Search

Understanding the hydrological processes of mountainous regions is crucial for watershed management. The increased frequency of floods in the Himalayan region emphasizes the need to set up a hydrological model in this region. The complex topography and climate patterns of the Himalayas, with a few hydrometeorological stations, make modeling the region challenging. Therefore, the study of hydrological responses using a fully distributed hydrological model in this region is very rare. This study aims to address the challenges of data scarcity in mountain regions using alternative data for observed discharge data for the calibration of the hydrological model. We assess the utility of reanalysis surface runoff data (RSRD) from ERA-5 in calibrating a fully distributed hydrological model WATFLOOD. Six water balance components at nine land-cover classes are analyzed using the WATFLOOD model. The results show that the RSRD can be used as an alternative for the discharge data for calibration of the hydrological model. The evaluation of water balance components shows changes corresponding to wet and dry years. We verified simulation results using observed data, revealing the limitations of calibrating a hydrological model with RSRD. [ABSTRACT FROM AUTHOR]

Cities are important actors in the global challenge of tackling climate change. They are not only responsible for the majority of emissions but also highly capable of taking action. An important precondition for effective climate mitigation is a city‐level greenhouse gas emission inventory to guide mitigation action. Yet, most cities in developing countries, in particular African cities, lack that crucial information. This study aims to assess the current state of the development of African cities' emission inventories. A total of 270 inventories from 137 cities across 54 African countries were identified from 15 research articles, 5 reports, and 3 data platforms. We find the lack of standardized protocols results in inventories that are often not comparable, while data scarcity emerged as a common problem. We observe that insufficient engagement from local governments impedes the creation of a data‐rich environment. Additionally, current inventory protocols do not fully address the data limitations faced by African cities, further hindering inventory development. To mitigate these challenges, multi‐agent collaboration is essential to enhance the accounting capabilities of local governments. Developing refined protocols that consider data constraints is necessary. Moreover, advanced technologies may provide opportunities to overcome data bottlenecks. [ABSTRACT FROM AUTHOR]

Like many small island communities, the U.S. Virgin Islands (USVI), an unincorporated territory of the U.S., is naturally freshwater scarce. In recent decades, rapid land development in the USVI has increased water demand considerably, exerting extra pressure on freshwater resources. Freshwater quantity and quality data for the USVI are very scarce and scattered, which limits freshwater management capabilities. We draw attention to this information deficit and discuss its implications by reviewing the current state of knowledge of surface and groundwater quantity and quality for the USVI. Our review confirms that long‐term records of surface and groundwater quantity and quality are limited and unreliable. For example, streamflow was most recently monitored in 2006, and the most extensive surface water quality records are from the 1960s and 1980s. Since 2016, mean groundwater levels have been recorded daily, but only for three wells (one on each island of the USVI). Importantly, this lack of information threatens water security for the territory and limits our understanding of how development has impacted water quality and availability over time. This could be addressed using models, such as a groundwater recharge model, in combination with remote sensing and updated field data (i.e., streamflow, groundwater, and ecohydrological characterizations of land use change). [ABSTRACT FROM AUTHOR]

Hydrometric data poverty compounds the challenge of accounting for uncertainties in non-stationary stage–discharge relationships. This paper builds on three methods to explore the integration of a dynamic approach to rating curve assessment and a physically based Bayesian framework for quantifying discharge amid geomorphologically induced rating shifts in a sparsely gauged alluvial river. The Modified GesDyn–FlowAM–BaRatin method entails sequentially segmenting gaugings according to residual indicators of riverbed instability and channel conveyance variability, leveraging cross-sectional surveys to augment calibration data, and eliciting hydraulic priors for probabilistic rating curve estimation. This method is applied to a Philippine watershed, where quarrying near the gauging station has ostensibly caused morphodynamic adjustments. Time-variable credible intervals for discharge are computed. The optimal estimates root mean square error (RMSE = 2.96 m3/s) from maximum a posteriori rating curves outperform the hydrographer's benchmark (RMSE = 5.00 m3/s), whose systematic errors from the gauged flows arise from lapses in shift detection. [ABSTRACT FROM AUTHOR]

This study advocates a multi-criteria approach to improve the streamflow predictions in a data-scarce catchment of Chennai metropolitan city of India using the Soil Water and Assessment Tool (SWAT). The remotely sensed evapotranspiration (ET) data, groundwater recharge estimation, and parameter regionalization were used to improve model prediction. Dynamic change of Land Use and Land Cover (LULC) was accounted for along with multi-parameter calibration for reducing the uncertainty in model parameters. The results revealed an improved streamflow prediction accuracy by 10%, especially in the prediction of medium and high flows with the Nash-Sutcliffe efficiency of 0.60. The enhanced parameters were regionalized to ungauged sub-basins and validated using a measured flow event downstream of regionalization with 15% prediction uncertainty. This semi-arid catchment is dominated by ET (58%) and runoff (27%) in the region's hydrology. The finding of this study can be applied to improve the hydrological modelling and predictions in data-scarce regions. [ABSTRACT FROM AUTHOR]

Power circuit breakers (CBs) are vital for the control and protection of power systems, yet diagnosing their faults accurately remains a challenge due to the diversity of fault types and the complexity of their structures. Traditional data-driven methods, although effective, require extensive labeled data for each fault class, limiting their applicability in real-world scenarios where many faults are unseen. This paper addresses these limitations by introducing symptom description transfer-based zero-shot fault diagnosis (SDT-ZSFD), a method that leverages zero-shot learning for fault diagnosis. Our approach constructs a fault symptom description (FSD) framework, which embeds a fault symptom layer between the feature layer and the label layer to facilitate knowledge transfer from seen to unseen fault classes. The method utilizes current and acceleration signals collected during CB operation to extract features. By applying sparse principal component analysis to these signals, we derive high-quality features that are mapped to the FSD framework, enabling effective zero-shot learning. Our method achieves a satisfactory recognition rate by accurately diagnosing unseen faults based on these symptoms. This approach not only overcomes the data scarcity problem but also holds potential for practical applications in power system maintenance. The SDT-ZSFD method offers a reliable solution for CB fault diagnosis and provides a foundation for future improvements in symptom-based zero-shot diagnostic mechanisms and algorithmic robustness. ● A novel method for addressing the challenge of diagnosing new, unknown faults. ● Proposes the concept of symptoms and a zero-shot learning method using symptom description transfer. ● Constructs symptom classifiers from accessible fault data, enabling cross-domain mapping. ● Demonstrates superior performance of SDT-ZSFD in diagnosing faults without prior samples. [ABSTRACT FROM AUTHOR]

Introduction: Developing automatic acne vulgaris grading systems based on machine learning is an expensive endeavor in terms of data acquisition. A machine learning practitioner will need to gather high-resolution pictures from a considerable number of different patients, with a well-balanced distribution between acne severity grades and potentially very tedious labeling. We developed a deep learning model to grade acne severity with respect to the Investigator's Global Assessment (IGA) scale that can be trained on low-resolution images, with pictures from a small number of different patients, a strongly imbalanced severity grade distribution and minimal labeling. Methods: A total of 1374 triplets of images (frontal and lateral views) from 391 different patients suffering from acne labeled with the IGA severity grade by an expert dermatologist were used to train and validate a deep learning model that predicts the IGA severity grade. Results: On the test set we obtained 66.67% accuracy with an equivalent performance for all grades despite the highly imbalanced severity grade distribution of our database. Importantly, we obtained performance on par with more tedious methods in terms of data acquisition which have the same simple labeling as ours but require either a more balanced severity grade distribution or large numbers of high-resolution images. Conclusions: Our deep learning model demonstrated promising accuracy despite the limited data set on which it was trained, indicating its potential for further development both as an assistance tool for medical practitioners and as a way to provide patients with an immediately available and standardized acne grading tool. Trial Registration: chinadrugtrials.org.cn identifier CTR20211314. [ABSTRACT FROM AUTHOR]

Automatic Grammar Error Correction (GEC) models identify and correct a wide range of grammatical errors. Various strategies have been proposed for GEC, with the Neural Machine Translation (NMT) approach being the most effective. However, NMT-based GEC models with encoderdecoder layers rely heavily on the highest layer, leading to potential inaccuracies. Additionally, during inference, exposure bias can cause the model to substitute previously targeted words with incorrect alternatives. Another challenge is data scarcity. This paper introduces a GEC model leveraging the seq-to-seq Transformer framework, specifically designed for low-resource languages like Arabic. We propose a method to generate noise in the text to create synthetic parallel data, addressing the data constraints. Inspired by Capsule Networks (CapsNet), we incorporate CapsNet in GEC to dynamically aggregate information from multiple layers. In order to mitigate exposure bias, we incorporated a bidirectional training approach and a regularization term using Kullback-Leibler divergence to align left-toright and right-to-left models. Experiments on two benchmarks demonstrate that our model outperforms current Arabic GEC models, achieving the highest scores. The code is available on GitHub (https://github.com/Zainabobied/ArabicGEC). [ABSTRACT FROM AUTHOR]

The main challenges in smart home systems and cyber-physical systems come from not having enough data and unclear interpretation; thus, there is still a lot to be done in this field. In this work, we propose a practical approach called Discrete Human Activity Recognition (DiscHAR) based on prior research to enhance Human Activity Recognition (HAR). Our goal is to generate diverse data to build better models for activity classification. To tackle overfitting, which often occurs with small datasets, we generate data and convert them into discrete forms, improving classification accuracy. Our methodology includes advanced techniques like the R-Frame method for sampling and the Mixed-up approach for data generation. We apply K-means vector quantization to categorize the data, and through the elbow method, we determine the optimal number of clusters. The discrete sequences are converted into one-hot encoded vectors and fed into a CNN model to ensure precise recognition of human activities. Evaluations on the OPP79, PAMAP2, and WISDM datasets show that our approach outperforms existing models, achieving 89% accuracy for OPP79, 93.24% for PAMAP2, and 100% for WISDM. These results demonstrate the model's effectiveness in identifying complex activities captured by wearable devices. Our work combines theory and practice to address ongoing challenges in this field, aiming to improve the reliability and performance of activity recognition systems in dynamic environments. [ABSTRACT FROM AUTHOR]

The scarcity of medical images significantly hampers the advancement of artificial intelligence techniques in the medical field. Medical images face the issue of inferior training accuracy and efficiency in classification tasks due to insufficient labeled data and privacy preserving demands. To address these issues, we propose a new and efficient federated learning (FL) scheme based on blockchain, called FedBG, to generate realistic images to expand the dataset for medical image classification. First, we design an enhanced classification-GAN (EC-GAN) algorithm based on cross-entropy, which expands the training target of the generator from a single improvement of image fidelity to a direction related to the classification tasks. EC-GAN increases the diversity of generated images and reduces the risk of pattern collapse. Second, we improve the generator’s loss function, which is guided by the cross-entropy loss term of the classifier model and the gradient information from the discriminator model, to generate medical images with enhanced category features. This design overcomes the limitation of traditional GAN that focuses only on image authenticity judgment, and achieves the dual optimization of synthetic image and classification accuracy. Finally, we propose an improved consensus mechanism based on maximum mean discrepancy and data contribution to ensure the consistency and security of data in the blockchain. The mechanism optimizes the fairness and efficiency of the model training by dynamically evaluating the contribution of each participating node, and guarantees the trust system of FL. Experiments are performed on two datasets and results demonstrate that FedBG reduces the training time by 27–38% and improves the accuracy by 0.9–2% compared to existing methods while ensuring data privacy and generating high-quality images. [ABSTRACT FROM AUTHOR]

[Display omitted] • We propose an unsupervised few-shot learning for landslide susceptibility mapping. • The meta-learning approach excels in identifying landslides in data-scarce regions. • The visualization of the proposed susceptibility map identified new landslides. Given the critical need to assess landslide hazards, producing landslide susceptibility map (LSM) in regions with scarce historical landslide inventories poses significant challenges. This study introduces a novel landslide susceptibility assessment framework that combines unsupervised learning strategies with few-shot learning methods to increase the accuracy of LSM in these areas. The framework has been practically validated in a representative geological disaster-prone area along the West-East Gas Pipeline in Shaanxi Province, China. We employed three advanced few-shot learning models: a support vector machine, meta -learning, and transfer learning. These models implement feature representation learning for weakly correlated influencing factors through an unsupervised approach, thereby constructing an effective landslide susceptibility assessment model. We compared traditional learning methods and used the receiver operating characteristic (ROC) curve and SHAP values to quantify the effectiveness of the models. The results indicate that the meta -learning algorithm outperforms both the SVM and transfer learning in areas with limited landslide data. The integration of unsupervised strategies significantly improves performance, achieving area under the curve (AUC) values of 0.9385 and 0.9861, respectively. Compared with using meta -learning alone, incorporating unsupervised learning strategies increased the AUC by 4.76%, enhancing both the predictive power of the model and the interpretability of the features. Meta-learning under unsupervised conditions effectively mitigates the evaluation difficulties caused by insufficient landslide records, providing a viable path and empirical evidence for performance improvement in similar data- scarce regions worldwide. [ABSTRACT FROM AUTHOR]

Abstract Grounded Situation Recognition (GSR) aims to generate structured image descriptions. For a given image, GSR needs to identify the key verb, the nouns corresponding to roles, and their bounding-box groundings. However, current GSR research demands numerous meticulously labeled images, which are labor-intensive and time-consuming, making it costly to expand detection categories. Our study enhances model accuracy in detecting and localizing under data scarcity, reducing dependency on large datasets and paving the way for broader detection capabilities. In this paper, we propose the Grounded Situation Recognition under Data Scarcity (GSRDS) model, which uses the CoFormer model as the baseline and optimizes three subtasks: image feature extraction, verb classification, and bounding-box localization, to better adapt to data-scarce scenarios. Specifically, we replace ResNet50 with EfficientNetV2-M for advanced image feature extraction. Additionally, we introduce the Transformer Combined with CLIP for Verb Classification (TCCV) module, utilizing features extracted by CLIP’s image encoder to enhance verb classification accuracy. Furthermore, we design the Multi-source Verb-Role Queries (Multi-VR Queries) and the Dual Parallel Decoders (DPD) modules to improve the accuracy of bounding-box localization. Through extensive comparative experiments and ablation studies, we demonstrate that our method achieves higher accuracy than mainstream approaches in data-scarce scenarios. Our code will be available at https://github.com/Zhou-maker-oss/GSRDS .

Grounded Situation Recognition (GSR) aims to generate structured image descriptions. For a given image, GSR needs to identify the key verb, the nouns corresponding to roles, and their bounding-box groundings. However, current GSR research demands numerous meticulously labeled images, which are labor-intensive and time-consuming, making it costly to expand detection categories. Our study enhances model accuracy in detecting and localizing under data scarcity, reducing dependency on large datasets and paving the way for broader detection capabilities. In this paper, we propose the Grounded Situation Recognition under Data Scarcity (GSRDS) model, which uses the CoFormer model as the baseline and optimizes three subtasks: image feature extraction, verb classification, and bounding-box localization, to better adapt to data-scarce scenarios. Specifically, we replace ResNet50 with EfficientNetV2-M for advanced image feature extraction. Additionally, we introduce the Transformer Combined with CLIP for Verb Classification (TCCV) module, utilizing features extracted by CLIP's image encoder to enhance verb classification accuracy. Furthermore, we design the Multi-source Verb-Role Queries (Multi-VR Queries) and the Dual Parallel Decoders (DPD) modules to improve the accuracy of bounding-box localization. Through extensive comparative experiments and ablation studies, we demonstrate that our method achieves higher accuracy than mainstream approaches in data-scarce scenarios. Our code will be available at https://github.com/Zhou-maker-oss/GSRDS. [ABSTRACT FROM AUTHOR]

The presence of "ill-posed samples" specifically in low-volume datasets leads to accuracy decrement in the learning procedure and the generalization of neural networks. Such samples can be caused by various reasons such as noise contamination, corrupted sensors, or even, the complex distribution of physical properties governing the problem. The peak ground acceleration (PGA) datasets are definitely among the last mentioned. Focusing on speed and accuracy, a method for calculating earthquake magnitude based on the PGA data recorded at a single station along with hypocentral information has been presented in this research. Here, after training a deep neural network, the regression errors of the training data samples are clustered into two groups, namely well and ill posed using the grey wolf optimization algorithm. Instead of being removed, the data samples with low learning rates are then modified using samples selected from the other cluster in a fusional form. Then, two separate models are used and trained independently for the clusters. Next, in addition to the routine procedure of network generalization, every new sample is first checked whether is more likely to belong to which group of the clustered data, and after processing, the corresponding trained model is used. The results of the experiments show that using the proposed method results in magnitude calculation with an error order of less than 0.212 units of moment magnitude with a probability of more than 99.7%, which is superior to the conventional methods some of which were reviewed in this research. [ABSTRACT FROM AUTHOR]

Credit risk refers to the possibility of borrower default, and its assessment is crucial for maintaining financial stability. However, the journey of credit risk data generation is often gradual, and machine learning techniques may not be readily applicable for crafting evaluations at the initial stage of the data accumulation process. This article proposes a credit risk modeling methodology, TED-NN, that first constructs an indicator system based on expert experience, assigns initial weights to the indicator system using the Analytic Hierarchy Process, and then constructs a neural network model based on the indicator system to achieve a smooth transition from an empirical model to a data-driven model. TED-NN can automatically adapt to the gradual accumulation of data, which effectively solves the problem of risk modeling and the smooth transition from no to sufficient data. The effectiveness of this methodology is validated through a specific case of credit risk assessment. Experimental results on a real-world dataset demonstrate that, in the absence of data, the performance of TED-NN is equivalent to the AHP and better than untrained neural networks. As the amount of data increases, TED-NN gradually improves and then surpasses the AHP. When there are sufficient data, its performance approaches that of a fully data-driven neural network model. [ABSTRACT FROM AUTHOR]

Copyright of Journal of National University of Defense Technology / Guofang Keji Daxue Xuebao is the property of NUDT Press and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Many fall detection systems are being used to provide real-time responses to fall occurrences. Automated fall detection is challenging because it requires very high accuracy to be clinically acceptable. Recent research has tried to improve sensitivity while reducing the high rate of false positives. Nevertheless, there are still limitations in terms of having efficient learning approaches and proper datasets to train. To reduce false alarms, one approach is to add more nonfall data as negative samples to train the deep learning model. However, this approach increases class imbalance in the training set. To tackle this problem, we propose a multi-task deep learning approach that divides datasets into multiple training sets for multiple tasks. We prove this approach gives better results than a single-task model trained on all datasets. Many experiments are conducted to find the best combination of tasks for multi-task model training for fall detection. [ABSTRACT FROM AUTHOR]

Load profiling for residential aggregates encounters challenges due to data scarcity and the inadequacy of standard profiles obtained from statistical analyses. In the absence of hourly data, many methods rely on standard profiles, which could lead to significant errors in consumption estimation, especially for evaluating specific aggregates. This article presents PyARC, a Python-based algorithm trainable with customizable consumption data, which addresses the problem related to evaluating the energy consumption of specific aggregates by using typological profiles extracted from similar users, thereby improving accuracy. The algorithm's innovative approach uses Association Rule Mining and Random Forest Classification to reconstruct the load profiles of aggregates, providing a more robust solution for estimating the electrical load with limited data.

Abstract Data scarcity in paleographic image datasets poses a significant challenge to researchers and scholars in the field. Unlike modern printed texts, historical manuscripts and documents are often scarce and fragile, making them difficult to digitize and create comprehensive datasets. Recently many innovations have been arrived on single image generative models for natural images but none of them are focused on paleographic character images and other handwritten datasets. In paleographic images like stone inscription characters, maintaining exact shape and structure of character is important unlike natural images. In this paper we propose an unconditional single image generative model, CharGAN for isolated paleographic character images. In the proposed system, augmented images are generated from a single image using generative adversarial networks, while maintaining their structure. Specifically, an external augmentation inducer is used to create higher-level augmentations in the generated images. In addition, the input to the generator is replaced with dynamic sampling from a Gaussian mixture model to make changes to the low-level features. From our experimental results, we infer that these two enhancements make single-image generative models suitable not only for natural images, but also for paleographic character images and other handwritten character datasets, the AHCD dataset, and EMNIST, where the global structure is important. Both the qualitative and quantitative results show that our approach is effective and superior in single-image generative tasks, particularly in isolated character image generation.

Many fall detection systems are being used to provide real-time responses to fall occurrences. Automated fall detection is challenging because it requires very high accuracy to be clinically acceptable. Recent research has tried to improve sensitivity while reducing the high rate of false positives. Nevertheless, there are still limitations in terms of having efficient learning approaches and proper datasets to train. To reduce false alarms, one approach is to add more nonfall data as negative samples to train the deep learning model. However, this approach increases class imbalance in the training set. To tackle this problem, we propose a multi-task deep learning approach that divides datasets into multiple training sets for multiple tasks. We prove this approach gives better results than a single-task model trained on all datasets. Many experiments are conducted to find the best combination of tasks for multi-task model training for fall detection.

Abstract Visual speech recognition (VSR) is a challenging task that has received increasing interest during the last few decades. Current state of the art employs powerful end-to-end architectures based on deep learning which depend on large amounts of data and high computational resources for their estimation. We address the task of VSR for data scarcity scenarios with limited computational resources by using traditional approaches based on hidden Markov models. We present a novel learning strategy that employs information obtained from previous acoustic temporal alignments to improve the visual system performance. Furthermore, we studied multiple visual speech representations and how image resolution or frame rate affect its performance. All these experiments were conducted on the limited data VLRF corpus, a database which offers an audio-visual support to address continuous speech recognition in Spanish. The results show that our approach significantly outperforms the best results achieved on the task to date.

Data scarcity in paleographic image datasets poses a significant challenge to researchers and scholars in the field. Unlike modern printed texts, historical manuscripts and documents are often scarce and fragile, making them difficult to digitize and create comprehensive datasets. Recently many innovations have been arrived on single image generative models for natural images but none of them are focused on paleographic character images and other handwritten datasets. In paleographic images like stone inscription characters, maintaining exact shape and structure of character is important unlike natural images. In this paper we propose an unconditional single image generative model, CharGAN for isolated paleographic character images. In the proposed system, augmented images are generated from a single image using generative adversarial networks, while maintaining their structure. Specifically, an external augmentation inducer is used to create higher-level augmentations in the generated images. In addition, the input to the generator is replaced with dynamic sampling from a Gaussian mixture model to make changes to the low-level features. From our experimental results, we infer that these two enhancements make single-image generative models suitable not only for natural images, but also for paleographic character images and other handwritten character datasets, the AHCD dataset, and EMNIST, where the global structure is important. Both the qualitative and quantitative results show that our approach is effective and superior in single-image generative tasks, particularly in isolated character image generation. [ABSTRACT FROM AUTHOR]

The accurate prediction of global horizontal irradiance (GHI) is crucial for optimizing solar power generation systems, particularly in mountainous areas with complex topography and unique microclimates. These regions face significant challenges due to limited reliable data and the dynamic nature of local weather conditions, which complicate accurate GHI measurement. The scarcity of precise data impedes the development of reliable solar energy prediction models, impacting both economic and environmental outcomes. To address these data scarcity challenges in solar energy prediction, this paper focuses on various locations in Europe and Asia Minor, predominantly in mountainous regions. Advanced machine learning techniques, including random forest (RF) and extreme gradient boosting (XGBoost) regressors, are employed to effectively predict GHI. Additionally, optimizing training data distribution based on cloud opacity values and integrating synthetic data significantly enhance predictive accuracy, with R2 scores ranging from 0.91 to 0.97 across multiple locations. Furthermore, substantial reductions in root mean square error (RMSE), mean absolute error (MAE), and mean bias error (MBE) underscore the improved reliability of the predictions. Future research should refine synthetic data generation, optimize additional meteorological and environmental parameter integration, extend methodology to new regions, and test for predicting global tilted irradiance (GTI). The studies should expand training data considerations beyond cloud opacity, incorporating sky cover and sunshine duration to enhance prediction accuracy and reliability. [ABSTRACT FROM AUTHOR]

[Display omitted] • We propose a feature-based transfer learning method for landslide susceptibility assessment. • The method effectively improves the cross-regional landslide susceptibility assessment. • Landslide wading elevation and geological lithology are the main influencing factors. Given the time-consuming nature of compiling landslide inventories, it is increasingly important to develop transferable landslide susceptibility models that can be applied to regions without existing data. In this study, we propose a feature-based domain adaptation method to improve the transferability of landslide susceptibility models, especially in "no sample" areas. Two typical landslide-prone areas in Fujian province, southeastern China, were chosen as research cases to test the practicality of the transfer effect. Five conventional machine learning algorithms (Support vector machines (SVM), Random Forest (RF), Logistic Regression (LOG), K-nearest neighbor (KNN), and Decision tree (C4.5)) are used to model landslide susceptibility in sampled areas (source domain), and a feature transfer-based landslide susceptibility evaluation model is constructed under coupled feature transfer methods to evaluate the susceptibility of landslide in un-sampled areas (target domain). The results showed that feature transfer can effectively improve the transferability of different machine learning models for cross-regional prediction (The indicators have improved overall by 8.49%), with SVM (increased by 13.68%) and LOG (increased by 10.19%) models showing the most significant improvements. The feature-based domain adaptive method can alleviate the burden of collecting and labeling new data, and effectively improve the assessment performance of machine learning-based landslide susceptibility models in un-sampled areas. This is a new solution for landslide susceptibility assessment in completely "no sample" areas. [ABSTRACT FROM AUTHOR]

Machine learning (ML) applications in hydrology are revolutionizing our understanding and prediction of hydrological processes, driven by advancements in artificial intelligence and the availability of large, high-quality datasets. This review explores the current state of ML applications in hydrology, emphasizing the utilization of extensive datasets such as CAMELS, Caravan, GRDC, CHIRPS, NLDAS, GLDAS, PERSIANN, and GRACE. These datasets provide critical data for modeling various hydrological parameters, including streamflow, precipitation, groundwater levels, and flood frequency, particularly in data-scarce regions. We discuss the type of ML methods used in hydrology and significant successes achieved through those ML models, highlighting their enhanced predictive accuracy and the integration of diverse data sources. The review also addresses the challenges inherent in hydrological ML applications, such as data heterogeneity, spatial and temporal inconsistencies, issues regarding downscaling the LSH, and the need for incorporating human activities. In addition to discussing the limitations, this article highlights the benefits of utilizing high-resolution datasets compared to traditional ones. Additionally, we examine the emerging trends and future directions, including the integration of real-time data and the quantification of uncertainties to improve model reliability. We also place a strong emphasis on incorporating citizen science and the IoT for data collection in hydrology. By synthesizing the latest research, this paper aims to guide future efforts in leveraging large datasets and ML techniques to advance hydrological science and enhance water resource management practices. [ABSTRACT FROM AUTHOR]

Detecting abnormal human behaviors in surveillance videos is crucial for various domains, including security and public safety. Many successful detection techniques based on deep learning models have been introduced. However, the scarcity of labeled abnormal behavior data poses significant challenges for developing effective detection systems. This paper presents a comprehensive survey of deep learning techniques for detecting abnormal human behaviors in surveillance video streams. We categorize the existing techniques into three approaches: unsupervised, partially supervised, and fully supervised. Each approach is examined in terms of its underlying conceptual framework, strengths, and drawbacks. Additionally, we provide an extensive comparison of these approaches using popular datasets frequently used in the prior research, highlighting their performance across different scenarios. We summarize the advantages and disadvantages of each approach for abnormal human behavior detection. We also discuss open research issues identified through our survey, including enhancing robustness to environmental variations through diverse datasets, formulating strategies for contextual abnormal behavior detection. Finally, we outline potential directions for future development to pave the way for more effective abnormal behavior detection systems. [ABSTRACT FROM AUTHOR]

Available data to depict socioeconomic realities are often scarce at the municipal level. Unlike recurring or continuous data, which are collected regularly or repeatedly, nonrecurrent data may be sporadic or irregular, due to significant costs for their compilation and limited resources at municipalities. To address regional data scarcity, we develop a bottom-up top-down methodology for constructing synthetic socioeconomic indicators combining a genetic algorithm and regression techniques. We apply our methodology for assessing income inequalities at 178 municipalities in Spain. The genetic algorithm draws the available data on circumstances or inequalities of opportunities that give birth to income disparities. Our methodology allows to mitigate the shortcomings arising from unavailable data. Thus, it is a suitable method to assess relevant socioeconomic conditions at a regional level that are currently obscured due to data unavailability. This is crucial to provide policymakers with an enhanced socioeconomic overview at regional administrative units, relevant to allocating public service funds. [ABSTRACT FROM AUTHOR]

Visual speech recognition (VSR) is a challenging task that has received increasing interest during the last few decades. Current state of the art employs powerful end-to-end architectures based on deep learning which depend on large amounts of data and high computational resources for their estimation. We address the task of VSR for data scarcity scenarios with limited computational resources by using traditional approaches based on hidden Markov models. We present a novel learning strategy that employs information obtained from previous acoustic temporal alignments to improve the visual system performance. Furthermore, we studied multiple visual speech representations and how image resolution or frame rate affect its performance. All these experiments were conducted on the limited data VLRF corpus, a database which offers an audio-visual support to address continuous speech recognition in Spanish. The results show that our approach significantly outperforms the best results achieved on the task to date. [ABSTRACT FROM AUTHOR]

Copyright of Computer Measurement & Control is the property of Magazine Agency of Computer Measurement & Control and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

The imbalance in global streamflow gauge distribution and regional data scarcity, especially in large transboundary basins, challenge regional water resource management. Effectively utilizing these limited data to construct reliable models is of crucial practical importance. This study employs a transfer learning (TL) framework to simulate daily streamflow in the Dulong-Irrawaddy River Basin (DIRB), a less-studied transboundary basin shared by Myanmar, China, and India. Our results show that TL significantly improves streamflow predictions: the optimal TL model achieves an average Nash-Sutcliffe efficiency of 0.872, showing a marked improvement in the Hkamti sub-basin. Despite data scarcity, TL achieves a mean NSE of 0.817, surpassing the 0.655 of the process-based model MIKE SHE. Additionally, our study reveals the importance of source model selection in TL, as different parts of the flow are affected by the diversity and similarity of data in the source model. Deep learning models, particularly TL, exhibit complex sensitivities to meteorological inputs, more accurately capturing non-linear relationships among multiple variables than the process-based model. Integrated gradients (IG) analysis further illustrates TL's ability to capture spatial heterogeneity in upstream and downstream sub-basins and its adeptness in characterizing different flow regimes. This study underscores the potential of TL in enhancing the understanding of hydrological processes in large-scale catchments and highlights its value for water resource management in transboundary basins under data scarcity. [ABSTRACT FROM AUTHOR]

Data scarcity in developing countries often poses significant challenges to the use of analytics in addressing development issues. In transportation research, digitized road data is one of the most fundamental data structures, and a poorly digitized road network significantly reduces the ability to optimize trade of micro-enterprises (SDG 8) and placement of hospitals (SDG 3). Unfortunately, current methods to enhance or create digital road networks are not well-adapted to regions with sparse geospatial data, often resulting in poor digital representations of road networks in less-developed regions such as rural areas of developing countries. We present a novel projection-based incremental insertion method that is well-suited to either enhance large existing road networks or combine multiple sources of road networks in regions with sparse geospatial data. In collaboration with PemPem and the World Bank, we perform two case studies that demonstrate the effectiveness of the proposed method. Together with PemPem, we show that our method significantly improves the digital road network for smallholder farmers in Indonesia, where only 40% of the origin–destination pairs in our dataset were previously digitized. Moreover, in a case study of optimizing geospatial accessibility to healthcare in Timor-Leste (World Bank), the improved digital road network detects an additional 5% of people to be in the vicinity of a hospital.

Developing countries face significant challenges in accessing sufficient and reliable hydro-meteorological data, hindering the implementation of effective disaster management strategies. This research proposes solutions for limitations on performing flood simulations through parameter sensitivity analysis and digital elevation model (DEM) modifications. The methodology provides alternatives to account for insufficient rainfall, roughness coefficient, infiltration data in simulating large-scale rainfall-runoff, and high-resolution DEMs incorporating road and building networks for urban flood modeling. By applying the method to an extreme flood event in the Marikina Basin, Philippines, a combination of ground-based and remotely retrieved rainfall data, roughness (n = 0.3861–0.5005), and infiltration parameters (Δθ = 0.326–0.505 and ψ = 0.4547–1.565) set at the maximum range were found to replicate the increase in the upstream water level. Simulations were able to accurately capture the flood propagation along the natural and artificial barriers in the urban area compared to untreated digital terrain and surface model (DTM and DSM) data, with root-mean-square error range improvements from 0–7.13 (DTM) and 0.29–4.20 (DSM) to 0–0.63 (modified DEM). The proposed methodology significantly improved the accuracy of the simulations, which is crucial for proposing adequate flood action plans, despite the lack of high-resolution data available for under-developed nations. HIGHLIGHTS Developed a GIS technique for modifying the DEM based on the DTM and the DSM.; Identified a method for adjusting flood model input parameter using sensitivity analysis.; Proposed solutions for the scarcity of high-resolution data for simulating urban floods in developing countries.; Identified the flood extent and flash-flood prone areas in the study area.;

For decades, there has been ongoing discussion about whether centralized or decentralized wastewater management systems are better. Decision-makers need to define the best option but do not always have the necessary tools to develop, compare, and identify the most appropriate solution. To address this, studies have been conducted on a settlement level. In this study, the main focus was to develop and optimize wastewater management scenarios for a region containing rural areas, where data scarcity was an issue, by extracting scenario-relevant information from the region using a satellite image and its calibration using locally available data. We selected a study region in India containing 184 villages with a total population of around 210,000 and covering an area of around 400 km2. The study considered three different scenarios for the study area: centralized, decentralized, and an optimized scenario, which consists of a hybrid system involving partly decentralized and partly semi-centralized (clustered) infrastructure. The study developed a systematic approach for defining an optimized cluster of villages by considering the cost trade-off between the wastewater treatment plant (WWTP) capacity and sewer network layout. The results showed that the clustered and decentralized scenarios were nearly equal in terms of cost (around EUR 118 million), while the centralized scenario showed a relatively high cost of EUR 168 million. Potential applications and further development of the method were also considered. The proposed methodology may aid global wastewater management by estimating and optimizing infrastructure costs needed to fulfill Sustainable Development Goal 6 (SDG#6) in rural regions. [ABSTRACT FROM AUTHOR]

Background and Objectives: The complex knowledge of local communities on the full cycle of disaster risk management has been proven valuable in various researches. However, the scientific literature still lacks studies that examine how to use Local Knowledge (LK) and the local people's capabilities for crowdsourcing in Flash Flood Early Warning Systems (FFEWS) studies. Hence the main target of this research is the investigation of the capacity of crowdsourcing for FFEWS and the identification of Flash Flood Hotspots (FFHs) by LK across a flood-prone area in northeast of Iran. Materials and Methods: In this study, a questionnaire with three different themes was designed. The first theme was related to the individual characteristics as independent variables, and the second theme addressed the residents' LK in determining the FFHs, the type and the predominant time of the flood occurrence in the region, by asking open-ended questions with short answers. The last theme addressed the assessment of people's capacity in Flash Flood Crowdsourcing (FFC), through asking questions with a Likert scale of 0-5. The face-to-face questionnaire administration mode was used for public surveys through conducting oral interviews and live discussions. Results: The results showed that there was no significant correlation between the individuals' characteristics and their willingness and motivation to participate in FFC. Comparing residents' LK with the 31- year flood report and literature review showed that the residents' LK about the flood occurrence location, time and type on a local scale was very promising. The research results indicated that the respondents showed the highest level of willingness to participate in the release of flood warning messages with an average score of 4.23 and the most important motivating factor for their willingness to participate was introducing saving relatives, fellow villagers, and human being from flood hazards with an average score of 4.84. Conclusion: In the most of the previous studies that have focused on the development of FFEWS, very little attention has been paid to understanding the needs of citizens and promoting their participation. There is a research gap regarding the method of citizen's participation and their potential support for FFEWS. Hence in this research, an attempt was made to take a small step towards filling this gap by investigating LK, motivation and willingness of local residents to participate in various aspects and steps of FFEWS. Our findings indicate that involving local people in FFEWS has various unknown aspects that should be explored through more extensive and detailed studies. [ABSTRACT FROM AUTHOR]

Determination of reference evapotranspiration (ETo) in required for design, management and scheduling of irrigation water in fan and pad greenhouses. In actual practice estimation of (ETo) in fan and pad greenhouses is often made using the Penman-Monteith FAO-56-PM; method from external meteorological data. This requires availability of accurate meteorological input data (temperature, relative humidity, wind speed, and solar radiation). This is constrained by lack of such data which is a common problem in developing countries. In this study the proposed procedure to estimate ETo is based on using limited data of outdoor historically recorded climate elements of only temperature wind speed, and site characteristics (altitude, latitude and sun shine hours). In the proposed method radiation is to be predicted from data of air temperature difference rather than its direct measurement. This because radiation measurement using pyranometers and net radiometers is borne to errors calibration errors commonly plagued by hysteresis, and nonlinearity. The obtained results of the proposed alternative procedure were statistically validated in comparison with the standard method (FAO 56 PM) using unlimited input data measured inside the greenhouse and in reference to a directly measured ETo values by class-A-evaporation pan. The performance of the developed model was evaluated by the determination coefficient of the regression "R² for goodness-of-fit" and by using the Root Mean Square Error (RMSE). The needed data is collected during three years in three sites in Khartoum North-Sudan El Alafoon, Halfaya, and Shambat. In each site three greenhouses were employed, and data is taken every three days for three months in each year. The obtained result reveals that the proposed limited data procedure to estimate the ETo inside greenhouses agree on statistical basis well with both pan measurement and PM estimation from measured indoor climate variables. The study reveals importance of temperature data for estimating ETo in greenhouses and calls for insuring high quality temperature data for calculating ETo in fan and pad greenhouses. [ABSTRACT FROM AUTHOR]

Study region: Afghanistan Study focus: Thirteen global gridded precipitation datasets (GGPDs) were evaluated against the historical records, based on the statistical indices and Scored-based system in four temporal and three spatial scales. The evaluation results reveal that across the country, the CFSR, PERSIANN, IMERG, PERSIANN_CCS, and GPCC datasets consistently demonstrate strong performance. The present study addresses the challenge of precipitation data scarcity by investigating reliable alternative data for diverse research and applications in Afghanistan. New hydrological insights for the region: GGPDs are reliable sources for water-related studies in regions with sparse and limited ground observations. This study, compared 13 GGPDs against 75 observation stations across Afghanistan between 2009 and 2020. Statistical analyses at daily, monthly, and seasonal resolutions assessed GGPDs using Pearson's correlation coefficient, root-mean-square error, and bias. To unify these statistical measures, a weighted scoring system was employed. Quantitative spatial assessment of GGPDs against observations considered factors such as river basins, climate zones, and elevations. Precipitation from GGPDs over same coordinates as the stations, were further analysed. Spatial data were generated from observations using inverse distance weighting. Discrepancies between observed and GGPDs were measured at seasonal and annual time scales. Analysis revealed mean annual precipitation values in observations ranging from 70 to 502 mm. Regarding mean annual, PERSIANN_CCS, GPCC, TRMM, and CFSR tended to overestimate, while the PERSIANN and CMORPH-v1 consistently underestimated precipitation across the country. CFSR and PERSIANN demonstrated higher reliability, averaging scores of 72 and 68, respectively. In contrast, GPCP and CMORPH-v1 displayed lower reliability, scoring averages of 56 and 57. The study area covers mountainous region of the world with complicated hydrology. The findings can provide a deeper insight into the performance of the GGPDs over a region of the world where less attention has been paid by the scholars. These findings guide researchers in selecting suitable GGPDs for specific applications, effectively mitigating data scarcity concerns through enhanced temporal and spatial coverage, especially in regions with limited observational data.