Hyperspectral indices for assessing damage by the solenopsis mealybug (Hemiptera: Pseudococcidae) in cotton

• This study was carried out to assess mealybug feeding damage in field cotton using hyperspectral reflectance data.
• Analyses indicate strong relationships between spectral vegetation indices and mealybug damage severity.
• New vegetation stress indices and multinominal logistic regression models were developed to quantify mealybug damage.
• In addition, significant reduction in chlorophyll and relative water content due to mealybug infestation were observed.

Phenacoccus solenopsis Tinsley, a native of North America, is a widespread exotic mealybug infesting cotton, Gossypium spp. in several countries. Monitoring of this pest is generally undertaken through regular field surveys, which is labour intensive, time consuming and error prone. Alternately, radiometry is a reliable technique for rapid and non-destructive assessment of plant health. Thus, a study was conducted to characterize reflectance spectra of cotton plants with known mealybug infestation levels (grade-0 is healthy and grade-4 is severe), and seek to identify specific narrow wavelengths sensitive to mealybug damage. Reflectance measurements were made in the spectral range of 350–2500 nm using a hyperspectral radiometer. Significant differences were found in green, near infrared and short wave infrared spectral regions for plants with early stages of P. solenopsis infestation, and for plants showing higher grades of infestation these differences extended to all the regions except blue. A significant reduction in total chlorophyll (12.83–35.83%) and relative water content (1.93–23.49%) was observed in the infested plants. Reflectance sensitivity analysis of the hyperspectral data revealed wavelengths centered at 492, 550, 674, 768 and 1454 nm as most sensitive to mealybug damage. Mealybug Stress Indices (MSIs) were developed using two or three wavelengths, tested using multinomial logistic regression (MLR) analysis and compared with other indices published earlier. Results showed that the MSIs were superior (R2 = 0.82) to all other spectral vegetation indices tested. Further, the proposed MLR models corresponding to each MSIs were validated using two independent field data sets. The overall percent correct classification of cotton plants into different mealybug damage severity grades was in the range of 38.3 and 54.9. High classification accuracy for grade-1 (81.8%) showed that models are capable of early detection of mealybug damage. Results of this study could suggest potential usage of remote sensing in monitoring spatial and temporal distribution of the solenopsis mealybug, and thereby enable effective planning and implementation of site-specific pest management practices.