Differential resilience of ancient sister lakes Ohrid and Prespa to environmental disturbances during the Late Pleistocene

Ancient lakes, such as lakes Ohrid and Prespa on the Balkan Peninsula, have become model systems for studying the link between geological and biotic evolution. Recently, the scientific deep-drilling project Scientific Collaboration on Past Speciation Conditions in Lake Ohrid (SCOPSCO) was initiated to better understand the environmental, climatic, and limnological evolution of the lake. It revealed that Lake Ohrid experienced a number of environmental disturbances during its ca. 2.0 million year long history. These are comprised of disturbances that lasted over longer periods of time (“press events”) such as glacial–interglacial cycles and Heinrich events, as well as sudden and short disturbances (“pulse events”) like the deposition of landslides, earthquakes, and volcanic ash depositions. The latter includes one of the most severe volcanic episodes during the Late Pleistocene: the eruption of the Campanian Ignimbrite (known as Y-5 marine tephra layer) from the Campi Flegrei caldera, dated to 39.6 ± 0.1 thousand years ago. The event is recorded by the deposition of a ca. 15 cm thick tephra layer in sediment cores of lakes Ohrid (DEEP-5045-1) and Prespa (Co1204). Coincidently, this pulse event is superimposed by the Heinrich H4 event, 40.4–38.4 thousand years ago. In the current paper, diatoms were used as proxies to compare the responses of these lakes to the Y-5 (pulse) and the H4 (press) disturbances. Based on stratigraphically constrained incremental sum of squares cluster (CONISS) and unconstrained Partitioning Around Medoids (PAM) analyses, we found little evidence that diatom community compositions in either lake responded to the H4 event. However, the Y5 influx caused clear and rapid diatom community changes. After the initial response, community compositions in Lake Ohrid and, to a lesser extent, in Lake Prespa slowly returned to their quasi pre-disturbance state. Moreover, there is no evidence for disturbance-related extinction events. The combined evidence from these findings suggests that lakes Ohrid and Prespa likely did not experience regime shifts. It is therefore concluded that both lakes show resilience to environmental disturbance. However, it seems that Lake Ohrid is more resilient than Lake Prespa, as the recovery of diatom communities is more pronounced and its estimated recovery time is only ca. 1100 years vs. ca. 4000 years in Lake Prespa. The reasons for the differential responses remain largely unknown, but differences in geology, lake age, limnology, and intrinsic parameters of the diatom proxies may play an important role.


Introduction 19
Ancient lakes, i.e., extant lakes that have continuously existed since before the last glacial 20 maximum (Albrecht and Wilke, 2008), have become model systems for studying the link 21 between geological and biological evolution over extended periods of time. For some ancient 22 lakes, such as Baikal (Russia) and Hövsgöl (Mongolia), it has been demonstrated that the 23 evolution of their species was largely shaped by massive environmental disturbances, like 24 extreme lake-level fluctuations and glacial-interglacial cycles (Karabanov et al., 2004). 25 However, for other ancient lakes, like the sister lakes Ohrid and Prespa on the Balkan 26 Peninsula, the link between geological and biotic evolution is not well understood. In order to 27 better understand the environmental, climatic, and limnological evolution of Lake Ohrid, the 28 SCOPSCO project was initiated. Early results revealed that the lake experienced a number of 29 environmental disturbances during its ca. 2.0 million year (Ma) long history (Lindhorst et al. , 30 2015). Some of these events lasted over longer periods of times and covered, for example, 31 glacial/interglacial cycles (Wagner et al., 2014) or Heinrich events (Wagner et al., 2010), i.e., 32 episodes of massive iceberg discharges that caused cooling of the North Atlantic during the 1 last glacial period (Bond et al., 1993). These events presumably intensified the aeolian 2 activity, lowered the temperature, and increased the aridity in the Ohrid region (Wagner et al., 3 2010). From a biological perspective, long-lasting disturbances (> several centuries) are 4 referred to as "press disturbances" (Niemi et al., 1990). In contrast, sudden disturbances with 5 a short and clearly defined duration (< few decades) are called "pulse disturbances" (Niemi et 6 al., 1990). Examples  It has been suggested that the interaction of volcanic ash deposition with a receiving lake 20 triggers perturbations, primarily through the effect of tephra weathering, but also through 21 changes in pH, mineral concentration, organic matter input, and short-term light deprivation 22 environmental disturbances. In fact, previous sediment core analyses suggest alterations in 2 diatom assemblage structure and abundances due to tephra influxes in both lakes (Cvetkoska 3 et al., 2012(Cvetkoska 3 et al., , 2014(Cvetkoska 3 et al., , 2015a. However, the low temporal resolution did not allow the diatom 4 data to be linked to distinct pulse events or used to estimate recovery periods (i.e., the time an 5 ecosystem needs to return to pre-disturbances conditions; the recovery period serves as 6 measure for resilience strength; Carpenter, 2013). Therefore, it remains unclear how the lakes 7 responded to such changes, and whether there were differences in response. 8 Given this lack of knowledge, the general goal of this paper is to use diatom community data 9 from the sediment records of lakes Ohrid (core DEEP-5045-1) and Prespa (core Co1204) as a 10 proxy to comparatively study the responses of these lakes to one of the most severe pulse 11 disturbance events during the late Pleistocene − the Y-5 tephra influx. Our specific objectives 12 are to study: 13

1)
Whether lakes Ohrid and Prespa had the resilience to tolerate this disturbance without 14 changing their regimes (i.e., without exceeding a critical threshold). Operational criteria 15 for resilience are the lack of disturbance-related extinction events in the diatom record 16 and a subsequent return of diatom communities to a quasi pre-disturbance state. 17

2)
If resilience can be demonstrated for one or both lakes, whether there are differences in 18 resilience strength between the two systems. The operational criterion for resilience 19 strength is the length of the recovery period, which is measured as the time the diatom 20 communities need to return to their quasi pre-disturbance state. 21 Lake Ohrid has long been considered to have a high level of ecosystem stability (sensu 22 Stanković, 1960;Föller et al., 2015), principally due to its depth, age, and peculiar karst 23 limnology. Hence, our working hypothesis is that Lake Ohrid is more resilient to 24 environmental disturbances than Lake Prespa. 25 Coincidently, the Y-5 tephra deposition (39.6±0.1 ka ago) is superimposed by the Heinrich opportunity to obtain information on the differential effect of a short pulse and a longer press 30 disturbance event. 31 35% H 2 O 2 and 10% HCl, and left overnight for the removal of carbonates. The samples were 23 then boiled in a water bath for 2 h in 37% HCl for oxidization of the organic matter (Renberg, 24 1990; Cvetkoska et al., 2012). The treated samples were rinsed several times with distilled 25 water and subsequently centrifuged for removing the products of the oxidation reaction. 26 Defined aliquots of the cleaned samples were settled onto coverslips and then mounted on 27 glass slides using Naphrax®. In each sample, random transects were selected and 200−400 28 diatom valves per slide were counted and identified by one of us (EJ) at 1000x magnification 29 with a Carl Zeiss, Axioplan 2 microscope equipped with a Nikon D5700 digital camera. All Recovery times were calculated by estimating the time differences between the same group-17 membership assigned by the PAM analyses before and after the tephra influx. As the diatom 18 communities sampled in Lake Ohrid are biased towards planktonic species due to the deep-19 water coring location, we determined recovery times both for planktonic and overall 20 communities in lakes Ohrid and Prespa. 21

Results 22
In total, 94 and 213 diatom species were identified in the cores of lakes Ohrid and Prespa, 23 respectively. Due to the difference in water depth of the coring locations (ca. 243 m for Lake 24 Ohrid vs. 14 m for Lake Prespa), planktonic species were dominant in Lake Ohrid, especially 25 members of the genus Cyclotella. Though many benthic species had been found, they only 26 occurred in low abundance. In Lake Prespa, planktonic and benthic species were roughly 27 balanced ( Figs. 2 and 3). 28 Some planktonic species showed a high morphological variability with respect to valve size, 29 shape of the central area, and number of ocelli in the central area (e.g., Cyclotella fottii and 30 Cyclotella ocellata). In order to fully cover the magnitude of potential community changes, 1 we assigned them to distinct morphotypes and identification units (see Figs. 2 and 3). 2

Identification of community response phases and diatom zones 3
The stratigraphically unconstrained PAM analyses identified three major community response 4 phases in lakes Ohrid and Prespa: A phase that corresponds to pre-disturbance conditions 5 (pre-tephra-disturbance phase; Fig. 4, also see the lower blue bars in Figs. 2 and 3), a distinct 6 disturbance phase (tephra-disturbance phase; Fig. 4, also see the green and yellow bars in 7 Figs. 2 and 3), and a phase in which communities had returned to quasi pre-disturbance 8 conditions (post-tephra-disturbance phase; Fig. 4, also see the upper blue bars in Figs. 2 and  9 3). 10 The stratigraphically constrained CONISS analyses identified three distinct diatom zones 11 together with several subzones each for lakes Ohrid (ODZs) and Prespa (PDZs). They largely 12 corresponded to the pre-tephra-disturbance phase (ODZ 3b−a and PDZ 3b−a), the tephra-13 disturbance phase (ODZ 2b−a and PDZ 2, 1d−b), and the post-tephra-disturbance phase (ODZ 14 1b−a and PDZ 1a) (see Figs. 2 and 3). 15 dominant with up to 50% relative abundance. In contrast, the benthic and facultative 20 planktonic species had abundances of up to 10% when taking the whole profile into account. 21

Community composition analyses and estimations of recovery times 6
The first ordination axis of the metric multidimensional scaling analyses indicates that the Y-7 5 tephra deposition caused very rapid changes in the diatom communities of lakes Ohrid (Fig.  8 4A) and Prespa (Fig. 4B). Given that the communities in Lake Ohrid's DEEP core were 9 dominated by planktonic species, the respective curves for overall (i.e., planktonic and 10 benthic communities) and planktonic communities in Fig. 4A showed similar patterns over 11 time. After the drastic change of community composition, coinciding with the tephra 12 deposition, communities reverted to a quasi pre-disturbance state (green bar in Fig. 2). 13 In Lake Prespa, where planktonic and benthic species were roughly balanced, the overall 14 community structure (grey curve in representative for the recovery period. For the planktonic communities of Lake Prespa, the 20 change coinciding with the tephra deposition was not as abrupt. However, the return to the 21 pre-eruption community state occurred even more gradually. 22 The diatom communities in both lakes Ohrid and Prespa did not display a strong response to 23 the onset of the H4 event 40.1 ka ago. In Lake Ohrid, H4 specific PAM clusters or CONISS 24 zones could not be detected. However, in Lake Prespa a distinct CONISS subzone coincides 25 with H4 (see Fig. 3). 26 The Ohrid communities had converted back to the quasi pre-disturbance state shortly before 27 the cessation of the H4 event 38.1 ka ago (grey and black dashed lines in Fig. 4A; also see the 28 upper blue bar in Fig. 2), whereas this process in the Prespa communities extended beyond 29 the end of the Heinrich event (grey and black dashed lines in Fig. 4B; also see the upper blue 30 bar in Fig. 3). The PAM analyses clearly show that the Ohrid and Prespa communities did 31 return to their quasi pre-disturbance states (see the upper blue bars in Figs. 2 and 3 and the 1 PAM clusters in Fig. 4), indicating that no regime shift occurred. 2 According to the age models of the two cores, the recovery times (i.e., the time differences 3 between the same group-membership assigned by the PAM analyses before and after the 4 tephra influx) for planktonic communities in lakes Ohrid and Prespa were ca. 1,100 and ca. 5 4,000 years, respectively (Fig. 4), following the Y-5 tephra influx. 6 4 Discussion 7 Our results indicated only mild effects of the H4-event on diatom community compositions in 8 lakes Ohrid and Prespa, though the impact is slightly greater in the latter one. In contrast, the 9 Y-5 influx caused clear and rapid responses in both lakes (Fig. 4). Whereas the overall 10 community composition in Lake Prespa partially recovered within a few decades, mostly 11 driven by benthic species, and then slowly returned to the quasi pre-disturbance state over an 12 extended period of time, the planktonic community needed a longer period of time for 13 recovery (compare the grey and black curves in Fig. 4B). 14 In Lake Ohrid, both overall and planktonic community composition indicated similar 15 reactions to the Y-5 tephra influx (Fig. 4A), owing the fact that planktonic communities 16 strongly dominated in the lake due to the depth of the drilling location. 17 When comparing changes in planktonic communities in lakes Ohrid and Prespa, overall 18 patterns are similar. An initial rapid response phase was followed by a phase in which 19 communities slowly returned to the quasi pre-disturbance state. However, as noted above, the 20 quasi pre-disturbance state in the Ohrid communities was reached shortly before the H4 21 cessation, whereas the Prespa communities recovered only long after the end of the Heinrich 22 event. 23

Diatom responses to disturbances in Lake Ohrid 24
The communities in the Ohrid core were mainly characterized by planktonic species (Fig. 2). 25 Although at low abundances, the benthic species likely indicate wind induced water currents, 26 water mixing, and/or sediment redistribution in the lake (cf. Vogel  Indications of these changes are the rapid replacement of the dominant hypolimnetic C. fottii 10 with the epilimnetic C. minuscula (Fig. 2). The latter species (only 3−7 µm in diameter) has 11 high silica incorporating rates and low transparency preferences, which makes it a strong The point of return to quasi pre-disturbance state was probably reached in subzone ODZ 1b, 20 when nutrient levels in the water column likely had recovered and silica levels had decreased. 21 This is indicated by the increase in abundances of the endemic C. fottii to pre-tephra-22 disturbance levels. As the recovery of planktonic communities was achieved prior to the end 23 of the H4 event (ca. 1,100 years), we here suggest that this press disturbance possibly 24 amplified the impact of the Y-5 and prolonged the recovery, but did not prevent it. 25

Diatom responses to disturbances in Lake Prespa 26
In contrast to the diatom communities in the Ohrid core, Prespa communities were 27 characterized by significant abundances of both planktonic and benthic species. During the 28 pre-tephra-disturbance phase (42.9−39.6 ka ago), the ordination (Fig. 4B) indicates only little 29 change in overall community composition. However, planktonic communities did show 30 moderate fluctuations in structure even before the onset of the H4 event 40.1 ka ago. 31 Moreover, the geochemical properties of the lake changed only moderately with the onset of 32 13 the H4 (Wagner et al., 2010). Therefore, it remains difficult to quantify the immediate 1 community impact of this press disturbance event. 2 The Y-5 associated silica fallouts (PDZ 2) rapidly altered the water chemistry by increasing 3 the silica content (ca. 60% SiO 2 in the tephra layer, Sulpizio et al., 2010) in the water column, 4 and likely affected the nutrient pool in the lake. The increased silica content favoured the 5 growth of planktonic species like C. minuscula, C. ocellata, C. paraocellata, and C. aff. 6 minuscula. The latter taxon has never been reported before. It occurs exclusively during the 7 recovery period and failed to establish permanently. 8 In contrast to the planktonic species, epiphytic and facultative planktonic species like 9 Cocconeis pseudothumensis, Staurosirella pinnata, and Pseudostaurosira brevistriata 10 temporally decreased in relative abundance (i.e., for a period of few decades). This may be 11 explained by a short-term destruction of the littoral macrophytic habitats as a result of the Y- 5 12 influx. The first specific objective of this study was to evaluate whether lakes Ohrid and Prespa had 6 the resilience to tolerate environmental disturbances without changing their regimes (i.e., 7 without exceeding a critical threshold sensu Scheffer and Carpenter, 2003). Our operational 8 criteria for assessing resilience were i) the lack of disturbance-related extinction events in the 9 diatom records and ii) a subsequent return of diatom communities to their quasi pre-10 disturbance state. 11 The data obtained are informative in this regard: we do not see extinction events directly 12 related to the H4 and/or Y-5 events (see Figs. 2 and 3). Moreover, community compositions 13 appear to subsequently return to their quasi pre-disturbance states (see Fig. 4A, B). However, 14 whereas the latter patterns are clear for both overall and planktonic communities in Lake 15 Ohrid as well as for overall communities in Lake Prespa, the return to the quasi pre-16 disturbance state in planktonic communities in Lake Prespa is less obvious (see the black 17 curve in Fig. 4B). Accordingly, neither lake underwent regime shifts. We, therefore, conclude 18 that lakes Ohrid and Prespa have a high ecosystem resilience. This is in contrast to findings 19 from some lakes where instability was hypothesized to increase susceptibility to regime shifts 20 (cf. Spanbauer et al., 2014). 21 However, the drivers for the resilience in lakes Ohrid and Prespa remain unclear at this stage. 22 They are likely multifactorial, involving parameters such as water depth, hydrological regime, 23 and chemical buffer processes. As the resilience of the lakes was indirectly inferred using 24 diatom communities as proxies, the results were likely also affected by intrinsic biotic 25 parameters of the diatoms. 26

Differential resilience in lakes Ohrid and Prespa 27
Given that ecosystem resilience has been demonstrated for both lakes, our second specific 28 objective was to investigate whether there were differences in resilience strength between the 29 two systems. As an operational criterion for resilience strength, we used the length of the 30 recovery periods (sensu Carpenter, 2013). Our working hypothesis was that Lake Ohrid is 1 more resilient to environmental disturbances than Lake Prespa. 2 Concluding from the length of the recovery periods, Lake Ohrid is more resilient than Lake 3 Prespa (ca. 1,100 years vs. ca. 4,000 years, respectively). The reasons for the differential 4 responses of the two neighbouring lakes remain less well understood (also see Wagner et al., 5 2010; Leng et al., 2013), but as discussed above, may be related to differences in their 6 geology, limnology, and lake age. 7

Limitations and outlook 8
We believe that the data and conclusions provided in the present paper are robust. The 9 analyses show that the diatom communities in both lakes recovered after major environmental 10 disturbances and that there are differences in recovery times between the two lakes. 11 Nonetheless, given the nature of our data, a number of limitations have to be noted. Firstly, 12 the resolution of the age models used and potential bioturbation may hamper the precise 13 estimation of community change above and below the actual tephra deposition. Additionally, 14 our findings are based on single core locations in lakes Ohrid and Prespa. Moreover, as 15 former littoral core sediments from Lake Ohrid were characterized by the presence of hiatuses 16 (e.g., Wagner et al., 2008; Vogel et al., 2010a), we had to use a core that was retrieved from a 17 greater water depth (see Fig. 1). This, in turn, resulted in a bias of the Ohrid communities 18 towards planktonic species. Finally, our study lacked high-resolution geochemical core data 19 for the timeframe of interest. 20 In order to mitigate these problems, we used relative time information (i.e., diatom zones) for 21 describing community changes, whenever possible. We focused in the comparative resilience 22 and recovery time analyses on changes in the planktonic communities, as they were directly 23 comparable in the two lakes (see black curves in Fig. 4). We also used previously published 24 Y-5 geochemical data, especially SiO 2 content in the tephra layers (Sulpizio et al., 2010). 25 Despite these limitations, the response curves for the planktonic diatom communities in Ohrid 26 and Prespa were similar. Differences mainly concerned the duration of the individual phases 27 of community response. We take this as another indication for the robustness of our data. 28 Nevertheless, given the interesting and partly unexpected patterns observed, we encourage 29 future projects that aim at studying resilience processes in lakes Ohrid and Prespa in more 30 detail. This would not only be of interest from a conceptual, but also from an applied point of 31 view relative to current and future human impact scenarios for these model lakes (e.g., 1 Kostoski et al., 2010). 2 In particular, we recommend high-resolution studies of more and/or other pulse and press 3 disturbance events (e.g., earthquakes, lake level fluctuations, orbital-suborbital climate 4 changes) in order to better understand the interplay of multiple disturbances. Given the 5 unexpectedly long recovery times found in this study, we also suggest studying post-6 disturbance patterns in higher resolution and over extended periods of time. 7

Conclusions 8
In the present study, we demonstrated that diatom communities in ancient lakes Ohrid and 9 Prespa reacted strongly to one of the most severe volcanic eruptions in the central 10 Mediterranean region during the Late Pleistocene − the Y-5 event (39.6±0.1 ka ago). After a 11 rapid initial response, community compositions slowly returned to their quasi pre-disturbance 12 states. In contrast to the Y-5 pulse disturbance event, signatures of the superimposed H4 press 13 disturbance event were less distinct. However, the latter likely contributed to the extended 14 recovery periods of > 1,000 years seen in both lakes. In the case of Lake Prespa, the H4 event 15 may have prolonged full recovery from the Y-5 pulse event until after the end of the H4. 16 Nonetheless, the data suggest that the communities in lakes Ohrid and Prespa likely did not 17 experience regime shifts (but see above for the complex pattern in planktonic communities in 18 Lake Prespa). We, therefore, conclude that both lakes show a high resilience to environmental 19 disturbances. However, the estimated recovery times, which can be used as measure for 20 resilience strength, differed between lakes Ohrid and Prespa (i.e., ca. 1,100 vs. ca. 4,000 21 years, respectively). This finding supports our working hypothesis that Lake Ohrid is more 22 resilient to environmental disturbances than Lake Prespa. The exact reasons for the 23 differential responses remain unknown, but differences in geology, lake age, limnology, as 24 well as intrinsic parameters of the diatom proxies may play an important role. 25 We do note some limitations of our study such as the resolution of the age models and the 26 different depths of the drilling locations, causing a bias towards planktonic species in Lake 27 Ohrid. Nonetheless, we believe that the results presented here are robust as indicated by 28 similar response curves for the overall communities in lakes Ohrid and Prespa. Yet, the curves 29 for the planktonic communities show no concurrence due to the complex response of Lake 30 Prespa. 31 We also believe that this study provides important new insights into the response of ancient 1 lakes to (multiple) environmental disturbances. Moreover, it contributes to one of the main 2 goals of the SCOPSCO deep drilling program -to evaluate the influence of major geological 3 events onto the evolution of endemic taxa in Lake Ohrid.  Age (ka) 5 15 25 35 45 55 65 75 85 95 5 15 5 15 25 35 45 55 65 75 5 15 25 5 15 5 15 25 35 5 5 5 5 5 5 5 5 5 5 5 5 5 20 40 60