Research report: Leveraging spatial scale and temporal variation to track the spread rate of invasive spotted lanternflies ( Lycorma delicatula )

A retrospective analysis of United States spotted lanternfly records from 2014 through 2023 found that estimated spread varied over time and increased with the spatial scale used to delineate the invaded range. Around the primary invasion, median annual boundary displacement reached about 20.3 to 20.7 kilometres in 2017–2019, while newly detected satellite populations generally expanded more slowly during their first year. The study used presence and absence records compiled in the anonymised lydemapr database from state agencies, the US Department of Agriculture and public reports.

The database contained 490,211 presence records and 441,012 absence records; 38,744 records from state or federal surveyors met the establishment criterion of at least two living insects or one egg mass. Analyses used biological years, which turned over on 1 May. By 2023, established populations extended from 35.63 to 43.21 degrees north and to 87.64 degrees west.

The main invasion covered large parts of Pennsylvania, New Jersey, Delaware and Maryland, with outlying populations established in several other states. Researchers delineated invaded areas at multiple spatial scales using square grids and α-convex hulls. A grid cell counted as invaded if it contained an established population in the focal year or any earlier year.

For the hulls, the researchers tested 19 α′ values from 1 to 100 kilometres, plus 1,000 kilometres to approximate the minimum convex hull. The tested scales ranged from much finer than county scale to much coarser, although county sizes vary geographically. Spread was estimated by converting delineated area to an effective range radius and fitting both simple linear regressions and logistic curves against time.

Model fit was compared using the small-sample corrected Akaike Information Criterion, or AICc. Logistic curves generally fitted better, consistent with a slow-fast-slow temporal pattern, but their changing slopes meant that a single spread-rate estimate was not appropriate. Across the tested grid scales, maximum logistic-model rates rose from 9.6 kilometres per year with 1-kilometre cells to 57.6 kilometres per year with 100-kilometre cells.

Corresponding α-convex hull estimates increased from 2.4 to 39.9 kilometres per year. Coarser scales also shifted the estimated peak from 2018 to 2020. Five-fold cross-validation was used to compare how well each boundary method separated presence from absence records.

Performance was assessed with precision, recall and Fβ scores, with β controlling the relative emphasis placed on recall. α-convex hulls generally outperformed grids when precision and recall received equal weight; the methods were more comparable at higher β values. The researchers selected α′ = 35 kilometres for subsequent satellite analyses from pooled cross-validation results at β = 2, which placed greater weight on capturing presence records and reducing the risk of omitting outlying populations. Using the 35-kilometre hulls, the researchers identified 153 distinct satellite populations during the invasion.

Of these, 46 remained separate from the primary invasion for at least one year. Only 29 one-year-old epicentres met the study’s sampling criterion of at least 20 recorded absences within a 5-kilometre buffer around their margins in both observed years. Median boundary displacement in the first year ranged from 0 to 10.6 kilometres and was generally larger in the second year, ranging from 0 to 15.4 kilometres.

Most satellite populations had first-year median displacement below 1.5 kilometres, compared with more than 20 kilometres annually around the primary invasion three to five years after detection. Direct measurements around the primary epicentre in Berks County, Pennsylvania, showed median boundary displacement of 20.7 kilometres in 2017, 20.3 kilometres in 2018 and 20.5 kilometres in 2019. The respective interquartile ranges were 14.5–30.9, 12.0–31.3 and 13.1–31.1 kilometres.

Boundary displacement slowed from 2020 through 2023, but the cause remains unclear. During that period, the invasion reached the eastern shoreline and extended outside Pennsylvania for the first time, state agencies used different sampling plans, and survey effort was reduced by the COVID-19 pandemic and associated shutdowns. The authors proposed climatic conditions, accumulating natural enemies and more frequent small populations near the invasion margin as possible explanations for the slowdown, but the study did not establish their causal contributions.

The analysis also has important limitations. Recorded absences may reflect failed detection rather than true absence, and public reports may include false positives. Boundaries were drawn from state and federal presence records, with absences used for evaluation rather than boundary construction.

Satellite estimates were based on relatively few sufficiently sampled epicentres, and the cross-validation did not test sensitivity to the proportion or spatial arrangement of withheld records. Overall, the findings show that spatial aggregation can substantially alter measured invasion rates and estimated boundary locations, making scale selection an essential part of spread analysis. The authors recommend examining land cover, transport flows and environmental conditions, incorporating temperature-dependent population dynamics, and improving estimates of detection and false-positive rates in future analyses.

They also argue that the observed pattern of slower initial expansion around satellite populations provides a historical basis for surveillance and early action.

Subject of Research: Temporal and spatial-scale analysis of spotted lanternfly spread in the United States

Article Title: Leveraging spatial scale and temporal variation to track the spread rate of invasive spotted lanternflies ( Lycorma delicatula )

Article References: Keller, J., De Bona, S., & Helmus, M. R. (2026). Leveraging spatial scale and temporal variation to track the spread rate of invasive spotted lanternflies (Lycorma delicatula). NeoBiota, 106, 287-307. https://doi.org/10.3897/neobiota.106.177041

Image Credits: AI Generated

DOI: 10.3897/neobiota.106.177041

Keywords: spotted lanternfly, Lycorma delicatula, invasive species, invasion spread, spatial scale, boundary displacement, cross-validation, surveillance, United States, Leveraging, spatial, scale