By Graham Harris

This blog is continued from Thinking Systems #10

In the last three blogs we saw that management and restoration of the entities we call ecosystems is problematic and driven by assumptions, myths and values. This is particularly so at the level of species and populations. Conservation biology and restoration ecology have had their local successes, but overall, the response to major initiatives has been poor. Global biodiversity continues to decline and, while there are strong calls to restore landscapes, going back to a prior state seems very difficult.

While at some level physiological responses do control species occurrences, internal dynamics also control the precise responses and so ensure that these systems are not entirely computable or predictable. The uncertainty in the response to management action is fundamental. There may be some organising principles that determine the persistent states that we observe but, at present, we have little knowledge of these. As John Lawton concluded some years back, community ecology is a “mess” (if you are a naïve realist and rationalist that is).

The ecosystems we have to deal with exhibit distributed robustness in the face of uncertainty – and may therefore be unpredictable at many levels – but they are constrained by external forces. This leads to what Tim Allen and others have called “supply side sustainability”. While the details of outcomes defined by species may elude us we can, and do, manage the context within which life operates. These constraints can be manipulated to provide a broad measure of remediation and there are short-term success stories to share.

What has been called the “ecological methodology” has been characterised by a kind of “black box” approach: a naïve realist approach to the representation of everything from populations to biomass. It has always been the practice to sample these systems and to assume that the sampling distributions were Gaussian and representative: therefore means and variances were meaningful statistics. Differences in age structures, population characteristics and species identities were glossed over. This has long been a matter of contention.

Consider the provenance of Raymond Lindeman’s (1942) paper on trophic dynamics; one of the most influential early papers on energy flows and food chains in lakes. Lindeman’s paper made a generalized argument based on averaged data from one lake. On submission, it was rejected by the editors of the journal because of insufficient data. One of the original referees of Lindeman’s paper, Chauncey Juday, one of the foremost limnologists of his age, wrote in his review that “a large proportion of the following discussion and argument is based on “belief, probability, possibility, assumption and imaginary lakes” rather than actual observation and data”. Juday, and other limnologists at the time, thought that all lakes were different. Juday continued, “according to our experiences, lakes are “rank individualists” and are very stubborn about fitting into mathematical formulae and artificial schemes proposed by man…” (his emphasis).

This statement has an interesting historical sequel. In the 1960s Richard Vollenweider organized and brought together a large OECD project on the responses of algal biomass in northern temperate lakes to differing nutrient loads. The objective was to seek general principles that could be used to manage nuisance algal blooms in these lakes by reductions in their phosphorus loads. Vollenweider discovered that there was a level of statistical predictability if strong system level drivers were present – in this case the phosphorus (P) load – despite the acknowledged differences between lakes in species composition and individual temporal and spatial dynamics.

Richard Vollenweider told me that he was very much aware of the “my lake is different to yours” view of most limnologists up to the 1960s; so he was deliberately looking for larger scale generalisations. Ensembles of deep northern temperate lakes (he had to be selective to get this to work) showed statistical responses to P loads – but even so the graphs and correlations were derived from log-log plots of annually averaged data. Variances were high but definable. Lake managers seemed to be quite happy to have a defined probability of success – and a major USA/Canada program of phosphorus removal in the Laurentian Great Lakes cleaned up Lake Erie and other lakes. These statistical relationships have been widely used for the management of lakes around the world although many forgot that the original relationships were based on a particular subset of lakes. As usual; the Messiah was correct: just be careful of the disciples.

We have always known that individual lakes did not necessarily follow the overall trend and might even trend entirely the wrong way for a time (i.e. algal biomass might actually increase for a few years despite a reduction in the P load). The work of Brian Moss, Geoff Phillips and others on Barton Broad in the UK is a splendid example of how internal ecosystem dynamics can override the response to a reduction in P load for many years. As we might expect, there is much individual working out of “ecosystem” dynamics; much as I had observed in my work on the Lund tubes in Blelham Tarn. The effects of nutrient reduction on algal biomass may take time and the responses of individual lakes may differ for some time but, in the end, the effects of the major nutrient driver are felt.

So Vollenweider was correct; there are indeed broad scale generalisations that can be made and successfully used for some aspects of lake management. But Vollenweider had, fortuitously, chosen a rather special case. When nutrient loads to lakes are increased, nutrient concentrations in the water rise and larger plankton cells are selected for: this is the basic physiological response kicking in.  Reduce nutrient levels and the reverse occurs. Even when driven strongly by nutrient additions from the bottom up, and in the face of weakening systems level interactions like grazing pressure (larger cells get too big to be eaten), the ecosystem response is highly variable and requires log-log plots to control the variances.

Colin Reynolds’ predictions of species occurrences based on growth form and physiology do also allow broad generalisations to be made about the species composition of plankton in the lakes. But these rules also largely apply to Volenweider’s special case: ungrazed larger cells where physiology rules. So, contextual drivers can be used as a form of ecosystem management only if physiology is the predominant determinant of outcomes over internal 2^nd order dynamics. Management of the context of one trophic level may be possible for a while but in the end, as we shall see, in the longer term other system level interactions can come to dominate. And that may be problematic.

Management of the overall context is what Allen, Tainter and Hoekstra (1999) called “Supply side sustainability”. In the face of uncertainty (and if the internal drivers are weakened) we can manage the constraints on systems to achieve broad scale, statistically predictable outcomes; not perhaps some of the outputs we might like at the levels of individual species – but some system level control nonetheless – and this has been used for management purposes.

This insight has led to the idea of allowing ecosystem dynamics to occur and only intervening when thresholds are approached or crossed. Thresholds of Potential Concern have been widely used as a management tool in places like Kruger National Park in South Africa. Similarly a critical nutrient load was set for the management of Port Phillip Bay near Melbourne in Victoria – and this concept led to the successful expenditure of over $250M on storm and wastewater control measures. Once again, however, the critical nutrient load was set largely on the basis of the physiological responses of key species in the Bay because of the suspicion that exceeding the critical load would cause irrevocable changes to the Bay ecosystem.

There is now a huge ecological literature on tipping points – this idea has been much in vogue in recent years. Much of this work has been based on traditional (trivial) ecological models – and, like the Port Phillip Bay model, these models can show hysteresis or tipping points. Likewise, some shallow lakes show tipping points – switching between domination by aquatic flowering plants and domination by planktonic algal blooms. Once again the drivers here are external nutrient loads and the physiological responses of dominant functional groups.

Again there is an important caveat: some ecosystem tipping points have been predicted; others have been entirely unexpected (driven by unknown internal 2^nd order interactions). Hidden internal flows of information and meaning can produce surprising effects. In some ecosystems space/time patchiness and heterogeneity clearly stops the entire ecosystem tipping over in one go; change is restricted to localised regions at first – and if the entire system is eventually tipped into a new state, a much smoother and more gradual response is observed – leaving time for management intervention. Not all the observed outcomes follow the canonical example.

The current fascination with “re-wilding” landscapes and ecosystems – reintroducing large herbivores and predators – is another form of contextual management. This time from the top down. Knowing what we now know we should not be surprised that, like the Yellowstone Park example I mentioned earlier, the outcomes are likely to be unexpected and they will be played out over the long term. Similarly, the introduction of exotic and alien species to ecosystems – a common occurrence these days – should be expected to produce long term and unforeseen effects as the internal dynamics of the system play out. Of course, what we think about what eventuates is heavily laden with culture and values. We are in the habit of invoking the aesthetics of some previous golden age and also of assuming that those ecosystems were in some kind of equilibrium before human interference. Not so.

Perhaps a good place to close is to go back to Vollenweider’s statistical tools for lake management. In the 1970s there was much concern about the status of Lake Erie and the fact that increased nutrient loads had led to algal blooms and the depletion of oxygen in bottom waters. The international Great Lakes clean-up program was largely driven by the declining status of Lake Erie. Success in controlling algal blooms in open waters ensured that Lake Erie became one of the canonical examples of phosphorus loading control as a means to clean up large lakes. The situation 40 years later is much more complex. Yes, phosphorus control was successful – water quality has improved overall – but now there is concern over the rapid decline of pelagic fish stocks, the continued occurrence of algal blooms in coastal waters, and the probable major effect of introduced species – in this case Zebra Mussels.

What the Great Lakes have experienced since the 1970s has been a “Smeltdown”. Smelt – the once abundant small pelagic fish – have disappeared in recent years. The Smelt are themselves an introduced species – they were first introduced into the Lakes in 1912 as food for another transplanted fish, the Atlantic salmon. Peak catches of smelt occurred in the 1940s and the population began to plummet in the 1990s – when the nutrient loading to the lake was also declining. The Lakes are now cleaner than they were, but what we now see is a complex set of internal interactions between native and introduced species that is having a major impact on an important commercial fishery.

The expectation from the fisheries scientists was that the diversity of fish species would increase as the Lakes were cleaned up. What has actually been observed in Lake Erie over the last 40 odd years was that fish diversity reached a peak at moderate levels of phosphorus in the Lake, and then declined as the phosphorus was further reduced. Walleye, the dominant predator in the lake and one of the most important commercial species, were at their most abundant at moderate nutrient levels. The other important commercial species – Smallmouth Bass – has increased as nutrient levels have been reduced.

So, like the outcomes of other large scale ecosystem management experiments, what we have seen is the strong, complex and surprising interaction of reduction of the overall nutrient status of the lake with the system level effects of natural internal dynamics, commercial fisheries and introduced species. Short term success in management at one level often leads into longer term/larger scale complexity at other levels. Ecosystems are, as Juday, wrote; ““rank individualists” and are very stubborn about fitting into mathematical formulae and artificial schemes proposed by man…”. This is the usual conundrum that we face.