, the pressure humans put on nature increased substantially with the intensification of urbanization and agriculture that generates the vanishing of many ecosystems.

The study of avian fauna of Southeast Asia reveals alarming trends: if the region hosts the highest mean proportion of endemic bird species at a national level, it also has the highest mean proportion of threatened bird species of all tropical regions. Despite this, the avifauna of Southeast Asia has been one of the least studied in the tropics (SODHI et al., 2006).

Thailand, as a country, holds 971 bird species (IUCN, 2013). 925 are native bird species while 1 has been introduced (Columbia Livia), 40 are vagrant species and 5 species are still uncertain data. Figure 4 shows a pie chart of Thailand bird species distribution through the IUCN Red List Categories.

Figure 4: Bird species distribution into the IUCN Red List Categories. CR= Critically Endangered, EN= Endangered,
VU=Vulnerable, NE=Near Threatened, LC= Least Concerned, DD= Data Deficient (Values sources: IUCN, 2013)

PHILLIP ROUND (2008) in his book, «The birds of the Bangkok area», reviewed the birds of the Central Plain of Thailand. He also put together existing behavioral, life-history and ecological data on birds around Bangkok. The following paragraphs give a short summary of the evolution of the avian state in Bangkok that ROUND (2008) developed in the introduction of his book.

Once, the Central Plain held a bird and mammal mega fauna that vanished because of the high rate of destruction due to large-scale historical transformations in Thailand. Unfortunately, historic surveys of the wildlife of Bangkok are poor. Some old records inform us about the previous presence of ibises, pelicans, adjutants, vultures and birds characteristic of open forests. All those species are not there anymore, only smaller and ecologically tolerant birds of forests and secondary growth persisted until the seventieth century. The gradual decrease in the number of resident bird species in the Central Plain is clear and keeps happening since the intensification of agriculture and the spread of housing and industry. The Asian Economic Crisis, from mid-1997^th onwards, gave the environment of the Chao Phraya Delta a brief respite from land speculation and uncontrolled development. However, today's economy is once again booming with all that is involved and no additional environment safeguards are put in place.

Nowadays, Bangkok's urban green areas are sparse compared with many other capitals and only the most ecologically tolerant species still survive well in inner city gardens and parks (Coppersmith Barbet, Common Iora, Pied Fantail, Oriental Magpie Robin, Streak-eared Bulbul, Common Tailorbird, Scarlet-backed Flowerpecker, Brown-throated Sunbird and Olive-backed Sunbird). Introduced bird species are also often seen in the city while they escape their cages, like the White-Crested Laughingthrush or the Red-Breasted Parakeet. Still, many birdwatchers are constantly amazed at how many species of birds they are able to see in the concrete jungle of Bangkok (pers. obs.).

Parks are of great importance in order to provide habitats for the birds but most of Bangkok's public parks have tended to be too deeply manicured to support many species. Native vegetation is disappearing and even water bodies are polluted with herbicides to prevent the colonization of aquatic vegetation. Though, the direct widespread impacts of pollution can't be estimated because of the absence of any systematic monitoring of the levels of toxic pollutants in birds.

An initiative came from the Bird Conservation Society of Thailand, which explored the possibility to implement an urban bird reserve in eastern Bangkok together with the Bangkok Metropolitan Administration (BMA). Regrettably, this effort fizzled out due to a change of BMA governor.

The previous trends highlighted by ROUND (2008) made that BirdLife international decided in 2004 to consider the inner gulf of Thailand (100,000 ha), including the Bangkok Metropolis, as an Important Bird Area (IBA) in order to control the over-exploitation of natural resources and promote compatible forms of land use across the whole area. The function of the IBA program worldwide is «to identify, protect and manage a network of sites that are significant for the longterm viability of naturally occurring bird populations, across the geographical range of those bird species for which a site-based approach is appropriate» (CHAN et al., 2004). However, this applies more to the coastal area where the actions are concentrated than to the Metropolis situation.

As there is much concern today about environmental changes, it is essential to know how those changes affect wildlife, and birds offer a great value as biological and environmental indicators (BIBBY, 1999; GOTTSCHALK et al., 2005).

In fact, they reflect well the global health of the surrounding biodiversity because they often have a high position in the trophic chain and also because they respond fast to landscape modifications. They are among the most conspicuous (POMEROY, 1992), indeed, compared to other animal taxa, birds are relatively easy to detect, identify and survey. They have been subject to numerous studies, especially in Europe and America and therefore their eco-ethology is generally well documented. Bird diversity was also found to be correlated with the diversity of other taxa (BLAIR, 1999, SATTLER et al., 2010) which means that they may be reliable indicators of the overall biodiversity. Additionally, many studies have shown that birds are particularly useful to detect unexpected changes, for example, due to the pesticides, as RACHEL CARSON (1962) denounced with her famous book, «Silent spring».

If birds are good environment indicators, it is also and primordially because of the relationship between a bird and its habitat. FULLER (2012) presents a good and recent synthesis of the multiple publications concerning the processes of habitat selection by birds and the following section is a brief description of these.

The habitat is the environment in which an individual lives, including biotic and abiotic features as climate, microclimate, soil type, topography, plant species and vegetation structure as well as the other animal species living in the same environment. The use of a habitat by a bird, meaning the way it uses the free spaces and the various resources it contains, differs obviously between every species but also in between the same species, for example, with the age or the sex of the animal. For birds in particular, the factors affecting the habitat evolve considerably along the year, especially for the migrant species or for the sedentary birds living in latitudes where the seasons are highly contrasted. Usually, it is during the breeding time that birds show the severest association with one habitat.

Mechanisms explaining how a bird chooses its habitat are better and better known. The notions of «ultimate factors» and «proximate factors» in habitat selection have been highly developed in the past (HILDÉN 1965², cited by FULLER, 2012) and are still universal.

- Ultimate factors: Basic factors defining the choice of habitat through its fitness potential (e.g. food-supply, shelter availability, territory space, structural and functional characteristics, other species...)

- Proximate factors: Immediate signs or stimuli that are not automatically of fitness value (e.g. landscape and microhabitat features, vegetation density or height, microhabitats or functional sites like song posts...)

Therefore, in order to allow the birds to select habitat that offers the best fitness, the «proximate factors» have to be correlated with the «ultimate factors». This is all the more important when habitat' quality can't be determined at the time the bird chose is territory, especially in the case of migratory birds which need to quickly select an area to stop.

Furthermore, the spatial scale is important as well to understand how birds select their habitat. Indeed, birds being very mobile animals and generally in need of several types of resources, the mechanisms of selection of their territories are often spatially hierarchized. Certain species start identifying a potential habitat using the general landscape' characteristics, perceived by flying over for example. Then, the exact location of their territory can be elected as a result of a finer scale' analysis. For example, the initial selection of a territory is based on the most common resource but the refinement to the final location will be done on the most limiting resource. In this case, the bird starts to take an interest at a finer scale and then starts checking other factors at a coarser scale. The spatial process encountered at various scales is all equally determinants in order to explain the choice of a habitat by a bird.

² HILDÉN O., 1965. Habitat selection in birds-a review. Ann. Zool. Fenn., 2, pp. 53-75.

Birds are also useful for monitoring and incorporating cumulative changes over long periods of time (BIBBY, 1999; KOSKIMIES, 1989). Birds counts conducted in a systematic and consistent way can provide an early-warning system in order to assess the health of an ecosystem. This is essential for the authorities to ensure that development is truly sustainable (POMEROY, 1992).

According to KOSKIMIES (1989), monitoring corresponds to «continuous and regular quantitative research using standardized methods, which reveal changes in the abundance and ecology of birds». In order to be well studied, the changes need to be clearly divided between those caused by human activities and those caused by the natural dynamics like the climatic changes, the geological processes or the biological evolution. Indeed, the last one affects the bird populations much more slowly than the human-caused ones. The first advantage of biological monitoring in opposition with non-biological monitoring is that environmental changes are detectable, especially for those that can't be observed or forecasted by the measurement of a set of pre-selected physical or chemical parameters. The second advantage is that biological monitoring makes it possible to detect and monitor cumulative and non-linear consequences of various environmental changes acting simultaneously. An integrated monitoring can allow to study cause-effect relationships which are truly important in order to decide the actions to be taken (KOSKIMIES, 1989).

The more changes the environment undergoes, the more it becomes necessary to learn how to manage it in order to take care of species conservation. Many cases need action but without precise data on numbers or trends, no useful recommendation for management action can be made. Furthermore, there is a need for a large ecological knowledge in many situations because we are continually changing our environment and in so doing, the birds with which we share it are inexorably affected (POMEROY, 1992)

In Thailand, assessment of the effect of various construction projects on biodiversity consists of little more than some unauthenticated lists of birds, mammals, or other taxa (ROUND, 2008). Indeed, a regrettably usual scheme in the case of birds is the statement that the impacts of their habitat damages will be minimal because the birds are able to fly to other areas.

«The effect of urbanization can be immense, yet our understanding is rudimentary»
(CHACE and WALSH, 2006).

Rapid urbanization has turned out to be one of the major concerns in conservation ecology (MILLER and HOBBS, 2002) and can be justified by the fact that the world urban population is expected to increase by 72 per cent by 2050 (UN, 2012). Moreover, cities occupy less than 3% of the global terrestrial surface, but account for 78% of carbon emissions, 60% of residential water use, and

76% of wood used for industrial purposes (BROWN, 2001). Still, the intensifying conflict between the economy and the ecosystem of which it is part is evident and undeniably, urbanization will keep having a significant impact on the ecology at local, regional and global scales (SINGH et al., 2010).

Uncontestably, urbanization and anthropogenic activities intensification in the landscapes change the ecosystem at many levels which lead to the homogenization of habitat structure and composition (FORMAN, 1995). For example, the urban sprawl is made of redundant artificial infrastructures that homogenize the urban landscape (MCKINNEY, 2002; 2006). This has a major negative impact on biodiversity and on the ecosystem capacity to ensure the expected services (FORMAN, 1995).

However, urban areas seem characterized by a more important species abundance than suburban areas for some biological groups (KÜHN et al., 2004; ARAÚJO, 2003; NIELSEN et al., 2013). Indeed, some species find alternative ecological niches in the cities and are able to develop important populations like it is the case for the well-known Columbia Livia (Rock Pigeon). Urban land-uses represent ideal habitats for the demographic explosion of those urban-exploiter species, able to use the abundant food resources associated to human litter (ORTEGA-ÁLVAREZ and MACGREGOR-FORS, 2009). Therefore, many studies claim the threat of the massive disturbances created by city growth on the habitat of native species (CONOLE and KIRKPATRICK, 2011; DEVICTOR et al., 2008; MCKINNEY, 2006). Indeed, those disturbances create a new habitat for few widespread non-native species that are easily adapting to urban conditions and enrich the local biodiversity while the global diversity is decreasing subsequently to the extinction of non-adapting local species (SAX and GAINES, 2003).

Those apparent contradictions probably result from differences between the geographic scales used, sampling bias, different contexts or else different biological responses (MACDONNELL and HAHS, 2008). Identifying the proximal factor of the urban diversity is relatively difficult as well as studying urbanizations gradients and biological response that are far from being linear (MACDONNELL and HAHS, 2008). Nevertheless, natural populations' extinction in the most urbanized parts of a city seems well established, especially in the new growth tropical cities. The geographic layout of the urban biodiversity hotspots is fundamental and there is a major lack of protected areas in the urban environment (BASTIN and THOMAS, 1999, SANDSTRÖM et al., 2006).

The «urban green spaces» comprises all urban parks, forests and related vegetation (SINGH et al., 2010); even cemeteries can be considered so (LUSSENHOP, 1977). According to the WHO (2008), at least 9 m² of urban green space per capita is recommended to alleviate undesirable

environmental effects and provide other benefits like a healthy population and a sustainable economy in any city (KUCHELMEISTER, 1998). However, the amount of required open green spaces per city dweller has remained controversial (SINGH et al., 2010). Those amounts have been estimated for some developing countries cities like Seoul that has 14 m² of urban green per capita (KUCHELMEISTER, 1998), Singapore, 10 m² (CHOW and ROTH, 2006), Beijing, 6 m² (DEMBNER, 1993), Mexico City, 1.9m² (DELOYA, 1993) and New Delhi, 0.12 m² (KUCHELMEISTER, 1998).

This is quite alarming while knowing that within municipal limits of 26 large European cities, the average of urban green space is estimated at 104 m² per inhabitant (KONIJNENDIJK, 2003). Indeed, urban green spaces are increasingly critical to healthy cities (WHO, 2008), even more in developing countries that include some of the world's biggest metropolitan areas and have the greater rate of urbanization (UN, 2012).

A functional network of urban green spaces can contribute to ecological diversity in a city (SANDSTRÖM et al., 2006). The major benefits of green spaces are (LAGHAI and BAHMANPOUR 2012):

- Assimilation of Carbon dioxide and other toxic gases as well as Oxygen production

- Prevention of unsuitable urban development and increase of the beautifulness of the city

Green spaces perform important functions and services worldwide (NIELSEN et al., 2013), their development has the potential to moderate the adverse effects of urbanization in a sustainable way, making cities more attractive to live in, reversing urban sprawl, and decreasing transport demand (DE RIDDER et al., 2004).

II.3.3. Conservation keys to reduce the urban effects on birds: state-of-the-art

Despite high stress ensuing from urban life features such as noise (KATTI and WARREN, 2004), air and soil pollution (MCKINNEY, 2002; ROUND, 2008) and high densities of domestic predators (ANDERIES et al., 2007; SORACE, 2002), urban areas throughout the world are characterized by high food resource abundance and high avian population densities but as developed before, lower species diversity is generally observed (MARZLUFF et al., 2001).

Many studies have attempted to determine the impacts of urbanization on birds worldwide together with solutions in order to alleviate the damages done to bird communities. Table 1 brings a state-of-the art giving conservation keys brought from multiple scientific studies.

Table 1: Synthesis table of the publications bringing conservation keys in order to alleviate the effects of urbanization on birds.

Integrating social and socio economic processes (e.g. poverty alleviation, public SODHI et al., 2004;

Using the habitat island ecological theory as a research framework for the FAHRIG, 2003;

Understanding better the degree to which species respond to local environmental conditions and landscape patterns

Using satellite-based remote sensing together with bird data in a GIS environment

Using a Community Specialization Index for measuring functional homogenization on both local and global scales across time

We carried out the study in Bangkok, Thailand's capital, both a city and a province. The metropolis of Bangkok covers a 1,568.7 km² area in the delta of the Chao Phraya River in Central Thailand. The study area was focused on green areas localized in the most densely urbanized area of the city (Figure 5).

Figure 5: Localization of the study area. a. Geographical location of Bangkok in Southeast Asia, b. localization of the study area

Since the 1960^th, Bangkok has known an astonishing physical growth from 6 km² to the current city area. Its major river, the Chao Phraya, has performed as the central artery of the whole city and has significantly influenced settlement formation and configuration (MATEO-BABIANO, 2012). The population of Bangkok was estimated to be around one million people in 1950 while today it is close to 12 million (FRASER, 2002). This number is continually increasing due to the excellent economic potential of the city that attracts people from the countryside as well as expatriates from all over the world. Furthermore, abundant tourists visit Bangkok every year, adding people in this already overpopulated city (THAIUTSA et al., 2008).

Bangkok presents the case of a dynamic city with competition between its traditions and the Western contemporary influences on its urban spaces. The city shows the same seemingly disorganized quality that characterizes the Asian space and therefore the diversity of the street space is alike the forest environment, where a cacophony of sounds, sights, smells, tastes and touch can be experienced altogether (MATEO-BABIANO, 2012).

Regrettably, the other side of the coin of a megacity like Bangkok is the multiple kinds of pollution that occur which include atmospheric, auditory and visual. Water pollution is also critical on the fact that the canals are like open sewer, the groundwater is therefore in really bad shape. The publications over this subject are countless and won't be developed through this work.

Bangkok has a seasonal monsoonal climate. According to the Köppen classification it is an Aw climate type (KHEDARI et al., 2002). The daily average temperature stays relatively constant over the year with a mean annual temperature of 28.1°C. The monsoonal rainy season stands from July to October while the dry season extends from November to June with the three first months (until February) cooler, making it be called the «cool» season. The last month of the dry season (from March until June) shows high solar intensity as well as high heat and longer days and is therefore called the hot season (Figure 6). The heat in this season is even more felt by the effect of pavement and buildings (THAIUTSA et al., 2008).

Climate change is a significant threat that will create, through the rising of the sea level, profound impacts on the Chao Phraya Delta (ROUND, 2008). As Bangkok is situated only 2 m above de sea level with some important parts of the city that are only at 0.5 m or less, clearly much of the inner city could be flooded by the turn of the century (JARUPONGSAKUL, 2000⁴; cited in ROUND, 2008). Over 14% of the city's total area is seasonally flooded during the wet season (THAIUTSA et al., 2008).

THAIUTSA et al. (2008) calculated the areas of the main land uses in Bangkok from GIS analysis of satellite imagery. As no more literature was found about the land use in Bangkok, the following information stems from this article.

Bangkok includes a quite large amount of unconstructed areas within its boundaries. As shown in Table 2, just over 50% of the metropolis is made of building, roads and other constructed surfaces, followed by an unexpected 26% of land used for food production, mainly farmland and shrimp farms on the periphery of the city boundaries. Finally, only 4.2 % of the city's total area is green space, if one excludes agricultural land. The green spaces are mostly made of trees found in the streets and naturalized areas with 1.2% of developed green spaces. The developed green spaces are nearly equally split between actual parks accessible to the entire population, sports field and golf courses which are not readily accessible to all the citizens of Bangkok.

Table 2: Area of main land uses in Bangkok (variables source: THAIUTSA et al., 2008)

The land use is very different in the central part of the city in comparison with its eastern and western edges that are adjoined with food producing areas and forest covered provinces. Thus, those districts have the highest rate of tree and food producing areas within Bangkok. On the other hand, the districts situated in the center of the city, have the highest population density with 14,000 persons per km² and 8000 persons per km² respectively, these amounts being 2 to 4 times higher

4 JARUPONGSAKUL T., 2000. Potential impacts of sea-level rise and the coastal zone management in the upper gulf of Thailand. pp. 138-151 in SINSAKUL S., CHAIMANEE N. and TIYAPAIRACH S. (eds.), Proceedingsof the Thai-Japanese geological meeting: The comprehensive assessment on impacts of sea-level rise. Geological Survey Division, Department of Mineral Resources, Bangkok.

than in the other districts groups. This is of greater importance as the study will focus on the green patches of this part of the city. Those variations in population density distort the per capita green space values. Hence, the per capita green space averages 2.8 m² in those two districts while it rises to a mean of 11.8 m² for the entire city. About 40% of those green spaces are actually park spaces (the rest being tree cover) which represent around 1.2 m² per inhabitant. The BMA tends to increase the park space per capita so it will reach 2.5 m² in 2023 with an ultimate goal of 4 m² per person.

IV. METHODOLOGY

As a reminder, the goal of this study is to investigate the ornithological characteristics, together with the environmental factors affecting them, within various vegetation patches in Bangkok in order to implement the basis for a long term monitoring.

In order to reach the fixed goal, a suite of processes was followed. First, the vegetation patches were sampled within the study area. A field phase was then realized within those green patches to collect the raw ornithological data. On the other hand, the environmental data were obtained from a digitalization out of satellite imagery using the GIS software ArcGIS(c) 10.1. Various variables were then calculated to permit the description of the urban green patches' ornithological and environmental characteristics. Different statistical methods (principally multivariate) were finally used in order to point out the bird communities formed and how the environmental features affect them.

The three main components of a landscape structure are (FORMAN and GODRON, 1986):

- The patches: functional units of the landscape which represent homogeneous environmental conditions and whose boundaries are distinguished by discontinuities in the state variable of a significant magnitude for the ecological process or the considered organism. All patches showing similar characteristics for the considered process is called «type» or «class».

- The corridors: units of a characteristic linear form. The «corridors» perform the ecological function of passage, filter or barrier. They are often present in a network-like landscape.

- The matrix: across types, the «matrix» is the most common and the less fragmented. This type can also be considered as the background of the landscape within which are situated the other elements.

This work will focus on the birds of the «urban green» patch type, the matrix in this case being the concrete of the urban zone. Unfortunately, the avian distribution in the corridors won't be approached because of the lack of time, although it can hold a high level of bird diversity (SODHI et al., 1999; WHITE et al., 2005).

The precise localization of the green spaces in the study area was a first necessary component for the proper achievement of the sampling. Like many cities, Bangkok has spatially explicit planning, land use, and land cover maps for the entire city. Regrettably, it was not possible to get those maps, especially in the context of the political crisis that occurred during the time of the field work.

A digitalization of those green patches was made possible thanks to the interpretation given out of Google Earth(c) satellite imagery (version 7.1.2.2041, accessed on February 2014), Google Maps(c) and various recent tourist maps of the city. A preliminary process of exploration was achieved ten days before the start of the survey to validate in the field the potential green areas.

Some factors affected the selection of the parks and restricted the sampling possibilities:

- we focused on green areas localized in the most densely urbanized area of the city - all the green patches of Bangkok were not open to public

Finally, we selected 25 green patches on a total sampling basis, which means that all the green areas identified accessible and public allowed were designated for the study. Most of them were urban public parks, with two crossroad greeneries, one temple, one University's park, and adding to that, a cemetery (Appendix 1). As we realized the study in all public areas, we assumed the anthropic effect as equal. Furthermore, as poacher's activities are unobserved and because people don't really hunt birds in the city (ROUND, 2008), the presence of people will not reduce the bird presence but is more likely to affect our chances to detect them.

Ideally, the order of visit of the different parks should be determined on a random basis. However, practically, in order to optimize the movements across the study area, small green patches, close to each other were all visited within a same survey period. Because some species have a more active acoustic activity in the early morning, the order of visit into the parks surveyed together was inversed every day. Thus, a green patch visited in the early morning one day, was visited at the end of the morning the next day in order to increase the probability of encountering all the present species. The same scheme was followed for the afternoon.

Figure 7: Map of the patches sampled in Central Bangkok RAW DATA COLLECTION

Counting the avian fauna can get quite complex and a review of the literature provides plenty of methods that were employed to do it (POMEROY, 1992). According to JONES (1998), «It is better to get reliable data using a simple method than unreliable data from a complex one, even if the latter (potentially at least) could provide more information». The major reason for adopting a simpler method is to make it easily repeatable, i.e. it allows the study to be significant even if the operators are different, don't have the same level of training or don't put the same amount of effort in the study (JONES, 1998). It is all the more important for this study as it implements the basis for a long term monitoring and others will repeat this work in the future.

The two basics aspects of counting birds are the number of species and the number of individuals (POMEROY, 1992). We will record both while making counts, but they are not necessarily related. Indeed, the importance of birds in the ecosystem varies from place to place because different places have variable numbers of species and individuals (POMEROY, 1992).

The counting method depends upon many things, such as why the counting is made, which level of accuracy needs to be achieved, which resource is available and what time of the day and of the year the survey is done (POMEROY, 1992). Moreover, many factors affect bird activity and behavior, which influences the chances of recording them. Among the more important factors are the season, the time of the day and weather condition (JONES, 1998).

Seasonal effects on birds can be difficult to cope with. JONES (1998) shows the case of a species which is breeding: while the males may be singing and calling to defend their territory, which makes it easy to record them, the females that are nesting won't be seen.

The ornithologist survey we realized within the field phase took place between February 17^th and June 4^th 2014 i.e. during Thailand's hot season. As it was spread over a short period of time, the seasonal effect was minimized. However, as the migration period occurred in April-May, and the non-breeding visitors leave Bangkok around this time as well, the bird's seasonal status (ROUND, 2008) were recorded on the bird species list created.

The ten days before the start of the survey, spent venturing in Bangkok urban green spaces, were also essential in order to get familiar with the bird species and to design the practical details of the bird survey. Moreover, we revisited the sixth first patches at the end of the field work period so the bird species knowledge can be assumed as equal in all the patches.

As the aim of a census is to record as many as possible of the birds that are present, as quickly as possible, and as most of the birds show trends of morning and late evening peaks of activity (JONES, 1998), we chose those times for our surveys. Consequently, to follow the common study design pointed out by JONES (1998), the data collection began 30 minutes after dawn (6.30 am) in order to avoid the saturated acoustic atmosphere, well known as the dawn chorus, when the louder birds would be over recorded at the expense of the quieter ones. The survey ended during the mid-morning when bird activity declines (9.00 or earlier depending on the length of the path). The second survey period occurred before dusk, at about 4.00 pm until 6.30 pm.

We avoided adverse weather conditions like rain in order to delete the bias it could lead to in the surveys. Indeed, the bird activity is generally highly affected by those conditions, as well as the observer's capacity to see or hear them (JONES, 1998). Therefore, surveys were sometimes postponed until the following day due to the weather while approaching the rainy season.

We calculated species richness and abundance from the data gathered from six repeated passages into the patches, 3 in the morning and 3 in the afternoon to optimize the detections and the number of events in order to increase the analytical precision. Such surveys offer baseline conservation data regarding the distribution of the species, the richness of the sites or habitats and allow comparisons to be made between areas (JONES, 1998).

The principle was to walk along the patches tracks at a mean pace of 25 meters per minute and to record and count all the birds seen or heard. Adding to the binoculars (Nikula (c) V061042) and a field guide of the birds of Thailand (ROBSON, 2002), we also used a GPS receiver (Garmin GPSMAP 62) to survey the length of the path followed and afterwards to deduct the time spent surveying.

We used this survey technique, at the expense of other widespread practices, after some tests in the field which demonstrated the best results of the latter. For this reason, we didn't proceed with the «point counts» technique even though it is the most common method used to study the link between birds and habitat (BIBBY et al., 1998b). Indeed, as the patches were only sampled during three consecutive days, there wasn't sufficient data collected for most species to use this technique (except for Rock Pigeon or Eurasian Tree Sparrow or other very common urban species). We chose not to apply the «distance» technique neither because it needed too much prerequisites.

While «simply» recording and counting all the birds seen or heard along the linear survey as described above, we observed for each species a relative abundance and not an absolute abundance. It is not suitable to compare on this basis the abundance of two species because one could be easier to detect than the other. Similarly, one species' abundances can't be compared between surveys made in two different environments. However, in this case, the different surveys were done in similar contexts - urban green spaces - and the hypothesis can be done that the detection probability of a same species is similar within all the patches (BIBBY and BUCKLAND, 1987; JONES, 1998).

It is important to note that we counted separately the birds only flying over the surveying zone.

Each of the 25 patches surveyed was characterized by different land covers. For reason of time availability and lack of permit from the Bangkok administration, local ecological variables like the tree cover or canopy height could not be surveyed directly within the urban green spaces visited during the field work.

Nevertheless, in order to get environmental variables for the purpose of data analysis, land cover features within the 25 green patches were digitalized from a ESRI basemap (world satellite imagery⁵) using the GIS software ArcGIS(c) 10.1 (Figure 8).

5 Sources: Esri, DigitalGlobe, Earthstar Geographics, CNES/Airbus DS, GeoEye, i-cubed, USDA FSA, USGS, AEX, Getmapping, Aerogrid, IGN, IGP, swisstopo, and the GIS User Community

DEFINITION AND CALCULATION OF THE VARIABLES IV.3.1. Ornithological variables

The number of species recorded in each patch was based on the total number of species that were observed on a track across the 6 sampling periods.

Moreover, the amount of bird individuals per species seen or heard around the track were also recorded. As double counting or omissions can occur while recording birds with this method, the abundance collected could not be assumed as real. Therefore, the field hours were recorded in conjunction with the number of individuals of each species observed. This allowed a relative frequency of abundance to be calculated for each species by dividing the number of birds recorded by the time spent on the field, giving a figure of birds per hour for each species (ROBERTSON and LILEY, 1998). We can therefore assume that there is an actual relationship between the observed frequencies and the bird densities into the patches. However, the abundance data obtained do not give precise indications of density but if the hypothesis is made that a species is as easy to detect at one site as another, then encounter rates can be compared for a species between sites by being split into crude ordinal categories of abundance (Table 4). The categories allowed abundance scores to be given and future surveys to detect large scale changes in the abundance of individual species (ROBERTSON and LILEY, 1998).

Table 4: Crude ordinal scale of abundance deducted from the encounter rate data (adapted from ROBERTSON and LILEY, 1998)

For the subsequent analyses, some birds encountered during the field survey were not counted. This applied for passage migrants only seen punctually in some patches as well as non-breeding visitors recorded during the first months of the survey that were gone at the end and also for presumed escaped captives birds. Similarly, the birds recorded as flying over the census area were not used in the analyses (principally Swifts, Asian Openbills or Barn Swallows).

The Island Biogeography Theory (IBT; MACARTHUR and WILSON, 1963; 1967) predicts how the area and isolation of oceanic islands affect their species richness through the variation of extinction and colonization rates. The IBT was later revisited to any «area where the species can exist, surrounded by an area in which the species can survive poorly or not at all and which consequently represents a distributional barrier» (DIAMOND, 1975). In our study, the green patches can be considered as `islands' of habitat in the inhospitable matrix of the concrete of Bangkok.

From the patches areas (AREA) of the 25 green spaces, we were able to calculate an isolation variable, the weighted connectivity (WCON) between the patches in order to combine one fragment area with the proximity of the other 24 fragments (BICKFORD et al., 2010). The weighted connectivity of focal fragment ?? was calculated as

where ???? is the area of fragment ??, ???? is the total area of all fragments and ?????? is the shortest distance between fragments ?? and ??.

The patch parameters allowed us to calculate four indices commonly used to characterize the landscape (BICKFORD et al., 2010; BASTIN and THOMAS, 1999; BUREL et al., 1999) presented in Table 6.

where ?? is the measured perimter of the patch divided by the circumference of a perfect circle of

the same area (2v?? * ??). So, a long and narrow patch will have a higher Shape Index than a more nearly circular patch. It is important in the context of the current work because the patch shape can affect its ecosystem vulnerability to external influences. As a reminder, a circular form offers the greatest possible area for a given perimeter.

The landscape data recorded in each patch allowed the quantification of the landscape composition that can be expressed by 3 types of indices (BUREL et al., 1999). The first one concerns the landscape richness and is expressed by the number of land cover classes it contains (nLC). The second index type refers to the equitability of the land cover classes it outlines, i.e. the proportion taken by the different land covers described earlier (WOOD, HERB, WATER, CONC). The third type of index of landscape composition is the heterogeneity, which combines the two previous ones. Heterogeneity is synonym of diversity and many indices allow its quantification. The one used in this work is the Shannon Index applied to the landscape (SHANL), defined as (SHANNON, 1948; SPELLERBERG and FEDOR, 2003)

where ???? = ???? / ?? is the relative frequency of each land cover within each patch. It gives an index of the combination of richness and evenness of the land cover.

Several spatial variables were considered in order to test if some of the previous landscape variables could be correlated with the patch position into the study area. Those spatial variables could as well explain the distribution of some species. They were defined on the basis of the WGS84 geographic coordinates of the central point of each park. In addition to the simple latitudes (X) and longitudes (Y) other spatial variables were calculated: X², Y², XY, X²Y, XY², X²+Y², X³, Y³, X³+Y³ (GAILLY, 2013).

A correlation matrix was realized in order to identify the highly redundant variables within the landscape and spatial variables. The goal hereby is to delete the correlated variables and therefore to decrease the number of variables in order to allow a better visibility of the results. The correlation coefficient of Pearson has to be higher than 70% in order to present a correlation.

The data analysis was split into four sections in order to achieve the four objectives of the thesis. We started with an evaluation of the bird diversity through the calculation of various parameters of the species assemblages in each patch. Following that, we analyzed the ornithological communities potentially found in the overall patches through multivariate analysis. The environment within the patches was then characterized and finally, the response of the bird assemblages to the driving forces affecting their habitat in Bangkok was investigated. Descriptive analysis and map of the distribution of those parameters across the study area were performed in order to illustrate the results before further interpretation.

The two first descriptive parameters of the bird distribution considered are species richness and abundance. In order to get closer with a major urban issue presented before, namely the biotic homogenization phenomenon, a Community Specialization Index (CSI) was calculated.

Species Richness is one of the basic measures of diversity. The species richness of a patch was defined in this work as being the total number of species recorded across the 6 sampling periods.

In order to allow a comparison between the assemblages, the sampling effort is assumed to be equal. However, the completeness of the surveys in the 25 patches must be considered. This calculation required the drafting of species saturation curves. Thus, for each patch, the random

settlement of the samples allowed the establishment of a mean cumulative richness curve. The most accurate the survey, the more this curve tends to an asymptote. The sample completeness is calculated as the ratio between the number of species observed and the real number of species within a patch. The EstimateS(c) 9.1.0 software was used to compute non-parametric, asymptotic species richness estimators for incidence-based data (COLWELL, 2013) and allowed the establishment of those mean curves of cumulative richness. The real number of species could be estimated via a first order Jack-knife extrapolation (GAILLY, 2013) which is function of the number of species that only appeared in one sample: «unique» species (HELTSHE and FORRESTER, 1983).

A map was created showing the species richness of each patch within the study area in order to obtain an overview of the richness distribution. Then, in order to show the species distribution in the overall patches, we estimated for each species a distribution range going from very low, while the species was only present in 1 to 5 patches to very high when the species was recorded in 20 to 25 patches. A column chart was drawn in order to illustrate the bird species distribution.

In order to get a bird abundance estimation within each patch, we decided to use a heterogeneity index since a simple sum of the species abundance scores for each patch doesn't take into account the individuals repartition between the different species. Thus, to synthetize the number of species and the equilibrium of the individuals' repartition into one quantified variable, we used the Shannon Index of Species Diversity (SHANNON, 1948; SPELLERBERG and FEDOR, 2003):

where pi = ⁿⁱ/_N is the proportion of species i in the green patch. This index gives an appropriate

index of bird diversity at a site because it takes into account the abundance of each species recorded at this site (SODHI et al., 1999).

Then, to show the species abundance distribution along the patches, we estimated for each species its total abundance on the study area going from very low, while the species had a cumulative abundance score ranging from 1 to 25, to very high when the species showed a cumulative abundance score ranging from 101 to 125. A column chart was drawn as well in order to illustrate the abundances distribution.

⁶ As a reminder, the abundance data was split into ordinal categories of abundance which were given an abundance score ranging from 1 to 5.

«Biotic homogenization» is a quite contemporary term traducing a biodiversity erosion process resulting from a diminution of the species community variability among sites or habitats and not from an impoverishment of the species communities in themselves (VAN TURNHOUT et al., 2007; SAX and GAINES, 2003).

This homogenization process generally appears while there is fragmentation or degradation of a habitat and is linked with an augmentation of generalist species to the detriment of more specialized species. Hence, the degree of specialization of a bird species to a given habitat class is positively related to the species abundance along that habitat class gradient (DEVICTOR et al., 2008). Therefore, a more specialized species will show higher densities with the augmentation of a particular resource while more generalist species will demonstrate little variation across habitats. The approach of JULLIARD et al. (2006) was used in order to quantify the Species Specialization Index (SSI) as the coefficient of variation (Standard Deviation / Average) of the species densities among the patches (DEVICTOR et al., 2008). The SSI allowed ranking all considered species from the most to the least specialized, whatever the species size, ecology, habitat preference and under any site classification (JULLIARD et al., 2006).

Then, a CSI was measured in order to estimate the functional homogenization. The CSI for a green patch ?? is given by the mean SSI of species present at a given site (DEVICTOR et al., 2008).

where ?? is the total number of species recorded, ?????? is the relative abundance of species ?? in patch ??, and ???????? its specialization index.

The research for species associations is one of the usual problems of community ecology (LEGENDRE and LEGENDRE, 2012). Until now, the analyses focused on the overall species diversity and did not take into account species assemblages. In order to include them, we need to use multivariate methods.

- Ordinations methods: identify gradients within the data set whilst opposing species and sites that are the most different (find continuities).

- Cluster methods: identify sites or species groups that share a maximum of similarities classification of the objects in groups (find discontinuities).

Those two approaches, while opposed, are in fact complementary to analyze multivariate data sets (DUFRÊNE, 2003). They will both be outlined in this work (Figure 9) and they were performed using the RStudio(c) software (Version 0.98.976). A research of indicator species into the different clusters was then operated using the same software.

The adopted methodology was set using the abundance scores matrix. Indeed, the abundance indices truly give more interpretation possibilities than presence/absence data. We thus assume that the necessary conditions (equal detectability), in order to compare the bird assemblages between the patches based on the abundance scores, are respected (JONES, 1998).

Figure 9: Method used for the ornithological communities analysis: Bray-Curtis = methods used for the creation of the distance matrix, Ward= Ward's minimum variance method used for the cluster analysis, PCoA= Principal Coordinate Analysis used for the ordination analysis, IndVal=Indicator Species Analysis.

Firstly, the data matrix was conversed into a distance matrix through the estimation of the sites «proximity» regarding the observed species and their abundance. It is therefore a multivariate measure of the differences existing between sites.

Among the distance calculation methods, that of Bray-Curtis (D 14) was selected for the analysis on the species communities based on the abundance data. Bray-Curtis index is a distance coefficient calculated from the Steinaus similarity coefficient (S17) that compares two sites (x1, x2) in terms of the minimum abundance of each species. The D14 formula is

where W is the sum of the minimum abundances of the different species, this minimum being defined as the abundance at the site where the species is the rarest. A and B are the sums of the abundances of all species at each of the two sites respectively (LEGENDRE and LEGENDRE, 2012). This index was chosen because it can be used on abundances frequencies and because a same difference between two sites for abundant of rare species has the same contribution to the similarity.

The cluster analysis gives a graphical representation (dendrogram) that was used to determine, from the Bray-Curtis distance matrix, the similar bird communities' compositions among the different sites. It also provided the opportunity to see how the sites rank in comparison to each other.

A hierarchical agglomerative clustering method was used to join together the more similar sites. The term «hierarchical» signifies that the position of a site was definitively imposed within a branch of the classification. The term «agglomerative» refers to the discontinuous partition of the objects where those are considered as being separate from one another. They are successively grouped into larger and larger clusters until a single, all-inclusive cluster is found (LEGENDRE and LEGENDRE, 2012).

The Ward's minimum variance method was chosen (e.g. CONOLE and KIRKPATRICK, 2011; GAILLY, 2013; LOUGBEGNON and CODJIA, 2011) because it aims at finding compact, spherical clusters. To form clusters, this method minimizes the variance across each group as the sites agglomeration progress in minimizing the sum of squared deviation distances from the centroid of each group. Compared to other clustering methods, the Ward's minimum variance method overestimates the distances between sites while the first groups are shaped (LEGENDRE and LEGENDRE, 2012).

The sites similarities calculated from the dendrogram and named cophenetic similarities are different from the original similarities that served to cluster the sites together. Hence, the calculation of the cophenetic correlation allowed us to measure the relationship between the distances of Ward's clustering and the original ones (LEGENDRE and LEGENDRE, 2012).

The Principal Coordinate Analysis (PCoA) allowed us to position the 25 patches, in a space of reduced dimensionality while preserving their distance relationships calculated above as well as possible. The PCoA used the distance matrix calculated before with the goal to find out the axes that maximized that distance matrix (LEGENDRE and LEGENDRE, 2012).

Some distance measures might be negative and did not allow a proper ordination of sites in a full Euclidean space. Therefore, they need a correction for the negative eigenvalues before being used for ordination by PCoA (LEGENDRE and LEGENDRE, 2012).

An indicator species must meet two specific criteria (DUFRÊNE, 2003): it must dominate into a group of sites (specificity index, percentage of individuals in the group) and it must occupy all of the sites within that group (fidelity index, percentage of occupied sites in the group). The IndVal method (DUFRÊNE and LEGENDRE, 1997) was used to identify the species that can be considered as associated to a site or a group of sites. Hence, this method takes into account the specificity and fidelity indices while calculating the species indicator value, giving a percentage proportional to the previous indices. The highest IndVal value obtained identifies the group where the species can be considered as indicator. In addition, the IndVal method allows to test statistically if a species can be considered as being indicator or not.

Moreover, the sum of the species indicator values can be assumed as a criterion in order to compare the clusters formed and choose which one explains the species distribution best (DUFRÊNE and LEGENDRE, 1997).

The following analyses aimed to achieve an investigation of the environmental factors affecting the different patches. It is indeed essential to examine those factors and their distribution within the study area as a first step to interpret the ornithological distribution. We used a multivariate approach using the RStudio(c) software (Version 0.98.976).

To determine the interactions that the environmental variables tended to have with each other in our study area, we performed a principal component analysis (PCA). The PCA is an ordination method that considers the variables altogether and allows the creation of a space of reduced dimensionality from an extraction of the axis that maximize the variance of multidimensional cloud of points shaped by the initial environmental variables. In other words, it condenses the dimension of the cloud of points and helps choose a point of view to look at this cloud of points. The two axes are perpendicular with each other (so they are not redundant) and each one explains

a part of the initial cloud information. Those new descriptors of the information are called principal component and are linear combinations of the initial variables. The roles of those new axes are interpreted while examining the initial variables that are the more correlated (ROBERTS, 2010).

Values of sites characteristics were correlated with the first two principal axes using Pearson and Spearman rank-order correlations coefficients. The reason we used the two methods is that Spearman's correlation coefficient is a non-parametric method unlike Pearson's. Non-parametric methods consist in finding a coefficient correlation, not between the values taken by the two variables, but between ranks of those values. It allows to find out monotone correlations although the variable distributions are skew. The Pearson method only consists of finding the linear relation between values and is easily biased if the two variables don't have a Gaussian distribution or show exceptional values.

The comparison of the two correlation coefficients values brought information about the bias of the correlation calculated. Hence, if Pearson coefficient was higher than the Spearman's one, this means that exceptional values could be present. This would increase the Pearson coefficient value but not modify the more robust Spearman's coefficient values.

IV.4.4. Environmental explicatory factors of the ornithological distribution analysis

For the thesis to be more than descriptive work, an investigation of the influence that the environmental factors have on the bird distribution was realized. The analyses were divided in two sections. The first analysis investigated the influence of the environmental factors on the previously calculated descriptive parameters of the ornithological through multivariate analysis (indirect gradient analysis). Then, Generalized Linear Models (GLM) were realized to demonstrate the relations existing between the environmental variables and our diversity measures. The third analysis directly confronted the bird abundance data matrix with the environmental factors through a multivariate analysis called the redundancy analysis (RDA; direct gradient analysis).

To determine how the ornithological descriptive parameters (species richness, CSI and Shannon index) were affected by site characteristics, the ornithological parameters were correlated with the previous PCA axes using Pearson and Spearman correlation coefficients.

Regressions equations allowed us to formalize the relationship existing between explicative variables and variables to explicate. We used GLMs in order to determine which variables or sets of variables had the largest influence on the overall species richness, Shannon index of heterogeneity and CSI.

For the following analyses, we transformed our previous set of explanatory environmental variables so they approach a normal distribution. A p-value was estimated for each variable through a Shapiro-Wilk test (ROYSTON, 1982) to confirm or not the normal distribution.

All GLM models were constructed in Rstudio(c). We used information-theoretic methods to compare models incorporating different environmental variables, because of the advantages of that method over stepwise approaches for this type of analysis (WHITTINGHAM et al., 2006). To identify the most biologically relevant models, we used the small sample size Akaike's Information Criterion (AICc). We identified the model showing the lowest AICc as the best model (BURNHAM and ANDERSON, 2004; WHITTINGHAM et al., 2006).

Two other criterions were important as well to give information on the model quality. The residual standard error (RSE) that describes the standard deviation of points formed around the linear function, and estimates the accuracy of the dependent variable being measured. And the coefficient of determination (Adjusted-R²) that describes the percentage of the observation variability that is explain by the regression equation.

Relationships between communities and the overall environmental variables were assessed with a canonical ordination method. The RDA is a multivariate analysis method commonly used in ecology in order to analyze simultaneously two data tables (e.g. GAILLY, 2013; SATTLER et al., 2010). It is the multivariate analog of simple linear regressions. The RDA analysis combines the ordination and regression concepts. It allows the ordination of the «species» variables constraints by a canonical axis that is maximally related to a linear combination of the environmental variables (LEGENDRE and LEGENDRE, 2012). It allows the representation at the same time of the species, the patches and the environmental variables in a space of reduced dimensionality.

We used a RDA in this thesis to link the species abundance data set together with the environmental variables. We also projected the Ward clusters obtained previously in the space of reduced dimensionality in order to characterize them.

The 6 respective visits into the 25 patches allowed us to survey 49 bird species. After deletion of the non-breeding visitors (9 species), passage migrants (2 species), presumed escaped birds (4 species) and a resident bird species that was every times recorded as flying over the census area (Hirundo Rustica), 34 species were kept for the analyses (Appendix 2). The following results give an overview of the ornithological descriptive parameters distribution in the 25 patches. The notations for the 3 variables will be as follow: bird_R (bird species richness), SHANB (Shannon index of bird diversity), CSI (community specialization index).

A conservation value was first planned to be calculated for each patch, however it has not been completed since out of all the bird recorded, none of the species were listed as «threatened» or «near threatened» in the «Asian Bird Red Data Book» (BIRDLIFE INTERNATIONAL, 2001).

All of the 34 species can be assumed to be synanthropic species in view of the situation of the study area. Two of the species kept for the analyses were identified as invasive species according to the GLOBAL INVASIVE SPESCIES DATABASE (2005): Columbia Livia (Rock Pigeon) as an alien invasive species and Acridotheres tristis (Common Myna) as a native invasive species. Nevertheless, the number of known alien species in Thailand was still far from being realistically estimated (NAPOMPETH, 2002).

Table 7 shows the species richness distribution within the 25 patches and its completeness while comparing them with the Jack-Knife estimated real richness.

All of the cumulative richness curves tend to an asymptote. Figure 10 shows two distinct curves got for patch No. 8 and patch No. 3. Their Jack-Knife estimated real richness tends to the same number of species (21.5 and 21.3 species respectively), however the two curves show different shapes. Indeed, the surveys in patch No.3 reached faster the real species richness than the surveys carried out in patch No. 8. Hence, at least four surveys were necessary in this case to get reliable data. Concerning the slopes the curves at the last survey, the No. 3 is 0.5 and No. 8 is 0.3. This signifies that in theory to get one new species, 2 supplementary surveys needs to be realized for patch No. 3 and 3 for patch No. 8. The samplings completeness is therefore acceptable.

Another observation made while comparing the species richness results is the time of the day of the survey. After a verification of the normality of the data, a paired t-test of Student was done in order to compare the mean bird richness observed in a park during the morning to the one observed in the afternoon. The test showed that the true difference in means was not equal to 0 (p-value= 8.627x10^-7). Thus, the average species found during the morning surveys appear to be greater than the ones during the afternoon surveys (see Appendix 3 for the entire results).

The Figure 11 below illustrates the previous richness distribution within the study area.

Figure 11: Map of the Species Richness per patch in the study area (bird_R=Bird Species Richness)

Regarding the species distribution, 12 of the 34 species were found in 21 to 25 of the patches surveyed while 11 of them were only found in 1 to 5 patches (Figure 12). Acridotheres grandis (White-vented Myna), Acridotheres tristis (Common Myna) and Copsychus saularis (Oriental Magpie Robin) had the higher range of distribution as they were found in the 25 patches surveyed.

Figure 12: Amount of species characterized by different distribution (number of records) in the study area. Distribution classes: very high (21 to 25 records), high (16 to 20 records), medium (11 to 15 records), low (6 to 10 records) and very low (1 to 5 records)

We generated a map (Figure 13) in order to compare the 25 patches Shannon index of diversity. We found that the patches No.21 and No.22 showed the lowest species diversity while the highest heterogeneity was situated in the patches No.4, No.6 and No.17.

Figure 13: Map of the Shannon Index of Diversity per patch in the study area (SHANB= Shannon index of bird diversity)

Concerning the individuals' relative abundance, it is not surprising that the two species that showed the highest count within the study area appear to be two widespread urban species: Passer montanus (Eurasian Tree Sparrow) and Columba livia (Rock Pigeon). On the other hand, 16 species show a very low abundance (Figure 14).

Figure 14: Amount of species individuals characterized by different relative densities (total of the abundance scores) in the study area. Relative abundance classes: very high (101 to 125), high (76 to 100), medium (51 to 75), low (26 to 50) and very low (1 to 25 records)

The species having the highest SSI (species specialization index) are Amaurornis phoenicurus (White-breasted Waterhen), Passer flaveolus (Plain-backed Sparrow)⁷. Conversely, Copsychus saularis (Oriental Magpie Robin), Acridotheres tristis (Common Myna) and Acridotheres grandis (White-breasted Myna) seems to be the more generalist species.

Concerning the CSI (community specialization index) calculated for all the patches, Figure 15 illustrates its distribution within the study area. The highest CSI was found in patch No.17 while the smallest was in patch No.22.

Figure 15: Community specialization Index (CSI) distribution in the study area

⁷ Apus affinis (House swift) was also listed a one of the most specialized species because it has been recorded only once perched in a patch while it was seen often flying over the patches. This species won't be taken into account in further statistical analyses.

In order to bring to a discussion of the previous parameters describing the ornithological data, it will be interesting to compare the evolution of their values in the 25 patches. Therefore, we decided to draw a graph of their evolution on Figure 16. The x-axis of the graph organizes the patches from the highest to the smallest bird species richness observed. We can see that the general decreasing trend is similarly for the three indices with the presence of peaks and troughs more contrasted for the Shannon index.

Figure 16: Comparison of the describing parameters of the ornithological data calculated in the 25 patches studied
(bird_R =bird species richness, CSI = community specialization index, SHANB= Shannon index of species diversity)

The Ward's minimum variance method allowed us to classify the sites in different groups on the basis of their species composition. The result of the cluster analysis for the abundance data is illustrated on the dendrogram on Figure 17 below.

Figure 17: Dendrogram formed out of the Ward's minimum variance method from the ornithological abundance dataset.
Four major groups were identified

The cophenetic correlation coefficient calculated for Figure 17 has a value of 0.523.

A clear differentiation was firstly made between group 1-2 and 3-4, showing us the fundamental species abundances distance between the patches of those groups. While comparing those results with the previous SHANB curve (Figure 16: Comparison of the describing parameters of the

ornithological data calculated in the 25 patches studied
(bird_R =bird species richness, CSI = community specialization index, SHANB= Figure 16), we can easily assume that the two first groups contain high species abundances while the groups 3 and 4 show lower abundances. Still, four groups were chosen in order to get more precise indication of the bird abundances distributions. This being said, group 4 separates then 2 patches (21 and 22) showing together similar low abundance trends. On the other side, the cut off between groups 1 and 2 is harder to describe.

The ordination method (PCoA in this case) offered the possibility to illustrate the result of the cluster method in a space of reduced dimensionality where the sites have been localized regarding their species abundance scores (Figure 18).

Figure 18: Factorial design created with the two first axis of the PCoA concerning the abundance data. The red ellipses allow the visualization of the four groups of sites defined earlier by the Ward's method.

The two first axes of the PCoA explain 37.09% of the point variability. The first axis (component 1) explains 24.03 % while the second (component 2) explains 13.05%.The first axis shows well the isolation of groups 3 and 4 from the other groups that are partially overlapping each other. Regarding the avian influence, it is quite hard to visualize bird assemblages at this point.

The IndVal method applied on the abundance data allowed the identification of 11 significant indicator species (p-value < 0.05) separating from each other two groups of sites. All the species listed in Table 8 were selected because they were indicators of the first and second group of sites defined by the Ward clusters (None of the species listed in the two other groups had a significant IndVal value). Some species were not significantly kept from the IndVal analysis due to their lack of specificity or their rarity in the surveys that didn't satisfy the two criterions of the method (i.e. specificity and fidelity).

Table 8: Species selected via the IndVal method as being significantly associated to a group of sites

The two groups obtained in the previous table showed various visible differences. First, considering the species record frequencies in the 25 patches, the first group holds widespread synanthropic species while compared with the second one. However, the species of the second group show higher IndVal indices than in the first one.

The correlation matrix (see Appendix 5) showed that a few of the environmental variables (SHANL and nLC; SHANL and WOOD) presented together a correlation coefficient higher than 70% and there was no correlation of the landscape variables with the spatial variables. The landscape richness (nLC) was eliminated from the landscape variables because the information given was poor, while we decided to keep all the others.

By contrast, the spatial variables were for the most part correlated. As their redundancy didn't offer real facilities for the subsequent interpretations, 9 of the 11 variables were deleted. The two left are X³+Y³ and Y³, as together, they were correlated with all of the other spatial variables.

The PCA (principal component analysis) reported the relationships between variables in the research of explanatory factors (for the numerical values see Appendix 6). Before the start of the analysis, the environmental factors have been log or squared-root transformed in order to tend as much as possible to a normal distribution. Hence, the variables AREA, HERB and WATE have been log-transformed while the WCON, SFACTOR, WOOD and Y3 were square root-transformed.

The two first axes of the PCA (PC1 and P) explain respectively 41% and 19% of the total variance of the data set, which means a total of 60%.

As we assume an axis is correlated if |r|>0.60 (Table 9), the axis 1 of the PCA appears positively correlated to SHANL (91%), logAREA (76%), logHERB (74%), SqrtSFACTOR (73%) and logWATE (68%) and negatively correlated to SqrtWOOD (80%) regarding the Pearson correlation coefficients (Figure 19). Concerning their Spearman correlation indices, they are quite close of the previous ones except for the SqrtFACTOR variable that doesn't appear correlated in this case (Figure 19). It is also remarkable that the variable WCON shows here an opposite sign. This is due to the presence of three exceptionally great values that overestimate the Pearson correlation coefficient value.

The landscape heterogeneity appears to parallel the all other factors on the first axis except for the percentage of wooded area and the connectivity of a site (the latter correlation is low). This axis opposes therefore the more open, large sites having a more heterogeneous landscape to the more wooded ones. It is true that within our study area, a bunch of very small patches were wood highly covered while the large patches contained more water and herbaceous cover, meaning more heterogeneous landscapes.

Axis 2 seems to be relatively correlated with the spatial variables (Y3: r =65%) regarding its Pearson coefficient, especially with the longitude which is a direction going from the South to the North. However, the Spearman correlation coefficient is lower than 0.6 because of three exceptional values (three patches being together far more North than the others). Moreover, the percentage of concrete in the sites seems to be negatively correlated to this axis (CONC: r =-66%).

Then, Figure 20 displays the representation of the patches along the two axes with the roles of those new axes interpreted.

Table 9: Pearson and Spearman correlation coefficients between the environmental variables and the two axes of the PCA.
The numbers highlighted show a correlation with the axis

Figure 19: Representation of the environmental variables in the Pearson and Spearman correlation circles formed by the two

The longer the red arrow is, the more variance is explained from the factorial design and the closer the arrow is from an axis, the more it contributes.

Figure 20: Factorial design created with the two first axis of the PCA concerning the environmental data.

ENVIRONMENTAL FACTORS EXPLAINING THE ORNITHOLOGICAL DISTRIBUTION V.4.1. Indirect gradient analysis

The study of the role played by the environmental factor on the ornithological distribution started with an analysis of the correlation of the ornithological descriptive parameters with the two axes of the previous PCA on the environmental factors.

Figure 21 Representation of the environmental variables (red arrows) together with the ornithological descriptive parameters (green arrows) in the Pearson and Spearman correlation circles formed by the two first axes of the PCA. Only one green arrow represented the three ornithological parameters in the Pearson correlation circle for a better visibility as the three arrows were merged together.

Figure 21 and Table 10 show both the positive correlation of the ornithological description variables with the first axis of the PCA. This means that species richness, Shannon index and CSI, became greater within patches showing larger areas as well as more open and heterogeneous landscapes.

Table 10: Pearson and Spearman correlation coefficients between the ornithological variables and the two axis of the PCA.
The numbers highlighted show a correlation with the specified axis.

For the following analysis, it is important to note that three environmental variables didn't reach a normal distribution (Shapiro-Wilk test: p<0.05): WCON, SHANL and Y3. We kept the same transformation as before for the other environmental variables. Table 11 presents a summary of the GLMs.

Table 11: General linear models and summary statistics for ornithological variables. Predictor environmental variables are patch area (logAREA, log-transformed), weighted connectivity (SqrtWCON, Square root-transformed), S-Factor (SqrtSFACTOR, Square root-transformed), Wood cover (SqrtWOOD, Square root-transformed), Herbaceous cover (logHERB, log-transformed), Water cover (logWATE, log-transformed), Concrete cover (CONC), Y³ (SqrtY3, Squared root transformed), X³+Y³(X3Y3). a= first parameter, b= second parameter (environmental factor coefficient), R²= coefficient of determination, RSE= Residual Standard Error, AICc = Akaike's information criterion corrected for small sample size. The best models are indicated in bold for each ornithological variable.

	~logAREA	-13.612	2.993	0.721	2.692	125.524
	~SHANL	8.531	10.176	0.467	3.720	141.691
	~logWATE	13.536	2.866	0.428	3.855	143.475
	~SqrtSFACTOR	-8.423	22.996	0.323	4.191	147.656
	~logHERB	13.484	2.126	0.236	4.453	150.683
bird_R
	~SqrtWOOD	28.371	-1.456	0.168	4.648	152.827
	~X3Y3	-2,087.000	2,066.000	0.098	4.840	154.849
	~SqrtY3	-507.424	10.305	0.058	4.946	155.934
	~CONC	18.208	-6.208	0.019	5.047	156.949
	~SqrtWCON	17.443	15.060	-0.017	5.138	157.836
	~SHANL	1.618	0.866	0.594	0.247	6.176
	~logWATE	2.062	0.232	0.490	0.277	11.861
	~logAREA	0.542	0.178	0.423	0.295	14.964
	~SqrtSFACTOR	0.471	1.699	0.302	0.324	19.722
	~SqrtWOOD	3.376	-0.134	0.263	0.333	21.062
SHANB
	~logHERB	2.091	0.157	0.218	0.343	22.545
	~X3Y3	-201.200	0.000	0.184	0.351	23.608
	~SqrtWCON	0.929	-0.054	-0.043	0.358	24.598
	~SqrtY3	-34.474	0.723	0.042	0.380	27.611
	~CONC	2.435	-0.366	-0.006	0.389	28.853
	~logAREA	-0.598	0.108	0.693	0.104	-37.202
	~logWATE	0.395	0.095	0.342	0.152	-18.141
	~SHANL	0.251	0.315	0.317	0.155	-17.225
	~SqrtSFACTOR	-0.271	0.709	0.214	0.166	-13.717
	~logHERB	0.396	0.070	0.178	0.170	-12.582
CSI
	~SqrtWOOD	0.925	-0.054	0.167	0.171	-12.256
	~SqrtY3	-20.240	0.408	0.073	0.181	-9.595
	~X3Y3	-69.200	0.000	0.071	0.181	-9.531
	~SqrtWCON	0.521	0.617	-0.011	0.189	-7.430
	~CONC	0.565	0.000	-0.038	0.191	-6.757

Considering concrete cover, weighted connectivity or the two spatial variables as predictors for the three ornithological variables respectively, provided the lowest relative statistical evidence and explanatory power.

Species richness and CSI increased with the patch area while Shannon's index Bird diversity increased with the water cover rate (Figure 22 and Figure 23).

Figure 22: Residuals plots of best GLM: bird species richness (top row) and log-transformed CSI (bottom row) as a function of the patches log-transformed area

Some of the patches deviate from the regression line in both plots (Figure 22). Patches No. 3, 5, 6, 10 and 19 showed higher bird richness while patches No. 7, 11, 18, 20 and 24 contain lower bird species than the regression line. Those trends need an interpretation regarding the factors affecting the deviation of the patches in comparison with the ones situated along the regression line (patches No. 15, 21, 6,1 for example). The same was completed with the CSI plot.

Figure 23: Residuals plots of best GLM: Shannon index of bird diversity as a function the patches log-transformed water cover.

In this case, the patches are more scattered. The patches No. 3 and 16 show both the higher deviance from the regression line, the first one showing a higher Shannon index of bird diversity and the second showing a lower one.

Finally, a redundancy analysis (RDA) was realized with the ornithological data (abundance matrix) and the environmental factors in order to identify the relationships between the two data sets.

As we have 10 environmental variables, we have 10 constrained ordination axes (RDA 1 to RDA 10). In total, the variance explained by the constrained axes (independent variables, i.e. environmental variables) is equal to 53% while the rest (47%) is explained by the unconstrained axis (dependent variables, i.e. bird abundance).

We will focus on the two first axes of the RDA (Figure 24) that show the highest variance explained in order to simplify the interpretation. The red names represent each individual species displayed in the RDA space and the blue vectors show how the environmental variables fall along that RDA space. The longest vectors along each RDA axis are the most influent in explaining variation of species abundance along that axis.

The first two constraints axes of the RDA explain respectively 37% and 16 % of the common variance among species and environmental factors (a total of 53% explained). While those two axes explain respectively 9% and 8% of the species abundance variance (17% in total). The plot (Figure 24) doesn't allow us to visualize the actual species impacts on the patches distribution between the two RDA axes. Regarding the environmental variables, three variables are negatively correlated (|r|>0.60) with the first axis of the RDA: logAREA, logWATE and SHANL (Table 12).

Figure 24: Representation of the species abundance and environmental variables in the plot formed by the two first axes of the RDA. The green ellipses show the Ward clusters obtained previously. An ellipse contains 80% of the patches of a group.

The four previous clusters obtained at section IV.4.2 via the Ward minimum variance method can be visualized on the factorial design shaped by the two first axes of the RDA (Figure 24). That allows their characterization. Group 1 and 2 are thus characterized by more open (more herbaceous and concrete cover), heterogeneous landscapes while group 3 and 4 seem characterized by a more wooded context. Group 2 is elongated by the spatial variables, situating those patches more in the North of the study area. They are also characterized by large patches with water.

Table 12: Pearson and Spearman correlation coefficients between the environmental variables and the two axes of the RDA.
The numbers highlighted show a correlation with the specified axis.

Over the century, as urbanization growth accelerates, urban green areas are facing severe recession. As the urban sprawl is continually extending due to land speculation and uncontrolled development, wildlife and people are getting closer and closer and the conservation of urban biodiversity emerges as a concern of rising importance (MCKINNEY, 2002).

The potential impact of urbanization on the avifauna has been largely studied worldwide. However, although Bangkok is a booming megacity where the environment has been severely damaged, there has not been any previous studies focusing on the overall urban bird species distribution (ROUND, 2008). Therefore, this thesis was the timely opportunity to report the situation of the birds in Central Bangkok so there can be basis implemented for a future monitoring.

We will guide the discussion through the answers to the two research questions, namely:

- How is the avifauna characterized and distributed into green patches situated in the center of the Bangkok Metropolis?

- How do the environmental parameters of those green patches influence the bird distribution?

In order to best inform the wildlife conservation strategies, a section regarding the conservation implications will then be highlighted. Finally, some result need to be nuanced and a section will focus on the limitations of this study.

HOW IS THE AVIFAUNA CHARACTERIZED AND DISTRIBUTED INTO GREEN PATCHES SITUATED IN THE CENTER OF THE BANGKOK METROPOLIS?

The results obtained through the ornithological distribution analysis bring a descriptive overlook of Bangkok's avifauna distribution. Those results will be discussed together with the literature review statements in order to be correctly replaced in the thesis context.

Firstly, the bird species sampled (49 species) represented 13 % of the birds assumed to be in the region of the Bangkok Metropolis all year round (LEPAGE, 2014). This is a quite poor amount but it was predictable as we focused our study area in the most urbanized part of the city. It confirmed the statements made by many previous studies, suggesting that species richness decreases as urban development increases (CHACE and WALSH, 2006; DEVICTOR et al., 2008; IMAI and NAKASHIZUKA, 2010; MCKINNEY, 2006; NIELSEN et al., 2013; ORTEGA-ÁLVAREZ and MACGREGOR-FORS, 2009; SANDSTRÖM et al., 2006).

A great part of the bird species found in the green patches sampled were common native synanthropic species in the Bangkok region like previously pointed out by ROUND (2008). A great part of those species were found in the first indicator species group (Table 8), those species being widespread, often abundant and easily detectable. Those common native species are of great importance for conservation purpose, since their presence contributes to the structure, biomass and energy turnover of the environment they live in (GASTON, 2010). It is all the more important as those species remain frequent victims of habitat loss as well as species invasion and this can have deep impacts on their environment and the ecosystem services they provide. Indeed, the interactions between those common species with city dwellers can influence positively their wellbeing and increase their relation with nature (MILLER and HOBBS, 2002). The latter argument is essential because it brings a socio-economic motivation for the Bangkok administration to invest in the support of avian biodiversity.

The species abundance structure (Figure 14) showed that most of the urban species recorded are non-abundant while only a few are very abundant. Indeed, some species are able to find alternative ecological niches in the cities and develop quite significant populations alike the common native synanthropic species listed before. However, those adaptions were also made by few well known urban-exploiter bird species (native or alien) that created excessive populations like Columbia Livia (Rock Pigeon, alien) and Acridotheres tristis (Common Myna, native) pointed out as invasive species in the GLOBAL INVASIVE SPECIES DATABASE (2005). Today's issue is that those abundant species tend to become overabundant and create competition with other native birds, forcing the decreasing of the latter's abundance (CONOLE and KIRKPATRICK, 2011; DEVICTOR et al., 2008; MCKINNEY, 2006). It is nevertheless important to nuance the statement that stipulates that a species is invasive at the whole study area scale. Indeed, even if some species were more widespread and abundant than others across the whole study area, it would be essential to look at the invasiveness problem at the patch scale and explore the metapopulation dynamics (ANDERIES et al., 2007).

The PCoA ordination axes (Figure 18) don't explain a high part of the variance between the patches regarding their bird assemblages based on the species abundance. External factors could better explain the bird abundance distribution within the study area. Indeed, neutral mechanisms like biotic processes such as scattering or competition, may play a subordinate role in structuring community composition in urban spaces. According to SATTLER et al. (2010), human disturbance happening in this case on a regular and frequent basis by the multiple human activities could cause inhibitions of both the development and installation of spatially organized biotic processes. Environmental parameters may play a role as well and their case is discussed below.

HOW DO THE ENVIRONMENTAL PARAMETERS OF THOSE GREEN PATCHES INFLUENCE THE BIRD DISTRIBUTION?

In order to go further than an assessment of the effect of urbanization on birds through the creation of species lists and to best inform wildlife-conservation strategies, it is crucial to understand the ornithological responses to the modified environmental features of Bangkok.

We know that Bangkok urban green patches are essential to provide habitat for birds (ROUND, 2008). It is true that the use of a habitat by a bird differs between every species as well as in between the same species (FULLER, 2012). However, for conservation purposes, it is important to attribute the species found in each patch to the way they use their environment through the evaluation of the observed differences in overall species abundances. The environmental features investigated in the urban green patches of central Bangkok supported various bird assemblages that may in part reflect the availability of different resources. For example, green patches with more water were more inclined to host bird characteristics of wetlands. Nevertheless, analyses of community-wide indices are complex to interpret because they contain a composite response of many individual bird species (PEARSON, 1993).

The previous tests realized through section V.4 showed that in the context of Central Bangkok, the size of the patches had the highest influence on bird species richness and CSI (Figure 22 and Table 11). This doesn't follow statements made by previous researchers in other regions, which demonstrated that green patches' internal habitat qualities are of greater importance than both park size and park isolation for the bird richness and composition (NIELSEN et al., 2013). We explained the patches showing higher species richness (or CSI) than the regression line (Figure 22) as being more heterogeneous patches. The ones below the line, especially patch No. 11 (Sanam Luang Park) are assumed as being more homogeneous, in this case with a high herbaceous cover rate.

The bird species heterogeneity (SHANB), on the other hand (Figure 23), was the most influenced by the rate of water cover in the patches. In this case, the patches below the regression lines seemed to be homogeneous but with high wooded cover rates in the case of patch 16, therefore, the detection of the birds is assumed as poorer than in other patches.

The following paragraphs study separately the environmental variables types regarding their influence on the ornithological dataset.

As suggested by other studies as well as by the IBT ( e.g. CASTELLETTA. et al., 2005; FERNÁNDEZ-JURICIC and JOKIMÄKI, 2001; MACARTHUR and WILSON, 1963, 1967, NIELSEN et al., 2013), a biogeographic variable, the area of the patches, was the environmental variable affecting the bird richness the most (Figure 22 and Table 11). The area of the patches was also

of greatest influence on the bird diversity estimated with the Shannon index of heterogeneity that takes into account abundance of each species recorded at one site. Bird communities at a site contained also more specialized species in larger area patches. The isolation index didn't show evident influence on the ornithological variables, contrary to what has previously been predicted by the IBT (MACARTHUR and WILSON, 1963, 1967, NIELSEN et al., 2013). The possible causes will be later discussed in section VI.3.3.

A first landscape factor explaining highly the bird diversity was the rate of water cover. It tended to bring more species to the patches including more wetland specialized species alike Anastomus oscitans (Asian Openbill), Egretta garzetta (Little Egret), Ardeola sp. (Pond Heron), Butorides striata (Little heron), or else ,Amaurornis phoenicurus (White-Breasted Waterhen).

Surprisingly, we found a negative influence of the percentage of wooded cover on the bird species diversity that doesn't corroborates some previous studies that found that trees play an important role in explaining urban bird diversity (e.g. EVANS et al., 2009; SANDSTRÖM et al., 2006; SATTLER et al., 2010). Two reasons can explain those trends; first, the detectability of birds decreased with an augmentation of wood cover and second, the larger patches of our study areas were mostly composed of heterogenetic landscapes while the smaller patches contained merely wood cover.

More herbaceous cover, which was highly correlated with the Shannon index of landscape heterogeneity seemed to have greater influence on the more synanthropic group of species (group 1). The concrete cover showed the same trends but has less influence.

Finally, even if environmental variables accounted for a large percentage of the explanatory power of the bird community models, an important part of the bird community composition seemed to be determined by environmental stochasticity, i.e. «random events such as habitat destruction by human activity, anthropogenic transportation or the introduction of exotic species» (SATTLER et al., 2010). Environmental variables can explain some variation in urban bird community composition, however, stochasticity, appeared to be more important in urban areas than in other habitat types, where avian species communities are far from stable, enduring constant change while adapting themselves to the disturbances and changes that constantly modify their habitat (SATTLER et al., 2010).

As the results obtained before are discussed, it is important to have a critical subsequent approach on the data quality and directions taken through the overall thesis.

VI.3.1. Limits regarding the study scope

Due to all of the constraints affecting our green patches sampling, the patches selected show trends only regarding public areas within the most urbanized part of Bangkok. Other green patches alike zoos, vacant lands, privates gardens or street trees haven't been taken into account although they can be considered as well as green patches and they include avifauna potential habitat. It is therefore important to ask ourselves the question of the chosen patches representativeness as they represent only a part of the green patches within the study area. Furthermore, as the field work depended on existing structures, the patches didn't range in area and land cover types and the representativeness of the patches panel can also be questioned.

A second important fact concerning bird counting is the difficulty of getting relevant abundance data. Indeed, it has never been difficult for an ornithologist to create a list of the bird species recorded. However, calculating species abundance is difficult considering such a mobile animal. Their mobility could lead us to believe that they could be everywhere although this is not the case.

It is true that birds are easily seen compared with other taxa but while counting multi bird species, many studies tend (us also) to assume them as similar. However, identifying the proximal factor of the urban bird diversity is relatively difficult as well as studying urbanizations gradients and biological response because they are far from being linear (MACDONNELL and HAHS, 2008). Indeed, many parameters affect the bird species differently according to one species eco-ethology: detectability, response to the environment changes, adaptation to human disturbances...It is therefore a limit to assume the anthropic effect or the habitats as equal in the overall patches while those factors have variable influences for each species. Statistical models for individual species would be best understood from the perspective of each species' natural history and habitat preferences but as we focused on the overall bird distribution and because the number of species was important, we chose not to proceed in this way (PEARSON, 1993).

Nevertheless, as the study was made in order to implement the basis for long-term monitoring, the data obtained will be compared to future data collected in a similar way. Therefore, while forgetting the bias brought by another observer, the abundance scores specified will allow future surveys to detect large scale changes in the abundance of individual species.

Concerning the identification process, the fact that the field work was a first ornithological survey experience and, in a foreign country, and despite the great motivation put into the bird visual and sound identification, a professional birdwatcher would surely have seen more birds. Hence, the species identifications we made were accurate concerning the abundant species while the multiplication of visual encounters allowed us to be sure of the species recognition. On the other hand, non-abundant less detectable species can lead to identification mistakes.

VI.3.3. Limits due to the choices of environmental indices

Some of the previous calculated environmental indices may not have any impact on the bird distribution. However, it is also possible that the indices used were not well adapted to the organism studied. For example, the study of isolation through the calculation of an index of weighted connectivity has many limits. First, as said before, the patches sampled were not as islands of habitats surrounded by a hostile urban matrix as presumed through the IBT. Indeed, street trees or public garden or other habitats could be find inbetween the patches which brings a relatively high bias to the index calculation. Second, from one bird species to another, the distances are not assumed as equal and the study of an isolation index should be done separately for each species regarding its eco-ethology.

«There is no catch-all conservation strategy for wildlife conservation. Targeting a specific species of concern, a functional group or the proper response variable will lead to greater gains in comparable conservation efforts.» (GALITSKY, 2012)

Despite the previous limitations, our study has major implications to help improve the efficiency of bird conservation efforts in Bangkok. A geographic layout of the urban biodiversity hotspots was set with the distribution of the ornithological parameters described throughout this thesis. To protect, preserve and restore functional green infrastructures in urban green spaces for birds, large areas should contain compact and clustered trees separated with sufficient amount and quality of open vegetation and water inbetween.

However, the implementation of such strategies faces significant challenges. Increasing the size of existing urban green patches is difficult, if not impossible in cities. Therefore, strategies to enhance habitat diversity and resource availability for birds within the patches appears to be a straightforward way of increasing urban bird diversity.

Furthermore, initiative like the one coming from the Bird Conservation Society of Thailand, which explored the possibility to implement an urban bird reserve in eastern Bangkok (ROUND , 2008), should be brought forward as well.

VII. CONCLUSION AND PERSPECTIVES

As there is much concern today about environmental changes, it is essential to know how those changes affect wildlife over time, and birds offer a great value as biological and environmental indicators (BIBBY, 1999; GOTTSCHALK et al., 2005).

This study allowed us to conclude that in Central Bangkok, the more an urban green space is large and with a heterogeneous landscape, the more bird species will be numerous and the more specialized species will be found. The influence of the water cover rates on the species abundance diversity was demonstrated as well.

The previous results describing the bird distribution are mostly descriptive and of little use in themselves but will be of high interest while compared with futures bird counts conducted in a similar way. Indeed, they will provide a quantification of urban bird diversity evolution across time since birds are useful indicators of changes within their environment (BIBBY, 1999; KOSKIMIES, 1989), their quantification is indispensable in order to implement management measures or to reach a priority for future actions. It is crucial for the Bangkok government to promote a sustainable development within the metropolis.

As the landscape is not supposed to change in those green areas, a landscape monitoring across time doesn't need to be implement. However, the environmental factors could be complemented by calculating more intrinsic environment structures within the patches with authorizations and maps from the BMA and a scale-dependent study could also be apply.

Another clue could be the implementation of volunteer-based surveys that provide sharing resources to facilitate science and management as well as an avenue for urban conservation to engage a broader audience. Since the interest for nature is growing more and more within the inhabitant of Bangkok (pers. Obs.), initiatives of counting projects from the Bird Conservation Society of Thailand need to be encouraged and may be successful.

The development of sustainable cities is today the major goal for urban landscape planners, government authorities and conservationists (WU, 2009). Programs that aim to find a sustainable balance between the today's traditional vegetation management and more natural management need to be implemented in order to contribute to the sustainable efforts (SHWARTZ et al., 2013). We shall not forget that each city is a unique system and the occurring management and planning actions should therefore be continuously evaluated to measure their effectiveness

VIII. BIBLIOGRAPHY

ANDERIES J. M., KATTI M. and SHOCHAT E., 2007. Living in the city: resource availability, predation, and bird population dynamics in urban areas. Journal of theoretical biology, 247(1), pp.36-49.

ANDRÉN H., 1994. Effects of habitat fragmentation on birds and mammals in landscapes with different proportions of suitable habitat: a review. Oikos, 71, pp.355-366.

ARAÚJO M.B., 2003. The coincidence of people and biodiversity in Europe. Global Ecology and Biogeography, 12, pp.5-12.

BASTIN L. and THOMAS C.D., 1999. The distribution of plant species in urban vegetation fragments. Landscape Ecology, 14, pp.493-507.

BEUDELS R. J., 2013. Protected Areas: the best tool we have against biodiversity loss. Yet, is it enough? Conference at Gembloux Agro Bio-Tech on October 10^th, 2013.

BIBBY C. J. and BUCKLAND S. T., 1987. Bias of bird census results due to detectability varying with habitat. Acta OEcologica: OEcol. Gener., 8, pp.103-112.

BIBBY C. J., JONES M. and MARSDEN S., eds., 1998a. Expedition Field Techniques. Birds Surveys. Geography Outdoors, London.

BIBBY C. J., MARSDEN S. and FIELDING A., 1998b. Bird-habitat studies. In: BIBBY C. J., JONES M. and MARSDEN S., eds. Expedition Field Techniques. Birds Surveys. Geography Outdoors, London. pp. 99-114.

BIBBY C.J., 1999. Making the most of birds as environmental indicators. Ostrich, 70, pp. 81-88.

BICKFORD D. et al., 2010. Forest Fragment and Breeding Habitat Characteristics Explain Frog Diversity and Abundance in Singapore. Biotropica, 42(1), pp.119-125.

BIRDLIFE INTERNATIONAL, 2001. Threatened birds of Asia: the BirdLife International Red Data Book. Birdlife International, Cambridge.

BLAIR R.B., 1999. Birds and Butterflies along an Urban Gradient : Surrogate Taxa for Assessing Biodiversity ? Ecological Applications, 9(1), pp.164-170.

BOLUND P. and HUNHAMMAR S., 1999. Ecosystem services in urban areas. Ecol. Econ., 29, pp.293- 301.

BRIGGS J. C. 1996. Tropical diversity and conservation. Conservation Biology, 10, pp. 713-718.

BROWN L.R., 2001. Eco-Economy: Building an Economy for the Earth. W. W. Norton & Co., New York.

BUREL F. et al., 1999. Ecologie du paysage : concepts, méthodes et applications. Éd. Tec & doc., Paris.

BURNHAM K. P. and ANDERSON D. R., 2004. Multimodel inference: Under- standing AIC and BIC in model selection. Sociological methods & research, 33, pp. 261-304.

CASTELLETTA M., THIOLLAY J.-M. and SODHI N. S., 2005. The effects of extreme forest fragmentation on the bird community of Singapore Island. Biological Conservation, 121(1), pp.135-155.

CHACE J. F. and WALSH J. J., 2006. Urban effects on native avifauna: a review. Landscape and Urban Planning, 74(1), pp.46-69.

CHAN S. et al., 2004. Important Bird Areas in Asia: Key Sites for Conservation. BirdLife International, Cambridge.

CHOW W.T.L. and ROTH M., 2006. Temporal dynamics of the urban heat island of Singapore. International Journal of Climatology, 26, pp. 2243-2260.

CLERGEAU P., JOKIMÄKI J. and SAVARD J.-P.L., 2002. Are urban bird communities influenced by the bird diversity of adjacent landscapes? Journal of Applied Ecology, 38(5), pp.1122-1134.

COLWELL R. K., 2013. EstimateS: Statistical estimation of species richness and shared species from samples. Version 9.1.0. User's Guide and application. « http://purl.oclc.org/estimates» (22/07/2014).

CONOLE L.E. and KIRKPATRICK J.B., 2011. Functional and spatial differentiation of urban bird assemblages at the landscape scale. Landscape and Urban Planning, 100(1-2), pp.11-23.

DE RIDDER K. et al., 2004. An integrated methodology to assess the benefits of urban green space. The Science of the total environment, 334-335, pp.489-497.

DELOYA M.C., 1993. Urban forestry activities in Mexico. Unasylva, 173(44), pp. 28-32. DEMBNER S.A., 1993. Urban forestry in Beijing. Unasylva, 173(44), pp. 13-18.

DEVICTOR V. et al., 2008. Functional biotic homogenization of bird communities in disturbed landscapes. Global Ecology and Biogeography, 17(2), pp.252-261.

DIAMOND J.M., 1975. The island dilemma: lessons of modern biogeographic studies for the design of natural reserves. Biological Conservation, 7, pp.129-146.

DOLMAN P.M., 2012. Mechanisms and processes underlying landscape structure effects on bird populations. In: FULLER R. J., 2012. Birds and habitat: relationships in changing landscapes. Cambridge University Press, New York. pp. 93-124.

DUFRÊNE M. and LEGENDRE P., 1997. Species Assemblages and Indicator Species: The Need for a Flexible Asymmetrical Approach. Ecological Monographs, 67(3), pp.345-366.

EVANS K.L., NEWSON S.E. and GASTON K.J., 2009. Habitat influences on urban avian assemblages. Ibis, 151(1), pp.19-39.

FAHRIG L., 2003. Effects of habitat fragmentation on biodiversity. Annual Review of Ecology, Evolution, and Systematics, 34(1), pp.487-515.

FERNÁNDEZ-JURICIC E. and JOKIMÄKI, J., 2001. A habitat island approach to conserving birds in urban landscapes: case studies from southern and northern Europe. Biodiversity and Conservation, 10, pp.2023-2043.

FOOD AND AGRICULTURE ORGANIZATION OF THE UNITED NATIONS (FAO), 2012. State of the World's Forests 2012, Food and Agriculture Organization of the United Nations, Rome, Italy.

FORMAN R.T.T. and GODRON M., 1986. Landscape Ecology. John Wiley and Sons Inc., New York.

FORMAN R.T.T., 1995. Land mosaics: the ecology of landscapes and regions. Cambridge University Press, Cambridge.

FRASER E.D.G., 2002. Urban Ecology in Bangkok, Thailand: Community Participation, Urban Agriculture and Forestry. Environments, 30(1), pp.37-50.

FULLER R. J., ed., 2012. Birds and habitat: relationships in changing landscapes. Cambridge University Press, New York.

GAILLY R., 2013. Impact des plantations de sapins de Noël sur l'avifaune des milieux ouverts en Ardenne occidentale. Travail de fin d'études : Gembloux Agro Bio-Tech/Université de Liège.

GALITSKY C., 2012. Effects of local vegetation and landscape patterns on avian biodiversity in the threatened oak habitat of the Willamette Valley, Oregon. Master Thesis: University of Washington.

GASTON K.J., 2010. Valuing Common Species. Science, 327(5962), pp.154-155.

GLOBAL INVASIVE SPECIES DATABASE, 2005. Invasive species of the organism type bird in Thailand. « http://www.issg.org/database/species/» (09/05/2014).

GOTTSCHALK T.K., HUETTMANNJ F. and EHLERS M., 2005. Thirty years of analyzing and modeling avian habitat relationships using satellite imagery data: a review. International Journal of Remote Sensing, 26(12), pp.2631-2656.

HELTSHE J.F. and FORRESTER N.E., 1983. Estimating Species Richness Using the Jackknife Procedure. Biometrics, 39(1), pp.1-11.

HOSTETLER M., 1999. Scale, birds, and human decisions: a potential for integrative research in urban ecosystems. Landscape and Urban Planning, 45(1), pp.15-19.

HUSTÉ A. and BOULINIER T., 2011. Determinants of bird community composition on patches in the suburbs of Paris, France. Biological Conservation, 144(1), pp.243-252.

IMAI H. and NAKASHIZUKA T., 2010. Environmental factors affecting the composition and diversity of avian community in mid- to late breeding season in urban parks and green spaces. Landscape and Urban Planning, 96(3), pp.183-194.

IUCN, 2013. The IUCN Red List of Threatened Species. Version 2013.2. « http://www.iucnredlist.org» (09/05/2014).

JONES M., 1998. Study design. In: BIBBY C. J., JONES M. and MARSDEN S., eds. Expedition Field Techniques. Birds Surveys. Geography Outdoors, London. pp. 15-34.

JULLIARD R. et al., 2006. Spatial segregation of specialists and generalists in bird communities. Ecology Letters, 9, pp. 1237-1244.

KATTI M. and WARREN P. S., 2004. Tits, noise and urban bioacoustics. Tree, 19, pp.109-110.

KHEDARI J., SANGPRAJAK A. and HIRUNLABH J., 2002. Thailand climatic zones. Renewable Energy, 25, pp.267-280.

KHERA N., MEHTA V. and SABATA B.C., 2009. Interrelationship of birds and habitat features in urban greenspaces in Delhi, India. Urban Forestry & Urban Greening, 8(3), pp.187-196.

KONIJNENDIJK C. C., 2003. A decade of urban forestry in Europe. Forest Policy and Economics, 5, pp. 175-186.

KOSKIMIES P., 1989. Birds as a tool in environmental monitoring. Ann. Zool. Fennici, 26, pp.153- 166.

KUCHELMEISTER G., 1998. Urban Forestry: Present Situation and Prospects in the Asia and Pacific region. FAO Asia-Pacific Forestry Sector Outlook Study, FAO Working Paper No.: APFSOS/WP/44, Food and Agriculture Organization of the United Nations, Rome.

KÜHN I., BRANDL R. and KLOTZ S., 2004. The flora of German cities is naturally species rich. Evolutionary Ecology Research, 6, pp. 749-764.

LAGHAI H. AND BAHMANPOUR H., 2012. GIS Application in Urban Green space Per Capita Evaluation (Case study: City of Tehran). Annals of Biological Research, 3(5), pp.2439-2446.

LEAKEY R.E. AND LEWIN R., 1999. La sixième extinction: évolution et catastrophes. Flammarion, Paris.

LEGENDRE P. and LEGENDRE L., 2012. Numerical ecology. 3^rd ed., Elsevier, Amsterdam.

LEPAGE D., 2014. Avibase - listes d'oiseaux mondiales : Bangkok Metropolis. « http://avibase.bsc-eoc.org/checklist.jsp?region=TH01bm&list=howardmoore» (10/08/2014).

LIM H.C. and SODHI N.S., 2004. Responses of avian guilds to urbanisation in a tropical city. Landscape and Urban Planning, 66(4), pp.199-215.

LOUGBEGNON T.O. and CODJIA T.C., 2011. Avifaune urbaine de Cotonou et sa distribution en relation avec les facteurs de l'habitat: implications pour l'aménagement écologique de la ville. Afrique Science, 7(1), pp.116-136.

LUSSENHOP J., 1977. Urban cemeteries as bird refuges. The Condor, 79, pp.456-461.

MA Z. et al., 2012. Use of localized descriptive statistics for exploring the spatial pattern changes of bird species richness at multiple scales. Applied Geography, 32(2), pp.185-194.

MACARTHUR R.H. and WILSON E.O., 1963. An Equilibrium Theory of Insular Zoogeography. Evolution, 17(4), pp.373-387.

MACARTHUR R.H. and WILSON E.O., 1967. The Theory of Island Biogeography. Princeton University Press, Princeton.

MACDONNELL M. J. and HAHS A. K., 2008. The use of gradient analysis studies in advancing our understanding of the ecology of urbanizing landscapes: current status and future directions. Landscape Ecology, 23, pp.1143-1155

MACGREGOR-FORS I., 2008. Relation between habitat attributes and bird richness in a western Mexico suburb. Landscape and Urban Planning, 84(1), pp.92-98.

MARZLUFF J.M., BOWMAN R. and DONNELLY R., 2001. Avian Ecology and Conservation in an Urbanizing World. Kluwer Academic Publishers, Dordrecht.

MATEO-BABIANO I.B., 2012. Public life in Bangkok's urban spaces. Habitat International, 36(4), pp.452-461.

MCKINNEY M.L., 2002. Urbanization, biodiversity, and conservation. BioScience, 52, pp.883-890.

MCKINNEY M.L., 2006. Urbanization as a major cause of biotic homogenization. Biological Conservation, 127(3), pp.247-260.

MILLENNIUM ECOSYSTEM ASSESSMENT (MEA), 2005. Ecosystems and Human Well-being: Biodiversity Synthesis. World Resources Institute, Washington.

MILLER J. and HOBBS R., 2002. Conservation where people live and work. Conservation Biology 16, pp.330-337.

NAPOMPETH B., 2002. Thailand. In: PALLETTAWA N., REASER J. K., and GUTIERREZ A. T., eds. Invasive Alien Species in South-Southeast Asia: National Reports & Directory of Resources. pp. 104-106.

NIELSEN A.B. et al., 2013. Species richness in urban parks and its drivers: A review of empirical evidence. Urban Ecosystems, 17(1), pp.305-327.

ORTEGA-ÁLVAREZ R. and MACGREGOR-FORS I., 2009. Living in the big city: Effects of urban land-use on bird community structure, diversity, and composition. Landscape and Urban Planning, 90(3-4), pp.189-195.

PEARSON S.M., 1993. The spatial extent and relative influence of landscape-level factors on wintering bird populations '. Landscape Ecology, 8(1), pp.3-18.

POMEROY D., 1992. Counting Birds. A guide to assessing numbers, biomass and diversity of Afrotropical birds. AWF, Nairobi.

PURVIS A. and HECTOR A., 2000. Getting the measure of biodiversity. Nature, 405(6783), pp.212- 219.

RAMALHO C.E. and HOBBS R.J., 2012. Time for a change: dynamic urban ecology. Trends in ecology & evolution, 27(3), pp.179-88.

ROBERTSON P. A. and LILEY D., 1998. Assessment of sites: Measurement of species richness and diversity. In: BIBBY C. J., JONES M. and MARSDEN S., eds. Expedition Field Techniques. Birds Surveys. Geography Outdoors, London. pp. 76-98.

ROSENZWEIG M.L., 2003. Win-Win Ecology: How Earth's Species can Survive in the Midst of Human Enterprise. Oxford University Press, New York.

ROYSTON P., 1982. Algorithm AS 181: The W test for Normality. Applied Statistics, 31, pp. 176- 180.

SANDSTRÖM U. G., ANGELSTAM P. and MIKUSIÑSKI G., 2006. Ecological diversity of birds in relation to the structure of urban green space. Landscape and Urban Planning, 77(1-2), pp.39-53.

SATTLER T. et al., 2010. Spider, bee, and bird communities in cities are shaped by environmental control and high stochasticity. Ecology, 91(11), pp.3343-53.

SAVARD J.-P.L., CLERGEAU P. and MENNECHEZ G., 2000. Biodiversity concepts and urban ecosystems. Landscape and Urban Planning, 48(3-4), pp.131-142.

SAX D.F. and GAINES S.D., 2003. Species diversity: from global decreases to local increases. Trends in Ecology & Evolution, 18(11), pp.561-566.

SHANNON C.E., 1948. A mathematical theory of communication. Bell System Technical Journal, 27, pp. 379-423.

SHWARTZ A., MURATET A., SIMON L. and JULLIARD R., 2013. Local and management variables outweigh landscape effects in enhancing the diversity of different taxa in a big metropolis. Biological Conservation, 157, pp. 285-292.

SINGH V. S., PANDEY D. N. and CHAUDHRY P., 2010. Urban forests and open green spaces: lessons for Jaipur, Rajasthan, India. Occasional paper from the Rajasthan State Pollution Control Board.

SODHI N.S. and BROOK B.W., 2006. Southeast Asian Biodiversity in Crisis. Cambridge University Press, Cambridge.

SODHI N.S. and LIOW L.H., 2000. Improving Conservation Biology Research in Southeast Asia, Conservation Biology, 14(4), pp.1211-1212.

SODHI N.S. et al., 1999. Bird use of linear areas of a tropical city: implications for park connector design and management. Landscape and Urban Planning, 45(2-3), pp.123-130.

SODHI N.S. et al., 2004. Southeast Asian biodiversity: an impending disaster. Trends in ecology and evolution, 19(12), pp.654-660.

SODHI N.S. et al., 2009. The state and conservation of Southeast Asian biodiversity. Biodiversity and Conservation, 19(2), pp.317-328.

SODHI N.S., KOH L.P. and BROOK B.W., 2006. Southeast Asian birds in peril. The Auk, 123(1), pp.275-277.

SORACE A., 2002. High density of bird and pest species in urban habitats and the role of predator abundance. Ornis Fenn. 79, pp.60-71.

SPELLERBERG I.A.N.F. and FEDOR P.J., 2003. A tribute to Claude Shannon (1916 - 2001) and a plea for more rigorous use of species richness, species diversity and the « Shannon - Wiener» Index. Global Ecology and Biogeography, 12, pp.177-179.

THAIUTSA B. et al., 2008. Urban green space, street tree and heritage large tree assessment in Bangkok, Thailand. Urban Forestry & Urban Greening, 7, pp.219-229

TURNER W. R., 2003. Citywide biological monitoring as a tool for ecology and conservation in urban landscapes: the case of the Tucson Bird Count. Landscape and Urban Planning, 65(3), pp.149-166.

UNITED NATIONS (UN), DEPARTMENT OF ECONOMIC AND SOCIAL AFFAIRS, POPULATION DIVISION, 2012. World Urbanization Prospects: The 2011 Revision: Highlights. New York.

UNITED NATIONS CONFERENCE ON ENVIRONMENT AND DEVELOPMENT (UNCED), 5^th of June 1992. Convention on Biological Diversity, Rio de Janeiro. 1760 UNTS 79; 31 ILM 818.

VAN TURNHOUT C. A. M. et al., 2007. Scale-dependent homogenization: Changes in breeding bird diversity in the Netherlands over a 25-year period. Biological Conservation, 134(4), pp.505- 516.

WHITE E.P. and HURLBERT A.H., 2010. The combined influence of the local environment and regional enrichment on bird species richness. The American naturalist, 175(2), pp.E35-E43.

WHITE J.G. et al., 2005. Non-uniform bird assemblages in urban environments: the influence of streetscape vegetation. Landscape and Urban Planning, 71(2-4), pp.123-135.

WHITTINGHAM M. J., STEPHENS P. A., BRADBURY R.B. and FRECKLETON R.P., 2006. Why do we still use stepwise modelling in ecology and behaviour? Journal of Animal Ecology, 75, pp. 1182- 1189.

WILLIAMS J.N., 2012. Humans and biodiversity: population and demographic trends in the hotspots. Population and Environment, 34(4), pp.510-523.

WORLD HEALTH ORGANIZATION (WHO), 2008. Our cities, our health, our future. Report to the WHO Commission on Social Determinants of Health from the Knowledge Network on Urban Settings.

WU J., 2009. Urban sustainability: an inevitable goal of landscape research. Landscape Ecology, 25(1), pp.1-4.

APPENDIX 3: PUNCTUAL RICHNESS COMPARED WITH THE RICHNESS OBSERVED IN THE MORNING AND IN THE AFTERNOON V

APPENDIX 4: COMPARISON BETWEEN THE PATCHES REGARDING THEIR ORNITHOLOGICAL CHARACTERISTICS. VI

APPENDIX 6: COMPARISON BETWEEN THE PATCHES REGARDING THEIR ENVIRONMENTAL CHARACTERISTICS VIII

APPENDIX 1: GAZETTEER OF THE GREEN PATCHES SAMPLED
PATCH N°	Name	Type	Longitude	Latitude⁸
1	Somdet Saranrat Maneerom Park	Park	100.590320	13.742434
2	Benjasiri Park	Park	100.567424	13.730513
3	Chuvit Garden	Park	100.557054	13.738443
4	Benjakitti Park	Park	100.558569	13.729384
5	Santiphap Park	Park	100.541323	13.762778
6	Chatuchak Park	Park	100.553783	13.804674
7	Pathum Wanaram Temple	Temple	100.536832	13.746267
8	Phanphirom Park	Park	100.591922	13.751225
9	Romaneenaart Park	Park	100.502551	13.749001
10	Santichai Prakan Public Park	Park	100.495482	13.764088
11	Sanam Luang Park	Park	100.493086	13.755204
12	Saranrom Park	Park	100.495169	13.748316
13	Nagaraphirom Park	Park	100.490174	13.746893
14	No Name (under Sirat Expressway)	Park	100.549435	13.757224
15	No Name (under Chalerm Maha Nakhorn Expressway)	Park	100.544085	13.758199
16	Vibhavadi Rangsit Forest Park	Park	100.553345	13.772275
17	Rot Fai and Queen Sirikit Park	Park	100.553254	13.811215
18	Princess Mother Garden	Park	100.560002	13.814623
19	Next to BTS Sathorn	Park	100.514444	13.718354
20	Chaloemprakiarti Forest Park	Park	100.510479	13.719703
21	No Name Next to BTS Thonburi (North)	Crossroad greenery	100.505818	13.721150
22	No Name Next to BTS Thonburi (South)	Crossroad greenery	100.505906	13.720329
23	Tae Chio Cemetery	Cemetery	100.523947	13.714271
24	Lumphini Park	Park	100.541316	13.730972
25	Chulalongkorn University	University	100.530864	13.738592

⁸ Longitude and Latitude are those of the approximate center of the green patches in decimal degrees

⁹ The non-breeding plumage of the Javan, Chinese and Indian pond herons are similar and virtually indistinguishable in the field. Furthermore, some hybrids have been observed by local birdwatchers.

¹⁰ Too far to be well identified. Presumed as being Merops philippinus (Blue-tailed Bee-eater)

¹¹ Too far to be well identified. Presumed as being Treron curvirostra (Thick-billed Green Pigeon)

APPENDIX 3: PUNCTUAL RICHNESS COMPARED WITH THE RICHNESS OBSERVED IN THE MORNING AND IN THE AFTERNOON

Patch n^°		Punctual Richness					Morning			Afternoon
Patch n^°	Average	S. D.	Minimum	Maximum	Average	S. D.	Minimum	Maximum	Average	S. D.	Minimum	Maximum
1	14.3	1.75	11	16	15.0	1.00	14	16	13.7	2.31	11	15
2	14.0	0.89	13	15	14.3	0.58	14	15	13.7	1.15	13	15
3	14.0	1.79	11	16	15.3	0.58	15	16	12.7	1.53	11	14
4	20.5	2.43	17	24	22.3	1.53	21	24	18.7	1.53	17	20
5	17.3	1.97	15	19	19.0	0.00	19	19	15.7	1.15	15	17
6	22.2	1.60	20	24	22.3	1.53	21	24	22.0	2.00	20	24
7	12.7	1.86	10	15	12.7	2.52	10	15	12.7	1.53	11	14
8	11.2	1.47	10	14	12.0	1.73	11	14	10.3	0.58	10	11
9	12.5	2.66	9	16	14.7	1.15	14	16	10.3	1.53	9	12
10	14.2	1.60	11	15	15.0	0.00	15	15	13.3	2.08	11	15
11	11.5	2.26	8	14	12.7	1.53	11	14	10.3	2.52	8	13
12	13.2	1.72	11	16	14.3	1.53	13	16	12.0	1.00	11	13
13	7.2	0.98	6	8	8.0	0.00	8	8	6.3	0.58	6	7
14	14.0	1.41	12	16	15.0	1.00	14	16	13.0	1.00	12	14
15	8.8	2.48	6	13	10.7	2.08	9	13	7.0	1.00	6	8
16	12.0	2.10	9	15	13.7	1.15	13	15	10.3	1.15	9	11
17	26.5	1.22	25	28	26.7	1.53	25	28	26.3	1.15	25	27
18	7.2	1.17	6	9	8.0	1.00	7	9	6.3	0.58	6	7
19	9.5	1.64	7	11	10.3	0.58	10	11	8.7	2.08	7	11
20	8.7	2.66	6	12	11.0	1.00	10	12	6.3	0.58	6	7
21	5.8	1.83	4	9	7.0	2.00	5	9	4.7	0.58	4	5
22	4.8	1.72	3	7	5.0	1.73	4	7	4.7	2.08	3	7
23	15.5	1.22	14	17	16.3	1.15	15	17	14.7	0.58	14	15
24	20.5	1.87	18	23	20.3	1.53	19	22	20.7	2.52	18	23
25	13.8	1.17	12	15	13.7	1.53	12	15	14.0	1.00	13	15

			PATCHES REGARDING THEIR ORNITHOLOGICAL
Patch n°	bird_R	Total sum of Abundance scores	Average Abundance score	Shannon Index	CSI
1	18	54	3.2	2.764	0.616
2	19	51	3.2	2.663	0.435
3	19	57	3.2	2.756	0.476
4	25	65	2.6	3.018	0.807
5	23	60	2.7	2.875	0.655
6	26	69	2.8	2.990	0.828
7	16	51	3.2	2.513	0.407
8	17	49	2.7	2.574	0.743
9	17	51	3.0	2.508	0.534
10	18	61	3.4	2.583	0.512
11	15	45	3.0	2.312	0.451
12	19	57	3.0	2.552	0.512
13	12	40	3.3	2.069	0.389
14	17	55	3.2	2.406	0.438
15	15	41	2.7	2.128	0.414
16	15	44	2.9	2.189	0.388
17	31	88	2.8	2.978	1.128
18	12	37	3.1	1.913	0.399
19	16	49	3.1	2.205	0.436
20	15	37	2.5	2.025	0.540
21	10	33	3.3	1.698	0.337
22	9	29	3.2	1.559	0.290
23	21	63	3.0	2.425	0.660
24	24	72	3.0	2.554	0.782
25	20	58	2.9	2.253	0.542

Cell content: Pearson's correlation (P-value) Highlighted: correlated variables (>0.7)

AREA	AREA 1	WCON	SFACTOR	WOOD	HERB	WATE	CONC	nLC	SHANL	X	Y	X²	Y²	XY	X²Y	XY²	X²+Y²
WCON	0.135732	1
SFACTOR	0.262775	0.52926	1
T	-0.2771	-0.23593	-0.44805	1
H	0.285552	0.441068	0.466348	-0.64643	1
W	0.380991	0.005337	0.191331	-0.35113	-0.10492	1
C	-0.18488	-0.10499	0.044278	-0.5801	-0.01223	-0.10255	1
nLC	0.221454	0.050828	0.518043	-0.65548	0.226946	0.335523	0.510409	1
SHANL	0.384442	0.054317	0.503373	-0.82851	0.436065	0.404938	0.494844	0.880369	1
X	0.180658	0.179194	0.293032	-0.13948	-0.06559	0.502741	-0.10811	0.329612	0.379773	1
Y	0.33663	0.542504	0.422484	-0.07662	0.191201	0.016192	-0.09327	0.219783	0.156515	0.282855	1
X²	0.180614	0.179171	0.292999	-0.13951	-0.06558	0.502736	-0.10807	0.32961	0.379793	1	0.282844	1
Y²	0.337018	0.542869	0.422473	-0.07638	0.191281	0.016144	-0.09369	0.219376	0.156143	0.282869	0.999999	0.282858	1
XY	0.345552	0.541243	0.442938	-0.09234	0.172704	0.085725	-0.10392	0.255137	0.201966	0.409065	0.990947	0.409054	0.990949	1
X²Y	0.348224	0.53183	0.454423	-0.10498	0.153543	0.14652	-0.11184	0.282811	0.239478	0.515537	0.967697	0.515527	0.967701	0.992784	1
XY²	0.342356	0.543469	0.43392	-0.08459	0.182216	0.051828	-0.09938	0.237973	0.179775	0.348254	0.997624	0.348243	0.997626	0.997841	0.982765	1
X²+Y²	0.214244	0.237966	0.332403	-0.14306	-0.03954	0.483911	-0.11503	0.342731	0.383126	0.993136	0.3931	0.993135	0.393113	0.512986	0.61222	0.455505	1
X³	0.18057	0.179148	0.292965	-0.13955	-0.06556	0.502731	-0.10804	0.329608	0.379812	1	0.282833	1	0.282847	0.409043	0.515517	0.348232	0.993134
Y³	0.337406	0.543234	0.422461	-0.07615	0.191361	0.016096	-0.0941	0.218969	0.155771	0.282884	0.999997	0.282872	0.999999	0.990949	0.967703	0.997626	0.393127	0.
X³+Y³	0.185507	0.187659	0.29881	-0.14016	-0.06191	0.500477	-0.10913	0.331742	0.380597	0.999862	0.298736	0.999862	0.29875	0.424153	0.529688	0.363765	0.994941	0.

APPENDIX 6: COMPARISON BETWEEN THE PATCHES REGARDING THEIR ENVIRONMENTAL CHARACTERISTICS

1	31630	1.92E-04	1.311	41	9	20	30	1.266	2595.32	1020410
2	43382	3.67E-04	1.343	55	6	15	24	1.122	2588.57	1019708
3	9849	5.49E-04	1.175	63	12	1	23	0.945	2593.06	1019398
4	247213	3.60E-04	1.266	28	3	52	18	1.097	2587.93	1019439
5	32738	3.29E-04	1.804	58	6	9	28	1.057	2606.86	1018935
6	284069	6.59E-02	2.231	16	54	11	20	1.185	2630.74	1019336
7	47085	4.54E-04	1.182	52	0	2	46	0.785	2597.49	1018789
8	23209	1.92E-04	1.377	35	23	9	32	1.292	2600.30	1020463
9	42631	2.66E-04	1.242	67	7	3	24	0.896	2599.04	1017751
10	14217	2.52E-04	1.369	46	13	2	39	1.071	2607.61	1017546
11	114859	2.29E-04	1.376	26	55	0	20	1.001	2602.56	1017468
12	39042	3.19E-04	1.062	70	5	5	20	0.867	2598.65	1017527
13	7805	2.81E-04	1.056	19	20	2	59	1.034	2597.85	1017375
14	39005	3.36E-04	1.180	35	15	11	39	1.268	2603.71	1019178
15	16701	3.29E-04	1.263	79	7	0	14	0.653	2604.26	1019016
16	22681	3.76E-04	1.858	75	1	16	8	0.754	2612.26	1019305
17	835221	2.99E-04	1.582	50	21	14	14	1.228	2634.48	1019324
18	11964	2.25E-02	1.066	86	0	1	14	0.439	2636.44	1019531
19	9920	3.96E-03	1.638	49	12	0	39	0.997	2581.70	1018095
20	30445	3.38E-04	1.336	52	2	0	46	0.767	2582.46	1017975
21	2929	2.74E-04	0.862	100	0	0	0	0.000	2583.28	1017835
22	3546	3.67E-04	0.783	100	0	0	0	0.000	2582.82	1017837
23	146990	3.05E-04	1.459	41	47	2	9	1.023	2579.40	1018380
24	586824	3.08E-04	1.217	44	14	18	24	1.289	2588.83	1018916
25	118390	2.69E-04	1.287	51	16	5	28	1.132	2593.14	1018604