Johnson Andrew J, Bednar David, Hulcr Jiri
School of Forest, Fisheries and Geomatics Sciences, University of Florida, Gainesville, Florida, USA.
Florida State Collection of Arthropods, FDACS-DPI, Gainesville, Florida, USA.
Ecol Appl. 2025 Jul;35(5):e70072. doi: 10.1002/eap.70072.
Pest risk assessment informs regulatory decisions to facilitate safe trade while also protecting a country's agricultural and environmental resources. The first step in pest risk assessment is pest categorization which can help determine whether an in-depth examination is needed. We created a model to predict the potential impact of non-indigenous bark and ambrosia beetles (Curculionidae: Scolytinae). This model uses biological variables derived from extensive assessment of alien species and produces a five-point scale of impact prediction. We accommodate uncertainty and missing data using random decision tree forests with Monte Carlo simulations. Non-indigenous bark beetles include both invasive species with significant ecological impacts, such as widespread tree death, and others that pose little risk. We assembled a comprehensive list of 60 introduced non-native bark beetle species in the continental United States as the training set. Forty-two potentially predictive variables were chosen from reports on behaviors, pestilence, recorded damage/interpretations in literature, biological traits, and interactions with fungi including plant pathogens. The model builds upon strategies used by USDA-APHIS in existing risk assessments, specifically the Objective Prioritization of Exotic Pests (OPEP) model, with changes in the following: (1) a transparent dataset for building and training the model enabling future updates and use in other systems, (2) uncertainty simulations using values derived from an extensive natural history matrix rather than an assumed equal distribution, and (3) predictions made on the probability of multiple impact levels, allowing users to decide based on acceptable risk. The model is designed for pest risk analysis for Scolytinae in the continental United States but can be adapted to other pests or regions. We tested the model's performance by iteratively removing each species from the training set and retraining the model. The retrained models accurately predicted the removed species. To demonstrate the model's application, we predicted the impact of scolytine beetles not yet present in the continental United States, Xylosandrus morigerus and Hypoborus ficus, plus an additional hypothetical species with no known data. Our model predicts that these species are likely to have moderate impacts and unlikely to have high impacts if they were introduced.
有害生物风险评估为监管决策提供依据，以促进安全贸易，同时保护一个国家的农业和环境资源。有害生物风险评估的第一步是有害生物分类，这有助于确定是否需要进行深入审查。我们创建了一个模型来预测非本土树皮甲虫和食菌小蠹（象甲科：小蠹亚科）的潜在影响。该模型使用从对外来物种的广泛评估中得出的生物学变量，并生成一个五点影响预测量表。我们使用带有蒙特卡洛模拟的随机决策树森林来处理不确定性和缺失数据。非本土树皮甲虫包括具有重大生态影响的入侵物种，如大面积树木死亡，以及其他风险较小的物种。我们整理了一份在美国大陆引入的60种非本土树皮甲虫物种的综合清单作为训练集。从行为、疫病、文献中记录的损害/解读、生物学特征以及与包括植物病原体在内的真菌的相互作用等报告中选择了42个潜在预测变量。该模型基于美国农业部动植物卫生检验局在现有风险评估中使用的策略构建，特别是外来有害生物目标优先级（OPEP）模型，在以下方面有所变化：（1）用于构建和训练模型的透明数据集，便于未来更新并用于其他系统；（2）使用从广泛的自然历史矩阵得出的值而非假定的均匀分布进行不确定性模拟；（3）对多个影响水平的概率进行预测，允许用户根据可接受风险做出决策。该模型专为美国大陆小蠹亚科的有害生物风险分析而设计，但可适用于其他有害生物或地区。我们通过从训练集中迭代移除每个物种并重新训练模型来测试模型的性能。重新训练的模型准确预测了被移除的物种。为了展示该模型的应用，我们预测了美国大陆尚未出现的小蠹甲虫——桑天牛和榕小蠹，以及另外一个无已知数据的假设物种的影响。我们的模型预测，如果引入这些物种，它们可能会产生中等影响，不太可能产生高影响。
