The collapse indicators of the dominant species outperform community-level critical slowing down indicators

We studied similarities of collapses and concluded that dominant species have a universal effect on the pattern of collapses. We used open paleoecological and modern data to detect ‘early warning signals’ of collapses. We tested and ranked collapse indicators at the community level (abundance, species richness, constancy, dominance, Shannon’s H, standard deviation, variance, lag-1 autocorrelation of community abundance, lag-1 autocorrelation of Shannon’s H) and at the level of the dominant species (total changes of pre-collapse and collapse dominant species, lag-1 autocorrelation of pre-collapse and collapse dominant species) based on their performances. We distinguished between small-scale signals (sharp drops and peaks) and large-scale signals (trend changes). Small-scale signals and large-scale signals can be at the same time, however, small-scale signals can also precede large-scale signals. Small-scale signals indicate environmental events. Large-scale trend changes refer to the decline and eventually the collapse of the community. Our results show that the collapse indicators of the dominant species outperform community-level critical slowing down indicators, which suggests that the dominant species probably have an important role in community-level collapses triggered by environmental events. We also concluded that unusual environmental events might be the number one cause of collapses therefore small-scale signals should be involved in analyses.


38
It has been long debated whether the different causes of events evoke different dynamics and 39 patterns of collapses. This study aims to find similarities among collapses. We share the views 40 by Erwin [1] that collapses have a primary abiotic and a secondary biotic aspect. Unusual

109
To detect universal signals of collapses, we included five data series (the relative abundances

113
Glossary 114 We introduce/reintroduce some phrases for community collapses in this study in our 115 understanding.

116
Community collapse -The collapse of the dominant species, the rise of a new dominant 117 species and the complete structural changes in the community indicate community collapse.

118
We initiate the use of 'community collapse' phrase for both community shifts and extinctions 119 based on common features.

200
The modern data series originates from a paleolimnological record of diatoms. The

214
To locate the collapses in the data series and to define the possible warning signals, we applied 215 both univariate and multivariate indicators at the community level and the level of the dominant 216 species.

217
Community and unusual event indicators 218 We applied the relative abundance of species (%), HCA (hierarchical cluster analysis) and where Collapse indicators at the level of the dominant species 318 We presume that collapse indicators at the level of the dominant species are more effective in

352
Communities and boundaries 353 We applied the HCA and the relative abundance of species to identify the clusters and the 354 boundaries between them. The clusters/sub-clusters represent the communities.

356
In the KPG data series, the pre-collapse community is in the interval between 66.14235-

EM1 and EM2
388 The HCA of the original Late Oligocene-early Miocene data series shows four communities 389 (community 1-4) ( Fig 5B). EM1 data series includes the collapse of community 1 (pre-collapse 390 community) (Fig 6).  The modern data series shows a recent diatom shift under global warming in Lake Hazen

425
The   indicator, has a huge peak, at the bolide event (Fig 3A). In the EM1 data series, the relative 446 abundance of D. antarcticus, the indicator of the cooling [49], also has a high peak at the 447 temporary glaciation event (Fig 6A).

448
We had two types of collapse triggering environmental events in the data series: sudden, sharply. In the HCA, the outlier 65.70326 Ma separates the clusters before and after the bolide 456 event (Fig 3B). In the PCA, the outlier is in the right-hand upper corner away from the clusters 457 ( Fig 3C) suggests a connection between these time steps (Fig 4C). A smaller environmental event 473 might have also occurred at 79.995 mbsf as it is shown by HCA, however, it is a weak sign, 474 and therefore we did not involve it in the analysis. We hypothesize that the event at 79.955 475 mbsf is a pulse event initiating/contributing to the collapse before the PETM.

EM1 and EM2
477 The HCA, the relative abundance of the Late Oligocene-early Miocene nannofossils (Figs 5A 478 and 5B) and the total change of the dominant species (Fig 9) help to identify the environmental 137.21 mbsf (community 3) (Fig 9).

493
EM1 data series covers a period at the beginning of Early Miocene when the climate started 494 to cool. It includes a temporary glaciation event at 180.51 mbsf (Fig 6). EM2 data series covers 495 a glaciation period in the Early Miocene (Fig 7). It probably has a temporary warming event at high peak (Fig 7). Coccolithus spp. are indicators of interglacial periods [51].

500
The main reason for the recent diatom shift in Lake Hazen is the warming climate [20].

501
Unfortunately, the study by Lehnherr et al. [20] does not include information on the local 502 temperature data of the Hazen Lake region for the whole diatom data series, therefore we 503 applied an indirect method to be able to detect environmental events. The increasing summer 504 air temperature, the expansion of the growing season and the earlier melting of the ice sheet

505
[20] are probably important contributors to Arctic melting. The reduced summer albedo due to 506 sea ice and snow cover loss may also be a key component of Arctic amplification [52].

507
According to Finkelstein and Gajewski [48], the lower ratios of Staurosirella pinnata to 508 Staurosira construens refers to warmer summer air temperatures, while greater ratios suggest 509 cooler summers. Using this information, we can learn about the local trends of temperature 510 and perhaps unusual environmental events. To make it more expressive, we use the reverse 511 relationship between these two species, which means that the higher ratios of Staurosira 512 construens to Staurosirella pinnata refer to warmer summers. In Fig 10,  and Gajewski [48]. The increasing ratios refer to warmer summers. Red solid line = critical 522 threshold. The ratio first rose above the threshold in 1924, and it stayed above it after 1939.

523
The probable temporary cooling in 1930 may be a local event (Fig 10). As the figure shows,   (Tables 3-8). We also indicated if an indicator give consistent signals (increase or 607 decrease). Small-scale signals are sharp decreases or increases at the environmental events.

608
In the assessment of small-scale signals, we did not include the modern data series because

620
We identified a collapse-warning zone based on the detected signals. The collapse-warning 621 zone starts at the environmental event and it includes the small-scale and large-scale signals.

622
It is shorter than the whole pre-collapse period. In this study, the warning signal zones consist 623 of three or four time steps (the environmental event plus two or three time steps).

624
First, we evaluated the indicators by the number of signs the indicators give close to the event 625 (Tables 3 and 4). The indicators with the highest number of signals get the best ranks. The

626
small-scale signals of the modern data series are not included.  Second, we evaluated the signals by their earliness (Tables 5 and 6). We detected the signals 632 of the collapse warning zones between time steps 0-3 (time step zero = environmental event).

633
Based on the earliness, the time steps get weights as follows: event = 4; step1 = 3; step2 = 2; 634 step3 = 1. The indicators with the highest weight of signals get the best ranks. The small-scale 635 signals of the modern data series are not included.