Structural and functional development of twelve newly established floodplain pond mesocosms

Abstract Ecosystems are complex structures with interacting abiotic and biotic processes evolving with ongoing succession. However, limited knowledge exists on the very initial phase of ecosystem development and colonization. Here, we report results of a comprehensive ecosystem development monitoring for twelve floodplain pond mesocosms (FPM; 23.5 m × 7.5 m × 1.5 m each) located in south‐western Germany. In total, 20 abiotic and biotic parameters, including structural and functional variables, were monitored for 21 months after establishment of the FPMs. The results showed evolving ecosystem development and primary succession in all FPMs, with fluctuating abiotic conditions over time. Principal component analyses and redundancy analyses revealed season and succession time (i.e., time since ecosystem establishment) to be significant drivers of changes in environmental conditions. Initial colonization of both aquatic (i.e., water bodies) and terrestrial (i.e., riparian land areas) parts of the pond ecosystems occurred within the first month, with subsequent season‐specific increases in richness and abundance for aquatic and terrestrial taxa over the entire study period. Abiotic environmental conditions and aquatic and terrestrial communities showed increasing interpond variations over time, that is, increasing heterogeneity among the FPMs due to natural environmental divergence. However, both functional variables assessed (i.e., aquatic and terrestrial litter decomposition) showed opposite patterns as litter decomposition rates slightly decreased over time and interpond differences converged with successional ecosystem developments. Overall, our results provide rare insights into the abiotic and biotic conditions and processes during the initial stages of freshwater ecosystem formation, as well as into structural and functional developments of the aquatic and terrestrial environment of newly established pond ecosystems.


| INTRODUC TI ON
Ecosystems are characterized by abiotic and biotic components that are linked by multiple processes and through nutrient cycles and energy flows. These systems are dynamic and developing entities controlled by external and internal factors. Mature (climax) ecosystems, developed over long-term periods, are often complex and relatively stable systems. In contrast, ecosystems at their initial developmental stages are far less complex and stable, with fewer biotic and abiotic components interacting, and characterized by high levels of stochasticity (Buma et al., 2019;Fath et al., 2004).
The period between the start of ecosystem development (i.e., the "point zero") and the quantitative establishment of a first dynamic equilibrium of element cycling can be defined as initial ecosystem stage (Elmer et al., 2013;Schaaf et al., 2011). Such initial systems are more homogenous and less structured than mature ecosystems, which are the results of previous ecosystem succession (Elmer et al., 2013;Odum, 1969). Generally, succession as an important concept in ecology encompasses the directional and continuous occurrence of a range of successional sequences within an ecosystem over varying time scales (Begon et al., 2006). With ongoing succession, the ecosystem complexity increases, with "opportunist" pioneer, shortlived species colonizing the ecosystem during initial stages (Connell & Slatyer, 1977), which are replaced at later stages by new erratic species with more passive colonization abilities (Ruhí et al., 2013). In addition to changes in community composition, also biomass, interactions, respiration, and information (e.g., genetic diversity) increase during ecosystem development, whereas entropy decreases (Fath et al., 2004;Odum, 1969).
Natural ponds as abundant and ecologically important freshwater ecosystems in Europe and globally (Downing et al., 2006;Kristensen & Globevnik, 2014) have been in the focus of succession research for more than a century (e.g., Allee, 1911). One prominent example is the Pinkhill Meadow pond complex, consisting of approximately 40 permanent, semipermanent and seasonal ponds in the River Thames Floodplain, Oxfordshire, UK. Selected ponds of this complex have been monitored since their construction in 1990 (Freshwater Habitats Trust, 2021). However, the knowledge on the abiotic conditions, biotic communities, and initial ecological processes, including their drivers, of newly established pond sites is generally limited (Miguel-Chinchilla et al., 2014;Williams et al., 2008). It is rarely possible to study the "point zero" of ecosystem development and primary succession in newly established ecosystems under natural conditions (Schaaf et al., 2011). Most ecological research is conducted in established, climax pond ecosystems, without investigating the abiotic and biotic conditions and processes that led from the initial stages to the development of mature systems (Raab et al., 2012). Also, available information on key environmental factors, such as physico-chemical water quality parameters and seasonality, are limited leaving a knowledge gap in what is driving primary succession and ecosystem development in newly established ponds (Williams et al., 2008). For this reason, monitoring of pond ecosystem development from the "point zero" provides valuable information on factors driving ecosystem structuring and functioning at initial development stages.
The colonization and subsequent community development during primary successional stages in new pond habitats is determined by stochastic primary effects, local environmental conditions, and biological interactions (Buma et al., 2019;Weidlich et al., 2021;Wiegleb et al., 2017). Along the initial structural developments, also ecosystem functioning evolves driven by the functional traits of the organisms present. The differences in aquatic taxa composition directly affect ecosystem processes and consequently functioning (de Bello et al., 2010;Herbert et al., 2016). Ecosystem functions such as litter breakdown (Gessner & Chauvet, 2002) also denote suitable indicators for the primary successional development and integrity of young emerging ecosystems.
In this study, we conducted an ecosystem primary succession monitoring considering a variety of physico-chemical parameters, aquatic and terrestrial taxa as well as variables to assess ecosystem functioning. This was done to investigate pond ecosystem developmental trajectories during 21 months from their establishment. The monitoring was conducted in a system of twelve natural floodplain pond mesocosms (FPM 1-12, natural and undisturbed lotic freshwater ecosystems planned and constructed at a research site and operated without anthropogenic manipulation) that have been established in 2017 at the Eußerthal Ecosystem Research Station (EERES; see https://www.uni-koble nz-landau.de/de/landa u/fb7/ umwel twiss ensch aften/ eeres) in south-western Germany. The FPMs used here represent systems at "point zero" of ecosystem development and enable repeated sampling and analyses of initial development stages in a fully replicated way. We used ordination methods (i.e., principal component analyses [PCA] and redundancy analyses [RDA]) to analyze pond environmental conditions and interpond variations during initial successional stages.
Besides insights into primary succession and initial pond ecosystem developments, our study contributes to the knowledge of experimental mesocosm studies used to disentangle stressor effects from other confounding factors (e.g., Finnegan et al., 2018;Hua & Relyea, 2014), as aquatic communities of these systems also generally are in very early stages of succession. Insights into primary succession of pond mesocosms, factors driving aquatic community development, and lotic ecosystem development trajectories as gained here may benefit the interpretation of findings from other studies focusing on artificially stressed mesocosm experiments. The objectives of this paper are (1) to monitor and describe the ecosystem development of

T A X O N O M Y C L A S S I F I C A T I O N
Agroecology; Biodiversity ecology; Ecosystem ecology; Population ecology twelve FPMs using 20 physico-chemical and biotic parameters; (2) to disentangle and analyze factors driving the development of environmental conditions within the pond ecosystems; and (3) to ascertain interpond variations (i.e., differences among ponds by natural heterogenization) over time.

| Study site and FPMs
This study has been conducted at the EERES located in Rhineland-Palatinate, south-west Germany (49°15′14″N, 7°57′42″E). The EERES site is situated in a small valley of the Franco-German Palatinate Forest-North Vosges UNESCO biosphere reserve, which is an extensively forested low-mountain range. In 2017, twelve FPMs have been constructed at the EERES site (see Figure S1 for an overview). Each FPM is connected to a small stream called Sulzbach via controllable inlets and overflows. The inlets, except for rare events (see below), were kept closed during the study period preventing the water from the Sulzbach to mix with pond water. The Sulzbach belongs to the stream type "small, fine-substrate dominated, siliceous highland rivers" (Westermann et al., 2011). The upstream catchment of the Sulzbach has only minor anthropogenic influences and the stream is characterized by a high structural and ecological quality.
Each FPM has a dimension of 23.5 m × 7.5 m with V-shaped banks at three sides and a flat water-land floodplain area at the inflow ( Figure   S2a). The depths of the FPMs increase towards the outflow area (bed slope 1:20), providing a maximum water storage level of 1.5 m.
The banks and the bottoms are impervious to water due to a PVC membrane covered by a sand layer of 15 cm; an additional 0.5 m gravel layer (uniform gravel diameter of >2 cm) covers half of each FPM pond bottom toward the outflow area ( Figure S2a). The upper half of the banks was seeded with a standard grass mixture (Festuca rubra ssp., Festuca trachyphylla, Poa pratensis) in October 2017 to prevent erosion. After flushing the newly constructed FPMs twice with water from the Sulzbach, the ponds were filled with Sulzbach water on 04 November 2017 to a water level of 30 cm at the outlet sites ( Figure S2b). This water level was kept constant at 30 cm (±10%) during the entire study period, meaning that losses due to evaporation were compensated by refilling with water from the Sulzbach.
However, refilling of the FPMs was needed rather infrequently (i.e., approximately every 2-3 months depending on meteorological conditions). During the refilling, in order to ensure reproducibility, the inlets of all FPMs were always opened at the same time and for the same brief (i.e., approximately 2 h) duration. Given that the volume of the inflow water was small compared with the water volume of the ponds and that all FPMs received water from the same stream, we estimate the influence of the stream water on the abiotic environmental conditions of the FPMs as marginal. Apart from refilling the ponds, no water inflow into the FPMs occurred. The FPMs thus represent small lentic water bodies with water-land floodplain areas.
It has to be noted that the water from the Sulzbach was not sterilized before it was used to initially fill or to refill the ponds. However, during the entire experimental phase, no anthropogenic manipulation or management of the twelve FPMs took place to allow natural primary succession and undisturbed development. The EERES research station is completely fenced off to avoid disturbances by human visitors or larger wild animals. Since October 2019, the FPMs are a key research site of the DFG Research Training Group 2360 "SystemLink" (https://syste mlink.uni-landau.de).

| Overview of the monitoring program and sampling methods
A monitoring program was launched immediately after the establishment of the FPMs. The initial ecosystem development and primary succession of the twelve FPMs from a defined "point zero" was monitored for 21 months from 06 November 2017 until 31 July 2019. Overall, 20 variables were surveyed: four parameters describing the general conditions at the EERES site and of the FPMs; five physico-chemical water quality parameters; nine biotic parameters; two functional parameters. Together, the 20 variables surveyed describe both the aquatic and adjacent terrestrial FPM ecosystem development (Table S1). Because of the high workload, it was not feasible to achieve the same temporal coverage and sampling dates for all parameters during the monitoring; not all parameters were surveyed regularly and during the entire study period. Specifically terrestrial species (i.e., terrestrial vegetation, ground beetles, leafhoppers) were assessed at a few distinct sampling dates only (see details on sampling design below). For this reason, absence of values in the figures and tables correspond to absence of data. During sampling campaigns, all sampling and measurement equipment was completely rinsed with tap water after usage in one FPM to prevent the transfer of organisms or biological material between the ponds.
In addition, all sampling and measurement equipment (e.g., polyethylene mesh bags, data logger) were leached for at least 24 h in tap water (with at least three water exchanges) before it was placed in the FPMs in order to prevent leaching of chemicals into the water.
The information on meteorological conditions, soil temperature, and water level fluctuations can be found in the Appendix S1 (supporting results). To track visible real-time changes during the FPM development, we recorded a time-lapse video of FPM9 using a daily camera trap (Bushnell Trophy Cam HD) (available at https://youtu. be/vfTNp wyYhFE).

| Physico-chemical water quality parameters
Water samples were taken monthly or bimonthly in each FPM 5-10 cm below the water surface using high-density polyethylene bottles. Water samples were analyzed by an accredited laboratory applying standard analytical methods (AGROLAB, Germany) for the following components: total organic carbon (TOC), dissolved organic carbon (DOC), fluoride, chloride, nitrite, nitrate, phosphate, sulfate, ammonium, sodium, potassium, calcium, and magnesium. The limits of quantification (LOQ) are given in Table S5.
Please note that due to analytical problems (i.e., equipment failure), not all components could be analyzed at all sampling dates.
Detected concentrations (i.e., concentration > detection limit) less than the LOQ were replaced by 0.5 × LOQ for data evaluation and visualization.
Specific conductivity and pH were measured monthly or weekly 5-10 cm below the water surface near the outlet sites of the FPMs using a portable analytical device (WTW Multi 3630 IDS). No data for conductivity are available for January 2019.
Water temperature and dissolved oxygen were recorded by a miniDOT ® logger (miniDOT ® USB Oxygen Logger 7392) installed 5 cm above the bottom gravel layer (i.e., at 25 cm water depth) near the outlet site of each FPM. The miniDOT ® logger recorded the water temperature and dissolved oxygen with a resolution of 10 min; daily averages of water temperature and dissolved oxygen were used for evaluations. Monitoring results for all physicochemical water quality parameters are presented in the Appendix S1 (supporting results).

| Aquatic macroinvertebrates
Benthic macroinvertebrates were sampled 15 times over the sampling campaign using pebble baskets as artificial substrates (see e.g., Brock et al. (1992) for details). Pebble baskets consist of a polyethylene mesh (length 15 cm; width 15 cm; height 7.5 cm) formed as a pyramid and were filled with the bottom substrate (gravel with a diameter of >2 cm) of the FPM. While the upper part of the pebble basket was made of a coarse mesh (2 cm aperture) to enable easy colonization, the lower section was made of a fine mesh of 0.5 cm aperture to prevent escaping of macroinvertebrates during sampling. In each pond, a pebble basket was placed at the bottom near the inlet and outlet sites, as well as in the middle of each FPM, given a total of three baskets per FPM. Before animal collection, baskets were left for colonization for three weeks. At the end of each sampling campaign, the pebble baskets were gently retrieved from each FPM with a dip net (500 µm mesh aperture). The macroinvertebrates present on the pebble basket substrates were removed and preserved in ethanol (70% vol.). All individuals were subsequently identified to the family level.

| Merolimnic insect emergence
We sampled the emergence of merolimnic insects from April to July 2019 using three floating pyramidal tents with a basal area of 0.25 m 2 as emergence traps (Cadmus et al., 2016). The emergence traps were made of nylon mesh and anchored to the pond bottom.
Emergence traps were installed from 24 April 2019 until 31 July 2019 and emptied once per week, that is 15 times in total. Three emergence traps per pond were placed at the water surface at locations comparable to the pebble basket sites. Emerged insects were collected in plastic bottles connected to the top of the trap and filled with ethylene glycol to preserve insects. Upon collection, aquatic insects were stored in ethanol (70% vol.) and subsequently identified to the family level. All results of the merolimnic insect emergence monitoring are available in the Appendix S1 (supporting results).

| Zooplankton
Zooplankton was sampled monthly in each FPM with a plankton net (20 cm diameter; 65 µm mesh aperture). Overall, 15 samples were taken in all FPMs during the study period, with no samples available for January and February 2018. For sampling, the plankton net was moved gently across 10 m horizontally through each FPM at a distance of approximately 1 m from the banks. After retrieval, samples were rinsed through a net (mesh aperture 120 µm) and stored in ethanol (70% vol.). Zooplankton abundances (individuals >250 µm) were assessed by the Institute for Fishery of the Bavarian State Research Center for Agriculture using the ZooScan V4 system (HydroptiC, version 2.4.0), a digital zooplankton image analysis (Gorsky et al., 2010). Please note that zooplankton individuals <250 µm could not be assessed by the ZooScan digital image analysis.

| Amphibians (tadpoles) and submerged vegetation
Amphibians and submerged vegetation, that is, filamentous algae and Elodea spec., were assessed by visual inspection of the per-

| Crayfish and fish
The presence of crayfish, which may dominate ecosystem development (e.g., Reynolds et al., 2013), in the FPMs were monitored using three baited standard crayfish traps (type: "Pirate") per FPM.

Overnight monitoring was conducted monthly between April and
October in 2018 and between March and July in 2019. The presence of fishes was also checked in the crayfish traps, as well as by visual inspection during each visit of the FPMs. No crayfish or fish were found during the entire study period.

| Ground beetles
Ground beetles were collected by pitfall traps (Schirmel, 2020) during seven sampling campaigns conducted in late spring and autumn 2018, as well as in late spring 2019. One pitfall trap was placed at the northern and southern bank of each FPM and two pitfall traps at a distance of two meters from the shore into the terrestrial part of each floodplain area. The pitfall traps consisted of a plastic cup (6.5 cm diameter) filled with ethylene glycol diluted 1:3 with water and a drop of detergent. At the end of each sampling campaign, the pitfall traps' contents were sieved and the ground beetles were stored in ethanol (70% vol.). Ground beetle individuals were subsequently identified to species level.

| Leafhoppers
Leafhoppers were surveyed on 26 June and 21 September 2019 at the northern and southern banks of FPM3, FPM5, FPM8, FPM10.
Leafhopper collection was conducted by suction sampling with a modified leaf blower (SH 56 Stihl, Waiblingen, Germany) using 80 suction pulses per sampling (Kormann et al., 2015). The leafhopper catches were transferred into a bucket and subsequently separated from the rest of the catch with an aspirator. Leafhopper samples were then stored in ethanol (70% vol.) and identified to species level.

| Aquatic and terrestrial litter decomposition
Aquatic leaf litter decomposition was approximated using the litterbag method (Benfield et al., 2017). In brief, coarse-(10 mm aperture) and fine-mesh (1 mm aperture) litterbags were filled with approximately 4 and 2 g (weighed to the nearest 0.001 g), respectively, of oven-dried (at 60°C for 24 h) alder leaves (Alnus glutinosa (L.) Gaertn.) (Voß et al., 2015). Fine-mesh and coarse-mesh bags were employed in the FPMs to separate the microbially and the shredder-mediated share of leaf litter decomposition. One coarse-mesh and one finemesh litterbag were submerged near the inlet and outlet sites, as well as in the middle part of each FPM for 20-25 days at nine dates between May 2018 and June 2019. After retrieval, the remaining leaf material was gently rinsed under running water to remove mineral particles and macroinvertebrates, oven-dried at 60°C for 24 h and reweighed to the nearest 0.001 g.
The litter decomposition rate k was calculated for each FPM and the respective sampling dates using the following formula (Benfield et al., 2017): where S(t) is the leaf mass as a function of deployment time t and S (0) is the initial mass of the coarse-and fine-mesh litterbags, respectively. The mass data from the fine litterbags were used to calculate microbial decomposition (k microbial ). For the calculation of the litter decomposition by shredders (k shredders ), S(t) of the coarse litterbag at a given sampling site was corrected for the mass losses due to microbial decomposition by adding to S(t) the amount degraded over time in the coarse litterbags (i.e., S(0) − S(t)) multiplied by the mean percentage microbial decomposition of the respective fine litterbags. For data evaluation, decomposition rates were averaged across the three sampling locations of a given pond.
The terrestrial litter decomposition was assessed using the tea bag approach (Keuskamp et al., 2013). In brief, one green tea and one rooibos tea bag (brand "Lipton") were buried pairwise at a depth of five cm at the middle of each the north and south banks (50 cm above water surface) of each FPM for approximately 21 d at twelve dates during the study period. After retrieval, adhered soil particles were removed and the filling of the tea bags was oven-dried for 24 h at 60°C and weighed to the nearest 0.001 g. The terrestrial litter decomposition was calculated as linear weight loss per day (mg/d) for both green and rooibos tea as follows: where M R is the reference mass derived from averaging the weight of ten green and rooibos tea bag fillings, respectively, and M t is the mass of the green and rooibos tea bag fillings after deployment time t. The mean value of the north and south bank sampling sites of a given FPM was calculated separately for green and rooibos tea bags, respectively, and used for data evaluation.

| Generalized additive models for temporal trend analyses
To visualize and assess overall temporal trends in biotic communities and ecosystem function, we fitted generalized additive models (GAM) using the mgcv R package (Wood, 2011

| PCA and RDA of pond ecosystem developments
We used PCA and RDA ordination analyses to analyze aquatic eco-  (Olsen et al., 2012;Ramette, 2007). Using multiple linear regression, RDA extends PCA by explaining variation between independent and dependent variables within an iterative process to find the best ordination (Ramette, 2007 (Table S7).

| Submerged vegetation
The percentage areal coverage of submerged vegetation (i.e., algae and  Table   S7). Apart from FPM3, algae and Elodea spec. vegetation coverage of the ponds aligned toward the end of the study period (Figure 1d).

| Terrestrial vegetation
During the first survey in November 2018, 53 grass and plant species were found at the 12 FPMs (see Table S8 for a full list of species).  (Table S8)

| Leafhoppers
Overall, 19 leafhopper species and 910 individuals were sampled in 2018 around FPM3, FPM5, FPM8, and FPM10 (please see Table   S10 for a full list of species). The species numbers did not differ much between the four FPMs; however, the number of individuals was twice as high at FPM3 (n = 326) compared with FPM8 (n = 163) (Table S10).  (Table S11).

| Aquatic and terrestrial litter decomposition
Nevertheless, larger differences in decomposition rates of the different FPMs existed particularly during the spring and summer months; differences between the ponds were moreover larger in 2018 and less pronounced in 2019 for both k microbial and k shredder (Figure 3a,b).
Concerning terrestrial litter decomposition, weight losses per day were higher for green tea compared with rooibos tea. However, the terrestrial litter decomposition for all ponds showed a comparable pattern over time for the two different tea bag types, with significant (p < .001) overall temporal trends for both parameters identified by GAM (Figure 3c  Terrestrial litter decomposition showed generally largest interpond variations during summer and early autumn months, but were overall comparable across the study period (Figure 3c,d; Table S12).
Particularly for green tea, the interpond variation was higher in 2018 compared with 2019.

| Statistical evaluation of pond ecosystem developments over time
Two PCA axes were selected by the broken stick approach, explaining a total variance of 48.2%. The first PCA axis, which explains 30.6% of the total variance, has significant loadings (correlation > .60) on specific conductivity, potassium, DOC, fluoride, as well as submerged vegetation and pH. The second PCA axis explains 17.6% of the total variance and correlates to sodium and sulfate, as well as to phosphate (Table S13).
For the entire study period, no linear pattern of pond environmental condition was identifiable along either the first or second PCs biplot for the second year shows less effects of the seasons in driving overall differences in environmental conditions, while environmental differences among ponds appear to become more influential ( Figure   S9b). In the same PCA ordination space, when the pond is selected as grouping factor, overall no large differences in environmental conditions are visible among ponds if the entire study period is considered ( Figure S10a). Also in this case, when the PCA is constrained to the first and second years, a high similarity of pond environmental conditions is shown until summer 2018 ( Figure S10b), whereas differences in environmental conditions among ponds increase substantially in the second year ( Figure S10c). This also confirms a lower interpond environmental diversity within seasons (i.e., less seasonal effects) in the second compared to the first year. A further PCA shows that the pond location (i.e., left, middle, right) along the experimental site (see Figure S1) had no effect on environmental conditions of the ponds (data not shown).
The results of the RDA and VPA overall confirm patterns observed in the PCA ordination. RDA and VPA show that for the entire study period successional time (i.e., time since the beginning of the study, VPA explained variance = 36.1%) and its combined effect with seasonality (VPA explained variance = 14.8%) have the largest effect on the environmental conditions of the FPMs (Table 1; Table   S14). Similarly to what was observed in the PCA, when the RDA is performed exclusively on the first year of ecosystem development, it identifies season (VPA explained variance = 22.6%) and its interaction with successional time (VPA explained variance = 38.7%) to explain most variance in the environmental conditions of the FPMs (Table 1), whereas the factor pond had no significant influence (Table   S15). On the other hand, when the RDA is performed on the second year, analysis identifies pond as significant factor explaining the highest proportion of variance (VPA explained variance = 33.5%) in environmental conditions, followed by the interaction of seasons with successional time (20.8%; Table 1; Table S16).

| Pond ecosystem developments and environmental conditions
Monitoring of the first 21 months of initial pond development and primary succession revealed dynamic ecosystem successional trajectories in the twelve FPMs (see also Figure S2 and the time-lapse video available at: https://youtu.be/vfTNp wyYhFE). Data from the TA B L E 1 VPA results for the factors successional time, season, and pond for the entire study period, as well as for the first and second year of the study The overall initial ecosystem development from the "point zero" observed here for the FPMs is in line with pond studies under field conditions. Similar developmental patterns were, in fact, also observed at a larger scale for the Chicken creek catchment in eastern Germany . Also, our results are comparable to the findings from the Pinkhill Meadow pond complex in the UK, in which a rapid pond colonization and ecosystem development was observed within the first years after construction (National River Authority, 1992).
Fluctuations of physico-chemical water quality parameters over time (Appendix S1, supporting results) likely resulted from effects caused by seasonality (effects on e.g., water temperature), meteorology (effects of precipitation on e.g., conductivity), and biology (e.g., effects of photosynthesis by aquatic vegetation on pH and dissolved oxygen concentrations). Overall low inorganic ion concentrations are related to the fact that ponds have been entirely filled with water from the Sulzbach, a nutrient-and ion-poor stream. Also, interpond variability of physico-chemical parameters could be explained by differences either in submerged vegetation coverage (pH values; dissolved oxygen) or differences in soil entries caused by bank erosion (e.g., specific conductivities; ion concentrations). Intersystem variability of physico-chemical parameters has also been observed in previous studies with replicated outdoor pond mesocosms (e.g., Caquet et al., 2001;Christman et al., 1994). However, care must be taken in interpreting the dissolved oxygen concentration data particularly of FPM9-12 as there are larger data gaps due to O 2 logger failures (i.e., defective devices). findings, comparable to those found in the Chicken creek catchment study (Elmer et al., 2013;Hüttl et al., 2013), confirm that while external factors dominate the very initial phase of ecosystem development, internal interacting abiotic and biotic factors gained importance in later stages. Increases in interactions and feedback, accompanied by, among other, an increase in biomass and species diversity (see below), are clear indicators of ecosystem development and succession (Fath et al., 2004;Odum, 1969).

| Pond colonization and aquatic and terrestrial community development
Data from our study show that new, undeveloped ponds are colonized quickly, that is, zooplankton and invertebrate species were recorded in six ponds already at the initial sampling dates ( This likely is a result of ecosystem divergence (see discussion of PCA and RDA results above), priority effects after the stochastic initial phase (Weidlich et al., 2021), as well as increasing biotic-abiotic interactions and feedback in the ponds during their advancing ecosystem development. Street and Titmus (1979) showed for macroinvertebrate communities of six newly constructed experimental ponds that interpond taxa variability was driven by increasing variability of pond environmental conditions and habitat structure (e.g., differences in aquatic macrophyte abundances) within two years of observation. Also, the importance of stochastic sequences of species arrival and subsequent priority effects, in which the initial species assemblages partly determines future community developments, has been largely proven for the colonization of new pond habitats (Chase, 2003;Louette & De Meester, 2007); priority effect-induced long-lasting differences in species dominance and overall community assemblages may thus lead to future increases in taxa variability among FPMs. Long-term pond monitoring in the UK showed larger variations of plant richness and macroinvertebrate richness in ponds after seven years compared to the first year of construction (Williams et al., 2008). Overall, as the FPM communities mature, individual ecosystem properties including abiotic and biotic interactions are expected to increase further in influence over time (Frisch et al., 2012), potentially resulting in further increase of interpond taxa variability.

| Pond ecosystem development and ecosystem functioning
Decomposition of plant litter denotes an important ecosystem function for the provisioning of energy and nutrients in aquatic and terrestrial environments (Benfield et al., 2017;Krishna & Mohan, 2017 (Martinez et al., 2014). However, in comparison to k microbial , k shredder was substantially less influenced by seasonal effects, which is possibly explained by the year-round presence of macroinvertebrates in the FPMs.
In contrast to the increase of numbers of aquatic species and abundances in subsequent seasons, a season-specific comparison shows that the ecosystem function leaf decomposition decreased over time, which is most pronounced for k microbial ; slightly lower water temperatures in spring and summer 2019 compared to 2018 (Appendix S1, supporting results) might be an explanation.
Interestingly, in contrast to what we found for all other abiotic and biotic variables (see above), the differences in leaf litter decomposition rates between the ponds also decreased over time. It follows that this ecosystem function converged with progressing ecosystem developments of the FPMs. One reason for the taxonomical divergence but functional convergence observed here for the second year might be that later arriving species are not yet able to fully compete with species already established in the FPMs after a largely stochastic initial colonization and niches (i.e., energy sources such as litter decomposition) partitioned already between the initial colonizers.
Data from the terrestrial parts of the FPMs in December 2017 revealed terrestrial litter to be efficiently decomposed already under initial ecosystem conditions. Moreover, terrestrial litter decomposition did not increase over time when comparing the respective seasons. This is in line with our findings on aquatic leaf litter decomposition and once again in contrast to our taxonomical findings on the structural ecosystem development, which showed a season-related increase in aquatic and terrestrial species presence and abundances between the FPMs. That said, ecosystem functions such as aquatic and terrestrial litter breakdown are, however, effective already in the very initial stages of freshwater ecosystem development, whereas ecosystem structure (i.e., diversity and abundance of different taxa) gradually increases over time. However, terrestrial litter decomposition was also strongly influenced by seasonal and meteorological effects, with higher soil temperature and soil moisture fostering an increase in litter decomposition due to elevated microbial activity in warmer months (Krishna & Mohan, 2017). Regarding the latter, the prolonged dry period in July 2018 (Appendix S1, supporting results) may explain the substantial reduction in litter degradation at the beginning of August 2018. The faster decomposition of green tea compared with rooibos tea is well known from the literature and explained by differences in litter composition, particularly cellulose and lignin contents (Didion et al., 2006;Keuskamp et al., 2013).

| CON CLUS I ON AND OUTLOOK
In our study, the initial freshwater ecosystem development and primary succession could successfully be observed at the EERES site.
Using an array of twelve FPMs denotes a novel and promising approach to observe and disentangle abiotic and biotic conditions including structural and functional ecosystem developments as well as their complex interactions. The FPMs at the EERES site offer unique research opportunities to study small freshwater environments, ecosystem development, and water-land interaction under field conditions in a replicated and controlled manner. The ecological research possibilities at the EERES site thus differ from other ecosystem research approaches that either analyze mature ecosystems, or do not feature repeated designs or sites with clear boundary conditions and fully controllable research environments. The findings on ecosystem primary succession and initial ecosystem developments from a "point zero" presented here may benefit planning, implementation, and ecological evaluation of freshwater restoration measurements, which are widely implemented in Germany and Europe (see e.g., http://www.europ eanpo nds.org; https://www.ecrr.org). Also, our monitoring data provide valuable information for the ecological assessment of temporary ponds with seasonal formation in river floodplain areas. Moreover, consequences from anthropogenic disturbances of natural pond ecosystems may become more predictable if fundamental mechanisms of initial ecosystem colonization are better understood. Our findings also support that creating small ponds denotes a simple and cost-effective tool to enhance landscapes by biodiversity-rich habitats within short periods of time; this is of particular importance considering the ongoing decline of global freshwater biodiversity (e.g., Albert et al., 2021;Tickner et al., 2020).
Last, our study shows that particularly interpond taxa variability among replicated outdoor pond mesocosms in early successional stages must be considered in case these aquatic systems are used to analyze responses to anthropogenic stressors, such as chemicals (Caquet et al., 2000(Caquet et al., , 2001. Although larger mesocosms such as the FPMs used here are more internally stable and sustainable, which reduces intersystem variability compared with smaller systems (Belanger, 1997;Caquet et al., 2000), larger experimental pond systems also need to be managed (e.g., inserting species into ponds; mixing water and sediments between ponds) to increase replicability between the experimental units. Writing -review & editing (equal).

DATA AVA I L A B I L I T Y S TAT E M E N T
All data of this study are provided in the main text and in the appendix.