Exploratory experimental study reveals soil moisture significantly alters apparent red and near-infrared reflectance in a soybean leaf model, highlighting background optical interference.
Vegetation reflectance under partial canopy cover can be influenced by soil background, complicating interpretation of red and near-infrared (NIR) signals. This exploratory study tested whether the apparent reflectance of a single soybean leaf differed among soil-background moisture treatments under a fixed leaf–soil–camera configuration. Images were acquired above clay soil at five moisture levels, expressed as percentages of maximum water-holding capacity, under natural illumination. Twenty trays were measured on 2 days; 10 images per tray and day were averaged as technical replicates, yielding 40 tray-by-day observations. NIR and red reflectance were analysed separately using repeated-measures generalized linear models with moisture group, day, band-specific illumination, acquisition cycle, and the moisture-group-by-day interaction as fixed effects. The primary NIR model used a Gamma distribution, log link, and compound-symmetry residual covariance; the red model used a Normal distribution, identity link, and diagonal residual covariance. Soil-moisture group was significant for NIR reflectance (F(4, 11.75) = 101.920, p < 0.001) and red reflectance (F(4, 20.32) = 19.922, p < 0.001). The moisture-group-by-day interaction was significant for red (p = 0.036); for NIR it did not reach 0.05 in the primary model (p = 0.063) and was specification-dependent in sensitivity analyses. A complementary random forest was evaluated using 10 repetitions of treatment-balanced, tray-grouped four-fold cross-validation. At the tray-by-day level, it achieved a coefficient of determination (R2) of 0.818 ± 0.021, a root-mean-square error (RMSE) of 0.034 ± 0.002, and a mean absolute error (MAE) of 0.025 ± 0.001, versus R2 = 0.780 ± 0.027 for a categorical linear baseline. The sampled soil-background moisture conditions were associated with apparent red and NIR reflectance differences in this leaf. The categories are experimental ranges rather than a physical threshold, and generality beyond this configuration remains unestablished.
No takes yet. Share an insight, caveat, or question.
Vasallo et al. (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: