Integrating multi-scale species distribution modeling and field validation to guide conservation of the endangered Hine’s Emerald Dragonfly (Somatochlora hineana)

Abstract The Hine’s Emerald Dragonfly ( Somatochlora hineana ; HED) is a federally endangered habitat specialist in the United States occurring in disjunct populations from the Great Lakes region to southeastern Missouri. Its elusive nature and specialized habitat requirements necessitate efficient methods for locating extant populations. Our study aimed to: (1) develop a detection probability model for HED to optimize survey efforts; (2) create and field-validate a regional Species Distribution Model (SDM) for the greater Chicago region (Northern Illinois/Southern Wisconsin) to guide surveys for new populations; and (3) develop and compare SDMs for the full U.S. range, Great Lakes region, and Missouri region to investigate broad-scale and regional habitat associations. All SDMs demonstrated good predictive performance (AUCs: 0.889–0.970). Field validation of the Chicago Region SDM resulted in discovery of a novel HED locality in Will County, Illinois, in atypical habitat lacking dolomite bedrock. Key predictors varied regionally: wetland soils and vegetation type were primary predictors rangewide, elevation and vegetation type were key in the Great Lakes, and wetland soils and depth to restrictive layer were most important in Missouri. The detection probability model indicated low overall detectability, mostly influenced by temperature, time of day, day of year, and survey effort, limiting robust validation of SDM-predicted absences. Integrating SDMs with occupancy-based modeling provides an effective, objective approach for prioritizing survey locations for rare species, as demonstrated by our discovery of a previously unknown population of HED, in non-traditional habitat. Regional variation in key predictors highlights the importance of local ecological context for management. However, survey protocols must account for imperfect detection when validating predictive models. Implications for Insect conservation Low detectability is common among imperiled insects, making surveys difficult and expensive. SDMs can reduce site-selection bias, furthering our understanding of species distributions. SDM-guided, occupancy-based surveys maximize limited conservation resources, while providing probabilistic species distribution estimates.