This study consisted of four chapters treating the question whether marine artificial reefs can effectively replace lost natural habitats. The underlying idea is straightforward: if marine artificial reefs can mimic efficiently natural reefs, they may help replacing lost natural habitats despite such a claim, evidences are lacking. This study aims to assess the conservation potential of artificial reefs with reference to fish assemblages and developed a new distinction method, Habitat Association Index. Using this new distinction method, this study examined the question by applying two different areas. The details are described according to the chapters as follows.

I. Introduction: defining subject and objectives

Despite the commonness of the concept, defining artificial reef remains an open challenge, as there is no widely accepted definition. In this study, the definition proposed by Seaman (2000) is adopted: "one or more objects of human origin deployed purposefully on the seafloor and which influence physical, biological or socioeconomic processes related to living marine resources".

Marine artificial reef studies commonly refer to the distinction between resident and non-resident species (e.g. D'Anna et al., 1994; Johnson et al., 1994; Moreno, 2002; Jan et al., 2003). However, it appears that there is no consensus for the number of subdivisions in artificial reef communities. While not focused on artificial reef communities, a research by Magurran and Henderson (2003) demonstrated that any community can fundamentally be divided into two groups: resident (or core) and transient (or vagrant) species. The suitability of the artificial reef as a replacement habitat may be hard to demonstrate if the species is transient, as there is no direct link between habitat presence and transient species occurrences. Nevertheless, if the species becomes resident, it indicates that the reef succeeded in providing a habitat, which can replace successfully natural reef habitat. As the concept of residence is essential to demonstrate the suitability of an artificial reef to replace successfully some lost natural habitats, distinguishing resident from transient species is essential.

The exclusive focus of this study on fish assemblages along marine artificial reefs is explained by two reasons. First, literature for marine artificial reef is more important for fish assemblages than for fouling organisms or crustaceans. Second, due to practical reasons, the field data collected in the scope of this study were exclusively related to fish assemblages.

This study has three main goals: 1/ demonstrate the absence of suitable method to distinguish accurately and objectively fish species along marine artificial reefs; 2/ develop suitable theoretical tools to overcome this distinction problem; 3/ apply these newly developed tools to a case study in order to compare fish assemblages along marine artificial reefs and surrounding natural reefs and consequently assess the potential for marine artificial reefs to replace lost natural habitats.

II. Why are available distinction methods not suitable?

As literature shows, artificial reef scientists refer to subjective methods to distinguish resident from transient species (i.e. no theoretical construct to justify their method). For instance, while Relini et al. (1994) requires a species to be present in at least 50 % of the total number of samples to be considered as resident, Costello and Myers (1996) categorise as resident any species which occur in all the samples but two whereas Talbot et al. (1978) define a species as resident if it appears in at least two consecutive samples. Not only do available definitions lack of theoretical construct but also the absence of common distinction method makes any comparison between studies an open challenge.

The core-satellite theory is a general theory in community ecology which aims to distinguish in any ecological region the core (i.e. the resident species) from the satellite species (i.e. the transient species). The distinction method proposed by this theory is based on a spatial approach: several locations within the eco-region are sampled simultaneously. This spatial approach, and its dedicated theoretical tools, is barely suitable for reef scientists who generally adopt a temporal approach (i.e. a single site sampled numerous times over time).

Contrary to the core-satellite theory, Magurran and Henderson (2003) propose a distinction method based on a temporal approach. The authors underline that, fundamentally, plotting species persistence against maximum abundance represents an efficient and objective way to distinguish resident from transient species. The gap, which appears along the persistence axis represents then the boundary between the resident and the transient species. Magurran and Henderson (2003) demonstrate the objectivity of this method by cross-checking results with three other methods: fitting a Poisson process to the species occurrence patterns; referring to the species' biological characteristics; calculating the reciprocal form of the Simpson Index (D) for each persistence level species group and looking at its relation with regard to the number of sample collected. When applied to four different fish assemblage datasets collected along four different artificial reefs, none of the four objective distinction tools successfully used by Magurran and Henderson (2003) is able to provide conclusive results (Boisnier et al., 2010). This lack of efficiency may be primarily related to the fact that Magurran and Henderson's tools were designed using exceptional datasets (samples collected over decades). As a result, objective theoretical tools capable of identifying accurately resident from transient species in relatively limited datasets have yet to be developed.

III. Developing an objective and accurate method to distinguish resident from transient fish species along marine artificial reefs

In practice, two elements are required to be able to distinguish efficiently and accurately resident from transient species: first, an accurate measurement of the habitat association level of every species present in the dataset; second, the ability to determine from which habitat association level a species passes from being transient status to resident status.

HAI has no unit and is calculated for every species present in the dataset. It is simply interpreted as the level of association between the species and the habitat where it appears. To be calculated, HAI simply requires the collection of presence-absence data. Hence, for any species j, this index is defined mathematically as following:

HAIj=In(AItj+e)×(MPj+MAj/APj+AAj)×(N/nj)

where:

Altj is the number of times presence and absence for species j alternate in the dataset;

e is the Napier's constant;

MPj is the maximum number of consecutive samples in which species j is present;

MAj is the maximum number of consecutive samples in which species j is absent;

APj is the average presence of species j in consecutive samples;

AAj is the average absence of species j in consecutive samples;

Nj is the total number of temporal units;

nj is the number of temporal units where species j appears.

First, one of the interesting properties of this index is to provide a similar output, whatever the temporal unit selected (samples, seasons or years). Second, it is worth pointing out that MAj and AAj calculations exclude both the absences occurring prior to the first appearance of the species in the dataset, and the absences occurring after the last occurrence of the species in the dataset, because it is important to remove absences that are not meaningful or misleading. Third, it is necessary to underline the key influence of sampling design on HAI calculation. Two species drawn from different assemblages may be associated to different HAI values if the two fish assemblages are sampled according to different designs. As a result, when sampling design differs, our index should never be used to compare species habitat association levels.

The more present a species is, the lower its HAI tends to be. Besides, the intensity of the relation between HAI and occurrence rate differs between the less present and the most present species. A reference to two logical assumptions enables to justify theoretically a deeper look at this pattern difference: first, the resident and the transient species constitute the two extremes of a continuum; second, the two extremes of a continuum are expected to display the most distinctive patterns (i.e. to display the maximum level of differences). As a result, amongst all the possible ways to classify species into two groups, looking for the combination offering the maximum difference between groups with reference to the relation HAI-occurrence rate may enable to distinguish resident from transient species. In order to highlight this particular combination, the following method is proposed: first, all the possible combinations to classify the occurring species into two groups are considered (minimum species number per group = one); second, for each combination, a multiple analysis of variance (MANOVA) is performed to test the effect of grouping species (fixed factor) on both occurrence rate and HAI (dependent variables). This effect can be easily assessed through the p value associated with the Pillai's trace. If significance appears (i.e. if the linear model is enhanced by introducing the variable "groups"), it indicates that the assemblage sub-division is meaningful; third, in order to highlight the particular combination which provides the most different groups with reference to the relation between HAI and occurrence rate, this study focuses on the partial η2 associated to the Pillai's trace and looks for the one closest to 1, as a partial η2 indicates the weight of this difference in the model beyond its simple statistical significance. The reliability of this three-step method is demonstrated by applying it to two different datasets (collected in Yamagata prefecture and Okayama prefecture respectively) and comparing the result with the result obtained by plotting persistence against maximum abundance (Boisnier et al., 2009).

IV. Can marine artificial reefs effectively replace lost natural marine habitats?

Various artificial reefs were deployed in Funakoshi bay in the 90's. This study focuses on the assemblages associated with small structures (triangle-shaped concrete blocks whose volume equals about 6 m3) which lie in the Southern part of the bay, at a depth of about 2.5 to 3 meters. Three species (Sebasticus marmoratus (Cuvier, 1829), Pagrus major (Temminck and Schlegel, 1843) and Ditrema temminckii (Bleeker, 1853)) were spotted as resident at both sites. The species similarity level (SSL) observed fluctuated significantly according to the reference. Hence, if the total number of species caught is used as a reference, the number of fish species common to both sites is clearly weak (always inferior to 50 %) and can even drop to levels as low as 20 %. Conversely, if we focus exclusively on the three resident species, the SSL equals 100 % throughout the entire period, indicating a perfect similarity between the two assemblages. This result clearly show that marine artificial reef are able to mimic surrounding natural reefs so efficiently that resident species becomes strictly identical at both site.

Numerous artificial reefs have been deployed off Okayama city in the last decade in order to provide alternative habitat to S. marmoratus whose natural habitat has been severely reduced. Different types of reefs have been built: from tiny reefs (the smallest is the "model 1.0" with a volume of 3.125m3) to huge structures such that the "model 10.0" with a volume of 646.4m3. It is worth wondering if all these structures are equally suitable for this particular species. Four different designs were compared ("models 1.8, 2.2, 6.0 and 7.0"). While S. marmoratus was recorded along the four reefs, it only became resident in a single case, suggesting that the "model 6.0" is closer from the S. marmoratus' natural habitat than the three others. This is consistent with previous studies and suggests that if artificial reef may effectively replace lost natural habitats, its efficiency will be related to the degree of structural similarity with the natural habitat.

水産資源涵養を目的に開発されてきた人工魚礁であるが、近年、効果的に天然魚礁の機能を模倣できるならば、消失した自然ハビタットの代替えが可能ではないかという考えから、自然保全ツールとして期待されている。そこで、この可能性について検討した。本論文は、次に述べる4章から構成されている。

第1章では、まず、人工魚礁についてSeaman (2000)の定義を採用し、海洋生物資源に関連した物理、生物、社会経済的過程に影響を及ぼす海底上に意図的に設置した1つあるいはそれ以上の人工物とした。天然魚礁に置き換わるハビタットとしての人工魚礁の好適性を判定する場合、人工魚礁と出現との間に直接的連関がない通過種と、密接な関係がある定住種とを区別し、定住種について比較することが必要となる。現在までの人工魚礁の研究では魚類群集を重視していること、本研究で解析するデータが魚類群集だけであることから、人工魚礁に集まる魚類を対象とした。

第2章では、定住種と通過種を区分する現状の方法がなぜ不適当か論じた。定住種を、Relini (1994)は全調査回数中、最低50%に出現する種、Costello and Myers (1996)は全調査時に出現する種、Talbot et al. (1978)は、少なくとも2回の連続した調査に出現する種と定義しており、共通の区分法がなかった。また、生態学理論に基づく区分法は長期間のデータセットが必要で、実際に適用するのは困難なことを述べた。

第3章では、人工魚礁に出現する定住種と通過種を区別する、効率的で客観的な、また、正確に、それぞれの種のハビタットとの連関のレベルと、通過種から定住種に変わるレベルとを決定できるHabitat Association Index (ハビタット連関指数:HAI)を考案した。ハビタットとの連関の強さを示すHSIを無次元の指数として次式で定義した。なお、添え字jはjという種を表す。

HAIj=In(AItj+e)×(MPj+MAj/APj+AAj)×(N/nj)

ここで、Altjは全調査回数におけるj種の出現と不出現が変わった回数、eは自然対数の底、MPjとMAjはj種の最大連続出現回数と最大連続不出現回数、APjとAAjはj種の平均連続出現回数と平均連続不出現回数、Nは全調査回数、njはj種の全出現回数である。この指数の特徴は、季節、年など時間スケールによらず、同じような結果が得られること、ある種が初めて出現する前の不出現および最後に出現してからの不出現の両方をMAj とAAjで排除でき、意味のない不出現や間違った解釈につながる不出現を排除できることである。HAIjは調査回ごとのj種の出現頻度が高くなれば低くなる。出現種を2つのグループに分ける組み合わせを考え、それぞれの組み合わせに多重分散解析を行い、出現率とHAIに及ぼす組み合わせの効果を、Pillaiのトレースと関係するp値を用いて評価し、有意性を判定できる。そして、最も異なるグループをPillaiのトレースのpartial η2に注目し、1に最も近いものを探すことで判定できる。なお、調査が異なる設計で行われた場合には、HAIの値を直接比較できない。

第4章では、HAIを2例の事例に適用し、検討した。岩手県船越湾南部の底深2.5-3 mに設置された6 m3の体積もつテトラポッドでつくられた人工魚礁と近くの天然魚礁とに出現する魚類を調べた。カサゴ、マダイ、ウミタナゴの3種がHAIにより定住種として区別された。漁獲された全種を用いると、人工魚礁と天然魚礁の間の種の類似レベルは常に50%以下と低かった。しかし、上述の3種に着目すると全調査で人工魚礁と天然魚礁の間の3種の出現は100%重なり、完全な類似を示した。この結果は、定住種は両方の場所で同一で、人工魚礁で天然魚礁を置き換え可能であることを意味していた。

岡山県沖合には、カサゴの減少したハビタットの代替のため、最小3.125m3から、最大646.4m3までの範囲にある4種の異なった構造の人工魚礁が設置されている。これらの魚礁のデータを比較したところ、自然ハビタットとの構造的類似性が高い1基のみカサゴが定住種となっており、人工魚礁の構造が消失した自然ハビタットに置き換わる能力に大きな違いを生じさせることを示唆していた。

以上、これらの成果は今後の海洋における環境修復に応用可能であり、環境学の研究として価値あるものである。なお、本論文第2章の一部は、佐川龍之、小松輝久、高木儀昌との共同研究であるが、論文提出者が主体となって分析及び検証を行ったもので、論文提出者の寄与が十分であると判断する。

したがって、博士(環境学)の学位を授与できると認める。