The suitability of weed risk assessment as a conservation tool to identify invasive plant threats in East African rainforests

We tested the ability of an international Weed Risk Assessment (WRA) protocol to predict invasion status of 230 alien plant species introduced via a botanical garden to tropical rainforest in Tanzania. The reliability and accuracy of WRA in discriminating between invaders and non-invaders was independently assessed using field data on species demography and distribution. The WRA rejected 83% of known invaders, and accepted 74% of non-invaders. Only 1% of accepted species were known invaders at the site. WRA performance varied among different growth forms, and underestimated the risks arising from palm species. Among those species that had naturalised, the WRA was better at identifying invaders of open rather than forest habitats. The WRA score was significantly correlated with how widespread species had become at the site, suggesting some capacity to predict spatial spread at a landscape scale. Knowledge of propagule pressure and residence time did not increase explanatory power. These results indicate that the WRA was able to discriminate between invaders and non-invaders with accuracy comparable to similar assessments in temperate and sub-tropical regions. It could be made more effective by weighting traits important in tropical forests e.g. certain growth forms, shade tolerance etc. more heavily. Such a modified WRA could be used successfully elsewhere in the palaeotropics as a screening tool to identify the risk of invasion arising from plants introduced for agroforestry, horticulture or landscaping. Given the increasing pressures on tropical forests and importance of agroforestry to local economies, the WRA protocol represents a useful conservation tool.