Ecological Networks as Conceptual Frameworks or Operational Tools in Conservation

Abstract:  The establishment of ecological networks (ENs) has been proposed as an ideal way to counteract the increasing fragmentation of natural ecosystems and as a necessary complement to the establishment of protected areas for biodiversity conservation. This conservation tool, which comprises core areas, corridors, and buffer areas, has attracted the attention of several national and European institutions. It is thought that ENs can connect habitat patches and thus enable species to move across unsuitable areas. In Europe, however, ENs are proposed as an oversimplification of complex ecological concepts, and we maintain that they are of limited use for biodiversity conservation for several reasons. The ENs are species specific and operate on species-dependent scales. In addition, the information needed for their implementation is only available for a handful of species. To overcome these limitations, ENs have been proposed on a landscape scale (and for selected "focal" species), but there is no indication that the structural composition of core areas, corridors, and buffer areas could ensure the functional connectivity and improve the viability of more than a few species. The theory behind ENs fails to provide sufficient practical information on how to build them (e.g., width, shape, structure, content). In fact, no EN so far has been validated in practice (ensuring connectivity and increasing overall biodiversity conservation), and there are no signs that validation will be possible in the near future. In view of these limitations, it is difficult to justify spending economic and political resources on building systems that are at best working hypotheses that cannot be evaluated on a practical level.     

Physical and Chemical Connectivity of Streams and Riparian Wetlands to Downstream Waters: A Synthesis

Streams, riparian areas, floodplains, alluvial aquifers, and downstream waters (e.g., large rivers, lakes, and oceans) are interconnected by longitudinal, lateral, and vertical fluxes of water, other materials, and energy. Collectively, these interconnected waters are called fluvial hydrosystems. Physical and chemical connectivity within fluvial hydrosystems is created by the transport of nonliving materials (e.g., water, sediment, nutrients, and contaminants) which either do or do not chemically change (chemical and physical connections, respectively). A substantial body of evidence unequivocally demonstrates physical and chemical connectivity between streams and riparian wetlands and downstream waters. Streams and riparian wetlands are structurally connected to downstream waters through the network of continuous channels and floodplain form that make these systems physically contiguous, and the very existence of these structures provides strong geomorphologic evidence for connectivity. Functional connections between streams and riparian wetlands and their downstream waters vary geographically and over time, based on proximity, relative size, environmental setting, material disparity, and intervening units. Because of the complexity and dynamic nature of connections among fluvial hydrosystem units, a complete accounting of the physical and chemical connections and their consequences to downstream waters should aggregate over multiple years to decades.     

多氯酚QSAR数值模型比较研究

应用多元线性回归分析和新近发展起来的人工神经网络方法进行了一类重要环境污染物多氯酚的定量构效关系研究，并用所建立的模型进行毒性预报，计算值与实验值的比较表明，前的相关系数约为０．９２，后的相关系数约为０．９９．后的百分误差地明显小于前，后的预报能力略好于前，中还讨论了后优于前的的原因。     

融合拓扑信息与气象信息的空气质量预测网络

通过空气质量监测数据对正在形成或即将到来的空气污染进行预测是一项具有重要意义的工作,而空气质量监测站只能检测其周围一定范围内的空气污染情况。为了衡量整个城市的空气污染情况,获取任意时间、任意位置的空气质量信息,结合交叉注意力机制,提出了一种融合拓扑信息与气象信息的空气质量预测网络(CGMIM)。将西安市空气质量监测数据与气象数据转换为图像拼接起来,作为输入信息。在高阶非线性时空动态神经网络(MIM)的基础上引入注意力机制,并增加拓扑图编码器模块,提高模型提取能力以及对空气质量监测数据中的空间特征的利用率。最后,使用时空损失函数替代传统的均方误差损失函数,提高模型对空间关系的关注。结果表明：CGMIM网络模型能够在准确预测的同时,对位置区域合理填充,能够有效提升空气质量监测数据的空间分辨率。     

Social network theory in the behavioural sciences: potential applications

Social network theory has made major contributions to our understanding of human social organisation but has found relatively little application in the field of animal behaviour. In this review, we identify several broad research areas where the networks approach could greatly enhance our understanding of social patterns and processes in animals. The network theory provides a quantitative framework that can be used to characterise social structure both at the level of the individual and the population. These novel quantitative variables may provide a new tool in addressing key questions in behavioural ecology particularly in relation to the evolution of social organisation and the impact of social structure on evolutionary processes. For example, network measures could be used to compare social networks of different species or populations making full use of the comparative approach. However, the networks approach can in principle go beyond identifying structural patterns and also can help with the understanding of processes within animal populations such as disease transmission and information transfer. Finally, understanding the pattern of interactions in the network (i.e. who is connected to whom) can also shed some light on the evolution of behavioural strategies.     

Analysis of pollutant levels in central Hong Kong applying neural network method with particle swarm optimization

Air pollution has emerged as an imminent issue in modernsociety. Prediction of pollutant levels is an importantresearch topic in atmospheric environment today. For fulfillingsuch prediction, the use of neural network (NN), and inparticular the multi-layer perceptrons, has presented to be acost-effective technique superior to traditional statisticalmethods. But their training, usually with back-propagation (BP)algorithm or other gradient algorithms, is often with certaindrawbacks, such as: 1) very slow convergence, and 2) easilygetting stuck in a local minimum. In this paper, a newlydeveloped method, particle swarm optimization (PSO) model, isadopted to train perceptrons, to predict pollutant levels, andas a result, a PSO-based neural network approach is presented. The approach is demonstrated to be feasible and effective bypredicting some real air-quality problems.     

Using a Bayesian network to clarify areas requiring research in a host–pathogen system

Bayesian network analyses can be used to interactively change the strength of effect of variables in a model to explore complex relationships in new ways. In doing so, they allow one to identify influential nodes that are not well studied empirically so that future research can be prioritized. We identified relationships in host and pathogen biology to examine disease‐driven declines of amphibians associated with amphibian chytrid fungus (Batrachochytrium dendrobatidis). We constructed a Bayesian network consisting of behavioral, genetic, physiological, and environmental variables that influence disease and used them to predict host population trends. We varied the impacts of specific variables in the model to reveal factors with the most influence on host population trend. The behavior of the nodes (the way in which the variables probabilistically responded to changes in states of the parents, which are the nodes or variables that directly influenced them in the graphical model) was consistent with published results. The frog population had a 49% probability of decline when all states were set at their original values, and this probability increased when body temperatures were cold, the immune system was not suppressing infection, and the ambient environment was conducive to growth of B. dendrobatidis. These findings suggest the construction of our model reflected the complex relationships characteristic of host–pathogen interactions. Changes to climatic variables alone did not strongly influence the probability of population decline, which suggests that climate interacts with other factors such as the capacity of the frog immune system to suppress disease. Changes to the adaptive immune system and disease reservoirs had a large effect on the population trend, but there was little empirical information available for model construction. Our model inputs can be used as a base to examine other systems, and our results show that such analyses are useful tools for reviewing existing literature, identifying links poorly supported by evidence, and understanding complexities in emerging infectious‐disease systems.     

Developing an empirical model of phytoplankton primary production: a neural network case study

We describe the development of a neural network model for estimating primary production of phytoplankton. Data from an enriched estuary in the eastern United States, Chesapeake Bay, were used to train, validate and test the model. Two error backpropagation multilayer perceptrons were trained: a simpler one (3-5-1) and a more complex one (12-5-1). Both neural networks outperformed conventional empirical models, even though only the latter, which exploits a larger suite of predictive variables, provided truly accurate outputs. The application of this neural network model is thoroughly discussed and the results of a sensitivity analysis are also presented.     

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.     

Using improved neural network model to analyze RSP,NOx and NO2 levels in urban air in Mong Kok,Hong Kong

As the health impact of air pollutants existing in ambient addresses much attention in recent years, forecasting of airpollutant parameters becomes an important and popular topic inenvironmental science. Airborne pollution is a serious, and willbe a major problem in Hong Kong within the next few years. InHong Kong, Respirable Suspended Particulate (RSP) and NitrogenOxides NO_x and NO₂ are major air pollutants due to thedominant diesel fuel usage by public transportation and heavyvehicles. Hence, the investigation and prediction of the influence and the tendency of these pollutants are ofsignificance to public and the city image. The multi-layerperceptron (MLP) neural network is regarded as a reliable andcost-effective method to achieve such tasks. The works presentedhere involve developing an improved neural network model, whichcombines the principal component analysis (PCA) technique and theradial basis function (RBF) network, and forecasting thepollutant levels and tendencies based in the recorded data. Inthe study, the PCA is firstly used to reduce and orthogonalizethe original input variables (data), these treated variables arethen used as new input vectors in RBF neural network modelestablished for forecasting the pollutant tendencies. Comparingwith the general neural network models, the proposed modelpossesses simpler network architecture, faster training speed,and more satisfactory predicting performance. This improvedmodel is evaluated by using hourly time series of RSP, NO_x and NO₂ concentrations collected at Mong Kok Roadside Gaseous Monitory Station in Hong Kong during the year 2000. By comparing the predicted RSP, NO_x and NO₂ concentrationswith the actual data of these pollutants recorded at the monitorystation, the effectiveness of the proposed model has been proven.Therefore, in authors' opinion, the model presented in the paper is a potential tool in forecasting air quality parameters and hasadvantages over the traditional neural network methods.