The Good and the Bad of the Hopkins Studies

The studies at HMS illustrate both the value and the challenges of long-term, historically based research in identifying biological responses to climate change. Perhaps the most important conclusion that can be drawn from these studies is that long-term monitoring programs are essential in attempting to reject alternative hypotheses to climate change as a driver of species or community changes. The dramatic range-related shift in species' abundances we observed, as well as the vertical shift in algal zonation, would not have appeared in a short-term study. Long-term monitoring of both biotic and climatic variables is a straightforward method of tracking changes that may be too subtle to be observed in the course of a single investigator's career. The National Science Foundation's Long-Term Ecological Research (LTER) program is an excellent step in this direction. Unfortunately, the program's coverage of nearshore coastal areas is still lacking. Out of 24 LTER sites, only six are coastal, and only one includes rocky shores, which are sites of high diversity and vulnerability to climate change. Ironically, long-term monitoring programs such as CalCOFI in southern California (used, for example, by Roemmich and McGowan 1995) and several programs in the United Kingdom (used by Southward et al. 1995) have suffered drastic scale reductions and budget cutbacks even as they contributed essential data to convincing studies of biological responses to climate change. It has been estimated that 40% of the long-term marine monitoring schemes in Europe were terminated in the late 1980s (Duarte et al. 1992), precluding study of some of the warmest years on record. These and other problems faced in the collection and use of long-term monitoring data sets are discussed in detail by Gross and Pake (1995) in a report for the Ecological Society of America.

An additional point illustrated by the HMS studies is that scientific reserves can play a vital role in long-term studies and the conclusions drawn from them. The designation of the Hopkins Marine Life Refuge in 1931 not only ensured that the study sites were protected from human disturbance, it also gave us a higher degree of certainty that the biological changes we observed were not due to the direct destructive practices of humans, such as foraging, collecting, or trampling.

From a sampling standpoint, the studies benefited from being replicated in the exact location of the original studies. Spatial variation in invertebrate abundance in the Hopkins intertidal zone is extreme, so that transects through the same intertidal zone just a few meters apart can show completely different patterns (R. Sagarin, pers. obs.). Likewise, algal tidal height distributions are extremely sensitive to aspect of the rocks. By sampling Endocladia height on different sides of rock faces at HMS (unpublished data), we found that height of this alga is consistently lower on south-facing rock slopes than on north-facing slopes. Thus, without knowing that Endocladia plots were surveyed in the precise location of Glynn's original plots, it would be difficult to rule out the effects of substratum aspect in the tidal height changes we observed.

Nonetheless, resampling in the same location raises its own problems in the lack of independence between repeated-measures of the same quadrat. One solution to this problem is to adjust the alpha value of statistical comparisons (such as a paired i-test) using the Bonferoni or Dunn-Sidak corrections (Sokal and Rohlf 1995). Unfortunately, these corrections severely impair the power to detect changes. For example, resampling a plot twice requires finding significance at the 0.025 level, based on the Dunn-Sidak correction, and additional re-surveys lower the alpha value further, making this impractical for data that would be collected many times from the same location. As an alternative, multivariate methods such as the paired T2 test (Morrison 1990) may be used, but they are computationally more complex. Other multivariate methods for dealing with repeated measures focus on response curves. For example, Scott (1993) proposed a time segment analysis for repeated measures of permanent quadrats which uses a combination of the mean of measured variable (e.g., percent cover of vegetation in a quadrat) and slope, or rate of change, between observations of the quadrat to generate curves showing the time trend in the data. Gurevitch and Chester (1986) and Potvin et al. (1990) discuss the benefits and limitations of such approaches and suggest methods for dealing with the problem of repeated measures. Lesica and Steele (1996) show through example plant populations that these methods are practical for long-term monitoring in the context of climate change studies. Thus, the benefits of permanent quadrats discussed above outweigh the statistical complexities of dealing with repeated measures data.

Our study of Hewatt's transect also illustrates the value of having quantitative data, rather than categorical data. Many ecological surveys rely on presence/absence data or categorical data (e.g., rare, common, abundant), making it difficult to determine any but the most extreme population changes. This point is illustrated dramatically by viewing our data from Hewatt's transect as if Hewatt had provided data only on the presence or absence of invertebrate species. In this analysis, the percentage of local deletions of species (found in Hewatt's study, but not in our re-surveys) is nearly identical across geographic range categories, whereas the number of local additions (found in our re-surveys, but not in Hewatt's survey) is only slightly greater for southern species (Fig. 3.14A). In other words, no range-related pattern of change could be found in this study in the absence of quantitative data. This surprising result is largely due to the effect of rare species on presence/absence changes in this comparative study. When sampling along a single transect in two separate points in time, rare species may easily be counted (or missed) by chance in one study but not the other. After removing the rare species (defined as those species with fewer than 10 individuals counted in both studies) from the presence/absence analysis, some range-related pattern of change can be observed (Fig. 3.14B), with a high percentage of southern species additions and no northern species additions. It should be stressed though, that the emergent pattern is not as clear or convincing as that shown with quantitative data (see Fig. 3.4), and furthermore, the elimination of w 'o

A. All species

80 60

southern northern cosmopolitan B. Non-rare species only all

CL 20

southern northern cosmopolitan all



Both studies

Figure 3.14. Changes in presence/absence of all species recorded by either Hewatt (1934) or Sagarin et al. (1999) in the 57 plots resurveyed between 1993 and 1995. "Additions" are species recorded by Sagarin et al., but not Hewatt. "Deletions" are species recorded by Hewatt, but not Sagarin et al. "Both studies" are species recorded by both Hewatt and Sagarin et al. (B) Changes in presence/absence of non-rare species only (defined as those with greater than 10 individuals recorded by either Hewatt or Sagarin et al.).

rare species was only possible because we had data on their abundances.

Despite the fortunate contributions of well-marked study sites and quantitative data from the original authors, the Hopkins studies' data were limiting for their lack of biological information beyond simple numbers and locations of individuals. Many sessile invertebrates are difficult to count on an individual basis. Sponges and most tunicates were ignored by both us and Hewatt for this reason. Information on their percent coverage, if taken by Hewatt, would have helped us determine whether changes had taken place in these taxa. Further, some biological responses to climate change may be more subtle, and not necessarily reflected in population density changes. For example, Tegner et al. (1996) found that number of stipes per plant, rather than traditional measures of whole plant survival, was a more sensitive indicator of kelp forest response to temperature shifts in southern California.

Measures of population and community changes beyond individual counts will become necessary as model predictions of species' responses to climate change become more sophisticated. For example, there is some evidence that climate change may intensify upwelling off the California coast (Bakun 1990). Indeed, recent analysis shows that upwelling increased along the coastline surrounding Monterey Bay during the period from 1946 to 1990 (Schwing and Mendelssohn 1997). Intertidal sites adjacent to upwelling zones in South Africa and Chile have shown enhanced algal productivity and increased herbivore biomass relative to non-upwelling sites, presumably due to nutrient enrichment during some upwelling periods (Bosman et al. 1987). This suggests the hypothesis that increased upwelling should have led to an increase in biomass of intertidal herbivores. Lacking historical information on animal biomass at HMS, we had no way of testing this hypothesis.

In a wider analysis, the Hopkins studies, like most studies of biological responses to climate change to date, are correlational in nature. They have drawn their strength from how well the relationship between observed biological changes and climate changes has matched predictions of these changes. Often the predictions themselves are quite simplistic and fail to take into account species interactions, different rates of migration between species or within a species range, and other "ecological noise" that will confound interpretation of range shifts. For example, Davis et al. (1998) used microcosm experiments on insects with various preferred temperature ranges to show that species interactions during periods of simulated warming result in range shifts that are different than expected from predictions that only take into account a species' preferred temperature range. As another example, Zimmerman et al. (1996) found that a southern limpet which had recently invaded Monterey Bay was responsible for a catastrophic decline in subtidal eelgrass beds. Although the limpet's northward expansion is consistent with the simple expectation from climate warming, the decline of the eelgrass is not, as seagrasses might be expected to show enhanced photosynthetic rates and growth in an era of increased atmospheric CO2 (Beardall et al. 1998).

Nevertheless, for the present, these correlational studies are a welcome addition to our small but growing body of data regarding biological responses to climate change. It is clear, nonetheless, that such studies would benefit from improved resolution in three important areas:

1. Temporal-Fine resolution in this area is essential to answer the question, How are both the biological and climatological signals changing through time? This calls for time series data, which are often available for climatic factors, but less common for biological variables. Notable exceptions are the work of Southward et al. (1995) and of Roemmich and McGowan (1995) (see earlier discussion). The advantages of such data are well-summarized by McGowan et al. (1996) who point out that time series data allow researchers to identify important frequencies of biological changes and to examine coherence between biological and physical events.

The obvious weakness of our "snapshot" historical data in terms of temporal coverage was pointed out by Denny and Paine (1998) who found that the 18.6-year oscillation in the moon's orbital inclination (preceding discussion) may be correlated with nearshore sea temperatures. Because we lack regular time series data on the biological changes at HMS it is difficult to know if the species changes we observed are more strongly related to the gradual rise in near-shore sea temperatures over 60 years, the shorter-term sea temperature changes associated with the lunar oscillation, or other changes that have occurred in the last 60 years (see Table 3.1).

2. Spatial-Strong spatial resolution is important in answering the question, Are the changes occurring throughout the species' range? This is critical in separating global changes from local perturbations. Parmesan's (1996) extensive study of Edith's checkerspot butterfly population spanned the entire North American range of the species and thus was able to convincingly show that populations were more likely to have gone extinct in the southern part of the range. In our study, by contrast, it was impossible without additional sources of information to separate abundance changes of species from actual shifts in their ranges because our data were gathered at a single site.

3. Taxonomic-Strong taxonomic resolution answers the question, Are these changes happening to everything? As our study of Hewatt's transect illustrates, this is an excellent tool for eliminating alternate hypotheses. A pattern (e.g., range-related abundance changes) is more likely to be a general biological response to climatic changes if it is consistent across taxonomic groups or life-history strategies. As a counterexample, if we found that only filter feeders (e.g., barnacles, mussels, etc.) had changed in abundance, we would suspect a more specific agent of change than a general temperature increase.

It is unlikely that any one study will adequately address these three areas. Thus investigators should be aware of weaknesses in these areas as they interpret their results. More importantly, investigators from different fields and different locations should compare results frequently, amalgamating their findings in order to overcome deficiencies in the spatial, temporal, or taxonomic resolution of any one study.

0 0

Post a comment