Dataset for A Sentinel-3-Based Structured Extraction and Deep Learning Framework for Daily Large-Scale Water-body Forecasting

Source
DataCite
Published
2026
Country
Kazakhstan
License
Creative Commons Attribution 4.0 International
Themes
Remote Sensing & Satellites, Machine Learning & AI, Fresh Water

Get the data from DataCite

About this dataset

Timely monitoring of surface-water dynamics is challenging because remote sensing methods often trade temporal frequency for spatial coverage. To address this issue, this study proposes an integrated framework for daily large-scale surface-water monitoring and basin-level next-day forecasting using Sentinel-3 sensor and multi-factor deep learning model. A daily water-area extraction workflow is developed by combining atmospheric correction, pixel-level quality control, structured point-set representation, and spectral-spatial water identification, in which water bodies are extracted as structured point sets to support rapid large-area mapping across Kazakhstan from 2020 to 2024. Based on the derived daily water series, a basin-level forecasting model is built using meteorological variables, temporal memory descriptors, seasonal indicators, and historical constraints. By constructing...

How the data was collected

Timely monitoring of surface-water dynamics is challenging because remote sensing methods often trade temporal frequency for spatial coverage. To address this issue, this study proposes an integrated framework for daily large-scale surface-water monitoring and basin-level next-day forecasting using Sentinel-3 sensor and multi-factor deep learning model. Based on the derived daily water series, a basin-level forecasting model is built using meteorological variables, temporal memory descriptors, seasonal indicators, and historical constraints.

Browse sustainability datasets on SusanHub