TSAE-UNet: A Novel Network for Multi-Scene and Multi-Temporal Water Body Detection Based on Spatiotemporal Feature Extraction DOI Creative Commons
Shuai Wang, Yu Chen, Yafei Yuan

et al.

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(20), P. 3829 - 3829

Published: Oct. 15, 2024

The application of remote sensing technology in water body detection has become increasingly widespread, offering significant value for environmental monitoring, hydrological research, and disaster early warning. However, the existing methods face challenges multi-scene multi-temporal detection, including diverse variations shapes sizes that complicate detection; complexity land cover types, which easily leads to false positives missed detections; high cost acquiring high-resolution images, limiting long-term applications; lack effective handling data, making it difficult capture dynamic changes bodies. To address these challenges, this study proposes a novel network based on spatiotemporal feature extraction, named TSAE-UNet. TSAE-UNet integrates convolutional neural networks (CNN), depthwise separable convolutions, ConvLSTM, attention mechanisms, significantly improving accuracy robustness by capturing multi-scale features establishing dependencies. Otsu method was employed quickly process Sentinel-1A Sentinel-2 generating high-quality training dataset. In first experiment, five rectangular areas approximately 37.5 km2 each were selected validate performance model across different scenes. second experiment focused Jining City, Shandong Province, China, analyzing monthly from 2020 2022 quarterly 2022. experimental results demonstrate excels achieving precision 0.989, recall 0.983, an F1 score 0.986, IoU 0.974, outperforming FCN, PSPNet, DeepLabV3+, ADCNN, MECNet.

Language: Английский

Forecast-Informed Reservoir Operations within a Satellite-Based Framework for Mountainous and High-Precipitation Regions: Case of the 2018 Kerala Floods DOI
Pritam Das, Sarath Suresh, Faisal Hossain

et al.

Journal of Hydrologic Engineering, Journal Year: 2025, Volume and Issue: 30(2)

Published: Jan. 24, 2025

Language: Английский

Citations

0
Land Use and Land Cover Change Dynamics in the Niger Delta Region of Nigeria from 1986 to 2024 DOI Creative Commons

Obroma O. Agumagu,

Rob Marchant, Lindsay C. Stringer

et al.

Land, Journal Year: 2025, Volume and Issue: 14(4), P. 765 - 765

Published: April 3, 2025

Land Use and Cover Change (LULCCs) shapes catchment dynamics is a key driver of hydrological risks, affecting responses as vegetated land replaced with urban developments cultivated land. The resultant risks are likely to become more critical in the future climate changes becomes increasingly variable. Understanding effects LULCC vital for developing management strategies reducing adverse on cycle environment. This study examines Niger Delta Region (NDR) Nigeria from 1986 2024. A supervised maximum likelihood classification was applied Landsat 5 TM 8 OLI images 1986, 2015, Five use classes were classified: Water bodies, Rainforest, Built-up, Agriculture, Mangrove. overall accuracy Kappa coefficients 93% 0.90, 91% 0.87, 84% 0.79 2024, respectively. Between built-up agriculture areas substantially increased by about 8229 6727 km2 (561% 79%), respectively, concomitant decrease mangrove vegetation 14,350 10,844 (−54% −42%), spatial distribution across NDR states varied, Delta, Bayelsa, Cross River, Rivers States experiencing highest rainforest, losses 64%, 55, 44%, 44% (5711 km2, 3554 2250 1297 km2), NDR’s mangroves evidently under serious threat. has important implications, particularly given role played forests regulating hazards. dramatic rainforest could exacerbate climate-related impacts. provides quantitative information that be used support planning practices well sustainable development.

Language: Английский

Citations

0
TSAE-UNet: A Novel Network for Multi-Scene and Multi-Temporal Water Body Detection Based on Spatiotemporal Feature Extraction DOI Creative Commons
Shuai Wang, Yu Chen, Yafei Yuan

et al.

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(20), P. 3829 - 3829

Published: Oct. 15, 2024

The application of remote sensing technology in water body detection has become increasingly widespread, offering significant value for environmental monitoring, hydrological research, and disaster early warning. However, the existing methods face challenges multi-scene multi-temporal detection, including diverse variations shapes sizes that complicate detection; complexity land cover types, which easily leads to false positives missed detections; high cost acquiring high-resolution images, limiting long-term applications; lack effective handling data, making it difficult capture dynamic changes bodies. To address these challenges, this study proposes a novel network based on spatiotemporal feature extraction, named TSAE-UNet. TSAE-UNet integrates convolutional neural networks (CNN), depthwise separable convolutions, ConvLSTM, attention mechanisms, significantly improving accuracy robustness by capturing multi-scale features establishing dependencies. Otsu method was employed quickly process Sentinel-1A Sentinel-2 generating high-quality training dataset. In first experiment, five rectangular areas approximately 37.5 km2 each were selected validate performance model across different scenes. second experiment focused Jining City, Shandong Province, China, analyzing monthly from 2020 2022 quarterly 2022. experimental results demonstrate excels achieving precision 0.989, recall 0.983, an F1 score 0.986, IoU 0.974, outperforming FCN, PSPNet, DeepLabV3+, ADCNN, MECNet.

Language: Английский

Citations

2