Mapping cation exchange capacity using a quasi-3d joint inversion of EM38 and EM31 data

•Separate LR could not be developed between ECa and measured CEC at different depths.•Combined EM38 and EM31 ECa were inverted by a quisi-3d joint-inversion algorithm.•A universal LR model was built to predict CEC at various depths using inverted electrical conductivity.•CEC was mapped at various de...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Soil & tillage research Ročník 200; s. 104618
Hlavní autoři: Zhao, Dongxue, Li, Nan, Zare, Ehsan, Wang, Jie, Triantafilis, John
Médium: Journal Article
Jazyk:angličtina
Vydáno: Elsevier B.V 01.06.2020
Témata:
algorithms
Cation exchange capacity
cations
computer software
Digital soil mapping
Electromagnetic inversion
EM38
Fertility
prediction
regression analysis
soil electrical conductivity
subsoil
topsoil
Cation exchange capacity
Digital soil mapping
Fertility
EM38
Electromagnetic inversion
ISSN:0167-1987, 1879-3444
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:•Separate LR could not be developed between ECa and measured CEC at different depths.•Combined EM38 and EM31 ECa were inverted by a quisi-3d joint-inversion algorithm.•A universal LR model was built to predict CEC at various depths using inverted electrical conductivity.•CEC was mapped at various depths (0–2.1 m with 0.3 m increment) across a 26 ha field, using the universal LR model. Cation exchange capacity (CEC, cmol(+) kg−1) is a measure of the capacity of soil to retain and exchange cations. However, it is expensive to sample and directly measure across a heterogenous field and at different depths. To add value to limited data, proximally sensed apparent soil electrical conductivity (ECa, mS m−1) from electromagnetic (EM) instruments has been coupled to CEC at each depth through a linear regression (LR) model. In this study, LR between ECa and depth specific CEC was compared with a LR developed between true electrical conductivity (σ, mS m−1), inverted from ECa, and CEC from various depths, including topsoil (0–0.3 m), subsurface (0.3–0.6 m), shallow subsoil (0.6–0.9 m) and deeper subsoil (0.9–2.1 m). We estimate σ using quasi-3d (q-3d) inversion software (EM4Soil) considering inversion of EM38 and EM31 ECa either alone or in combination (joint inversion), in horizontal (ECah) and vertical (ECav) modes, and EM38 at two different heights (i.e. 0.2 or 0.4 m). The calibration results showed LR between ECa and depth specific CEC in the topsoil (R2 = 0.31), subsurface (0.37) and shallow subsoil (0.52) was unsatisfactory. Stronger LR could be established for deeper subsoil CEC (> 0.60). However, a single LR could be developed between CEC at all depths with σ (R2 = 0.72) estimated by jointly inverting EM38 (0.2 m) and EM31 ECa in both modes using a forward model (CF), inversion algorithm (S2) and small damping factor (λ = 0.03). A leave-one-out-cross-validation showed CEC prediction was precise (RMSE, 2.39 cmol(+) kg−1), unbiased (ME, -0.01 cmol(+) kg−1) with good concordance (Lin’s = 0.82). To improve areal prediction closer spaced transects are required, while to improve vertical resolution of prediction we recommend the use of a single-frequency multi-coil array DUALEM-421.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 23
ISSN:0167-1987
1879-3444
DOI:10.1016/j.still.2020.104618