Predictive vegetation modeling for conservation: impact of error propagation from digital elevation data.

The effect of digital elevation model (DEM) error on environmental variables, and subsequently on predictive habitat models, has not been explored. Based on an error analysis of a DEM, multiple error realizations of the DEM were created and used to develop both direct and indirect environmental variables for input to predictive habitat models. The study explores the effects of DEM error and the resultant uncertainty of results on typical steps in the modeling procedure for prediction of vegetation species presence/absence. Results indicate that all of these steps and results, including the statistical significance of environmental variables, shapes of species response curves in generalized additive models (GAMs), stepwise model selection, coefficients and standard errors for generalized linear models (GLMs), prediction accuracy (Cohen's kappa and AUC), and spatial extent of predictions, were greatly affected by this type of error. Error in the DEM can affect the reliability of interpretations of model results and level of accuracy in predictions, as well as the spatial extent of the predictions. We suggest that the sensitivity of DEM-derived environmental variables to error in the DEM should be considered before including them in the modeling processes.     

The role of local and landscape level measures of greenness in modelling boreal plant species richness

Measurements of primary productivity and its heterogeneity based on satellite images can provide useful estimates of species richness and distribution patterns. However, species richness at a given site may depend not only on local habitat quality and productivity but also on the characteristics of the surrounding landscape. In this study we investigated whether the predictions of species richness of plant families in northern boreal landscape in Finland can be improved by incorporating greenness information from the surrounding landscape, as derived from remotely sensed data (mean, maximum, standard deviation and range values of NDVI derived from Landsat ETM), into local greenness models. Using plant species richness data of 28 plant families from 440 grid cells of 25 ha in size, generalized additive models (GAMs) were fitted into three different sets of explanatory variables: (1) local greenness only, (2) landscape greenness only, and (3) combined local and landscape greenness. The derived richness–greenness relationships were mainly unimodal or positively increasing but varied between different plant families, and depended also on whether greenness was measured as mean or maximum greenness. Incorporation of landscape level greenness variables improved significantly both the explanatory power and cross-validation statistics of the models including only local greenness variables. Landscape greenness information derived from remote sensing data integrated with local information has thus the potentiality to improve predictive assessments of species richness over extensive and inaccessible areas, especially in high-latitude landscapes. Overall, the significant relationship between plants and surrounding landscape quality detected here suggests that landscape factors should be considered in preserving species richness of boreal environments, as well as in conservation planning for biodiversity in other environments.     

Relative influence of environmental factors and fishing on coral reef fish assemblages

Understanding whether assemblages of species respond more strongly to bottom-up (availability of trophic resources or habitats) or top-down (predation pressure) processes is important for effective management of resources and ecosystems. We determined the relative influence of environmental factors and predation by humans in shaping the density, biomass, and species richness of 4 medium-bodied (10–40 cm total length [TL]) coral reef fish groups targeted by fishers (mesopredators, planktivores, grazer and detritivores, and scrapers) and the density of 2 groups not targeted by fishers (invertivores, small fish ≤10 cm TL) in the central Philippines. Boosted regression trees were used to model the response of each fish group to 21 predictor variables: 13 habitat variables, 5 island variables, and 3 fishing variables (no-take marine reserve [NTMR] presence or absence, NTMR size, and NTMR age). Targeted and nontargeted fish groups responded most strongly to habitat variables, then island variables. Fishing (NTMR) variables generally had less influence on fish groups. Of the habitat variables, live hard coral cover, structural complexity or habitat complexity index, and depth had the greatest effects on density, biomass, and species richness of targeted fish groups and on the density of nontargeted fishes. Of the island variables, proximity to the nearest river and island elevation had the most influence on fish groups. The NTMRs affected only fishes targeted by fishers; NTMR size positively correlated with density, biomass, and species richness of targeted fishes, particularly mesopredatory, and grazing and detritivorous fishes. Importantly, NTMRs as small as 15 ha positively affected medium-bodied fishes. This finding provides reassurance for regions that have invested in small-scale community-managed NTMRs. However, management strategies that integrate sound coastal land-use practices to conserve adjacent reef fish habitat, strategic NTMR placement, and establishment of larger NTMRs will be crucial for maintaining biodiversity and fisheries.     

Spatial residual analysis of six modeling techniques

In recent years alternative modeling techniques have been used to account for spatial autocorrelations among data observations. They include linear mixed model (LMM), generalized additive model (GAM), multi-layer perceptron (MLP) neural network, radial basis function (RBF) neural network, and geographically weighted regression (GWR). Previous studies show these models are robust to the violation of model assumptions and flexible to nonlinear relationships among variables. However, many of them are non-spatial in nature. In this study, we utilize a local spatial analysis method (i.e., local Moran coefficient) to investigate spatial distribution and heterogeneity in model residuals from those modeling techniques with ordinary least-squares (OLS) as the benchmark. The regression model used in this study has tree crown area as the response variable, and tree diameter and the coordinates of tree locations as the predictor variables. The results indicate that LMM, GAM, MLP and RBF may improve model fitting to the data and provide better predictions for the response variable, but they generate spatial patterns for model residuals similar to OLS. The OLS, LMM, GAM, MLP and RBF models yield more residual clusters of similar values, indicating that trees in some sub-areas are either all underestimated or all overestimated for the response variable. In contrast, GWR estimates model coefficients at each location in the study area, and produces more accurate predictions for the response variable. Furthermore, the residuals of the GWR model have more desirable spatial distributions than the ones derived from the OLS, LMM, GAM, MLP and RBF models.     

Testing the Generality of Bird-Habitat Models

Bird-habitat models are frequently used as predictive modeling tools—for example, to predict how a species will respond to habitat modifications. We investigated the generality of the predictions from this type of model. Multivariate models were developed for Golden Eagle (Aquila chrysaetos), Raven (Corvus corax), and Buzzard (Buteo buteo) living in northwest Scotland. Data were obtained for all habitat and nest locations within an area of 2349 km². This assemblage of species is relatively static with respect to both occupancy and spatial positioning. The area was split into five geographic subregions: two on the mainland and three on the adjacent Island of Mull, which has one of United Kingdom's richest raptor fauna assemblages. Because data were collected for all nest locations and habitats, it was possible to build models that did not incorporate sampling error. A range of predictive models was developed using discriminant analysis and logistic regression. The models differed with respect to the geographical origin of the data used for model development. The predictive success of these models was then assessed by applying them to validation data. The models showed a wide range of predictive success, ranging from only 6% of nest sites correctly predicted to 100% correctly predicted. Model validation techniques were used to ensure that the models' predictions were not statistical artefacts. The variability in prediction success seemed to result from methodological and ecological processes, including the data recording scheme and interregional differences in nesting habitat. The results from this study suggest that conservation biologists must be very careful about making predictions from such studies because we may be working with systems that are inherently unpredictable.     

Comparative performance of generalized additive models and multivariate adaptive regression splines for statistical modelling of species distributions

Two statistical modelling techniques, generalized additive models (GAM) and multivariate adaptive regression splines (MARS), were used to analyse relationships between the distributions of 15 freshwater fish species and their environment. GAM and MARS models were fitted individually for each species, and a MARS multiresponse model was fitted in which the distributions of all species were analysed simultaneously. Model performance was evaluated using changes in deviance in the fitted models and the area under the receiver operating characteristic curve (ROC), calculated using a bootstrap assessment procedure that simulates predictive performance for independent data. Results indicate little difference between the performance of GAM and MARS models, even when MARS models included interaction terms between predictor variables. Results from MARS models are much more easily incorporated into other analyses than those from GAM models. The strong performance of a MARS multiresponse model, particularly for species of low prevalence, suggests that it may have distinct advantages for the analysis of large datasets. Its identification of a parsimonious set of environmental correlates of community composition, coupled with its ability to robustly model species distributions in relation to those variables, can be seen as converging strongly with the purposes of traditional ordination techniques.     

Multiple stressors and amphibian declines: dual impacts of pesticides and fish on yellow-legged frogs.

More than 40% of Earth's 5700+ amphibian species have undergone recent declines. Despite the likely involvement of multiple factors in driving these declines, most studies continue to focus on single stressors. In California (USA), separate studies have implicated either introduced fish or pesticides as causal agents. To date, however, no study has simultaneously evaluated the respective roles of these two potential stressors nor attempted to assess their relative importance, information critical for the development of effective conservation efforts and environmental policies. We examined the role and relative effect of fish and pesticides on the mountain yellow-legged frog (Rana muscosa) using unusually detailed data sets for a large portion of R. muscosa's historic range in California's Sierra Nevada. Habitat characteristics and presence/absence of R. muscosa and fish were quantified at each of 6831 sites during field surveys. Pesticide use upwind of each site was calculated from pesticide application records and predominant wind directions. Using generalized additive models, we found that, after accounting for habitat effects, the probability of R. muscosa presence was significantly reduced by both fish and pesticides, with the landscape-scale effect of pesticides much stronger than that of fish. The degree to which a site was sheltered from the predominant wind (and associated pesticides) was also a significant predictor of R. muscosa presence. Taken together, these results represent the strongest evidence to date that windborne pesticides are contributing to amphibian declines in pristine locations. Our results suggest that amphibian declines may have complex multi-factorial causes, and caution that single-factor studies that demonstrate the importance of one factor should not be used as evidence against the importance of other factors.     

Sensitivity of ecological models to their climate drivers: statistical ensembles for forcing.

Global and regional numerical models for terrestrial ecosystem dynamics require fine spatial resolution and temporally complete historical climate fields as input variables. However, because climate observations are unevenly spaced and have incomplete records, such fields need to be estimated. In addition, uncertainty in these fields associated with their estimation are rarely assessed. Ecological models are usually driven with a geostatistical model's mean estimate (kriging) of these fields without accounting for this uncertainty, much less evaluating such errors in terms of their propagation in ecological simulations. We introduce a Bayesian statistical framework to model climate observations to create spatially uniform and temporally complete fields, taking into account correlation in time and space, spatial heterogeneity, lack of normality, and uncertainty about all these factors. A key benefit of the Bayesian model is that it generates uncertainty measures for the generated fields. To demonstrate this method, we reconstruct historical monthly precipitation fields (a driver for ecological models) on a fine resolution grid for a climatically heterogeneous region in the western United States. The main goal of this work is to evaluate the sensitivity of ecological models to the uncertainty associated with prediction of their climate drivers. To assess their numerical sensitivity to predicted input variables, we generate a set of ecological model simulations run using an ensemble of different versions of the reconstructed fields. We construct such an ensemble by sampling from the posterior predictive distribution of the climate field. We demonstrate that the estimated prediction error of the climate field can be very high. We evaluate the importance of such errors in ecological model experiments using an ensemble of historical precipitation time series in simulations of grassland biogeochemical dynamics with an ecological numerical model, Century. We show how uncertainty in predicted precipitation fields is propagated into ecological model results and that this propagation had different modes. Depending on output variable, the response of model dynamics to uncertainty in inputs ranged from uncertainty in outputs that matched that of inputs to those that were muted or that were biased, as well as uncertainty that was persistent in time after input errors dropped.     

Relative and interactive effects of plant and grazer richness in a benthic marine community

The interactive effects of changing biodiversity of consumers and their prey are poorly understood but are likely to be important under realistic scenarios of biodiversity loss and gain. We performed two factorial manipulations of macroalgal group (greens, reds, and browns) and herbivore species (amphipods, sea urchin, and fish) composition and richness in outdoor mesocosms simulating a subtidal, hard-substratum estuarine community in North Carolina, U.S.A. In the experiment where grazer richness treatments were substitutive, there were no significant effects of algal or herbivore richness on final algal biomass. However, in the experiment in which grazer treatments were additive (i.e., species-specific densities were held constant across richness treatments), we found strong independent and interactive effects of algal and herbivore richness. Herbivore polycultures reduced algal biomass to a greater degree than the sum of the three herbivore monocultures, indicating that the measured grazer richness effects were not due solely to increased herbivore density in the polycultures. Taking grazer density into account also revealed that increasing algal richness dampened grazer richness effects. Additionally, the effect of algal richness on algal biomass accumulation was far stronger when herbivores were absent, suggesting that grazers can utilize the increased productivity and mask the positive effects of plant biodiversity on primary production. Our results highlight the complex independent and interactive effects of biodiversity between adjacent trophic levels and emphasize the importance of performing biodiversity-ecosystem functioning experiments in a realistic multi-trophic context.     

Coupling ecology and GIS to evaluate efficacy of marine protected areas in Hawaii.

In order to properly determine the efficacy of marine protected areas (MPAs), a seascape perspective that integrates ecosystem elements at the appropriate ecological scale is necessary. Over the past four decades, Hawaii has developed a system of 11 Marine Life Conservation Districts (MLCDs) to conserve and replenish marine resources around the state. Initially established to provide opportunities for public interaction with the marine environment, these MLCDs vary in size, habitat quality, and management regimes, providing an excellent opportunity to test hypotheses concerning MPA design and function using multiple discrete sampling units. Digital benthic habitat maps for all MLCDs and adjacent habitats were used to evaluate the efficacy of existing MLCDs using a spatially explicit stratified random sampling design. Analysis of benthic cover validated the a priori classification of habitat types and provided justification for using these habitat strata to conduct stratified random sampling and analyses of fish habitat utilization patterns. Results showed that a number of fish assemblage characteristics (e.g., species richness, biomass, diversity) vary among habitat types, but were significantly higher in MLCDs compared with adjacent fished areas across all habitat types. Overall fish biomass was 2.6 times greater in the MLCDs compared to open areas. In addition, apex predators and other species were more abundant and larger in the MLCDs, illustrating the effectiveness of these closures in conserving fish populations within their boundaries. Habitat type, protected area size, and level of protection from fishing were all important determinates of MLCD effectiveness with respect to their associated fish assemblages. Although size of these protected areas was positively correlated with a number of fish assemblage characteristics, all appear too small to have any measurable influence on the adjacent fished areas. These protected areas were not designed for biodiversity conservation or fisheries enhancement yet still provide varying degrees of protection for fish populations within their boundaries. Implementing this type of biogeographic process, using remote sensing technology and sampling across the range of habitats present within the seascape, provides a robust evaluation of existing MPAs and can help to define ecologically relevant boundaries for future MPA design in a range of locations.     

Multi-scale marine biodiversity patterns inferred efficiently from habitat image processing

Cost-effective proxies of biodiversity and species abundance, applicable across a range of spatial scales, are needed for setting conservation priorities and planning action. We outline a rapid, efficient, and low-cost measure of spectral signal from digital habitat images that, being an effective proxy for habitat complexity, correlates with species diversity and requires little image processing or interpretation. We validated this method for coral reefs of the Great Barrier Reef (GBR), Australia, across a range of spatial scales (1 m to 10 km), using digital photographs of benthic communities at the transect scale and high-resolution Landsat satellite images at the reef scale. We calculated an index of image-derived spatial heterogeneity, the mean information gain (MIG), for each scale and related it to univariate (species richness and total abundance summed across species) and multivariate (species abundance matrix) measures of fish community structure, using two techniques that account for the hierarchical structure of the data: hierarchical (mixed-effect) linear models and distance-based partial redundancy analysis. Over the length and breadth of the GBR, MIG alone explained up to 29% of deviance in fish species richness, 33% in total fish abundance, and 25% in fish community structure at multiple scales, thus demonstrating the possibility of easily and rapidly exploiting spatial information contained in digital images to complement existing methods for inferring diversity and abundance patterns among fish communities. Thus, the spectral signal of unprocessed remotely sensed images provides an efficient and low-cost way to optimize the design of surveys used in conservation planning. In data-sparse situations, this simple approach also offers a viable method for rapid assessment of potential local biodiversity, particularly where there is little local capacity in terms of skills or resources for mounting in-depth biodiversity surveys.     

Trait groups as management entities in a complex,multispecies reef fishery

Localized stressors compound the ongoing climate-driven decline of coral reefs, requiring natural resource managers to work with rapidly shifting paradigms. Trait-based adaptive management (TBAM) is a new framework to help address changing conditions by choosing and implementing management actions specific to species groups that share key traits, vulnerabilities, and management responses. In TBAM maintenance of functioning ecosystems is balanced with provisioning for human subsistence and livelihoods. We first identified trait-based groups of food fish in a Pacific coral reef with hierarchical clustering. Positing that trait-based groups performing comparable functions respond similarly to both stressors and management actions, we ascertained biophysical and socioeconomic drivers of trait-group biomass and evaluated their vulnerabilities with generalized additive models. Clustering identified 7 trait groups from 131 species. Groups responded to different drivers and displayed divergent vulnerabilities; human activities emerged as important predictors of community structuring. Biomass of small, solitary reef-associated species increased with distance from key fishing ports, and large, solitary piscivores exhibited a decline in biomass with distance from a port. Group biomass also varied in response to different habitat types, the presence or absence of reported dynamite fishing activity, and exposure to wave energy. The differential vulnerabilities of trait groups revealed how the community structure of food fishes is driven by different aspects of resource use and habitat. This inherent variability in the responses of trait-based groups presents opportunities to apply selective TBAM strategies for complex, multispecies fisheries. This approach can be widely adjusted to suit local contexts and priorities.     

Habitat biodiversity as a determinant of fish community structure on coral reefs

Increased habitat diversity is often predicted to promote the diversity of animal communities because a greater variety of habitats increases the opportunities for species to specialize on different resources and coexist. Although positive correlations between the diversities of habitat and associated animals are often observed, the underlying mechanisms are only now starting to emerge, and none have been tested specifically in the marine environment. Scleractinian corals constitute the primary habitat-forming organisms on coral reefs and, as such, play an important role in structuring associated reef fish communities. Using the same field experimental design in two geographic localities differing in regional fish species composition, we tested the effects of coral species richness and composition on the diversity, abundance, and structure of the local fish community. Richness of coral species overall had a positive effect on fish species richness but had no effect on total fish abundance or evenness. At both localities, certain individual coral species supported similar levels of fish diversity and abundance as the high coral richness treatments, suggesting that particular coral species are disproportionately important in promoting high local fish diversity. Furthermore, in both localities, different microhabitats (coral species) supported very different fish communities, indicating that most reef fish species distinguish habitat at the level of coral species. Fish communities colonizing treatments of higher coral species richness represented a combination of those inhabiting the constituent coral species. These findings suggest that mechanisms underlying habitat-animal interaction in the terrestrial environment also apply to marine systems and highlight the importance of coral diversity to local fish diversity. The loss of particular key coral species is likely to have a disproportionate impact on the biodiversity of associated fish communities.     

Two-way coupling versus one-way forcing of plankton and fish models to predict ecosystem changes in the Benguela

‘End-to-end’ models have been adopted in an attempt to capture more of the processes that influence the ecology of marine ecosystems and to make system wide predictions of the effects of fishing and climate change. Here, we develop an end-to-end model by coupling existing models that describe the dynamics of low (ROMS–N₂P₂Z₂D₂) and high trophic levels (OSMOSE). ROMS–N₂P₂Z₂D₂ is a biogeochemical model representing phytoplankton and zooplankton seasonal dynamics forced by hydrodynamics in the Benguela upwelling ecosystem. OSMOSE is an individual-based model representing the dynamics of several species of fish, linked through opportunistic and size-based trophic interactions. The models are coupled through a two-way size-based predation process. Plankton provides prey for fish, and the effects of predation by fish on the plankton are described by a plankton mortality term that is variable in space and time. Using the end-to-end model, we compare the effects of two-way coupling versus one-way forcing of the fish model with the plankton biomass field. The fish-induced mortality on plankton is temporally variable, in part explained by seasonal changes in fish biomass. Inclusion of two-way feedback affects the seasonal dynamics of plankton groups and usually reduces the amplitude of variation in abundance (top-down effect). Forcing and coupling lead to different predicted food web structures owing to changes in the dominant food chain which is supported by plankton (bottom-up effect). Our comparisons of one-way forcing and two-way coupling show how feedbacks may affect abundance, food web structure and food web function and emphasise the need to critically examine the consequences of different model architectures when seeking to predict the effects of fishing and climate change.     

Appraisal of visual assessments of habitat complexity and benthic composition on coral reefs

Visual assessments of topographic habitat structure and benthos on coral reefs were appraised using quantitative data collected from 16 replicate surveys within each of 21 sites on Seychelles reefs. Results from visual assessments of reef benthos were similar to those obtained using techniques frequently used to assess benthic complexity and composition. Visual estimates of habitat topography were correlated with rugosity, reef height and holes of 10–70 cm diameter, whilst visual estimates of benthic composition were very similar to those obtained from line intercept transects. Visual estimates of topography correlated strongly with species richness of fish communities and explained 42% of the variation in these data. The relationship between visual estimates of topography and species richness is strongest with fish 10–30 cm total length (TL), abundance of fish within this size category also correlating positively with topographic visual assessments. Visual techniques are prone to observer bias, however with regular training they can be used to quickly provide a reliable and effective means of assessing habitat complexity and benthos on coral reefs.     

Flows, droughts, and aliens: factors affecting the fish assemblage in a Sierra Nevada, California, stream

The fishes of Martis Creek, in the Sierra Nevada of California (USA), were sampled at four sites annually over 30 years, 1979-2008. This long-term data set was used to examine (1) the persistence and stability of the Martis Creek fish assemblage in the face of environmental stochasticity; (2) whether native and alien fishes responded differently to a natural hydrologic regime (e.g., timing and magnitude of high and low flows); and (3) the importance of various hydrologic and physical habitat variables in explaining the abundances of native and alien fish species through time. Our results showed that fish assemblages were persistent at all sample sites, but individual species exhibited marked interannual variability in density, biomass, and relative abundance. The density and biomass of native fishes generally declined over the period of study, whereas most alien species showed no significant long-term trends. Only alien rainbow trout increased in both density and biomass at all sites over time. Redundancy analysis identified three hydrologic variables (annual 7-day minimum discharge, maximum winter discharge, and number of distinct winter floods) and two habitat variables (percentage of pool habitat and percentage of gravel substrate) that each explained a significant portion of the annual variation in fish assemblage structure. For alien taxa, their proportional contribution to the total fish assemblage was inversely related to mean annual streamflow, one-day maximum discharge in both winter and spring, and the frequency of springtime floods. Results of this study highlight the need for continuous annual monitoring of streams with highly variable flow regimes to evaluate shifts in fish community structure. Apparent successes or failures in stream management may appear differently depending on the time series of available data.     

Going with the flow: using species-discharge relationships to forecast losses in fish biodiversity

In response to the scarcity of tools to make quantitative forecasts of the loss of aquatic species from anthropogenic effects, we present a statistical model that relates fish species richness to river discharge. Fish richness increases logarithmically with discharge, an index of habitat space, similar to a species-area curve in terrestrial systems. We apply the species-discharge model as a forecasting tool to build scenarios of changes in riverine fish richness from climate change, water consumption, and other anthropogenic drivers that reduce river discharge. Using hypothetical reductions in discharges (of magnitudes that have been observed in other rivers), we predict that reductions of 20-90% in discharge would result in losses of 2-38% of the fish species in two biogeographical regions in the United States (Lower Ohio-Upper Mississippi and Southeastern). Additional data on the occurrence of specific species relative to specific discharge regimes suggests that fishes found exclusively in high discharge environments (e.g., Shovelnose sturgeon) would be most vulnerable to reductions in discharge. Lag times in species extinctions after discharge reduction provide a window of opportunity for conservation efforts. Applications of the species-discharge model can help prioritize such management efforts among species and rivers.     

Effects of Human Population Density and Proximity to Markets on Coral Reef Fishes Vulnerable to Extinction by Fishing

Coral reef fisheries are crucial to the livelihoods of tens of millions of people; yet, widespread habitat degradation and unsustainable fishing are causing severe depletion of stocks of reef fish. Understanding how social and economic factors, such as human population density, access to external markets, and modernization interact with fishing and habitat degradation to affect fish stocks is vital to sustainable management of coral reef fisheries. We used fish survey data, national social and economic data, and path analyses to assess whether these factors explain variation in biomass of coral reef fishes among 25 sites in Solomon Islands. We categorized fishes into 3 groups on the basis of life‐history characteristics associated with vulnerability to extinction by fishing (high, medium, and low vulnerability). The biomass of fish with low vulnerability was positively related to habitat condition. The biomass of fishes with high vulnerability was negatively related to fishing conducted with efficient gear. Use of efficient gear, in turn, was strongly and positively related to both population density and market proximity. This result suggests local population pressure and external markets have additive negative effects on vulnerable reef fish. Biomass of the fish of medium vulnerability was not explained by fishing intensity or habitat condition, which suggests these species may be relatively resilient to both habitat degradation and fishing. Efectos de la Densidad de Poblaciones Humanas y la Proximidad del Mercado sobre Peces de Arrecifes de Coral Vulnerables a la Extinción     

High seagrass diversity and canopy-height increase associated fish diversity and abundance

Seagrass species function as typical foundation species that unifies most ecosystem processes. This ecosystem role depends largely on the morphological characteristics and structural complexity of seagrass beds, including their ecological importance for fish species. This study examined relationships between seagrass bed characteristics and associated fish communities in mixed seagrass beds. Correspondence analysis (CA) and canonical correlation analysis (CCoA) were performed to estimate relationships for individual seagrass bed characteristics. The CCoA results revealed that species richness and three-dimensional structure of seagrass had great effect on the biomass and richness of the associated fish community. The CA results revealed that the relative importance of seagrass bed characteristics differed among fish functional groups including fishes appearing on the surface of, inside, and on the bottom of seagrass beds. The fishes found on the surface of the beds preferred beds with low seagrass biomass and high three-dimensional structure, those inside the beds preferred beds with high seagrass biomass and high three-dimensional structure, and those on the bottom of the beds preferred locations with low seagrass biomass and low three-dimensional structure. The results of this study provide compelling evidence that seagrass beds with high species diversity and high three-dimensional structure, but intermediate biomass, may provide the great benefit to the associated fish community. Such niche complementarity among fishes may be a process facilitated by seagrass diversity for secondary production as an ecosystem functioning.