Soil moisture forecasting in wireless sensor networks via spatiotemporal graph convolutional networks

Abstract Wireless sensor networks enable long‐term, automated, networked monitoring of soil moisture, an indispensable tool in soil moisture sensing research and application. The growing abundance of soil moisture data has increased interest in using historical data to forecast future soil moisture...

Full description

Saved in:
Bibliographic Details
Main Authors: Weixuan Wang, Yating Wei, Longfei Hao, Zushuai Wei, Tianjie Zhao
Format: Article
Language:English
Published: Wiley 2025-01-01
Series:Vadose Zone Journal
Online Access:https://doi.org/10.1002/vzj2.70000
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849721856870842368
author Weixuan Wang
Yating Wei
Longfei Hao
Zushuai Wei
Tianjie Zhao
author_facet Weixuan Wang
Yating Wei
Longfei Hao
Zushuai Wei
Tianjie Zhao
author_sort Weixuan Wang
collection DOAJ
description Abstract Wireless sensor networks enable long‐term, automated, networked monitoring of soil moisture, an indispensable tool in soil moisture sensing research and application. The growing abundance of soil moisture data has increased interest in using historical data to forecast future soil moisture variations effectively. However, due to the combined effects of internal factors like soil types, terrain, and vegetation cover, as well as external factors such as precipitation and temperature, soil moisture data in wireless sensor networks exhibit complex spatial and temporal interdependencies. Consequently, developing predictive models that can accurately capture these dependencies and enable precise forecasts of soil moisture within these networks poses a significant challenge. To address this problem, a graph was used to express the network topology among sensors. Graph neural network is adept at capturing and utilizing the complex spatiotemporal relationships and patterns inherent in graph‐structured data, thereby enhancing prediction accuracy. This study incorporates soil temperature and precipitation as external factors into the model predictions. The performance of the prediction model was evaluated by comparing its results with those of two baseline methods across different time intervals. The results indicated that the model is suitable for long‐term forecasting of soil moisture. The model's accuracy showed improvements of 49.53%, 34.86%, and 29.73% at the 48‐h, 144‐h, and 192‐h time steps, respectively. The correlation coefficient between predicted and actual values was 0.94. The model contributes to the scientific management of water resources, facilitates the rational regulation of irrigation water volume and timing, and enhances agricultural yield.
format Article
id doaj-art-0209fd352bc74ec1bc7c2d44e60b4f93
institution DOAJ
issn 1539-1663
language English
publishDate 2025-01-01
publisher Wiley
record_format Article
series Vadose Zone Journal
spelling doaj-art-0209fd352bc74ec1bc7c2d44e60b4f932025-08-20T03:11:31ZengWileyVadose Zone Journal1539-16632025-01-01241n/an/a10.1002/vzj2.70000Soil moisture forecasting in wireless sensor networks via spatiotemporal graph convolutional networksWeixuan Wang0Yating Wei1Longfei Hao2Zushuai Wei3Tianjie Zhao4School of Remote Sensing and Information EngineeringWuhan UniversityWuhanChinaSchool of Remote Sensing and Information EngineeringWuhan UniversityWuhanChinaSchool of Surveying and Land Information EngineeringHenan Polytechnic UniversityJiaozuoChinaSchool of Artificial IntelligenceJianghan UniversityWuhanChinaState Key Laboratory of Remote Sensing Science, Aerospace Information Research InstituteChinese Academy of SciencesBeijingChinaAbstract Wireless sensor networks enable long‐term, automated, networked monitoring of soil moisture, an indispensable tool in soil moisture sensing research and application. The growing abundance of soil moisture data has increased interest in using historical data to forecast future soil moisture variations effectively. However, due to the combined effects of internal factors like soil types, terrain, and vegetation cover, as well as external factors such as precipitation and temperature, soil moisture data in wireless sensor networks exhibit complex spatial and temporal interdependencies. Consequently, developing predictive models that can accurately capture these dependencies and enable precise forecasts of soil moisture within these networks poses a significant challenge. To address this problem, a graph was used to express the network topology among sensors. Graph neural network is adept at capturing and utilizing the complex spatiotemporal relationships and patterns inherent in graph‐structured data, thereby enhancing prediction accuracy. This study incorporates soil temperature and precipitation as external factors into the model predictions. The performance of the prediction model was evaluated by comparing its results with those of two baseline methods across different time intervals. The results indicated that the model is suitable for long‐term forecasting of soil moisture. The model's accuracy showed improvements of 49.53%, 34.86%, and 29.73% at the 48‐h, 144‐h, and 192‐h time steps, respectively. The correlation coefficient between predicted and actual values was 0.94. The model contributes to the scientific management of water resources, facilitates the rational regulation of irrigation water volume and timing, and enhances agricultural yield.https://doi.org/10.1002/vzj2.70000
spellingShingle Weixuan Wang
Yating Wei
Longfei Hao
Zushuai Wei
Tianjie Zhao
Soil moisture forecasting in wireless sensor networks via spatiotemporal graph convolutional networks
Vadose Zone Journal
title Soil moisture forecasting in wireless sensor networks via spatiotemporal graph convolutional networks
title_full Soil moisture forecasting in wireless sensor networks via spatiotemporal graph convolutional networks
title_fullStr Soil moisture forecasting in wireless sensor networks via spatiotemporal graph convolutional networks
title_full_unstemmed Soil moisture forecasting in wireless sensor networks via spatiotemporal graph convolutional networks
title_short Soil moisture forecasting in wireless sensor networks via spatiotemporal graph convolutional networks
title_sort soil moisture forecasting in wireless sensor networks via spatiotemporal graph convolutional networks
url https://doi.org/10.1002/vzj2.70000
work_keys_str_mv AT weixuanwang soilmoistureforecastinginwirelesssensornetworksviaspatiotemporalgraphconvolutionalnetworks
AT yatingwei soilmoistureforecastinginwirelesssensornetworksviaspatiotemporalgraphconvolutionalnetworks
AT longfeihao soilmoistureforecastinginwirelesssensornetworksviaspatiotemporalgraphconvolutionalnetworks
AT zushuaiwei soilmoistureforecastinginwirelesssensornetworksviaspatiotemporalgraphconvolutionalnetworks
AT tianjiezhao soilmoistureforecastinginwirelesssensornetworksviaspatiotemporalgraphconvolutionalnetworks