How many trees are there in Amazonia: two recent studies reached very different conclusions – UPDATED

The region of South America that we know as “Amazonia” has arguably the greatest biological diversity of any part of the planet, certainly as far as plants are concerned.  In some places the number of tree species per hectare exceeds 400, an order of magnitude greater than the number for the whole of the British Isles.  However estimating the total number of even the described plant species in this vast area has proven controversial, as two recent studies exemplify.  The first study was by ter Steege et al. (2016) and entitled “The discovery of the Amazonian tree flora with an updated checklist of all known tree taxa“, whilst the second is from just last month: Cardoso et al. (2017) “Amazon plant diversity revealed by a taxonomically verified species list“.  Both of them are open access so click on the links if you want to read the full studies.

One might expect that two such studies focused on Amazonia, both using vouchered herbarium records, would reach broadly similar conclusions as to the number of tree species in the region.  Not a bit of it: ter Steege et al. (2016) report 11,676 species, whilst Cardoso et al. (2017) say that the figure is 6,727.  That’s almost a two-fold difference!  Why the discrepancy?  Inspired by an initial tweet by University of Glasgow taxonomist Roderic Page, I downloaded the data from both studies and looked at it closely.

Here’s a scatter plot of the number of tree species per plant family reported by both studies:

Amazon tree diversity


The red line shows where we would expect the data points to lie if both studies had reported the same number of tree species per family.  Clearly few families lie on this line and most are above it as we might expect: as I’ve said, ter Steege et al. (2016) concluded that there were far more tree species overall and this is reflected at the family level.  Note that I’ve graphed this using a log scale and what might seem to be small differences are actually very large indeed.

Although the findings from two studies are highly correlated (diverse families are diverse in both studies, ditto families with low diversity) the actual level of that species richness is very different.  For example, in the Annonaceae, ter Steege et al. report  480 species, Cardoso et al. report 388; in the Clusiaceae the figures are 247 versus 135.  Other families are excluded from one data set or the other: ter Steege et al. reckon there 7 species of trees in the Dilleniaceae whereas Cardoso et al. cite zero.  Here’s a link to the data set if you want to explore further.  

So what’s going on here?  Why do two studies with similar aims, published about 12 months apart, come to such different conclusions.  As far as I can see there are three reasons for this.

First of all, the studies used slightly different taxonomies when it came to considering families and species.  So for example, Cardoso et al. recognise the family Peraceae which ter Steege et al. do not.  Although I haven’t done it, I’m sure that if one were to dig down to the species level there would be differences in which species were accepted and which were considered synonyms.

Secondly, the exact definition of what constitutes a “tree” varies between botanists, and the non-botanists who are no doubt responsible for some of the plant collections: some consider anything to be woody and tall-ish to be a “tree”, others have more strict definitions.  Notes about growth form taken in the field consequently get included in herbarium databases and may be inaccurate, especially for the uncommon species that have rarely been seen in the field.

The final reason, and the one that seems to be responsible for most of the discrepancy, is the definition of what constitutes “Amazonia”.  In the first study ter Steege et al. defined it as including the “forests and savannahs of the Amazon basin and Guiana Shield”.  In contrast Cardoso et al. considered only “lowland Amazon rain forests”.  That’s a big difference as there’s lot of savannah in this region, as well as other habitat types.  When we did field work in Guyana some years ago we could travel very quickly between savannah and rainforest.  It was clear to us that there is a range of trees that are restricted to one habitat or another, including species of Dilleniaceae (mentioned above) that are savannah specialists (hence the family’s exclusion from the Cardoso et al. study).

Now neither of these studies is “wrong” in the sense of being inaccurate or misguided: both are great studies involving a huge effort on the part of the authors.  But the limitations and definitions of geography and taxonomy that I’ve highlighted do mean that they need to be treated as rather different and not directly comparable.

So how many tree species are there in Amazonia?  If we consider just the rainforest then it’s 6,727 (Cardoso et al. 2017).  If we consider all habitats in the region, including rainforest plus savannah etc., then the figure is 11,676 species (ter Steege et al. 2016).  One of the implications of this is that the non-rainforest “Amazonian” habitats collectively contain 4949 tree species.  Thus a large proportion of the diversity of the region is in habitats, such as savannah, which are less of a focus for conservation efforts and not as well known to the general public, but are at least as threatened by agriculture and mining as rainforest.

Thanks to Roderic Page for initially highlighting this on Twitter, and Sandy Knapp for discussion.

UPDATE:  In retrospect my conclusion above regarding the proportion of trees in non-lowland rainforest habitats was much too high, as a couple of commenters have noted below.  It’s worth reading what they have to say, and my responses.  It’s likely that the taxonomic differences between the two studies are at least as great as the geographical ones, but then taxonomic opinions vary hugely.  Just serves to emphasise what a controversial and problematic question this is!



7 thoughts on “How many trees are there in Amazonia: two recent studies reached very different conclusions – UPDATED

  1. Alessandro Rapini

    Hi Jeff,

    I am glad that you have discussed these results and are apparently happy with a compromise between the two counts. Let us follow your logic and believe that the difference between them is simply result of different concepts of Amazonia and therefore of their universe. Then, according to your reasoning, the non-rainforest “Amazonian” habitats, which represent approximately 10% of the Amazonia (you may check their small representation in Fig. 2) and where trees are not the main or the most diverse component, house 4949 tree species, being therefore responsible for the difference between the two counts. Like that, more than 42% of the TREE species in the Amazonia are exclusively found in NON-FOREST habitats and only 58% of them in the rainforest habitats. Is that right?!

    In case you have missed, besides the list of seed plants, Cardoso et al. (2017) also indicated 3,794 corrections in the ter Steege´s et al. (2016) checklist, based on 80% of their list, corresponding to the families they checked. This represents 40% of wrong names. An extrapolation of these corrections to the whole list would reduce their number of tree species to approximately 7000, a value closer to the 6727 reached by taxonomists.

    1. jeffollerton Post author

      Hi Alessandro – thanks for the comments. Looking again at the maps in the two papers it’s clear that the ter Steege et al. study also included regions further south that Cardoso et al. did not, a lot of which is forested but not defined as “lowland Amazon rainforest” by Cardoso et al. So I think that the non-lowland rainforest habitats, collectively, are going to be greater than 10% – perhaps closer to 20%? Plus they include some montane rainforest in the north, and cover a large geographical range. But I agree, the 42% of species outside lowland rainforest is going to be too and I will add an update to the post to that effect.

      With regard to the taxonomy, I chose not to get too involved with that as it’s such a minefield and opinions (as you know) vary hugely 🙂

      1. Alessandro Rapini

        I see… I agree that taxonomy is hugely affected by opinions; other fields are hugely affected by oversimplifications. Yet, we may have good and bad opinions and, in many case, taxonomic differences are not only a simple question of opinions. I think that a minimally rigorous researcher should not use and any “accurate” study should be based on a (taxonomically blind) checklist of terrestrial mammals of Amazonia that includes, for instance, leopard (Panthera pardus) and boto (Inia geoffrensis). But this is only an opinion (not a taxonomic one, I think). Well, I don´t want to push you to the minefield; I am happy that you thought about these studies and would suggest another one that I saw today and that also question the taxonomy and the implications of it in a “big data” study: “A matter of weight: Critical comments on the basic data analysed by Maestri et al. (2016) in Journal of Biogeography, 43, 1192–1202” DOI: 10.1111/jbi.13098.


  2. CassieFreund (@CassieFreund)

    How many of the 4949 species do you think are different due to the different taxonomies vs. grow in the savannah? Your conclusion that the non-rainforest Amazonian habitats have ~5000 species seems oversimplified given your previous argument about mis-naming, synonyms, etc.

    1. jeffollerton Post author

      Thanks for the comment Cassie. Yes, I agree, it is rather over-simplified in retrospect and I am going to add an update. See my reply below in response to Alessandro’s comments.


Leave a Reply

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

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

Google photo

You are commenting using your Google 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 )

Connecting to %s