Research on Hyperspectral Inversion of Soil Salinity Based on Fractional-Order Differentiation and Multidimensional Spectral Indices: A Case Study of Pingluo County, Ningxia
Soil salinization is a key issue of global soil degradation. To improve the inversion accuracy of soil salinity, this study takes Pingluo County in Ningxia as the study area. Based on the spectral data collected from spectrometers, the spectral characteristics of soils with different salinization degrees are systematically analyzed. The fractional-order (0-2.00 order) differentiation method is employed to enhance the effective information of the spectra, combined with two-dimensional and three-dimensional spectral indices and the Boruta feature selection algorithm to construct three types of soil salinity inversion models: Random Forest(RF), Extremely Randomized Trees (ERT) and Extreme Gradient Boosting (XGBoost). The results show that typical absorption valleys appear in the soil spectra around 1,400 nm, 1,900 nm and 2,200 nm. The fractional-order differentiation method effectively enhances the spectral absorption features, with the 0.75-order differentiation transformation showing the strongest correlation between the two-dimensional and three-dimensional spectral indices and soil salinity, with correlation coefficients (r) of -0.786 8 and 0.849, respectively. The Boruta algorithm identifies 27 key features. Comparative analysis of the models shows that among the four modeling strategies, the RF model is the most stable and accurate, achieving optimal fitting results (R²=0.92, erms=0.84) when SG, FOD, Boruta, and Spectral Index are used as input features. Feature importance analysis indicates that feature TBI2 is the primary driving factor for model predictions (importance=0.14), and SHAP explanations further confirm that TBI2 contributes positively to soil salinity predictions (mean SHAP value=0.56), while feature RI contributes negatively. Partial Dependence Plots (PDP) and Individual Conditional Expectation(ICE) curves reveal that when TBI2 values are around 2, the model’s predictions exhibit a stepwise positive response, while R1590 shows a smooth negative response. The analyses from various interpretative methods are consistent. This research provides important references for hyperspectral inversion of salinity in salinized soils.
LalR. Soil carbon sequestration impacts on global climate change and food security[J]. Science,2004, 304:1623-1627.
[2]
ChiYuan, FanMinglei, ZhangZhiwei, et al. Zoning the soil salinization levels in the northern China's coastal areas based on high-resolution soil mapping [J/OL]. Ecological Indicators,2025, 172:113303.
[3]
DengMeihua, ZhangChao, TangMin, et al. Stacking ensemble learning coupled with multi-source remote sensing data: Enhancing soil salinity inversion accuracy in barley-cultivated salinized soils[J/OL]. Agricultural Water Management,2025, 322:109959.
[4]
IvushkinK, BartholomeusH, BregtA K, et al. Global mapping of soil salinity change[J/OL]. Remote Sensing of Environment,2019, 231:111260.
[5]
ShokriN, HassaniA, SahimiM. Multi-scale soil salinization dynamics from global to pore scale: A review[J/OL]. Reviews of Geophysics, 2024,62(4).
[6]
MetternichtG I, ZinckJ A. Remote sensing of soil salinity: Potentials and constraints[J]. Remote Sensing of Environment,2003, 85(1):1-20.
[7]
YuBo, YanChangxiang, YuanJing, et al. Prediction of soil properties based on characteristic wavelengths with optimal spectral resolution by using Vis-NIR spectroscopy[J/OL]. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy,2023,293:122452.
[8]
DuRuiqi, ChenJunying, XiangYouzhen, et al. Timely monitoring of soil water-salt dynamics within cropland by hybrid spectral unmixing and machine learning models[J]. International Soil and Water Conservation Research,2024, 12(3):726-740.
[9]
BaoYilin, UstinS, MengXiangtian, et al. A regional-scale hyperspectral prediction model of soil organic carbon considering geomorphic features[J/OL]. Geoderma,2021, 403:115263.
[10]
JiaPingping, HeWei, HuYi, et al. Inversion of coastal cultivated soil salt content based on multi-source spectra and environmental variables[J/OL]. Soil Tillage Research,2024, 241:106124.
[11]
Ben-DorE, MetternichtG, GoldshlegerN, et al. Review of remote sensing-based methods to assess soil salinity [M/OL].Metternicht G,Zinck J A.Remote Sensing of Soil Salinization: Impact on Land Management. Boca Raton,CRC press,2008.
LeydenK, GoodwineB. Fractional-order system identification for health monitoring[J]. Nonlinear Dynamics,2018, 92:1-18.
[15]
XiaoChao, JiQingyuan, ChenJunqing, et al. Prediction of soil salinity parameters using machine learning models in an arid region of northwest China[J/OL]. Computers and Electronics in Agriculture,2023, 204:107512.
[16]
MoradpourM, Rostami PaydarG, Beiranvand PourA, et al. Landsat-7 and ASTER remote sensing satellite imagery for identification of iron skarn mineralization[J]. Geocarto International,2022, 37(7):1971-1998.
GeXiangyu, WangJinzhe, DingJianli, et al. Combining UAV-based hyperspectral imagery and machine learning algorithms for soil moisture content monitoring[J/OL]. PeerJ,2019, 7:e6926.
[22]
WangNan, PengJie, ChenSongchao, et al. Improving remote sensing of salinity on topsoil with crop residues using novel indices of optical and microwave bands[J/OL]. Geoderma,2022, 422:115935.
ElithJ, LeathwickJ R, HastieT. A working guide to boosted regression trees[J]. Journal of Animal Ecology,2008, 77(4): 802-813.
[28]
LundbergS M, LeeS I. A unified approach to interpreting model predictions[PP/OL].V2.arXiv(2017-11-25)[2026-04-22].
[29]
FriedmanJ H. Greedy function approximation: A gradient boosting machine[J]. Annals of Statistics,2001, 29:1189-1232.
[30]
GoldsteinA, KapelnerA, BleichJ, et al. Peeking inside the black box: Visualizing statistical learning with plots of individual conditional expectation[J]. Journal of Computational and Graphical Statistics,2015, 24(1): 44-65.
[31]
LiashchynskyiP, LiashchynskyiP. Grid search, random search, genetic algorithm: A big comparison for NAS[PP/OL]. V1.arXiv(2019-12-12)[2026-04-22].
Ben-DorE, TaylorR G, HillJ, et al. Imaging spectrometry for soil applications[J]. Advances in Agronomy,2009, 97: 321-392.
[34]
FariftehJ F, van der MeerM, van der MeijdeC, et al. Spectral characteristics of salt-affected soils: A laboratory experimen[J], Geoderma,2008, 145(3/4):196-206.
XuXibo, WangXiaoguang, YangPeijie, et al. Strategy for mapping soil salt contents during the bare soil period through a satellite image: Optimal calibration set combined with random forest[J/OL]. Catena,2023, 223:106900.
AllbedA, KumarL. Soil salinity mapping and monitoring in arid and semi-arid regions using remote sensing technology: A review[J]. Advances in Remote Sensing,2013, 2:373-385.
[40]
MetternichtG, ZinckJ A. Remote sensing of soil salinity: Potentials and constraints[J]. Remote Sensing of Environment,2003, 85(1):1-20.
[41]
HuangFeini, WeiShangguan, LiQingliang, et al. Beyond prediction: An integrated post-hoc approach to interpret complex model in hydrometeorology[J/OL]. Environmental Modelling & Software,2023, 167:105762.
[42]
MolnarC. Interpretable machine learning: A guide for making black box models explainable [M/OL].Raleigh:Lulu.com, 2020.