Soil characteristics and soil moisture of a gradient from bare soil to vegetated site at grid points (a) K5 and (b) N5.
Landscapes and ecosystems are complex systems with many feedback mechanisms acting between the various abiotic and biotic components. The knowledge about these interacting processes is mainly derived from mature ecosystems. The initial development of ecosystem complexity may involve state transitions following catastrophic shifts, disturbances, or transgression of thresholds. We propose a conceptual framework of feedback processes in early states of ecosystem development affected by spatiotemporal environmental drivers. To test this concept, we used 10-year time series of hydrological, biological, geomorphological, and soil data from the constructed catchment Chicken Creek.” The 6ha site was left to unrestricted development since 2005 and was intensively monitored. The data showed a very rapid development of the site with an increasing complexity and heterogeneity. In the first years, stochastic signals like the initial substrate conditions and external drivers like extreme weather events were the most important factors resulting in abiotic/abiotic feedback mechanisms shaping the morphology of the site and creating site diversity. Initial abiotic feedback mechanisms between water and substrate were soon followed by abiotic/biotic feedbacks between biological soil crusts, invading vegetation, geomorphology, and hydrology resulting in state transitions of catchment functioning.
- state transition
- feedback mechanisms
In 2001, the US National Academy of Sciences defined the understanding of the so-called Critical Zone as one of the most challenging central research topics. The Critical Zone is the near-surface part of the Earth’s crust, which sustains all terrestrial life . During the last decade, a large number of landscape observatories have been implemented in order to study the Critical Zone and its behavior under global change condition. First Critical Zone Observatories (CZOs) were established in the USA; later, additional sites have been launched all over the world . This paper introduces one of these CZOs in Germany with very unique conditions. The site represents a constructed watershed designed to analyze the ecological feedback mechanisms during the initial development of an ecosystem.
Ecosystems have been shown in many studies to be utterly complex systems with many feedback mechanisms interacting between compartments [3, 4]. However, most of these studies were carried out in mature systems that had evolved over long time periods (e.g., [5, 6, 7]). Therefore, very little is known about how this complexity evolves over time and what role these feedbacks play during the development of ecosystems . Most knowledge about early states of ecosystems was derived either from case studies after natural or anthropogenic disturbances (e.g., [9, 10, 11, 12, 13, 14]) or from chronosequence studies (e.g., [15, 16, 17, 18]).
In recent years, landscape development and ecosystem development have been viewed as complex interacting processes leading to self-organization and state transitions [19, 20]. Within this context the concept of multiple stable states depending on environmental settings and on the possibility for catastrophic shifts in ecosystem composition and functioning, especially after disturbances or after passing ecological thresholds, was created [21, 22, 23, 24, 25, 26]. Unfortunately, most of the data supporting these concepts were recorded only post-event.
The role of feedback mechanisms for the development of ecosystems and their functioning may differ depending on internal dynamics and external drivers. To better understand the underlying mechanisms, we developed a simple conceptual framework of how feedbacks between the major system components substrate, water, and biota interact and are affected by stochastic spatiotemporal drivers. Studies from arid environments indicate that ecosystems under water limitation may display alternative stable states, in which the interactions between abiotic and biotic processes determine whether or not a degraded state will prevail after disturbance . The interaction between substrate and water is clearly driven by rainfall and evaporative demand, in which soil particles are mobilized on the surface (erosion) and within the substrate (soffusion) and subsequently immobilized during dry periods in which, for example, seals are transformed into physical surface crusts irreversibly changing runoff properties. Such interactions represent abiotic feedbacks that can either yield into self-stabilizing surface structure or a continuously changing erosive surface. At the same time, the formation of biological crusts and the germination of seeds will generate feedback processes that will affect substrate properties. In more humid and temperate climates, eventually feedback mechanisms between abiotic components and the biota will dominate the system, which was shown in many ecosystem studies.
To study the role of feedbacks in state transitions during ecosystem development, we used 10-year time series of data from the constructed catchment Chicken Creek. This unique site offers the chance to observe state transitions in a relatively simple ecosystem from a very initial state with hardly any internal ecological memory  to more complex states.
2. Material and methods
The Chicken Creek (“Hühnerwasser”) catchment is located within the lignite mining district of Lusatia in Northeastern Germany, about 150 km southeast from Berlin. It was constructed in 2004–2005 in the lignite mining area of Welzow-Süd, 30 km south of Cottbus, Germany (Figure 1) . The region is characterized by temperate seasonal climate (563 mm mean annual precipitation, 8.9°C mean annual air temperature). The catchment was constructed by means of large mining techniques to establish a headwater for the Hühnerwasser, a small stream, which was destroyed by mining activities in the 1980s. This stream has to be restored during the reclamation process of the post-mining landscape. The site consists of two layers and has an inclination of 2.0–3.5% and a southeastern exposition. At the lowest part of the site, a basin was formed allowing the formation of small pond (originally about 60–70 m in diameter and 3 m depth in 2005/2006). Details of the construction process and of the substrates used are described .
The construction started with a 1–2-m-thick clay layer 8 (Figure 2a), which forms an aquiclude underlying the whole area of 60,000 m2 (400 × 150 m). The material was separated from the tertiary overburden layers by bucket wheel excavators and dumped by a stacker. The clay surface was then leveled by bulldozers – but not compacted, as this material tends to self-seal when considerable swelling occurs. Initially, the freshly dumped clay consisted of large aggregates, but these vanished after wetting and subsequent swelling. In this state, the clay layer had an extremely low permeability (ksat. ~10−9 m s−1). The clay layer was shaped into a shallow basin ascending from the center to the edges to form the subsurface boundaries of the catchment (Figure 2b). Belowground in the lower part of the catchment, additional clay dams were constructed on top of the clay layer perpendicular to the slope as a stabilization barrier to prevent the sandy substrate of the aquifer from sliding downhill on the clay layer and as a central groundwater discharge unit for the creation of an artificial spring for the reconstructed creek. A clay wall at the southern edge of the pond defines the lower boundary of the catchment and has a single defined outlet (Figure 2c).
On top of this clay layer, 117,500 m3 of Pleistocene sandy material taken from the forefield of the open-cast mine (i.e., mainly C horizon substrates from the former landscape) was dumped to form the 2–3 m aquifer of the watershed (Figure 2a). The construction of this layer began in August 2004 with the eastern area of the hillside adjacent to the later hydrological catchment, followed by the central parts of the catchment during the next 2 months. The area of the catchment generally consists of three different sections which can be distinguished in terms of the overall construction procedure. The sandy material in the eastern part was dumped in August/September 2004. During the next construction phase in September/October 2004, the western part of the clay layer was completed and immediately covered with sandy substrate material. The central part of the site was left open as a “central trench” for a period of 7 months, before finally being filled in by bulldozers in May 2005 with substrate material from the eastern and western sections (Figure 2b). The surface level of the eastern part of the catchment was lowered in order to remove surface substrate that had been exposed to the atmosphere for more than 1 year and to restore the surface to an initial state. The surface layer was flattened and shaped into a shallow basin in order to define clear hydrological boundaries at the surface. As a final step, the surface of the sandy layer was homogenized, and the remaining surface structures from the construction were removed as good as possible.
The hillslope-shaped site with defined boundary conditions and well-documented inner structures allows for studying ecosystem development ab initio at the catchment scale (Figure 2d). No amelioration measures, fertilization, or planting was carried out. Since 2005, the unrestricted, unmanaged development of the catchment was intensively monitored. Sensors and monitoring plots were originally oriented along a regular 20 × 20 m grid and were successively complemented with more structure and pattern-oriented instrumentation adapted to the development of the catchment (Figure 3). In total, 3 flumes in the main erosion gullies and 2 weirs equipped with automated sampling devices, 3 weather stations, 42 groundwater wells, 9–18 deposition samplers, 16 suction plates, 88 FDR probes, 40 pF meter, and 1 multiparameter probe in the pond are installed. Vegetation is surveyed every summer at four 1 × 1 m plots around each of the 119 grid points. More details on installed sensors and analytical methods are described [31, 32].
2.1. Results and discussion
Initial soil sampling from all 120 grid points revealed some slight spatial variation of the Pleistocene material with respect to texture and chemical parameters reflecting the natural variability of these postglacial deposits. The western part of the catchment had more loamy sands, whereas in the eastern part, pure sands dominated. The vertical distribution was very homogenous. Due to the carbonate contents of the material (0.6–1.1%), pH values were uniformly between 7 and 8. Organic carbon (Corg.) content was very low (1–2 mg g−1).
Time series of meteorological, hydrological, biological, and soil data revealed a fast colonization of the catchment with invading vegetation that transformed the site from an initially abiotic system similar to arid systems to a state where abiotic/biotic feedback processes dominate [33, 34]. Aerial images of the catchment documented this rapid development and showed the increasing heterogeneity and the formation of surface structures and patterns (Figure 4). From 2010 woody plants increasingly formed clearly visible patches particularly in the eastern part of the catchment and in the surroundings of the pond. In 2015, more than 30 tree species were detected in the catchment.
Total vascular plant cover reached 58% in 2015, composed of over 170 plant species . During the first 2–3 years, surface runoff, gully erosion, and sediment translocation were the dominating processes at the sparsely vegetated site that were triggered by single episodic events like heavy thunderstorms . The slight textural variations in the initial substrates resulted in differences in morphological features of these surface structures, e.g., the frequency, depth, and width of erosion channels . Surface runoff was promoted by the development of mechanical and biological surface crusts (BSC, ) sealing the soil surface. This sealing effect affected the catchment hydrology and runoff behavior far more than was predicted by hydrological models using mainly the textural composition of the substrate [38, 39]. Even though the establishment of BSC and its effects on infiltration and runoff is well described for arid and semiarid regions [40, 41, 42, 43, 44], the effect of BSC on hydraulic soil properties is discussed controversially, e.g., . They conclude that the impact of BSC on water infiltration and conductivity is depending on surface properties of the crusts (roughness) as well as on specific crustal components. Due to regional climatic differences, BSC may either promote or inhibit infiltration and conductivity in soils.  showed the effects of different successional stages of soil crusts. They found that physical crusts, directly formed after a disturbance, lead to a homogenous and smooth soil surface and a promotion of runoff and soil erosion. The further development of BSC caused an increasing surface roughness (particularly well-developed BSC including mosses and lichens) and a decreasing potential for surface runoff. This is supported by soil moisture measurements in the topsoil along two gradients in the catchment showing highly significant increased soil moisture under BSC compared to bare soil (e.g., Table 1).
This was most prominent for the mean and maximum values in both 3 and 10 cm soil depth, but not for minimum soil moisture. At both sites, the moss cover of the BSC was very high. At the same time, the gradients showed increased silt, clay, and Corg. contents below the BSC. These findings indicate that surface stability is probably one of the key factors for BSC establishment. Once established, they initially promoted surface runoff due to hydrophobicity and pore clogging . During further development, the BSC were more and more covered by mosses, which provide a high potential for water storage and eventually for higher infiltration rates due to increasing surface roughness . During recent years, most of the BSC areas were overgrown by higher vegetation. This development provides a good example for abiotic/biotic feedbacks and their spatiotemporal functioning.
Other examples for these kinds of feedback mechanisms are the spatial spreading of plant species within the catchment once they have established and the effect of surface structures like erosion gullies on plant species distribution. After the first years that were dominated by a disperse colonization of the site by prolific spreader species like Conyza canadensis , most of the newly arriving species first established in the western part of the catchment with the more loamy sands and then spread from there to other parts of the catchment [50, 51]. In some parts of the catchment, incision of erosion gullies increased to the point that the gully floor intersected with the groundwater surface leading to permanent groundwater seepage into these gullies. This transformed the ephemeral channels into more permanent stream networks . Phragmites australis, a reed species that first established around the pond area in the lower part of the catchment used these stream networks to extend its distribution uphill preferentially along the channels (Figure 5) probably due to better water availability caused by more favorable physical properties of the stream bed sediments and by a closer proximity to the groundwater surface.
Despite the high surface runoff in the first years, an unconfined aquifer was formed above the clay layer across the hillslope (Figure 6). The groundwater table showed a seasonal and spatial variation but an overall increasing trend during the first 5 years.
In 2010, a very wet year with more than 900 mm precipitation compared to the long-term average of 563 mm, groundwater levels peaked and reached almost the soil surface in many parts of the catchment. Afterward, levels decreased again. The overall discharge from the catchment is controlled by the water level in the pond above the outlet weir. Until spring 2007, no discharge was recorded, due to the filling processes at the hillslope and the pond. Then, spiky, episodic discharge events were induced by rain, surface runoff, and snowmelt events (Figure 7).
Discharge also peaked in 2010 (max. 21 L s−1) and became more continuous afterward but still with long periods without discharge during summer. Since summer 2013 only very low discharge <1 L s−1 was recorded. Data from groundwater levels at the hillslope and water level in the pond that determines discharge showed no significant correlation during the first years (Figure 8).
With time, both the correlation between the coefficient and the slope of the linear regression increased. This indicated that discharge was not controlled by groundwater flow in the beginning but mainly by surface runoff from the hillslope. With development of the catchment, especially the buildup of groundwater and filling of the aquifer, base flow became likely the dominating source of discharge as is typical for many catchments . Time series of R2 values and slope of the regression between the groundwater and pond level showed a similar trend as the cover of vascular plants in the catchment (Figure 9).
These results indicate a transition of the hydrological regime from a surface runoff dominated discharge state to a state in which evapotranspiration and groundwater flow dominate the catchment response to rainfall.
Over the period of 10 years, we could define three phases of feedback controls on the catchment hydrology (Figure 10). In the very initial phase, these controls were mainly abiotic feedbacks between the initial substrate properties and rainfall events resulting in surface runoff, erosion, and gully formation (Figure 10a). These processes controlled discharge in the catchment and resulted in geomorphological changes of the hillslope surface. Surface runoff was promoted by the almost complete absence of vegetation and the sealing of surfaces due to the formation of various soil crusts. In this phase, substrate was transported from the hillslope downhill forming a sedimentation fan in the lower part above and in the pond. This resulted not only in a sorting of particle sizes within the fan but also in a residual accumulation of coarse-textured fragments at the soil surface uphill. Due to the various filling processes in the Pleistocene sediment body above the clay layer and in the pond, catchment discharge in this phase was very low and was dominated by single episodic events of surface runoff triggered by rainfall events or snowmelt.
After the rapid invasion and establishment of vascular plants, the second phase was mainly controlled by abiotic/biotic feedbacks (Figure 10b). As the vegetation cover increased, the soil surface was stabilized, and surface roughness increased. Groundwater levels increased and promoted plant growth especially within the gullies due to better water availability. Surface runoff and erosion were reduced and occurred only at episodic events after heavy rains or thunderstorms. After the pond was filled very quickly and the groundwater had increased to high levels, catchment discharge was high and more continuous in this phase.
In the third phase, catchment hydrology was still controlled by abiotic/biotic feedback mechanism, but biotic components had an increasing impact (Figure 10c). The vegetation and especially the woody species clearly profited from the high groundwater table and allowed a strong growth. The BSC developed to a moss dominated state, which changed their role in hydrological functioning from reducing infiltration and surface runoff promotion to increasing water storage, infiltration, and higher soil moisture below the BSC. At the same time, the spatial distribution of BSC was reduced due to overgrowing by higher vascular plants. Both vegetation and BSC altered the soil properties of the uppermost topsoil by increasing the organic matter contents and by accumulation of surface litter, which resulted in better infiltration. Evapotranspiration increased both due to evaporation in periods with groundwater tables close to the surface and the increase of vegetation cover throughout the catchment, which in turn lowered the groundwater. Especially, the higher woody vegetation increased canopy interception. Some of the mainly inactive erosion gullies that were incised deep enough to drain groundwater contributed to a higher groundwater discharge in this phase. Surface runoff was almost completely stopped. Due to the elevated evapotranspiration demand, catchment discharge decreased again and was mainly fed by groundwater.
These three phases of feedback controls on catchment hydrology have to be seen as an idealization of governing processes and the transition between them is gradual. The overall fast development within the catchment also increased spatial heterogeneity and diversity. Therefore, different parts and patches within the catchment may be in different phases at the same time or may remain in single phases over different periods. This behavior depends mainly on initial variations in substrate properties, on vegetation patterns formed and on spatial extent and intensity of the described feedback processes.
Ten years of data from the Chicken Creek catchment showed a very rapid development of the site with an increasing complexity and heterogeneity. In the first years, stochastic signals like the initial substrate conditions and external drivers like extreme weather were the most important factors resulting in abiotic/abiotic feedback mechanisms shaping the morphology of the site and creating site diversity. Invading vegetation from the regional species pool (representing external ecological memory) over time increased the role of abiotic/biotic feedbacks as could be shown in many examples.
Sites like Chicken Creek with known boundary conditions and structure information could help in disentangling feedback mechanisms between hydrological, pedogenic, biological, and geomorphological processes. Also, a more integrative view of succession and its drivers during the transition from initial, less complex systems to more mature ecosystems can be derived from such experimental sites developing from defined starting conditions [54, 55, 56]. Long-term time series of data are a key for a better understanding of these processes and the effects on ecosystem resilience [3, 57] and self-organization  as well as past and future effects of disturbance and global change .
This study was part of the Transregional Collaborative Research Centre 38 (CRC/TR 38), which was financially supported by the Deutsche Forschungsgemeinschaft (DFG, Bonn). The monitoring was funded by the Brandenburg University of Technology (BTU) and the Brandenburg Ministry of Science, Research, and Culture (MWFK, Potsdam). The authors thank Vattenfall Europe Mining AG for providing the research site and the many helping hands at the BTU contributing to the field and lab work.