Estimation of population thresholds of plant-parasitic nematodes on cotton using hyperspectral remotely sensed data

Hand-held hyperspectral data were collected from Rotylenchulus reniformis infected cotton plants in field microplots and cotton production fields to investigate the utility of remotely sensed data in detecting R. reniformis population numbers. Hyperspectral reflectance data were collected from individual cotton leaves and cotton plant canopies in combination with soil samples infested by R. reniformis. Nematode population numbers and corresponding hyperspectral plant data were analyzed using a Matlab hyperspectral toolkit employing the use of self-organizing maps to extract relevant wavelengths used in correlation classifications and with linear discriminant analysis to test toolkit classification accuracies.Relevant wavelengths for discriminating R. reniformis populations were within 451--949 nm comprising the visible and near-infrared (NIR) regions of the electromagnetic spectrum (EMS). Matlab hyperspectral toolkit correlation classifications ranged from 60--100% and were more accurate for single locations than for multiple locations in most instances. Linear discriminant analysis accuracies were higher in 2001 than 2002 possibly due to better spectral collection conditions. Reflectance data collected from the plant's single leaf was the best reference target for acquiring reflectance information relative to the presence of R. reniformis and population estimation on production field cotton. However, plant canopy provided the best reference (100% classification accuracies) for reflectance analysis of cotton grown in field microplots. For multiple location analysis, the single leaf target was the best reference. The best time to acquire cotton plant reflectance data and soil samples for R. reniformis analysis was approximately 40 days after planting. Data acquired during this time using the single leaf in production field analysis or plant canopy in microplot analysis, provide the highest classification accuracies in accurately detecting and estimating R. reniformis on cotton plants.