Mapping Cation Exchange Capacity (CEC) Across Sugarcane Fields with Different Comparisons by Using DUALEM Data

The sugarcane growing soil in far-north Queensland is sandy, and infertile. To ensure productivity, nutrient guidelines recommend lime application rates based on soil cation exchange capacity (CEC). However, laboratory determination of CEC is expensive. Because CEC is often correlated with soil apparent electrical conductivity (ECa, mS/m) measured from electromagnetic induction (EM) instruments, ECa can be used to predict CEC. Using ECa may lead to uncertainty in prediction, but estimates of true electrical conductivity (σ, mS/m) generated from inversion of ECa, can be correlated with depth-specific CEC. In this study, we compared linear regression (LR) between ECa from a DUALEM-421S and CEC at specific depths ( i.e., topsoil [0–0.3 m], subsurface [0.3–0.6 m], subsoil [0.6–0.9 m] and deep subsoil [0.9–1.2 m]), with a LR of σ using a quasi-2d (q-2d) or quasi-3d (q-3d) inversion of DUALEM-421S ECa and CEC at all depths. The use of a multiple linear regression (MLR) to predict CEC, using σ with depth and location ( i.e., Easting and Northing) is also explored along with σ from the other EM products ( i.e., DUALEM-1S and DUALEM-21S). The minimum number of calibration sample locations ( i.e., n = 165, 150,…, 15) is also investigated. The LR between ECa ( e.g., 1mPcon) and CEC were very weak (R2 = 0.27) and weak (0.36) in the topsoil and subsurface, respectively, but moderate in the subsoil (0.57) and deep subsoil (0.67). The LR between σ, estimated from q-2d (R2 = 0.66) and q-3d (0.64) inversion of DUALEM-421S ECa, and CEC at all depths was moderate. In terms of prediction agreement, the Lin's concordance correlation coefficient (LCCC) was moderate for q-2d (0.79) and q-3d (0.75). Using a MLR with depth, coordinates and σ, led to an improvement in calibration using q-2d (R2 = 0.71) or q-3d (0.67), with prediction agreement substantial for q-2d (LCCC = 0.83) and moderate for q-3d (0.78), with comparable agreement from DUALEM-1S and DUALEM-2S (0.77) estimates of σ. The minimum number of calibration samples for a strong MLR R2 (>0.7) and substantial and good agreement was 15 for q-2d and 30 for q-3d, respectively. The final digital soil mapping of topsoil CEC developed using MLR and σ estimated from q-3d of DUALEM-421S ECa could be used to apply the Australian sugarcane industry lime application guidelines with areas with intermediate (3–6 cmol[+]/kg) and small (<3 cmol[+]/kg) topsoil CEC requiring 4 and 2.25 t/ha, respectively.