In this study, the most influential bands used to partition the observations at the root nodes of MTRT models are located in the red edge region more specifically at 694 nm, 661 nm, and 689 nm for the low, medium, and high infestation levels, respectively.
Caption: Figure 6: An example of a MTRT to simultaneously predict FSD and RLCC from canopy spectral reflectance for the high infestation level.
Multitarget regression trees (MTRTs) predict several numeric response variables simultaneously  and offer several advantages over STRTs.
More specifically, the objectives of this study are to (1) compare the interpretability of STRTs and MTRTs to predict FSD and RLCC of variable infection levels on water hyacinth plants and (2) compare the predictive performance of STRTs, MTRTs, and RF-MTRTs to predict FSD and RLCC of variable infection levels on water hyacinth plants.
The random forest algorithm constructs an ensemble of individually grown MTRTs with the prediction of response variables (i.e., FSD and RLCC) based on an average prediction of the response variables for all the regression trees in the forest [46-48].
The predictive performance (correlation coefficient and RMSE) of the MTRTs and RF-MTRTs for the three infestation levels is presented in Table 3.
RF-MTRTs consistently achieved a higher predictive performance than single MTRTs when predicting FSD and RLCC for the three infestation levels (Table 3).
MTRTs are the best regression tree models to interpret and understand the relationship between reflectance spectra and biocontrol measures.
Overall, STRTs are larger, less interpretable, and less informative than MTRTs to predict biocontrol measures from hyperspectral canopy reflectance.
The predictive performance of both MTRTs and STRTs for all infection levels is relatively strong.
In this study, RF-MTRTs perform better than both STRTs and MTRTs for all infestation levels (Table 2; Table 3).
RF-MTRTs predicted both biocontrol measures for low infection levels with a relatively high accuracy and better than MTRTs. The ability to predict biocontrol measures accurately at the initial stages of the infestation is highly beneficial to biocontrol initiatives in determining if weevil populations are alive and establishing within water hyacinth infestations.