Inversion Methods of Soil Hydraulic Parameters Based on Hyperspectral Characteristics
To enrich remote inversion methods of soil hydraulic parameters and achieve a fast and nondestructive prediction, 59 high-spectral in-situ soil samples in Xiangzhou District, Zhuhai City, were collected using Field Spec4 spectrometer. The remote hydraulic parameter models were established for two la...
Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | zho |
| Published: |
Editorial Office of Pearl River
2024-01-01
|
| Series: | Renmin Zhujiang |
| Subjects: | |
| Online Access: | http://www.renminzhujiang.cn/thesisDetails?columnId=79074178&Fpath=home&index=0 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | To enrich remote inversion methods of soil hydraulic parameters and achieve a fast and nondestructive prediction, 59 high-spectral in-situ soil samples in Xiangzhou District, Zhuhai City, were collected using Field Spec4 spectrometer. The remote hydraulic parameter models were established for two land types: grassland and bare land (saturated water content <italic>θ</italic><sub>s</sub>, residual water content <italic>θ</italic><sub>r</sub>, inverse of air entry value <italic>α</italic>, shape coefficient <italic>n</italic>, saturated hydraulic conductivity <italic>K<sub>s</sub></italic>, and soil moisture content <italic>θ</italic>). The results are as follows. ① The spectral curves exhibit clear linearity in three bands: 700–750 nm, 830–1 100 nm, and 1 520–1 620 nm, with the mean values of coefficient of determination for the linear fitting <italic>R<sup>2</sup></italic><sup> </sup>all over 0.94. ② Among gradient boosting regression (GBR), partial least squares regression (PLSR), and random forest (RF), GBR performs the best and shows high sensitivity to the linear fitting parameter lg<italic>a</italic><sub>2</sub> (the logarithm of the slope) in the 830-1100 nm band. It accurately predicts <italic>θ</italic><sub>s</sub>, <italic>α</italic>, <italic>n</italic>, and <italic>θ</italic> for grassland types (Relative Percent Deviation, RPD> 1.4). For the bare land types, the coefficient of determination <italic>R</italic><sub>P</sub>² for all hydraulic characteristic parameters, except soil moisture content <italic>θ</italic>, is greater than 0.9, and RPD exceeds 2.0. ③ In most hydraulic characteristic models predicting the grassland types, the feature importance of the leaf area index (LAI) is the highest. Therefore, an effective inversion model is built based on field spectral data and LAI, with simple access to input parameters and good prediction effect, providing a potential for wide application in the future. |
|---|---|
| ISSN: | 1001-9235 |