bioscience – aarhus university pin-point plant cover data christian damgaard bioscience aarhus...
TRANSCRIPT
Bioscience – Aarhus University
Pin-point plant cover data
Christian Damgaard
Bioscience
Aarhus University
Bioscience – Aarhus University
Hierarchical plant cover data
Large-scale ecological processes (among-sites): environmental drivers extinction / colonization of sites
Small-scale ecological process (within-sites): size of individuals density-dependent population growth inter-specific competition
Plant cover data has too many zero values and too much variance compared to a binomial distribution
U- shaped distributions of plant cover are typical
Bioscience – Aarhus University
The pin-point (point-intercept) methodObjective method for measuring plant cover
Place a frame with a grid pattern
A pin is inserted vertically through one of the grid points into the vegetation
The pin will typically touch a number of plants and the different species are recorded
This procedure is repeated for each grid point
Bioscience – Aarhus University
Distribution of pin-point cover data within a site
The distribution of plant cover is modelled using a generalised binomial distribution with two parameters q : mean plant cover d : intra-plot correlation
The probability density function of the generalised binomial distribution (=Pólya-Eggenberger distribution) is equal to the beta-binomial mixture distribution but is somewhat more general in that negative intra-plot correlation is allowed (Qu et al. 1993)
Bioscience – Aarhus University
Generalised binomial distribution
n
ynq
yq
y
nqnyfY
,11
,)1)(1(
,11
),,;(
)1()1()()(),( nxxxxnxnx
qnYE )(
))1(1()1()( nqqnYVar
q : mean cover at the site d : intra-plot correlation
Bioscience – Aarhus University
The parameters
q : mean plant cover – may be regressed to environmental gradients test hypotheses on the effect of environmental gradients on plant
abundance (Damgaard 2008, Damgaard 2012)
d : intra-plot correlation - depends on size of individuals and spatial arrangements test hypotheses on the effect of environmental gradients on the spatial
variation The intra-plot correlation parameter may be generalized to a
multispecies case and be used to test for different hypotheses on the level of the community test of neutrality (Damgaard and Ejrnæs 2009)
does climate change or nitrogen deposition change the spatial structure or increase size of plants? (Damgaard et al. 2012)
Bioscience – Aarhus University
Relationship between plant cover and intra-plot correlation in dune grasslands
Bioscience – Aarhus University
Important to include spatial correlation
The cover of Calluna vulgaris and Deschampsia flexuosa on dry heathlands
Bioscience – Aarhus University
Important to include spatial correlation
If d was set to zero (no spatial correlation) then there was a strong significant effect (grey line, P < 0.0001)If d was allowed to vary then the effect was found to be insignificant (black line, d = 0.57, P = 0.19)
Bioscience – Aarhus University
Among site-variation in cover
The zero-inflated beta-distribution is a good candidate distribution to model among-site variation in plant cover (Damgaard, in prep)
Bioscience – Aarhus University
Analysing trends in plant cover Framework for analysing trends of plant cover data has been
developed Bayesian state-space model Separation of process and sampling error Efficient treatment of missing values Effect of treatment or co-variable on change in cover Possible auto-correlation
Published in Ecology 93(6), 2012, pp. 1269–1274
Bioscience – Aarhus University
Conclusions The parameters in the generalised binomial distribution may
be used to test ecological hypotheses
Erroneous conclusions if the spatial variation is ignored also when assuming a naïve model with normally distributed
data