Environ Monit Assess (2011) 178:455–460 DOI 10.1007/s10661-010-1704-0
Sampling state and process variables on coral reefs Roger H. Green · Brian A. McArdle · Robert van Woesik
Received: 7 May 2009 / Accepted: 6 September 2010 / Published online: 21 September 2010 © Springer Science+Business Media B.V. 2010
Abstract Contemporary coral reefs are forced to survive through and recover from disturbances at a variety of spatial and temporal scales. Understanding disturbances in the context of ecological processes may lead to accurate predictive models of population trajectories. Most coral-reef studies and monitoring programs examine state variables, which include the percentage coverage of major benthic organisms, but few studies examine the key ecological processes that drive the state variables. Here we outline a sampling strategy that captures both state and process variables, at a spatial scale of tens of kilometers. Specifically, we are interested in (1) examining spatial and temporal patterns in coral population size-frequency distributions, (2) determining major population
processes, including rates of recruitment and mortality, and (3) examining relationships between processes and state variables. Our effective sampling units are randomly selected 75 × 25 m stations, spaced approximately 250–500 m apart, representing a 103 m spatial scale. Stations are nested within sites, spaced approximately 2 km apart, representing a 104 m spatial scale. Three randomly selected 16 m2 quadrats placed in each station and marked for relocation are used to assess processes across time, while random belt-transects, re-randomized at each sampling event, are used to sample state variables. Both quadrats and belt-transects are effectively subsamples from which we will derive estimates of means for each station at each sampling event. This nested sampling strategy allows us to determine critical stages in populations, examine population performance, and compare processes through disturbance events and across regions.
Based on a platform presentation at the 11th International Coral Reefs Symposium (ICRS), July 7–11, 2008, Fort Lauderdale, Florida.
Keywords Coral reef · Climate change · Populations · Sampling · Processes
R. H. Green University of Western Ontario, London, Canada B. A. McArdle University of Auckland, Auckland, New Zealand R. van Woesik (B) Florida Institute of Technology, Melbourne, FL, USA e-mail:
[email protected]
Introduction Projected climate change is set to drive temperature and seawater chemistry to levels outside of modern reef experience (Kleypas et al. 1999; Hoegh-Guldberg 1999). Over merely three
456
decades we have witnessed increases is aboveaverage water temperatures that have caused unprecedented coral bleaching and mortality (Glynn 1993; Loya et al. 2001; Hoegh-Guldberg et al. 2007). There is also clear evidence that atmospheric carbon dioxide increases have caused shifts toward ocean acidification (IPCC 2007), which are recorded as reduced calcification in coral skeletons (Jury et al. 2010; Ries et al. 2010). One of the primary goals of contemporary coral reef ecology is to understand the dynamics of reef corals in regard to this global climate change, and determine which coral populations are destined to become the ‘winners’ and which populations are destined to become the ‘losers’? (Loya et al. 2001). The best means to address this question is through long-term assessments of permanent sites. Most reef studies and monitoring programs assess the state of the reefs by estimating the percentage coverage of major benthic organisms or by examining coral size–frequency distributions. Few studies examine the key ecological processes that drive the state variables. Processes of major interest include recruitment, coral growth, rates of partial colony mortality, fission, and mortality (Roth et al. 2010). These rates are dependent on macro-processes, such as predation and herbivory (Mumby et al. 2006). For example, large aggregations of the coral-eating Acanthaster planci will quickly change the composition of coral assemblages (Done 1987), and will essentially mask all other, more subtle, processes influencing the coral populations. Similarly, regional oceanographic thermal stress events may shift coral species composition within one season (Loya et al. 2001). Moreover, a recent study showed that sites dominated by high-frequency temperature variability over the past 300 years, will also experience greater thermal and irradiance stress during future regional events, but corals at these site are also more likely to undergo rapid directional selection compared with coral populations at lowfrequency-dominated sites (Thompson and van Woesik 2009). Therefore, when comparing population trajectories across regions and oceans, it is critical to understand the key processes within a regional and historical context. Here we outline a sampling strategy that captures both state and process variables, at a spatial
Environ Monit Assess (2011) 178:455–460
scale of tens of kilometers. Specifically, our main interests are (1) examining spatial and temporal patterns in coral population size–frequency distributions; (2) determining vital processes, including rates of recruitment, post-settlement mortality, growth, partial mortality, and total-colony mortality; and (3) examining relationships between processes and state variables and whether size distributions reflect population performance. The proposed sampling strategy will allow us to answer the following questions: Which processes are primarily responsible for population change? Which process best reflects the state of the reef? Which processes best reflect the change in, or stability of, size–frequency distributions? In locations where thermal stress occurs, the following questions can be addressed: Which processes are most sensitive to thermal stress? Is recruitment reduced after a thermal stress event, and if so for how many years? What role do coralcolony remnants play in recovery processes? Does differential population response to thermal stress vary among habitats, with different hydrodynamics? Do some habitats recover more rapidly than others? Can differential and local management practices influence thermal stress response and recovery?
Proposed design Our effective sampling units are randomly selected 75 × 25 m stations, spaced approximately 250–500 m apart, representing a 103 m spatial scale. Stations are nested within sites, which are spaced approximately 2 km apart, representing a 104 m spatial scale (Table 1). Three randomly selected 16 m2 quadrats placed in each station, and permanently marked for resampling, are used to assess processes across time (repeated measures design), while random belt-transects, which are rerandomized at each sampling event, are used to sample state variables. Both quadrats and belttransects are effectively sub-samples from which we will derive estimates of means for each station at each sampling event. The analysis of such data will be by a repeated measures design (RMD), which recognizes that locations are re-sampled over time, and thus data
Environ Monit Assess (2011) 178:455–460 Table 1 The spatial scales of sampling
Name
457 Within
Locations (top) Sites Locations Stations Sites Subsamples Stations 16 m2 quadrats Belt transects
Scale
Distance apart Allocation
Region 10 km 1 km 100 m
>100 km ca. 2 km 250–500 m 10s of meters
for the same location at different times are not independent (Crowder and Hand 1990; Green 1993; Green and Smith 1997). Two traditional statistical analysis approaches are commonly used in ecology. One is a multivariate analysis where the Y observations are visualized as distributed in a space of as many dimensions as there are times. In environmental studies there are rarely enough replicates relative to the number of times for this approach to be feasible, i.e., there are not enough (if any) error degrees of freedom. The other approach is the usual one in ecology— a univariate split plot design where Y observations at times are nested within treatments/groups. This approach has some assumptions about the error
Re-randomized
Arbitrary No Arbitrary (systematic) No Random No Random No Yes
distribution of observations over times, but (a) the assumption is less severe because the observations are independent, (b) the assumptions can be tested, and (c) even if the assumptions are not satisfied there are analysis and interpretation techniques to satisfy the assumptions. It is also possible for RMDs to be quite complex, for example hierarchical or several-way factorial, or blocked designs in which one of the dimensions is a repeated measure. Standard statistical packages will analyze RMDs (see Green 1993 for a worked example using Minitab). The SAS package has a REPEATED subcommand which somewhat automates the process. Linear mixed models (Littell et al. 2006),
Fig. 1 Study locations at the coral-reef targeted research capacity-building centers of excellence
458
generalized estimating equations (Diggle et al. 1994), or Bayesian hierarchical methods (Gelman et al. 2004) can also be used for these kinds of design. The specific approach taken and models used is dependent on the situation and philosophical approach. With a hierarchical mixed model design a Bayesian approach has certain advantages, especially if there are informative priors. If the data are normal, and there are no informative priors, a Bayesian analysis will produce the same answer as a traditional method, for example general linear mixed models. Our design is hierarchical, with Stations nested within Sites, which are in turn nested within Locations. At the Station level there is a two-tiered approach, with state variables sampled by randombelt transects and process variables sampled by permanent quadrats, which are re-sampled over time (repeated measures; Table 1). Stations are the basic sampling units in the design and they are on a ca. 1 km scale, 250–500 m apart. Stations (∼1,875 m2 ) are randomly allocated, then permanently fixed for re-location purposes. Subsamples are within Stations, but are not the basic sampling units. Herein, replicated, and re-randomized (at every time interval) belt transects assess state variables, and permanent quadrats assess process variables. These subsamples are on a 100 m scale, ca. tens of meters apart. Sites are arbitrarily (systematically) located on a 10 km scale, ca. 2 km apart (Table 1). Locations are the largest spatial level, arbitrarily located on a region scale that initially coincided with coral-reef centers of excellence (Fig. 1). All locations are permanent (i.e., not randomized each time) and arbitrarily defined (i.e., not randomly; Table 1).
Discussion This nested sampling strategy is allowing us to examine population dynamics, determine critical stages in demographic performance, and assess vital population rates across locations. For example in Zanzibar, this nested sampling strategy was employed to examine coral species diversity across three spatial scales (Zvuloni et al. 2010). Nonrandom outcomes of the partitioning analyses,
Environ Monit Assess (2011) 178:455–460
using individual-based null models, suggested that coral process studies may best focus among sites (i.e., tens of kilometers). One site supported (regionally) rare species and some unique taxa; this site was recommended as a marine protected area (Zvuloni et al. 2010). Still unknown is whether marine protected areas change any processes that have consequences to the state of reefs in the Indian Ocean. In the Caribbean, Mumby et al. (2006) argued that marine protected areas increase the density of herbivorous fishes, and these fishes cropped macroalgae that smothered coral recruits. Therefore, marine protected areas are suggested to have cascading effects on coral recruitment. The densities of major predators, such as A. planci and herbivorous urchins and fishes, are easily captured using urchin and fish counts within the rerandomized belt-transects. Such studies can be expanded, if necessary, to include caging/noncaging experiments that examine bioerosion rates through time (Hibino and van Woesik 2000). A necessary constraint in our design results from our knowledge that coral reefs differ with respect to depth. We have limited our sampling focus to shallow (2–5 m) sites to emphasize effort at a large (horizontal) spatial scale. This approach can be combined with an effective water-quality program, where funding is available. We add a note of caution because the spatial and temporal scales at which many water-quality parameters vary (in the water column) is poorly understood. Many programs do not capture the inherent scale at which nutrient parameters vary (Wagner et al. 2008). For example, near-substrate water temperatures show homogeneous patches at a 1 km scale, whereas ammonia (NH4 ), nitrate (NO3 ), total nitrogen (TN), total organic nitrogen (TON), and turbidity are more predictable at tens of meters (Wagner et al. 2008). Still, our multi-scale approach, using a randomized (re-randomized) design has the ability to measure the rates of the primary benthic processes, as well as measure the key state variables. Although, it is not feasible to measure all relevant processes on coral reefs, our design is a step in the direction of comparing the major processes using standard units. Such an approach will allow us to go beyond simply describing state variables through time.
Environ Monit Assess (2011) 178:455–460
A hierarchically nested sampling design seems most appropriate for coral-reef systems that are not only predicted to undergo substantial change over the coming decades but also because state and process variables need to be understood at a variety of scales. Most recently, the very reason for the global success of reef corals, which rely in large part on the photosynthetic capacity of their symbionts, has made them particularly vulnerable to slight water temperatures rises that are related to global climate change (Hoegh-Guldberg 1999). Particularly sensitive are the species rich habitats that have not been subjected to high selective pressures in the past (McClanahan et al. 2005; Thompson and van Woesik 2009). Unfavorable physical and chemical environments act as strong selective forces removing the most fragile species, and such processes may be scale invariant. Indeed, recent studies have shown that harsh and highly selective habitats support fewer coral species compared with benign habitats whose composition approaches regional species pool richness (van Woesik 2002). Yet, these species-rich, benign environments may be most vulnerable to climate change. Examining these key ecological processes, and understanding disturbances in context of background processes, will lead to accurate predictive models of population trajectories over time and determine what type of management intervention, and at what scale, will really facilitate reef resilience in the face of rapid climate change. Acknowledgements We would particularly like to thank Andy Hooten for bringing us together, and constant support and encouragement through that period. This program was funded by the Coral Reef Targeted Research (CRTR) Program, a partnership between the Global Environment Facility, the World Bank, The University of Queensland (Australia), the United States National Oceanic and Atmospheric Administration (NOAA).
References Crowder, M. J., & Hand, D. J. (1990). Analysis of repeated measures. New York: Chapman and Hall. Diggle, P. J., Liang, K. Y., & Zeger, S. L. (1994). Analysis of longitudinal data. Oxford: Clarendon Press.
459 Done, T. J. (1987). Simulation of the effects of Acanthaster planci on the population structure of massive corals in the genus Porites: Evidence of population resilience? Coral Reefs, 6, 75–90. Gelman, A., Carlin, J., Stern, H., & Rubin, D. (2004). Bayesian data analysis, 2nd Edn. London: Chapman & Hall. Glynn, P. W. (1993). Coral reef bleaching: Ecological perspectives. Coral Reefs, 12, 1–17. Green, R. H. (1993). Application of repeated measures designs in environmental impact and monitoring studies. Australian Journal of Ecology, 18, 81–98. Green, R. H., & Smith S. R. (1997). Sample program design and environmental impact assessment on coral reefs. In J. B. C. Jackson (Ed.), Proceedings of the international coral reefs symposium. Hibino, K., & van Woesik, R. (2000). Spatial differences and seasonal changes of net carbonate accumulation on some corals reefs of the Ryukyu Islands, Japan. Journal of Experimental Marine Biology and Ecology, 252(1), 1–14. Hoegh-Guldberg, O. (1999). Climate change, coral bleaching and the future of the world’s coral reefs. Marine and Freshwater Research, 50, 839–866. Hoegh-Guldberg, O., Mumby, P. J., Hooten, A. J., Steneck, R. S., Greenfield, P., Gomez, E., et al. (2007). Coral reefs under rapid climate change and ocean acidification. Science, 318, 1737–1742. IPCC (2007). Climate Change 2007: Synthesis report. In Core Writing Team, R. K. Pachauri, & A. Reisinger (Eds.), Contribution of working groups I, II and III to the fourth assessment report of the Intergovernmental Panel on Climate Change (104 pp.). Geneva, Switzerland: IPCC. Jury, C. P., Whitehead, R. F., & Smant, A. M. (2010). Effects of variations in carbonate chemistry on calcification rates of Madracis auretenra (=Madracis mirabilis sensu Wells, 1973): Bicarbonate concentrations best predict calcification rates. Global Change Biology, 16, 1632–1644. Kleypas, J. A., Buddemeier, R. W., Archer, D., Gattuso, J.-P., Langdon, C., & Opdyke, B. N. (1999). Geochemical consequences of increased atmospheric CO2 on coral reefs. Science, 284, 118–120. Littell, R. C., Milliken, G. A., Stroup, W. W., Wolfinger, R. D., & Schabenberger, O. (2006). SAS for mixed models, 2nd Edn. Cary: SAS Institute Inc. Loya, Y., Sakai, K., Yamazato, K., Nakano, Y., Sambali, H., & van Woesik, R. (2001). Coral bleaching: The winners and the losers. Ecology Letters, 4, 122–131. McClanahan, T. M., Maina, J., Moothien-Pillay, R., & Baker, A. C. (2005). Effects of geography, taxa, water flow, and temperature variation on coral bleaching intensity in Mauritius. Marine Ecology Progress Series, 298, 131–142. Mumby, P. J., Dahlgren, C. P., Harborne, A. R., Kappel, C. V., Micheli, F., Brumbaugh, D. R., et al. (2006). Fishing, trophic cascades, and the process of grazing on coral reefs. Science, 311, 98–101. Ries, J. B., Cohen, A. L., & McCorkle, D. C. (2010). A nonlinear calcification response to CO2 -induced ocean
460 acidification by the coral Oculina arbuscula. Coral Reefs, 29, 661–674. Roth, L., Koksal, S., & van Woesik, R. (2010). Effects of thermal stress on key processes driving coral population dynamics. Marine Ecology Progress Series, 411, 73–87. Thompson, D., & van Woesik, R. (2009). Corals escape bleaching in regions that recently and historically experienced frequent thermal stress. Proceedings of the Royal Society B, 276(1669), 2893–2901.
Environ Monit Assess (2011) 178:455–460 van Woesik, R. (2002). Processes regulating coral communities. Comments on Theoretical Biology, 7, 201–214. Wagner, D., Mielbrecht, E., van Woesik, R. (2008). Application of landscape ecology to spatio-temporal variance of water-quality parameters along the Florida Keys reef tract. Bulletin of Marine Science, 83(3), 553–569. Zvuloni, A., van Woesik, R., & Loya, Y. (2010). Diversity partitioning of stony corals across multiple spatial scales around Zanzibar Island, Tanzania. PLoS One, 5(3), e9941. doi:10.1371/journal.pone.0009941.
