Abundance and Abundance Change in The World's Parrots
Abundance and Abundance Change in The World's Parrots
Abundance and Abundance Change in The World's Parrots
Accepted Article
Accepted Date : 27-Dec-2014
parrots
Division of Biology & Conservation Ecology, School of Science & the Environment, Manchester
* Corresponding author.
Email: s.marsden@mmu.ac.uk
This article has been accepted for publication and undergone full peer review but has not
been through the copyediting, typesetting, pagination and proofreading process, which may
lead to differences between this version and the Version of Record. Please cite this article as
doi: 10.1111/ibi.12236
This article is protected by copyright. All rights reserved.
Estimates of population density and abundance change (differences in density or
Accepted Article
encounter rates across land uses or time periods) form the cornerstone of much of our
10 000 bird species and monitoring trends in the light of rapidly changing
review literature on population densities and abundance changes across habitats in one
of the world’s largest and most threatened bird families, the parrots (Psittaciformes), in
abundance and seek patterns that might guide thinking for data-deficient taxa and
situations. Density estimates were found for only 25% of 355 parrot species.
Abundance change data were similarly limited and most came from logged forest, with
very few comparisons across different anthropogenic habitats. Threatened species were
no more likely to have a density estimate than non-threatened species, and were less
densities are most influenced by genus, and are generally higher within protected areas
than outside. It is unclear whether the latter effect stems from habitat protection, a
reduction in poaching, or both, but protected areas appear to be beneficial for parrots.
Individual members of the ‘parakeet’ genera (e.g. Pyrrhura and Eos) were predictably
abundant, while within larger-bodied genera such as Ara (macaws), species were
predictably uncommon (< 10 individuals per km2), and there was a long tail of extreme
rarity. Responses of parrots to habitat change were highly variable, with natural
variation in parrot abundance across different primary forests as great as that between
primary forest and human-altered forests. The speed at which environmental change is
affecting the world’s parrots far outstrips that of our current capacity to track their
abundance and we assess the likely scale of data deficiency in this and other bird
sampling and the integration of niche modelling with point density estimation may
increase our effectiveness in monitoring parrots and other important and threatened
bird groups.
Measures of population size, density and abundance change underpin much of our
understanding of the extinction risk that taxa or populations face (e.g. Collen et al.
2011), and contribute to the criteria on which the IUCN Red List extinction risk
classification system are based (Mace et al. 2008). While the quantitative data we are
able to gather on animal and plant abundance or abundance change are not without
problems (e.g. Wilson et al. 2011), they are often comparable across studies, species and
years by having an indication of precision (e.g. Bibby et al. 2000) and being open to
statistical testing (e.g. Maxwell & Jennings 2005), thus allowing more of the IUCN Red
be difficult and costly to obtain for rare species of great conservation interest,
especially in remote or conflict-ridden regions (e.g. Buckland et al. 2008, Rovero &
mark-recapture/camera trapping methods (e.g. MacKenzie et al. 2006, Royle et al. 2009,
Thomas et al. 2010). Occupancy estimation, while not in itself yielding abundance
estimates, may require less effort than abundance estimation and, as such, can be
are more often done latitudinally (e.g. Gray et al. 2006), and thus involve an
assumption that treatment and control sites were identical before the land use change
occurred.
We review knowledge of population densities of one of the world’s largest but most
threatened bird families – the parrots - and assess how habitat and land use change
affect their abundance. First, we review the available data and identify where there are
important gaps in knowledge. Then we look for patterns in abundance and abundance
species, based on information from those species for which there are data. Finally, we
compare our knowledge of parrots with that for other bird groups and assess prospects
for generating sufficient data for parrots and other birds to allow data availability to
Parrots are a group of approximately 350 mainly tropical and sub-tropical species,
occurring across much of the world, but with centres of high species richness in South
Eos semilarvata) and a few species with unusual partial associations with grasslands
(e.g. Eastern Ground Parrot Pezoporus wallicus) or deserts (e.g. Princess Parrot Polytelis
Parrots are well known to the general public as pets, exhibits in zoo collections,
and as flagships in conservation efforts (Verissimo et al. 2011), and they have a high
media profile. Some aspects of their behaviour are very well studied (e.g. Pepperberg
1999). Many parrot species, however, have not been the subject of dedicated ecological
study (Collar 1998, Masello & Quillfeldt 2002), and this lack of knowledge is a
compared to other bird families (Bennett & Owens 1997). Forty-two per cent of the
largely due to habitat alteration and loss, and excessive capture for the pet trade
(Collar & Juniper 1992, Wright et al. 2001). Parrots have long been traded, and largely
uncontrolled capture for both the international bird trade (Pires 2012) and growing
domestic markets (Gastañaga et al. 2011) is a great cause for concern. Capture pressure
is, of course, uneven across years, taxa and geographical areas. Heavily traded groups
Amazona spp (15 892 ‘wild-originated’ individuals for the five year period 2008-2012),
Ara spp (34 410 individuals), and Cacatua spp (15 694, most from Indonesia). Such
apps.org/citestrade).
METHODS
We searched for abundance data on all 356 extant parrot species currently recognised
pristine habitats and one or more anthropogenic habitats/land uses, between two or
more pristine habitats (in different areas or across natural habitat types), or between
anthropogenic habitats. These data, which came from latitudinal, or, more rarely,
longitudinal studies (see above), formed the basis of our inferences about the responses
of parrot species to land use change. There is debate about the extent of pristine forests
(e.g. Willis et al. 2004) but here we consider the most pristine habitats available as being
using combinations of key phrases (density, parrot, abundance, habitat change, land
use) and on the scientific name of every parrot genus and species. A few older studies
12 Nov 2002). We excluded density estimates from this source if they were based on
were derived from distance sampling, for which field methods both for points and line
transects have been largely standardised since methodological work in the early 1990s
(Jones et al. 1995, Marsden 1999). We located unpublished PhD theses and unpublished
reports by searching the reference lists of relevant published papers, by searching the
reference given in support of information on the species account for each parrot species
around 2000 references held within the extensive parrot literature library of the
We did not include papers that compared parrot abundance between seasons but not
habitats (e.g. Renton 2002). We included studies that presented a single density
estimate, but not those that presented a single encounter rate from just one site/habitat
(e.g. Smyth et al. 2002). We sometimes had a choice between density estimates given in
a publication and, in such cases, we chose the estimate based on the largest sample
size. Studies of parrot populations outside their natural range (e.g. Rose-ringed
were excluded.
The IUCN threat status of each species was recorded from BirdLife
International’s Data Zone, along with the region where the bulk of its natural range
currently occurs: Africa (AF) including Madagascar and Mauritius; Asia (AS);
Australasia (AU) including New Guinea, Australia, New Zealand, Fiji and the Pacific
islands; South America (SA) and Central America and the Caribbean (CA). The last two
regions were kept separate because, unlike SA, CA contains a number of small island
species, which are relatively well studied. The regions are also administered separately
within the Parrot Researchers Group. A few species ranged across more than one
region (mostly South and Central America, but also Australasia and Asia). These
species were assigned to the region in which the bulk of their range falls (from Juniper
& Parr 1998). Inevitably, there would be benefits to having a finer resolution to our
geographical categories, but small sample sizes (see later) meant that we were obliged
publication; survey method used and whether the estimate was a density estimate or
an encounter rate. Survey methods were classified as line transect distance sampling
(DLT); point count distance sampling (DPC); line transect encounter rate (ERLT); point
Nestor notabilis and Eastern Ground Parrot) were excluded from analyses except those
each study were assigned to one of the following habitat categories according to
information given in those studies: primary (PR) including natural and little-disturbed
forest, woodland or savannah; selectively logged (SL) forests that had been recently
totally cleared land, or regrown in areas heavily altered in structure and floristics due
to small-scale logging and other forest uses; agricultural systems (AGR) including
small- and medium-scale systems (Scales & Marsden 2008); and plantation (PLAN)
including any scale of monocultural arboreal cultivation (e.g. Eucalyptus, coconut Cocos
nucifera, pine Pinus etc). Forest fragments (FRAG) were included if parrot abundance in
them was compared explicitly with abundance in non-fragmented forest, or with other
anthropogenic habitats.
Proportions of species with and without population density estimates were tested
across regions using a G-test, and across IUCN threat status (Threatened and Near-
selectively logged forests) were tested with Wilcoxon signed ranks tests as they were
encounter rates not considering the direction of change) between primary forest and
each of the other habitat types (e.g. primary versus secondary tested against primary
versus agriculture) were examined using Mann-Whitney U-tests as they were not
paired by study.
habitat. We regard this analysis as exploratory for several reasons (Zuur et al. 2010). We
had relatively few data points for the same species, and data were sparse compared
whether or not the site was protected. Survey method was coded as one of the
following: DLT; DPC; and OTHMETH (other methods) which included spot mapping,
roost counts, long watches and other miscellaneous methods. Genera were included if
there were > 5 species cases within them - if not, they were included as OTHGEN
(other genera). A few closely-related genera were merged to increase sample sizes for
Cacatua, Charmosyna was merged with Trichoglossus, and Pionetes and Pyrilia were
merged with Pionus. Abundance estimates from primary habitat were coded as either
PROTECTED or NOT PROTECTED. The former included national parks and other
nature reserves, forest reserves or wildlife management areas, this information being
abundance estimates within studies. We combined a few studies (e.g. Riley 2002, 2003)
under the same study term if they used the same methods in nearby areas. Studies
reporting fewer than five density estimates were combined and included in the study
random terms as OTHSTUD (i.e. other studies). We used a random intercept model
(e.g. Zuur et al. 2009) as we were not so much interested in the intercept values of
Analysis was conducted within the package nmle in R (R Development Core Team
corrected for low sample sizes (AICc) because of the relatively large number of
parameters per data point, and identified the variables most likely to appear in reliable
models by summing the Akaike weights of each model in which a variable appeared
(Symonds & Moussalli 2011). We present all models and AICc values in Appendix S2
abundance change data for parrots that fitted all our criteria for inclusion. Of these, 72
dealt with parrots in wooded habitats. Of 300 density estimates found, 253 (84%) were
derived using distance sampling, with 197 (66%) from point count distance sampling,
56 (19%) from distance line transects, and the remainder derived using spot-mapping
24 (8%) and a variety of other methods including nest or roost counting. Mean total
length of transects used for density estimates derived from distance-sampled line
transects was 676 ± 1001 km (SD; some surveys > 3000 km of effort; n = 48), and for
distance-sampled point count surveys was 190 ± 164 km (SD; n = 148). All density
confidence intervals. Density estimates derived from line transects had mean
coefficients of variation (CVs) of 31% ± 15 (SD; n = 41), while for point counts the mean
At least one density estimate was available for 90 species (c. 25% of all 356
species). Most species had density estimates available from a single study but 22
species had estimates from two studies, 12 from three studies, and one species, Eclectus
Parrot Eclectus roratus, from five studies. The proportions of species with at least one
density estimate did not differ significantly across biogeographical region (G = 8.6, df =
4, P = 0.10), although Asia had a relatively high proportion of species with density
estimates (22 from 58) and Africa and related islands a relatively low proportion (two
= 0.76, df = 1, P = 0.38).
Density or encounter rate data that allowed comparisons of abundance across space or
time were available for 80 species (23% of all species) from 40 genera (Appendix S1).
Paired abundance estimates across habitat or across site were available for 179
habitat/site pairs. The most common comparisons were between primary forest and
selectively logged forest, and across different primary forest types/sites (Table 1). Most
comparisons were between primary forest and an anthropogenic habitat – there were
species were more likely to have a comparison of abundances across habitats than were
Density estimates in primary forest/woodland habitats for those parrot genera with
more than eight density estimates are shown in Figure 1. Most Asian/Australasian
genera had higher estimated densities than those from the Neotropics – an exception
was the Asian genus Tanygnathus. Genera of larger-bodied parrots such as Ara,
Aratinga, Pionus and Amazona) were positively skewed – i.e. most species were rare but
there was a long tail of ‘commonness’ in a few species. Species with surprisingly high
Blue-eyed Cockatoo Cacatua ophthalmica of New Britain and the St Lucia Amazon
Amazona versicolor, one of the few island species in the Neotropical sample. In three
genera (Eos, Tanygnathus and Ara), abundances were negatively skewed, with most
on average negative in all land uses except agriculture (Table 1). Paired tests of
forest and secondary forest, between primary forest and plantation and, marginally,
primary forest and secondary forest were modest compared to those of some other
land uses (e.g. absolute difference =1.70 as compared to 4.08 for selectively logged
compared with logged forest. Natural variation in parrot abundance across different
primary habitats was substantial and comparable with abundance changes between
appeared randomly distributed along the abundance axis. The distribution of residuals
in different genera appeared unremarkable but, as expected, was diverse for the group
OTHGEN. Genus appeared in almost all of the top models and had the highest Akaike
variable weight (Table 2; Appendix S2). Within the best supported model, coefficients
were significantly different from zero in six of eleven genera (OTHGEN excluded).
Protection was also a strong and significant candidate variable appearing in the top
three models and having an Akaike weight of 0.71. Densities tended to be higher in
abundance variation.
DISCUSSION
The lack of data is serious for all parrot groups, and acute for some, and we question
whether it will ever be possible able to gather enough data on parrots to inform
available. Even if our extensive data trawl had detected only half of the estimates
estimates would still only be occurring at the rate of 12 species per year. If we then
This is a serious degree of data deficiency when one compares this rate of data
generation to the timescales considered in Red List decline criteria or the periods over
which trade can escalate. For example, one Red List criterion is the percentage decline
over three generations, which is 9-48 years in parrots based on minimum and
whilst exports of African Grey Parrot Psittacus erithacus increased from under 18 000 in
Scarlet Macaw Ara macao has an extent of occurrence well in excess of 5 million km2
and has just four density estimates from a single study (Lee & Marsden 2012). There is
likely to be great variation in abundance across a species’ range, for example between
densities at the periphery of its range compared to that in the centre (e.g. Gaston et al.
1997). There are big differences in parrot abundance across discrete geographical units
such as islands (e.g. Lambert 1993, Marsden 1999). Eclectus Parrot is the best studied
species, with estimates from five studies, but it occurs on over 50 islands across its
oriented scientific papers per species), the Psittaciformes are unremarkable, being
ranked 13th of 25 bird orders by Brito and Oprea (2009). Our findings, in terms of
availability of abundance estimates for parrots, resonate with the findings of de Lima et
Globally Threatened species were no more likely to have a density estimate than non-
threatened species, and were less likely to have a comparison of abundance changes
across habitats. Generally, threatened bird species are no more the focus of scientific
research than are non-threatened species (Brooks et al. 2008). Of course, there are also
Houston et al. 2007), but such cases are not the norm across the order. With such huge
data gaps and with no evidence for any recent increase in research output (98 density
estimates were published during the 1990s, and 66 from the period 2000-2009), there
environmental changes.
In some genera, we can be reasonably sure of approximate density estimates for most
species in pristine habitats. For example, Tanygnathus, Aratinga and Ara appear to
comprise invariably low density species. In other groups, such as Pionus, Cacatua and
Amazona, densities of 10 per km2 are the norm, but there are individual taxa or
situations in which species can achieve unusually high densities. In the case of these
taxa, it was species such as Purple-bellied Lory, Blue-eyed Cockatoo and St Lucia
Amazon that were particularly common. All of these species occur on islands, but
In a conservation context, errors arising from assuming rarity but finding abundance
are not a great problem. However, there appear to be other genera (e.g. Eos, Lorius)
within which most species in most situations are common but a few are unusually rare.
means we may overlook species with very low populations. Examples include the
North Maluku (Lambert 1993) compared to this and other Lorius species elsewhere (e.g.
Models suggested that parrot abundance in primary habitat was most strongly
linked to two predictors - genus and forest protection. The former again lends weight
to the idea that there is a phylogenetic signature in parrot abundance, as has been
found in other taxonomic groups (e.g. Kelly et al. 2008). There has been much recent
Andam et al. 2008, Rayner et al. 2014). Protection, including national park status and
other designations, does not always provide strict protection to either parrots or the
forest in which they live (e.g. Metzger et al. 2010), so it is encouraging that protection
was an important predictor of higher parrot densities. How much of this is due to
specific resources such as nest holes (e.g. Gibbons & Lindenmayer 1996), or how much
environmental changes, meant that we had little power to predict species’ responses to
secondary forest or through logging were highly variable but tended to be negative.
more species than it negatively affected. As with the other across-habitat abundance
species were included in the analysis due to a lack of data. Studies of parrots in small-
scale agriculture have found high food availability in this habitat as compared with
primary forest (Marsden & Pilgrim 2003) and a high usage by frugivores of agricultural
plots which border more pristine forests (Marsden & Symes 2008). The variation in
parrot responses to land use change mirrors the diversity of abundance changes noted
within some taxa across land uses found by others (Gray et al. 2007, Scales & Marsden
2008). Many Red List classifications are built on assumptions that a taxon will react to a
given land use change at an unstudied site in a certain way based on reactions of the
same or related species elsewhere or to a different land use change. Clearly, further
work to make sense of the responses of parrots and other groups to land use changes,
plantations vs. small-scale agroforests) could yield enormous benefits, if clear patterns
can be found.
of parrots, and indeed many other families of birds (Buckland et al. 2008). While not
always well suited to tropical conditions and parrots, we suggest that they remain the
best option for parrot ecologists outside situations in which total counts can be made,
or perhaps where roost counts are feasible (e.g. Matuzak & Brightsmith 2007). There
have also been recent advances in tailoring distance sampling methods to individual
situations such as tall, structurally complex Amazonian forests, through the use of a
cue-counting variant of line transects (Lee & Marsden 2012). Recent work has also
African grey parrots can be estimated from simpler encounter rates (CITES 2013).
Importantly, choice of survey method (distance line transects, distance point counts
and ‘other’ including roost counts) did not have a consistent influence on parrot
density estimates. This is reassuring for two reasons. First, point counts have
sometimes been found to inflate bird densities over those of line transect and other
methods (e.g. Casagrande & Beissinger 1997). Second, several reference works suggest
the use of line transects rather than point counts where the target species is likely to be
rarely encountered (e.g. Bibby et al. 2000; Buckland et al. 2008). This may mean that
an effect of method choice, and while we caution that some method-related influence
might be associated with regional or taxon-related bias in method choice, we can report
Population studies of many threatened parrots and other birds are frustrated by
low population densities, which make accumulation of sufficient records for reliable
abundance estimation very time- and resource-consuming (Marsden 1999). There is, of
course, no easy solution to this problem, but there are procedures that can allow
distance sampling to be used reasonably effectively with rare species. Espanola et al.
(2013) combined the 63 records of four rare parrots (genera Tanygnathus and
probability and derived species-specific density estimates, including those for rare
species, using the multiplier function in DISTANCE. A variant of this procedure would
of species (body size, call rate, behaviour, stratum used) to pair up species and use the
many records of the more common species to model detection in the rarer one (with
the assumption that detectability is indeed equivalent; Buckland et al. 2008). With the
are being increasingly used to predict ranges and range changes (e.g. Marin-Togo et al.
2012), and examine habitat suitability and likely areas for reintroduction (Thorn et al.
2009). While not useful per se for abundance measurement, they can feed into other
corresponding point estimates of density from field surveys (e.g. van der Wal et al.
2009). There are issues to overcome, particularly with rare species with few or highly
clustered location data, and with predicting accurately actual local density rather than
maximum potential density (Torres et al. 2012), but the methods show promise. If a
workable relationship between occurrence probability and local density is found for
parrots, then coupling the two methods, while not negating the need for field surveys,
may greatly increase our ability to predict abundance and abundance changes for a
Our thanks go to those authors who kindly shared copies of their published papers
with us, and to Carlos de Araújo and Juan Masello for making the virtual library of the
general, and Stuart Butchart in particular, for making available to us their database of
REFERENCES
Beissinger, S.R., Wunderle, J.M., Meyers, J.M., Saether B.E. & Engen, S. 2008.
401-408.
Bibby, C.J., Burgess, N.D., Hill, D.A. & Mustoe, S.H. 2000. Bird Census
Bolker, B.M., Brooks, M.E. Clark, C.J. Geange, S.W. Poulsen, J.R., Stevens,
practical guide for ecology and evolution. Trends Ecol. Evol. 24: 127-135.
Brito, D. & Oprea, M. 2009. Mismatch of research effort and threat in avian
Brooks, T.M., Collar, N.J., Green, R.E., Marsden, S.J. & Pain, D.J. 2008. The
Buckland, S.T., Marsden, S.J. & Green, R.E. 2008. Estimating bird abundance:
Casagrande, D.G. & Beissinger, S.R. 1997. Evaluation of four methods for
Collen, B., McRae, L., Deinet, S., De Palma, A., Carranza, T., Cooper, N., Loh, J. &
Collar, N.J. 1997. Psittacidae (Parrots). In: del Hoyo, J., Elliott, A. and Sargatal, J.
(ed.), Handbook of the birds of the world, 280-477. Barcelona: Lynx Edicions.
Collar, N.J. 1998. Information and ignorance concerning the world’s parrots: an
235.
Collar, N.J. & Juniper, A.T. 1992. Dimensions and causes of the parrot
conservation crisis. In Beissinger, S.R. & Snyder, N.F.R. eds. New World
de Lima, R.F., Bird, J.P. & Barlow, J. 2011. Research effort allocation and the
627-632.
Espanola, C., Collar, N.J. & Marsden, S.J. 2013. Are populations of large-
the parrot trade in Peru and the potential importance of internal trade for
Gaston, K.J., Blackburn, T.M. & Gregory, R.D. 1997. Interspecific abundance-
20: 390–399.
Gibbons, P. & Lindenmayer, D.B. 1996. Issues associated with the retention of
Gray, M.A., Baldauf, S.L. Mayhew, P.J. & Hill, J.K. 2007. The response of
avian feeding guilds to tropical forest disturbance. Conserv. Biol. 21: 133-
141.
Houston, D., Mcinnes, K., Elliott, G., Eason, D., Moorhouse, R. & Cockrem, J.
Kelly, C.K., Bowler, M.G., Pybus, O. & Harvey, P.H. 2008. Phylogeny, niches,
Lambert, F.R. 1993. Trade, status and management of three parrots in the North
garrulus and violet-eared lory Eos squamata. Bird Conserv. Int. 3: 145-168.
Lee, A.K.T. & Marden, S.J. 2012. The influence of habitat, season and
Lloyd, B.D. & Powlesland, R.G. 1994. The decline of Kakapo Strigops
69: 75-85.
Mace, G.M., Collar, N.J., Gaston, K.J., Hilton-Taylor, C., Akcakaya, H.R.,
MacKenzie, D.I., Nichols, J.D., Royle, J.A., Pollock, K.H., Bailey, L.L. & Hines,
coast: potential impacts of habitat loss and capture for trade. Biodivers.
Marsden, S.J. 1999. Estimation of parrot and hornbill densities using a point
Marsden, S.J. & Pilgrim, J.D. 2003. Factors influencing the abundance of
Marsden, S.J. & Symes, C.T. 2008. Bird richness and community composition
33: 784-793.
decline and recovery of rare and vulnerable fish. J. App. Ecol. 42: 25-37.
Metzger, K.L., Sinclair, A.R.E., Hilborn, R., Hopcraft, J.G.C. & Mduma, S.A.R. 2010.
Nepstad, D., Schwartzman, S., Bamberger, B., Santilli, M., Ray, D.,
Schlesinger, P., Lefebvre, P., Alencar, A., Prinz, E., Fiske, G. & Rolla, A.
Pires, S.F. 2012. The illegal parrot trade : a literature review. Global Crime 13:
176-190.
Riley, J. 2002. Population sizes and the status of endemic and restricted-range
bird species on Sangihe Island, Indonesia. Bird Conserv. Int. 12: 53-78.
Riley, J. 2003. Population sizes and the conservation status of endemic and
46: 118-127.
Smyth, A., Mac Nally, R. & Lamb, D. 2002. Comparative influence of forest
Snyder, N., McGowan, P., Gilardi, J. & Grajal, A. (eds.) 2000. Parrots. Status
Thorn, J.S., Nijman, V., Smith, D. & Nekaris, K.A.I. 2009. Ecological niche
15: 289-298.
Torres, N.M., De Marco, P., Santos, T., Silveira, L., Jacomo, A.T.D. &
Verissimo, D., MacMillan, D.C. & Smith, R.J. 2011. Toward a systematic
van der Wal, J., Shoo, L.P., Johnson, C.N. & Williams, S.E. 2009. Abundance
niche models predicts the upper limit of local abundance. Am. Nat. 174: 282-291.
Trujillo, A., Beissinger, S.R., Berovides, V., Galvez, X., Brice, A.T.,
Joyner, K., Eberhard, J., Gilardi, J., Koenig, S.E., Stoleson, S.,
Asanza, A.C., Vilella, F.J. & Wiley, J.W. 2001. Nest poaching in
Zuur, A.F., Ieno, E.N. & Elphick, C.S. 2010. A protocol for data exploration to
PR SL SEC AGR
PLAN FRAG
Directional changes
n= 31 n = 47 ns n = 24 ** n = 21 ns
n = 12 ** n = 18 ns
SL n=0 n=3
n=0 n=0
n = 9 ns
n=2 n=0
AGR
n=2 n=0
n=6*
Absolute differences
PR SL SEC AGR
PLAN FRAG
Abundance
* * * 0
* * 0.01*
* * 1.55
Coeff ± SE