Search

To link to full-text access for this article, visit this link: http://dx.doi.org/10.1016/j.jenvman.2007.09.010 Byline: M.J. Aitkenhead (a), I.H. Aalders (b) Abstract: Modelling land cover change from existing land cover maps is a vital requirement for anyone wishing to understand how the landscape may change in the future. In order to test any land cover change model, existing data must be used. However, often it is not known which data should be applied to the problem, or whether relationships exist within and between complex datasets. Here we have developed and tested a model that applied evolutionary processes to Bayesian networks. The model was developed and tested on a dataset containing land cover information and environmental data, in order to show that decisions about which datasets should be used could be made automatically. Bayesian networks are amenable to evolutionary methods as they can be easily described using a binary string to which crossover and mutation operations can be applied. The method, developed to allow comparison with standard Bayesian network development software, was proved capable of carrying out a rapid and effective search of the space of possible networks in order to find an optimal or near-optimal solution for the selection of datasets that have causal links with one another. Comparison of land cover mapping in the North-East of Scotland was made with a commercial Bayesian software package, with the evolutionary method being shown to provide greater flexibility in its ability to adapt to incorporate/utilise available evidence/knowledge and develop effective and accurate network structures, at the cost of requiring additional computer programming skills. The dataset used to develop the models included GIS-based data taken from the Land Cover for Scotland 1988 (LCS88), Land Capability for Forestry (LCF), Land Capability for Agriculture (LCA), the soil map of Scotland and additional climatic variables. Author Affiliation: (a) Department of Plant and Soil Science, University of Aberdeen, St. Machar Drive, Aberdeen AB24 3UU, Scotland, UK (b) The Macaulay Institute, Craigiebuckler, Aberdeen AB15 8QH, Scotland, UK Article History: Received 12 April 2006; Revised 16 July 2007; Accepted 14 September 2007

Abstract: Effective land cover mapping often requires the use of multiple data sources and data interpretation methods, particularly when no one data source or interpretation method provides sufficiently good results. Method-oriented approaches are often only effective for specific land cover class/data source combinations, and cannot be applied when different classification systems or data sources are required or available. Here we present a method, based on Endorsement Theory, of pooling evidence from multiple expert systems and spatial datasets to produce land cover maps. Individual ‘experts’ are trained to produce evidence for or against a class, with this evidence being categorised according to strength. An evidence integration rule set is applied to evidence lists to produce conclusions of different strength regarding individual classes, and the most likely class identified. The only expert system design implemented currently within the methodology is a neural network model, although the system has been designed to accept information from decision trees, fuzzy k-means and Bayesian statistics as well. We have used the technique to produce land cover maps of Scotland using three classification systems of varying complexity. Mapping accuracy varied between 52.6% for a map with 96 classes to 88.8% for a map with eight classes. The accuracy of the maps generated is higher than when individual datasets are used, showing that the evidence integration method applied is suitable for improving land cover mapping accuracy. We showed that imagery was not necessarily the most important data source for mapping where a large number of classes are used, and also showed that even data sources that produce low accuracy scores when used for mapping by themselves do improve the accuracy of maps produced using this integrative approach. Future work in developing the method is identified, including the inclusion of additional expert systems and improvement of the evidence integration, and evaluation carried out of the overall effectiveness of the approach. [Copyright &y& Elsevier]

Abstract: Decision tree classification provides a rapid and effective method of categorising datasets. Many algorithmic methods exist for optimising decision tree structure, although these can be vulnerable to changes in the training dataset. An evolutionary method is presented which allows decision tree flexibility through the use of co-evolving competition between the decision tree and the training data set. This method is tested using two different datasets and gives results comparable with or superior to other classification methods. A final discussion argues for the utility of decision trees over algorithmic or other alternative methods such as neural networks, particularly in situations where a large number of variables are being considered. [Copyright &y& Elsevier]

A wide variety of watershed-scale attributes can be used as predictors of the export of dissolved organic carbon (DOC) from a watershed. However, the complexity and number of relationships makes the development of generally applicable mechanistic models for prediction of DOC export based on measurement of factors difficult. Here we have applied neural network modelling methods to the prediction of stream flux and daily DOC export from several watersheds of varying size within the Dee valley, in north-east Scotland. A two-stage process was carried out in which first a model was developed which used a large number of variables thought to be relevant to DOC export, and then the possibility of using a restricted set of variables was investigated in order to reduce the amount of analysis required in order to produce accurate DOC export predictions. The results showed that it is possible to predict DOC export using input variables corresponding broadly to the factors responsible for soil formation, and that a single sample site may provide enough information to allow prediction for an entire watershed. However, in order to achieve a model with statistically significant results, it is necessary to use multiple sample sites per watershed, and to use measured rather than modelled flow values. Discussion is made of the effectiveness of the neural network method in developing models of DOC export, and of problems with the method (particularly in the inability to use NN models for process-based models). [Copyright &y& Elsevier]

Abstract: Modelling land use change is often constrained by imperfect and incomplete data sources. This paper explores three modelling methodologies and their ability to predict agricultural land use on the basis of information from the Scottish Agricultural Census. This dataset, which contains information on ownership, land use and employment statistics for the majority of Scotland, is restricted by law concerning the level of detail which can be provided, and as such is both the best available source of information for agricultural practice in Scotland and is partial and incomplete. It is demonstrated that the methodologies applied to the problem (neural network, Bayesian network and decision tree), with a limited number of relevant drivers included in the modelling process, are capable of use for the prediction of changes in land use, suitable for policy analysis. The reasons for selecting these particular modelling approaches included a need to deal with a large amount of noisy, inaccurate data, and the fact that each is capable of successfully investigating and quantifying unknown relationships between dataset variables. The greatest success, measured as a combination of accuracy, data-handling flexibility and ease of model comprehension by the user, was achieved by the decision tree method. [Copyright &y& Elsevier]

A convincing argument has been made in the literature for the advantages gained by using modular neural networks (NN) instead of homogeneous structures. Here, a modular NN design was used in conjunction with evolutionary algorithm methods to evolve an animat capable of learning behavioural patterns at several levels of complexity. A parallel was drawn between the training of the animat and the stages of learning experienced by the young of many animals. Following movement learning, the animat was eventually capable of navigating an environment and avoiding obstacles. Discussion is made of how such an animat with many degrees of freedom can develop complex behavioural patterns. [Copyright &y& Elsevier]

A neural network based facial recognition program (FADER—FAce DEtection and Recognition) was developed and tested. The hardware and software components were all standard commercial design, allowing the system to be built for minimal cost. Using a set of 1000 face and 1000 ‘no-face’ images, we achieved 94.7% detection rate, and a 0.6% false positive rate. Three different neural network models were applied to face recognition, using single images of each subject to train the system. A novel adaptation of the Hebbian connection strength adjustment model gave the best results, with 74.1% accuracy achieved. Each of the system''s components, including an intermediate substructure detection network, was subject to evolutionary computation in order to optimise the system performance. [Copyright &y& Elsevier]

Recently, much research has been done on the possibilities of autonomous robots navigating and performing obstacle avoidance. Vision is an important method of gathering information about obstacles and other objects within an environment. A virtual environment was created, in which an animat (virtual organism) navigates using a visual system comparable to that used by biological organisms. A spike-based neural network model was applied to learning. Various applicable rules on network topology were used, and examination was made of how well the animat learnt under varying sizes of visual input layer and intermediate ‘processing’ layer. In addition, three different training regimes were applied and their different merits discussed. It was discovered that increasing layer size in general improved the performance of the animat, provided that the sizes of the input array and the processing arrays correspond. The learning curve of the animat over time was investigated, allowing an optimal training time to be determined. These findings allow an insight into how well such a system can perform and how its design and training may be optimized. [Copyright &y& Elsevier]