Development of Continuous AMSR-E/2 Soil Moisture Time Series by Hybrid Deep Learning Model (ConvLSTM2D and Conv2D) and Transfer Learning for Reanalyses

Surface soil moisture (SSM) is a crucial climate variable of the Earth system that regulates water and energy exchanges between the land and atmosphere, directly influencing hydrological, biogeochemical, and energy cycles. However, satellite-derived SSM, particularly from the Advanced Microwave Scan...

Full description

Saved in:
Bibliographic Details
Main Authors: Visakh Sivaprasad, Mehdi Rahmati, Anne Springer, Harry Vereecken, Carsten Montzka
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10949744/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849314136539791360
author Visakh Sivaprasad
Mehdi Rahmati
Anne Springer
Harry Vereecken
Carsten Montzka
author_facet Visakh Sivaprasad
Mehdi Rahmati
Anne Springer
Harry Vereecken
Carsten Montzka
author_sort Visakh Sivaprasad
collection DOAJ
description Surface soil moisture (SSM) is a crucial climate variable of the Earth system that regulates water and energy exchanges between the land and atmosphere, directly influencing hydrological, biogeochemical, and energy cycles. However, satellite-derived SSM, particularly from the Advanced Microwave Scanning Radiometer AMSR-E/2, is limited by radio frequency interference, vegetation effects, frozen ground, and significant spatial and temporal data gaps. By excluding data points affected by these problems, we are able to train an unaffected system and fill the gaps with high accuracy predictions. We developed a sophisticated deep learning ConvLSTM model, that combines convolutional long short-term memory (ConvLSTM2D) layers and convolutional neural network (CNN) layers. The model initially enhances AMSR-2 SSM values across time and space using Advanced SCATterometer (ASCAT) SSM as input. The ConvLSTM model, trained to enhance AMSR-2 SSM, is then fine-tuned by using the transfer learning technique to enhance AMSR-E data. The enhanced AMSR-2 data is used as a target to guide the enhancement of AMSR-E. This approach ensures that gaps in AMSR-E data are filled, while aligning the characteristics with the more consistent AMSR-2 SSM, resulting in a seamless AMSR-E/2 dataset from 2003 to 2023. Unlike previous studies incorporating additional datasets like precipitation, temperature, and digital elevation models, our approach avoids these to prevent redundancy and potential inaccuracies when generating land surface reanalyzes based on data assimilation, since such data are already integrated into the land surface model. The ConvLSTM model achieved a lower root mean squared error of 0.07 for AMSR-2 prediction and 0.04 for AMSR-E via transfer learning demonstrating significant gap-filling accuracy. The enhanced SSM demonstrated a 26% improvement in the correlation with in situ SSM measurements, while maintaining accuracy and consistency in spatial and temporal patterns.
format Article
id doaj-art-bb7a48c90f5740f9acb4da1341f5eda9
institution Kabale University
issn 1939-1404
2151-1535
language English
publishDate 2025-01-01
publisher IEEE
record_format Article
series IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
spelling doaj-art-bb7a48c90f5740f9acb4da1341f5eda92025-08-20T03:52:32ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing1939-14042151-15352025-01-0118111691118210.1109/JSTARS.2025.355795610949744Development of Continuous AMSR-E/2 Soil Moisture Time Series by Hybrid Deep Learning Model (ConvLSTM2D and Conv2D) and Transfer Learning for ReanalysesVisakh Sivaprasad0https://orcid.org/0000-0002-9861-083XMehdi Rahmati1https://orcid.org/0000-0001-5547-6442Anne Springer2Harry Vereecken3https://orcid.org/0000-0002-8051-8517Carsten Montzka4https://orcid.org/0000-0003-0812-8570Institute of Bio- and Geosciences, Agrosphere, Forschungszentrum Jülich, Jülich, GermanyInstitute of Bio- and Geosciences, Agrosphere, Forschungszentrum Jülich, Jülich, GermanyInstitute of Geodesy and Geoinformation, Agricultural faculty, University of Bonn, Bonn, GermanyInstitute of Bio- and Geosciences, Agrosphere, Forschungszentrum Jülich, Jülich, GermanyInstitute of Bio- and Geosciences, Agrosphere, Forschungszentrum Jülich, Jülich, GermanySurface soil moisture (SSM) is a crucial climate variable of the Earth system that regulates water and energy exchanges between the land and atmosphere, directly influencing hydrological, biogeochemical, and energy cycles. However, satellite-derived SSM, particularly from the Advanced Microwave Scanning Radiometer AMSR-E/2, is limited by radio frequency interference, vegetation effects, frozen ground, and significant spatial and temporal data gaps. By excluding data points affected by these problems, we are able to train an unaffected system and fill the gaps with high accuracy predictions. We developed a sophisticated deep learning ConvLSTM model, that combines convolutional long short-term memory (ConvLSTM2D) layers and convolutional neural network (CNN) layers. The model initially enhances AMSR-2 SSM values across time and space using Advanced SCATterometer (ASCAT) SSM as input. The ConvLSTM model, trained to enhance AMSR-2 SSM, is then fine-tuned by using the transfer learning technique to enhance AMSR-E data. The enhanced AMSR-2 data is used as a target to guide the enhancement of AMSR-E. This approach ensures that gaps in AMSR-E data are filled, while aligning the characteristics with the more consistent AMSR-2 SSM, resulting in a seamless AMSR-E/2 dataset from 2003 to 2023. Unlike previous studies incorporating additional datasets like precipitation, temperature, and digital elevation models, our approach avoids these to prevent redundancy and potential inaccuracies when generating land surface reanalyzes based on data assimilation, since such data are already integrated into the land surface model. The ConvLSTM model achieved a lower root mean squared error of 0.07 for AMSR-2 prediction and 0.04 for AMSR-E via transfer learning demonstrating significant gap-filling accuracy. The enhanced SSM demonstrated a 26% improvement in the correlation with in situ SSM measurements, while maintaining accuracy and consistency in spatial and temporal patterns.https://ieeexplore.ieee.org/document/10949744/AMSR-E/2Advanced SCATterometerCNNConvLSTMdeep learningremote sensing
spellingShingle Visakh Sivaprasad
Mehdi Rahmati
Anne Springer
Harry Vereecken
Carsten Montzka
Development of Continuous AMSR-E/2 Soil Moisture Time Series by Hybrid Deep Learning Model (ConvLSTM2D and Conv2D) and Transfer Learning for Reanalyses
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
AMSR-E/2
Advanced SCATterometer
CNN
ConvLSTM
deep learning
remote sensing
title Development of Continuous AMSR-E/2 Soil Moisture Time Series by Hybrid Deep Learning Model (ConvLSTM2D and Conv2D) and Transfer Learning for Reanalyses
title_full Development of Continuous AMSR-E/2 Soil Moisture Time Series by Hybrid Deep Learning Model (ConvLSTM2D and Conv2D) and Transfer Learning for Reanalyses
title_fullStr Development of Continuous AMSR-E/2 Soil Moisture Time Series by Hybrid Deep Learning Model (ConvLSTM2D and Conv2D) and Transfer Learning for Reanalyses
title_full_unstemmed Development of Continuous AMSR-E/2 Soil Moisture Time Series by Hybrid Deep Learning Model (ConvLSTM2D and Conv2D) and Transfer Learning for Reanalyses
title_short Development of Continuous AMSR-E/2 Soil Moisture Time Series by Hybrid Deep Learning Model (ConvLSTM2D and Conv2D) and Transfer Learning for Reanalyses
title_sort development of continuous amsr e 2 soil moisture time series by hybrid deep learning model convlstm2d and conv2d and transfer learning for reanalyses
topic AMSR-E/2
Advanced SCATterometer
CNN
ConvLSTM
deep learning
remote sensing
url https://ieeexplore.ieee.org/document/10949744/
work_keys_str_mv AT visakhsivaprasad developmentofcontinuousamsre2soilmoisturetimeseriesbyhybriddeeplearningmodelconvlstm2dandconv2dandtransferlearningforreanalyses
AT mehdirahmati developmentofcontinuousamsre2soilmoisturetimeseriesbyhybriddeeplearningmodelconvlstm2dandconv2dandtransferlearningforreanalyses
AT annespringer developmentofcontinuousamsre2soilmoisturetimeseriesbyhybriddeeplearningmodelconvlstm2dandconv2dandtransferlearningforreanalyses
AT harryvereecken developmentofcontinuousamsre2soilmoisturetimeseriesbyhybriddeeplearningmodelconvlstm2dandconv2dandtransferlearningforreanalyses
AT carstenmontzka developmentofcontinuousamsre2soilmoisturetimeseriesbyhybriddeeplearningmodelconvlstm2dandconv2dandtransferlearningforreanalyses