Yazar "Beklioglu, Meryem" seçeneğine göre listele

Listeleniyor 1 - 2 / 2
  • [ X ]
    Öğe
    Modeling complex nonlinear responses of shallow lakes to fish and hydrology using artificial neural networks
    (Elsevier, 2006) Tan, Can Ozan; Beklioglu, Meryem
    Mathematical abstractions may be useful in providing insight that is otherwise very difficult to obtain due to complex interactions in the ecosystems. The difficulty associated with the nonlinearity and complexity of ecological processes and interactions can be avoided with artificial neural networks (ANN) and generalized logistic models (GLMs) which are practically ANNs without hidden layer. An ANN and a GLM were developed to determine the probability of submerged plant occurrence in five shallow Anatolian lakes, and both models were tested on an spatially and temporally independent test set consisting of data collected from another Anatolian lake, Lake Mogan. Independent variables included the ratio of carp biomass to total fish biomass (carp ratio), the amplitude of water levels, water level z-scores, morphology indices of the lakes and a period index. We optimized different ANN architectures to optimize the performance, and used bootstrapping to determine the maximum number of epochs to train the model. Cross-entropy measure (c-index) was used to assess the performance of the models, which is approximately the area under the receiver-operator curve (RCC) and equivalent to log likelihood measure for binary outcome events. Following optimization, both ANN and GLM models explained the data effectively (corrected c-indices 0.99 and 0.95, respectively), and both models explained the independent test data set completely (c-indices 1.00 for both models). Both models predicted a very strong impact of carp ratio on the occurrence of submerged plants, and relatively strong impacts of the amplitude, water level z-score and morphology index. In that sense, the predictions of the models were consistent with the field observations regarding the study lakes, as well as with alternative stable states theory of shallow lakes. in general, both models have been successfully predicted the state shifts in shallow lakes included in the model, and identified the thresholds for inducing those shifts from mainly hydrological variables. However, ANN model was more successful in capturing the relationships and interactions among input variables compared to the GLM model. To best of our knowledge, current study is the first ANN-based approach to predict the state shifts in shallow lakes and to identify the corresponding threshold value of the control factors. The model has been robust in being generalizable over distinct shallow lake ecosystems, at least in the same climatic zone as the study lakes.
  • [ X ]
    Öğe
    Predictive models in ecology
    (Elsevier, 2006) Tan, Can Ozan; Ozesmi, Uygar; Beklioglu, Meryem; Per, Esra; Kurt, Bahtiyar
    Ecological systems are governed by complex interactions which are mainly nonlinear. In order to capture the inherent complexity and nonlinearity of ecological, and in general biological systems, empirical models recently gained popularity. However, although these models, particularly connectionist approaches such as multilayered backpropagation networks, are commonly applied as predictive models in ecology to a wide variety of ecosystems and questions, there are no studies to date aiming to assess the performance, both in terms of data fitting and generalizability, and applicability of empirical models in ecology. Our aim is hence to provide an overview for nature of the wide range of the data sets and predictive variables, from both aquatic and terrestrial ecosystems with different scales of time-dependent dynamics, and the applicability and robustness of predictive modeling methods on such data sets by comparing different empirical modeling approaches. The models used in this study range from predicting the occurrence of submerged plants in shallow lakes to predicting nest occurrence of bird species from environmental variables and satellite images. The methods considered include k-nearest neighbor (k-NN), linear and quadratic discriminant analysis (LDA and QDA), generalized linear models (GLM) feedforward multilayer backpropagation networks and pseudo-supervised network ARTMAP. Our results show that the predictive performances of the models on training data could be misleading, and one should consider the predictive performance of a given model on an independent test set for assessing its predictive power. Moreover, our results suggest that for ecosystems involving time-dependent dynamics and periodicities whose frequency are possibly less than the time scale of the data considered, GLM and connectionist neural network models appear to be most suitable and robust, provided that a predictive variable reflecting these time-dependent dynamics included in the model either implicitly or explicitly. For spatial data, which does not include any time-dependence comparable to the time scale covered by the data, on the other hand, neighborhood based methods such as k-NN and ARTMAP proved to be more robust than other methods considered in this study. in addition, for predictive modeling purposes, first a suitable, computationally inexpensive method should be applied to the problem at hand a good predictive performance of which would render the computational cost and efforts associated with complex variants unnecessary. (c) 2006 Elsevier B.V. All rights reserved.