Synergistic use of Sentinel-1 and Sentinel-2 data for Fall Armyworm infestation detection and mapping in maize croplands

Fall Armyworm (FAW) is a widespread invasive pest in maize crops. This study aimed at detecting and mapping FAW infestations in maize fields across Bangladesh, using freely available Sentinel-1 and Sentinel-2 data. Field observations were conducted during the 2019–2020 maize growing season in 579 maize fields across six administrative divisions of Bangladesh. The study covered both infested and non-infested sites across four crop growth phases, namely vegetative phases 9 (V9) and 12 (V12), as well as the silking and maturing phases. Synthetic Aperture Radar backscatter values, spectral reflectance profiles, and eight vegetation indices were extracted from the Sentinel data and analysed using non-parametric statistical tests to identify differences between infested and non-infested fields. Machine learning models, specifically Random Forest - and Support Vector Machine, were then used to classify infestation severity based on five model input data combinations: (i) Sentinel-1, (ii) Sentinel-2, (iii) Sentinel-2 with vegetation indices, (iv) Sentinel-1 and Sentinel-2, and (v) Sentinel-1, Sentinel-2, and vegetation indices. The results indicated that infested maize fields exhibited reduced near-infrared reflectance and distinct backscatter patterns in σVH o, with notable variations at silking and maturity phases. The red edge (740 nm), near-infrared (865 nm) and shortwave infrared (1610–2190 nm) bands were particularly effective in distinguishing infestation levels across all growing phases. Among the studied vegetation indices, the Normalized Difference Vegetation Index , Modified Chlorophyll Absorption in Reflectance Index, Red Edge Simple Ratio, and Modified Simple Ratio - were identified as the most significant indicators for discriminating between non-infested and infested maize classes at all growing phases. RF achieved 94–96% accuracy (96% in V9) versus SVM’s 78–80% using only Sentinel‐1 data. Multi‐source (Sentinel-1, Sentinel-2 and Vegetation Indices) integration improved both models in most cases. These results underscore the potential of integrating multi-source remote sensing data for scalable and accurate pest detection. Freely available Sentinel data is a valuable source of information for early pest detection and management aiding policymakers in identifying high-risk areas, implementing timely interventions, and promoting sustainable pest management strategies to protect maize production and reduce economic losses.