Aboveground physiological response and yield prediction of Chloris gayana and Digitaria eriantha grown in rehabilitated coal mined soils using random forest algorithm

A recent study demonstrated that a blend of amendments improved both the physical and hydraulic properties of reclaimed mine soils more effectively than standard mine treatments, suggesting further research on its impact on plant growth. Additionally, there is currently no published research that has examined the potential of the random forest (RF) algorithm for predicting the aboveground yield of Chloris gayana (Rhodes grass) and Digitaria eriantha (Smutsfinger grass) grown in reclaimed mine soils. To address this, a field trial of 36 bins consisting of nine treatments and four replications each was conducted in a randomized block design at the experimental farm of the University of Pretoria. The results showed that the dry matter yield, leaf area index, and leaf water potential were all significantly ( p < 0.05) affected by the treatment. The blend of amendments increased aboveground dry matter yield by 70%–150% and leaf area index by 60%–95%. These improvements significantly enhanced productivity and, consequently, the carrying capacity of the rehabilitated land compared to the standard mine treatment of liming and fertilization. The most important wavelengths for predicting aboveground yield were located in the visible (400–700 nm) region of the electromagnetic spectrum, yielding an r 2 of 0.90, mean absolute error of 0.183 t ha −1 and root mean square error of 0.255 t ha −1 . These findings demonstrate that a blend of amendments can enhance the production potential of these grasses by improving soil nutrient availability. However, the longevity of these effects needs to be verified through long‐term studies. The results also indicate that RF algorithm can accurately predict aboveground yield of C. gayana and D. eriantha accurately based on changes in the plant canopy spectral signature.