How foreign is the past?

The past is a foreign country: they do things differently there.
L. P. Hartley

Wooly mammoths. Saber-toothed cats. Giant sloths. You don’t need to think very long before some differences between modern ecology and the ecology of the past become apparent. Altered nutrient cycles. Introduced species. Altered soundscapes.

One of the important rationales for studying palaeoecology is that knowledge of past ecosystems can help understand modern ecosystems and make predictions about future ecosystems on our warming planet. This skill might be limited if, as Lyons et al suggested in Nature last week, the rules that govern how plant and animal communities are structured have been altered by the pervasive influence of humans on Earth’s ecosystems.

Lyons et al compile many datasets of species presence and absence for modern and fossil data over the last 300 million years and test whether pairs of species tend to aggregate together, segregate or co-occur at random.

In the modern datasets, most species pairs randomly co-occur. Species with non-random patterns tend to segregate. Lyons et al find the converse pattern in fossil data: pairs of species tend to aggregate. The switch from aggregation to segregation occurred about 6000 years ago, coincident with the expansion of agriculture.

It would be interesting to consider how the relatively limited impact of mid-Holocene agriculture could have such profound consequences. Could the processes, for example habitation fragmentation, suggested by Lyons et al change the rules of species assembly.

We could test the results of Lyons et al with alternative data, for example the European pollen database which is analogous to the North American pollen database used by Lyons et al.

But first we need, as Lyons et al do, to consider the possibility that the results are an artefact of some property of the data. Lyons et al first show that the result is not dependent on the modern data: a decline in aggregated species pairs is seen even if only the fossil data are analysed.

Loess curve weighted by number of sites with shaded 95% confidence intervals illustrates the reduction in the proportion of aggregated species pairs towards the present. Data are analysed with (black line and shading) and without (red line and shading) the modern data. Colours indicate continent: North America (green), Eurasia (purple), Australia (dark grey), South America (dark blue), Africa (orange). Point shapes indicate type of data: pollen (square), mammals (triangle), macroplants (circle). Data on terrestrial communities from ref. 2 are diamonds. Only mainland assemblages were included in the calculation for the weighted Loess curve and the density plots here and in Fig. 1.

Loess curve weighted by number of sites with shaded 95% confidence intervals illustrates the reduction in the proportion of aggregated species pairs towards the present. Data are analysed with (black line and shading) and without (red line and shading) the modern data. Colours indicate continent: North America (green), Eurasia (purple), Australia (dark grey), South America (dark blue), Africa (orange). Point shapes indicate type of data: pollen (square), mammals (triangle), macroplants (circle). Data on terrestrial communities from ref. 2 are diamonds. Only mainland assemblages were included in the calculation for the weighted Loess curve and the density plots here and in Fig. 1.

The Holocene decline in the proportion of aggregated species pairs is driven by a small number of 20th century “fossil” data sets. Quote marks on “fossil” as delving into the extensive supplementary material shows that only one of the 2oth century data sets actually comprises fossils. Regardless, I’m not convinced this decline is particularly robust.

Lyons et al test whether there is a relationship between the proportion of aggregated pairs and the spatial or temporal grain and extent of the datasets. They find no significant patterns, but the relationship with spatial extent is suggestive (p = 0.066). Lyons et al only test for these patterns in the fossil data sets, so the modern-fossil contrast is missing. The obvious thing to do is to add the modern data to this analysis, but that would take a lot of digging.

Lyons et al is interesting and provocative, but I am inclined to think that the result is probably an artefact. This paper will be the subject of our next palaeoecology journal club meeting; I’ll update this post afterwards.

Advertisements

About richard telford

Ecologist with interests in quantitative methods and palaeoenvironments
This entry was posted in Peer reviewed literature and tagged . Bookmark the permalink.

5 Responses to How foreign is the past?

  1. Are humans included in the taxon pairs? From the freely available material it would seem we’re excluded, but it would be interesting to see how many pairs to which we belong. For many animals domestication has probably ensured their survival.

    I remember when underwater noise first became a growing ecological concern and now that we realize many mammals can see in UV our generation of UV light on power lines helps explain some of the habitat fragmentation we observe. It will be interesting to see if there are other human activities that we never considered to be detrimental that wind up having significant effects.

    • I think the data are from (semi-) natural systems. Including strongly modified systems in the analysis could change the results.
      Humans have immense impacts on natural systems, but I’m not yet convinced the impact was immense at 6000BP.

  2. Andrew Millard says:

    I can’t see how they justify their conclusion of Holocene timing for the shift. The error bars in Extended Data Figure 2 show a 95% CI for the breakpoint of 10^1.606 to 10^5.951 years i.e. 40 years to 900,000 years. While this clearly “encompasses the beginning of agriculture in North America around 8,000 years ago and the increase in human populations during the Holocene” it also encompasses the Second World War and the appearance of Neanderthals. The best argument they manage is “it is difficult to pinpoint the exact time of the shift, but a closer examination of the data suggests that placing it within the Holocene is reasonable”, followed by a discussion of average values which ignores the uncertainty in the averages and the uncertainty in where to put the breakpoint and thus gives no reason for placing it in the Holocene.

    I’m also worried that I can’t see any justification given for analysing the breakpoint on a logarithmic scale which weights data from the current interglacial equally with the data for the preceding 100 million years.

    The analysis also ignores the clear structure in their ‘proportion of significant pairs that are
    aggregated’ data which follows some zero-inflated distribution and is far from the normal distribution assumed by the breakpoint analysis in Extended Data Figure 2 and the loess fitted in Extended Data Figure 3.

    • Yes, the non-Gaussian data are a problem. Some of the datasets have only a single non-random species pair. I would have been inclined to drop these, or use a different distribution, but that would certainly make the analysis more complicated (but probably also more robust.

  3. Pingback: Denying the “undeniable reality” | Musings on Quantitative Palaeoecology

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s