Resource-efficient surveillance of invasive plants using Sentinel-2 phenology and street view validation

Invasive plants can undermine the ecological integrity and long-term benefits of urban greening. Yet municipal agencies often lack scalable tools for targeted surveillance. We evaluated a resource-efficient workflow in Lucknow, India, combining Sentinel-2 Normalized Difference Vegetation Index (NDVI) time series, MaxEnt modelling, field checks and Google Street View (GSV). Three invasive plants representing contrasting growth forms were selected: Neltuma juliflora (Sw.) Raf. (syn. Prosopis juliflora (Sw.) DC.; tree), Lantana camara L. (shrub) and Parthenium hysterophorus L. (herb). Monthly predictor contributions varied among species, identifying candidate surveillance windows in late post-monsoon and late-winter periods. Field surveys confirmed species presence at 26/30, 27/30 and 24/30 selected high-suitability sites for L. camara, N. juliflora and P. hysterophorus, respectively. Across 90 GSV sites per species, detections increased from low to high suitability, with each class increase associated with 3.46-4.11-fold higher odds of detection. Inter-observer agreement was lower for the herb than for the woody species. Land-cover overlays showed the highest mean suitability in forest for the woody species and in built-up land for the herb. Intersecting the maximum-suitability screening layer with low Dynamic World vegetation probability identified 1,000.61 ha of potential conflict zones, including 361.37 ha within mapped wards. Overall, the workflow supports targeted surveillance by prioritising areas for field verification. With local recalibration, it could inform municipal policies for invasive-plant surveillance and the allocation of management resources within urban greening programmes in other cities also.