物种潜在分布地预测系统PSDS的设计与实现

本研究设计并实现了一个基于地理信息系统(GIS)的仅用物种已知分布点数据预测物种潜在分布地的PSDS系统.在这一系统中,通过层次聚类算法对物种已知分布点数据进行处理,减少了异常值对预测结果的影响,从而解决了环境包络模型预测结果过于乐观的问题,在物种已知分布数据较少时也能取得较好的结果.该系统实现了数据加载与导出、图层浏览与显示、生态因子分析与分布地预测、结果展示等功能,操作方便,简单易用.本文以白冠长尾雉(Syrmaticus reevesii)为例,根据4个省的少量已知分布点数据对其在国内的潜在分布地进行了预测,获得了较理想的结果,可为该物种的保护提供科学依据.     

使用伪氨基酸组成和模糊支持向量机预测蛋白质结构类

蛋白质结构类预测是生物信息和蛋白质科学中重要的研究领域.基于Chou提出的伪氨基酸离散模型框架,从蛋白质序列出发,设计一种新的伪氨基酸组成方法表示蛋白质序列样本.抽取氨基酸组合(10-D)在序列中出现的频率和疏水氨基酸模式(6-D)表示蛋白质序列的附加特征,用和传统的氨基酸组成(20-D)一起构成的36维的伪氨基酸组成向量来表示蛋白质序列的特征.使用遗传算法来优化附加特征的权重系数.伪氨基酸组成向量作为输入数据,模糊支持向量机作为预测工具.使用三个常用的标准数据集来验证算法的性能.Jack-knife检验结果说明本方法具有较高的准确率,有望成为潜在的预测蛋白质功能的工具.     

生境概率预测值转换为二元值过程中4个阈值选择方法的比较评估——以珙桐和杉木生境预估为例

物种生境模型预测结果通常是概率性的,然而在具体的保护管理等实践应用过程中通常需要基于二元值(存在/不存在)的分布图,此时就需要把概率性的预测结果转化为二元值,在此转化过程中就涉及阈值选择问题。此外,在评估模型预测准确度的时候,多数评估指标也需要选择一个阈值用于转化概率预测结果,这个阈值选择对于模型预测准确度也会有极大的影响。然而阈值选择却是物种生境模拟不确定性研究中较少涉及的领域。"随机森林"既可以生成物种生境概率分布图(回归算法)也可以生成二元分布图(分类算法),然而还未见对两种预测方式的比较研究。该文以珙桐(Davidia involucrata)和杉木(Cunninghamia lanceolata)为例,分别采用"随机森林"的分类算法和回归算法预测其生境二元分布图和概率分布图,通过4个不同阈值选择方法(默认值0.5、MaxKappa、MaxTSS和MaxACC)把概率预测图转换为二元分布图,进而比较分析转换结果对模型预估的影响。珙桐不同阈值选择方法所确立的阈值之间存在显著差异,而杉木没有显著差异;两物种模型准确度之间没有显著差异;在预测两物种未来气候条件下的生境面积变化、生境分布区迁移方向和距离以及最适宜海拔分布高度变化时,二元值转换后的回归算法与分类算法之间存在显著差异,但回归算法中各阈值选择方法之间没有显著差异。空间生境分布图的相似性分析表明MaxKappa和MaxTSS法具有最大相似性,分类算法与4种阈值选择方法之间具有最大差异。     

基于地理环境相似度的长江经济带入侵物种虚拟负样本生成方法

入侵物种空间分布建模的核心数据源来源于物种多样性采样(物种出现点和未出现点),然而,大多数入侵物种标本库只记录物种出现点样本信息,缺乏对未出现点(负样本)位置的记录。因此,生成有效的入侵物种虚拟负样本是建立物种空间分布模型的关键。本文提出了一种基于地理环境相似度的虚拟负样本生成方法。首先利用主成分分析(PCA)方法对地理环境原始变量进行线性相关性建模,基于提取的主成分,采用K-means算法对入侵物种样本进行聚类分析并计算各样本的地理环境相似度。在此基础上,通过建立基于主成分的入侵物种相似性度量与可信度计算框架来识别虚拟负样本。以长江经济带入侵物种一年蓬(Erigeron annuus)数据集为例,分析了整个区域虚拟负样本的可信度。结果表明,与空间随机采样和单类支持向量机采样相比,用本研究提出的方法生成的样本数据建立的logistic回归和支持向量机预测结果更优,验证了该方法的可行性与有效性。基于地理环境相似度的虚拟负样本抽样策略有助于解决由于随机采样而引起的误采样到潜在入侵点的难题,同时负样本的可信度能有助于识别不同等级的入侵物种适应区。     

利用癌功能类从癌基因组突变谱中识别癌基因

从大规模癌样本基因突变扫查数据中识别癌基因具有重要的意义. 一些重要功能的改变对于癌的发生发展是必需的, 因此将它们定义为癌功能类, 并从GO(Gene Ontology)中选择一组显著富集已知癌基因的细致功能类来代表它们. 为了评价以癌相关功能类作为特征识别癌基因的效果, 将已知的蛋白激酶癌基因定义为阳性金标准, 而将其他的蛋白激酶基因定义为阴性金标准. 结果表明, 与利用选择压力作为特征的方法比较, 利用癌相关功能类作为特征的方法可以更有效地识别癌基因. 进一步结合癌相关功能类与基因非同义突变个数可以产生更可靠的预测结果. 最后, 将46个注释到癌相关功能类并且其非同义突变个数至少为3的蛋白激酶基因预测为癌基因, 预测精确率达到0.42.     

Maxent模型复杂度对物种潜在分布区预测的影响

生态位模型在入侵生物学和保护生物学中具有广泛的应用, 其中Maxent模型最为流行, 被越来越多地应用在预测物种的现实分布和潜在分布的研究中。在Maxent模型中, 多数研究者采用默认参数来构建模型, 这些默认参数源自早期对266个物种的测试, 以预测物种的现实分布为目的。近期研究发现, Maxent模型采用复杂机械学习算法, 对采样偏差敏感, 易产生过度拟合, 模型转移能力仅在低阈值情况下较好。基于默认参数的Maxent模型不仅预测结果不可靠, 而且有时很难解释。在本研究中, 作者以入侵害虫茶翅蝽(Halyomorpha halys)为例, 采用经典模型构建方案(即构建本土模型然后将其转移至入侵地来评估), 利用ENMeval数据包来调整本土Maxent模型调控倍频和特征组合参数, 分析各种参数条件下模型的复杂度, 然后选取最低复杂度的模型参数(即为最优模型), 综合比较默认参数和调整参数后Maxent模型的响应曲线和预测结果, 探讨Maxent模型复杂度对预测结果的影响及Maxent模型构建时所需注意事项, 以期对物种潜在分布进行合理的预测, 促进Maxent模型在我国的合理运用和发展。作者认为, 环境变量的选择至关重要, 需要综合分析其对所模拟物种分布的限制作用和环境变量之间的空间相关性。构建Maxent模型前需对物种分布采样偏差及模型的构建区域进行合理地判断, 模型构建时需要比较不同参数下模型的预测结果和响应曲线, 选取复杂度较低的模型参数来最终建模。在茶翅蝽的分析中, Maxent模型的默认参数和最优模型参数不同, 与Maxent模型默认参数相比, 采用调整参数后所构建的模型预测效果较好, 响应曲线较为平滑, 模型转移能力较高, 能够较为合理反映物种对环境因子的响应和准确地模拟该物种的潜在分布。     

预测物种潜在分布区——比较SVM与GARP

 物种分布与环境因子之间存在着紧密的联系,因此利用环境因子作为预测物种分布模型的变量是当前最普遍的建模思路,但是绝大多数物种分 布预测模型都遇到了难以解决的“高维小样本"问题。该研究通过理论和实践证明,基于结构风险最小化原理的支持向量机(Support vector machine, SVM)算法非常适合“高维小样本"的分类问题。以20种杜鹃花属(Rhododendron)中国特有种为检验对象,利用标本数据和11个1 km×1 km的栅格环境数据层作为模型变量,预测其在中国的潜在分布区,并通过全面的模型评估——专家评估,受试者工作特征(Receiver operator characteristic, ROC)曲线和曲线下方面积(Area under the curve, AUC)——来比较模型的性能。我们实现了以SVM为核心的物种分布预测 系统,并且通过试验证明其无论在计算速度还是预测效果上都远远优于当前广泛使用的规则集合预测的遗传算法(Algorithm for rule-set prediction, GARP)预测系统。     

蛋白质网络聚类算法分析平台的设计与实现

蛋白质网络聚类是识别功能模块的重要手段,不仅有利于理解生物系统的组织结构,对预测蛋白质功能也具有重要的意义。针对目前蛋白质网络聚类算法缺乏有效分析软件的事实,本文设计并实现了一个新的蛋白质网络聚类算法分析平台ClusterE。该平台实现了查全率、查准率、敏感性、特异性、功能富集分析等聚类评估方法,并且集成了FAG-EC、Dpclus、Monet、IPC-MCE、IPCA等聚类算法,不仅可以对蛋白质网络聚类分析结果进行可视化,并且可以在不同聚类分析指标下对多个聚类算法进行可视化比较与分析。该平台具有良好的扩展性,其中聚类算法以及聚类评估方法都是以插件形式集成到系统中。     

心脏手术风险评估欧洲系统在OPCAB术后ICU中的应用

现在对于不停跳冠脉旁路移植术(OPCAB)的患者术后的预测的模型有很多种,这些模型大多用于预测术后死亡率、术后并发症,手术方式的选择、手术资源的应用价值的评估等。心脏手术风险评估欧洲系统(Euro SCORE)也是其中一种,它对于现代OPCAB术后死亡率的预测比较合理。但是随着手术外科的发展,Euro SCORE模型在中、低危组过高估计术后的死亡率,而在高危组又过低估计术后死亡率。此外,Euro SCORE模型也应用于预测术后并发症、住院费用多少、在ICU住院时间及机械通气时间,得到广泛应用,并在世界范围内得到推广,包括欧美等国家。最近在中国,也开始对心脏手术风险评估欧洲系统大量了的数据研究,并发展到对其他手术术后的预测及治疗。     

红外光谱结合化学计量学快速检测熊胆粉真伪及掺伪问题CSCD

利用红外光谱技术对熊胆粉正品、伪品及掺伪品进行快速鉴别并预测掺伪比例,建立熊胆粉的快速质量评价方法。采集94批熊胆粉正品、70批伪品(猪胆粉、牛胆粉、羊胆粉、兔胆粉、鸡胆粉、鸭胆粉、鹅胆粉)及180批掺伪品(掺伪猪胆粉、掺伪牛胆粉)的红外光谱图,利用正交偏最小二乘判别分析(OPLS-DA)及偏最小二乘回归(PLSR)分别建立熊胆粉正品、伪品和掺伪品的定性校正模型及不同类别掺伪品掺伪比例的定量校正模型。熊胆粉正品、伪品及掺伪品的定性校正模型对样本的判别准确率分别为99.64%(校正集)和95.65%(验证集);通过进一步判别分析,可鉴别熊胆粉伪品及掺伪品种类,准确率均大于95%;在2个不同类别掺伪品定量校正模型中,验证集相关系数(R_(V)^(2))和预测均方根误差(RMSEP)分别为0.9874、2.6300%(熊胆粉掺伪猪胆粉)和0.9826、3.1887%(熊胆粉掺伪牛胆粉);3个定性模型及2个定量模型均表现出优秀的预测能力。本研究建立的红外光谱结合化学计量学方法可实现对熊胆粉正品、伪品和掺伪品的快速鉴别及掺伪比例确定,为熊胆粉的质量控制及评价提供参考。     

利用分布有/无数据预测物种空间分布的研究方法综述

详细的物种地理分布信息是生态学研究和制定保护策略的基础。相比较于直接估测种群数量,获取物种分布的有/无数据更为实用。因此,利用分布有/无数据并结合环境变量建立模型预测物种空间分布的方法在近年来得到了长足发展,并被广泛应用。利用分布有/无数据预测物种分布,关键的步骤包括：1)构建总体概念模型,2)收集物种分布有/无数据,并准备环境变量图层;3)选择合适的统计模型和算法,以及4)对模型进行评估。概念模型提出研究假设,并确定数据收集及模型方法。收集物种分布数据有系统调查及非系统调查方法。筛选并准备与物种分布相关的环境变量,利用GIS工具处理,使之成为符合模型条件的具有合适的空间尺度的数字化图层。利用环境变量和物种分布有/无的数据,选择合适的方法及软件建立模型,并对模型进行检验和评估。我们总结了用于构建物种分布模型的不同算法和软件。本文将针对以上各个环节,阐述利用物种分布有/无数据进行研究所需要的技术细节,以期望为读者提供借鉴。     

The MIAmaxent R package: Variable transformation and model selection for species distribution models

The widely used “Maxent” software for modeling species distributions from presence‐only data (Phillips et al., Ecological Modelling, 190, 2006, 231) tends to produce models with high‐predictive performance but low‐ecological interpretability, and implications of Maxent's statistical approach to variable transformation, model fitting, and model selection remain underappreciated. In particular, Maxent's approach to model selection through lasso regularization has been shown to give less parsimonious distribution models—that is, models which are more complex but not necessarily predictively better—than subset selection. In this paper, we introduce the MIAmaxent R package, which provides a statistical approach to modeling species distributions similar to Maxent's, but with subset selection instead of lasso regularization. The simpler models typically produced by subset selection are ecologically more interpretable, and making distribution models more grounded in ecological theory is a fundamental motivation for using MIAmaxent. To that end, the package executes variable transformation based on expected occurrence–environment relationships and contains tools for exploring data and interrogating models in light of knowledge of the modeled system. Additionally, MIAmaxent implements two different kinds of model fitting: maximum entropy fitting for presence‐only data and logistic regression (GLM) for presence–absence data. Unlike Maxent, MIAmaxent decouples variable transformation, model fitting, and model selection, which facilitates methodological comparisons and gives the modeler greater flexibility when choosing a statistical approach to a given distribution modeling problem.     

GLM versus CCA spatial modeling of plant species distribution

Despite the variety of statistical methods available for static modeling of plant distribution, few studies directly compare methods on a common data set. In this paper, the predictive power of Generalized Linear Models (GLM) versus Canonical Correspondence Analysis (CCA) models of plant distribution in the Spring Mountains of Nevada, USA, are compared. Results show that GLM models give better predictions than CCA models because a species-specific subset of explanatory variables can be selected in GLM, while in CCA, all species are modeled using the same set of composite environmental variables (axes). Although both techniques can be readily ported to a Geographical Information System (GIS), CCA models are more readily implemented for many species at once. Predictions from both techniques rank the species models in the same order of quality; i.e. a species whose distribution is well modeled by GLM is also well modeled by CCA and vice-versa. In both cases, species for which model predictions have the poorest accuracy are either disturbance or fire related, or species for which too few observations were available to calibrate and evaluate the model. Each technique has its advantages and drawbacks. In general GLM will provide better species specific-models, but CCA will provide a broader overview of multiple species, diversity, and plant communities.     

Performance evaluation of cetacean species distribution models developed using generalized additive models and boosted regression trees

Species distribution models (SDMs) are important management tools for highly mobile marine species because they provide spatially and temporally explicit information on animal distribution. Two prevalent modeling frameworks used to develop SDMs for marine species are generalized additive models (GAMs) and boosted regression trees (BRTs), but comparative studies have rarely been conducted; most rely on presence‐only data; and few have explored how features such as species distribution characteristics affect model performance. Since the majority of marine species BRTs have been used to predict habitat suitability, we first compared BRTs to GAMs that used presence/absence as the response variable. We then compared results from these habitat suitability models to GAMs that predict species density (animals per km²) because density models built with a subset of the data used here have previously received extensive validation. We compared both the explanatory power (i.e., model goodness of fit) and predictive power (i.e., performance on a novel dataset) of the GAMs and BRTs for a taxonomically diverse suite of cetacean species using a robust set of systematic survey data (1991–2014) within the California Current Ecosystem. Both BRTs and GAMs were successful at describing overall distribution patterns throughout the study area for the majority of species considered, but when predicting on novel data, the density GAMs exhibited substantially greater predictive power than both the presence/absence GAMs and BRTs, likely due to both the different response variables and fitting algorithms. Our results provide an improved understanding of some of the strengths and limitations of models developed using these two methods. These results can be used by modelers developing SDMs and resource managers tasked with the spatial management of marine species to determine the best modeling technique for their question of interest.     

Modeling bird species distribution change in fire prone Mediterranean landscapes: incorporating species dispersal and landscape dynamics

Species distribution models (SDMs) have traditionally been founded on the assumption that species distributions are in equilibrium with environmental conditions and that these species–environment relationships can be used to estimate species responses to environmental changes. Insight into the validity of this assumption can be obtained from comparing the performance of correlative species distribution models with more complex hybrid approaches, i.e. correlative and process‐based models that explicitly include ecological processes, thereby accounting for mismatches between habitat suitability and species occupancy patterns. Here we compared the ability of correlative SDMs and hybrid models, which can accommodate non‐equilibrium situations arising from dispersal constraints, to reproduce the distribution dynamics of the ortolan bunting Emberiza hortulana in highly dynamic, early successional, fire driven Mediterranean landscapes. Whereas, habitat availability was derived from a correlative statistical SDM, occupancy was modeled using a hybrid approach combining a grid‐based, spatially‐explicit population model that explicitly included bird dispersal with the correlative model. We compared species occupancy patterns under the equilibrium assumption and different scenarios of species dispersal capabilities. To evaluate the predictive capability of the different models, we used independent species data collected in areas affected to different degree by fires. In accordance with the view that disturbance leads to a disparity between the suitable habitat and the occupancy patterns of the ortolan bunting, our results indicated that hybrid modeling approaches were superior to correlative models in predicting species spatial dynamics. Furthermore, hybrid models that incorporated short dispersal distances were more likely to reproduce the observed changes in ortolan bunting distribution patterns, suggesting that dispersal plays a key role in limiting the colonization of recently burnt areas. We conclude that SDMs used in a dynamic context can be significantly improved by using combined hybrid modeling approaches that explicitly account for interactions between key ecological constraints such as dispersal and habitat suitability that drive species response to environmental changes.     

From topography to hydrology—The modifiable area unit problem impacts freshwater species distribution models

Species distribution models (SDMs) are statistical tools to identify potentially suitable habitats for species. For SDMs in river ecosystems, species occurrences and predictor data are often aggregated across subcatchments that serve as modeling units. The level of aggregation (i.e., model resolution) influences the statistical relationships between species occurrences and environmental predictors—a phenomenon known as the modifiable area unit problem (MAUP), making model outputs directly contingent on the model resolution. Here, we test how model performance, predictor importance, and the spatial congruence of species predictions depend on the model resolution (i.e., average subcatchment size) of SDMs. We modeled the potential habitat suitability of 50 native fish species in the upper Danube catchment at 10 different model resolutions. Model resolutions were derived using a 90‐m digital‐elevation model by using the GRASS‐GIS module r.watershed. Here, we decreased the average subcatchment size gradually from 632 to 2 km². We then ran ensemble SDMs based on five algorithms using topographical, climatic, hydrological, and land‐use predictors for each species and resolution. Model evaluation scores were consistently high, as sensitivity and True Skill Statistic values ranged from 86.1–93.2 and 0.61–0.73, respectively. The most contributing predictor changed from topography at coarse, to hydrology at fine resolutions. Climate predictors played an intermediate role for all resolutions, while land use was of little importance. Regarding the predicted habitat suitability, we identified a spatial filtering from coarse to intermediate resolutions. The predicted habitat suitability within a coarse resolution was not ported to all smaller, nested subcatchments, but only to a fraction that held the suitable environmental conditions. Across finer resolutions, the mapped predictions were spatially congruent without such filter effect. We show that freshwater SDM predictions can have consistently high evaluation scores while mapped predictions differ significantly and are highly contingent on the underlying subcatchment size. We encourage building freshwater SDMs across multiple catchment sizes, to assess model variability and uncertainties in model outcomes emerging from the MAUP.     

Wrong,but useful: regional species distribution models may not be improved by range‐wide data under biased sampling

Species distribution modeling (SDM) is an essential method in ecology and conservation. SDMs are often calibrated within one country's borders, typically along a limited environmental gradient with biased and incomplete data, making the quality of these models questionable. In this study, we evaluated how adequate are national presence‐only data for calibrating regional SDMs. We trained SDMs for Egyptian bat species at two different scales: only within Egypt and at a species‐specific global extent. We used two modeling algorithms: Maxent and elastic net, both under the point‐process modeling framework. For each modeling algorithm, we measured the congruence of the predictions of global and regional models for Egypt, assuming that the lower the congruence, the lower the appropriateness of the Egyptian dataset to describe the species' niche. We inspected the effect of incorporating predictions from global models as additional predictor (“prior”) to regional models, and quantified the improvement in terms of AUC and the congruence between regional models run with and without priors. Moreover, we analyzed predictive performance improvements after correction for sampling bias at both scales. On average, predictions from global and regional models in Egypt only weakly concur. Collectively, the use of priors did not lead to much improvement: similar AUC and high congruence between regional models calibrated with and without priors. Correction for sampling bias led to higher model performance, whatever prior used, making the use of priors less pronounced. Under biased and incomplete sampling, the use of global bats data did not improve regional model performance. Without enough bias‐free regional data, we cannot objectively identify the actual improvement of regional models after incorporating information from the global niche. However, we still believe in great potential for global model predictions to guide future surveys and improve regional sampling in data‐poor regions.     

The effect of sample size and species characteristics on performance of different species distribution modeling methods

Species distribution models should provide conservation practioners with estimates of the spatial distributions of species requiring attention. These species are often rare and have limited known occurrences, posing challenges for creating accurate species distribution models. We tested four modeling methods (Bioclim, Domain, GARP, and Maxent) across 18 species with different levels of ecological specialization using six different sample size treatments and three different evaluation measures. Our assessment revealed that Maxent was the most capable of the four modeling methods in producing useful results with sample sizes as small as 5, 10 and 25 occurrences. The other methods compensated reasonably well (Domain and GARP) to poorly (Bioclim) when presented with datasets of small sample sizes. We show that multiple evaluation measures are necessary to determine accuracy of models produced with presence-only data. Further, we found that accuracy of models is greater for species with small geographic ranges and limited environmental tolerance, ecological characteristics of many rare species. Our results indicate that reasonable models can be made for some rare species, a result that should encourage conservationists to add distribution modeling to their toolbox.     

A global organism detection and monitoring system for non-native species

Harmful invasive non-native species are a significant threat to native species and ecosystems, and the costs associated with non-native species in the United States is estimated at over $120 Billion/year. While some local or regional databases exist for some taxonomic groups, there are no effective geographic databases designed to detect and monitor all species of non-native plants, animals, and pathogens. We developed a web-based solution called the Global Organism Detection and Monitoring (GODM) system to provide real-time data from a broad spectrum of users on the distribution and abundance of non-native species, including attributes of their habitats for predictive spatial modeling of current and potential distributions. The four major subsystems of GODM provide dynamic links between the organism data, web pages, spatial data, and modeling capabilities. The core survey database tables for recording invasive species survey data are organized into three categories: “Where, Who & When, and What.” Organisms are identified with Taxonomic Serial Numbers from the Integrated Taxonomic Information System. To allow users to immediately see a map of their data combined with other user's data, a custom geographic information system (GIS) Internet solution was required. The GIS solution provides an unprecedented level of flexibility in database access, allowing users to display maps of invasive species distributions or abundances based on various criteria including taxonomic classification (i.e., phylum or division, order, class, family, genus, species, subspecies, and variety), a specific project, a range of dates, and a range of attributes (percent cover, age, height, sex, weight). This is a significant paradigm shift from “map servers” to true Internet-based GIS solutions. The remainder of the system was created with a mix of commercial products, open source software, and custom software. Custom GIS libraries were created where required for processing large datasets, accessing the operating system, and to use existing libraries in C++, R, and other languages to develop the tools to track harmful species in space and time. The GODM database and system are crucial for early detection and rapid containment of invasive species.     

Projecting future expansion of invasive species: comparing and improving methodologies for species distribution modeling

Modeling the distributions of species, especially of invasive species in non‐native ranges, involves multiple challenges. Here, we developed some novel approaches to species distribution modeling aimed at reducing the influences of such challenges and improving the realism of projections. We estimated species–environment relationships for Parthenium hysterophorus L. (Asteraceae) with four modeling methods run with multiple scenarios of (i) sources of occurrences and geographically isolated background ranges for absences, (ii) approaches to drawing background (absence) points, and (iii) alternate sets of predictor variables. We further tested various quantitative metrics of model evaluation against biological insight. Model projections were very sensitive to the choice of training dataset. Model accuracy was much improved using a global dataset for model training, rather than restricting data input to the species’ native range. AUC score was a poor metric for model evaluation and, if used alone, was not a useful criterion for assessing model performance. Projections away from the sampled space (i.e., into areas of potential future invasion) were very different depending on the modeling methods used, raising questions about the reliability of ensemble projections. Generalized linear models gave very unrealistic projections far away from the training region. Models that efficiently fit the dominant pattern, but exclude highly local patterns in the dataset and capture interactions as they appear in data (e.g., boosted regression trees), improved generalization of the models. Biological knowledge of the species and its distribution was important in refining choices about the best set of projections. A post hoc test conducted on a new Parthenium dataset from Nepal validated excellent predictive performance of our ‘best’ model. We showed that vast stretches of currently uninvaded geographic areas on multiple continents harbor highly suitable habitats for parthenium. However, discrepancies between model predictions and parthenium invasion in Australia indicate successful management for this globally significant weed.