Gunnar Lischeid

Nonlinear dimensionality reduction: Alternative ordination approaches for extracting and visualizing biodiversity patterns in tropical montane forest vegetation data

Abstract Ecological patterns are difficult to extract directly from vegetation data. The respective surveys provide a high number of interrelated species occurrence variables. Since often only a limited number of ecological gradients determine species distributions, the data might be represented by much fewer but effectively independent variables. This can be achieved by reducing the dimensionality of the data. Conventional methods are either limited to linear feature extraction (e.g., principal component analysis, and Classical Multidimensional Scaling, CMDS) or require a priori assumptions on the intrinsic data dimensionality (e.g., Nonmetric Multidimensional Scaling, NMDS, and self organized maps, SOM). In this study we explored the potential of Isometric Feature Mapping (Isomap). This new method of dimensionality reduction is a nonlinear generalization of CMDS. Isomap is based on a nonlinear geodesic inter-point distance matrix. Estimating geodesic distances requires one free threshold parameter, which defines linear geometry to be preserved in the global nonlinear distance structure. We compared Isomap to its linear (CMDS) and nonmetric (NMDS) equivalents. Furthermore, the use of geodesic distances allowed also extending NMDS to a version that we called NMDS-G. In addition we investigated a supervised Isomap variant (S-Isomap) and showed that all these techniques are interpretable within a single methodical framework. As an example we investigated seven plots (subdivided in 456 subplots) in different secondary tropical montane forests with 773 species of vascular plants. A key problem for the study of tropical vegetation data is the heterogeneous small scale variability implying large ranges of β -diversity. The CMDS and NMDS methods did not reduce the data dimensionality reasonably. On the contrary, Isomap explained 95% of the data variance in the first five dimensions and provided ecologically interpretable visualizations; NMDS-G yielded similar results. The main shortcoming of the latter was the high computational cost and the requirement to predefine the dimension of the embedding space. The S-Isomap learning scheme did not improve the Isomap variant for an optimal threshold parameter but substantially improved the nonoptimal solutions. We conclude that Isomap as a new ordination method allows effective representations of high dimensional vegetation data sets. The method is promising since it does not require a priori assumptions, and is computationally highly effective.

Trends in deposition and canopy leaching of mineral elements as indicated by bulk deposition and throughfall measurements

In the past three decades, numerous studies on the biogeochemistry of forested ecosystems in Europe and North America have shown that the deposition of mineral elements from the atmosphere strongly influences their functioning. Acidification of soils, surface- and groundwaters, N saturation and forest decline are key processes that change with rates of deposition of mineral elements (Ulrich 1994; Fenn et al. 1998; Evans et al. 2001). As an example of ecosystem functioning, the losses of elements from the ecosystem by seepage and runoff can be considered. On a European-wide, scale the deposition of S and N was shown to determine the Al losses from seepage and runoff in acid forest soils (Dise et al. 2001), the N deposition to determine the NO3 losses (MacDonald et al. 2002) and the Mg deposition to largely determine the Mg losses (Armbruster et al. 2002).

Grasping the heterogeneity of kettle hole water quality in Northeast Germany

In the young moraine landscape in Northeast Germany, small glacially created ponds, the so-called kettle holes, are very abundant. They exhibit large spatial heterogeneity, seemingly rendering each kettle hole unique. However, this would not be consistent with any scientific approach. Thus, a classification scheme has been developed for kettle holes in Northeast Germany based on morphology, hydrodynamics and connection to stream networks of the kettle holes as well as size, topography and land use of the respective catchment. These indices are assumed to be related both to water quality as well as to biological issues of the kettle holes. Starting in the mid-1990s, an extensive monitoring program has been established in the federal state of Brandenburg, Germany. In this study, a subset comprising 1,316 samples from 79 kettle holes was analysed, where 21 parameters had been determined. Sampling intervals varied widely, and were between bi-weekly and three-monthly at most sites. A nonlinear principal component analysis was performed. The first four components explained 90% of the variance. These components seem to provide quantitative measures of phosphorus release from the sediments during hypoxic periods, agricultural solute input, algae primary production, and geogenic compounds. This allowed differentiating between the natural and anthropogenic impact factors on water quality. In addition, scores of single components were related to properties of the kettle holes and their environments. The results contribute to a better understanding of biological and biogeochemical processes and can be used to verify the effects of conservation and management strategies for kettle holes.

Investigating short-term dynamics and long-term trends of SO4 in the runoff of a forested catchment using artificial neural networks

Abstract The impact of long-lasting non-point emissions on groundwater and streamwater in remote watersheds has been studied at numerous sites. In spite of substantially decreasing emissions in the last decade, recovery has not yet been observed in all cases. This trend might be masked by the considerable short-term variability of the chemical hydrographs. In this study, artificial neural networks are applied to investigate the SO 4 dynamics in the runoff of a small forested catchment susceptible to SO 4 deposition. Empirical models are fitted to the short-term dynamics at a time step of one day. About 75% of the variance of the SO 4 data is explained by the instantaneous discharge, short-term history of discharge and the moving average of SO 4 concentration in throughfall. In contrast, neither air temperature as an indicator for biological activity nor a snowmelt indicator based on the temperature sum increase the performance of the model. The model is used to investigate long-term trends in sub-regions of the phase space spanned by the identified input variables. According to the model, decreasing emissions have a significant effect on runoff SO 4 concentration only during the first severe storms at the end of the vegetation period. This suggests to focus on these events as indicators for recovery of the topsoil layers.

Water flow paths and residence times in a small headwater catchment at Gårdsjön, Sweden, during steady state storm flow conditions

Since April 1991 the small forested headwater catchment Gl at Gardsjon (Sweden) has been covered by a roof underneath which natural throughfall is replaced by artificial irrigation with a controlled chemical composition. Here this unique experimental setup was used for a tracer experiment with LiBr. The tracer pulse was applied to a subcatchment of approximately 1000 m2 that was maintained at steady state flow conditions throughout the experiment. Except these steady state flow conditions, the irrigation rates corresponded to a typical storm flow episode. Infiltration of event water was confined to the steep slope of a subcatchment of G1; no water was applied at the boggy valley bottom or close to the weir. An array of groundwater wells, suction lysimeters, and surface water sampling plots was used to document the soil passage of this pulse. Breakthrough in runoff (i.e., streamflow from the weir) occurred in a single peak within about 17 hours when less than 15% of the estimated total soil water in the subcatchment was replaced, indicating a relatively small fraction of mobile water. Tracer concentrations in groundwater wells and surface water in the catchment revealed some shallow, locally confined flow paths through the lower parts of the subcatchment. However, 4 days after application of the tracer the runoff concentration already reached the preevent background level for Br−. Only about 14% of the applied tracer was recovered by this time. Taking this into account, i.e., considering bromide as a nonconservative tracer, a one-dimensional model application (two-region convection-dispersion approach) successfully reproduced breakthrough curves at various places in the catchment. Thus the small portion of mobile water has intensive contact with the resident immobile water along the one-dimensional flow paths yielding an extremely long tail in the residence time distribution by back diffusion from immobile water. Results of this experiment qualitatively confirm earlier tracer studies under less controlled conditions but are virtually impossible to extrapolate quantitatively to transient flow conditions.

The effects of climate and changing land use on the discharge regime of a small catchment in Tanzania

Increasing pressure on water resources makes it necessary to understand the reasons for the changes in the run-off characteristic of the Ngerengere River in Tanzania during recent years. Changing land use and changes in climate boundaries are identified as effects. A combination of statistical analysis and the use of the hydrological model SWAT were chosen to handle the problem of poor data quantity and quality with non-overlapping periods. Changes in the discharge regime were identified with the 5th percentile of the flow duration curve as an indicator for high-flow events, with an indicator for low-flow duration and with the base flow index. The analysis showed that climate boundaries and changing land use do not have a uniform effect on discharge in the catchment. Changing land use affects surface run-off and increases floods in the mountainous areas. Changes in climate boundaries increase the duration of low flow and no flow in the Ngerengere catchment. Changes in climate conditions and land use had antipodal effects on parts of the discharge regime. Thus, the observed changes in land use and climate conditions partially compensate for each other.

Carbon and nutrient cycling in kettle hole sediments depending on hydrological dynamics: a review

Kettle holes as a specific group of isolated, small lenticfreshwatersystems (LFS) often are (i) hot spots of biogeochemical cycling and (ii) exposed to frequent sediment desiccation and rewetting. Their ecological functioning is greatly determined by immanent carbon and nutrient transformations. The objective of this review is to elucidate effects of a changing hydrological regime (i.e., dry–wet cycles) on carbon and nutrient cycling in kettle hole sediments. Generally, dry–wet cycles have the potential to increase C and N losses as well as P availability. However, their duration and frequency are important controlling factors regarding direction and intensity of biogeochemical and microbiological responses. To evaluate drought impacts on sediment carbon and nutrient cycling in detail requires the context of the LFS hydrological history. For example, frequent drought events induce physiological adaptation of exposed microbial communities and thus flatten metabolic responses, whereas rare events provoke unbalanced, strong microbial responses. Different potential of microbial resilience to drought stress can irretrievably change microbial communities and functional guilds, gearing cascades of functional responses. Hence, dry–wet events can shift the biogeochemical cycling of organic matter and nutrients to a new equilibrium, thus affecting the dynamic balance between carbon burial and mineralization in kettle holes.

Investigating trends of hydrochemical time series of small catchments by artificial neural networks

Abstract The short-term variation of discharge and solute concentration of the runoff of small catchments generally reflects the interplay of a variety of different processes. This makes the investigation of anthropogenic impacts on the catchments runoff often rather difficult. On the other hand, short-term dynamics at the output boundary provide information about the system. This information can be used, in principle at least, to assess its long-term behaviour more precisely. In this paper examples of time series of sulphate and nitrate in the runoff of two small forested catchments are presented. To minimise the danger of over-parametrisation, the objective was to find a very simple empirical model to map a substantial portion of the observed variance (daily values). Here artificial neural networks were applied. They yield an efficiency of more than 0.7 for the solutes investigated, based on discharge depth and air temperature as input variables only. As a next step, the invariance of these relationships was investigated. In the case of sulphate, a significant trend is observed. However, it differs considerably for different subregions of the regression plane. Thus the neural network approach reveals a much more detailed insight into temporal shifts of the dynamics than an overall trend analysis.

Temporal variability of the optimal monitoring setup assessed using information theory

Hydrology is rich in methods that use information theory to evaluate monitoring networks. Yet in most existing studies, only the available data set as a whole is used, which neglects the intraannual variability of the hydrological system. In this paper, we demonstrate how this variability can be considered by extending monitoring evaluation to subsets of the available data. Therefore, we separately evaluated time windows of fixed length, which were shifted through the data set, and successively extended time windows. We used basic information theory measures and a greedy ranking algorithm based on the criterion of maximum information/minimum redundancy. The network investigated monitored surface and groundwater levels at quarter-hourly intervals and was located at an artificially drained lowland site in the Spreewald region in north-east Germany. The results revealed that some of the monitoring stations were of value permanently while others were needed only temporally. The prevailing meteorological conditions, particularly the amount of precipitation, affected the degree of similarity between the water levels measured. The hydrological system tended to act more individually during periods of no or little rainfall. The optimal monitoring setup, its stability, and the monitoring effort necessary were influenced by the meteorological forcing. Altogether, the methodology presented can help achieve a monitoring network design that has a more even performance or covers the conditions of interest (e.g., floods or droughts) best.

Effective modelling of percolation at the landscape scale using data-based approaches

Process-based models have been extensively applied to assess the impact of landuse change on water quantity and quality at landscape scales. However, the routine application of those models suffers from large computational efforts, lack of transparency and the requirement of many input parameters. Data-based models such as Feed-Forward Multilayer Perceptrons (MLP) and Classification and Regression Trees (CART) may be used as effective models, i.e. simple approximations of complex process-based models. These data-based approaches can subsequently be applied for scenario analysis and as a transparent management tool provided climatic boundary conditions and the basic model assumptions of the process-based models do not change dramatically. In this study, we apply MLP, CART and Multiple Linear Regression (LR) to model the spatially distributed and spatially aggregated percolation in soils using weather, groundwater and soil data. The percolation data is obtained via numerical experiments with Hydrus1D. Thus, the complex process-based model is approximated using simpler data-based approaches. The MLP model explains most of the percolation variance in time and space without using any soil information. This reflects the effective dimensionality of the process-based model and suggests that percolation in the study area may be modelled much simpler than using Hydrus1D. The CART model shows that soil properties play a negligible role for percolation under wet climatic conditions. However, they become more important if the conditions turn drier. The LR method does not yield satisfactory predictions for the spatially distributed percolation however the spatially aggregated percolation is well approximated. This may indicate that the soils behave simpler (i.e. more linear) when percolation dynamics are upscaled.