Premium
Field level digital mapping of soil mineralogy using proximal and remote‐sensed data
Author(s) -
Nagra G.,
Burkett D.,
Huang J.,
Ward C.,
Triantafilis J.
Publication year - 2017
Publication title -
soil use and management
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.709
H-Index - 81
eISSN - 1475-2743
pISSN - 0266-0032
DOI - 10.1111/sum.12353
Subject(s) - illite , kaolinite , soil science , mineralogy , environmental science , compositional data , clay minerals , linear regression , quartz , remote sensing , geology , mathematics , statistics , paleontology
Primary (e.g., quartz) and secondary (clay) minerals are key factors determining the physical and chemical characteristics of soil. Understanding spatial distribution of minerals at the field scale would, therefore, be of potential benefit for soil management. However, current analysis requires time‐consuming laboratory procedures and computational quantification analysis (e.g., SIROQUANT). Furthermore, mineral composition (e.g., quartz, kaolinite, illite and expandable clay minerals) must sum to 100. We aimed to add value to laboratory data by developing multiple linear regression (MLR) relationships between mineralogy and ancillary data such as digital numbers (DNs) (i.e., Red [R], Green [G] and Blue [B]) acquired from remotely sensed air‐photographs and soil apparent electrical conductivity (EC a – mS/m) measured from proximal sensing electromagnetic (EM) instruments (i.e., EM38 and EM31). To account for composition, we compare results from the MLR approach with those from additive log‐ratio (ALR) transformation of mineralogy prior to MLR modelling. This approach together with various ancillary data and trend surface parameters (i.e., scaled Easting and Northing) has greater precision and less bias of prediction than the MLR approach using untransformed data. Our approach also enables predictions to sum to 100. We conclude that the most useful ancillary data to predict the abundance of quartz, kaolinite and illite are B DNs and EM31, while expandable clays are best predicted with R DNs, EM38 and scaled Northing. The use of ancillary data to map mineralogical components combined with ALR‐MLR is an effective approach, with resulting maps providing insights into soil and water management issues consistent with farmer experience.