1 School of Smart City, Chongqing Jiaotong University, Chongqing 400074, China
2 School of Traffic and Transportation, Chongqing Jiaotong University, Chongqing 400074, China
3 School of Geography and Information Engineering, China University of Geosciences (Wuhan), Wuhan 430074, China
Show less
文章历史+
Received
Published
Revised
2025-08-23
2026-08-08
2025-12-16
Online Date
2026-09-03
PDF (16309K)
摘要
土壤湿度是农业、环境科学、水文学等领域中非常重要的参数之一,通常用体积含水量来表示,即土壤中水的体积与总体积之比。随着机器学习的发展,尤其是卷积神经网络在土壤湿度预测等领域的广泛应用,提出了深度-浅层耦合卷积神经网络(shallow deep coupled convolutional neural network,SD-CNN)。该模型通过对遥感影像或已有土壤湿度数据进行处理,可自主学习数据中的深浅层特征,并利用这些特征对未知区域的土壤湿度进行精确预测。相比传统的观测方法和传感器获取方式,SD-CNN 能够在较短时间内获取大范围、连续性更强的土壤湿度信息,并有效捕捉其空间变异性,从而显著提升预测的准确性和可靠性。将 SD-CNN 应用于山地复杂地形环境,并通过迭代次数和参数设置进行区域适配,可基于历史数据实现对未来一年土壤湿度的高精度预测,并通过实测数据验证其相关性与稳定性。实验结果表明该土壤湿度预测能力不仅能提高精准灌溉和农业生产效率,降低水资源消耗,还能增强干旱监测、洪涝预警和生态调控的科学性,从而为农业稳产增产、水资源优化配置以及生态安全保障提供关键技术支撑,对区域乃至国家经济的可持续发展具有重要推动作用。
Abstract
Soil moisture is one of the most important parameters in agriculture, environmental science, and hydrology. It is commonly expressed as volumetric water content, defined as the ratio of the volume of water in the soil to the total soil volume. With the rapid development of machine learning, particularly the widespread application of convolutional neural networks in soil moisture prediction, SD-CNN (shallow-deep coupled convolutional neural network) was proposed. By processing remote sensing imagery or historical soil moisture data, the model automatically learns both shallow and deep features embedded in the data and utilizes these features to accurately predict soil moisture in unobserved regions. Compared with traditional observation methods and sensor-based measurements, SD-CNN can obtain wide-coverage, continuous soil moisture information within a shorter time and effectively capture its spatial variability, thereby significantly improving prediction accuracy and reliability. By applying SD-CNN to complex mountainous terrain and performing regional adaptation through iterative optimization and parameter tuning, the model achieves high-precision predictions of soil moisture for the upcoming year and demonstrates strong correlation and stability when validated with field measurements. Experimental results show that this soil moisture prediction capability not only improves precise irrigation and agricultural production efficiency, but also strengthens drought monitoring, flood forecasting, and ecological regulation. These advancements provide essential technical support for stable agricultural output, optimized water resource allocation, and ecological security, ultimately contributing to sustainable regional and national economic development.
ZHOU Lian-jie, RUAN Ting-sheng, LANG Dan-ning, LI Jia, LIU Qian, WANG Yong, DENG Tian-min, ZHANG Xiang.
Soil Moisture Prediction Algorithm Based on Shallow Deep Coupled Convolutional Neural Network: Typical Regions in Abaga Banner and Banan District[J].
Science Technology and Engineering, 2026, 26(22): 9629-9638 DOI:10.12404/j.issn.1671-1815.2506139
激励层指在每个卷积层之后添加的一层,用于对卷积层输出的特征图进行非线性变换。采用线性整流函数(rectified linear unit, ReLU)或其变种作为激活函数,将卷积层的输出转化为具有非线性特征的特征图。激励层主要有两个方面的作用:第一,它能够增强网络的非线性表示能力,使得网络能够学习到更加复杂和抽象的特征;第二,它能够避免梯度消失的问题,使得网络的训练更加稳定和快速。
所构建的SD-CNN模型融合浅层时序与深度卷积结构,能有效捕捉土壤湿度的时空特征,具有较强的自适应性和稳定性。模型输入变量(NDVI、降雨量、日期等)均具有普适性,从而具备跨地区和多气候条件下的推广潜力。实验结果显示,10 cm 与 40 cm 土层预测精度差异较小,很好地体现出一定的泛化能力。通过在内蒙古阿巴嘎旗与重庆巴南区两类气候与生态条件显著差异区域的对比验证,模型在不同区域均保持稳定预测性能,进一步证明了其良好的跨区域普适性与稳健性。SD-CNN 同时捕获空间、地形和时序信息的能力,大幅提升土壤湿度预测精度与稳定性。其高分辨率预测可支持精准灌溉、水资源调度、生态修复与干旱预警,为农业、水利和生态管理提高可靠数据支撑,从而促进社会经济可持续发展。
WangChen, LiuSiyuan, ZhaoTingting, et al. Spatiotemporal varia-tion characteristics of soil moisture and its response to precipitation in mountainous ecological regions[J]. Science Technology and Engineering, 2025, 25(7): 3124-3132.
YangXuemei, WangRanghui. Analysis of soil moisture variation characteristics and driving forces in the Kokkayar Project Area based on geophysical Detectors[J]. Science Technology and Engineering, 2024, 24(26): 11148-11157.
[7]
WangY, ShiL, HuY, et al. A comprehensive study of deep learning for soil moisture prediction[J]. Hydrology and Earth System Sciences, 2024, 28(4): 917-943.
LuZhao, WangDaqing, ShiYue, et al. Research on soil moisture retrieval in desert-oasis areas based on multi-source remote sensing[J]. Surveying and Spatial Geographic Information, 2024, 47(12): 10-13.
[10]
HemmatiE, SahebiM R. Surface soil moisture retrieval based on transfer learning using SAR data on a local scale[J]. International Journal of Remote Sensing, 2024, 45(7): 2374-2406.
[11]
AbbesA B, JarrayN, FarahI R. Advances in remote sensing based soil moisture retrieval: applications, techniques, scales and challenges for combining machine learning and physical models[J]. Artificial Intelligence Review, 2024, 57(3): 477-505.
WangJingzhe, DinJianli, GeXiangyu, et al. Progress and prospect of soil salinization monitoring based on satellite-ground sensing technology[J]. Journal of Remote Sensing, 2024, 28(9): 2187-2208.
WangLei, ZhangSiyu, LiMing. Cloud and cloud-shadow detection method based on a lightweight convolutional neural network model[J]. Science Technology and Engineering, 2023, 23(32): 13681-13687.
WangXunhong, HaoTongzheng, MaCong. New energy vehicle sales prediction model based on PSO-BP neural network[J]. Scie-nce Technology and Engineering, 2024, 24(31): 13467-13474.
[18]
VahidiM, ShafianS, FrameW H. Multi-depth soil moisture estimation via 1D convolutional neural networks from drone-mounted ground penetrating Radar data[J]. Computers and Electronics in Agriculture, 2025, 232(1): 110104.
[19]
MisbahI, LeeC K, KeungK L. Fault diagnosis in rotating machines based on transfer learning: literature review[J]. Knowle-dge-Based Systems, 2024, 283(1): 111158.
ZhangLei, GuoAnhong, SongYingbo, et al. Soil moisture data fusion and its application in multi-dimensional assessment of drought and flood disasters[J]. Transactions of the Chinese Society of Agricultural Engineering, 2024, 40(13), 68-76.
LiQiaoling, LiXiaomei, QiuJuanjuan, et al. Research on CNN-based downscaling of multi-source soil moisture data in river basins[J]. Water Resources Protection, 2026, 42(2): 100-106.
LiShunyong, LiQinghui, XingYuman. Intelligent rock recognition based on lightweight network and transfer learning[J]. Scie-nce Technology and Engineering, 2025, 25(5): 1774-1782.
[30]
XuM, YangH, HuA, et al. A deep learning approach for SMAP soil moisture downscaling informed by thermal inertia theory[J]. International Journal of Applied Earth Observation and Geoinformation, 2025, 136(1): 104370.
[31]
XuH, SunH, XuZ, et al. kNDMI: a kernel normalized differe-nce moisture index for remote sensing of soil and vegetation moisture[J]. Remote Sensing of Environment, 2025, 319(1): 114621.
JinYan, XuLing, ZhouQun, et al. Monitoring of agricultural drought in Yunnan based on CLDAS soil relative moisture[J]. Agricultural Meteorology in China, 2024, 45(1): 79-90.
ShaoYu. The impact of intelligent farmland irrigation systems on water resource utilization efficiency[J]. Agricultural Engineering Technology, 2024, 44(26): 50-51.
[36]
LiJ, ChengY, LuY, et al. From ReLU to GeMU: activation functions in the lens of cone projection[J]. Neural Networks, 2025 (1): 107654.
[37]
YangY, LiD, LüY, et al. Multispectral and hyperspectral images fusion based on subspace representation and nonlocal low-rank regularization[J]. International Journal of Remote Sensing, 2024, 45(9): 2965-2984.
[38]
FarhangmehrV, ImanianH, MohammadianA, et al. A spatiotemporal CNN-LSTM deep learning model for predicting soil temperature in diverse large-scale regional climates[J]. Science of the Total Environment, 2025, 968: 178901.
WuShangrong, ZhaoRongkun, CaoHong, et al. Research progress on SAR crop and soil parameter inversion based on microwave scattering mechanism[J]. Journal of Radar, 2025, 14(3): 1-19.