The research, led by Prem Prakash Singh and Dibyendu Adhikari of the CSIR-National Botanical Research Institute in collaboration with the University of Lucknow, set out to build a resource-efficient surveillance workflow that municipal agencies could realistically adopt. The team focused on three invasive plants representing sharply different growth forms: Neltuma juliflora, a deep-rooted tree also known by its earlier scientific name Prosopis juliflora; Lantana camara, a dense woody shrub notorious across the tropics; and Parthenium hysterophorus, an aggressive annual herb blamed for crop losses and human health problems, including contact dermatitis and respiratory irritation. Choosing one tree, one shrub and one herb was deliberate, because the spectral and phenological signatures of these life forms differ in ways that matter for remote sensing.

The technical backbone of the approach is the Normalized Difference Vegetation Index, or NDVI, a measure derived from the way vegetation reflects red and near-infrared light. Healthy, photosynthetically active foliage reflects near-infrared strongly and absorbs red light, producing a characteristic ratio that satellites can measure from orbit. Sentinel-2, a pair of twin satellites operated by the European Space Agency’s Copernicus programme, revisits every point on Earth roughly every five days and captures imagery at a resolution fine enough to resolve urban vegetation patches. Rather than relying on a single snapshot, the researchers assembled monthly NDVI time series, exploiting phenology, the seasonal rhythm of leafing, flowering and senescence, to separate invasive species from the surrounding green matrix.

That phenological logic is what makes the method clever. Invasive plants often green up earlier, hold their leaves longer or peak at different times than native vegetation, creating temporal fingerprints that a single-date image would miss. The team fed these monthly spectral metrics, alongside environmental and land-cover predictors, into MaxEnt, a maximum-entropy species distribution model widely used in ecology to estimate habitat suitability from presence-only records. MaxEnt compares the environmental conditions at known occurrence points against the background landscape and produces a continuous suitability surface. Crucially, the model also reports how much each predictor contributes to the final prediction, and here the monthly variables told a species-specific story. The contributions of different months varied among the three species, and this variation pointed to candidate surveillance windows concentrated in the late post-monsoon period and in late winter, the seasons when each invader’s spectral behaviour diverges most clearly from its surroundings.

A model is only as good as its validation, and this is where Google Street View entered the picture, a pairing that seems almost whimsical until you consider the economics. Street View offers panoramic, street-level imagery of entire cities without a single kilometre of walking. The researchers sampled 90 Street View locations per species and scored them for visible detection of each plant. The results tracked modelled suitability strikingly well: detections rose consistently from low to high suitability classes, and each step up in suitability class was associated with 3.46 to 4.11 times higher odds of detecting the target species. In statistical terms, the MaxEnt surfaces were not just plausible maps; they carried real predictive information about where a street-level observer would actually encounter these plants.

Ground truthing in the field added a second layer of confirmation. Surveyors visited 30 high-suitability sites per species and confirmed presence for 27 of 30 sites for Lantana camara, 26 of 30 for Neltuma juliflora and 24 of 30 for Parthenium hysterophorus. Those hit rates, between 80 and 90 percent, are exactly the kind of efficiency gain that matters for city agencies. Instead of searching blindly, field crews can be dispatched to locations where the model says they are most likely to succeed, effectively multiplying the return on every hour of survey effort. The study also revealed a human factor worth noting: agreement between independent observers was lower for the herbaceous Parthenium than for the two woody species, a reminder that small, low-growing plants are harder to spot consistently, whether through a car-mounted camera or from the roadside.

Land-cover overlays added ecological texture to the results. The woody invaders showed their highest mean suitability in forested land, consistent with their need for established canopy and soil conditions, while the herb Parthenium peaked in built-up areas, thriving in the disturbed, nutrient-poor waste ground that cities produce in abundance. This distinction matters for management strategy, because a shrub lurking at the forest edge and a weed colonising roadside verges require entirely different detection and removal tactics. The team went further by integrating Dynamic World, a near real-time, 10-metre-resolution global land-cover product developed by Google and the World Resources Institute. By intersecting the maximum-suitability screening layer with areas flagged as having low vegetation probability in Dynamic World, they identified potential conflict zones where urban development and invasive habitat overlap.

The scale of that overlap is striking. Across Lucknow, the analysis flagged 1,000.61 hectares as potential conflict zones, land where high modelled suitability for invasive plants meets low vegetation cover, the conditions under which urban greening investments are most likely to be undermined. Within the city’s formally mapped administrative wards, 361.37 hectares of such zones fall squarely inside areas where municipal management decisions apply. For a city planning new parks, avenue plantings or restoration schemes, those hectares represent a prioritised to-do list: places where the money spent on greening is most at risk of being swallowed by invaders that outcompete planted seedlings and degrade the ecological integrity of the finished project.

What elevates the study beyond a local case study is its cost profile. Every major input is free or nearly free: Sentinel-2 imagery through Copernicus, MaxEnt as open-source software, Google Earth Engine for planetary-scale computation, Dynamic World as a public dataset, and Street View as a pre-existing photographic archive. The workflow requires no drones, no hyperspectral aircraft campaigns and no proprietary analytics, only the technical capacity to run the models and the local botanical knowledge to interpret them. The authors note that with local recalibration, including species-specific occurrence records and phenological tuning to regional climate, the same framework could inform surveillance policies and the allocation of management resources in other cities with expanding urban greening programmes.

The implications reach into some of the thorniest debates in invasion ecology. Trees like Neltuma juliflora occupy contested territory because they deliver ecosystem services, such as shade and fuelwood, even as they displace native communities and alter soils. Surveillance tools that map where such species are spreading, rather than debating whether they are good or bad in the abstract, give city planners a way to manage the trade-off spatially. There is also a public-health dimension: Parthenium is a recognised allergenic weed, and its affinity for built-up land means that dense neighbourhoods may face disproportionate exposure. In an era when climate change is shifting phenologies and expanding the ranges of opportunistic species, the Lucknow study suggests that the most powerful early-warning systems may come not from expensive bespoke technology but from creatively combining the planetary data infrastructure that already exists, orbiting overhead and parked on the street corner alike.

Subject of Research: Resource-efficient surveillance of invasive plants using Sentinel-2 phenology-based habitat modelling and Google Street View validation in Lucknow, India

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

Article References: Resource-efficient surveillance of invasive plants using Sentinel-2 phenology and street view validation. (n.d.). https://doi.org/10.1007/s10661-026-16063-w

Image Credits: AI Generated

DOI: 10.1007/s10661-026-16063-w

Keywords: invasive plants, Sentinel-2, NDVI, phenology, MaxEnt, Google Street View, urban greening, habitat suitability, Dynamic World, Lantana camara, Parthenium hysterophorus, Lucknow