Hyperspectral datasets contain spectral noise, the presence of which adversely affects the classifier performance to generalize accurately. Despite machine learning algorithms being regarded as robust classifiers that generalize well under unfavourable noisy conditions, the extent of this is poorly understood. This study aimed to evaluate the influence of simulated spectral noise (10%, 20%, and 30%) on random forest (RF) and oblique random forest (oRF) classification performance using two node-splitting models (ridge regression (RR) and support vector machines (SVM)) to discriminate healthy and low infested water hyacinth plants. Results from this study showed that RF was slightly influenced by simulated noise with classification accuracies decreasing for week one and week two with the addition of 30% noise. In comparison to RF, oRF-RR and oRF-SVM yielded higher test accuracies (oRF-RR: 5.36%–7.15%; oRF-SVM: 3.58%–5.36%) and test kappa coefficients (oRF-RR: 10.72%–14.29%; oRF-SVM: 7.15%–10.72%). Notably, oRF-RR test accuracies and kappa coefficients remained consistent irrespective of simulated noise level for week one and week two while similar results were achieved for week three using oRF-SVM. Overall, this study has demonstrated that oRF-RR can be regarded a robust classification algorithm that is not influenced by noisy spectral conditions.

Unfettered agricultural activities have severely degraded vast areas of grasslands over the last decade. To rehabilitate and restore the productivity in affected grasslands, rangeland management practices still institute vast nitrogen-based fertilization regimes. However, excessive fertilization can often have damaging environmental effects. Over-fertilization can lead to nitrogen saturation. Although early indicators of nitrogen saturation have been documented, research detailing the near-real-time nitrogen saturation status of grasslands is required to better facilitate management protocols and optimize biomass production within degraded grasslands. Hence, the aim of this study was to discriminate nitrogen-saturated tropical grasses grown under a diverse fertilization treatment trial, using Worldview-3 satellite imagery and decision tree techniques. To accomplish this, nitrogen-saturated plots were first identified through specific physiological-based criteria. Thereafter, Worldview-3 satellite imagery (400–1040 nm) and decision tree techniques were applied to discriminate between nitrogen-saturated and -unsaturated grassland plots. The results showed net nitrate (NO₃⁻-N) concentrations and net pH levels to be significantly different (α = 0.05) between saturated and non-saturated plots. Moreover, the random forest model (overall accuracy of 91%) demonstrated a greater ability to classify saturated plots as opposed to the classification and regression tree method (overall accuracy of 79%). The most important variables for classifying saturated plots were identified as: the Red-Edge (705–745 nm), Coastal (400–450 nm), Near-Infrared 3 (838–950 nm), Soil-Adjusted Vegetation Index (SAVI) and the Normalized Difference Vegetation Index 3 (NDVI3). These results provide a framework to assist rangeland managers in identifying grasslands within the initial stages of nitrogen saturation. This will enable fertilization treatments to be adjusted in near-real-time according to ecosystem demand and thereby maintain the health and longevity of Southern African grasslands.

Eucalyptus scab disease (Elsinoë) is a harmful plant fungus that can disrupt various ecological and economic services provided by commercial forests. To effectively control and monitor the occurrence of forest pathogens, it is important to understand their spatial distribution within the infected area. Consistent monitoring, together with high-resolution imagery obtained from unmanned aerial vehicles (UAVs), has become important in forest management. Therefore, this study focuses on detecting and mapping the spatial distribution of E. masingae within commercial forests using image texture and vegetation indices (VIs) computed from a UAV sensor with machine learning (ML) and deep learning (DL) models. The fast large margin (FLM), random forest (RF), and deep learning (DL) models were used to determine which model effectively mapped the spatial distribution of the disease. The results indicated that image texture significantly increased the model accuracies (FLM = 94.8%; RF = 98.9%; DL = 98.9%) as opposed to the results without the use of image texture (FLM = 84.4%; RF = 76.1%; DL = 81.7%). Since the DL model obtained the fastest model run time and was proven to be the most significant model, it selected the mean, homogeneity, second moment, and correlation texture parameters which were predominantly determined from the red and blue bands of the UAV sensor containing visible bands. Additionally, the 3 × 3 moving window size was ideal for detecting E. masingae since the estimation of texture parameters was reduced efficiently. Overall, this study showcases the ability of UAVs to effectively map forest disease. Together with that, it has proven that the DL model outperformed the conventional ML models.

Water hyacinth (Eichhornia crassipes) is an exotic plant species that is effectively controlled byNeochetinaspp. weevils. This study is aimed at determining if spectroscopic data may be utilized to predict insect-induced stress on water hyacinth plants. Single target regression trees (STRTs), multitarget regression trees (MTRTs), and random forest multitarget regression trees (RF-MTRTs) have been used to predict feeding scar damage (FSD) and relative leaf chlorophyll content (RLCC) from hyperspectral canopy reflectance data. Results from this study show that the correlation coefficient of STRTs (training accuracy: 76%–97%; validation accuracy: 47%–86%) performs better than MTRTs (training accuracy: 74%–90%; validation accuracy: 45%–77%) for all infestation levels but are difficult to interpret simultaneously. In contrast, MTRTs (size: 23–35 nodes) are much smaller and more interpretable than STRTs (size: 11–47 nodes) because they predict FSD and RLCC simultaneously. Importantly, RF-MTRTs (training accuracy: 95%–98%; validation accuracy: 55%–88%) yield better predictive performance than STRTs and MTRTs for all infestation levels. It is concluded that MTRTs can be utilized for model interpretation as they are more interpretable; however, RF-MTRTs offer an improved predictive performance.