Ensemble Model Results Research Articles

Predicting the Potential Distribution of Rare and Endangered Emmenopterys henryi in China Under ClimateChange.

Climate change has a pivotal impact on the potential distribution of endangered and relic tree species. Probably due to unrepresentative sampling and single algorithm, at present, there are different views on the potential range of the endangered tree Emmenopterys henryi, endemic to China. Here, we first collated 612 occurrence records and 22 environmental variables including climate, topography, and soil. Combined the Biomod2 with MaxEnt, we then predicted its past, current, and future potential suitable area in China, and determined the key factors influencing its distribution. The ensemble model results showed that the main environmental variables affecting this species were the minimum temperature of the coldest month (BIO6), precipitation of warmest quarter (BIO18), and temperature seasonality (BIO4). Its current potential distribution area was 176.53 × 104 km2, mainly concentrated in eastern, central, and southwestern China. Collectively, the suitable area of E. henryi would averagely decrease by 3.90% in all 16 future scenarios, with its centroid largely migrating northeastward. Our findings indicate that the endangered E. henryi covered 18 provinces in China, having a larger area than known. Moreover, climate change may have an adverse effect on its potential distribution. In addition, the ensemble model can produce more effective prediction outcomes than MaxEnt for such endemic tree species with large environmental range. We recommend increasing sample representativeness by analyzing the completeness properties of sample coverage, and simultaneously selecting appropriate algorithms to ensure the reliability of distribution prediction for endangered and relict tree species.

Research and development of algorithms for the formation of an effective ensemble of convolutional neural networks for image classification

The object of the research is artificial neural networks (ANN) with convolutional architecture for image classification. The subject of the research is the study and development of algorithms for constructing ensembles of convolutional neural networks (SNS) in conditions of limited training sample. The aim of the study is to develop an algorithm for the formation of an effective model based on an ensemble of convolutional SNS using methods of averaging the results of each model, capable of avoiding overfitting in the process of improving the accuracy of the forecast and trained on a small amount of data, less than 10 thousand examples. As a basic network, an effective SNA architecture was developed as part of the ensemble, which showed good results as a single model. The article also examines methods for combining the results of ensemble models and provides recommendations for the formation of the SNA architecture. The research methods used are the theory of neural networks, the theory of machine learning, artificial intelligence, methods of algorithmization and programming of machine learning models, a comparative analysis of models based on different algorithms using classical ensembling with simple averaging and combining the results of basic algorithms in conditions of limited sampling, taking into account weighted average. The field of application of the obtained algorithm and model is medical diagnostics in medical institutions, sanatoriums during primary diagnostic admission, using the example of a research task, the model is trained to classify dermatological diseases according to input photographs. The novelty of the study lies in the development of an effective algorithm and image classification model based on an ensemble of convolutional NS that exceed the prediction accuracy of basic classifiers, the process of retraining an ensemble of classifiers with deep architecture on a small sample volume is investigated, from which conclusions are drawn on the design of an optimal network architecture and the choice of methods for combining the results of several basic classifiers. As a result of the research, an algorithm has been developed for the formation of an ensemble of SNS based on an effective basic architecture and weighted average averaging of the results of each model for the classification task of image recognition in conditions of limited sampling.

The possible impact of human activity and climate change on the potential suitable habitats of Taxus wallichianavar.mairei (Taxaceae) evaluated by ensemble modeling techniques

In the current work, we examined the possible changes in potential suitable habitats of Taxus wallichiana var. mairei (Taxaceae) under the influence of human activity and climate change using ensemble modeling techniques. T. wallichiana var. mairei is an endangered subtropical coniferous evergreen tree and is often used in medical applications. It is mainly distributed in southern China and has experienced several degenerations due to human activity in recent decades. Its growth is also very sensitive to climate change. The area of its highly suitable habitat under current climate conditions was approximately 2.31 × 106 km2, as estimated by our ensemble models, accounting for 81.85% of its total suitable habitat. The ensemble model results showed that the mean diurnal range (bio2) was the most critical environment variable affecting the performance of the ensemble models. Human activity had an overall negative influence on the suitable habitats of T. wallichiana var. mairei under current climate conditions. Under the influence of future climate change, the ensemble model predicted that the highly suitable habitat would convert to low- to medium-suitability or be lost, with the predicted loss of highly suitable habitat at the end of this century ranging from 30.40% to 96.55%. The predicted changes were more intense with increases in the severity of global warming. In addition, the ensemble model also predicted a shift in the distribution of total suitable habitats toward higher elevations. The results of this study should provide information for conservation purposes.

Dynamic Analysis of Regional Wheat Stripe Rust Environmental Suitability in China

Stripe rust is one of the most destructive wheat diseases in China, negatively affecting the production safety and causing yield losses of wheat. Thus, it is important to analyze the environmental suitability and dynamic changes of wheat stripe rust in China. The occurrence of stripe rust is affected by multiple factors. Therefore, this study combined data from various disciplinary fields such as remote sensing, meteorology, biology, and plant protection to evaluate the environmental suitability of stripe rust in China using species distribution models. The study also discusses the importance and effect of various variables. Results revealed that meteorological factors had the greatest impact on the occurrence of stripe rust, especially temperature and precipitation. Wheat growth factors have a greater impact from April to August. Elevation has a greater impact in summer. The ensemble model results were better than the single model, with TSS and AUC greater than 0.851 and 0.971, respectively. Overlapping analysis showed that the winter stripe rust suitable areas were mainly in the Sichuan Basin, Northwestern Hubei, Southern Shaanxi, and Southern Henan wheat areas. In spring, the suitable areas of stripe rust increased in Huang-Huai-Hai and the middle and lower reaches of the Yangtze River and Guanzhong Plain, and the development of northwestern wheat areas such as Xinjiang and Gansu slightly lagged behind. In summer, wheat threatened by stripe rust is mainly in late-ripening spring wheat areas in Gansu, Ningxia, Qinghai, and Xinjiang. This study can provide a scientific basis for optimizing and improving the comprehensive management strategy of stripe rust.

Assessment of Ensemble Learning to Predict Wheat Grain Yield Based on UAV-Multispectral Reflectance

Grain yield is increasingly affected by climate factors such as drought and heat. To develop resilient and high-yielding cultivars, high-throughput phenotyping (HTP) techniques are essential for precise decisions in wheat breeding. The ability of unmanned aerial vehicle (UAV)-based multispectral imaging and ensemble learning methods to increase the accuracy of grain yield prediction in practical breeding work is evaluated in this study. For this, 211 winter wheat genotypes were planted under full and limited irrigation treatments, and multispectral data were collected at heading, flowering, early grain filling (EGF), and mid-grain filling (MGF) stages. Twenty multispectral vegetation indices (VIs) were estimated, and VIs with heritability greater than 0.5 were selected to evaluate the models across the growth stages under both irrigation treatments. A framework for ensemble learning was developed by combining multiple base models such as random forest (RF), support vector machine (SVM), Gaussian process (GP), and ridge regression (RR). The R2 values between VIs and grain yield for individual base models were ranged from 0.468 to 0.580 and 0.537 to 0.598 for grain yield prediction in full and limited irrigation treatments across growth stages, respectively. The prediction results of ensemble models were ranged from 0.491 to 0.616 and 0.560 to 0.616 under full and limited irrigation treatments respectively, and were higher than that of the corresponding base learners. Moreover, the grain yield prediction results were observed high at mid grain filling stage under both full (R2 = 0.625) and limited (R2 = 0.628) irrigation treatments through ensemble learning based stacking of four base learners. Further improvements in ensemble learning models can accelerate the use of UAV-based multispectral data for accurate predictions of complex traits like grain yield in wheat.

Cost-sensitive probability for weighted voting in an ensemble model for multi-class classification problems

Ensemble learning is an algorithm that utilizes various types of classification models. This algorithm can enhance the prediction efficiency of component models. However, the efficiency of combining models typically depends on the diversity and accuracy of the predicted results of ensemble models. However, the problem of multi-class data is still encountered. In the proposed approach, cost-sensitive learning was implemented to evaluate the prediction accuracy for each class, which was used to construct a cost-sensitivity matrix of the true positive (TP) rate. This TP rate can be used as a weight value and combined with a probability value to drive ensemble learning for a specified class. We proposed an ensemble model, which was a type of heterogenous model, namely, a combination of various individual classification models (support vector machine, Bayes, K-nearest neighbour, naïve Bayes, decision tree, and multi-layer perceptron) in experiments on 3-, 4-, 5- and 6-classifier models. The efficiencies of the propose models were compared to those of the individual classifier model and homogenous models (Adaboost, bagging, stacking, voting, random forest, and random subspaces) with various multi-class data sets. The experimental results demonstrate that the cost-sensitive probability for the weighted voting ensemble model that was derived from 3 models provided the most accurate results for the dataset in multi-class prediction. The objective of this study was to increase the efficiency of predicting classification results in multi-class classification tasks and to improve the classification results.

Application of machine learning techniques in groundwater potential mapping along the west coast of India

ABSTRACT Groundwater potential mapping (GWPM) in the coastal zone is crucial for the planning and development of society and the environment. The current study is aimed to map the groundwater potential zones of Sindhudurg coastal stretch on the west coast of India, using three machine learning models: random forest (RF), boosted regression tree (BRT), and the ensemble of RF and support vector machine (SVM). In order to achieve the objective, 15 groundwater influencing factors including elevation, slope, aspect, slope length (LS), profile curvature, plan curvature, topographical wetness index (TWI), distance from streams, distance from lineaments, lithology, geomorphology, soil, land use, normalized difference vegetation index (NDVI), and rainfall were considered for inter-thematic correlations and overlaid with spring and well occurrences in a spatial database. A total of 165 spring and well locations were identified, which had been divided into two classes: training and validation, at the ratio of 70:30, respectively. The RF, BRT, and RF-SVM ensemble models have been applied to delineate the groundwater potential zones and categorized into five classes, namely very high, high, moderate, low, and very low. RF, BRT, and ensemble model results showed that 33.3%, 35.6%, and 36.8% of the research area had a very high groundwater potential zone. These models were validated with area under the receiver operating characteristics (AUROC) curve. The accuracy of RF (94%) and hybrid model (93.4%) was more efficient than BRT (89.8%) model. In order to further evaluate and validate, four different sites were subsequently chosen, and we obtained similar results, ensuring the validity of the applied models. Additionally, ground-penetrating radar (GPR) technique was applied to predict the groundwater table and validated by measured wells. The mean difference between measured and GPR predicted groundwater table was 14 cm, which reflected the importance of GPR to guide the location of new wells in the study region. The outcomes of the study will help the decision-makers, government agencies, and private sectors for sustainable planning of groundwater in the area. Overall, the present study provides a comprehensive high-precision machine learning and GPR-based groundwater potential mapping.

Analyzing the Performance and Application of CERES-Wheat and APSIM in the Guanzhong Plain, China

HighlightsAPSIM and CERES-Wheat were calibrated and evaluated for winter wheat in the Guanzhong Plain, China.Both models performed well as the result of the calibration based on comprehensive field data.CERES-Wheat showed high sensitivity to field capacity.APSIM showed high sensitivity to nitrate at sowing.A crop model ensemble is better than a single model application, particularly for grain yield.Abstract. Cropping system models are useful tools to estimate the impact of climate and environment on agricultural production and to improve the management of agricultural systems. Numerous crop models are used worldwide to simulate wheat crop growth and predict yield. In this study, two process-based crop models (APSIM and CERES-Wheat) were assessed for simulating wheat crop growth in the Guanzhong Plain, China. The main objectives were to compare the performance of both models and perform sensitivity analysis. The possibility of reducing the model output uncertainty by estimating ensemble means was also explored. Crop models were calibrated and evaluated using local experimental data from three growing seasons (2009 to 2012), which encompassed 27 crop managements based on different levels of irrigation and nitrogen application. APSIM and CERES-Wheat performed well after calibration. The nRMSE for aboveground biomass, grain yield, LAI, canopy nitrogen, cumulative evapotranspiration, water use efficiency, and nitrogen fertilizer productivity was less than 21%. Sensitivity analysis of both models showed that grain yield in CERES-Wheat was more sensitive to field capacity. In APSIM, grain yield was sensitive to nitrogen application rate. Mean ensemble model results showed reduced uncertainty for simulated crop parameters, especially for differences between observed and simulated grain yield, which were reduced by up to 4.1%. Overall, the results suggest that APSIM and CERES-Wheat are both suitable for simulating wheat crop growth, development, and yield in the Guanzhong Plain. Furthermore, by using the ensemble modeling approach, decision-makers can be more confident in their planning, as the information will be based on multiple models’ results. Keywords: Crop model ensemble, Model calibration, Multiple crop models, Sensitivity analysis, Winter wheat.

A living environment prediction model using ensemble machine learning techniques based on quality of life index

The living environment is an area that has the features necessary for all people to keep living their lives. Nevertheless, as time has progressed, needs and economic capacities of people have changed; consequently, the living environment has begun to express different meanings for different people. This differentiation brings the following question to mind: Can every person live in the same living environment under the same conditions? Of course not! People differ from each other in their preferences, needs, economic capacities, and many other aspects, so they must be assigned living environments that best fit their unique features. In this study, machine learning (ML) techniques and ensemble machine learning (EML) techniques are used for the establishment of a living environment prediction model. The data required for the quality of life index are obtained from a questionnaire. This questionnaire was prepared to consider people’s desires and economic capacities. Quality of life index is computed taking the weighted arithmetic mean of questionnaire data. Calculated quality of life indexes of the individuals is assigned to available quality of life classes of the cities. ML-based classification techniques have been used for prediction of assigned classes of the cities. A living environment prediction model is proposed in this study by using an ML and EML techniques-based quality of life index. For predicting living environment, ML-based methodologies include an artificial neural network (ANN), support vector machines (SVM) and EML-based methodologies include stacking and voting. The prediction results of ensemble models are better than prediction results of individual ML models, especially the stacking-based model composed of SMO + LMT and which reached the best performance values using the 80% split method.

Torrential rainfall-triggered shallow landslide characteristics and susceptibility assessment using ensemble data-driven models in the Dongjiang Reservoir Watershed, China

This study investigated the characteristics of rainfall-triggered landslides during the Typhoon Bilis in the Dongjiang Reservoir Watershed, China. The comparative shallow landslide susceptibility mappings (LSMs) were produced by the ensemble data-driven statistical models in a GIS environment. At first, the landslide inventory for the study area was prepared from the high-resolution QuickBird images, and China–Brazil Earth Resources Satellite images, and field survey. Other necessary data for landslide susceptibility analysis such as the amount of rainfall, geology, and topography were also collected from the respective agencies. Twelve predisposing factors were then prepared using this available dataset. To reduce the subjectivity of models and caution in the selection of predisposing factors, and to avoid the spatial autocorrelation redundancy, certainty factor approach was attempted to optimize these twelve set of parameters. For validating the accuracy of the model, the original landslide data were randomly divided into two parts: 70% (1545 landslides) for training the model and the remaining 30% (662 landslides) for validation. The verified results showed that using the optimized predisposing factors has a higher performance than using all the original twelve factors. The results of ensemble models also showed that LSM maps prepared using binary logistic regression (accuracy is 0.848) model are more accurate than those prepared using bivariate statistical analysis (accuracy is 0.837) model. Additionally, our analysis concludes that the short duration and high-intensity rainfall, drainage density, lithology, and curvature are the major influencing factors for landslide occurrences in this case study area. This research provides an improved understanding of the mechanism of landslides caused by the typhoons for the adjoining watersheds nearby the reservoir. The preliminary understandings and approach could also be applied in similar geological and rainfall-triggered case study sites in the other parts of the world for risk mitigation.

Will the California Current lose its nesting Tufted Puffins?

Tufted Puffin (Fratercula cirrhata) populations have experienced dramatic declines since the mid-19th century along the southern portion of the species range, leading citizen groups to petition the United States Fish and Wildlife Service (USFWS) to list the species as endangered in the contiguous US. While there remains no consensus on the mechanisms driving these trends, population decreases in the California Current Large Marine Ecosystem suggest climate-related factors, and in particular the indirect influence of sea-surface temperature on puffin prey. Here, we use three species distribution models (SDMs) to evaluate projected shifts in habitat suitable for Tufted Puffin nesting for the year 2050 under two future Intergovernmental Panel on Climate Change (IPCC) emission scenarios. Ensemble model results indicate warming marine and terrestrial temperatures play a key role in the loss of suitable Tufted Puffin nesting conditions in the California Current under both business-as-usual (RCP 8.5) and moderated (RCP 4.5) carbon emission scenarios, and in particular, that mean summer sea-surface temperatures greater than 15 °C are likely to make habitat unsuitable for breeding. Under both emission scenarios, ensemble model results suggest that more than 92% of currently suitable nesting habitat in the California Current is likely to become unsuitable. Moreover, the models suggest a net loss of greater than 21% of suitable nesting sites throughout the entire North American range of the Tufted Puffin, regardless of emission-reduction strategies. These model results highlight continued Tufted Puffin declines—particularly among southern breeding colonies—and indicate a significant risk of near-term extirpation in the California Current Large Marine Ecosystem.

Online Ensemble Modeling for Real Time Water Level Forecasts

Accurate and reliable flood forecasting is essential to mitigate the threats brought by floods. Ensemble approaches have been used in limited studies to improve the forecasts of component models. In this paper an ensemble model based on neural-fuzzy inference system (NFIS) and three real time updating approaches were used to synthesize the water level forecasts from a Adaptive-Network-based Fuzzy Inference System (ANFIS) model and the Unified River Basin Simulator (URBS) model for three stations in Lower Mekong. The NFIS ensemble model results are compared with the simple average model (SAM) which is adopted as a benchmark ensemble model. The ensemble model of offline learning without real time updating (EN-OFF), ensemble model with real time updating using offline learning (EN-RTOFF), ensemble model with real time updating using online learning (EN-RTON1) and ensemble model with real time updating using online learning and sub-models (EN-RTON2) were studied in this paper. Statistical analysis of the models for all the three stations indicated the superiority of the EN-RTON2 model over EN-RTOFF, EN-RTON1 models, SAM and the EN-OFF model. Not only the spikes in the URBS model were eliminated, but also the time shift problems in the ANFIS model results were decreased.

On the value of combining different modelled soil moisture products for European drought monitoring

summary In the context of evaluating the occurrence of drought events over Europe, soil moisture maps provide an invaluable resource to quantify the effects of rainfall deficits on vegetated lands. Spatially distributed models represent one of the main options, alongside satellite remote sensing, to successfully monitor this quantity over large areas in a cost effective way. This work has the double aim of: (i) intercomparing three soil moisture outputs obtained by different land-surface models (LISFOOD, CLM and TESSEL) through long (at least 6 years of data between 2001 and 2011) in-situ measured datastreams, and (ii) quantifying the added value of combining the estimates of these three models by means of a simple ensemble approach. Generally, the three models return similar soil moisture anomalies over most of Europe, with few notable exceptions during summer in Mediterranean regions. The comparison with in-situ data suggests no substantial differences among the models, with LISFLOOD slightly outperforming the other two in terms of correlation as also supported by a pairwise comparison. The combined soil moisture anomalies obtained via the ensemble-mean approach are characterized by an increase of both the correlation and the accuracy in retrieving extreme events compared to the single models; however, the number of observed extreme events actually captured by the ensemble model does not increase significantly if compared to the single models. Overall, the ensemble model results are skillful, with an all site average skill score of about 0.4.

Ensemble Forecasting of Volcanic Emissions in Hawai’i

<div class="page" title="Page 1"><div class="section"><div class="layoutArea"><div class="column"><p><span>Deterministic model forecasts do not convey to the end users the forecast uncertainty the models possess as a result of physics parameterizations, simplifications in model representation of physical processes, and errors in initial conditions. This lack of understanding leads to a level of uncertainty in the forecasted value when only a single deterministic model forecast is available. Increasing computational power and parallel software architecture allows multiple simulations to be carried out simultaneously that yield useful measures of model uncertainty that can be derived from ensemble model results. The Hybrid Single Particle Lagrangian Integration Trajectory and Dispersion model has the ability to generate ensemble forecasts. A meteorological ensemble was formed to create probabilistic forecast products and an ensemble mean forecast for volcanic emissions from the Kilauea volcano that impacts the state of Hawai’i. The probabilistic forecast products show uncertainty in pollutant concentrations that are especially useful for decision-making regarding public health. Initial comparison of the ensemble mean forecasts with observations and a single model forecast show improvements in event timing for both sulfur dioxide and sulfate aerosol forecasts. </span></p></div></div></div></div><p> </p>