Improved prediction of soil exchangeable sodium percentage (ESP) using wavelet

In Australia, the soil is considered sodic when the top meter has an exchangeable sodium percentage (ESP) greater than six (6%). Sodicity is a natural feature of Queensland, particularly the case in northeast Queensland where sugarcane is grown. To ameliorate sodicity effectively, a digital soil mapping (DSM) approach with digital data (e.g. elevation, gamma-ray [gamma-ray] spectrometry, and soil apparent electrical conductivity [(ECa)]) could be used. However, when a relationship between digital and soil data does not exist, wavelet transformation could be employed. Firstly, we aim to see if digital data, alone or in combination, can be used to predict topsoil (0-0.3 m) ESP using a linear mixed model (LMM). Secondly, we decompose the digital data using wavelet transform into components (i.e. detail [D] and approximate [A]) at different scales (i.e. 10, 20, 40, and 80 m). We then aim to develop an LMM to predict topsoil ESP using D or A or in combination. Results show that prediction of topsoil ESP by LMM using raw digital data either alone (Lin's = 0.27) or in combination (0.37) was poor (Lin's < 0.65). The prediction of topsoil ESP by LMM using D components alone was unsatisfactory at small scales, however, moderate agreement (0.65 < Lin's < 0.8) was achieved at scales of 80 m (0.66). In combination with D and A, a moderate agreement was achieved at 40 m (0.71) and a good agreement at 80 m (0.85). When considering the decomposed digital data individually, and at the same scale, the prediction of topsoil ESP using decomposed ECa (0.56) was better than gamma-ray (0.38). Interestingly, the decomposed elevation cannot be used individually to predict ESP. However, adding decomposed elevation to gamma-ray (0.58) and ECa (0.67), the prediction of ESP was enhanced. The best prediction of ESP was achieved using all decomposed data at 80 m (0.85). However, further work in validation would be required to confirm topsoil ESP was non-sodic (<5 %) and where it was deemed to be strongly sodic (>20 %) as these were not measured. Nevertheless, the DSM offers the real prospect of applying the six-easy-steps ameliorant management guidelines (Schroeder et al., 2006) to apply suitable rates of gypsum, including 2 t/ha for sodic (ESP; 5-10 %), 4 t/ha for moderately (10-15 %) and 6 t/ha for strongly (>15 %) sodic topsoil.

Automated summary

In Australia, sodic soil is defined as having an exchangeable sodium percentage (ESP) greater than 6% in the top meter. Sodicity is common in Queensland, especially in areas where sugarcane is grown. Digital soil mapping (DSM) with digital data such as elevation, gamma-ray spectrometry, and soil apparent electrical conductivity (ECa) can be used to effectively improve sodicity. When no relationship between digital and soil data exists, wavelet transformation can be employed. The study aims to predict topsoil ESP using a linear mixed model (LMM) with digital data and decomposed components obtained through wavelet transform at different scales. The results show that the LMM prediction using raw digital data alone or in combination was poor, but moderate to good agreement was achieved when using decomposed data at certain scales. Additionally, the combination of decomposed elevation with gamma-ray and ECa further enhanced the prediction of ESP. Overall, DSM offers the potential for effective management of sodicity and application of appropriate ameliorants.