Can the collapse of an ecosystem be foreseen?

A while back, I met theoretical ecologist Vishwesha Guttal of the Indian Institute of Science’s Centre for Ecological Sciences and discussed his work on early-warning signals for ecological changes. Guttal’s research is fascinating, and forms part of a growing body of work on tipping points. The upcoming issue of Theoretical Ecology, where Guttal’s latest paper appears, is devoted to this area of research.

Guttal’s work revolves around the concept of ‘catastrophic regime changes’ (CRC) in ecological systems. While the term may sound like the subject of a Hollywood disaster film about tsunamis and earthquakes, a catastrophic regime change need not be as dramatic.  It is simply a sudden change in an ecosystem from one state to another in a relatively short time (as opposed to a slow and gradual change). When a semi-arid region turns into a desert over a short period of time, an ecologist would call it a CRC.

Another example of a CRC is eutrophication---the reason why several of Bangalore’s lakes, such as Varthur and Bellandur are dying out.  When too many nutrients like nitrogen and phosphorous are pumped into lakes (through sewage, detergent, effluents etc.), algae multiplies,  the lake loses its transparency and aquatic plants under the surface are not clearly visible anymore. As oxygen levels drop because of excess algal growth, some important species of fish die away, while others begin to dominate.  

Ecologists say eutrophication can often be a *sudden* phenomenon---that is, even though the  nutrients are being pumped into the lake steadily over many years, the lake resists change and remains in its clear state. This goes on until a tipping point occurs. At this point, a mere incremental increase in nutrient input leads to a dramatic change in the lake, which clouds over in mere months.

The concept of CRC doesn’t apply to lakes and deserts alone.  Some researchers have speculated that the Indus Valley civilization could have come to an end due to changes in monsoon patterns, causing its rivers to dry up suddenly. A 2009 review paper in Nature by Martin Scheffer et al. draws dramatic parallels between epileptic fits, asthma attacks, stock market crashes and ecosystem collapses, describing these as systemic failures that are triggered by similar mechanisms.

But can scientists predict such failures? 

While research in this area is still nascent, scientists such as Guttal have suggested several statistical predictors for CRCs. In a 2008 paper published by Guttal and C Jayaprakash in Ecology Letters, they suggested that “changes in the asymmetry, quantified by changes in the skewness of time series data, can be a generic indicator of an impending regime shift.”

                                                                  

In the context of lake eutrophication, the time series data would be phosphorous levels at various points in the lake, collected over a period of time. As these levels fluctuate more and more against their mean (increased skewness), one can expect eutrophication to occur.

                                                                  

In the past, researchers would typically measure mean phosphorous levels in a lake and plot it over time. As the mean increased with time, they would take it as an indicator of an impending CRC.  This is a simple enough conclusion, since higher phosphorous levels lead to eutrophication.

What Guttal and Jayaprakash’s paper says, however, is that the mean phosphorous level alone cannot tell much. Instead of the average, it makes more sense to track phosphorous levels at various points in the lake and calculate how much they vary against the mean. Measuring the asymmetry of these fluctuations (skewness)  would indicate an upcoming tipping point.

Can these methods be used to predict monsoon failiures?

The statistical methods described above may someday be useful in predicting even monsoons. This is a long shot, but researchers are already working along these lines. Guttal points to a 2005 paper in Geophysical Research Letters by K Zickfield et al., which suggests that the Indian monsoon may have two stable states (a wet monsoon and a low precipitation monsoon). Further, changes in sulphur emmissions, land use, insolation and CO2 concentrations, driven by human activity, could trigger a transition from one stable state to another. If this is true, then the statistical observations by Guttal and other scientists working on tipping points could help predict monsoon failures by studying factors such as rainfall patterns.

Extending this further, the desertification of semi-arid areas could be predicted by examining vegetation patterns. Vegetation can grow in various patterns, such as spots and stripes. Guttal is working with the Department of Aerospace engineering on the new IISc campus near Chitradurga district of Karnataka, as well as in the deserts of Rajasthan, to deploy unmanned aerial vehicles to capture these patterns and tease out a connection between them and ecosystem changes.

As I said earlier in this post, this field of research is still nascent. To successfully apply these statistical predictors to real life would require a tremendous amount of data and several other variables. Guttal adds that in addition to data, a systematic methodology and toolkits are needed to help foresters and ecosystem managers  apply these methods. A 2012 paper by Dakos et al (including Guttal) in PLOS One makes an effort in this direction. This paper comes with a software package in programming language R, that accepts inputs of time series and quantifies early warnings.  

Meanwhile, policy makers are also looking at such statistical methods for predicting catastrophic risks, as this paper from the International Risk Governance Council shows.