A multiscale framework for deconstructing the ecosystem physical template of high altitudes lakes
=================================================================================================

* Dragos G. Zaharescu
* Peter S. Hooda
* Carmen I. Burghelea
* Antonio Palanca-Soler

## ABSTRACT

An ecosystem is generally sustained by a set of integrated physical elements forming a functional landscape unit-ecotope, which supplies nutrients, microclimate, and exchanges matter and energy with the wider environment. To better predict environmental change effects on ecosystems, particularly in critically sensitive regions such as high altitudes, it is imperative to recognise how their natural landscape heterogeneity works at different scales to shape habitats and sustain biota communities prior to major changes.

We conducted a comprehensive survey of catchment physical, geological and ecological properties of 354 high altitude lakes and ponds in the axial Pyrenees at a variety of scales, in order to formulate and test an integrated model encompassing major flows and interactions that drive lake ecosystems.

Three composite drivers encompassed most of the variability in lake catchment characteristics. In order of their statistical weight they were: (i) hydrology/hydrodynamics-responsible for type and discharge of inlets/outlets, and for water body size; (ii) bedrock geomorphology, summarizing geology, slope and fractal order-all dictating vegetation cover of catchment slope and lake shore, and the presence of aquatic vegetation. And, (iii) topography, i.e. catchment formation type-driving lakes connectivity, and the presence of summer snow deposits. While (i) appeared to be local, (ii) and (iii) showed gradient changes along altitude and latitude. These three drivers differentiated several lake ecotopes based on their landscape similarities. The three-driver model was successfully tested on a riparian vegetation composition dataset, further illustrating the validity of the concept.

The findings inform on the relative contribution of scale-dependent catchment physical elements to lake ecotope and ecosystem formation, which represents vital information about main factors predicting the natural functioning of high altitude lakes, which should inform any assessment of potentially major deleterious effects due to environmental/climate change.

Keywords
*   high altitude lakes
*   ecotope
*   ecosystem
*   scale
*   landscape function
*   lake classification
*   categorical principal component analysis
*   fuzzy set ordination

## 1. INTRODUCTION

One of the first conceptual ideas illustrating ecosystem-landscape interdependence was Vernadsky's theory of Earth's surface evolution, which recognized the synergetic relationships and transfer of nutrients between geosphere and biosphere (Vernadsky 1926). Recent research in the critical zone framework (i.e. Earth’s near-surface environment influenced by life; Richter and Billings 2015) advances this understanding by providing high spatial and temporal resolution details of landscape physiology at a variety of scales.

From a landscape perspective, a lake is a structural and physiological unit that draws energy and nutrients from its surrounding catchment. A lake ecosystem is therefore sustained by its physical template (ecotope, the lake’s life support system), which incorporates elements of catchment geomorphology, land cover and climate, all directly and indirectly affecting the flows of water and nutrients resulting from bedrock weathering. Predicting how changes in physical environment control ecosystems in high altitude catchments is generally challenging, due to their remoteness, the complexity of their landscape, and the many direct and indirect linkages between landscape features and processes operating at different scales. For example abiotic factors such as water resilience and cycling, primary productivity and nutrient availability are all key aquatic factors shaping community/ ecosystem development (Van der Molen et al. 2003).

Of more than 300 million lakes on the Earth’s surface, a great abundance occur at mid-to-high altitudes (Downing et al. 2006). Only in the Pyrenees, a relatively low-density lacustric region, there are an estimated 1030 lakes > 0.05ha above 1000m altitude (Castillo-Jurado 1992), meaning that high altitude lakes mediate a great portion of ecological and geochemical processes in their catchments. Due to their remoteness and high topography, most of these lakes host pristine or semi-pristine ecosystems, and are under increasing attention worldwide as clean water repositories, hotspots of biodiversity (Gopal et al. 2000), sensors of long-range transported pollutants (Andrea et al. 2007) and global climate change (Williamson et al. 2009). Moreover, their location in headwater basins, imply that they are the first to collect and redistribute bedrock-derived nutrients to the wider biosphere. These waterbodies and surrounding catchments are therefore ideal for studying how physical environment sustains their ecosystems, before climate change can induce major deleterious effects.

Environmental influence on species richness in mountain-top lakes has been discussed in the conceptual framework of Equilibrium Theory of Island Biogeography (Vuilleumier 1970; Barbour and Brown 1974; Brown and Dinsmore 1988). The theory predicts species composition at equilibrium, in a suitable habitat, being a function of habitat isolation, size and composition (MacArthur and Wilson 1963; Losos and Ricklefs 2009). For example general trends in fauna and flora functional composition can be predicted by local physical characteristics, including geology, geomorphology, waterbody size and slope, and land cover (Della Bella et al. 2005; Mazerolle et al. 2005; Goebel et al. 2006).

At any given time, a lake/pond can be assumed to support a type of vegetation and fauna whose composition is constrained by substrate/ecotope characteristics. This could result in a particular configuration of nutrient distribution, microclimate, and a local ecosystem succession/evolution in time. It is therefore critical to understand the relative contribution of the physical elements of an ecosystem to supporting biota development, and how they connect to regional, continental and global gradients in substrate and climate. This could address a major need in ecology, to better model how physical heterogeneity within an ecosystem predicts current and future ecological dynamics, particularly in human-induced climate and habitat stress scenarios.

We will use the term “ecotope” to represent the lake/pond and its proximal catchment area as an integrated physiological unit that supports an ecosystem. Similar to spatial patches in landscape ecology (Forman 1995), we assume this unit to represent unique combinations of hierarchically organised abiotic drivers that interact and drive the flow of energy/nutrients at multiple spatial scales, ultimately feeding and shaping the development of an lake ecosystem. The ecotope concept allows to considering all such features and their spatial heterogeneity, including how it may be connected to large-scale gradients in substrate and climate, and thus has the potential to incorporate and predict their function.

The main aim of this work was to identify the main landscape elements assumed to sustain a lake ecosystem in high altitude basins, and model how they organise at different scales to produce coherent ecosystem function. We also postulated that a lake’s physical template is not formed randomly. Rather, it is a geomorphic inheritance left by the past major transformations of the landscape, particularly following last glaciation. The work is based on a survey of 354 waterbodies in the axial Pyrenees. The strong E-W orientation of this mountain range, together with large blocks of distinct geology provide sharp contrasts in climate and biogeography, that makes the concept easier to test against large geographical gradients.

A secondary aim was to identify and define a number of ecotope types, supporting distinct lake ecosystems, which integrate related physical drivers. The lake ecotope concept is ultimately validated by showing its effect on lake riparian ecosystem composition.

## 2. METHODS

### 2.1 Study area and geology

The Pyrenees extend over roughly 430 km from the Atlantic to the Mediterranean Sea and separate the Iberian Peninsula from the rest of continental Europe. The area under study extends over about 80 km in the axial part of Pyrénées National Park (Atlantic Pyrenees), France (Fig. 1). This area, under reinforced protection, is restricted to recreational hiking, angling, and seasonal livestock grazing. Due to their location the majority of the waterbodies could be considered to reflect mostly natural processes.

![Figure 1.](http://biorxiv.org/https://www.biorxiv.org/content/biorxiv/early/2015/12/15/034405/F1.medium.gif)

[Figure 1.](http://biorxiv.org/content/early/2015/12/15/034405/F1)

Figure 1. 
Study area (axial Pyrénées National Park, outlined in dark green) together with major hydrological and geological formations on a digital elevation model. Lake representations are after LANDSAT imagery; inset radar map is after JPL (2000); Pyrenees digital elevation model is after Geoportail ([http://www.geoportail.fr/](http://www.geoportail.fr/)); geological representation is after SGN (1996).

Bedrock geology is marked by the outcrop of Cauterets-Panticosa igneous (granitic) batholith in the central part, massively flanked by metasedimentary (shale) and sedimentary (limestone) materials (Fig. 1). The abundance of granite, which is particularly resistant to erosion, gives the region a characteristic steep-sloping aspect. The contact zone between Cauterets granitic outcrop and the low-grade Cambrian-Carboniferous metamorphic material yields ore deposits, some of which have been exploited for metalliferous mining in the past (Paegelow 2008). Mineral springs are abundant in this area, particularly the hot springs at the contact of granite with the stratified rocks.

### 2.2 Climatology

The main air masses are from the W - NW, bringing precipitation (i.e. rain, snow and moist air) mainly from the Atlantic and the Bay of Biscay (oceanic-suboceanic climate; Mate 2002). This leads to a marked contrast between different sections/valleys of the region with glacial formations being present mainly on the N-oriented slopes of the western and the central parts of the range. Some of the glaciers are still active and are the source of major torrents. Precipitation averages 100-160cm year-1 in the area while mean annual temperature is 13-14°C (0°C isotherm oscillating between 1200m in January-3300m in July/August). Tree line varies between 2000-2500m a.s.l. The snow cover above 2000m settles down in November and starts to thaw in April. The glacier forming line is relatively high, ranging between 2500 to 3200m a.s.l. (Kessler and Chambraud 1990).

### 2.3 Hydrology

There are more than 400 lakes and ponds within the boundaries of the Pyrénées National Park. The great majority of the lakes are of post-glacial origin and they are generally formed at the head of the valleys, in the axial part of the mountain range. At global scale they are relatively small water-bodies as aree >90% of the lakes on the Earth’s surface (Downing et al. 2006). A large number of mountain torrents (>210), locally called *gaves*, drain the lake catchments, and give a generally dendritic structure to the hydrological network (Fig. 1). These ‘gaves’ subdivide the area in six major units: Aspe, Ossau, Azun, Cauterets, Luz and Aure (Fig. 1 and Appendix S1). A number of lakes in the major valleys were transformed into reservoirs and are used to generate hydroelectricity and supply water to human populations further downstream (Mate 2002).

### 2.4. Sampling and statistical methodology

A total of 354 lakes/ponds were surveyed during the month of July in 2000, 2001 and 2002. The sampling was aimed to represent the majority of mountain lakes in the area. The survey of lakes was undertaken in an east-westward direction to minimize the possible bias induced by a generally late snow thaw in the western side. Appendix S1 lists the name and location of the surveyed water-bodies. At each location a number of major landscape factors considered to influence ecotope/habitat processes were visually approximated and scored according to dominant units. A detailed description of the variables surveyed is presented in Table 1. Lakes’ size/type categories were estimated from their surface area. This was calculated as the surface of an ellipse whose major and minor diameters were measured in the field. A digital laser telemeter was used for this purpose. Furthermore, a portable GPS device helped record the geo-position coordinates, i.e. latitude, longitude and altitude, at each location.

View this table:
[Table 1:](http://biorxiv.org/content/early/2015/12/15/034405/T1)

Table 1: 
Description of geographical and ecological variables used in the analysis of 354 altitude lakes from the central Pyrenees.

Principal Component Analysis (PCA) was used to reduce the landscape variables to a small number of composite variables (factors) that represent the major environmental trends/ processes in the dataset. PCA is suitable for multivariate data which finds independent sets (principal components, PC) of linearly related variables. Also, PCA is a relatively robust tool for datasets which are not normally distributed. A Varimax rotation was applied to the extracted axes (components) in order to maximize the captured variance. Any considered variable was excluded if the model was not improved by its inclusion in a principal component (Table 1).

To help identify ecotope units, the interaction between variable categories of each extracted PC and projected variables’ vectors on lakes ordination space were evaluated. A categorical principal component analysis (CATPCA) was applied in this case. CATPCA is a nonparametric approach appropriate to find relationships between variables which span over multiple scales (e.g. numerical, categorical and nominal). CATPCA, however, may be influenced by the sample characteristics. For this, the stability of CATPCA results from our data (the degree of sensitivity to changes in the data) was tested by bootstrap procedure. It implied 1000 sets of bootstrap samples with replacement being taken randomly from the original dataset and repeating CATPCA on each set. This procedure determined the constancy of assignment (correlation) of the variables to the component vectors and produced 90% confidence regions of component loadings. If the results provided by CATPCA are stable, we expect narrow confidence ellipses.

The reliability of ecotope factors for their influence on riparian vegetation composition was tested using the logic of (Multidimensional) Fuzzy Set Ordination (Roberts 2008). Briefly, this approach related the explanatory PCA-derived composite ecotope factors (summarized into regression factor score variables) to riparian vegetation structure (response variables) by using a distance matrix of species incidence (calculated on Sørensen similarity index). Generally this matrix gives a measure of similarity between sites based solely on biota composition.

Linear regression analysis was used to determine the behaviour of catchment-scale ecotope properties along large scale geographical gradients. The variables were summarised as regression factor scores of the extracted principal components (PCs) before being used as response variables to geographical predictors in the regression analysis. Statistical treatment of the data was conducted in SPSS for Windows. Bootstrap procedure was computed with macro file Categories CATPCA Bootstrap for PASW developed by Linting et al. (2007), available online at [http://www.spss.com/devcentral/](http://www.spss.com/devcentral/). Multidimensional Fuzzy Set Ordination was computed in R statistical language, using FSO (Roberts 2007) and LabDSV (Roberts, D 2012) packages. Step-across function was performed in VEGAN package (Oksanen et al. 2012) for R.

## 3. RESULTS AND DISCUSSION

The surveyed water-bodies spanned from 1161 to 2747 m altitude. Figure 2 presents the exploratory statistics of the assessed landscape variables. As can be observed from this, most of the water-bodies can be included into pond and small lake categories. These water-bodies are mostly located on relatively flat surfaces at the head of glacial valleys; they have granite-dominated bedrock, and a great number are connected in chain with other lakes/ponds within their area. Likewise, altitude lakes in the central Pyrenees typically have feebly developed riparian zones (as shown by a high frequency of lakes with low fractal order, Fig. 2), which correspond to a relatively young age on a lake evolutionary time scale. Aquatic vegetation was largely absent at the time of sampling. Regarding the hydrological dynamics most of the lakes/ponds are fed by precipitation or small surface streams of very low discharge, which is typical of high altitudes. Accordingly, a great number of them have visibly absent or small outputs. Water flowing from springs, on the other hand, seems to have very little importance in their hydrodynamics. Shore/slopes vegetation coverage for most of the water-bodies was < 10%, and a relatively mixed snow coverage was recorded in their near-catchment during the month of surveying, i.e. July.

![Figure 2.](http://biorxiv.org/https://www.biorxiv.org/content/biorxiv/early/2015/12/15/034405/F2.medium.gif)

[Figure 2.](http://biorxiv.org/content/early/2015/12/15/034405/F2)

Figure 2. 
Frequency distribution (%) of sampled landscape variables at 354 altitude lake/pond locations from the Central Pyrenees.

### 3.1 Deconstructing the main drivers of a lake physical template

The intersection between climate and geomorphology can potentially shape the formation of an ecotope. To examine the influence of landscape components on the structure of lake ecotopes at catchment-scale, a PCA of all assessed variables (Table 1) was carried out. This reduced the variables to a limited number of key components which can explain the main environmental processes. Three components accounted for more than 58% of the total variance in the lake characteristics (Fig. 3). The first component (PC1) accounted for 21.3% of the variation (Fig. 3). It, i.e. PC1 (interpreted hereafter as hydrodynamics), indicates a strong association between waterbody size and lake hydrology (type and volume of water input/output). This is important as wetland macrophyte and invertebrate richness are likely to vary with the size of a lake/pond (Oertli et al. 2002; Biggs et al. 2005), a core idea in the “ecological theory of island biogeography” (MacArthur and Wilson 1963; Losos and Ricklefs 2009).

![Figure 3.](http://biorxiv.org/https://www.biorxiv.org/content/biorxiv/early/2015/12/15/034405/F3.medium.gif)

[Figure 3.](http://biorxiv.org/content/early/2015/12/15/034405/F3)

Figure 3. 
Relationships between landscape variables in their projections on principal components 1-2 and 1-3 of principal component analysis (PCA). Variables clustering with the PCs are enclosed. Figure symbols represent the variables with high loading on: (□) PC1, (○) PC2 and (Δ) PC3. These variables are from a sample pool of 354 altitude waterbodies (lakes, ponds and pools) from the central Pyrenees. Rotation method: Varimax with Kaiser normalization. Kaiser-Meyer-Olkin measure of sampling adequacy= 0.72 Bartlett's test of sphericity: approx. *χ*2 = 1398.2 (P<0.001). Inset plot displays the number of extracted components.

The second component (PC2, explaining additional 19.2% of the total variance) had high loadings for the variables that would be determined by the main bedrock geology/ geomorphology, i.e. geology, shore sloping, % of slope/shore covered by grass, fractal order and the presence of aquatic vegetation (Fig. 3). Geerling et al. (2006) have shown that ecotope composition (i.e. riparian surface, vegetation coverage and composition) can change during rejuvenating hydro-geomorphological processes of rivers, i.e., meander progression, meander interruption and channel shift. Likewise, substrate geology and slope are recognised physical factors that can influence the characteristics of a lake through their effects on hydraulics, weathering and nutrient cycling processes which together shape its biological structure (EC 2000; Kamenik et al. 2001). It seems therefore that geo-morphology is a second major driver of an altitude lake ecotope development and can influence not only the topographically-related high energy processes, such as slope erosion and runoff, but also the riparian development, its vegetation coverage and the development of aquatic vegetation. Lake shores’ vegetation coverage is a crucial ecotope factor in high altitude waterbodies which has been found to control nutrient cycling in a lake and therefore its biotic composition (Kopacek et al. 2000).

Finally, the third PC axis accounted for further 17.8% of the variability in the lakes’ characteristics. The variables grouped under PC3 were: presence of snow deposits at shore level and in the near catchment, catchment type and visible connectivity with other lakes, together being interpreted as topographical formation (Fig. 3). The PC3 findings suggest that topography also has significant control over ecotopic processes by its influence on important factors such as habitat connectivity and habitat snow coverage, the latter being important in shaping land-water processes during the large periods a mountain lake catchment is snow-covered (Edwards et al. 2007).

Indeed, the patterns of snow distribution in rugged alpine terrain are the most visible consequence of topography and its interaction with climatic variables like precipitation, solar radiation and wind (Körner 1992; Gottfried et al. 1999; Körner 2003). The seasonal cycles of snow accumulation and ablation as well as snow coverage can have a crucial influence on high altitude ecosystems’ composition at a variety of scales, with species capable of coping with the environmental conditions/stresses becoming more abundant (Walker et al. 1993; Keller et al. 2005). Habitat connectivity, on the other hand, is an important factor in maintaining the integrity of metapopulations of plant (Biggs et al. 2005) and animal (Richards-Zawacki 2009) species, with species assemblages likely to be richer in areas that facilitate propagule dispersal and colonisation. This is a second important aspect of “island biogeography” theory (MacArthur and Wilson 1963) which predicts an increase in species number with a decrease in remoteness of an island ecotope.

The remaining 42% variability in the dataset is accounted for by other numerous assessed/not assessed small factors, individually each accounting for an insignificant amount of the variability (Fig. 3 inset).

### 3.2 Integrated physical drivers determine lake basin types

Further analysis of the three PCs, individually, can help uncover the influence they have on ecotope development. To classify the waterbodies into ecotope types we studied the interaction between the variable categories within individual PCs previously discussed, i.e. hydrodynamics (PC1, Fig. 4), geo/morphology (PC2, Fig. 5A and B) and topographical formation (PC3, Fig. 5C and D). The plot of individual PC variables yielded a considerable degree of stability, as shown by relatively narrow 90% confidence ellipses of the bootstrap component loadings (Appendix S2). We can therefore confidently use CATPCA to uncover relationships between variable vectors.

![Figure 4.](http://biorxiv.org/https://www.biorxiv.org/content/biorxiv/early/2015/12/15/034405/F4.medium.gif)

[Figure 4.](http://biorxiv.org/content/early/2015/12/15/034405/F4)

Figure 4. 
(A) Interaction between categorical variables of the first principal component (i.e. hydrodynamics-Fig. 2) of PCA and their projection on lake ordination space, according to CATPCA. Interacting categories are enclosed in grey. The association between variables and lakes is enclosed in dashed polygons, while lake classes and examples are illustrated in B (lake coding corresponds to Appendix S1). Figure legend: ○, water-body size; ⋄, nature of water output; □, tributary discharge; ▼ nature of water input. N=354 water-bodies.

![Figure 5.](http://biorxiv.org/https://www.biorxiv.org/content/biorxiv/early/2015/12/15/034405/F5.medium.gif)

[Figure 5.](http://biorxiv.org/content/early/2015/12/15/034405/F5)

Figure 5. 
Nonlinear interactions between categorical variables of the second (geo-morphology, A and B) and the third (topographical formation, C and D) principal components of PCA (Fig. 2), and their associations to lakes. Lake coding is detailed in Appendix S1.

As displayed in Figure 4A, the interaction between hydrodynamics variables (PC1) shows that small waterbodies such as pools and ponds are fed principally by meteoric water, e.g. snow and rain, and such water-bodies either lack or have temporary tributaries/output. They represent a lake ecotope category. A second category is represented by small and medium-size lakes. They are characterized by various forms of water input, e.g. springs and streams/waterfalls of low to high discharge; this category is also associated to a diverse output nature, e.g. surface and subterranean (Fig. 4A). On the other side, large lakes plot further apart and are represented by dam lakes (Fig. 4A). The analysis also shows the cross-point where major lake properties change, with variable vectors plotting onto two well-defined waterbody clusters; first cluster, pools and ponds of low water turnover, plotting on the negative side of the first dimension, and second cluster, represented by small to large lakes of relatively large tributary/output, plotting on the positive side in the ordination space (Fig. 4B). This is an important finding since waterbodies which receive significant runoff can have different biotic composition compared with the mainly rain-fed ones, as they will receive more nutrients from the catchment (EC 2000; Kamenik et al. 2001). For example Riera et al. (2000), Saros et al. (2005) and Robinson and Kawecka (2005) provide illustrative cases of how nutrient availability/drainage type can shape phytoplankton, crayfish and fish development in oligotrophic alpine lakes.

The plot of interaction between PC2 variables, representing geo/morphological processes, shows that landscape categories such as limestone/sandstone/conglomerates associate with lakes surrounded by relatively flat topography, >50% grass covered shore/slopes, a highly developed riparian zone and the presence of aquatic vegetation (Fig. 5A). On the other hand, granite-schist bedrock plots together with medium to steep lake shore slopes, <20% grass covered shore/slopes, a poorly developed riparian zone and lack of aquatic vegetation (Fig. 5A). These two lake categories, i.e. formed on limestone and granite, point out to a spatial segregation of lake ecotopes according to the two main geomorphological units in the Pyrenees. That is, the Paleozoic-Mezozoic sedimentary/ limestone bedrock and the granitic outcrops (Fig. 1) which can influence biota composition at these sites. Plotting of the surveyed sites, however, did not form well-defined clusters, suggesting rather transient ecotope differences between the two main categories, i.e. on limestone and granite bedrock (Fig. 5B), possibly owing to the influence of mixed geological materials in them.

An analysis of the third composite factor, i.e. topographical formation (PC3; Fig. 5C and D) reveals two major ecotope forms. On one hand, there are lakes at the head of glacial valleys. They are generally either interconnected in chain with other lakes, or are in a basin in the vicinity of a major lake, and have a high proportion of summer snow deposits on their shores/near-catchment (Fig. 5C). Secondly, there are lakes on flat terrain, mountain passes and V/U shaped valleys, which are generally isolated or connected to a neighbouring lake, and have very scarce or no summer snow cover in their surroundings (Fig. 5C). The geomorphic changes resulted from the last glaciation and their location in the landscape (through the extent of their influence) are likely the main drivers of this factor. For example, differences in biota assemblages in different geomorphic settings have been found for the Northern Highland Lake District, Wisconsin, USA (Riera et al. 2000).

The analysis helped individualise and classify key physical drivers in terms of their influence on lake ecotope development. A conceptualised form of the analysis’s outcome can help simplify ecotope processes/forms that may be used to assess the relationship between key ecotope drivers and ecosystem functioning, e.g. vegetation structure (see below).

### 3.3 Conceptualisation and testing of a lake ecotope and its drivers

A conceptualization of the three major ecotope factors, i.e. hydrodynamics, geo/morphology and topography is presented in Fig. 6. This figure illustrates differences in climate, geological and topographical conditions on the mountain terrain that are responsible for the development of different ecotopes. For example, differences in precipitation received by two slopes of a mountain as a result of Foehn cloud formation, typical of high altitudes (Fig. 6A), influence the amount of water that a lake receives as a result of a sharp drop in air moisture and an elevation of the cloud as it meets dry, warm air masses from the opposite slopes. This is a typical phenomenon found along the wet Atlanticdry Mediterranean climate gradient (N-S) in the Pyrenees. Similarly, contrasting differences in substratum geo/morphology, i.e. limestone and siliceous, influence lake ecotope development (Fig. 6B). Such influence on zoobenthos assemblages has been exemplified by Borderelle et al. (2005). A conceptualisation of the third composite factor, i.e., topography formation (Fig. 6C) shows in a simplified way that different topographical forms or glacial formations determine different lake types.

![Figure 6.](http://biorxiv.org/https://www.biorxiv.org/content/biorxiv/early/2015/12/15/034405/F6.medium.gif)

[Figure 6.](http://biorxiv.org/content/early/2015/12/15/034405/F6)

Figure 6. 
Conceptualization of lake ecotope development at high altitudes and its principal drivers, i.e. hydrodynamics (A), geo-morphology (B) and topography (C). A digital elevation model of upper Tena Valley, central Pyrenees exemplifies this concept. A hydrodynamic gradient is represented here as a typical mountain Foehn cloud leaves the precipitation on one slope, then it is dispersed in contact with warm, dry air masses on the opposite slopes. Colours in geo-morphology and topography models represent distinct bedrock geology and glacial valley sections, respectively. A test of the conceptualized model performed by Multidimensional Fuzzy Set Ordination (D) shows the influence of the 3 ecotope formation factors on riparian vegetation species composition recorded at each lake (D). Number of permutations in this model = 1000.

This conceptual model was tested on a dataset representing a complete survey of riparian vegetation composition in the study area (Zaharescu 2011), which is briefly illustrated as follows. Riparian plant species composition was assessed against the three identified principal ecotope drivers by forward stepwise multidimensional fuzzy set ordination (MFSO) with step-across improvement. The MFSO used as predictor variables the three PCs (their regression scores) and as response variables a distance matrix of species incidence data. The statistical significance was tested by 1000 permutations. The effect magnitude of these variables is cumulatively presented in Fig. 6D. It shows that all three principal drivers (PCs) identified and conceptualised in this work had an overwhelming influence in determining the riparian vegetation composition (cumulative *r*=0.64, *p*<0.05). The strongest influence on vegetation composition clearly came from composite factor topography formation (*r*=0.43); this was independently followed by hydrodynamics (*r*=0.12) and geo-morphology (*r*=0.09).

This clearly illustrates that the conceptualised model is based on valid identification of the key ecotope forming factors and their influence in determining ecosystem development. The approach thus has the potential to be used as a tool, e.g. to predict response of vegetation or other ecosystem components to change in the physical environment.

### 3.3 Connection with large geographical gradients

The three ecotope forming factors while are fundamental to ecotope development, they may be influenced by large scale variation in altitude, latitude and longitude. An analysis of the three composite factors, i.e. hydrology/hydrodynamics, PC1, geo-morphology, PC2, and topographical formation, PC3 (Fig. 3) along altitudinal, latitudinal and longitudinal (continentality) gradients (Table 2) identifies elevation as a primary gradient explaining lake ecotopes development, with local effects of the variables associated with topography, i.e. PC3 (Fig. 7A). Altitude is a geographical constraint with known influences on catchment development through its main effects on glacial processes, such as cirque and valley formation.

![Figure 7.](http://biorxiv.org/https://www.biorxiv.org/content/biorxiv/early/2015/12/15/034405/F7.medium.gif)

[Figure 7.](http://biorxiv.org/content/early/2015/12/15/034405/F7)

Figure 7. 
(A) Relationship (linear) between topographical formation (i.e. regression factor scores of PC 3: catchment type, connectivity with other lakes, catchment and shore snow coverage - see Fig. 2) and altitudinal gradient. Slope equation is: y= 0.0029 × x − 6.3451. (B) Relationship between geo-morphology (i.e. regression factor scores of PC 2: dominant bedrock geology, % grass covered slopes, % grass covered shore, aquatic vegetation and fractal development) and latitude. Slope equation is: y= −1.6403 × + 4744.0970. Confidence intervals (95%) are dashed.

View this table:
[Table 2:](http://biorxiv.org/content/early/2015/12/15/034405/T2)

Table 2: 
Relationship (Spearman rank correlation coefficients) between geo-position variables and summarised landscape variables (i.e. regression factor scores of principal components) resulting from PCA. They represent: hydrodynamics (PC1), geo-morphology (PC2) and, topographical formation (PC3). Variables summarised by these composite factors are presented in Figure 3.

This can influence water and nutrient cycling and photosynthesis, and can lead to biota compositional differences along aquatic gradients at high altitudes. Examples of altitudinal effect on biota composition have been reported for various taxa, including zoobenthos, macrophyte and amphibian species (Hinden et al. 2005).

Latitude was the second most important broad-scale gradient for lake ecotope variation, with local effects of variables related to bedrock geo-morphology (regression factor score of the second PC) (Fig. 7B). Latitude apparently also had a broad effect on the variables associated to lake hydrodynamics, as shown by its relatively weak, but significant relationship with PC1 regression score (Spearman ρ=0.26; Table 2). The association of latitude to geological constraints may unveil a major N-S geomorphological gradient involved in lakes ecotope development across the mountain range. However, the variation in lake hydrodynamics across latitude could be explained by the rates at which the catchments are fed by the dominant Atlantic air masses (i.e. N-S direction), which lose moisture as they advance toward the (drier) axial part of the mountain range.

## 4. CONCLUSIONS

In elevated headwater basins, the development of a lake ecosystem’s physical support was scale dependent, and was driven primarily by basin’s hydrology/hydrodynamics, seconded by substrate geo/morphology and topographical formation. These major drivers resulted in a number of lake types, sharing similarities in their catchment physical properties, which provided distinctive abiotic settings for riparian plant communities.

Except hydrodynamics, which appeared to be mostly a local factor, the identified drivers were connected to large-scale geographical gradients, of which altitude and latitude were the most influential. The effect of lake physical template on its ecosystem is therefore expected to change along large horizontal and vertical gradients, in connection to major substrate units, and continental-to-global climate gradients. Changes in climate factors are therefore expected to affect not only lake ecosystem composition, as previously shown, but also many of its physical and chemical processes, such as water energy and weathering, that feed and shape fauna and flora development and structure. Our work provides compelling empirical confirmation of these cross-scale linkages in remote close-to-natural catchments. We interpret this as major confirmation of local-to-large scale landscape evolution in the postglacial period (Holocene) starting 11,000 years ago, which created the major elements of the physical landscape that drove biota setting.

We conceptualised and successfully tested how hydrodynamics, geo/morphology and topography support ecotope and riparian vegetation composition development. Our conceptualised template could be a common feature in other similar mountain ranges, therefore provide an integrated conceptual framework for hypothesis testing and experimentation in ecological modelling studies where scale and landscape properties/fluxes are important.

## SUPPLEMENTARY INFORMATION

View this table:
[Appendix S1:](http://biorxiv.org/content/early/2015/12/15/034405/T3)

Appendix S1: Lakes surveyed in this study
Water-bodies of central Pyrenees (The Pyrénées National Park) used in the present study. Altitude is in m a.s.l.; latitude and longitude are in decimal coordinates. Main valleys, locally called *gaves*, give the structure of the topography.

![Appendix S2:](http://biorxiv.org/https://www.biorxiv.org/content/biorxiv/early/2015/12/15/034405/F8.medium.gif)

[Appendix S2:](http://biorxiv.org/content/early/2015/12/15/034405/F8)

Appendix S2: Model testing
Plots showing the stability of CATPCA results (i.e. variables loading on first 2 extracted dimensions), for hydrodynamics, geo-morphology and topography factors (as summarized by PCA), as resulting from Bootstrap procedure. Component loadings are displayed together with 90% confidence intervals. The procedure shows a generally good level of stability, as illustrated by generally narrow confidence intervals.

## ACKNOWLEDGEMENTS

The study was developed with the financial support of Pyrenees National Park, France. The field work was supported over three years by an international team from Spain, UK, Romania and New Zeeland, including the late Richard Lester, as well as Javier Fernandez-Fañanas, Cristina Castan-Lanaspa, David Rodríguez-Vieites, Manuel Domínguez-Rey, Ana Quintillán-Cortiñas, Belén Cirujano-Díaz, Roberto García-Carrera, Jorge Diez-Dieguez, Jorge Rodriguez-Vila, Nicolas Palanca-Castán, Claudia Toda-Castán, Jesús Giraldez-Moreira, Juan Fernández-Rodríguez, Catalin Tanase, Andreea Vasiloiu, Carles Roselló-Vila, Carlos Tur-Lahiguera, Nuria Marti, Maria José Ferrus-Leiva, Julio Palanca-Castán, Bruce Dudley and José Martín-Gallardo – they are all gratefully acknowledged.

*   Received December 14, 2015.
*   Accepted December 15, 2015.


*   © 2015, Posted by Cold Spring Harbor Laboratory

This pre-print is available under a Creative Commons License (Attribution-NonCommercial-NoDerivs 4.0 International), CC BY-NC-ND 4.0, as described at [http://creativecommons.org/licenses/by-nc-nd/4.0/](http://creativecommons.org/licenses/by-nc-nd/4.0/)

## References

1.  Andrea L, Tartari GA, Musazzi S, et al. (2007) 21 High altitude lakes: limnology and paleolimnology. Dev. Earth Surf. Process. 10: 155–170.
    
    

2.  Barbour CD, Brown JH (1974) Fish species diversity in lakes. Am Nat 108:473–489. doi: 10.1086/282927
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1086/282927&link_type=DOI) 
    
    [Web of Science](http://biorxiv.org/lookup/external-ref?access_num=A1974T647300006&link_type=ISI) 

3.  Biggs J, Williams P, Whitfield M, et al. (2005) 15 Years of pond assessment in Britain: Results and lessons learned from the work of Pond Conservation. In: Aquatic Conservation: Marine and Freshwater Ecosystems. pp 693–714
    
    

4.  Brown M, Dinsmore JJ (1988) Habitat islands and the equilibrium theory of island biogeography: testing some predictions. Oecologia 75:426–429. doi: 10.1007/BF00376947
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1007/BF00376947&link_type=DOI) 

5.  Castillo-Jurado M (1992) Morfometría De Lagos Una Aplicación a Los Lagos. University of Barcelona
    
    

6.  Della Bella V, Bazzanti M, Chiarotti F (2005) Macroinvertebrate diversity and conservation status of Mediterranean ponds in Italy: Water permanence and mesohabitat influence. In: Aquatic Conservation: Marine and Freshwater Ecosystems. pp 583–600
    
    

7.  Downing J a., Prairie YT, Cole JJ, et al. (2006) The global abundance and size distribution of lakes, ponds, and impoundments. Limnol Oceanogr 51:2388–2397. doi: 10.4319/lo.2006.51.5.2388
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.4319/lo.2006.51.5.2388&link_type=DOI) 
    
    [GeoRef](http://biorxiv.org/lookup/external-ref?access_num=2010025273&link_type=GEOREF) 
    
    [Web of Science](http://biorxiv.org/lookup/external-ref?access_num=000240673800041&link_type=ISI) 

8.  EC (2000) Directive 2000/60/EC of the European Parliament and of the Council of 23 October 2000 establishing a framework for Community action in the field of water policy. Off J Eur Parliam L327:1–82. doi: 10.1039/ap9842100196
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1039/ap9842100196&link_type=DOI) 

9.  Edwards AC, Scalenghe R, Freppaz M (2007) Changes in the seasonal snow cover of alpine regions and its effect on soil processes: A review. 163:172–181. doi: 10.1016/j.quaint.2006.10.027
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1016/j.quaint.2006.10.027&link_type=DOI) 

10. Forman RTT (1995) Land Mosaics: The Ecology of Landscapes and Regions, 1st edn. Cambridge University Press
    
    

11. Geerling GW, Ragas AMJ, Leuven RSEW, et al. (2006) Succession and rejuvenation in floodplains along the river Allier (France). In: Hydrobiologia. pp 71–86
    
    

12. Goebel PC, Pregitzer KS, Palik BJ (2006) Landscape hierarchies influence riparian ground-flora communities in Wisconsin, USA. For Ecol Manage 230:43–54. doi: 10.1016/j.foreco.2006.04.035
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1016/j.foreco.2006.04.035&link_type=DOI) 

13. Gopal B, W.J. J, Davis. JA (eds) (2000) Biodiversity in wetlands: assessment, function and conservation. Backhuys Publishers, Leiden, Netherlands
    
    

14. Gottfried M, Pauli H, Reiter K, Grabherr G (1999) A fine-scaled predictive model for changes in species distribution patterns of high mountain plants induced by climate warming. Divers Distrib 5:241–251. doi: 10.1046/j.1472-4642.1999.00058.x
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1046/j.1472-4642.1999.00058.x&link_type=DOI) 

15. Hinden H, Oertli B, Menetrey N, et al. (2005) Alpine pond biodiversity: What are the related environmental variables? In: Aquatic Conservation: Marine and Freshwater Ecosystems. pp 613–624
    
    

16. Kamenik C, Schmidt R, Kum G, Psenner R (2001) The Influence of Catchment Characteristics on the Water Chemistry of Mountain Lakes. Arctic, Antarct Alp Res 33: 404–409.
    
    

17. Keller F, Goyette S, Beniston M (2005) Sensitivity analysis of snow cover to climate change scenarios and their impact on plant habitats in alpine terrain. Clim Change 72:299–319. doi: 10.1007/s10584-005-5360-2
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1007/s10584-005-5360-2&link_type=DOI) 
    
    [Web of Science](http://biorxiv.org/lookup/external-ref?access_num=000232908800002&link_type=ISI) 

18. Kessler J, Chambraud A (1990) Météo de la France. Tous les climats, localité par localité. Paris, France (in French)
    
    

19. Kopacek J, Stuchlik E, Straskrabova V, Psenakova P (2000) Factors governing nutrient status of mountain lakes in the Tatra Mountains. Freshw Biol 43:369–383. doi: 10.1046/j.1365-2427.2000.00569.x
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1046/j.1365-2427.2000.00569.x&link_type=DOI) 

20. Körner C (1992) Response of alpine vegetation to global climate change. Catena Verlag Supplement:85–96.
    
    

21. Körner C (2003) Alpine plant life: functional plant ecology of high mountain ecosystems.
    
    

22. Linting M, Meulman JJ, Groenen PJF, van der Koojj AJ (2007) Nonlinear principal components analysis: introduction and application. Psychol Methods 12:336–358. doi: 10.1037/1082-989X.12.3.336
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1037/1082-989X.12.3.336&link_type=DOI) 
    
    [PubMed](http://biorxiv.org/lookup/external-ref?access_num=17784798&link_type=MED&atom=%2Fbiorxiv%2Fearly%2F2015%2F12%2F15%2F034405.atom) 
    
    [Web of Science](http://biorxiv.org/lookup/external-ref?access_num=000249296300006&link_type=ISI) 

23. Losos JB, Ricklefs RE (eds) (2009) The Theory of Island Biogeography Revisited. Princeton University Press
    
    

24. MacArthur RH, Wilson EO (1963) An Equilibrium Theory of Insular Zoogeography. Evolution (N Y) 17:373–387. doi: 10.2307/2407089
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.2307/2407089&link_type=DOI) 
    
    [Web of Science](http://biorxiv.org/lookup/external-ref?access_num=A1963P196500011&link_type=ISI) 

25. Mate (2002) Atlas du Parc Natioal des Pyrénées. In: GIP ATEN, Morgan Multimedia, EDATER.
    
    

26. Mazerolle MJ, Desrochers A, Rochefort L (2005) Landscape characteristics influence pond occupancy by frogs after accounting for detectability. Ecol Appl 15:824–834. doi: 10.1890/04-0502
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1890/04-0502&link_type=DOI) 

27. Oertli B, Joye DA, Castella E, et al. (2002) Does size matter? The relationship between pond area and biodiversity. Biol Conserv 104:59–70. doi: 10.1016/S0006-3207(01)00154-9
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1016/S0006-3207(01)00154-9&link_type=DOI) 
    
    [Web of Science](http://biorxiv.org/lookup/external-ref?access_num=000174356200006&link_type=ISI) 

28. Oksanen J, Blanchet FG, Kindt R, et al. (2012) vegan: Community Ecology Package. R Packag. version 1:R package version 2.0-4.
    
    

29. Paegelow C (2008) Pyrenäen Bibliografie. Andorra, spanische & französische Pyrenäen. Pyrenees Bibliography. Andorra, Spain & French Pyrenees. Verlag Claus Paegelow, Bremen
    
    

30. Richards-Zawacki CL (2009) Effects of slope and riparian habitat connectivity on gene flow in an endangered Panamanian frog, Atelopus varius. Divers Distrib 15:796–806. doi: 10.1111/j.1472-4642.2009.00582.x
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1111/j.1472-4642.2009.00582.x&link_type=DOI) 
    
    [Web of Science](http://biorxiv.org/lookup/external-ref?access_num=000269264300007&link_type=ISI) 

31. Richter DD, Billings SA (2015) One physical system: Tansley’s ecosystem as Earth’s critical zone, Tansley review. New Phytol 206: 900–912.
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1111/nph.13338&link_type=DOI) 
    
    [PubMed](http://biorxiv.org/lookup/external-ref?access_num=25731586&link_type=MED&atom=%2Fbiorxiv%2Fearly%2F2015%2F12%2F15%2F034405.atom) 

32. Riera JL, Magnuson JJ, Kratz TK, Webster KE (2000) A geomorphic template for the analysis of lake districts applied to the Northern Highland Lake District, Wisconsin, U.S.A. Freshw Biol 43:301–318. doi: 10.1046/j.1365-2427.2000.00567.x
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1046/j.1365-2427.2000.00567.x&link_type=DOI) 

33. Roberts, D (2012) labdsv: Ordination and Multivariate Analysis for Ecology.
    
    

34. Roberts DW (2007) FSO: fuzzy set ordination. R package.
    
    

35. Roberts DW (2008) Statistical analysis of multidimensional fuzzy set ordinations. Ecology 89:1246–1260. doi: 10.1890/07-0136.1
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1890/07-0136.1&link_type=DOI) 
    
    [PubMed](http://biorxiv.org/lookup/external-ref?access_num=18543619&link_type=MED&atom=%2Fbiorxiv%2Fearly%2F2015%2F12%2F15%2F034405.atom) 

36. Robinson CT, Kawecka B (2005) Benthic diatoms of an Alpine stream/lake network in Switzerland. Aquat Sci 67:492–506. doi: 10.1007/s00027-005-0783-4
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1007/s00027-005-0783-4&link_type=DOI) 

37. Saros JE, Interlandi SJ, Doyle S, et al. (2005) Are the Deep Chlorophyll Maxima in Alpine Lakes Primarily Inducedby Nutrient Availability, not UV Avoidance? Arctic, Antarct. Alp. Res. 37: 557–563.
    
    

38. Van der Molen DT, Geilen N, Backx J, et al. (2003) Water Ecotope Classification for integrated water management in the Netherlands. Eur Water Manag Online 2003: 1–14.
    
    

39. Vernadsky V (1926) Biosfera, 1st edn. Nauch, Leningrad, Russia
    
    

40. Vuilleumier F (1970) Insular biogeography in contenintal regions. I. The northern Andes of South America. Am Nat 104: 373–388.
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.1086/282671&link_type=DOI) 
    
    [Web of Science](http://biorxiv.org/lookup/external-ref?access_num=A1970H251700007&link_type=ISI) 

41. Walker DA, Walker DA, Halfpenny JC, et al. (1993) Long-term studies of snow-vegetation interactions. Bioscience 43:287–301. doi: 10.2307/1312061
    
    [CrossRef](http://biorxiv.org/lookup/external-ref?access_num=10.2307/1312061&link_type=DOI) 
    
    [Web of Science](http://biorxiv.org/lookup/external-ref?access_num=A1993KX57500002&link_type=ISI) 

42. Williamson CE, Saros JE, Schindler DW (2009) Sentinels of Change. Science (80-) 323: 887–888.
    
    [Abstract/FREE Full Text](http://biorxiv.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6Mzoic2NpIjtzOjU6InJlc2lkIjtzOjEyOiIzMjMvNTkxNi84ODciO3M6NDoiYXRvbSI7czozNzoiL2Jpb3J4aXYvZWFybHkvMjAxNS8xMi8xNS8wMzQ0MDUuYXRvbSI7fXM6ODoiZnJhZ21lbnQiO3M6MDoiIjt9) 

43. Zaharescu DG (2011) Landscape ecology and geochemistry of high altitude lakes. Insights from the central Pyrenees. University of Vigo