TY - JOUR
T1 - Aggregate size distribution of arid and semiarid laboratory soils (<2 mm) as predicted by VIS-NIR-SWIR spectroscopy
AU - Ben Dor, Eyal
AU - Francos, Nicolas
AU - Ogen, Yaron
AU - Banin, Amos
N1 - Publisher Copyright:
© 2022 Elsevier B.V.
PY - 2022/6/15
Y1 - 2022/6/15
N2 - Soil aggregation is an important property, affecting issues such as soil structure and dust load. To investigate the potential to evaluate soil aggregation status via proximal sensing, we conducted a comprehensive study using the legacy soil spectral library (SSL) of Israel representing arid and semiarid environments. The SSL was segregated into six soil aggregate size fractions as follows: 2–1.4 mm, F1; 1.4–1.0 mm, F2; 1.0–0.5 mm, F3; 0.5–0.25 mm, F4; 0.25–0.1 mm, F5; <0.1 mm, F6; and the average aggregate size (AVG, in mm) was calculated. In addition, another 74 soil attributes were measured along with their sample reflectance spectra across the 0.4–2.5-μm spectral region. A comprehensive correlation matrix between all soil attributes enabled isolating four cementing agents (CAs) that bind the primary particles into aggregates: clay content, clay mineral (smectite), organic matter and free iron oxide content. Generating pedotransfer functions (PTFs) with these CAs revealed equations that fairly predicted the aggregate size fractions F1 (R2 = 0.68), F2 (R2 = 0.79), F3 (R2 = 0.67), F5 (R2 = 0.61) and AVG (R2 = 0.78) with high accuracy. The six aggregate size fractions and AVG were divided into three groups based on their relation to the CAs: group A (F1, F2, F3, AVG) presenting positive correlations with the CAs, group B (F4, F6) presenting poor relationships with the CAs, and group C (F5) presenting negative correlations with the CAs. As the CAs were found to be chromophoric substances, it was possible to predict each CA from spectral-based models. A separate spectral-based analysis was also performed to evaluate the aggregate size fractions directly with no a priori information or PTF adoption. This analysis revealed high statistical agreement with spectral assignments for the four selected CAs. Whereas groups A and C were successfully predicted in the validation stage spectral-based models (F1 [R2 = 0.68], F2 [R2 = 0.79], F3 [R2 = 0.7], F5 [R2 = 0.57] and AVG [R2 = 0.67]), the predictions of group B were poorer relationships against the selected CAs that present important spectral assignments. We concluded that soil aggregation stage can be assessed directly or indirectly (via PTF) using spectral analysis and a data-mining approach. Assuming that the reflectance information from hyperspectral remote-sensing means, such as EMIT (NASA initiative), will soon be available from orbit (2022), this approach may pave the way for monitoring soil aggregation status from afar, to determine the soil's potential as a dust source.
AB - Soil aggregation is an important property, affecting issues such as soil structure and dust load. To investigate the potential to evaluate soil aggregation status via proximal sensing, we conducted a comprehensive study using the legacy soil spectral library (SSL) of Israel representing arid and semiarid environments. The SSL was segregated into six soil aggregate size fractions as follows: 2–1.4 mm, F1; 1.4–1.0 mm, F2; 1.0–0.5 mm, F3; 0.5–0.25 mm, F4; 0.25–0.1 mm, F5; <0.1 mm, F6; and the average aggregate size (AVG, in mm) was calculated. In addition, another 74 soil attributes were measured along with their sample reflectance spectra across the 0.4–2.5-μm spectral region. A comprehensive correlation matrix between all soil attributes enabled isolating four cementing agents (CAs) that bind the primary particles into aggregates: clay content, clay mineral (smectite), organic matter and free iron oxide content. Generating pedotransfer functions (PTFs) with these CAs revealed equations that fairly predicted the aggregate size fractions F1 (R2 = 0.68), F2 (R2 = 0.79), F3 (R2 = 0.67), F5 (R2 = 0.61) and AVG (R2 = 0.78) with high accuracy. The six aggregate size fractions and AVG were divided into three groups based on their relation to the CAs: group A (F1, F2, F3, AVG) presenting positive correlations with the CAs, group B (F4, F6) presenting poor relationships with the CAs, and group C (F5) presenting negative correlations with the CAs. As the CAs were found to be chromophoric substances, it was possible to predict each CA from spectral-based models. A separate spectral-based analysis was also performed to evaluate the aggregate size fractions directly with no a priori information or PTF adoption. This analysis revealed high statistical agreement with spectral assignments for the four selected CAs. Whereas groups A and C were successfully predicted in the validation stage spectral-based models (F1 [R2 = 0.68], F2 [R2 = 0.79], F3 [R2 = 0.7], F5 [R2 = 0.57] and AVG [R2 = 0.67]), the predictions of group B were poorer relationships against the selected CAs that present important spectral assignments. We concluded that soil aggregation stage can be assessed directly or indirectly (via PTF) using spectral analysis and a data-mining approach. Assuming that the reflectance information from hyperspectral remote-sensing means, such as EMIT (NASA initiative), will soon be available from orbit (2022), this approach may pave the way for monitoring soil aggregation status from afar, to determine the soil's potential as a dust source.
KW - Dust source
KW - Pedotransfer function
KW - Soil aggregation
KW - Soil spectroscopy
UR - http://www.scopus.com/inward/record.url?scp=85125952662&partnerID=8YFLogxK
U2 - 10.1016/j.geoderma.2022.115819
DO - 10.1016/j.geoderma.2022.115819
M3 - ???researchoutput.researchoutputtypes.contributiontojournal.article???
AN - SCOPUS:85125952662
SN - 0016-7061
VL - 416
JO - Geoderma
JF - Geoderma
M1 - 115819
ER -