N-mixture models for population estimation: Application in spotted lanternfly egg mass survey

Population density and structure are critical to nature conservation and pest management. Traditional sampling methods such as capture-mark-recapture and catch-effort can't be used in situations where catching, marking, or removing individuals are not feasible. N-mixture models use repeated count data to estimate population abundance based on detection probability. They are widely adopted in wildlife surveys in recent years to account for imperfect detection. However, its application in entomology is relatively new. In this paper, we describe the general procedures of N-mixture models in population studies from data collection to model fitting and evaluation. Using Lycorma delicatula egg mass survey data at 28 plots in seven sites from the field, we found that detection probability (p) was negatively correlated with tree diameter at breast height (DBH), ranged from 0.516 [95 % CI: 0.470-0.561] to 0.614 [95 % CI: 0.566-0.660] between the 1st and the 3rd sample period. Furthermore, egg mass abundance (λ) was positively associated with basal area (BA) for the sample unit (single tree), with more egg masses on tree of heaven (TOH) trees. More egg masses were also expected on trees of other species in TOH plots. Predicted egg mass density (masses/100 m²) ranged from 5.0 (95 % CI: 3.0-16.0) (Gordon) to 276.9 (95 % CI: 255.0-303.0) (Susquehannock) for TOH plots, and 11.0 (95 % CI: 9.00-15.33) (Gordon) to 228.3 (95 % CI: 209.7-248.3) (Burlington) for nonTOH plots. Site-specific abundance estimates from N-mixture models were generally higher compared to observed maximum counts. N-mixture models could have great potential in insect population surveys in agriculture and forestry in the future.