Scientists map threatened orchids using free data, cutting conservation costs
Researchers created a predictive model for tracking a protected Swedish orchid using only publicly available forest data—no expensive field surveys needed. The approach achieved 88% accuracy and could help governments and land managers identify conservation priorities at a fraction of traditional costs.
Originaltitel: Predicting the occurrence of a protected orchid and other old-growth specialists: A cost-effective approach using open geodata
<p>Predicting species occurrence across landscapes is fundamental to conservation planning, yet many approaches rely on costly field data. We developed and validated a model for predicting the occurrence of Goodyera repens, a threatened orchid associated with old-growth coniferous forests and legally protected in Sweden. Using 4805 plots from the Swedish National Forest Inventory (2013-2022), we modelled occurrence probability with logistic regression based on three predictors derived from publicly available geodata: basal area, soil moisture, and proportion of old forest in the surrounding landscape. Basal area had the strongest positive effect on occurrence (beta = 1.01, Delta AIC = 61.5), followed by proportion of old forest (&gt;= 80 years) within 100 m (beta = 0.61, Delta AIC = 30.5), while soil moisture had a negative effect (beta =-0.37, Delta AIC = 5.5). Model validation in southeastern Sweden using 227 independent records from the Swedish Species Observation System yielded an AUC of 0.88, with predicted probabilities at validation locations nearly 12 times higher than at random background points. The model also performed well for other conifer-associated species of conservation concern (AUC = 0.79), suggesting it captures general characteristics of high-quality conifer forest habitat. Performance was comparable in independent validations in central and northern Sweden (AUC = 0.91 for G. repens and 0.77 for indicator species), demonstrating geographic transferability. Our results show that simple geodata-based models can effectively predict the distribution of declining old-forest specialists, offering a cost-effective tool for habitat screening and forest management decisions in production landscapes.</p>