Correction: Nonparametric conditional density estimation in a deep learning framework for short-term forecasting

Environmental and Ecological Statistics -     

A hybrid Wavelet-CNN-LSTM deep learning model for short-term urban water demand forecasting

● A novel deep learning framework for short-term water demand forecasting. ● Model prediction accuracy outperforms other traditional deep learning models. ● Wavelet multi-resolution analysis automatically extracts key water demand features. ● An analysis is performed to explain the improved mechanism of the proposed method. Short-term water demand forecasting provides guidance on real-time water allocation in the water supply network, which help water utilities reduce energy cost and avoid potential accidents. Although a variety of methods have been proposed to improve forecast accuracy, it is still difficult for statistical models to learn the periodic patterns due to the chaotic nature of the water demand data with high temporal resolution. To overcome this issue from the perspective of improving data predictability, we proposed a hybrid Wavelet-CNN-LSTM model, that combines time-frequency decomposition characteristics of Wavelet Multi-Resolution Analysis (MRA) and implement it into an advanced deep learning model, CNN-LSTM. Four models - ANN, Conv1D, LSTM, GRUN - are used to compare with Wavelet-CNN-LSTM, and the results show that Wavelet-CNN-LSTM outperforms the other models both in single-step and multi-steps prediction. Besides, further mechanistic analysis revealed that MRA produce significant effect on improving model accuracy.     

Lattice-based methods for regression and density estimation on complicated multidimensional regions

Environmental and Ecological Statistics - This paper illustrates the use of diffusion kernels to estimate smooth density and regression functions defined on highly complex domains. We generalize...     

A flocculation tensor to monitor water quality using a deep learning model

The increasing quantities of polluted waters are calling for advanced purification methods. Flocculation is an essential component of the water purification process, yet flocculation is commonly not optimal due to our poor understanding of the flocculation process. In particular, there is little knowledge on the mechanisms ruling the migration of pollutants during treatment. Here we have created the first tensor diagram, a mathematical framework for the flocculation process, analyzed its properties with a deep learning model, and developed a classification scheme for its relationship with pollutants. The tensor was constructed by combining pixel matrices from a variety of floc images, each with a particular flocculation period. Changing the factors used to make flocs images, such as coagulant dose and pH, resulted in tensors, which were used to generate matrices, that is the tensor diagram. Our deep learning algorithm employed a tensor diagram to identify pollution levels. Results show tensor map attributes with over 98% of sample images correctly classified. This approach offers potential to reduce the time delay of feedback from the flocculation process with deep learning categorization based on its clustering capabilities. The advantage of the tensor data from the flocculation process improves the efficiency and speed of response for commercial water treatment.

     

A framework for investigating illegal wildlife trade on social media with machine learning

Article impact statement: Machine learning can be used to automatically monitor and assess illegal wildlife trade on social media platforms.     

Nonparametric Ranked-set Sampling Confidence Intervals for Quantiles of a Finite Population

Ranked-set sampling from a finite population is considered in this paper. Three sampling protocols are described, and procedures for constructing nonparametric confidence intervals for a population quantile are developed. Algorithms for computing coverage probabilities for these confidence intervals are presented, and the use of interpolated confidence intervals is recommended as a means to approximately achieve coverage probabilities that cannot be achieved exactly. A simulation study based on finite populations of sizes 20, 30, 40, and 50 shows that the three sampling protocols follow a strict ordering in terms of the average lengths of the confidence intervals they produce. This study also shows that all three ranked-set sampling protocols tend to produce confidence intervals shorter than those produced by simple random sampling, with the difference being substantial for two of the protocols. The interpolated confidence intervals are shown to achieve coverage probabilities quite close to their nominal levels. Rankings done according to a highly correlated concomitant variable are shown to reduce the level of the confidence intervals only minimally. An example to illustrate the construction of confidence intervals according to this methodology is provided.     

Manipulating decay time for efficient large-mammal density estimation: gorillas and dung height

Large-mammal surveys often rely on indirect signs such as dung or nests. Sign density is usually translated into animal density using sign production and decay rates. In principle, such auxiliary variable estimates should be made in a spatially unbiased manner. However, traditional decay rate estimation methods entail following many signs from production to disappearance, which, in large study areas, requires extensive travel effort. Consequently, decay rate estimates have tended to be made instead at some convenient but unrepresentative location. In this study we evaluated how much bias might be induced by extrapolating decay rates from unrepresentative locations, how much effort would be required to implement current methods in a spatially unbiased manner, and what alternate approaches might be used to improve precision. To evaluate the extent of bias induced by unrepresentative sampling, we collected data on gorilla dung at several central African sites. Variation in gorilla dung decay rate was enormous, varying by up to an order of magnitude within and between survey zones. We then estimated what the effort-precision relationship would be for a previously suggested "retrospective" decay rate (RDR) method, if it were implemented in a spatially unbiased manner. We also evaluated precision for a marked sign count (MSC) approach that does not use a decay rate. Because they require repeat visits to remote locations, both RDR and MSC require enormous effort levels in order to gain precise density estimates. Finally, we examined an objective criterion for decay (i.e., dung height). This showed great potential for improving RDR efficiency because choosing a high threshold height for decay reduces decay time and, consequently, the number of visits that need to be made to remote areas. The ability to adjust decay time using an objective decay criterion also opens up the potential for a "prospective" decay rate (PDR) approach. Further research is necessary to evaluate whether the temporal bias inherent in such an approach is small enough to ignore, given the 10-20-fold increases in precision promised by a PDR approach.     

MSWNet: A visual deep machine learning method adopting transfer learning based upon ResNet 50 for municipal solid waste sorting

● MSWNet was proposed to classify municipal solid waste. ● Transfer learning could promote the performance of MSWNet. ● Cyclical learning rate was adopted to quickly tune hyperparameters. An intelligent and efficient methodology is needed owning to the continuous increase of global municipal solid waste (MSW). This is because the common methods of manual and semi-mechanical screenings not only consume large amount of manpower and material resources but also accelerate virus community transmission. As the categories of MSW are diverse considering their compositions, chemical reactions, and processing procedures, etc., resulting in low efficiencies in MSW sorting using the traditional methods. Deep machine learning can help MSW sorting becoming into a smarter and more efficient mode. This study for the first time applied MSWNet in MSW sorting, a ResNet-50 with transfer learning. The method of cyclical learning rate was taken to avoid blind finding, and tests were repeated until accidentally encountering a good value. Measures of visualization were also considered to make the MSWNet model more transparent and accountable. Results showed transfer learning enhanced the efficiency of training time (from 741 s to 598.5 s), and improved the accuracy of recognition performance (from 88.50% to 93.50%); MSWNet showed a better performance in MSW classsification in terms of sensitivity (93.50%), precision (93.40%), F1-score (93.40%), accuracy (93.50%) and AUC (92.00%). The findings of this study can be taken as a reference for building the model MSW classification by deep learning, quantifying a suitable learning rate, and changing the data from high dimensions to two dimensions.     

Sex-specific conditional mating preferences in a cichlid fish: implications for sexual conflict

Conditional mating strategies enable individuals to modulate their mating behaviour depending on ‘individual status’ to maximise fitness. Theory predicts that variation in individual quality can lead to differences in mating preferences. However, empirical evidence is scarce particular in terms of variation in male and female strategies. Here, we experimentally investigated quality-dependent variation in mating preferences concerning reliable quality indicators in Pelvicachromis taeniatus, a colourful cichlid fish with mutual mate choice and ornamentation. Males as well as females were artificially manipulated in phenotypic quality by different feeding regimes. Ornamentation was connected to individual quality in both sexes. Males and females showed conditional mating strategies in different directions. Males showed prudent choice by preferring females of similar quality. In contrast to males, low-quality females preferred highly ornamented males, whereas high-quality females showed neither preferences for high- nor for low-quality males. The results suggest that individuals aim for specific benefits depending on individual quality. Furthermore, the conflicting conditional mating preferences of males and females might lead to sexual conflict, implicating a highly dynamical mating system that evolves even in absence of environmental changes.     

Leveraging social media and deep learning to detect rare megafauna in video surveys

Deep learning has become a key tool for the automated monitoring of animal populations with video surveys. However, obtaining large numbers of images to train such models is a major challenge for rare and elusive species because field video surveys provide few sightings. We designed a method that takes advantage of videos accumulated on social media for training deep-learning models to detect rare megafauna species in the field. We trained convolutional neural networks (CNNs) with social media images and tested them on images collected from field surveys. We applied our method to aerial video surveys of dugongs (Dugong dugon) in New Caledonia (southwestern Pacific). CNNs trained with 1303 social media images yielded 25% false positives and 38% false negatives when tested on independent field video surveys. Incorporating a small number of images from New Caledonia (equivalent to 12% of social media images) in the training data set resulted in a nearly 50% decrease in false negatives. Our results highlight how and the extent to which images collected on social media can offer a solid basis for training deep-learning models for rare megafauna detection and that the incorporation of a few images from the study site further boosts detection accuracy. Our method provides a new generation of deep-learning models that can be used to rapidly and accurately process field video surveys for the monitoring of rare megafauna.     

Towards a co-evolutionary framework for implementing sustainable development

SUMMARY

A quantitative method which can be used to gauge the sustainability of existing socio-cultural, economic and environmental constructs is proposed. An underlying assumption of the approach is that a co-evolutionary relationship exists between social, cultural, economic and environmental systems and that these cannot be addressed in a reductionist and deterministic manner. The method focuses on both broad trends and patterns of variance in a multivariate data matrix, it allows the identification of dominant trends as well as underlying tensions, does not treat socio-cultural, economic and environmental heterogeneity as statistical noise and can be applied at different administrative scales. Furthermore, it may be used to identify potential sources of human insecurity, factors that may affect the probability of positive human survival, developmental patterns at odds with the established values system of civil society that can be targeted for political action, and to inform public debate. In addition, the method can be used to identify unexpected synergies between variables and may serve as an early warning system for regional planners.     

Forced copulation as a conditional alternative strategy in camel crickets

In the animal kingdom, males typically acquire reproductive access to females through courtship, nuptial gifts, or combat. In many species, small or poor-condition males that are unable to produce an attractive display, proffer an adequate mating gift, or dominate in male–male competition may resort to forced copulation. Forced copulation has been previously demonstrated in the camel cricket Pristoceuthophilus marmoratus and is uniquely facilitated by male hind leg weaponry that functions in both intrasexual competition and sexual coercion. Because male leg armaments are positively allometric, the largest males would seem best equipped to force-copulate, although they might be the most attractive to females, and so least likely to adopt a forced copulation strategy. In order to resolve this potential paradox and determine which males are in fact most likely to force-copulate, we manipulated male body size using diet and performed mating trials. Results showed that (1) male size increased with diet quality, (2) females appear to prefer larger males, and (3) smaller males were more likely to attempt forced copulation. Our results thus suggest that forced copulation is a condition-dependent alternative mating tactic used by small males unattractive to females.     

Accounting for large-scale factors in the study of understory vegetation using a conditional logistic model

Local-scale and large-scale factors can affect the presence of a species of understory vegetation in the forest. Local-scale factors may be the influence of surrounding trees, while climate and latitude are typically considered large-scale factors. A model for the presence of a species needs to take into account both scales. A conditional logistic model is proposed for those studies where only the local-scale factors are of interest and that avoids estimating the large-scale parameters. Conditioning is carried out by the number of quadrats in the plot where the vegetation is found. As the latter is a sufficient statistic for the large-scale factors, a model free from these parameters is obtained. Data gathered in the permanent sample plots of the 1985–1986 National Forest Inventory of Finland is used for illustration, where the local-scale factor of interest is the influence of the trees, quantified by an index based on the size and location of the trees. The model fitted to Vaccinium vitis-idaea showed a significant and positive influence of Scots pine on the presence of this species, while for Calamagrostis arundinacea, a decrease in the odds ratio was observed due to the influence of Norway spruce.     

Semi-parametric Bayesian density estimation using ranked set sample in the presence of ranking error

In this paper, we propose a Bayesian method to estimate the underlying density function of a study variable Y using a ranked set sample in which an auxiliary variable X is used to rank the sampling units. The amount of association between X and Y is not known, resulting in an unknown degree of ranking error. We assume that (X, Y) follows a Morgenstern family of distributions. The study variable Y is assumed to have a parametric distribution, with the distribution of the parameters having a Dirichlet process prior. A Markov chain Monte Carlo procedure is developed to obtain a Bayesian estimator of the desired density function as well as of the ranking error. A simulation study is used to evaluate the performance of the proposed method. An example from forestry is used to illustrate a real-life application of the proposed methodology.     

Towards a common Mediterranean framework for beach nourishment projects

Integrated Coastal Zone Management as a strategy for achieving conservation and sustainable multiple use of the coastal zone includes various types of management initiatives. Due to natural phenomena such as tides and winds and to social and economic activities, coastal areas undergo transformation. Coastal erosion and the disappearance of beaches as a result of wrong planning decisions and lack of effective legislation are among the most damaging effects and to reverse them requires application of a series of engineering techniques. Beach nourishment projects as a way towards shore protection and utilization through recreational purposes in the Spanish and Italian coasts are noteworthy in this respect. In beach nourishment projects, the roles of various entities, both public and private, should be clearly indicated and in the evaluation and execution stage a series of questions should be answered for the successful completion of any nourishment project. Past projects in the Mediterranean and experiences from recent Italian projects in Anzio and Nettuno confirm this. For example, dredging of the entrance channel of the port of Anzio enabled middle-sized ferryboats once again to enter the port (which had not been possible before). This will in turn increase the tourism potential of the town. Nourishment of two beaches at these sites prevented further erosion and provided more area for recreational purposes. Expected economic contribution of beach nourishments to the regional authority was estimated for Nettuno and Anzio; the resulting theoretical payback period was found to be 3 yr for the former and 15 yr for the latter.     

New multidimensional functional diversity indices for a multifaceted framework in functional ecology

Functional diversity is increasingly identified as an important driver of ecosystem functioning. Various indices have been proposed to measure the functional diversity of a community, but there is still no consensus on which are most suitable. Indeed, none of the existing indices meets all the criteria required for general use. The main criteria are that they must be designed to deal with several traits, take into account abundances, and measure all the facets of functional diversity. Here we propose three indices to quantify each facet of functional diversity for a community with species distributed in a multidimensional functional space: functional richness (volume of the functional space occupied by the community), functional evenness (regularity of the distribution of abundance in this volume), and functional divergence (divergence in the distribution of abundance in this volume). Functional richness is estimated using the existing convex hull volume index. The new functional evenness index is based on the minimum spanning tree which links all the species in the multidimensional functional space. Then this new index quantifies the regularity with which species abundances are distributed along the spanning tree. Functional divergence is measured using a novel index which quantifies how species diverge in their distances (weighted by their abundance) from the center of gravity in the functional space. We show that none of the indices meets all the criteria required for a functional diversity index, but instead we show that the set of three complementary indices meets these criteria. Through simulations of artificial data sets, we demonstrate that functional divergence and functional evenness are independent of species richness and that the three functional diversity indices are independent of each other. Overall, our study suggests that decomposition of functional diversity into its three primary components provides a meaningful framework for its quantification and for the classification of existing functional diversity indices. This decomposition has the potential to shed light on the role of biodiversity on ecosystem functioning and on the influence of biotic and abiotic filters on the structure of species communities. Finally, we propose a general framework for applying these three functional diversity indices.     

Connectivity for conservation: a framework to classify network measures

Graph theory, network theory, and circuit theory are increasingly being used to quantify multiple aspects of habitat connectivity and protected areas. There has been an explosive proliferation of network (connectivity) measures, resulting in over 60 measures for ecologists to now choose from. Conceptual clarification on the ecological meaning of these network measures and their interrelationships is overdue. We present a framework that categorizes network measures based on the connectivity property that they quantify (i.e., route-specific flux, route redundancy, route vulnerability, and connected habitat area) and the structural level of the habitat network to which they apply. The framework reveals a lack of network measures in the categories of "route-specific flux among neighboring habitat patches" and "route redundancy at the level of network components." We propose that network motif and path redundancy measures can be developed to fill the gaps in these categories. The value of this framework lies in its ability to inform the selection and application of network measures. Ultimately, it will allow a better comparison among graph, network, and circuit analyses, which will improve the design and management of connected landscapes.     

Long-term effects of short-term perturbation in a subalpine grassland

Theoretical advances and short-term experimental studies have furthered our understanding of how ecosystems respond to perturbation. However, there are few well-replicated experimental studies that allow an assessment of long-term responses. Results from a controlled, large-scale field experiment in a subalpine grassland near Interlaken, Switzerland, show that 2-4 years of liming (Ca: 40 g x m(-2) x yr(-1)) still significantly affected the composition of the vegetation and the soil microbial community nearly 70 years after the treatments were imposed, whereas NPK fertilization (8 g x m(-2) x yr(-1)) only marginally affected vegetation composition. The exchangeable content of Ca ions and soil pH were higher in limed plots but were unaffected in fertilized plots. Plant species and PLFAs (phospholipid fatty acids) indicating low pH values were found in higher abundance in the unlimed plots, suggesting that the long-lasting effects of liming on the above- and belowground communities were mediated through changes in soil pH. The results of this long-term study indicate that the resilience of mountain ecosystems may be particularly low in response to perturbations that substantially alter soil pH or other key determinants of belowground processes.