ensemble model

In this study, an Ensemble Model approach consisting of three different deep learning-based
architectures was used to classify white blood cells. The networks employed are MobileNetV2,
ResNet50, and EfficientNet-B0. MobileNetV2 is a lightweight model optimized for high accuracy and
fast operation on mobile devices and low-resource systems. ResNet50 is a widely preferred network in
image classification tasks, utilizing residual connections to mitigate learning problems in deep networks.
EfficientNet-B0 is a model with high parameter efficiency, offering a balanced approach to both
accuracy and computational efficiency through model scaling. Features obtained from these networks
were combined (feature fusion) for classification, leveraging the strengths of each network.
Additionally, the k-fold cross-validation method was applied to enhance the generalization capability of
the model. This ensemble approach provided a combination of strengths to achieve higher accuracy and
generalization compared to a single model.

In this paper, historical data from a wholesale alcoholic beverage distributor was used to forecast sales demand. Demand forecasting is a vital part of the sale and distribution of many goods. Accurate forecasting can be used to optimize inventory, improve cash flow, and enhance customer service. However, demand forecasting is a challenging task due to the many unknowns that can impact sales, such as the weather and the state of the economy. While many studies focus effort on modeling consumer demand and endpoint retail sales, this study focused on demand forecasting from the distributor perspective. An ensemble approach was applied using traditional statistical univariate time series models, multivariate models, and contemporary deep learning-based models. The final ensemble models for the most sold product and highest revenue grossing product were able to reduce sales forecasting error by nearly 50% and 33.5%, respectively, in comparison to a statistical näıve model. Additionally,...

Proactive management should be applied within a forest conservation context to prevent extinction or degradation of those forest ecosystems that we suspect will be affected by global warming in the next century. The aim of this study is to estimate the vulnerability under climate change of a localized and endemic tree species Pinus cembra that occurs in the alpine timberline. We used the Random Forest ensemble classifier and available bioclimatic and ecological data to model present and future suitable areas for P. cembra and estimate its current and future vulnerability. Future projections for years 2020, 2050 and 2080 were simulated using two IPCC Special Report on Emission Scenarios run under four global climate models. The suitability model described the optimal environmental conditions for P. cembra. Model scores (k = 0.77, sensitivity = 0.99 and specificity = 0.80) are robust. The main factors defining the model were Kira's warmth index and summer temperatures. Results show that there is potential for P. cembra to regenerate and persist in currently suitable areas. Future trends analysis suggested a cumulated mean loss of suitable areas of between 53% and 72% for different scenarios. All modeled projections predicted an upslope shift of the optimally suitable P. cembra belt and no downslope shift. We discuss environmental factors/plant interactions, the theoretical assumptions behind the model, model strengths and limitations, and we highlight the conservative traits of our analysis. The results suggest that forest management practices will play a fundamental role in the conservation of P. cembra habitats in the Alps.

The tree rhododendrons include the most widely distributed Himalayan Rhododendron species belonging to the subsection Arborea. Distributions of two members of this subspecies were modelled using bioclimatic data for current conditions (1950-2000). A subset of the least correlated bioclimatic variables was used for ecological niche modelling (ENM). We used an ENM ensemble method in the BiodiversityR R-package to map the suitable climatic space for tree rhododendrons based on 217 point location records. Ensemble bioclimatic models for tree rhododendrons had high predictive power with bioclimatic variables, which also separated the climatic spaces for the two species. Tree rhododendrons were found occurring in a wide range of climate and the distributional limits were associated with isothermality, temperature ranges, temperature of the wettest quarter, and precipitation of the warmest quarter of the year. The most suitable climatic space for tree rhododendrons was predicted to be in western Yunnan, China, with suitability declining towards the west and east. Its occurrence in a wide range of climatic settings with highly dissected habitats speaks to the adaptive capacity of the species, which might open up future options for their conservation planning in regions where they are listed as threatened.

Forecasting of electricity prices is important in deregulated electricity markets for all of the stakeholders: energy wholesalers, traders, retailers and consumers. Electricity price forecasting is an inherently difficult problem due to its special characteristic of dynamicity and non-stationarity. In this paper, we present a robust price forecasting mechanism that shows resilience towards the aggregate demand response effect and provides highly accurate forecasted electricity prices to the stakeholders in a dynamic environment. We employ an ensemble prediction model in which a group of different algorithms participates in forecasting 1-h ahead the price for each hour of a day. We propose two different strategies, namely, the Fixed Weight Method (FWM) and the Varying Weight Method (VWM), for selecting each hour's expert algorithm from the set of participating algorithms. In addition, we utilize a carefully engineered set of features selected from a pool of features extracted from the past electricity price data, weather data and calendar data. The proposed ensemble model offers better results than the Autoregressive Integrated Moving Average (ARIMA) method, the Pattern Sequence-based Forecasting (PSF) method and our previous work using Artificial Neural Networks (ANN) alone on the datasets for New York, Australian and Spanish electricity markets.

Coastal ecosystems are experiencing degradation from compound impacts of climate change and multiple anthropogenic disturbances. These pressures often act synergistically and complicate designing effective conservation measures; consequently, large-scale coastal restoration actions become a wicked problem. The purpose of this study was to use two different food web models in a coordinated manner to inform resource managers in their assessment of the ecological effects of a large-scale marsh restoration project. A team was formed that included the model developers and outside scientists, who were asked to use available model results of the calibrated simulations of an Ecopath with Ecosim (EwE) model and a Comprehensive Aquatic Systems Model (CASM), both designed to describe the structure and energetics of the Barataria Bay, Louisiana, United States food web. Both models offer somewhat different depictions of the predator-prey and competitive interactions of species within the food we...

The Mw 7.8 Gorkha earthquake of 25 April 2015 triggered thousands of landslides in the central part of the Nepal Himalayas. The main goal of this study was to generate an ensemble-based map of co-seismic landslide susceptibility in Sindhupalchowk District using model comparison and combination strands. A total of 2194 co-seismic landslides were identified and were randomly split into 1536 (~70%), to train data for establishing the model, and the remaining 658 (~30%) for the validation of the model. Frequency ratio, evidential belief function, and weight of evidence methods were applied and compared using 11 different causative factors (peak ground acceleration, epicenter proximity, fault proximity, geology, elevation, slope, plan curvature, internal relief, drainage proximity, stream power index, and topographic wetness index) to prepare the landslide susceptibility map. An ensemble of random forest was then used to overcome the various prediction limitations of the individual models. The success rates and prediction capabilities were critically compared using the area under the curve (AUC) of the receiver operating characteristic curve (ROC). By synthesizing the results of the various models into a single score, the ensemble model improved accuracy and provided considerably more realistic prediction capacities (91%) than the frequency ratio (81.2%), evidential belief function (83.5%) methods, and weight of evidence (80.1%).

A scientific environmental investment prediction plays a crucial role in controlling environmental pollution and avoiding the blind investment of environmental management. However, effective environmental investment prediction usually has to fact three challenges about diversiform indicators, insufficient data, and the reliability of prediction models. In the present study, a new prediction model is proposed using the extended belief rule-based system (EBRBS) and evidential reasoning (ER) rule, called ensemble EBRBS model, with the aim to overcome the above challenges for better environmental investment prediction. The proposed ensemble EBRBS model consists of two components: 1) multiple EBRBSs, which are constructed on the basis of not only using various feature selection methods to select representative indicators but also data increment transformation to enrich the training data; 2) an ER rule-based combination method, which utilizes the ER rule to accommodate the weights and reliabilities of different EBRBSs with the predicted outputs of these EBRBSs to have an integrated environmental investment prediction. A detailed case study is then provided for validating the proposed model via extensive experimental and comparison analysis based on the real-world environmental data about 25 environmental indicators for 31 provinces in China ranged from 2005 to 2018. The results demonstrate that the ensemble EBRBS model can be used as an effective model to accurately predict environmental investments. More importantly, the ensemble EBRBS model not only obtains a high accuracy better than some existing prediction models, but also has an excellent robustness compared with others under the situations of excessive indicators and insufficient data.

Heart disease is one of the most common diseases in middle-aged citizens. Among the vast number of heart diseases, coronary artery disease (CAD) is considered a common cardiovascular disease with a high death rate. The most popular tool for diagnosing CAD is the use of medical imaging, e.g., angiography. However, angiography is known for being costly and also associated with a number of side effects. Hence, the purpose of this study is to increase the accuracy of coronary heart disease diagnosis by selecting significant predictive features in order of their ranking. In this study, we propose an integrated method using machine learning. The machine learning methods of random trees (RTs), the decision tree of C5.0, support vector machine (SVM), the decision tree of Chi-squared automatic interaction detection (CHAID) are used in this study. The proposed method shows promising results and the study confirms that the RTs model outperforms other models.

In an effort to improve tools for effective flood risk assessment, we applied machine learning algorithms to predict flood-prone areas in Amol city (Iran), a site with recent floods (2017–2018). An ensemble approach was then implemented to predict hazard probabilities using the best machine learning algorithms (boosted regression tree, multivariate adaptive regression spline, generalized linear model, and generalized additive model) based on a receiver operator characteristic-area under the curve (ROC-AUC) assessment. The algorithms were all trained and tested on 92 randomly selected points, information from a flood inundation survey, and geospatial predictor variables (precipitation, land use, elevation, slope percent, curve number, distance to river, distance to channel, and depth to groundwater). The ensemble model had 0.925 and 0.892 accuracy for training and testing data, respectively. We then created a vulnerability map from data on building density, building age, population d...

The distributions of animal populations change and evolve through time. Migratory species exploit different habitats at different times of the year. Biotic and abiotic features that determine where a species lives vary due to natural and anthropogenic factors. This spatiotemporal variation needs to be accounted for in any modeling of species' distributions. In this paper we introduce a semiparametric model that provides a flexible framework for analyzing dynamic patterns of species occurrence and abundance from broadscale survey data. The spatiotemporal exploratory model (STEM) adds essential spatiotemporal structure to existing techniques for developing species distribution models through a simple parametric structure without requiring a detailed understanding of the underlying dynamic processes. STEMs use a multi-scale strategy to differentiate between local and globalscale spatiotemporal structure. A user-specified species distribution model accounts for spatial and temporal patterning at the local level. These local patterns are then allowed to ''scale up'' via ensemble averaging to larger scales. This makes STEMs especially well suited for exploring distributional dynamics arising from a variety of processes. Using data from eBird, an online citizen science bird-monitoring project, we demonstrate that monthly changes in distribution of a migratory species, the Tree Swallow (Tachycineta bicolor), can be more accurately described with a STEM than a conventional bagged decision tree model in which spatiotemporal structure has not been imposed. We also demonstrate that there is no loss of model predictive power when a STEM is used to describe a spatiotemporal distribution with very little spatiotemporal variation; the distribution of a nonmigratory species, the Northern Cardinal (Cardinalis cardinalis).

Advances in close-range and remote sensing technologies drive innovations in forest resource assessments and monitoring at varying scales. Data acquired with airborne and spaceborne platforms provide us with higher spatial resolution, more frequent coverage and increased spectral information. Recent developments in ground-based sensors have advanced three dimensional (3D) measurements, low-cost permanent systems and community-based monitoring of forests. The REDD+ mechanism has moved the remote sensing community in advancing and developing forest geospatial products which can be used by countries for the international reporting and national forest monitoring. However, there still is an urgent need to better understand the options and limitations of remote and close-range sensing techniques in the field of degradation and forest change assessment. This Special Issue contains 12 studies that provided insight into new advances in the field of remote sensing for forest management and RE...

The tree rhododendrons include the most widely distributed Himalayan Rhododendron species belonging to the subsection Arborea. Distributions of two members of this subspecies were modelled using bioclimatic data for current conditions (1950-2000). A subset of the least correlated bioclimatic variables was used for ecological niche modelling (ENM). We used an ENM ensemble method in the BiodiversityR R-package to map the suitable climatic space for tree rhododendrons based on 217 point location records. Ensemble bioclimatic models for tree rhododendrons had high predictive power with bioclimatic variables, which also separated the climatic spaces for the two species. Tree rhododendrons were found occurring in a wide range of climate and the distributional limits were associated with isothermality, temperature ranges, temperature of the wettest quarter, and precipitation of the warmest quarter of the year. The most suitable climatic space for tree rhododendrons was predicted to be in western Yunnan, China, with suitability declining towards the west and east. Its occurrence in a wide range of climatic settings with highly dissected habitats speaks to the adaptive capacity of the species, which might open up future options for their conservation planning in regions where they are listed as threatened.

In this study paper, the feasibility of constructing a complete smart system for anticipating electrical power consumption is created, as electricity's market share is expected to expand over the future decades. Smart grids and smart meters will help utility companies and their customers soon. New services and businesses in energy management need software development and data analytics skills. New services and enterprises are competitive. The project's electricity consumers are categorized by their hourly power usage percentage. This classification was done using data mining (five algorithms in specific) and data analysis theory. This division aims to help each group minimize energy use and expenditures, encourage energy-saving activities, and promote consumer involvement by giving tailored guidance. The intended segmentation is done through an iterative process using a computer classification computation, post-analysis, and data mining with visualization and statistical methodologies.

Groundwater is an important natural resource in arid and semi-arid environments, where discharge from karst springs is utilized as the principal water supply for human use. The occurrence of karst springs over large areas is often poorly documented, and interpolation strategies are often utilized to map the distribution and discharge potential of springs. This study develops a novel method to delineate karst spring zones on the basis of various hydrogeological factors. A case study of the Bojnourd Region, Iran, where spring discharge measurements are available for 359 sites, is used to demonstrate application of the new approach. Spatial mapping is achieved using ensemble modelling, which is based on certainty factors (CF) and logistic regression (LR). Maps of the CF and LR components of groundwater potential were generated individually, and then, combined to prepare an ensemble map of the study area. The accuracy (A) of the ensemble map was then assessed using area under the receiv...

Predicting the potential distribution of medicinal plants in response to climate change is essential for their conservation and management. Contributing to the management program, this study aimed to predict the distribution of two threatened medicinal plants, Fritillaria cirrhosa and Lilium nepalense. The location of focal species gathered from herbarium specimen housed in different herbaria and online databases were geo-referenced and checked for spatial autocorrelation. The predictive environmental variables were selected, and MaxEnt software was used to model the current and future distributions of focal species. Four Representative Concentration Pathway (RCP) trajectories of the BCC-CSM1.1 model were used as the future (2050) projection layer. The MaxEnt modelling delineated the potential distribution of F. cirrhosa and L. nepalense. The current suitability is projected towards Central and Eastern Hilly/Mountainous regions. Both species gain maximum suitability in RCP 4.5 which decline towards other trajectories for L. nepalense. Overall, both the focal species shift towards the northwest , losing their potential habitat in hilly and lower mountainous regions by 2050 across all trajectories. Our results highlight the impact of future climate change on two threatened and valuable species. The results can be further useful to initiate farming of these medicinally and economically important species based on climatically suitable zone and for designing a germplasm conservation strategy.

Over the previous decade, energy usage has increased exponentially all over the world. The machine learning algorithms are used to classify the demand and  requirement  of  off  and  evening  peak  load  of  southern  regional  load dispatch  centre  (SRLDC)  data.  In  this  paper,  data  are  classified  based  on demand and requirement of both evening and off peak of day wise southern regional grid of Andhra Pradesh, Karnataka, Kerala, Tamilnadu, and Pondicherry of different states are proposed. The machine learning algorithms  like  k-nearest  neighbors  (KNN),  random  forest,  and  logistic regression  have  been  adopted  to classify  the  model. To  improve  this  model efficiency,  an  ensemble  learning  method  is  used  to  increase  the  accuracy. The performance measure of state-wise outcome is determined by classifying its demand and requirement needs over its state energy consumption  and  with  different  classification  algorithms  and  it  is  improved by using a combined method of ensemble model with accuracy of 86%.

Pandanus unguifer is a threatened species endemic to Sikkim and Darjeeling district of West Bengal. The species bears mildly fragrant creamy white colored flower and is the only Pandanus species that produces flowers in potted condition and hence, is of ornamental value. We used field collected-coordinates of five presence occurrence localities of the species for habitat distribution modeling. Additionally, downloaded environmental data from worldclim database were used for correlation. Amongst the input environmental variables, Mean diurnal range in temperature (Bio2) and Precipitation seasonality (CV) (Bio15) were influential factors and contributed 48.7% and 40.8% respectively to the MaxEnt model. The Rest of the remaining layers collectively contributed 10.5% to the final habitat model. Of the total 5963 sq.km suitable area only 153 sq. km has a very high probability of having P. unguifer which is only 2.57% of the total area. The study revealed the highest potential distributio...

In this study, the self-adaptive evolutionary (SaE) agent is employed to structure the contributing elements to process the management of extreme learning machine (ELM) architecture based on a logical procedure. In fact, the SaE algorithm is utilized for possibility of enhancing the performance of the ELM to estimate daily soil temperature (ST) at 6 different depths of 5, 10, 20, 30, 50 and 100 cm. In the developed SaE-ELM model, the network hidden node parameters of the ELM are optimized using SaE algorithm. The precision of the SaE-ELM is then compared with the ELM model. Daily weather data sets including minimum, maximum and average air temperatures (T min , T max and T avg), atmospheric pressure (P) and global solar radiation (R S) collected for two Iranian stations of Bandar Abbas and Kerman with different climate conditions have been utilized. After primary evaluation, T min , T max and T avg are considered as final inputs for the ELM and SaE-ELM models due to their high correlations with ST at all depths. The achieved results for both stations reveal that both ELM and SaE-ELM models offer desirable performance to estimate daily ST at all depths; nevertheless, a slightly more precision can be obtained by the SaE-ELM model. The performance of the ELM and SaE-ELM models are verified against genetic programming (GP) and artificial neural network (ANN) models developed in this study. For Bandar Abbass station, the obtained mean absolute bias error (MABE) and correlation coefficient (R) for the ELM model at different depths are in the range of 0.9116-1.5988°C and 0.9023-0.9840, respectively while for the SaE-ELM model they are in the range of 0.8660-1.5338°C and 0.9084-0.9893, respectively. In addition, for Kerman Station the attained MABE and RMSE for the ELM model vary from 1.6567 to 2.4233°C and 0.8661 to 0.9789, respectively while for the SaE-ELM model they vary from 1.5818 to 2.3422°C and 0.8736 to 0.9831, respectively.

The tree rhododendrons include the most widely distributed Himalayan Rhododendron species belonging to the subsection Arborea. Distributions of two members of this subspecies were modelled using bioclimatic data for current conditions . A subset of the least correlated bioclimatic variables was used for ecological niche modelling (ENM). We used an ENM ensemble method in the BiodiversityR R-package to map the suitable climatic space for tree rhododendrons based on 217 point location records. Ensemble bioclimatic models for tree rhododendrons had high predictive power with bioclimatic variables, which also separated the climatic spaces for the two species. Tree rhododendrons were found occurring in a wide range of climate and the distributional limits were associated with isothermality, temperature ranges, temperature of the wettest quarter, and precipitation of the warmest quarter of the year. The most suitable climatic space for tree rhododendrons was predicted to be in western Yunnan, China, with suitability declining towards the west and east. Its occurrence in a wide range of climatic settings with highly dissected habitats speaks to the adaptive capacity of the species, which might open up future options for their conservation planning in regions where they are listed as threatened.

This study focused on the rare and threatened plant species eastern turkeybeard (Xerophyllum asphodeloides (L.) Nutt.) and its presence or absence in the Talladega National Forest in Alabama, USA. An ensemble suitable habitat map was developed using four different modeling methods (MaxEnt, Generalized Linear Model, Generalized Additive Model, and Random Forest). AUC evaluation scores for each model were 0.99, 0.96, 0.98, and 0.99, respectively. Biserial correlation scores for models ranged from 0.71 (GLM) to 0.94 (RF). The four different models agreed suitable habitat was found to cover 159.57 ha of the land. The ground slope variable was the most contributive variable in the MaxEnt and RF models and was also significant in the GLM and GAM models. The knowledge gained from this research can be used to establish and implement habitat suitability strategies across the Talladega National Forest and similar ecosystems in the southern United States.

The vital essence of evolutionary learning consists of information flows between the environment and the entities differentially surviving and reproducing therein. Gain or loss of information in individuals and populations due to evolutionary steps should be considered in evolutionary algorithm theory and practice. Information theory has rarely been applied to evolutionary computation – a lacuna that this dissertation addresses, with an emphasis on objectively and explicitly evaluating the ensemble models implicit in evolutionary learning. Information theoretic functionals can provide objective, justifiable, general, computable, commensurate measures of fitness and diversity. We identify information transmission channels implicit in evolutionary learning. We define information distance metrics and indices for ensembles. We extend Price‟s Theorem to non-random mating, give it an effective fitness interpretation and decompose it to show the key factors influencing heritability and evo...

This paper presents an information theoretic perspective on design and analysis of evolutionary algorithms. Indicators of solution quality are developed and applied not only to individuals but also to ensembles, thereby ensuring information diversity. Price's Theorem is extended to show how joint indicators can drive reproductive sampling rate of potential parental pairings. Heritability of mutual information is identified as a key issue.

The research upon which this dissertation is based has occupied many years and consumed a significant share not only of my life but also of the lives of several other individuals who have very patiently guided and supported me. First are my parents, Philip and Marilyn Card, who provided my earliest education, imbued me with a love of learning, and by example invested me with a desire to serve my fellow creatures by contributing to the store of human knowledge and its practical applications. Second collectively are the many teachers in schools both formal and informal who labored to inform me in the full sense of the word. Third are my coworkers at Critical Technologies Inc., who have understood my need to pursue this research, despite the time it has taken me away from work having been greatly to the detriment of our business, and who have struggled to fill the gaps my frequent absences have left. Fourth is my advisor Professor Chilukuri Mohan, who has understood my need to try to fulfill my obligations to my coworkers and others outside academia, despite the time these have taken me away from this research having greatly delayed this dissertation, and who never failed in any of our meetings to encourage me and lead me to some new insight that I would not have found without his guidance. Fifth and always last because she makes that sacrifice for me but never least because I love her for that and more, is my wife Brenda Card, who has worked outside our home to support us and done all the tedious chores of our life together while I have done my work. While I have enjoyed the thrill of learning, they have borne my burdens for many years, and I can thank them only by dedicating my work to free minds.

In this study, we propose a novel nonlinear ensemble forecasting model integrating generalized linear autoregression (GLAR) with artificial neural networks (ANN) in order to obtain accurate prediction results and ameliorate forecasting performances. We compare the new model's performance with the two individual forecasting models-GLAR and ANN-as well as with the hybrid model and the linear combination models. Empirical results obtained reveal that the prediction using the nonlinear ensemble model is generally better than those obtained using the other models presented in this study in terms of the same evaluation measurements. Our findings reveal that the nonlinear ensemble model proposed here can be used as an alternative forecasting tool for exchange rates to achieve greater forecasting accuracy and improve prediction quality further.

In this study, we propose a novel nonlinear ensemble forecasting model integrating generalized linear autoregression (GLAR) with artificial neural networks (ANN) in order to obtain accurate prediction results and ameliorate forecasting performances. We compare the new model's performance with the two individual forecasting models-GLAR and ANN-as well as with the hybrid model and the linear combination models. Empirical results obtained reveal that the prediction using the nonlinear ensemble model is generally better than those obtained using the other models presented in this study in terms of the same evaluation measurements. Our findings reveal that the nonlinear ensemble model proposed here can be used as an alternative forecasting tool for exchange rates to achieve greater forecasting accuracy and improve prediction quality further.

Due to recent financial crisis and regulatory concerns of Basel II, credit risk assessment is becoming one of the most important topics in the field of financial risk management. Quantitative credit scoring models are widely used tools for credit risk assessment in financial institutions. Although single support vector machines (SVM) have been demonstrated with good performance in classification, a single classifier with a fixed group of training samples and parameters setting may have some kind of inductive bias. One effective way to reduce the bias is ensemble model. In this study, several ensemble models based on least squares support vector machines (LSSVM) are brought forward for credit scoring. The models are tested on two real world datasets and the results show that ensemble strategies can help to improve the performance in some degree and are effective for building credit scoring models.

The tree rhododendrons include the most widely distributed Himalayan Rhododendron species belonging to the subsection Arborea. Distributions of two members of this subspecies were modelled using bioclimatic data for current conditions (1950-2000). A subset of the least correlated bioclimatic variables was used for ecological niche modelling (ENM). We used an ENM ensemble method in the BiodiversityR R-package to map the suitable climatic space for tree rhododendrons based on 217 point location records. Ensemble bioclimatic models for tree rhododendrons had high predictive power with bioclimatic variables, which also separated the climatic spaces for the two species. Tree rhododendrons were found occurring in a wide range of climate and the distributional limits were associated with isothermality, temperature ranges, temperature of the wettest quarter, and precipitation of the warmest quarter of the year. The most suitable climatic space for tree rhododendrons was predicted to be in western Yunnan, China, with suitability declining towards the west and east. Its occurrence in a wide range of climatic settings with highly dissected habitats speaks to the adaptive capacity of the species, which might open up future options for their conservation planning in regions where they are listed as threatened.

Boreal species sensitive to the timing and duration of snow cover are particularly vulnerable to global climate change. Recent work has shown a link between wolverine (Gulo gulo) habitat and persistent spring snow cover through 15 May, the approximate end of the wolverine's reproductive denning period. We modeled the distribution of snow cover within the Columbia, Upper Missouri, and Upper Colorado River Basins using a downscaled ensemble climate model. The ensemble model was based on the arithmetic mean of 10 global climate models (GCMs) that best fit historical climate trends and patterns within these three basins. Snow cover was estimated from resulting downscaled temperature and precipitation patterns using a hydrologic model. We bracketed our ensemble model predictions by analyzing warm (miroc 3.2) and cool (pcm1) downscaled GCMs. Because Moderate-Resolution Imaging Spectroradiometer (MODIS)-based snow cover relationships were analyzed at much finer grain than downscaled GCM output, we conducted a second analysis based on MODIS-based snow cover that persisted through 29 May, simulating the onset of spring two weeks earlier in the year. Based on the downscaled ensemble model, 67% of predicted spring snow cover will persist within the study area through 2030-2059, and 37% through 2070-2099. Estimated snow cover for the ensemble model during the period 2070-2099 was similar to persistent MODIS snow cover through 29 May. Losses in snow cover were greatest at the southern periphery of the study area (Oregon, Utah, and New Mexico, USA) and least in British Columbia, Canada. Contiguous areas of spring snow cover become smaller and more isolated over time, but large (.1000 km 2) contiguous areas of wolverine habitat are predicted to persist within the study area throughout the 21st century for all projections. Areas that retain snow cover throughout the 21st century are British Columbia, north-central Washington, northwestern Montana, and the Greater Yellowstone Area. By the late 21st century, dispersal modeling indicates that habitat isolation at or above levels associated with genetic isolation of wolverine populations becomes widespread. Overall, we expect wolverine habitat to persist throughout the species range at least for the first half of the 21st century, but populations will likely become smaller and more isolated.

The distributions of animal populations change and evolve through time. Migratory species exploit different habitats at different times of the year. Biotic and abiotic features that determine where a species lives vary due to natural and anthropogenic factors. This spatiotemporal variation needs to be accounted for in any modeling of species' distributions. In this paper we introduce a semiparametric model that provides a flexible framework for analyzing dynamic patterns of species occurrence and abundance from broadscale survey data. The spatiotemporal exploratory model (STEM) adds essential spatiotemporal structure to existing techniques for developing species distribution models through a simple parametric structure without requiring a detailed understanding of the underlying dynamic processes. STEMs use a multi-scale strategy to differentiate between local and globalscale spatiotemporal structure. A user-specified species distribution model accounts for spatial and temporal patterning at the local level. These local patterns are then allowed to ''scale up'' via ensemble averaging to larger scales. This makes STEMs especially well suited for exploring distributional dynamics arising from a variety of processes. Using data from eBird, an online citizen science bird-monitoring project, we demonstrate that monthly changes in distribution of a migratory species, the Tree Swallow (Tachycineta bicolor), can be more accurately described with a STEM than a conventional bagged decision tree model in which spatiotemporal structure has not been imposed. We also demonstrate that there is no loss of model predictive power when a STEM is used to describe a spatiotemporal distribution with very little spatiotemporal variation; the distribution of a nonmigratory species, the Northern Cardinal (Cardinalis cardinalis).

A genetic-based neural network ensemble (GNNE) is applied for estimation of daily soil temperatures (DST) at distinct depths. A sequential genetic-based negative correlation learning algorithm (SGNCL) is adopted to train the GNNE parameters. CLMS algorithm is used to achieve the optimum weights of components. Recorded data for two different stations located in Iran are used for the development of the GNNE models. Furthermore, the GNNE predictions are compared with the existing machine-learning models. The results demonstrate that GNNE outperforms other methods for the prediction of DSTs.