Comparison of Random Forest, k-Nearest Neighbor, and Support Vector Machine Classifiers for Land Cover Classification Using Sentinel-2 Imagery

Integrating multisource data and machine learning for supraglacial lake detection: Implications for environmental management and sustainable development goals in high mountainous regions

The accurate detection and monitoring of supraglacial lakes in high mountainous regions are crucial for understanding their dynamic nature and implications for environmental management and sustainable development goals. In this study, we propose a novel approach that integrates multisource data and machine learning techniques for supra-glacial lake detection in the Passu Batura glacier of the Hunza Basin, Pakistan. We extract pertinent features or parameters by leveraging multisource datasets such as radar backscatter intensity VH and VV parameters from Sentinel-1 Ground Range Detected (GRD) data, near-infrared (NIR), NDWI_green, NDWI_blue parameters from Sentinel-2 Multi-spectral Instrument (MSI) data, and surface slope, aspect, and elevation parameters from topographic data. The entire dataset is partitioned into training and testing sets, with machine learning models including the artificial neural network (ANN), the support vector machine (SVM), logistic regression (LR), random forest (RF), and K-nearest neighbour (KNN) trained on the training data (70%). Accuracy assessment employs testing data and involves the evaluation of metrics such as ROC curves and confusion matrices. The best-performing model, ANN, is validated against manually digitized lake polygons derived from Sentinel-2 and Google Earth Pro imagery. Furthermore, the digitized lake polygons are used to analyze glacial lake dynamics from 2016 to 2022. Key findings of this research presented that the NDWI_green, Sigma0_VH, and elevation are the most significant predictors in detecting supra-glacial lakes. Among the various trained and evaluated models, the Artificial Neural Network (ANN) achieved the highest performance (accuracy: 95%, AUC: 0.99) and accurately mapped supra-glacial lakes regardless of their small size. The findings have significant implications for understanding glacial lake behavior in the context of climate change and informing future research and monitoring efforts.

Predicting agricultural and meteorological droughts using Holt Winter Conventional 2D-Long Short-Term Memory (HW-Conv2DLSTM)

Drought is an extended shortage of rainfall resulting in water scarcity and affecting a region's social and economic conditions through environmental deterioration. Its adverse environmental effects can be minimised by timely prediction. Drought detection uses only ground observation stations, but satellite-based supervision scans huge land mass stretches and offers highly effective monitoring. This paper puts forward a novel drought monitoring system using satellite imagery by considering the effects of droughts that devastated agriculture in Thanjavur district, Tamil Nadu, between 2000 and 2022. The proposed method uses Holt Winter Conventional 2D-Long Short-Term Memory (HW-Conv2DLSTM) to forecast meteorological and agricultural droughts. It employs Climate Hazards Group InfraRed Precipitation with Station (CHIRPS) data precipitation index datasets, MODIS 11A1 temperature index, and MODIS 13Q1 vegetation index. It extracts the time series data from satellite images using trend and seasonal patterns and smoothens them using Holt Winter alpha, beta, and gamma parameters. Finally, an effective drought prediction procedure is developed using Conv2D-LSTM to calculate the spatiotemporal correlation amongst drought indices. The HW-Conv2DLSTM offers a better R2 value of 0.97. It holds promise as an effective computer-assisted strategy to predict droughts and maintain agricultural productivity, which is vital to feed the ever-increasing human population.

Glove-Net: Enhancing Grasp Classification with Multisensory Data and Deep Learning Approach

Grasp classification is pivotal for understanding human interactions with objects, with wide-ranging applications in robotics, prosthetics, and rehabilitation. This study introduces a novel methodology utilizing a multisensory data glove to capture intricate grasp dynamics, including finger posture bending angles and fingertip forces. Our dataset comprises data collected from 10 participants engaging in grasp trials with 24 objects using the YCB object set. We evaluate classification performance under three scenarios: utilizing grasp posture alone, utilizing grasp force alone, and combining both modalities. We propose Glove-Net, a hybrid CNN-BiLSTM architecture for classifying grasp patterns within our dataset, aiming to harness the unique advantages offered by both CNNs and BiLSTM networks. This model seamlessly integrates CNNs' spatial feature extraction capabilities with the temporal sequence learning strengths inherent in BiLSTM networks, effectively addressing the intricate dependencies present within our grasping data. Our study includes findings from an extensive ablation study aimed at optimizing model configurations and hyperparameters. We quantify and compare the classification accuracy across these scenarios: CNN achieved 88.09%, 69.38%, and 93.51% testing accuracies for posture-only, force-only, and combined data, respectively. LSTM exhibited accuracies of 86.02%, 70.52%, and 92.19% for the same scenarios. Notably, the hybrid CNN-BiLSTM proposed model demonstrated superior performance with accuracies of 90.83%, 73.12%, and 98.75% across the respective scenarios. Through rigorous numerical experimentation, our results underscore the significance of multimodal grasp classification and highlight the efficacy of the proposed hybrid Glove-Net architectures in leveraging multisensory data for precise grasp recognition. These insights advance understanding of human-machine interaction and hold promise for diverse real-world applications.

Predicting Tumor Dynamics Post-Staged GKRS: Machine Learning Models in Brain Metastases Prognosis

This study assesses the predictive performance of six machine learning models and a 1D Convolutional Neural Network (CNN) in forecasting tumor dynamics within three months following Gamma Knife radiosurgery (GKRS) in 77 brain metastasis (BM) patients. The analysis meticulously evaluates each model before and after hyperparameter tuning, utilizing accuracy, AUC, and other metrics derived from confusion matrices. The CNN model showcased notable performance with an accuracy of 98% and an AUC of 0.97, effectively complementing the broader model analysis. Initial findings highlighted that XGBoost significantly outperformed other models with an accuracy of 0.95 and an AUC of 0.95 before tuning. Post-tuning, the Support Vector Machine (SVM) demonstrated the most substantial improvement, achieving an accuracy of 0.98 and an AUC of 0.98. Conversely, XGBoost showed a decline in performance after tuning, indicating potential overfitting. The study also explores feature importance across models, noting that features like "control at one year", "age of the patient", and "beam-on time for volume V1 treated" were consistently influential across various models, albeit their impacts were interpreted differently depending on the model's underlying mechanics. This comprehensive evaluation not only underscores the importance of model selection and hyperparameter tuning but also highlights the practical implications in medical diagnostic scenarios, where the accuracy of positive predictions can be crucial. Our research explores the effects of staged Gamma Knife radiosurgery (GKRS) on larger tumors, revealing no significant outcome differences across protocols. It uniquely considers the impact of beam-on time and fraction intervals on treatment efficacy. However, the investigation is limited by a small patient cohort and data from a single institution, suggesting the need for future multicenter research.

Remote sensing-based seasonal surface urban heat island analysis in the mining and industrial environment

Cooling spaces have an optimistic influence on surface urban heat islands (SUHI). Blue spaces benefit from balancing the changing climate and heat variations. Because of the rapid deforestation and SUHI increase, the climate is gradually changing in Paschim Bardhhaman, West Bengal state, India. Paschim Bardhhaman has two sectors: specifically, Durgapur is the main industrial centre and Asansol has coal mines. This investigation aims to categorize spatiotemporal variations and seasonal differences in cooling spaces and their influence on SUHI, land use and land cover (LULC), and thermal differences using Landsat datasets for the years 1992, 2004, 2012, and 2022 in summer and winter. The coal mining and industrial range decreased from 10,391.92 (1992) to 3591.1 ha (2022), respectively. Open pit mining distresses fresh water by heavy water uses in ore processing, and mining water was applied to excerpt minerals. Among the two sub-divisions, the blue space amount was higher in Asansol because mining actions were higher in Asansol than in Durgapur. The open vegetation volume has reduced from 46,441.03 (1992) to 25,827.55 ha (2022) and dense vegetation has erased from 7368.02 (1992) to 15,608.56 ha (2022). Dense vegetation improved because of heavy precipitation in those regions. Mostly, Raghunathpur, Saraswatiganja, Bhagabanpur, Bistupur, Paschim Gangaram, Garkilla Kherobari, and Gourbazar have dense vegetation. The outcomes similarly demonstrate that the total built-up part has increased by 8412.82 ha in between 30 years. The built-up zone changes near the southeast and western Paschim Bardhhaman district. Those region needs appropriate attention and planning to survive soon.

Automated Mapping of Land Cover Type within International Heterogenous Landscapes Using Sentinel-2 Imagery with Ancillary Geospatial Data

A near-global framework for automated training data generation and land cover classification using shallow machine learning with low-density time series imagery does not exist. This study presents a methodology to map nine-class, six-class, and five-class land cover using two dates (winter and non-winter) of a Sentinel-2 granule across seven international sites. The approach uses a series of spectral, textural, and distance decision functions combined with modified ancillary layers (such as global impervious surface and global tree cover) to create binary masks from which to generate a balanced set of training data applied to a random forest classifier. For the land cover masks, stepwise threshold adjustments were applied to reflectance, spectral index values, and Euclidean distance layers, with 62 combinations evaluated. Global (all seven scenes) and regional (arid, tropics, and temperate) adaptive thresholds were computed. An annual 95th and 5th percentile NDVI composite was used to provide temporal corrections to the decision functions, and these corrections were compared against the original model. The accuracy assessment found that the regional adaptive thresholds for both the two-date land cover and the temporally corrected land cover could accurately map land cover type within nine-class (68.4% vs. 73.1%), six-class (79.8% vs. 82.8%), and five-class (80.1% vs. 85.1%) schemes. Lastly, the five-class and six-class models were compared with a manually labeled deep learning model (Esri), where they performed with similar accuracies (five classes: Esri 80.0 ± 3.4%, region corrected 85.1 ± 2.9%). The results highlight not only performance in line with an intensive deep learning approach, but also that reasonably accurate models can be created without a full annual time series of imagery.

Evaluating Machine Learning Methods of Analyzing Multiclass Metabolomics

Multiclass metabolomic studies have become popular for revealing the differences in multiple stages of complex diseases, various lifestyles, or the effects of specific treatments. In multiclass metabolomics, there are multiple data manipulation steps for analyzing raw data, which consist of data filtering, the imputation of missing values, data normalization, marker identification, sample separation, classification, and so on. In each step, several to dozens of machine learning methods can be chosen for the given data set, with potentially hundreds or thousands of method combinations in the whole data processing chain. Therefore, a clear understanding of these machine learning methods is helpful for selecting an appropriate method combination for obtaining stable and reliable analytical results of specific data. However, there has rarely been an overall introduction or evaluation of these methods based on multiclass metabolomic data. Herein, detailed descriptions of these machine learning methods in multiple data manipulation steps are reviewed. Moreover, an assessment of these methods was performed using a benchmark data set for multiclass metabolomics. First, 12 imputation methods for imputing missing values were evaluated based on the PSS (Procrustes statistical shape analysis) and NRMSE (normalized root-mean-square error) values. Second, 17 normalization methods for processing multiclass metabolomic data were evaluated by applying the PMAD (pooled median absolute deviation) value. Third, different methods of identifying markers of multiclass metabolomics were evaluated based on the CWrel (relative weighted consistency) value. Fourth, nine classification methods for constructing multiclass models were assessed using the AUC (area under the curve) value. Performance evaluations of machine learning methods are highly recommended to select the most appropriate method combination before performing the final analysis of the given data. Overall, detailed descriptions and evaluation of various machine learning methods are expected to improve analyses of multiclass metabolomic data.

Spatio-temporal prediction of land use and land cover change in Bahi (Manyoni) Catchment, Tanzania, using multilayer perceptron neural network and cellular automata-Markov chain model

Evaluation of land use and land cover (LULC) change is among vital tools used for tracking environmental health and proper resource management. Remote sensing data was used to determine LULC change in Bahi (Manyoni) Catchment (BMC) in central Tanzania. Landsat satellite images from Landsat 5 TM and Landsat 8 OLI/TIRS were used, and support vector machine (SVM) algorithm was applied to classify the features of BMC. The obtained kappa values were 0.74, 0.83 and 0.84 for LULC maps of 1985, 2005 and 2021, respectively, which indicates the degree of accuracy from produced being substantial to almost perfect. Classified maps along with geospatial, socio-economic and climatic drivers with sufficient explanatory power were incorporated into MLP-NN to produce transition potential maps. Transition maps were subsequently used in cellular automata (CA)-Markov chain model to predict future LULC for BMC in immediate-future (2035), mid-future (2055) and far-future (2085). The findings indicate BMC is expected to experience significant expansion of agricultural lands and built land from 31.89 to 50.16% and 1.48 to 9.1% from 2021 to 2085 at the expense of open woodland, shrubland and savanna grassland. Low-yield crop production, water scarcity and population growth were major driving forces for rapid expansion of agricultural lands and overall LULC in BMC. The findings are essential for understanding the impact of LULC on hydrological processes and offer insights for the internal drainage basin (IDB) board to make necessary measures to lessen the expected dramatic changes in LULC in the future while sustaining harmonious balance with livelihood activities.

Identification of gastric signet ring cell carcinoma based on endoscopic images using few-shot learning

BACKGROUND

Deep learning (DL) models perform poorly when there are limited gastric signet ring cell carcinoma (SRCC) samples. Few-shot learning (FSL) can address the small sample problem.

METHODS

EfficientNetV2-S was first pretrained on ImageNet and then pretrained on esophageal endoscopic images (i.e., base classes: normal vs. early cancer vs. advanced cancer) using transfer learning. Second, images of gastric benign ulcers, adenocarcinoma and SRCC, i.e., novel classes (n = 50 per class), were included. Image features were extracted as vectors using the dual pretrained EfficientNetV2-S. Finally, a k-nearest neighbor classifier was used to identify SRCC. The above proposed 3-way 3-shot FSL framework was conducted in three rounds.

RESULTS

Dual pretrained FSL performed better than single pretrained FSL, endoscopists and traditional EfficientNetV2-S models. Dual pretrained FSL obtained the highest accuracy (79.4%), sensitivity (68.8%), recall (68.8%), precision (69.3%) and F1-score (0.691), and the senior endoscopist achieved the highest specificity of 93.6% when identifying SRCC. The macro-AUC and F1-score for dual pretraining (0.763 and 0.703, respectively) were higher than those for single pretraining (0.656 and 0.537, respectively), along with endoscopists and traditional EfficientNetV2-S models. The 2-way 3-shot FSL also performed better.

CONCLUSION

The proposed FSL framework showed practical performance in the differentiation of SRCC on endoscopic images, suggesting the potential of FSL in the computer-aided diagnosis for rare diseases.

Quantitative assessment of Land use/land cover changes in a developing region using machine learning algorithms: A case study in the Kurdistan Region, Iraq

The identification of land use/land cover (LULC) changes is important for monitoring, evaluating, and preserving natural resources. In the Kurdistan region, the utilization of remotely sensed data to assess the effectiveness of machine learning algorithms (MLAs) for LULC classification and change detection analysis has been limited. This study monitors and analyzes LULC changes in the study area from 1991 to 2021 using a quantitative approach with multi-temporal Landsat imagery. Five MLAs were applied: Support Vector Machine (SVM), Random Forest (RF), Artificial Neural Network (ANN), K-Nearest Neighbor (KNN), and Extreme Gradient Boosting (XGBoost). The results showed that the RF algorithm produced the most accurate maps of the three-decade study period, accompanied by a high kappa coefficient (0.93-0.97) compared with the SVM (0.91-0.95), ANN (0.91-0.96), KNN (0.92-0.96), and XGBoost (0.92-0.95) algorithms. Consequently, the RF classifier was implemented to categorize all obtainable satellite images. Socioeconomic changes throughout these transition periods were revealed by the change detection results. Rangeland and barren land areas decreased by 11.33 % (-402.03 km2) and 6.68 % (-236.8 km2), respectively. The transmission increases of 13.54 % (480.18 km2), 3.43 % (151.74 km2), and 0.71 % (25.22 km2) occurred in agricultural land, forest, and built-up areas, respectively. The outcomes of this study contribute significantly to LULC monitoring in developing regions, guiding stakeholders to identify vulnerable areas for better land use planning and sustainable environmental protection.

Land cover and crop types mapping using different spatial resolution imagery in a Mediterranean irrigated area

Crop type identification is critical for agricultural sustainability policy development and environmental assessments. Therefore, it is important to obtain their spatial distribution via different approaches. Medium-, high- and very high-resolution optical satellite sensors are efficient tools for acquiring this information, particularly for challenging studies such as those conducted in heterogeneous agricultural fields. This research examined the ability of four multitemporal datasets (Sentinel-1-SAR (S1), Sentinel-2-MSI (S2), RapidEye (RE), and PlanetScope (PS)) to identify land cover and crop types (LCCT) in a Mediterranean irrigated area. To map LCCT distribution, a supervised pixel-based classification is adopted using Support Vector Machine with a radial basis function kernel (SVMRB) and Random Forest (RF). Thus, LCCT maps were generated into three levels, including six (Level I), ten (Level II), and fourteen (Level III) classes. Overall, the findings revealed high overall accuracies of >92%, >83%, and > 81% for Level I, Level II, and Level III, respectively, except for Sentinel-1. It was found that accuracy improves considerably when the number of classes decreases, especially when cropland or non-cropland classes are grouped into one. Furthermore, there was a similarity in performance between S2 alone and S1S2. PlanetScope LCCT classifications outperform other sensors. In addition, the present study demonstrated that SVM achieved better performances against RF and can thereby effectively extract LCCT information from high-resolution imagery as PlanetScope.

Automated segmentation of meningioma from contrast-enhanced T1-weighted MRI images in a case series using a marker-controlled watershed segmentation and fuzzy C-means clustering machine learning algorithm

INTRODUCTION AND IMPORTANCE

Accurate segmentation of meningiomas from contrast-enhanced T1-weighted (CE T1-w) magnetic resonance imaging (MRI) is crucial for diagnosis and treatment planning. Manual segmentation is time-consuming and prone to variability. To evaluate an automated segmentation approach for meningiomas using marker-controlled watershed segmentation (MCWS) and fuzzy c-means (FCM) algorithms.

CASE PRESENTATION AND METHODS

CE T1-w MRI of 3 female patients (aged 59, 44, 67 years) with right frontal meningiomas were analyzed. Images were converted to grayscale and preprocessed with Otsu's thresholding and FCM clustering. MCWS segmentation was performed. Segmentation accuracy was assessed by comparing automated segmentations to manual delineations.

CLINICAL DISCUSSION

The approach successfully segmented meningiomas in all cases. Mean sensitivity was 0.8822, indicating accurate identification of tumors. Mean Dice similarity coefficient between Otsu's and FCM1 was 0.6599, suggesting good overlap between segmentation methods.

CONCLUSION

The MCWS and FCM approach enables accurate automated segmentation of meningiomas from CE T1-w MRI. With further validation on larger datasets, this could provide an efficient tool to assist in delineating meningioma boundaries for clinical management.

Research on land cover classification of multi-source remote sensing data based on improved U-net network

In recent years, remote sensing images of various types have found widespread applications in resource exploration, environmental protection, and land cover classification. However, relying solely on a single optical or synthetic aperture radar (SAR) image as the data source for land cover classification studies may not suffice to achieve the desired accuracy in ground information monitoring. One widely employed neural network for remote sensing image land cover classification is the U-Net network, which is a classical semantic segmentation network. Nonetheless, the U-Net network has limitations such as poor classification accuracy, misclassification and omission of small-area terrains, and a large number of network parameters. To address these challenges, this research paper proposes an improved approach that combines both optical and SAR images in bands for land cover classification and enhances the U-Net network. The approach incorporates several modifications to the network architecture. Firstly, the encoder-decoder framework serves as the backbone terrain-extraction network. Additionally, a convolutional block attention mechanism is introduced in the terrain extraction stage. Instead of pooling layers, convolutions with a step size of 2 are utilized, and the Leaky ReLU function is employed as the network's activation function. This design offers several advantages: it enhances the network's ability to capture terrain characteristics from both spatial and channel dimensions, resolves the loss of terrain map information while reducing network parameters, and ensures non-zero gradients during the training process. The effectiveness of the proposed method is evaluated through land cover classification experiments conducted on optical, SAR, and combined optical and SAR datasets. The results demonstrate that our method achieves classification accuracies of 0.8905, 0.8609, and 0.908 on the three datasets, respectively, with corresponding mIoU values of 0.8104, 0.7804, and 0.8667. Compared to the traditional U-Net network, our method exhibits improvements in both classification accuracy and mIoU to a certain extent.

Modeling of the Sintered Density in Cu-Al Alloy Using Machine Learning Approaches

In powder metallurgy materials, sintered density in Cu-Al alloy plays a critical role in detecting mechanical properties. Experimental measurement of this property is costly and time-consuming. In this study, adaptive boosting decision tree, support vector regression, k-nearest neighbors, extreme gradient boosting, and four multilayer perceptron (MLP) models tuned by resilient backpropagation, Levenberg-Marquardt (LM), scaled conjugate gradient, and Bayesian regularization were employed for predicting powder densification through sintering. Yield strength, Young's modulus, volume variation caused by the phase transformation, hardness, liquid volume, liquidus temperature, the solubility ratio among the liquid phase and the solid phase, sintered temperature, solidus temperature, sintered atmosphere, holding time, compaction pressure, particle size, and specific shape factor were regarded as the input parameters of the suggested models. The cross plot, error distribution curve, and cumulative frequency diagram as graphical tools and average percent relative error (APRE), average absolute percent relative error (AAPRE), root mean square error (RMSE), standard deviation (SD), and coefficient of correlation (R) as the statistical evaluations were utilized to estimate the models' accuracy. All of the developed models were compared with preexisting approaches, and the results exhibited that the developed models in the present work are more precise and valid than the existing ones. The designed MLP-LM model was found to be the most precise approach with AAPRE = 1.292%, APRE = -0.032%, SD = 0.020, RMSE = 0.016, and R = 0.989. Lately, outlier detection was applied performing the leverage technique to detect the suspected data points. The outlier detection discovered that few points are located out of the applicability domain of the proposed MLP-LM model.

Forecasting harmful algae blooms: Application to Dinophysis acuminata in northern Norway

Dinophysis acuminata produces Diarrhetic Shellfish Toxins (DST) that contaminate natural and farmed shellfish, leading to public health risks and economically impacting mussel farms. For this reason, there is a high interest in understanding and predicting D. acuminata blooms. This study assesses the environmental conditions and develops a sub-seasonal (7 - 28 days) forecast model to predict D. acuminata cells abundance in the Lyngen fjord located in northern Norway. A Support Vector Machine (SVM) model is trained to predict future D. acuminata cells abundance by using the past cell concentration, sea surface temperature (SST), Photosynthetic Active Radiation (PAR), and wind speed. Cells concentration of Dinophysis spp. are measured in-situ from 2006 to 2019, and SST, PAR, and surface wind speed are obtained by satellite remote sensing. D. acuminata only explains 40% of DST variability from 2006 to 2011, but it changes to 65% after 2011 when D. acuta prevalence reduced. The D. acuminata blooms can reach concentration up to 3954 cells l-1 and are restricted to the summer during warmer waters, varying from 7.8 to 12.7 °C. The forecast model predicts with fair accuracy the seasonal development of the blooms and the blooms amplitude, showing a coefficient of determination varying from 0.46 to 0.55. SST has been found to be a useful predictor for the seasonal development of the blooms, while the past cells abundance is needed for updating the current status and adjusting the blooms timing and amplitude. The calibrated model should be tested operationally in the future to provide an early warning of D. acuminata blooms in the Lyngen fjord. The approach can be generalized to other regions by recalibrating the model with local observations of D. acuminata blooms and remote sensing data.

Identifying key multi-modal predictors of incipient dementia in Parkinson's disease: a machine learning analysis and Tree SHAP interpretation

Background

Persons with Parkinson's disease (PD) differentially progress to cognitive impairment and dementia. With a 3-year longitudinal sample of initially non-demented PD patients measured on multiple dementia risk factors, we demonstrate that machine learning classifier algorithms can be combined with explainable artificial intelligence methods to identify and interpret leading predictors that discriminate those who later converted to dementia from those who did not.

Method

Participants were 48 well-characterized PD patients (M_{baseline age} = 71.6; SD = 4.8; 44% female). We tested 38 multi-modal predictors from 10 domains (e.g., motor, cognitive) in a computationally competitive context to identify those that best discriminated two unobserved baseline groups, PD No Dementia (PDND), and PD Incipient Dementia (PDID). We used Random Forest (RF) classifier models for the discrimination goal and Tree SHapley Additive exPlanation (Tree SHAP) values for deep interpretation.

Results

An excellent RF model discriminated baseline PDID from PDND (AUC = 0.84; normalized Matthews Correlation Coefficient = 0.76). Tree SHAP showed that ten leading predictors of PDID accounted for 62.5% of the model, as well as their relative importance, direction, and magnitude (risk threshold). These predictors represented the motor (e.g., poorer gait), cognitive (e.g., slower Trail A), molecular (up-regulated metabolite panel), demographic (age), imaging (ventricular volume), and lifestyle (activities of daily living) domains.

Conclusion

Our data-driven protocol integrated RF classifier models and Tree SHAP applications to selectively identify and interpret early dementia risk factors in a well-characterized sample of initially non-demented persons with PD. Results indicate that leading dementia predictors derive from multiple complementary risk domains.

Research on Rice Fields Extraction by NDVI Difference Method Based on Sentinel Data

To meet the challenge of food security, it is necessary to obtain information about rice fields accurately, quickly and conveniently. In this study, based on the analysis of existing rice fields extraction methods and the characteristics of intra-annual variation of normalized difference vegetation index (NDVI) in the different types of ground features, the NDVI difference method is used to extract rice fields using Sentinel data based on the unique feature of rice fields having large differences in vegetation between the pre-harvest and post-harvest periods. Firstly, partial correlation analysis is used to study the influencing factors of the rice harvesting period, and a simulation model of the rice harvesting period is constructed by multiple regression analysis with data from 32 sample points. Sentinel data of the pre-harvest and post-harvest periods of rice fields are determined based on the selected rice harvesting period. The NDVI values of the rice fields are calculated for both the pre-harvest and post-harvest periods, and 33 samples of the rice fields are selected from the high-resolution image. The threshold value for rice field extraction is determined through statistical analysis of the NDVI difference in the sample area. This threshold was then utilized to extract the initial extent of rice fields. Secondly, to address the phenomenon of the "water edge effect" in the initial data, the water extraction method based on the normalized difference water index (NDWI) is used to remove the pixels of water edges. Finally, the extraction results are verified and analyzed for accuracy. The study results show that: (1) The rice harvesting period is significantly correlated with altitude and latitude, with coefficients of 0.978 and 0.922, respectively, and the simulation model of the harvesting period can effectively determine the best period of remote sensing images needed to extract rice fields; (2) The NDVI difference method based on sentinel data for rice fields extraction is excellent; (3) The mixed pixels have a large impact on the accuracy of rice fields extraction, due to the water edge effect. Combining NDWI can effectively reduce the water edge effect and significantly improve the accuracy of rice field extraction.

Habitat selection of wintering cranes in typical wetlands in the middle and lower reaches of the Yangtze River over the past 20 years, China

The wetlands in the middle and lower reaches of the Yangtze River are the main overwintering and perching places for cranes. To examine the habitat selection mechanism of cranes in this area, two natural wetland reserves, Shengjin Lake and Poyang Lake, which are the main habitats of typical cranes, were selected as the study area. Using 20 years of Landsat satellite image data (between 1999 and 2019), the vegetation cover index was calculated from a pixel dichotomy model, and the landscape pattern index was obtained through Fragstats. The entropy method was adopted to determine the weight of the landscape index, and then, the habitat suitability index was calculated. Combined with the number of typical crane populations in the reserve, the selection mechanism of overwintering habitat of cranes was revealed. On the change of land-use type, the crane habitat of Shengjin Lake transferred more to non-crane habitat, and other land types increased, resulting in the decrease of crane habitat area. However, the change of crane habitat in Poyang Lake Reserve was small, so it can accommodate more cranes to overwintering here. In terms of vegetation coverage, most of the vegetation cover areas of Shengjin Lake were woodland near or far from the lake, but the woodland was not the habitat of cranes. Most of the vegetation-covered areas of Poyang Lake are grassland near the lake, which provide rest and foraging places for cranes. In the landscape pattern, the number of landscape patches in Shengjin Lake was large, the degree of landscape fragmentation was higher than that in Poyang Lake, the landscape complexity was higher, and the landscape diversity was simpler. This is not conducive to the maintenance of crane habitat, but also reduces the attractiveness of overwintering cranes, while the landscape suitability of crane habitat in Poyang Lake was higher than that in Shengjin Lake, and cranes were more likely to choose Poyang Lake as their overwintering habitat.

Mapping abandoned cropland using Within-Year Sentinel-2 time series

Against the background of the COVID-19 pandemic and various armed conflicts, the world is experiencing an unprecedented food crisis. The reclamation of abandoned cropland with food production potential may increase the global food supply in a short period of time, ensuring food security. At present, the extraction of abandoned cropland is mainly based on low- and medium-resolution remote sensing image data, making it difficult to extract fragmented areas in mountainous regions and to distinguish between abandoned cropland and transitional classes (such as fallow cropland). We developed a change-detection method based on within-year Sentinel-2 time series to extract cropland abandoned from 2018 to 2021 and defined four types of croplands, namely spontaneously abandoned, induced abandoned, fallow, and lost cropland, using Linxia County in mountainous China as the study region. First, cropland objects were generated from multi-temporal Sentinel-2 images using the multi-resolution segmentation method, and the land use map of Linxia County from 2017 to 2021 was drawn using random forest classifier. Second, through defining and identifying different cropland types, the interannual dynamic changes in cropland from 2018 to 2021 were extracted by analyzing the annual land use change trajectory. Third, by analyzing the normalized difference vegetation index (NDVI) time series of cropland within-year, the active and cultivated cropland sites within-year were extracted by threshold segmentation. Finally, the changes in the four cropland types were extracted by intersecting the two result types. Our method captured the object level changes well (overall mapping accuracy = 93 ± 5 %), and the extraction accuracy of abandoned cropland reached 81 ± 2 %. Abandoned cropland was mostly located in areas of medium quality and with a moderate distance from rural settlements. Reclamation can potentially increase the grain production in Linxia County by at least 3.6 % and needs to be combined with the local natural geography and human activities. Our method is a robust method for extracting abandoned cropland and may be applied to other research related to land use change.

Consecutive multiscale feature learning-based image classification model

Extracting useful features at multiple scales is a crucial task in computer vision. The emergence of deep-learning techniques and the advancements in convolutional neural networks (CNNs) have facilitated effective multiscale feature extraction that results in stable performance improvements in numerous real-life applications. However, currently available state-of-the-art methods primarily rely on a parallel multiscale feature extraction approach, and despite exhibiting competitive accuracy, the models lead to poor results in efficient computation and low generalization on small-scale images. Moreover, efficient and lightweight networks cannot appropriately learn useful features, and this causes underfitting when training with small-scale images or datasets with a limited number of samples. To address these problems, we propose a novel image classification system based on elaborate data preprocessing steps and a carefully designed CNN model architecture. Specifically, we present a consecutive multiscale feature-learning network (CMSFL-Net) that employs a consecutive feature-learning approach based on the usage of various feature maps with different receptive fields to achieve faster training/inference and higher accuracy. In the conducted experiments using six real-life image classification datasets, including small-scale, large-scale, and limited data, the CMSFL-Net exhibits an accuracy comparable with those of existing state-of-the-art efficient networks. Moreover, the proposed system outperforms them in terms of efficiency and speed and achieves the best results in accuracy-efficiency trade-off.

Classification of land use/land cover using artificial intelligence (ANN-RF)

Because deep learning has various downsides, such as complexity, expense, and the need to wait longer for results, this creates a significant incentive and impetus to invent and adopt the notion of developing machine learning because it is simple. This study intended to increase the accuracy of machine-learning approaches for land use/land cover classification using Sentinel-2A, and Landsat-8 satellites. This study aimed to implement a proposed method, neural-based with object-based, to produce a model addressed by artificial neural networks (limited parameters) with random forest (hyperparameter) called ANN_RF. This study used multispectral satellite images (Sentinel-2A and Landsat-8) and a normalized digital elevation model as input datasets for the Sana'a city map of 2016. The results showed that the accuracy of the proposed model (ANN_RF) is better than the ANN classifier with the Sentinel-2A and Landsat-8 satellites individually, which may contribute to the development of machine learning through newer researchers and specialists; it also conventionally developed traditional artificial neural networks with seven to ten layers but with access to 1,000's and millions of simulated neurons without resorting to deep learning techniques (ANN_RF).

Dimensionality reduction strategies for land use land cover classification based on airborne hyperspectral imagery: a survey

Hyperspectral image (HSI) contains hundreds of adjacent spectral bands, which can effectively differentiate the region of interest. Nevertheless, many irrelevant and highly correlated spectral bands lead to the Hughes phenomenon. Consequently, hyperspectral image dimensionality reduction is necessary to select the most informative and significant spectral band and eliminate the redundant spectral band. To this end, this paper represents an extensive and systematic survey of hyperspectral dimensionality reduction approaches for land use land cover (LULC) classification. Moreover, this paper reviewed the following important points: (1) hyperspectral imaging data acquisition methods, (2) the difference between hyperspectral and multispectral images, (3) hyperspectral image dimensionality reduction based on machine learning (ML) and deep learning (DL) techniques, (4) the popular benchmark hyperspectral datasets with the performance metrics for LULC classification, and (5) the significant challenges with the future trends for hyperspectral dimensionality reduction.

Temperature Field Reconstruction Method for Acoustic Tomography Based on Multi-Dictionary Learning

A reconstruction algorithm is proposed, based on multi-dictionary learning (MDL), to improve the reconstruction quality of acoustic tomography for complex temperature fields. Its aim is to improve the under-determination of the inverse problem by the sparse representation of the sound slowness signal (i.e., reciprocal of sound velocity). In the MDL algorithm, the K-SVD dictionary learning algorithm is used to construct corresponding sparse dictionaries for sound slowness signals of different types of temperature fields; the KNN peak-type classifier is employed for the joint use of multiple dictionaries; the orthogonal matching pursuit (OMP) algorithm is used to obtain the sparse representation of sound slowness signal in the sparse domain; then, the temperature distribution is obtained by using the relationship between sound slowness and temperature. Simulation and actual temperature distribution reconstruction experiments show that the MDL algorithm has smaller reconstruction errors and provides more accurate information about the temperature field, compared with the compressed sensing and improved orthogonal matching pursuit (CS-IMOMP) algorithm, which is an algorithm based on compressed sensing and improved orthogonal matching pursuit (in the CS-IMOMP, DFT dictionary is used), the least square algorithm (LSA) and the simultaneous iterative reconstruction technique (SIRT).

Land-Use and Land-Cover Dynamics in the Brazilian Caatinga Dry Tropical Forest

The use of remote sensing to determine land-use and land-cover (LULC) dynamics is often applied to assess the levels of natural forest conservation and monitor deforestation worldwide. This study examines the loss of native vegetation in the Campo Maior Complex (CMC), in the Brazilian Caatinga dry tropical forest, from 2016 to 2020, considering the temporal distribution of rainfall and discussing the trends and impacts of forest-degradation vectors. The Google Earth Engine (GEE) platform is used to obtain the rainfall data from the CHIRPS collection and to create the LULC maps. The random forest classifier is used and applied to the Landsat 8 collection. The QGIS open software and its SPC plugin are used to visualize the LULC dynamics. The results show that the months from June to October have the lowest average rainfall, and that 2019 is the year with the highest number of consecutive rainy days below 5 mm. The LULC maps show that deforestation was higher in 2018, representing 20.19%. In 2020, the proportion of deforestation was the lowest (11.95%), while regeneration was the highest (20.33%). Thus, the characterization of the rainfall regime is essential for more accurate results in LULC maps across the seasonally dry tropical forests (SDTF).

Land-Use and Land-Cover Classification in Semi-Arid Areas from Medium-Resolution Remote-Sensing Imagery: A Deep Learning Approach

Detailed Land-Use and Land-Cover (LULC) information is of pivotal importance in, e.g., urban/rural planning, disaster management, and climate change adaptation. Recently, Deep Learning (DL) has emerged as a paradigm shift for LULC classification. To date, little research has focused on using DL methods for LULC mapping in semi-arid regions, and none that we are aware of have compared the use of different Sentinel-2 image band combinations for mapping LULC in semi-arid landscapes with deep Convolutional Neural Network (CNN) models. Sentinel-2 multispectral image bands have varying spatial resolutions, and there is often high spectral similarity of different LULC features in semi-arid regions; therefore, selection of suitable Sentinel-2 bands could be an important factor for LULC mapping in these areas. Our study contributes to the remote sensing literature by testing different Sentinel-2 bands, as well as the transferability of well-optimized CNNs, for semi-arid LULC classification in semi-arid regions. We first trained a CNN model in one semi-arid study site (Gujranwala city, Gujranwala Saddar and Wazirabadtownships, Pakistan), and then applied the pre-trained model to map LULC in two additional semi-arid study sites (Lahore and Faisalabad city, Pakistan). Two different composite images were compared: (i) a four-band composite with 10 m spatial resolution image bands (Near-Infrared (NIR), green, blue, and red bands), and (ii) a ten-band composite made by adding two Short Wave Infrared (SWIR) bands and four vegetation red-edge bands to the four-band composite. Experimental results corroborate the validity of the proposed CNN architecture. Notably, the four-band CNN model has shown robustness in semi-arid regions, where spatially and spectrally confusing land-covers are present.

Exploring Natura 2000 habitats by satellite image segmentation combined with phytosociological data: a case study from the Čierny Balog area (Central Slovakia)

Natura 2000 is a network of protected areas covering Europe's most valuable and threatened species and habitats. Recently, biota belonging to these networks have been threatened by both climate change and various human impacts. Regular monitoring is needed to ensure effective protection and proper management measures in these sites and habitats, but conventional field approaches are often time-consuming and inaccurate. New approaches and studies with different focuses and results are being developed. Our approach includes point data from field research and phytosociological databases as starting points for automatic segmentation, which has been developed just recently as a novel method that could help to connect ground-based and remote sensing data. Our case study is located in Central Slovakia, in the mountains around the village of Čierny Balog. The main aim of our case study is to apply advanced remote sensing techniques to map the area and condition of vegetation units. We focus on forest habitats belonging mainly to the Natura 2000 network. We concentrated on the verification of the possibilities of differentiation of various habitats using only multispectral Sentinel-2 satellite data. Our software NaturaSat created by our team was used to reach our objectives. After collecting data in the field using phytosociological approach and segmenting the explored areas in the program NaturaSat, spectral characteristics were calculated within identified habitats using software tools, which were subsequently processed and tested statistically. We obtained significant differences between forest habitat types. Also, segmentation accuracy was tested by comparing closed planar curves of ground based filed data and software results. This provided promising results and validation of the methods used. The results of this study have the potential to be used in a wider area to map the occurrence and quality of Natura 2000 habitats.

Inversion of Soil Organic Matter Content Based on Improved Convolutional Neural Network

Soil organic matter (SOM) is an important source of nutrients required during crop growth and is an important component of cultivated soil. In this paper, we studied the possibility of using deep learning methods to establish a multi-feature model to predict SOM content. Moreover, using Nong'an County of Changchun City as the study area, Sentinel-2A remote sensing images were taken as the data source to construct the dataset by using field sampling and image processing. The LeNet-5 convolutional neural network model was chosen as the deep learning model, which was improved based on the basic model. The evaluation metrics were selected as the root mean square error (RMSE) and the coefficient of determination R². Through comparison, the R² of the improved model was found to be higher than that of the linear regression method, Support Vector Machines (SVM) (RMSE = 2.471, R² = 0.4035), and Random Forest (RF) (RMSE = 2.577, R² = 0.4913). The result shows that: (1) It is feasible to use the multispectral data extracted from remote sensing images for soil organic matter content inversion based on the deep learning model with a minimum RMSE of 2.979 and with the R² reaching 0.89. (2) The choice of features has an impact on the prediction of the model to a certain extent. After ranking the importance of features, selecting the appropriate number of features for inversion provides better results than full feature inversion, and the computational speed is improved.

Advanced machine learning algorithms for flood susceptibility modeling - performance comparison: Red Sea, Egypt

Floods are among the most devastating environmental hazards that directly and indirectly affect people's lives and activities. In many countries, sustainable environmental management requires the assessment of floods and the likely flood-prone areas to avoid potential hazards. In this study, the performance and capabilities of seven machine learning algorithms (MLAs) for flood susceptibility mapping were tested, evaluated, and compared. These MLAs, including support vector machine (SVM), random forest (RF), multivariate adaptive regression spline (MARS), boosted regression tree (BRT), functional data analysis (FDA), general linear model (GLM), and multivariate discriminant analysis (MDA), were tested for the area between Safaga and Ras Gharib cities, Red Sea, Egypt. A geospatial database was developed with eleven flood-related factors, namely altitude, slope aspect, lithology, land use/land cover (LULC), slope length (LS), topographic wetness index (TWI), slope angle, profile curvature, plan curvature, stream power index (SPI), and hydrolithology units. In addition, 420 actual flooded areas were recorded from the study area to create a flood inventory map. The inventory data were randomly divided into training group with 70% and validation group with 30%. The flood-related factors were tested with a multicollinearity test, the variance inflation factor (VIF) was less than 2.135, the tolerance (TOL) was more than 0.468, and their importance was evaluated with a partial least squares (PLS) method. The results show that RF performed the best with the highest AUC (area under curve) value of 0.813, followed by GLM with 0.802, MARS with 0.801, BRT with 0.777, MDA with 0.768%, FDA with 0.763, and SVM with 0.733. The results of this study and the flood susceptibility maps could be useful for environmental mitigation, future development activities in the area, and flood control areas.

Spatial point patterns generation on remote sensing data using convolutional neural networks with further statistical analysis

Continuous technological growth and the corresponding environmental implications are triggering the enhancement of advanced environmental monitoring solutions, such as remote sensing. In this paper, we propose a new method for the spatial point patterns generation by classifying remote sensing images using convolutional neural network. To increase the accuracy, the training samples are extended by the suggested data augmentation scheme based on the similarities of images within the same part of the landscape for a limited observation time. The image patches are classified in accordance with the labels of previously classified images of the manually prepared training and test samples. This approach has improved the accuracy of image classification by 7% compared to current best practices of data augmentation. A set of image patch centers of a particular class is considered as a random point configuration, while the class labels are used as marks for every point. A marked point pattern is regarded as a combination of several subpoint patterns with the same qualitative marks. We analyze the bivariate point pattern to identify the relationships between points of different types using the features of a marked random point pattern.

Landscape Analysis of Cobalt Mining Activities from 2009 to 2021 Using Very High Resolution Satellite Data (Democratic Republic of the Congo)

The cobalt mining sector is well positioned to be a key contributor in determining the success of the Democratic Republic of the Congo (DRC) in meeting the Sustainable Development Goals (SDGs) by 2030. Despite the important contribution to the DRC’s economy, the rapid expansion of mining operations has resulted in major social, health, and environmental impacts. The objective of this study was to quantitatively assess the cumulative impact of mining activities on the landscape of a prominent cobalt mining area in the DRC. To achieve this, an object-based method, employing a support vector machine (SVM) classifier, was used to map land cover across the city of Kolwezi and the surrounding mining areas, where long-term mining activity has dramatically altered the landscape. The research used very high resolution (VHR) satellite imagery (2009, 2014, 2019, 2021) to map the spatial distribution of land cover and land cover change, as well as analyse the spatial relationship between land cover classes and visually identified mine features, from 2009 to 2021. Results from the object-based SVM land cover classification produced an overall accuracy of 85.2–90.4% across the time series. Between 2009 and 2021, land cover change accounted to: rooftops increasing by 147.2% (+7.7 km2); impervious surface increasing by 104.7% (+3.35 km2); bare land increasing by 85.4% (+33.81 km2); exposed rock increasing by 56.2% (+27.46 km2); trees decreasing by 4.5% (−0.34 km2); shrub decreasing by 38.4% (−26.04 km2); grass and cultivated land decreasing by 27.1% (−45.65 km2); and water decreasing by 34.6% (−3.28 km2). The co-location of key land cover classes and visually identified mine features exposed areas of potential environmental pollution, with 91.6% of identified water situated within a 1 km radius of a mine feature, and vulnerable populations, with 71.6% of built-up areas (rooftop and impervious surface class combined) situated within a 1 km radius of a mine feature. Assessing land cover patterns over time and the interplay between mine features and the landscape structure allowed the study to amplify the findings of localised on-the-ground research, presenting an alternative viewpoint to quantify the true scale and impact of cobalt mining in the DRC. Filling geospatial data gaps and examining the present and past trends in cobalt mining is critical for informing and managing the sustainable growth and development of the DRC’s mining sector.

Irrigated Crop Types Mapping in Tashkent Province of Uzbekistan with Remote Sensing-Based Classification Methods

Appropriate crop type mapping to monitor and control land management is very important in developing countries. It can be very useful where digital cadaster maps are not available or usage of Remote Sensing (RS) data is not utilized in the process of monitoring and inventory. The main goal of the present research is to compare and assess the importance of optical RS data in crop type classification using medium and high spatial resolution RS imagery in 2018. With this goal, Landsat 8 (L8) and Sentinel-2 (S2) data were acquired over the Tashkent Province between the crop growth period of May and October. In addition, this period is the only possible time for having cloud-free satellite images. The following four indices "Normalized Difference Vegetation Index" (NDVI), "Enhanced Vegetation Index" (EVI), and "Normalized Difference Water Index" (NDWI1 and NDWI2) were calculated using blue, red, near-infrared, shortwave infrared 1, and shortwave infrared 2 bands. Support-Vector-Machine (SVM) and Random Forest (RF) classification methods were used to generate the main crop type maps. As a result, the Overall Accuracy (OA) of all indices was above 84% and the highest OA of 92% was achieved together with EVI-NDVI and the RF method of L8 sensor data. The highest Kappa Accuracy (KA) was found with the RF method of L8 data when EVI (KA of 88%) and EVI-NDVI (KA of 87%) indices were used. A comparison of the classified crop type area with Official State Statistics (OSS) data about sown crops area demonstrated that the smallest absolute weighted average (WA) value difference (0.2 thousand ha) was obtained using EVI-NDVI with RF method and NDVI with SVM method of L8 sensor data. For S2-sensor data, the smallest absolute value difference result (0.1 thousand ha) was obtained using EVI with RF method and 0.4 thousand ha using NDVI with SVM method. Therefore, it can be concluded that the results demonstrate new opportunities in the joint use of Landsat and Sentinel data in the future to capture high temporal resolution during the vegetation growth period for crop type mapping. We believe that the joint use of S2 and L8 data enables the separation of crop types and increases the classification accuracy.

The Influence of Data Density and Integration on Forest Canopy Cover Mapping Using Sentinel-1 and Sentinel-2 Time Series in Mediterranean Oak Forests

Forest canopy cover (FCC) is one of the most important forest inventory parameters and plays a critical role in evaluating forest functions. This study examines the potential of integrating Sentinel-1 (S-1) and Sentinel-2 (S-2) data to map FCC in the heterogeneous Mediterranean oak forests of western Iran in different data densities (one-year datasets vs. three-year datasets). This study used very high-resolution satellite images from Google Earth, gridded points, and field inventory plots to generate a reference dataset. Based on it, four FCC classes were defined, namely non-forest, sparse forest (FCC = 1–30%), medium-density forest (FCC = 31–60%), and dense forest (FCC > 60%). In this study, three machine learning (ML) models, including Random Forest (RF), Support Vector Machine (SVM), and Classification and Regression Tree (CART), were used in the Google Earth Engine and their performance was compared for classification. Results showed that the SVM produced the highest accuracy on FCC mapping. The three-year time series increased the ability of all ML models to classify FCC classes, in particular the sparse forest class, which was not distinguished well by the one-year dataset. Class-level accuracy assessment results showed a remarkable increase in F-1 scores for sparse forest classification by integrating S-1 and S-2 (10.4% to 18.2% increased for the CART and SVM ML models, respectively). In conclusion, the synergetic use of S-1 and S-2 spectral temporal metrics improved the classification accuracy compared to that obtained using only S-2. The study relied on open data and freely available tools and can be integrated into national monitoring systems of FCC in Mediterranean oak forests of Iran and neighboring countries with similar forest attributes.

Fine-Scale Classification of Urban Land Use and Land Cover with PlanetScope Imagery and Machine Learning Strategies in the City of Cape Town, South Africa

Urban land use and land cover (LULC) change can be efficiently monitored with high-resolution satellite products for a variety of purposes, including sustainable planning. These, together with machine learning strategies, have great potential to detect even subtle changes with satisfactory accuracy. In this study, we used PlaneScope Imagery and machine learning strategies (Random Forests, Support Vector Machines, Naïve Bayes and K-Nearest Neighbour) to classify and detect LULC changes over the City of Cape Town between 2016 and 2021. Our results showed that K-Nearest Neighbour outperformed other classifiers by achieving the highest overall classification of accuracy (96.54% with 0.95 kappa), followed by Random Forests (94.8% with 0.92 kappa), Naïve Bayes (93.71% with 0.91 kappa) and Support Vector Machines classifiers with relatively low accuracy values (92.28% with 0.88 kappa). However, the performance of all classifiers was acceptable, exceeding the overall accuracy of more than 90%. Furthermore, the results of change detection from 2016 to 2021 showed that the high-resolution PlanetScope imagery could be used to track changes in LULC over a desired period accurately.

Density Prediction in Powder Bed Fusion Additive Manufacturing: Machine Learning-Based Techniques

Machine learning (ML) is one of the artificial intelligence tools which uses past data to learn the relationship between input and output and helps to predict future trends. Powder bed fusion additive manufacturing (PBF-AM) is extensively used for a wide range of applications in the industry. The AM process establishment for new material is a crucial task with trial-and-error approaches. In this work, ML techniques have been applied for the prediction of the density of PBF-AM. Density is the most vital property in evaluating the overall quality of the AM building part. The ML techniques, namely, artificial neural network (ANN), K-nearest neighbor (KNN), support vector machine (SVM), and linear regression (LR), are used to develop a model for predicting the density of the stainless steel (SS) 316L build part. These four methods are validated using R-squared values and different error functions to compare the predicted result. The ANN and SVM model performed well with the R-square value of 0.95 and 0.923, respectively, for the density prediction. The ML models would be beneficial for the prediction of the process parameters. Further, the developed ML model would also be helpful for the future application of ML in additive manufacturing.

Multi-Crop Classification Using Feature Selection-Coupled Machine Learning Classifiers Based on Spectral, Textural and Environmental Features

Machine learning (ML) classifiers have been widely used in the field of crop classification. However, having inputs that include a large number of complex features increases not only the difficulty of data collection but also reduces the accuracy of the classifiers. Feature selection (FS), which can availably reduce the number of features by selecting and reserving the most essential features for crop classification, is a good tool to solve this problem effectively. Different FS methods, however, have dissimilar effects on various classifiers, so how to achieve the optimal combination of FS methods and classifiers to meet the needs of high-precision recognition of multiple crops remains an open question. This paper intends to address this problem by coupling the analysis of three FS methods and six classifiers. Spectral, textual, and environmental features are firstly extracted as potential classification indexes from time-series remote sensing images from France. Then, three FS methods are used to obtain feature subsets and combined with six classifiers for coupling analysis. On this basis, 18 multi-crop classification models (FS–ML models) are constructed. Additionally, six classifiers without FS are constructed for comparison. The training set and the validation set for these models are constructed by using the Kennard-Stone algorithm with 70% and 30% of the samples, respectively. The performance of the classification model is evaluated by Kappa, F1-score, accuracy, and other indicators. The results show that different FS methods have dissimilar effects on various models. The best FS–ML model is RFAA+-RF, and its Kappa coefficient can reach 0.7968, which is 0.33–46.67% higher than that of other classification models. The classification results are highly dependent on the original classification index sets. Hence, the reasonability of combining spectral, textural, and environmental indexes is verified by comparing them with the single feature index set. The results also show that the classification strategy combining spectral, textual, and environmental indexes can effectively improve the ability of crop recognition, and the Kappa coefficient is 9.06–65.52% higher than that of the single unscreened feature set.

Image Classification and Land Cover Mapping Using Sentinel-2 Imagery: Optimization of SVM Parameters

Land use/cover (LU/LC) classification provides proxies of the natural and social processes related to urban development, providing stakeholders with crucial information. Remotely sensed images combined with supervised classification are common to define land use, but high-performance classifiers remain difficult to achieve, due to the presence of model hyperparameters. Conventional approaches rely on manual adjustment, which is time consuming and often unsatisfying. Therefore, the goal of this study has been to optimize the parameters of the support vector machine (SVM) algorithm for the generation of land use/cover maps from Sentinel-2 satellite imagery in selected humid and arid (three study sites each) climatic regions of Iran. For supervised SVM classification, we optimized two important parameters (gamma in kernel function and penalty parameter) of the LU/LC classification. Using the radial basis function (RBF) of the SVM classification method, we examined seven values for both parameters ranging from 0.001 to 1000. For both climate types, the penalty parameters (PP) showed a direct relationship with overall accuracy (OA). Statistical results confirmed that in humid study regions, LU/LC maps produced with a penalty parameter >100 were more accurate. However, for regions with arid climates, LU/LC maps with a penalty parameter >0.1 were more accurate. Mapping accuracy for both climate types was sensitive to the penalty parameter. In contrast, variations of the gamma values in the kernel function had no effect on the accuracy of the LU/LC maps in either of the climate zones. These new findings on SVM image classification are directly applicable to LU/LC for planning and environmental and natural resource management.

Multi-Temporal LiDAR and Hyperspectral Data Fusion for Classification of Semi-Arid Woody Cover Species

Mapping the spatial distribution of woody vegetation is important for monitoring, managing, and studying woody encroachment in grasslands. However, in semi-arid regions, remotely sensed discrimination of tree species is difficult primarily due to the tree similarities, small and sparse canopy cover, but may also be due to overlapping woody canopies as well as seasonal leaf retention (deciduous versus evergreen) characteristics. Similar studies in different biomes have achieved low accuracies using coarse spatial resolution image data. The objective of this study was to investigate the use of multi-temporal, airborne hyperspectral imagery and light detection and ranging (LiDAR) derived data for tree species classification in a semi-arid desert region. This study produces highly accurate classifications by combining multi-temporal fine spatial resolution hyperspectral and LiDAR data (~1 m) through a reproducible scripting and machine learning approach that can be applied to larger areas and similar datasets. Combining multi-temporal vegetation indices and canopy height models led to an overall accuracy of 95.28% and kappa of 94.17%. Five woody species were discriminated resulting in producer accuracies ranging from 86.12% to 98.38%. The influence of fusing spectral and structural information in a random forest classifier for tree identification is evident. Additionally, a multi-temporal dataset slightly increases classification accuracies over a single data collection. Our results show a promising methodology for tree species classification in a semi-arid region using multi-temporal hyperspectral and LiDAR remote sensing data.

Evaluation of Food Security Based on Remote Sensing Data—Taking Egypt as an Example

Egypt, a country with a harsh natural environment and rapid population growth, is facing difficulty in ensuring its national food security. A novel model developed for assessing food security in Egypt, which applies remote sensing techniques, is presented. By extracting the gray-level co-occurrence matrix (GLCM) mean texture features from Sentinel-1 and Landsat-7 images, the arable land used to grow grain crops was first classified and extracted using a support vector machine. In terms of the classified results, meteorological data, and normalized difference vegetation index (NDVI) data, the Carnegie–Ames–Stanford approach (CASA) model was adopted to compute the annual net primary production (NPP). Then, the NPP yield conversion formula was used to forecast the annual grain yield. Finally, a method for evaluating food security, which involves four dimensions, i.e., quantity security, economic security, quality security, and resource security, was established to evaluate food security in Egypt in 2010, 2015, and 2020. Based on the proposed model, a classification accuracy of the crop distribution map, which is above 82%, can be achieved. Moreover, the reliability of yield estimation is verified compared to the result estimated using statistics data provided by Food and Agriculture Organization (FAO). Our evaluation results show that food security in Egypt is declining, the quantity and quality security show large fluctuations, and economic and resource security are relatively stable. This model can satisfy the requirements for estimating grain yield at a wide scale and evaluating food security on a national level. It can be used to provide useful suggestions for governments regarding improving food security.

Encoding Geospatial Vector Data for Deep Learning: LULC as a Use Case

Geospatial vector data with semantic annotations are a promising but complex data source for spatial prediction tasks such as land use and land cover (LULC) classification. These data describe the geometries and the types (i.e., semantics) of geo-objects, such as a Shop or an Amenity. Unlike raster data, which are commonly used for such prediction tasks, geospatial vector data are irregular and heterogenous, making it challenging for deep neural networks to learn based on them. This work tackles this problem by introducing novel encodings which quantify the geospatial vector data allowing deep neural networks to learn based on them, and to spatially predict. These encodings were evaluated in this work based on a specific use case, namely LULC classification. We therefore classified LULC based on the different encodings as input and an attention-based deep neural network (called Perceiver). Based on the accuracy assessments, the potential of these encodings is compared. Furthermore, the influence of the object semantics on the classification performance is analyzed. This is performed by pruning the ontology, describing the semantics and repeating the LULC classification. The results of this work suggest that the encoding of the geography and the semantic granularity of geospatial vector data influences the classification performance overall and on a LULC class level. Nevertheless, the proposed encodings are not restricted to LULC classification but can be applied to other spatial prediction tasks too. In general, this work highlights that geospatial vector data with semantic annotations is a rich data source unlocking new potential for spatial predictions. However, we also show that this potential depends on how much is known about the semantics, and how the geography is presented to the deep neural network.

Mapping Spatiotemporal Changes in Forest Type and Aboveground Biomass from Landsat Long-Term Time-Series Analysis—A Case Study from Yaoluoping National Nature Reserve, Anhui Province of Eastern China

A natural reserve’s forest is an important base for promoting natural education, scientific research, biodiversity conservation and carbon accounting. Dynamic monitoring of the forest type and forest aboveground biomass (AGB) in a nature reserve is an important foundation for assessing the forest succession stage and trend. Based on the Landsat images covering the National Nature Reserve of Yaoluoping in Anhui province spanning from 1987 to 2020, a total of 42 Landsat scenes, the forest cover product set was first developed by using the well-established vegetation change tracker (VCT) model. On this basis, a new vegetation index, NDVI_DR, which considers the phenological characteristics of different forest types, was proposed to distinguish coniferous forest from broad-leaved forest. Next, multiple modeling factors, including remote sensing spectral signatures, vegetation indices, textural measures derived from gray level co-occurrence matrix and wavelet analysis and topographic attributes, were compiled to model the AGB in 2011 by forest type separately by using the stochastic gradient boosting (SGB) algorithm. Then, using the 2011 Landsat image as the base, all the Landsat images in the other years involved in the modelling were relatively normalized by using the weighted invariant pixels (WIP) method, followed by an extrapolation of the 2011 AGB model to other years to create a time-series of AGB. The results showed that the overall accuracy of the VCT-based forest classification products was over 90%. The annual forest type classifications derived from NDVI_DR thresholding gained an overall accuracy above 92%, with a kappa coefficient above 0.8. The 2011 forest-type-dependent SGB-based AGB estimation model achieved an independent validation R2 at 0.63 and an RMSE at 11.18 t/ha for broad-leaved forest, and 0.61 and 14.26 t/ha for coniferous forest. The mapped time-series of AGB showed a gradual increasing trend over the past three decades. The driving factors responsible for the observed forest cover and AGB changes were analyzed to provide references for reasonable protection and development. The proposed methodology is a reliable tool for evaluating the management status, which can be extended to other similar regions.

Limiting the Collection of Ground Truth Data for Land Use and Land Cover Maps with Machine Learning Algorithms

Land use and land cover (LULC) classification maps help understand the state and trends of agricultural production and provide insights for applications in environmental monitoring. One of the major downfalls of the LULC technique is inherently linked to its need for ground truth data to cross-validate maps. This paper aimed at evaluating the efficiency of machine learning (ML) in limiting the use of ground truth data for LULC maps. This was accomplished by (1) extracting reliable LULC information from Sentinel-2 and Landsat-8 s images, (2) generating remote sensing indices used to train ML algorithms, and (3) comparing the results with ground truth data. The remote sensing indices that were tested include the difference vegetation index (DVI), the normalized difference vegetation index (NDVI), the normalized built-up index (NDBI), the urban index (UI), and the normalized bare land index (NBLI). Extracted vegetation indices were evaluated on three ML algorithms, namely, random forest (RF), k-nearest neighbour (K-NN), and k dimensional-tree (KD-Tree). The accuracy of these algorithms was assessed with standard statistical measures and ground truth data randomly collected in Prince Edward Island, Canada. Results showed that high kappa coefficient values were achieved by K-NN (82% and 74%), KD-Tree (80% and 78%), and RF (83% and 73%) for Sentinel-2A and Landsat-8 imagery, respectively. RF was a better classifier than K-NN and KD-Tree and had the highest overall accuracy with Sentinel-2A satellite images (92%). This approach provides the basis for limiting the collection of ground truth data and thus reduces the labour cost, time, and resources needed to collect ground truth data for LULC maps.

Urban Land Use and Land Cover Change Analysis Using Random Forest Classification of Landsat Time Series

Efficient implementation of remote sensing image classification can facilitate the extraction of spatiotemporal information for land use and land cover (LULC) classification. Mapping LULC change can pave the way to investigate the impacts of different socioeconomic and environmental factors on the Earth’s surface. This study presents an algorithm that uses Landsat time-series data to analyze LULC change. We applied the Random Forest (RF) classifier, a robust classification method, in the Google Earth Engine (GEE) using imagery from Landsat 5, 7, and 8 as inputs for the 1985 to 2019 period. We also explored the performance of the pan-sharpening algorithm on Landsat bands besides the impact of different image compositions to produce a high-quality LULC map. We used a statistical pan-sharpening algorithm to increase multispectral Landsat bands’ (Landsat 7–9) spatial resolution from 30 m to 15 m. In addition, we checked the impact of different image compositions based on several spectral indices and other auxiliary data such as digital elevation model (DEM) and land surface temperature (LST) on final classification accuracy based on several spectral indices and other auxiliary data on final classification accuracy. We compared the classification result of our proposed method and the Copernicus Global Land Cover Layers (CGLCL) map to verify the algorithm. The results show that: (1) Using pan-sharpened top-of-atmosphere (TOA) Landsat products can produce more accurate results for classification instead of using surface reflectance (SR) alone; (2) LST and DEM are essential features in classification, and using them can increase final accuracy; (3) the proposed algorithm produced higher accuracy (94.438% overall accuracy (OA), 0.93 for Kappa, and 0.93 for F1-score) than CGLCL map (84.4% OA, 0.79 for Kappa, and 0.50 for F1-score) in 2019; (4) the total agreement between the classification results and the test data exceeds 90% (93.37–97.6%), 0.9 (0.91–0.96), and 0.85 (0.86–0.95) for OA, Kappa values, and F1-score, respectively, which is acceptable in both overall and Kappa accuracy. Moreover, we provide a code repository that allows classifying Landsat 4, 5, 7, and 8 within GEE. This method can be quickly and easily applied to other regions of interest for LULC mapping.

Machine learning approaches for biomolecular, biophysical, and biomaterials research

A fluent conversation with a virtual assistant, person-tailored news feeds, and deep-fake images created within seconds-all those things that have been unthinkable for a long time are now a part of our everyday lives. What these examples have in common is that they are realized by different means of machine learning (ML), a technology that has fundamentally changed many aspects of the modern world. The possibility to process enormous amount of data in multi-hierarchical, digital constructs has paved the way not only for creating intelligent systems but also for obtaining surprising new insight into many scientific problems. However, in the different areas of biosciences, which typically rely heavily on the collection of time-consuming experimental data, applying ML methods is a bit more challenging: Here, difficulties can arise from small datasets and the inherent, broad variability, and complexity associated with studying biological objects and phenomena. In this Review, we give an overview of commonly used ML algorithms (which are often referred to as "machines") and learning strategies as well as their applications in different bio-disciplines such as molecular biology, drug development, biophysics, and biomaterials science. We highlight how selected research questions from those fields were successfully translated into machine readable formats, discuss typical problems that can arise in this context, and provide an overview of how to resolve those encountered difficulties.

PlanetScope, Sentinel-2, and Sentinel-1 Data Integration for Object-Based Land Cover Classification in Google Earth Engine

PlanetScope (PL) high-resolution composite base maps have recently become available within Google Earth Engine (GEE) for the tropical regions thanks to the partnership between Google and the Norway’s International Climate and Forest Initiative (NICFI). Object-based (OB) image classification in the GEE environment has increased rapidly due to the broadly recognized advantages of applying these approaches to medium- and high-resolution images. This work aimed to assess the advantages for land cover classification of (a) adopting an OB approach with PL data; and (b) integrating the PL datasets with Sentinel 2 and Sentinel 1 data both in Pixel-based (PB) or OB approaches. For this purpose, in this research, we compared ten LULC classification approaches (PB and OB, all based on the Random Forest (RF) algorithm), where the three satellite datasets were used according to different levels of integration and combination. The study area, which is 69,272 km2 wide and located in central Brazil, was selected within the tropical region, considering a preliminary availability of sample points and its complex landscape mosaic composed of heterogeneous agri-natural spaces, including scattered settlements. Using only the PL dataset with a typical RF PB approach produced the worse overall accuracy (OA) results (67%), whereas adopting an OB approach for the same dataset yielded very good OA (82%). The integration of PL data with the S2 and S1 datasets improved both PB and OB overall accuracy outputs (82 vs. 67% and 91 vs. 82%, respectively). Moreover, this research demonstrated the OB approaches’ applicability in GEE, even in vast study areas and using high-resolution imagery. Although additional applications are necessary, the proposed methodology appears to be very promising for properly exploiting the potential of PL data in GEE.

Effects on Fluvial Geomorphology and Vegetation Cover following Hydroelectric Power Plant Operation: A Case Study in the Maule River (Chile)

The installation of hydroelectric plants has generated multiple environmental impacts on the world’s river systems. In central Chile, the impacts of hydroelectric reservoir operation have been documented in ecological and hydrologic regime terms. This investigation assesses the changes in channel morphology, vegetation distribution, and flows in the middle section of the Maule River during the period following the start-up of a hydroelectric plant. Changes in fluvial morphology (active area) and land cover are quantified using LANDSAT images, contrasted with a vegetation sampling and flow analysis. The results show a 12% decrease in active areas of the river, indicating a loss of geomorphological diversity. Within the active channel, there was a gradual increase in plant-covered surface area, which reached 159% between 1989 and 2018, mainly due to reductions in water (−61%), active bar (−35%), and bare soil surface areas (−29%). The changes were evident ten years after plant operations began and intensified during the period known as the megadrought in central Chile (2008–2018). The flow magnitudes present a decrease for exceedance probabilities (P) below 85% in the period after 1985, with a slight increase recorded for low flows (P > 85%). In the segments with superior stabilization, invasive species such as Acacia dealbata (silver wattle) predominated, which are specialists at taking advantage of disturbances to settle and stabilize active areas, narrowing the possibilities for morphological change.

A Remote Sensing-Based Analysis of the Impact of Syrian Crisis on Agricultural Land Abandonment in Yarmouk River Basin

In this study, we implemented a remote sensing-based approach for monitoring abandoned agricultural land in the Yarmouk River Basin (YRB) in Southern Syria and Northern Jordan during the Syrian crisis. A time series analysis for the Normalized Difference Vegetation Index (NDVI) and Normalized Difference Moisture Index (NDMI) was conducted using 1650 multi-temporal images from Landsat-5 and Landsat-8 between 1986 and 2021. We analyzed the agricultural phenological profiles and investigated the impact of the Syrian crisis on agricultural activities in YRB. The analysis was performed using JavaScript commands in Google Earth Engine. The results confirmed the impact of the Syrian crisis on agricultural land use. The phenological characteristics of NDVI and NDMI during the crisis (2013-2021) were compared to the phenological profiles for the period before the crisis (1986-2010). The NDVI and NDMI profiles had smooth, bell-shaped, and single beak NDVI and NDMI values during the period of crisis in comparison to those irregular phenological profiles for the period before the crisis or during the de-escalation/reconciliation period in the study area. The maximum average NDVI and NDMI values was found in March during the crisis, indicating the progress of natural vegetation and fallow land, while they fluctuated between March and April before the crisis or during the de-escalation/reconciliation period, indicating regular agricultural and cultivation practices.