Diffuse reflectance spectroscopy for assessing soil quality indicators in cocoa production system validated using multi‐year data

Abstract

Diffuse reflectance spectroscopy is emerging as a reliable soil testing approach. However, few studies reported on the applicability of DRS for estimating the overall soil quality index (SQI) in smallholder farms. Specifically, the DRS approach has not been evaluated in cocoa production systems with multi‐year soil data. We evaluated the DRS approach in typical cocoa production systems from Papua New Guinea (PNG) by collecting soil samples before and after implementing soil test‐based nutrient recommendations. Specifically, 379 soil samples were collected in 2017 from cocoa farms in East New Britain, New Ireland, and Bouganville provinces of PNG. A second soil testing campaign was undertaken in 2019 to analyse 432 soil samples from the same farms to test the consistency of the DRS approach for SQI assessment. In all these samples, 15 soil parameters were measured using the conventional wet chemistry approach. Reflectance spectra in each sample were also measured over the visible to near‐infrared (wavelength: 350–2500 nm). A minimum dataset approach was followed to estimate SQIs from laboratory‐measured soil properties. With available N, clay content, available P, and available Zn adjudged as significant soil quality parameters for the cocoa plantations of PNG, estimated mean SQI values ranged from 0.30 in the cocoa farm from Bouganville province (BOKA‐PAN) in 2017 to 0.46 for the New Ireland province (LAU‐PAN) soil samples collected during 2019. Overall, SQI values increased from about 13%–21% following the nutrient applications. Based on the validation of two commonly‐used chemometric models (PLSRFS: partial‐least‐squares regression with feature selection, SVR: support vector regression), the SQI values were estimated with the coefficient of determination (R2) values as high as .94. Our results showed that a DRS model calibrated in soil samples collected in a specific year can be used for estimating SQI values in soil samples collected in different years. These results suggest that the principal component analysis (PCA) based SQI may be estimated from the DRS data with good accuracy and SQI values may potentially capture the management effects in smallholder farms.