Assessing the risks and opportunities of presence‐only data for conservation planning

AIM: Presence‐only data represent a significant source of information for quantifying biodiversity distributions and provide opportunities for use in conservation planning. The large databases of presence‐only records that are available and the lower cost of acquisition could help overcome the traditional problem of lack of data for conservation. However, there are risks associated with the use of presence‐only data inherent with the lack of true absences that might cause omission errors (species are erroneously thought to be absent) and loss of efficiency (more areas are thought to be necessary than needed). These errors could constrain the economic viability of conservation plans and thus the success of conservation practice. We therefore evaluated the opportunities and risks of using presence‐only data for conservation planning. LOCATION: Northern Australia. METHODS: The effects of using two different types (presence‐only and presence–absence) and different quantities of data were simulated by building predictive models on different subsets of data with increasing numbers of presence–absence or presence‐only records or a combination of both, for 80 freshwater fish species. We then compared the performance of conservation planning outcomes with the best information attainable (a true model built on the complete set of presence–absence data). We measured omission and commission errors in conservation planning outcomes, and the efficiency of and return on the investment in data acquisition. RESULTS: Including presence‐only data helped reduce commission and omission errors in conservation planning outcomes, but only when used in combination with at least some presence–absence data. The use of just a large quantity of presence‐only data resulted in significant reductions in the efficiency of conservation planning outcomes, as more areas than actually needed were required to achieve conservation targets. This reduction in efficiency was mainly related to inflated omission errors. MAIN CONCLUSIONS: We recommend using presence‐only data cautiously if this is the only source of data available; whenever possible, presence‐only data should be complemented with presence–absence data.