Remote sensing approaches to assess the Fall Armyworm (Spodoptera frugiperda) infestation on maize crop: a case study from Sri Lanka

Abstract

Managing pests and diseases of crops is important to ensuring global food security. Fall armyworm (Spodoptera frugiperda) is an invasive pest in Sri Lanka, causing significant damage to maize cultivation on the island. Continuous monitoring of fall armyworms is essential to ensure the high productivity of the crop. Traditional investigation methods used to examine fall armyworm incidence, such as field surveys, are time-consuming and labor- intensive. With the rapid development of remote sensing satellites, spectral reflectance measurements and vegetation indices have been used widely to monitor crop conditions. The present study was initiated to detect the spatial distribution of maize and the fall armyworm incidence in the Moneragala district, Sri Lanka with sentinel-2 multispectral images. The supervised maximum likelihood classification method was performed to determine the extent and spatial distribution of maize in the Moneragala district. Furthermore, remote sensing spectral vegetation indices, i.e., NDVI, SAVI, and NDRE and field surveys were performed to investigate the crop status and disease severity of fall armyworm. In the present study, three disease severity classes were recognized in terms of damage to the leaves, i.e., healthy (no visible leaf damage or less than 5% damage), slightly damaged (5% to 30% damage), and severely damaged (over 30% damage). The results revealed that NDVI, SAVI, and NDRE for healthy maize vegetation are 0.66±0.06, 0.88±0.03 and 0.41±0.02, respectively. Moreover, the disease severity classes of NDVI, SAVI, and NDRE were compared with One-way Analysis of variance (ANOVA) with the Tukey test for multiple comparisons. The results indicated a statistically significant difference between disease severity classes of NDRE (p<0.05), suggesting NDRE provides a more accurate measurement to detect fall armyworm incidence. The overall accuracy of the supervised image classification techniques was 89.78%, with a kappa coefficient of 0.88. Results was validated using statistics of maize extent data obtained from the Department of Census and Statistics, Sri Lanka, demonstrating significant accuracy (p<0.05). Therefore, present study revealed that remote sensing is an effective tool for mapping maize vegetation cover and early identification of fall armyworm incidence, making it a more economical and effective alternative to conventional methods.