Review and comparison of methods to study the contribution of variables in artificial neural network models

Convinced by the predictive quality of artificial neural network (ANN) models in ecology, we have turned our interests to their explanatory capacities. Seven methods which can give the relative contribution and/or the contribution profile of the input factors were compared: (i) the ‘PaD’ (for Partial Derivatives) method consists in a calculation of the partial derivatives of the output according to the input variables; (ii) the ‘Weights’ method is a computation using the connection weights; (iii) the ‘Perturb’ method corresponds to a perturbation of the input variables; (iv) the ‘Profile’ method is a successive variation of one input variable while the others are kept constant at a fixed value; (v) the ‘classical stepwise’ method is an observation of the change in the error value when an adding (forward) or an elimination (backward) step of the input variables is operated; (vi) ‘Improved stepwise a’ uses the same principle as the classical stepwise, but the elimination of the input occurs when the network is trained, the connection weights corresponding to the input variable studied is also eliminated; (vii) ‘Improved stepwise b’ involves the network being trained and fixed step by step, one input variable at its mean value to note the consequences on the error. The data tested in this study concerns the prediction of the density of brown trout spawning redds using habitat characteristics. The PaD method was found to be the most useful as it gave the most complete results, followed by the Profile method that gave the contribution profile of the input variables. The Perturb method allowed a good classification of the input parameters as well as the Weights method that has been simplified but these two methods lack stability. Next came the two improved stepwise methods (a and b) that both gave exactly the same result but the contributions were not sufficiently expressed. Finally, the classical stepwise methods gave the poorest results.     

Solving multidimensional bioeconomic problems with singular-perturbation reduction methods: Application to managing pest resistance to pesticidal crops

We investigate the application of ‘singular-perturbation’ reduction methods from the dynamical-systems literature to solve continuous-time multidimensional bioeconomic models resulting from integrating economics with increasingly complex biological structures. These methods reduce multidimensional solution space to the lower-dimensional subspace confining long-term dynamics. They arise naturally in problems with state variables evolving on widely disparate time scales. In particular, we demonstrate how the methods reduce the solution space of a linear-control specification—characterized by two state variables adjusting at widely disparate rates—to a single differential equation in the slow variable. All other system variables are determined by algebraic equations. We apply singular-perturbation methods to investigate the optimal management of pest resistance to pesticidal crops. The pest population evolves on a fast-time scale, while the population's genetic composition evolves on a slow-time scale. In comparison with past work, we can more fully characterize the continuous-time dynamics associated with a complex genetic specification.     

A link between ecological diversity indices and measures of biodiversity

The practice of environmental planning and protection frequently necessitates the quantification of ecological diversity. Traditional ‘ecological diversity indices’ are based on the abundances of species present. However, such indices are insensitive to taxonomic or similar differences. With equal species abundances they measure the species richness (species number) only. Conversely, so-called ‘biodiversity indices’ are based on species differences, but are insensitive to the abundance conditions. The quadratic entropy index is the only ecological diversity index, the value of which reflects both the differences ‘and’ abundances of the species. When a species list is given without abundance data, then, using the quadratic entropy index and postulating equal abundances, one gets the only biodiversity index derived from a traditional ecological index of diversity. Its extensive form is identical with the sum of differences or distances between the species present. This index trivially satisfies set monotonicity, an important property for biodiversity indices.     

Assessment of the contribution of land use pattern and intensity to landscape quality: use of a landscape indicator

The objective of this research work is the evaluation of the impact of landuse pattern and intensity on landscape by means of an indicator. The method used to calculate a ‘landscape indicator’ (I_land) allows to take into account the objective as well as the subjective approach of landscape. I_land corresponds to the degree of agreement between landscape supply by farmers and landscape demand by the social groups. The supply and the demand are evaluated through four criteria: ‘diversity’, ‘upkeep’, ‘openness’ and ‘heritage’. The landscape supply is calculated from data of landscape objects (punctual, linear and spatial) for each criterion recorded at the field level. The values of the four criteria for the landscape demand are allocated by the user(s) of the indicator (decision makers, regional council, social groups…) into five classes (0–4). The value of the landscape indicator is the least favourable difference between supply and demand for the four criteria. An example of calculation of the ‘landscape indicator’ for an arable farm is given. The collection of data needs 2 h with the farmer and 2 h for a survey of the farm land.     

Comparing discriminant analysis, neural networks and logistic regression for predicting species distributions: a case study with a Himalayan river bird

We assessed the occurrence of a common river bird, the Plumbeous Redstart Rhyacornis fuliginosus, along 180 independent streams in the Indian and Nepali Himalaya. We then compared the performance of multiple discrimant analysis (MDA), logistic regression (LR) and artificial neural networks (ANN) in predicting this species’ presence or absence from 32 variables describing stream altitude, slope, habitat structure, chemistry and invertebrate abundance. Using the entire data (=training set) and a threshold for accepting presence in ANN and LR set to P≥0.5, ANN correctly classified marginally more cases (88%) than either LR (83%) or MDA (84%). Model performance was assessed from two methods of data partitioning. In a ‘leave-one-out’ approach, LR correctly predicted more cases (82%) than MDA (73%) or ANN (69%). However, in a holdout procedure, all the methods performed similarly (73–75%). All methods predicted true absence (i.e. specificity in holdout: 81–85%) better than true presence (i.e. sensitivity: 57–60%). These effects reflect species’ prevalence (=frequency of occurrence), but are seldom considered in distribution modelling. Despite occurring at only 36% of the sites, Plumbeous Redstarts are one of the most common Himalayan river birds, and problems will be greater with less common species. Both LR and ANN require an arbitrary threshold probability (often P=0.5) at which to accept species presence from model prediction. Simulations involving varied prevalence revealed that LR was particularly sensitive to threshold effects. ROC plots (received operating characteristic) were therefore used to compare model performance on test data at a range of thresholds; LR always outperformed ANN. This case study supports the need to test species’ distribution models with independent data, and to use a range of criteria in assessing model performance. ANN do not yet have major advantages over conventional multivariate methods for assessing bird distributions. LR and MDA were both more efficient in the use of computer time than ANN, and also more straightforward in providing testable hypotheses about environmental effects on occurrence. However, LR was apparently subject to chance significant effects from explanatory variables, emphasising the well-known risks of models based purely on correlative data.     

Enforcement missions: Targets vs budgets

Enforcement of policy is typically delegated. What sort of mission should the head of an enforcement program be given? When there is more than one firm being regulated the firms’ decision problems—otherwise completely separate—become linked in a way that depends on that mission. Under some sorts of missions firms compete to avoid the attention of the enforcer by competitive reductions in the extent of their non-compliance, in others the interaction encourages competitive expansions. We develop a general model that allows for the ordering of some typical classes of missions. We find that in plausible settings ‘target-driven’ missions (that set a hard target in terms of environmental outcome but flexible budget) achieve the same outcome at lower cost than ‘budget-driven’ ones (that fix the enforcement budget). Inspection of some fixed fraction of firms is never optimal.     

Integrated ecological models: simulation of socio-cultural constraints on ecological dynamics

We suggest that general systems theory provides a common philosophical basis for dialog between ecological and social scientists interested in studying the reciprocal interactions of humans and their environment. We (1) provide a synopsis of the ‘systems approach' as viewed from the biological and social sciences, respectively; (2) develop a conceptual framework for the explicit linking of ecological and social variables, and (3) draw upon game theoretic results of the Prisoner's Dilemma to represent human decision-making quantitatively in a model that simulates the tragedy of the commons. The model consists of 5 submodels that represent the ‘observers world' and each of 4 ‘participant's worlds.' The observer's-world represents the decision processes, either Optimize or Tit-for-Tat, by which each of 2 users decides to add or remove animals. The 4 perceived worlds represent hypothetical situations in which (1) persons A and B both add an animal; (2) A adds and B does not; (3) B adds and A does not, and (4) neither A nor B add an animal. Simulation results indicate that net worth of the community and of each person individually under Tit-for-Tat is more than double the net worth attained under Optimize. Replacement of the static payoff matrix assumed in game theory with a dynamic quantitative model illustrates how ‘norm-based' approaches to ecosystem management can outperform optimizing approaches based on predicted outcomes. Although ‘soft systems' techniques may better help decision-makers reach norm-based agreements on ecosystem management, quantitative models have more explanatory value, and if developed sufficiently such models could incorporate complex social dimensions that would enhance further their explanatory value.     

Using social network analysis tools in ecology: Markov process transition models applied to the seasonal trophic network dynamics of the Chesapeake Bay

Ecosystem components interact in complex ways and change over time due to a variety of both internal and external influences (climate change, season cycles, human impacts). Such processes need to be modeled dynamically using appropriate statistical methods for assessing change in network structure. Here we use visualizations and statistical models of network dynamics to understand seasonal changes in the trophic network model described by Baird and Ulanowicz [Baird, D., Ulanowicz, R.E., 1989. Seasonal dynamics of the Chesapeake Bay ecosystem. Ecol. Monogr. 501 (59), 329–364] for the Chesapeake Bay (USA). Visualizations of carbon flow networks were created for each season by using a network graphic analysis tool (NETDRAW). The structural relations of the pelagic and benthic compartments (nodes) in each seasonal network were displayed in a two-dimensional space using spring-embedder analyses with nodes color-coded for habitat associations (benthic or pelagic). The most complex network was summer, when pelagic species such as sea nettles, larval fishes, and carnivorous fishes immigrate into Chesapeake Bay and consume prey largely from the plankton and to some extent the benthos. Winter was the simplest of the seasonal networks, and exhibited the highest ascendency, with fewest nodes present and with most of the flows shifting to the benthic bacteria and sediment POC compartments. This shift in system complexity corresponds with a shift from a pelagic- to benthic-dominated system over the seasonal cycle, suggesting that winter is a mostly closed system, relying on internal cycling rather than external input. Network visualization tools are useful in assessing temporal and spatial changes in food web networks, which can be explored for patterns that can be tested using statistical approaches. A simulation-based continuous-time Markov Chain model called SIENA was used to determine the dynamic structural changes in the trophic network across phases of the annual cycle in a statistical as opposed to a visual assessment. There was a significant decrease in outdegree (prey nodes with reduced link density) and an increase in the number of transitive triples (a triad in which i chooses j and h, and j also chooses h, mostly connected via the non-living detritus nodes in position i), suggesting the Chesapeake Bay is a simpler, but structurally more efficient, ecosystem in the winter than in the summer. As in the visual analysis, this shift in system complexity corresponds with a shift from a pelagic to a more benthic-dominated system from summer to winter. Both the SIENA model and the visualization in NETDRAW support the conclusions of Baird and Ulanowicz [Baird, D., Ulanowicz, R.E., 1989. Seasonal dynamics of the Chesapeake Bay ecosystem. Ecol. Monogr. 501 (59), 329–364] that there was an increase in the Chesapeake Bay ecosystem's ascendancy in the winter. We explain such reduced complexity in winter as a system response to lowered temperature and decreased solar energy input, which causes a decline in the production of new carbon, forcing nodes to go extinct; this causes a change in the structure of the system, making it simpler and more efficient than in summer. It appears that the seasonal dynamics of the trophic structure of Chesapeake Bay can be modeled effectively using the SIENA statistical model for network change.     

‘Optimal’ pollution abatement—whose benefits matter, and how much?

We examine the determinants of environmental regulatory activity (inspections and enforcement actions) and levels of air and water pollution for 409 US pulp and paper mills, using data for 1985–1997. We focus on the benefits to the surrounding population from pollution abatement. Plants with larger benefits emit less pollution, as do those with more kids and elders nearby. Plants in poor areas emit more pollution, though (surprisingly) we find less pollution in minority areas. Out-of-state neighbors seem to count less than in-state ones, although this effect diminishes if the bordering state's Congressional delegation is strongly pro-environment. We use ‘spatially lagged’ instrumental variables to control for the potential endogeneity of which individuals choose to locate near the plant. The results for regulatory activity are noticeably less significant than the emissions results.     

Environ indicator sensitivity to flux uncertainty in a phosphorus model of Lake Sidney Lanier, USA

Effective environmental impact assessment and management requires improved understanding of the organization and transformation of ecosystems in which independent agents are linked through an intricate network of energy, matter, and informational interactions. While advances have been made, we still lack a complete understanding of the processes that create, constrain, and sustain ecosystems. Network environ analysis (NEA) provides one approach for building novel ecosystem insights, but it is model dependent. As ecological modeling is an imprecise art, often complicated by inadequate empirical data, the utility of NEA may be limited by model uncertainty. Here, we investigate the sensitivity of NEA indicators of ecosystem growth and development to flow and storage uncertainty in a phosphorus model of Lake Sidney Lanier, USA. The indicators are total system throughflow (TST), total system storage (TSS), total boundary input (Boundary), Finn cycling index (FCI), ratio of indirect-to-direct flows (Indirect/Direct), indirect flow index (IFI), network aggradation (AGG), network homogenization (HMG), and network amplification (AMP). Our results make two primary contributions. First, they demonstrate that five of the indicators – FCI, Indirect/Direct, IFI, AGG and HMG – are relatively robust to the flow and storage uncertainty in the Lake Lanier model. This stability lets us draw robust conclusions about the Lake Lanier ecosystem organization (e.g., phosphorus flux in the lake is dominated by internal processes) in spite of uncertainties in the model. Second, we show that the majority of the indicators co-vary and that most of their common variation could be mapped onto two latent factors, which we interpret as (1) system integration and (2) boundary influences.     

Diatom Thalassiosira weissflogii in oligotrophic versus eutrophic culture: models and ultrastructure

Populations of marine diatom Thalassiosira weissflogii were grown in continuous cultures enriched with f/2 medium. One of the two contrasting cultures (‘eutrophic’) received 5.6 times more nutrients than the other (‘oligotrophic’). Two mathematical models are analyzed to estimate eutrophication differences. The second model based on the Michaelis–Menten uptake and Droop growth shows that cells in the eutrophic culture should have about 56% higher content of silica which is the limiting nutrient. Diatom samples were prepared for the transmission electron microscopy after cells have been kept in chemostats for 37 days. The structure of diatom cells was investigated and a comparison is made between cells grown in oligotrophic and eutrophic conditions. In eutrophic culture, dividing cells were encountered more frequently while cell concentration was approximately equal in both chemostats. The central vacuole of cells in eutrophic culture accumulated dispersed and compact material from amorphous to spherical shape. In some cells the large central vacuole had fibrilar and peppered dense materials in addition to translucent granules, vesicules and multivesicular bodies. In the cytoplasm we found increased number of multivesicular bodies, dense and lucent granules some of which enclose membrane particles and lucent vesicules. Dense material depositions observed in the vacuole are also seen in the cytoplasm associated with organelles, mitochondria and plasmalemma. Cells have well-developed, active and slightly increased number of dictyosomes (5–6). Some dictyosomes with dense secretory material in the cistern are apparently engaged in a granule formation process. Functional significance of dense material in the central vacuole, which has not been observed in cells grown in oligotrophic condition, is discussed.     

A study on the Spatial Distribution of Nitrogen and Phosphorus Balance, and Regional Nitrogen Flow Through Crop Production

The spatial distribution patterns of the nitrogen and phosphorus input/intake amounts in crop production within two small basins are examined, based upon a cropping unit distribution map that is obtained from remote sensing data analysis. Firstly, we examine the availability and suitability of approaches to the spatial distribution analysis of cultivation patterns classified from material flow characteristics of crop production using seasonal remote-sensing data. Secondly, material flow units in crop production are grouped according to the cultivation patterns obtained from the remote-sensing data analysis. Consequently, the spatial patterns of the amounts of both nitrogen and phosphorus inputs/intakes through crop production on farmland are examined and their spatial distribution maps are prepared according to the material flow units. In addition, we developed a nitrogen flow and runoff model and the model is simulated based on the examination of the results of spatial distribution patterns of the material flow units. The annual nitrogen runoff from small catchments, where various crops are cultivated, varies from 2.7 kg ha^–1 year^–1 to 108 kg ha^–1 year^–1 and the annual balanced losses of nitrogen in small catchments varied from –30 kg ha^–1 year^–1 to 101 kg ha^–1 year^–1. Also, the monthly changes in soil nitrogen of each material flow unit is estimated at –55 kg ha^–1 as a maximum decrease and 114 kg ha^–1 as a maximum increase. These results indicate that the spatial distribution patterns of nutrient input and intake through agricultural activities should be considered when analyzing the material flows and nutritient movement in soil–water systems in rural areas for watershed environmental control and regional agricultural management.     

Efficient and sustainable management of complex forest ecosystems

A large range of models has been developed for the analysis of optimal forest management strategies, with the well-known Faustmann models dating back to the mid-19th century. To date, however, there has been relatively little attention for the implications of complex ecosystem dynamics for optimal forest management. This paper examines the implications of irreversible ecosystem responses for efficient and sustainable forest management. The paper is built around two forest models that comprise two ecosystem components, forest cover and topsoil, the interactions between these components, and the supply of the ecosystem services ‘wood’ and ‘erosion control’. The first model represents a forest that responds in a reversible way to overharvesting. In the second model, an additional ecological process has been included and the ecosystem irreversibly collapses below certain thresholds in forest cover and topsoil depth. The paper presents a general model, and demonstrates the implications of pursuing efficient as well as sustainable forest management for the two forest ecosystems. Both fixed and variable harvesting cycles are examined. Efficient and sustainable harvesting cycles are compared, and it is shown that irreversible ecosystem behaviour reduces the possibilities to reconcile efficient and sustainable forest management through a variable harvesting cycle.     

Ecosystem services as a counterpart of emergy flows to ecosystems

A generic input-state-output scheme has been used to represent ecosystem dynamics. Systemic approaches to ecosystems use functions that are based either on inputs, state or outputs of the system. Some examples of approaches that use a combination of functions have been recently proposed. For example the use of eco-exergy to emergy flow can be seen as a mixed input-state approach; more recently, to connect the state to the output of the ecosystem, the relation of eco-exergy and ecosystems services has been proposed. This paper studies the link between the useful output of an ecosystems and its input through the relation between ecosystem services and emergy flow, in a kind of grey/black box scheme (i.e., without considering the state and the structure of the ecosystem). No direct connection between the two concepts can be determined, but identifying and quantifying the emergy flows feeding an ecosystem and the services to humans coming from them facilitate the sustainable conservation of Nature and its functions. Furthermore, this input-output relation can be established in general by calculating the ratio of the value of the ecosystem services to the emergy flow that supports the system. In particular, the ratio of the world ecosystem services to the emergy flow supporting the entire biosphere has been calculated showing that, at least at the global level, Nature is more efficacious in producing “money” (in form of ecosystem services) than economic systems (e.g., national economies and their GDP).     

Optimization of exergy and implications of body sizes of phytoplankton and zooplankton in an aquatic ecosystem model

Size appears to be an important parameter in ecological processes. All physiological processes vary with body size ranging from small microorganisms to higher mammals. In this model, five state variables — phosphorus, detritus, phytoplankton, zooplankton and fish are considered. We study the implications of body sizes of phytoplankton and zooplankton for total system dynamics by optimizing exergy as a goal function for system performance indicator. The rates of different sub-processes of phytoplankton and zooplankton are calculated, by means of allometric relationships of their body sizes. We run the model with different combinations of body sizes of phytoplankton and zooplankton and observe the overall biomass of phytoplankton, zooplankton and fish. The highest exergy values in different combinations of phytoplankton and zooplankton size indicate the maximum biomass of fish with relative proportions of phytoplankton and zooplankton. We also test the effect of phosphorus input conditions corresponding to oligotrophic, mesotrophic, eutrophic system on its dynamics. The average exergy to be maximized over phytoplankton and zooplankton size was computed when the system reached a steady state. Since this state is often a limit cycle, and the exergy copies this behaviour, we averaged the exergy computed for 365 days (duration of 1 year) in the stable period of the run. In mesotrophic condition, maximum fish biomass with relative proportional ratio of phytoplankton, zooplankton is recorded for phytoplankton size class 3.12 (log V μm³ volume) and zooplankton size 4 (log V μm³ volume). In oligotrophic condition the highest average exergy is obtained in between phytoplankton size 1.48 (log V μm³ volume) and zooplankton size 4 (log V μm³ volume), whereas in eutrophic condition the result shows the highest exergy in the combination of phytoplankton size 5.25 (log V μm³ volume) and zooplankton size 4 (log V μm³ volume).     

Potential impact model to assess agricultural pressure to landscape ecological functions

Conservation and protection of soil and water resources and visual aspects of landscape, as well as the promotion of biodiversity, are some of the central tasks of environmental policy development and social politics in the future. One of the main questions is: ‘which agricultural systems are able to guarantee sustained resource-conserving land use?’ Based on the ecological risk concepts of the 1970s and 1980s, a potential impact model was developed using a universal assessment algorithm derived from fuzzy logic. The model estimated the potential impact of agricultural land use on ecosystem function using a few resource indicators. Intervention intensities of agricultural land-use are set in relation to site conditions and aggregated for each of several defined potential impact categories. The interpretation with respect to risk and the calculation of potential impact values are explained.     

Use of genetic algorithms to select input variables in decision tree models for the prediction of benthic macroinvertebrates

Predicting freshwater organisms based on machine learning is becoming more and more reliable due to the availability of appropriate datasets, advanced modelling techniques and the continuously increasing capacity of computers. A database consisting of measurements collected at 360 sampling sites in non-navigable watercourses in Flanders was applied to predict the absence/presence of benthic macroinvertebrate taxa by means of decision trees. The measured variables were a combination of physical–chemical (temperature, pH, dissolved oxygen concentration, conductivity, total organic carbon, Kjeldahl nitrogen and total phosphorus), structural (granulometric analysis of the sediment, width, depth and flow velocity of the river) and two ecotoxicological variables. The predictive power of decision trees was assessed on the basis of the number of Correctly Classified Instances (CCI). A genetic algorithm was introduced to compare the predictive power of different sets of input variables for the decision trees. The number of input variables was reduced from 15 to 2–8 variables without affecting the predictive power of the decision trees significantly. Furthermore, reducing the number of input variables allowed to ease the identification of general data trends.     

Power-law behaviour and parametric models for the size-distribution of forest fires

This paper examines the distribution of areas burned in forest fires. Empirical size distributions, derived from extensive fire records, for six regions in North America are presented. While they show some commonalities, it appears that a simple power-law distribution of sizes, as has been suggested by some authors, is too simple to describe the distributions over their full range. A stochastic model for the spread and extinguishment of fires is used to examine conditions for power-law behaviour and deviations from it. The concept of the extinguishment growth rate ratio (EGRR) is developed. A null model with constant EGRR leads to a power-law distribution, but this does not appear to hold empirically for the data sets examined. Some alternative parametric forms for the size distribution are presented, with a four-parameter ‘competing hazards’ model providing the overall best fit.     

Dynamics of soil organic matter in primary and secondary forest succession on sandy soils in The Netherlands: An application of the ROMUL model

We applied the simulation model ROMUL of soil organic matter dynamics in order to analyse and predict forest soil organic matter (SOM) changes following stand growth and also to identify gaps of data and modelling problems. SOM build-up was analysed (a) from bare sand to forest soil during a primary succession in Scots pine forest and (b) on mature forest soil under Douglas fir plantations as an example of secondary succession in The Netherlands. As some of the experimental data were unreliable we compiled a set of various scenarios with different soil moisture regime, initial SOM pools and amount and quality of above and below ground litter input. This allowed us to find the scenarios that reflect the SOM dynamics more realistically. In the Scots pine forest, total litter input was estimated as 0.50 kg m⁻² year⁻¹. Two scenarios were defined for the test runs: (a) forest floor moisture regimes—‘dry, mesic and hydric’ and (b) augmenting a root litter pool with three ratios of needles and branches to roots: 1:1, 1:1.5 and 1:2.0. The scenario finally compiled had the following characteristics: (a) climate for dry site with summer drought and high winter moisture of forest floor; (b) a litter input of 0.25 kg m⁻² year⁻¹ above ground and 0.50 kg m⁻² year⁻¹ below ground; (c) a low nitrogen and ash content in all litter fall fractions. The test runs for the estimation of the initial SOM pools and the amount and proportion of above and below ground litter fall were also performed in the Douglas fir plantation. The inputs of above ground litter tested in various combinations were 0.30 and 0.60 kg m⁻² year⁻¹, and below ground litter 0.30, 0.60 and 0.90 kg m⁻² year⁻¹. The scenario that fitted the experimental data had an SOM pool of 20–25 kg m⁻², an aboveground litter input of 0.6 kg m⁻² year⁻¹and a below ground litter input of 0.9 kg m⁻² year⁻¹. The long-term simulation corresponded well with the observed patterns of soil organic matter accumulation associated with the forest soil development in primary and secondary succession. During primary succession in Scots pine forest on dry sand there is a consistent accumulation of a raw humus forest floor. The soil dynamics in the Douglas fir plantation also coincide with the observed patterns of SOM changes during the secondary succession, with SOM decreasing significantly under young forest, and SOM being restored in the older stands.     

Assessment of nitrogen leaching from arable land in large river basins: Part I. Simulation experiments using a process-based model

A two-step procedure for analysing nitrogen leaching from arable land in large river basins is suggested: (1) application of a process-based dynamic model for a set of representative conditions in a large river basin to simulate water and nitrogen fluxes and (2) development of a fuzzy-rule based metamodel using the simulated nitrogen fluxes in Step 1 as a training set. After that the metamodel can be used for rapid assessment of water quality inside the considered ranges of parameters, describing natural conditions and management practices. This paper describes Step 1 of the procedure. Step 2 is described in an accompanying paper (Haberlandt et al., Ecological Modelling 150 (3) (2002) 277–294). The advantage of this approach is that it combines the ‘process-based foundation’ with the resulting simplicity of the metamodel. Simulation experiments for analysing nitrogen (N) leaching from arable land were performed using the Soil and Water Integrated Model (SWIM) for a set of representative conditions in the Saale basin (23 687 km²) in Central Europe. The Saale River is one of the main tributaries of the Elbe. In advance, hydrological validation of the model was done for the whole Saale basin and validation of nitrogen dynamics was fulfilled in two mesoscale sub-basins of the Elbe. For the simulation experiments the drainage basin area was sub-divided into five climate zones and nine representative soil classes were chosen. The basic rotation and fertilisation schemes were established using regional information obtained from literature. In addition, the effects of changing the basic rotation to more/less intensive ones and changing fertilisation rates by 50% increase/decrease were studied. The ranges of simulated nitrogen fluxes for the basic rotation and fertilisation schemes are comparable to available regional estimates and differences between sub-regions and soils are plausible. The relative importance of natural and anthropogenic factors affecting nitrogen leaching for the Saale River basin was as follows: (1) soil, (2) climate, (3) fertilisation rate and (4) crop rotation. The simulation experiments provide a basis for a fuzzy-rule based metamodel approach, which aims at rapid water quality assessment of large regions.