Soil salinization is one of the most important forms of land degradation in arid and semi-arid regions, placing pressure on regional food security and ecological security. The rapid and accurate identification of the spatial distribution and dynamics of salinized soils is essential to the effectiveness of cultivated land protection and ecological restoration. A single sensor has inherent limitations in spectral separability and spatial coverage. Meanwhile, the spectral response of salinization exhibits strong regional heterogeneity. These two factors jointly restrict the application of conventional remote sensing identification methods across large regional scales. This paper therefore proposes an intelligent identification framework for regional soil salinization that integrates multi-source remote sensing and transfer learning systematically examines its scale effects. Sentinel-1 synthetic aperture radar and Sentinel-2 multispectral imagery serve as the core data sources, and vegetation indices, salinity indices and topographic factors are jointly incorporated into a multi-dimensional feature space to enrich the discriminative information of the features. Sample scarcity exists in the target region. To tackle this problem, this paper adopts a transfer learning strategy based on convolutional neural networks. The pre-trained model from sample-rich regions is transferred to the target region with limited samples, which improves the model’s generalization ability. Random forest, support vector machine and deep learning methods are compared within the same framework, and the variation of identification accuracy is analyzed at the pixel, field and regional scales. Multi-source remote sensing feature fusion markedly improves the overall accuracy of salinization identification, transfer learning relieves the pressure of insufficient samples in the target region, and the model shows good spatial transferability. Identification accuracy first rises and then stabilizes as the spatial scale increases, and the optimal identification scale is jointly determined by the degree of landscape fragmentation and image resolution. This study can provide methodological support and a scientific basis for the remote sensing monitoring, farmland management and ecological governance of regional soil salinization.
References
[1] Ivushkin, K., Bartholomeus, H., Bregt, A. K., Pulatov, A., Kempen, B., De Sousa, L. (2019) Global mapping of soil salinity change. Remote Sensing of Environment, 231, 111260.
[2] Tola, D., Satgé, F., Pillco Zolá, R., Sainz, H., Condori, B., Miranda, R., Espinoza-Villar, R. (2024) Soil salinity mapping of plowed agriculture lands combining radar sentinel-1 and optical sentinel-2 with topographic data in machine learning models. Remote Sensing, 16(18), 3456.
[3] He, Y., Zhang, Z., Xiang, R., Ding, B., Du, R., Yin, H., Ba, Y. (2023) Monitoring salinity in bare soil based on sentinel-1/2 image fusion and machine learning. Infrared Physics & Technology, 131, 104656.
[4] Erkin, N., Zhu, L., Gu, H., Tusiyiti, A. (2019) Method for predicting soil salinity concentrations in croplands based on machine learning and remote sensing techniques. Journal of Applied Remote Sensing, 13(3), 034520.
[5] Sulieman, M. M., Kaya, F., Elsheikh, M. A., Başayiğit, L., Francaviglia, R. (2023) Application of machine learning algorithms for digital mapping of soil salinity levels and assessing their spatial transferability in arid regions. Land, 12(9), 1680.
[6] Garajeh, M. K., Malakyar, F., Weng, Q., Feizizadeh, B., Blaschke, T., Lakes, T. (2021) An automated deep learning convolutional neural network algorithm applied for soil salinity distribution mapping in Lake Urmia, Iran. Science of the Total Environment, 778, 146253.
[7] Padarian, J., Minasny, B., McBratney, A. B. (2019) Transfer learning to localise a continental soil vis-NIR calibration model. Geoderma, 340, 279-288.
[8] Li, Y., Chang, C., Wang, Z., Zhao, G. (2023) Upscaling remote sensing inversion and dynamic monitoring of soil salinization in the Yellow River Delta, China. Ecological Indicators, 148, 110087.
[9] Yu, X., Yang, J., Graf, T., Koneshloo, M., O’Neal, M. A., Michael, H. A. (2016) Impact of topography on groundwater salinization due to ocean surge inundation. Water Resources Research, 52(8), 5794-5812.
[10] Boateng, E. Y., Otoo, J., Abaye, D. A. (2020) Basic tenets of classification algorithms k-nearest-neighbor, support vector machine, random forest and neural network: a review. Journal of Data Analysis and Information Processing, 8(4), 341-357.
Share and Cite
Sun, Q. (2026) Intelligent Identification and Scale Effect of Regional Soil Salinization Based on Multi-source Remote Sensing and Transfer Learning. Scientific Research Bulletin, 3(4), 41-45. https://doi.org/10.71052/srb2024/GSBZ472