滨海湿地土壤质地高光谱估测模型对比分析

doi:10.3964/j.issn.1000-0593(2024)09-2568-09

Abstract
References
Related Citation (15)

Download: PDF (15521 KB)
Export: BibTeX | EndNote (RIS)

Abstract Soil texture affects many physical, chemical, biological, and hydrological characteristics and processes, such as vegetation distribution, soil and water conservation capacity, and microbial activity. Accurate acquisition of soil texture is of great significance for wetland ecological restoration and protection. Based on 57 measured surface soil texture and visible-near-infrared hyperspectral data in Tianjin coastal wetland, the soil samples were smoothed by S-G and transformed by first derivative (FD), reciprocal transformation (RT), reciprocal first derivative (RTFD), square root (SR), square root first derivative (SRFD), logarithm of reciprocal (LR) and logarithm of reciprocal first derivative (LRFD),the characteristics and correlations of spectral curves of different soil texture categories were analyzed. A competitive adaptive reweighting algorithm (CARS) was used to select the characteristic bands, and partial least square regression (PLSR), random forest regression (RFR), and support vector machineregression (SVR) algorithms were combined to compare the modeling effects of different spectral transformations. The results show that: (1) The texture categories of wetland soil are mainly silty loam and silt. The spectral reflectance of silt is the highest in the 400~2 400 nm band, and the spectral reflectance of sandy soil is the lowest in the 400~2 000 nm band. The correlation between the spectral reflectance of FD, RTFD, and SRFD and the soil particle size content has significantly increased. The absolute value of the maximum correlation coefficient is above 0.58, and the highest is 0.70. (2) The feature band number of eight spectral transforms screened by the CARS algorithm is 1.05%~6.15% of the total band number, effectively reducing the information redundancy of spectral data. (3) Compared with the three estimation models for particle size content, the SVR model of SRFD and RTFD spectral transformation had the best accuracy and was superior to the other two models, the clay (SRFD) test set (R²=0.72, RMSE=1.86%, nRMSE=11.33%), the silt (SRFD) test set (R²=0.72, RMSE=2.82%, nRMSE=7.30%) and the sand (RTFD) test set (R²=0.71, RMSE=5.75%, nRMSE=5.91%). The results of this study can provide a basis and technical support for the accurate monitoring of soil texture in coastal wetland areas with hyperspectral data.

Key words：Coastal wetland; Soil texture; Spectral transformation; Competitive adaptive reweighted sampling; Machine learning

Received: 2023-05-21 Accepted: 2023-11-23

ZTFLH:

S155

Corresponding Authors: LIU Ming-yue E-mail: liumy917@ncst.edu.cn

Cite this article:

URL:

https://www.gpxygpfx.com/EN/10.3964/j.issn.1000-0593(2024)09-2568-09 OR https://www.gpxygpfx.com/EN/Y2024/V44/I09/2568

[1] Hossain M S, Rahman G K M M, Alam M S, et al. Soil Research, 2018, 56(4): 421.
[2] Leone D, Mirabile M, Altieri G M, et al. Ecological Indicators, 2023, 153: 110428.
[3] Swetha R K, Chakraborty S. Geoderma, 2021, 382: 114775.
[4] Hobley E U, Prater I. European Journal of Soil Science, 2019, 70(1): 83.
[5] QIAO Tian, LÜ Cheng-wen, XIAO Wen-ping, et al(乔天, 吕成文, 肖文凭, 等). Chinese Journal of Soil Science(土壤通报), 2018, 49(4): 773.
[6] Gou Y, Wei J, Li J, et al. Journal of Mountain Science, 2020, 17(9): 2214.
[7] Ba Y, Liu J, Han J, et al. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 2020, 229: 117863.
[8] DIWU Peng-yao, BIAN Xi-hui, WANG Zi-fang, et al(第五鹏瑶，卞希慧，王姿方，等). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2019, 39(9): 2800.
[9] Bahrami A, Danesh M, Bahrami M. Infrared Physics & Technology, 2022, 122: 104056.
[10] Zhang X, Xue J, Xiao Y, et al. Remote Sensing, 2023, 15(2): 465.
[11] ZHENG Yu-zhen, CHEN Yi-yun, CHEN Min, et al(郑嵛珍, 陈奕云, 陈敏, 等). Hubei Agricultural Sciences(湖北农业科学), 2021, 60(10): 38.
[12] Ge Y, Morgan C L S, Wijewardane N K. Soil Science Society of America Journal, 2020, 84(5): 1495.
[13] Lazaar A, Pradhan B, Naiji Z, et al. Kuwait Journal of Science, 2021, 48(1): 127.
[14] Nawar S, Buddenbaum H, Hill J, et al. Soil and Tillage Research, 2016, 155: 510.
[15] SHI Wen-jiao, ZHANG Mo(史文娇, 张沫). Acta Geographica Sinica(地理学报), 2022, 77(11): 2890.
[16] Shahrayini E, Noroozi A A, Eghbal M K. Eurasian Soil Science, 2020, 53: 1760.
[17] YIN De-jie, JING Rui, GUAN Hai-yan, et al(尹德洁, 荆瑞, 关海燕, 等). Journal of Beijing Forestry University(北京林业大学学报), 2018, 40(8): 103.
[18] JIANG Jian-ping, ZHU Han-hua, WU Li-jian(蒋建平, 朱汉华, 吴立键). Yangtze River(人民长江), 2020, 51(7): 89.
[19] Veloso M F, Rodrigues L N, Fernandes Filho E I. Geoderma Regional, 2022, 30: e00569.
[20] Ludwig B, Murugan R, Parama V R R, et al. Journal of Plant Nutrition and Soil Science, 2018, 181(5): 704.
[21] Sakhaee A, Gebauer A, Ließ M, et al. Soil, 2022, 8(2): 587.