Predicting host associations of the invasive spotted lanternfly on trees across the USA

Abstract Global impacts of invasive insect pests cost billions of dollars annually, but the impact of any individual pest species depends on the strength of associations with economically important plant hosts. Estimating host associations for a pest requires surveillance field surveys that observe pest association on plant species within an invaded area. However, field surveys often miss rare hosts and cannot observe associations with plants found outside the invaded range. Associations for these plants instead are estimated with experimental assays such as controlled feeding trials, which are time consuming and for which few candidate hosts can be tested logistically. For emerging generalist pests, these methods are unable to rapidly produce estimates for the hundreds of potential suitable hosts that the pest will encounter as it spreads within newly invaded regions. In such cases, association data from these existing methods can be statistically leveraged to impute unknown associations. Here we use phylogenetic imputation to estimate potential host associations in an emergent generalist forest pest in the U.S., the spotted lanternfly ( Lycorma delicatula ; SLF). Phylogenetic imputation works when closely related plants have similar association strengths, termed phylogenetic signal in host association, which is common in phytophagous insects. We first aggregated known SLF host associations from published studies. Existing research has estimated association strengths for 144 species across both the invaded and native range of SLF. These known associations exhibited phylogenetic signal. We then developed two protocols that combined known host association data and fit phylogenetic imputation models based on hidden state prediction algorithms to estimate association strength for 569 candidate tree species found across the continental U.S. Of candidate species considered, 255 are predicted to have strong associations with SLF in the U.S. and can be found in several clades including Juglandaceae, Rutaceae, Salicaceae, and Sapindaceae. Uninvaded regions with the highest numbers of these strongly associated species include midwestern and west coast states such as Illinois and California. Survey efforts for SLF should be focused on these regions and predicted species, which should also be prioritized in experimental assays. Phylogenetic imputation scales up existing host association data, and the protocols we present here can be readily adapted to inform surveillance and management efforts for other invasive generalist plant pests.