Global climate change will probably exacerbate crop losses from insect pests, reducing agricultural production, and threatening food security. To predict where crop losses will occur, scientists have mainly used correlative models of species' distributions, but such models are unreliable when extrapolated to future environments. To minimize extrapolation, we developed mechanistic and hybrid models that explicitly capture range-limiting processes, and we explored how incorporating mechanisms altered the projected impacts of climate change for an agricultural pest, the South American locust (Schistocerca cancellata). Because locusts are generalist herbivores surrounded by food, their population growth may be limited by thermal effects on digestion more than food availability. To incorporate this mechanism into a distribution model, we measured the thermal effects on the consumption and defecation of field-captured locusts and used these data to model energy gain in current and future climates. We then created hybrid models by using outputs of the mechanistic model as predictor variables in correlative models, estimating the potential distribution of gregarious outbreaking locusts based on multiple predictor sets, modeling algorithms, and climate scenarios. Based on the mechanistic model, locusts can assimilate relatively high amounts of energy throughout temperate and tropical South America; however, correlative and hybrid modeling revealed that most tropical areas are unsuitable for locusts. When estimating current distributions, the top-ranked model was always the one fit with mechanistic predictors (i.e., the hybrid model). When projected to future climates, top-ranked hybrid models projected range expansions that were 23%–30% points smaller than those projected by correlative models. Therefore, a combination of the correlative and mechanistic approaches bracketed the potential outcomes of climate change and enhanced confidence where model projections agreed. Because all models projected a poleward range expansion under climate change, agriculturists should consider enhanced monitoring and the management of locusts near the southern margin of the range.

中文翻译：

---

气候变化有望提高消化率并引发爆发蝗虫的范围扩大

全球气候变化可能会加剧害虫造成的农作物损失，减少农业产量，并威胁粮食安全。为了预测作物损失将在何处发生，科学家们主要使用物种分布的相关模型，但这些模型在外推到未来环境时并不可靠。为了最大限度地减少外推，我们开发了明确捕获范围限制过程的机械模型和混合模型，并且我们探索了结合机制如何改变气候变化对农业害虫南美蝗虫（Schistocerca cancellata ）的预期影响). 由于蝗虫是被食物包围的通才食草动物，它们的种群增长可能受到消化热效应的限制，而不是食物供应的限制。为了将这种机制纳入分布模型，我们测量了对野外捕获的蝗虫的消耗和排便的热效应，并使用这些数据来模拟当前和未来气候下的能量增益。然后，我们通过使用机械模型的输出作为相关模型中的预测变量来创建混合模型，基于多个预测变量集、建模算法和气候情景估计群居爆发蝗虫的潜在分布。根据机械模型，蝗虫可以在南美洲温带和热带地区吸收相对大量的能量；然而，相关和混合模型表明，大多数热带地区不适合蝗虫。在估计当前分布时，排名靠前的模型始终与机械预测器（即混合模型）相吻合。当预测未来气候时，排名靠前的混合模型预测的范围扩展比相关模型预测的范围小 23%–30%。因此，相关方法和机制方法的结合将气候变化的潜在结果括起来，并增强了模型预测一致的信心。由于所有模型都预测在气候变化的情况下，范围会向极地扩展，因此农业学家应该考虑加强对范围南缘附近蝗虫的监测和管理。排名靠前的模型始终与机械预测器（即混合模型）相吻合。当预测未来气候时，排名靠前的混合模型预测的范围扩展比相关模型预测的范围小 23%–30%。因此，相关方法和机制方法的结合将气候变化的潜在结果括起来，并增强了模型预测一致的信心。由于所有模型都预测在气候变化的情况下，范围会向极地扩展，因此农业学家应该考虑加强对范围南缘附近蝗虫的监测和管理。排名靠前的模型始终与机械预测器（即混合模型）相吻合。当预测未来气候时，排名靠前的混合模型预测的范围扩展比相关模型预测的范围小 23%–30%。因此，相关方法和机制方法的结合将气候变化的潜在结果括起来，并增强了模型预测一致的信心。由于所有模型都预测在气候变化的情况下，范围会向极地扩展，因此农业学家应该考虑加强对范围南缘附近蝗虫的监测和管理。排名靠前的混合模型预测的范围扩展比相关模型预测的范围小 23%–30%。因此，相关方法和机制方法的结合将气候变化的潜在结果括起来，并增强了模型预测一致的信心。由于所有模型都预测在气候变化的情况下，范围会向极地扩展，因此农业学家应该考虑加强对范围南缘附近蝗虫的监测和管理。排名靠前的混合模型预测的范围扩展比相关模型预测的范围小 23%–30%。因此，相关方法和机制方法的结合将气候变化的潜在结果括起来，并增强了模型预测一致的信心。由于所有模型都预测在气候变化的情况下，范围会向极地扩展，因此农业学家应该考虑加强对范围南缘附近蝗虫的监测和管理。