Rangeland condition assessment in the Gobi Desert: A quantitative approach that places stakeholder evaluations front and Centre

There is widespread concern that the condition of rangelands in the Gobi Desert is declining. Opinions differ about how to translate these concerns into a defensible assessment of condition. Finding common ground is essential because condition measurements influence land-use decisions on large scales. We created ‘condition metrics’ for three Gobi Desert ecosystems. We do not use the word ‘condition’ to mean simply ‘the measurable conditions’ (attributes, observable state) of a site. Rather, we use it to describe the evaluated or judged ‘condition’ (health, desirability, goodness). The metrics explicitly represent the consensus view of a large (n = 92) and diverse stakeholder group, including nomadic pastoralists, botanists, wildlife ecologists and policymakers. The metrics were created by training models (regression trees) to predict stakeholder evaluation scores from site variables. These models can be used as metrics to produce a score for any site, on a scale of 0–100. We demonstrate using field tests that the metrics are practical to implement, sensitive to changes caused by management intervention, and produce scores which approximate the consensus view among stakeholders. There is a high level of redundancy among site variables, suggesting the metrics could be simplified for remote-sensing applications, where only some attributes are detectable. We conclude that the metrics are useful for evaluating rangeland condition in the Gobi Desert, and that the metrics represent the consensus opinion of a range of stakeholders. Our methods are applicable to ecosystem evaluation worldwide. •Quantitative judgements underpin ecological condition metrics for the Gobi Desert.•Machine learning methods found an explicit consensus among diverse stakeholders.•Field tests show the metrics are practical and useful.