A multidimensional framework for measuring biotic novelty: How novel is a community?

Anthropogenic changes in climate, land use and disturbance regimes, as well as introductions of non-native species can lead to the transformation of many ecosystems. The resulting novel ecosystems are usually characterized by species assemblages that have not occurred previously in a given area. Quantifying the ecological novelty of communities (i.e. biotic novelty) would enhance the understanding of environmental change. However, quantification remains challenging since current novelty metrics, such as the number and/or proportion of non-native species in a community, fall short of considering both functional and evolutionary aspects of biotic novelty. Here, we propose the Biotic Novelty Index (BNI), an intuitive and flexible multidimensional measure that combines (1) functional differences between native and non-native introduced species with (2) temporal dynamics of species introductions. We show that the BNI is an additive partition of Rao’s quadratic entropy, capturing the novel interaction component of the community’s functional diversity. Simulations show that the index varies predictably with the relative amount of functional novelty added by recently arrived species, and they illustrate the need to provide an additional standardized version of the index. We present a detailed R-code and two applications of the BNI by (1) measuring changes of biotic novelty of dry grassland plant communities along an urbanization gradient in a metropolitan region and (2) determining the biotic novelty of plant species assemblages at a national scale. Results illustrate the applicability of the index across scales and its flexibility in the use of data of different quality. Both case studies revealed strong connections between biotic novelty and increasing urbanization, a measure of abiotic novelty. We conclude that the BNI framework may help in building a basis for a better understanding of the ecological and evolutionary consequences of global change.


Introduction
native and non-native species may gradually adapt to their new interaction partner(s) . We argue that a species 107 that is functionally dissimilar from the resident species represents greater biotic 108 novelty than one that is similar to the pre-existing community. We describe how to calculate the BNI from various data sources, and how it 148 associates with traditional measures of biotic novelty, abiotic novelty, species 149 richness and functional diversity. By presenting simulations and two case studies, we 150 show that this new method to quantify biotic novelty is intuitive and versatile, as it is 151 easily adaptable to datasets of different scale, scope and resolution. We demonstrate 152 in this paper that the BNI framework is a helpful tool whenever the assessment of 153 novel species assemblages or communities is needed, which may not only be useful  The functional diversity component 174 The general rule to calculate functional diversity indices is that traits must be linked to 175 the function(s) of interest. For instance, specific leaf area, maximum growth rate and about which traits to include and how to weigh them depends on the purpose to 185 which the index will be applied and should rely on expert knowledge of the system 186 (Laliberté & Legendre 2010 be constructed with the same dimension as the trait distance matrix described before.

234
The values of the temporal matrix range between 0 and 1 (due to the normalization 235 step given in equation 2) and functions as weighting factor for the trait distance ma-236 trix. In this way, trait differences between species with low coexistence time are 237 weighted heavily, whereas trait differences between species coexisting for millennia 238 (such as a pair of native species) will be given no weight in the BNI.

239
The BNI as a framework 240 The BNI is in essence the sum of two components: the mean functional distance be- cies, or 1 for pairs involving at least one non-native species.

288
The BNI as described above is a multispecies approach since it captures the func- Simulations of plant communities were used to explore the behavior of the BNI in dif-296 ferent scenarios of functional diversity and biological invasion. We randomly generat-297 ed a regional pool of 250 species, with 70 % natives and 30 % non-natives. In order 298 to spread the simulated residence times realistically, we followed the three-level clas- (2) traits of neophytes have on average higher values than the residents (i.e. different

395
Simulations showed that the BNI varies broadly with the proportion of non-native 396 species and with the size of trait differences between species (Fig. 3). Overall, as 397 long as neophytes made up less than 50 % of the relative abundance of species in 398 the community, the BNI increased monotonously as more neophytes were added.

399
Beyond this point, however, the BNI did not always increase with the proportion of 400 neophytes. Its behavior depended on how much pairwise trait variance the neo-401 phytes were bringing to the community, relative to the resident species.

402
In scenario 1, when neophytes were not on average functionally different from na-403 tives, the BNI increased monotonously with the proportion of neophytes (Fig. 3a).

404
This is because, in this scenario, the mean pairwise trait differences (i.e. Rao's Q) remained constant, while the contribution of neophytes increased with their relative 406 abundance in the community. The BNI simulation curve tended to saturate at high 407 neophyte proportions as new neophyte species were less likely to add new trait dif-408 ferences.

409
In scenario 2 and 4, when neophytes were on average functionally different from the 410 residents, the simulated BNI often showed a humped-shaped curve, with a maximum 411 at intermediate proportions of neophytes (Fig. 3c, g). This pattern is due to the fact 412 that the BNI is based on mean pairwise differences between species, which reaches

494
The BNI captures novelty in both functional diversity and introduction history 495 We designed the BNI to combine two aspects of ecological novelty: historical novelty, native species (Fig. 5d).

567
Further, by applying the standardization of the BNI (the BNI in proportion to Rao's Q), 568 we showed in the first case study that the BNI was not driven by the inherent varia-569 tion in functional diversity along the urbanity gradient (since BNI and BNIs varied to a 570 very similar extent along the gradient). As shown in our methods section, this stand-571 ardization of the BNI can be easily applied by the user for a validation of the BNI re-572 sults.

573
The second case study demonstrated the applicability of the BNI to nationwide da-

594
We are aware that analyzing a dataset with the extent of our second case study is 595 not free of concerns. For example, the large grid-cell size (11 x 11 km) and the spa-  However, since it is the scope of this paper to demonstrate possible applications of 601 the BNI rather than disentangling various factors that structure biotic novelty, we re-602 frained to perform complex statistical analysis and chose to present a map without 603 underlying models. Therefore, it is up to future studies to focus on this demanding 604 task.  introduced non-native plants (neophytes). Scenarios explore different parameters (mean and 873 SD) of the normal distribution from which species traits for neophytes were sampled. In the 874 first scenario (a, b), traits of native and non-native species follow the same normal distribu-875 tion (trait mean = 0, SD = 1). In scenario 2 (c, d), the mean trait values of neophytes are in-876 creasingly different from the natives (colors represent variation in neophyte trait mean from 0 877 to 10; SD = 1). In the third scenario (e, f), natives and neophytes have the same trait mean 878 (mean = 0), but neophyte trait SD increases from 0 to 10. In the fourth scenario (g, h), both 879 the mean and SD of neophyte trait distributions increase together from 0 to 10 and 0 to 5, 880 respectively. Lines represent LOESS regressions fitted on the 100 simulated points corre-881 sponding to one simulation run. 882 ized BNI. Asterisks indicate statistical significance using linear models ('***' = P < 0.001, '**' = 887 P < 0.01, '*' = P < 0.05, 'n.s.' = P ≥ 0.05). 888