Visible/NIR reflectance spectroscopy for twospotted spider mite (Tetranychus urticae Koch) detection and prediction on strawberry leaves

The feasibility of visible/near infrared (Vis/NIR) reflectance spectroscopy to detect and predict twospotted spider mite [TSSM] (Tetranychus urticae Koch) damage on strawberry leaves was investigated using spectral changes induced by TSSM injury. Two greenhouse experiments were carried out in spring 2009 and 2012 using strawberry varieties, Chandler and Florida Festival, respectively. Each variety was replicated 10 times. Strawberry plants were artificially inoculated with 0, 5, 15, 25 TSSM per leaf, to achieve a control, low, medium and high mite infestation levels, respectively. At 14, 21, 28, and 35 days after TSSM inoculation, two strawberry leaves were collected from each plant, mite numbers enumerated and respective spectral reflectance of each leaf scanned from 200 to 2500 nm at every wavelength using a spectrophotometer. Partial Least Squares (PLS) regression was used to relate mite numbers on the leaves with their corresponding spectral readings. In addition, the ability to accurately predict unknown samples was evaluated based on coefficient of determination (R²) and residual predictive deviation (RPD). Results indicate that TSSM injury on the leaves resulted in spectral changes within the visible and NIR wavebands. Overall, the best prediction model for TSSM numbers for Festival variety had a root mean square error (RMSE) of 17 mites per leaflet with an R² of 0.6. Predictive models generated for the Chandler variety were ‘poor’ (RPD < 1.29). Predictive models for the varieties were different indicating that in a mixed strawberry planting different models must be developed for different varieties.