Modelling moisture‐induced soil reflectance changes in cultivated sandy soils: a case study in citrus orchards

Abstract

The value of hyperspectral imagery in agricultural management has been amply demonstrated. Despite this, image interpretation is often drastically impeded by changes in soil moisture content (SMC). Soil moisture variations dominate the spectral reflectance in the 350–2500‐nm wavelength domain and affect the effectiveness of spectral indices used to monitor variations in soil and vegetation properties. Consequently, removing soil moisture effects in spectral images is critical and would provide a significant breakthrough for agricultural remote sensing. Yet, current available soil moisture reflectance models fail to properly address the moisture‐induced reflectance changes occurring within soils of the same texture class. This very much limits the operational implementation of these models, particularly in agricultural fields where within‐field variations in soil characteristics such as organic matter and clay content prevail. In this study, the effect of SMC on the reflectance in the 400–2500‐nm spectral domain was studied for six sandy soils located in citrus orchards in the Western Cape Province, South Africa. In a series of experiments, novel insights into soil moisture reflectance modelling of sandy cultivated soils are provided. The wavelength and soil‐specific variations in model parameters are mechanistically modelled and a general model for moist reflectance of cultivated sand soils is presented and successfully tested. Model comparison and validation demonstrate that the model fit of soil‐type‐specific models can be approximated (R2 = 0.82; RRMSE = 0.14) while a significant increase in model fit compared with the traditional general models (R2 = 0.59; RRMSE = 0.28) was achieved.