Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Enhanced multi-mission remote sensing of inland water surface elevation using Sentinel-3, Sentinel-6, and SWOT satellite altimeters and an environmentally informed LSTM-based neural network

Rezapour, Mahdis ; Taheri Dehkordi, Alireza LU ; Valadan Zoej, Mohammad Javad ; Khesali, Elahe ; Naghibi, Amir LU and Hashemi, Hossein LU orcid (2026) In Remote Sensing Applications: Society and Environment 42.
Abstract

Frequent and accurate measurements of inland Water Surface Elevation (WSE) are essential for effective water resource management. However, single-mission satellite altimetry often lacks the temporal resolution needed to capture detailed WSE changes. While multi-mission integration can improve temporal coverage, it is hindered by inter- and intra-mission biases arising from variations in sensor design, orbital characteristics, atmospheric effects, and environmental conditions. These biases, which have been insufficiently addressed in previous studies, are typically nonlinear, spatiotemporally variable, and require advanced methods for correction. This study proposes EILSTMNet, an Environmentally Informed Long Short-Term Memory... (More)

Frequent and accurate measurements of inland Water Surface Elevation (WSE) are essential for effective water resource management. However, single-mission satellite altimetry often lacks the temporal resolution needed to capture detailed WSE changes. While multi-mission integration can improve temporal coverage, it is hindered by inter- and intra-mission biases arising from variations in sensor design, orbital characteristics, atmospheric effects, and environmental conditions. These biases, which have been insufficiently addressed in previous studies, are typically nonlinear, spatiotemporally variable, and require advanced methods for correction. This study proposes EILSTMNet, an Environmentally Informed Long Short-Term Memory (LSTM)-based Neural Network that enables multi-mission synergy of satellite altimetry data (Sentinel-3, Sentinel-6, and SWOT) by correcting altimetric measurements of WSE through the integration of environmental variables such as precipitation, temperature, and evapotranspiration. EILSTMNet employs stacked LSTM layers to capture temporal dependencies in environmental drivers, combined with a fully connected neural network that incorporates static inputs such as altimetric WSE, day of year, and satellite-specific identifiers. The proposed approach is validated over three U.S. lakes, Michigan, Ontario, and Winnebago, using in-situ gauge measurements. Results show that EILSTMNet-based estimates are significantly improved compared to altimeter-derived WSE measurements, reducing the Root Mean Squared Error from 0.31 m to 0.09 m and increasing the Pearson correlation coefficient from 0.69 to 0.93. Furthermore, the model demonstrates strong generalization to unseen time periods, highlighting its temporal transferability. The proposed approach refines multi-mission altimetric measurements, yielding temporally frequent, higher-accuracy WSE observations, thereby enhancing water resource management and advancing the understanding of hydrological processes.

(Less)
Please use this url to cite or link to this publication:
author
; ; ; ; and
organization
publishing date
type
Contribution to journal
publication status
published
subject
keywords
Bias correction, Deep learning, Machine learning, Multi-source data, Satellite altimeters
in
Remote Sensing Applications: Society and Environment
volume
42
article number
102019
publisher
Elsevier
external identifiers
  • scopus:105035397960
ISSN
2352-9385
DOI
10.1016/j.rsase.2026.102019
project
The United Nations University Hub: Water in a Changing Environment (WICE)
language
English
LU publication?
yes
additional info
Publisher Copyright: © 2026 The Authors.
id
03b7c1f6-4093-4aba-8d78-bb83ea306c56
date added to LUP
2026-05-30 19:14:18
date last changed
2026-06-02 13:28:16
@article{03b7c1f6-4093-4aba-8d78-bb83ea306c56,
  abstract     = {{<p>Frequent and accurate measurements of inland Water Surface Elevation (WSE) are essential for effective water resource management. However, single-mission satellite altimetry often lacks the temporal resolution needed to capture detailed WSE changes. While multi-mission integration can improve temporal coverage, it is hindered by inter- and intra-mission biases arising from variations in sensor design, orbital characteristics, atmospheric effects, and environmental conditions. These biases, which have been insufficiently addressed in previous studies, are typically nonlinear, spatiotemporally variable, and require advanced methods for correction. This study proposes EILSTMNet, an Environmentally Informed Long Short-Term Memory (LSTM)-based Neural Network that enables multi-mission synergy of satellite altimetry data (Sentinel-3, Sentinel-6, and SWOT) by correcting altimetric measurements of WSE through the integration of environmental variables such as precipitation, temperature, and evapotranspiration. EILSTMNet employs stacked LSTM layers to capture temporal dependencies in environmental drivers, combined with a fully connected neural network that incorporates static inputs such as altimetric WSE, day of year, and satellite-specific identifiers. The proposed approach is validated over three U.S. lakes, Michigan, Ontario, and Winnebago, using in-situ gauge measurements. Results show that EILSTMNet-based estimates are significantly improved compared to altimeter-derived WSE measurements, reducing the Root Mean Squared Error from 0.31 m to 0.09 m and increasing the Pearson correlation coefficient from 0.69 to 0.93. Furthermore, the model demonstrates strong generalization to unseen time periods, highlighting its temporal transferability. The proposed approach refines multi-mission altimetric measurements, yielding temporally frequent, higher-accuracy WSE observations, thereby enhancing water resource management and advancing the understanding of hydrological processes.</p>}},
  author       = {{Rezapour, Mahdis and Taheri Dehkordi, Alireza and Valadan Zoej, Mohammad Javad and Khesali, Elahe and Naghibi, Amir and Hashemi, Hossein}},
  issn         = {{2352-9385}},
  keywords     = {{Bias correction; Deep learning; Machine learning; Multi-source data; Satellite altimeters}},
  language     = {{eng}},
  publisher    = {{Elsevier}},
  series       = {{Remote Sensing Applications: Society and Environment}},
  title        = {{Enhanced multi-mission remote sensing of inland water surface elevation using Sentinel-3, Sentinel-6, and SWOT satellite altimeters and an environmentally informed LSTM-based neural network}},
  url          = {{http://dx.doi.org/10.1016/j.rsase.2026.102019}},
  doi          = {{10.1016/j.rsase.2026.102019}},
  volume       = {{42}},
  year         = {{2026}},
}