Overview
Many measures of community structure, such as diversity indices and rank-abundance curves, represent “snapshots in time” - which do not capture the temporal dynamics of ecological systems. For example, species richness (i.e., the number of species present) is a common metric to characterize ecological communities. However, species richness may be a poor indicator of community change if there is turnover in the identity but not the number of species present over time (Collins et al. 2008; Cleland et al. 2013). Similarly, species rank abundances can reorder through time, even when the pool of species present remains relatively stable. Within-year rank abundance curves fail to capture this dynamic, and other tools are needed to quantify within-community reordering without masking informative temporal shifts in internal dynamics (Collins et al. 2008). codyn includes three functions to characterize temporal shifts in species identity and rank abundances over time:
turnovercalculates total turnover as well as the proportion of species that either appear or disappear between time points.rank_shiftquantifies relative changes in species rank abundances by taking the sum difference of species ranks in consecutive time points. This metric goes hand-in-hand with “rank clocks,” a useful visualization tool of shifts in species ranks.rate_changeanalyzes differences in species composition between samples at increasing time lags. It reflects the rate of directional change in community composition. The associated functionrate_change_intervalreturns the full set of community distance values and associated time lag intervals to use in visualization.
## Example dataset
`codyn` utilizes a subset of the data presented by Collins et al. [-@Collins2008]. The `collins08` dataset includes plant composition from the Konza Prairie Long-term Ecological Research site in Manhattan, KS. It spans 18 years and includes data collected from two locations, one that is annually burned and one that is unburned.
|replicate | year|species | abundance|
|:---------------|----:|:--------|---------:|
|annually burned | 1984|achimill | 0.050|
|annually burned | 1984|ambrpsil | 1.350|
|annually burned | 1984|amorcane | 1.325|
|annually burned | 1984|andrgera | 76.000|
|annually burned | 1984|antenegl | 0.050|
|annually burned | 1984|apoccann | 0.075|
## Species turnover
Species turnover represents a temporal analog to species richness. It is described by the total turnover in species (appearances and disappearances) as well as the proportion of species that either appear or disappear between time points [@Collins2008; @cleland2013].
### Total turnover
The function `turnover` calculates three metrics of species turnover: total turnover, appearances, and disappearances.
The default metric `total` refers to total turnover, which calculates the proportion of species that differ between time points as:
$$ Total\; turnover = \frac{Species\; gained\; +\; Species\; lost}{Total\; species\; observed\; in\; both\; timepoints} $$
The metric was introduced by MacArthur and Wilson [-@MacArthur1963] and modified by Diamond [-@diamond1969] to be expressed as a proportion in order to compare turnover between sites that differ in species richness.
`turnover` requires a data frame with columns for species, time and abundance, and includes an optional argument to specify a column for spatial replicates.
```r
KNZ_turnover <- turnover(df = collins08,
time.var = "year",
species.var = "species",
abundance.var = "abundance",
replicate.var = "replicate")| total | year | replicate |
|---|---|---|
| 0.237 | 1985 | annually burned |
| 0.302 | 1986 | annually burned |
| 0.242 | 1987 | annually burned |
| 0.230 | 1988 | annually burned |
| 0.185 | 1989 | annually burned |
| 0.224 | 1990 | annually burned |
Note that if the replicate.var column is not specified but the dataset in fact includes replicates, calling turnover will result in an error. For example, forgetting to specify the replicate.var, allowing it to assume the default value of NA, yields an error because multiple abundance values are included for the same species within a single year.
KNZ_turnover_agg <- turnover(df = collins08,
species.var = "species",
time.var = "year",
abundance.var = "abundance")## Error in check_single_onerep(df, time.var, species.var): Either data span multiple replicates with no replicate.var specified or multiple records within years for some speciesAppearances and disappearances
Total turnover incorporates both species appearances and disappearances, but sometimes it is useful to parse their relative contribution. For example, a time point in which many species appear may reflect a different ecological story than a time point in which many species drop from the system, but the total turnover in both scenarios may be similar.
Specifying metric="appearance" will return the proportion of species that appeared relative to the total number of species observed in both time points. As before, spatial replicates can be specified with the replicate argument; setting replicate.var=NA will calculate across the full dataset.
KNZ_appearance <- turnover(df = collins08,
time.var = "year",
species.var = "species",
abundance.var = "abundance",
replicate.var = "replicate",
metric = "appearance")Similarly, specifying metric="disappearance" will return the proportion of species that disappeared relative to the total number of species observed in both time points.
Turnover at Konza
Total turnover indicates that in general there were greater fluctuations in the species present in the unburned than burned location at Konza.
Parsing total turnover by appearances and disappearances indicates that there were generally a higher proportion of appearances in the unburned than burned location. In addition, the strong divergence in total turnover between locations in 1995 was because the unburned location lost a high proportion of species that year, while the annually burned locations lost none.
