LAUSR

Abstract Permanent grasslands produce highly diverse ecosystem goods and services, which need to be easily assessed by decision makers. Naturalists and agronomists classify grasslands in different ways to predict ecological, agronomic and environmental characteristics of the grasslands. However, few studies have compared the prediction abilities of these different classifications using the same botanical releves, and none has explored the utility of combining classifications. In this study, we attributed a grassland class from each of three classifications (phytosociological, agronomic and functional) to 250 permanent grasslands in north-eastern France to predict 16 characteristics: nine ecological, three agronomic and four environmental. We used statistical model selection to identify the classification or combination of classifications that best predicted each characteristic. Our results showed great prediction ability of agronomic classification, which created the best models for predicting agronomic (yield) and environmental (management, elevation) characteristics. We also identified a strong prediction ability of combining two or all three classifications to predict seven other grassland characteristics. However, grassland classifications did not predict most of the ecological characteristics well. We can assume that phytosociological classification, despite its mainstream use, predicts grassland characteristics less well than agronomic classification. We recommend combining grassland classifications to improve rapid prediction abilities. This study provides new knowledge useful for developing grassland classifications which meet the needs of agronomists and naturalists.