* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download D3.3 Annex 1 review pest risk mapping best practice
Climate engineering wikipedia , lookup
Climate governance wikipedia , lookup
Climate change and agriculture wikipedia , lookup
Climatic Research Unit documents wikipedia , lookup
Attribution of recent climate change wikipedia , lookup
Climate sensitivity wikipedia , lookup
Media coverage of global warming wikipedia , lookup
Solar radiation management wikipedia , lookup
Scientific opinion on climate change wikipedia , lookup
Public opinion on global warming wikipedia , lookup
Global Energy and Water Cycle Experiment wikipedia , lookup
Atmospheric model wikipedia , lookup
IPCC Fourth Assessment Report wikipedia , lookup
Effects of global warming on humans wikipedia , lookup
Climate change and poverty wikipedia , lookup
Surveys of scientists' views on climate change wikipedia , lookup
Years of Living Dangerously wikipedia , lookup
ENHANCEMENTS OF PEST RISK ANALYSIS TECHNIQUES Milestone 3.5 Best practice for mapping endangered areas identified Author(s): Richard Baker, Christelle Robinet, David Makowski, Alain Roques, Sylvie Augustin, Sarah Brunel, Philippe Reynaud, Maxime Dupin, Darren Kriticos, Vojtech Jarosik, Sue Worner Partner(s): Fera, INRA, EPPO, LNPV, CRCNPB, IBOT, Bio-Protection Submission date: November 2008 EU Framework 7 Research Project Enhancements of Pest Risk Analysis Techniques (Grant Agreement No. 212459) PRATIQUE No. 212459 Deliverable number: Date: DD/MM/YYYY _____________________________________________________________________ PROJECT OVERVIEW: PRATIQUE is an EC-funded 7th Framework research project designed to address the major challenges for pest risk analysis (PRA) in Europe. It has three principal objectives: (i) to assemble the datasets required to construct PRAs valid for the whole of the EU, (ii) to conduct multi-disciplinary research that enhances the techniques used in PRA and (iii) to provide a decision support scheme for PRA that is efficient and userfriendly. For further information please visit the project website or e-mail the project office using the details provided below: Email: [email protected] Internet: www.pratiqueproject.eu Authors of this report and contact details Name: Richard Baker Partner: Fera E-mail: [email protected] Name: Christelle Robinet Partner: INRA E-mail: [email protected] Name: Christelle Robinet Partner: INRA E-mail: [email protected] Name: Alain Roques Partner: INRA E-mail: [email protected] Name: Sylvie Augustin Partner: INRA E-mail: [email protected] Name: Sarah Brunel Partner: EPPO E-mail: [email protected] Name: Philippe Reynaud Partner: CIRAD (LNPV) E-mail: [email protected] Name: Maxime Dupin Partner: CIRAD (LNPV) E-mail: [email protected] Name: Darren Kriticos Partner: CRCNPB E-mail: [email protected] Name: Vojtĕch Jarošík Partner: IBOT E-mail: [email protected] Name: Sue Worner Partner: Bioprotection E-mail: [email protected] Page 2 of 30 Disclaimer: This publication has been funded under the small collaborative project PRATIQUE, an EU 7th Framework Programme for Research, Technological Development and Demonstration addressing theme: [kbbe-2007-1-2-03: development of more efficient risk analysis techniques for pests and pathogens of phytosanitary concern call: fp7- kbbe-2007-1]. Its content does not represent the official position of the European Commission and is entirely under the responsibility of the authors. The information in this document is provided as is and no guarantee or warranty is given that the information is fit for any particular purpose. The user thereof uses the information at its sole risk and liability. PRATIQUE No. 212459 Deliverable number: Date: DD/MM/YYYY _____________________________________________________________________ Milestone 3.5 Best practice for mapping endangered areas identified (month 6) (Subtask 3.3.1) CONTENTS 1. Task objective 2. Best practice on mapping endangered areas from existing PRA standards and schemes 3. Best practice on mapping endangered areas from the literature (including other EU projects) 4. Key issues to address 5. Conclusions Annex 1: Profile of risk mapping software packages Annex 2: Mapping the potential distribution of insect species: what is the best practice? 1. Task Objective Text from the Description of Work: Best practice worldwide for mapping endangered areas under current and future climates will be determined by a review of PRA schemes and the literature. Particular importance will be given to evaluating the work undertaken by EU funded projects, e.g. ALARM, that have assembled and mapped relevant datasets. CSL, CIRAD, Bio-Protection, CRCNPB, INRA, JKI, IBOT and UPAD will undertake this work. 2. Best practice on mapping endangered areas from existing PRA standards and schemes 2.1 Defining endangered areas in International Standards The following text is the only guidance given on defining endangered areas in the International Standards for Phytosanitary Measures (ISPMs): ISPM5 (IPPC, 2006) Endangered Area: An area where ecological factors favour the establishment of a pest whose presence in the area will result in economically important loss [FAO, 1995] ISPM11 (IPPC, 2004) 2.2.4.1 Conclusion regarding endangered areas The part of the PRA area where ecological factors favour the establishment of the pest should be identified in order to define the endangered area. This may be the whole of the PRA area or a part of the area. Page 2 of 30 2.3.3.1 Endangered area The part of the PRA area where presence of the pest will result in economically important loss should be identified as appropriate. This is needed to define the endangered area. It is clearly describing a two step process to define the area: (i) where the organism can establish based on ecological factors and (ii) where the organism will cause economic damage Like all ISPMs it does not describe or recommend the particular methods that should be used. Even the term “mapping” is not included. However in paragraph 2.2.2, it does state that: “Climatic modelling systems may be used to compare climatic data on the known distribution of a pest with that in the PRA area.” 2.2 Defining endangered areas in the EPPO Risk Analysis Scheme The following text is in the current EPPO PRA scheme (EPPO, 2007): EPPO PRA SCHEME Conclusion regarding endangered areas 1.35 Based on the answers to questions 1.16 to 1.34 identify the part of the PRA area where presence of host plants or suitable habitats and ecological factors favour the establishment and spread of the pest to define the endangered area. Note: The PRA area may be the whole EPPO region or part of it. The endangered area may be the whole of the PRA area, or part or parts of the area (i.e. the whole EPPO region or whole or part of several countries of the EPPO region). It can be defined ecoclimatically, geographically, by crop or by production system (e.g. protected cultivation such as glasshouses) or by types of ecosystems. 2.16 Referring back to the conclusion on endangered area (1.35), identify the parts of the PRA area where the pest can establish and which are economically most at risk. This text mirrors that in ISPM11. It states that the endangered area can be defined by, e.g.: Ecoclimatic zones Geographic area Crop distribution Production systems (e.g. protected cultivation) Ecosystem However, like ISPM11, it does not describe or recommend any particular method that should be used. Pragmatically, because no clear methods are available for defining the endangered areas where economically important loss will occur and there may not be agreement, EPPO has suggested (EPPO, 2007) that: “The EPPO PRA process should identify the endangered area whenever possible, i.e. when there is a consensus that the presence of the pest in the area may result in economically important loss. When there is no consensus that the presence of the pest in the area may result in economically important loss, the delimitation of the endangered area is not possible at the EPPO level. In this case only the area of potential establishment can be determined.” 2.3 Defining endangered areas in other PRA Schemes Australia: no specific guidance given New Zealand: no specific guidance given Canada: no specific guidance given beyond what is already given in the ISPMs (Karen Castro, personal communication 22nd August, 2008) USA: No specific guidance is provided in the PRA scheme but detailed risk maps with detailed instructions are available particularly as output from NAPPFAST that combines pest (degree day) and disease models (generic infection model) with host distributions to map pest risk in the USA (Magarey et al., 2007). Other schemes: no risk mapping guidance given 2.4. Conclusions on best practice in defining endangered areas in International Standards and PRA Schemes Although ISPM11 states that the endangered area should be based on where (a) ecological factors favour the establishment of the pest and (b) the presence of the pest will result in economically important loss, existing PRA schemes provide little or no guidance on how this should be done. This is primarily because risk mapping tends to be confined to detailed PRAs, undertaken, for example, to combat a specific new threat, to determine whether expensive/stringent phytosanitary measures are justified or to respond to legal/trade challenges. As such, best practice in mapping endangered areas can generally only be inferred by examining these detailed PRAs. Examples: Phytophthora ramorum (RAPRA EU project in prep; NAPPFAST USA; Canada) Tilletia indica (Karnal Bunt Risks EU project; USA) Diabrotica virgifera virgifera (DIABRACT EU project; UK) Anoplophora glabripennis (Europe) Eichornia crassipes (EPPO) Features in common: Emphasis on identifying areas of climatic suitability Some mapping of host distribution 3. Best practice in mapping endangered areas from the literature (including other EU projects) 3.1 Mapping endangered areas based on ecological (including human) factors Ecological and human factors that can be mapped include (EPPO PRA scheme question in brackets): Abiotic o Climate (1.19) o Soils (1.20) o Pollution (1.20) o Topography (1.20) o Aquatic factors (1.20) Biotic o Hosts (1.17) o Habitats (1.17) o Alternate hosts, vectors, root symbionts, pollinators, seed dispersers etc (1.18) o Competitors (1.22) o Natural enemies (1.23) Human o Entry points (1.13) o Cultivation/management practices (1.24) o Commodity, conveyance, human movement (1.33) 3.1.1 Climate Modelling and mapping the climatic factors suitable for establishment is generally the most frequently attempted. Magarey et al. (2007) list 18 methods (see Annex 1). These can be loosely divided into: Climatic mapping based on Deterministic Models, e.g: o Degree Day and phenology model mapping o NAPPFAST (generic pest/pathogen models with interpolated weather data from ZedX) Climatic mapping based on Inductive Techniques, e.g: o Climate envelope models o BioSim o MaxEnt o CLIMEX: Match Climates Climatic mapping based on Combined Techniques, e.g: o CLIMEX: Compare locations A detailed review of the different climatic risk mapping methods has been undertaken by Christelle Robinet, Alain Roques and Sylvie Augustin for this project (see Annex 2). Descriptions of best practice for applying and interpreting CLIMEX are provided in the “EPPO Instructions for the Use and Interpretation of CLIMEX” (EPPO, 2007; 2008 in prep) and in the CLIMEX manual (Sutherst et al., 2007) 3.1.2 Hosts and habitats In general, other factors, such as suitable hosts/habitats, soils etc are simply overlaid on the map of suitable climate. If displayed in a GIS, masking techniques can be used so that only the areas where suitable climate, soils, hosts/habitats etc are present are displayed. The principal difficulties arise when the datasets are at different spatial and temporal resolutions and collected at different time periods. Upscaling and downscaling methods exist for resolving resolution issues. Only very rarely is there sufficient information to map factors other than the distribution of suitable hosts and habitats. Robinet et al (Annex 2, Part 4) summarise the issues in part 4 of their review. 3.2 Mapping endangered areas based on economic, environmental and social impacts The economic, environmental and social factors that can be mapped include (EPPO PRA scheme question in brackets): Crop area (2.2) Crop value (2.2) Vulnerable species, habitats and ecosystems (2.7) Vulnerable people and communities (2.9) 3.2.1 Mapping the risk of plant invasions in Europe based on habitat invasibility A specific approach to mapping the risk from invasions by alien plants was recently undertaken (Chytrý et al. 2009) with funding from the EU ALARM project. This approach focuses not on individual species but on the overal invasion load in European habitats. Recent studies analysing plots used for sampling vegetation at the scale of tens to hundreds of square metres have demonstrated that habitats differ considerably in their invasibility. The differences in the level of invasion (expressed as the proportion of alien to all species in the plot) between Central European habitats are mainly caused by inherent habitat properties, and to a lesser extent by propagule pressure and climatic differences between regions. Therefore, habitat type is a good predictor of the level of plant invasion (Chytrý et al. 2008a). It has been also shown that patterns of habitat invasion are consistent among European regions with contrasting climates, biogeographical affinities, history and socio-economic background (Chytrý et al. 2008b). These findings provided a solid background for mapping the level of plant invasion, based on the projection of the habitat-specific levels of invasion onto land-cover maps. More than 50,000 vegetation plots were classified in EUNIS habitat categories and used to quantify levels of invasion for each habitat. The spatially non-explicit EUNIS based data were transformed into the spatially explicit CORINE land-cover classes, based on estimated proportion of each of EUNIS habitat types in CORINE land-cover classes, and the level of invasion was calculated for each class. Sampling was done in three European regions, representing Mediterranean, temperate and Atlantic climate and extrapolated to other regions on the basis of climatic similarities. The resulting map reflects the risk from invasions by alien plants at the European scale (Chytrý et al. 2009). An approach linking large sets of spatially explicit data from vegetation survey plots can produce robust information on macroecological patterns of plant invasions. Spatially explicit information on habitat invasions can be used to identify the areas of highest risk of invasion so as to support effective monitoring and management of alien plants; combined with scenarios of future land-use change, it may also be used for prediction of invasion risks in the future (Pyšek et al., 2008). 3.2.2 Mapping other impact factors A preliminary review of the literature has not provided good examples. Some maps, e.g. of D. virgifera virgifera in the USA (http://www.entm.purdue.edu/wcr/ ) and Europe, are based on known impacts. Others are based on maps of highly suitable ecological conditions (see above). At the pest risk mapping workshop in Minneapolis, Richard Baker summarised the challenge of mapping potential impacts as follows: o Predicting establishment endangered areas based on climatic suitability and host/habitat range is already very difficult o Predicting spread very challenging even using diffusion models o Modelling population dynamics in relationship to an economic injury level even more difficult o Can we assign some a priori vulnerability index for economic, environmental social receptors? o How do we take time and climate/landuse change into account? The difference between the establishment and impacts endangered areas can be seen in this table: Topic Establishment Spread Population density, inoculum level Key factors Establishment Endangered Area Possible Not necessarily Sufficient to maintain presence Suitable climate, available hosts/habitats Impacts Endangered Area Very likely Very likely Above economic injury level Very suitable climate, many hosts/habitats, vulnerable receptors of high value Some possible ways forward include: o Splitting establishment endangered area into grid cells and assessing spread, population dynamics and impact vulnerabilities for each cell o Assuming most species can be spread rapidly and long distances by man and assigning risk by distance from ports, nurseries, habitation, existing outbreaks as appropriate o Using climatic suitability indices, e.g. growth/ecoclimatic indices, degree days, generation time, generation number and generic infection index as a surrogate for population/inoculum density. o Determining the relative vulnerablity of receptors (by value, size, rarity, control efficacy etc) o Estimating the change in impacts over time (shape of the curve) rather than an overall value o Studying historic invasions The most vulnerable crop types/areas are, for example, those with: o Favoured host status o Especially high value, e.g. seed potatoes o Very high quality standards, e.g. dessert fruit o Long replacement time, e.g. timber & top fruit trees o Pest friendly management practices, e.g. no rotation for D. virgifera virgifera o High vector densities o Significant proportion of national production o Significant proportion of the export market o Heritage varieties o Organic status and/or biological control systems o No effective control methods available The most vulnerable environmental receptors include: o Keystone species o Rare and endemic species o Nature reserves and special areas of conservation under, e.g. the EC Habitats Directive o Islands and other isolated habitats o High amenity value o Important ecosystem services 4. Key issues to address Limitations of current methods Data requirements Representing current and future climate change and land use Handling, displaying and communicating uncertainty No specific guidance on best practice in pest risk mapping exists. This is one of the reasons for holding the first international Pest Risk Mapping Workshop in June 2007 (Magarey et al., 2007). This meeting identified the following ten critical issues in building risk models and creating risk maps, ranked as follows: 1. Model assessment, validation and documentation 2. Map representation and visualization of uncertainty 3. Availability and accessibility of primary data 4. Best practice guide for modeling (including toolkit)* 5. Communication, interpretation and use of risk maps by decision-makers* 6. Impact mapping 7. International/online collaboration* 8. Climate change 9. Gap in how human and biological dimensions interact 10. Training in modelling practice* The issues marked with an asterisk are primarily organisational issues. While modelling and risk mapping are taken together, clearly, a best practice guide for mapping endangered areas must cover both aspects. In PRATIQUE, issue 3 is being dealt with by WP1. 5. Conclusions on best practice in pest risk mapping from existing schemes and the literature There is no existing guide to best practice in mapping endangered areas. Currently, best practice can only be inferred by analysis of the different methods used and examples available. This issue was discussed at the international pest risk mapping workshop in September 2008 and coordinated approach will be adopted with contributions from the USA, Canada, Australia, New Zealand and Europe. The European component will be provided by PRATIQUE. It is clear that mapping and spatial analysis in general could play a much greater role in all sections of pest risk analysis since almost all the questions asked have a spatial component and rely on datasets with a spatial reference. Maps communicate risk in a much more direct and understandable manner than any risk rating method. Its therefore vital that best practice is described and followed by all. In some areas, further work needs to be undertaken to determine best practice. These include: The most appropriate methods for mapping endangered areas for species with poorly known distributions and/or climatic responses. Guidance on the choice of models in particular situations Mapping pathogens with complex life cycles Taking climate change into account Mapping economic, environmental and social impacts (impact endangered areas) Communicating uncertainty in risk maps References Baker, R.H.A. 2002. Predicting the limits to the potential distribution of alien crop pests. In: Invasive Arthropods in Agriculture. Problems and Solutions, Hallman, G.J. & Schwalbe, C.P. (Eds). pp. 207-241. Science Publishers Inc. Enfield USA. Baker, R.H.A., Brunel, S., MacLeod, A. & Kriticos, D. J. 2007. Instructions for the use and interpretation of CLIMEX. Draft EPPO Document: 07-13301. Chytrý M., Jarošík V., Pyšek P., Hájek O., Knollová I., Tichý L. & Danihelka J. 2008b. Separating habitat invasibility by alien plants from the actual level of invasion. Ecology 89: 1541–1553. Chytrý M., Maskell L., Pino J., Pyšek P., Vila M., Font X. & Smart S. 2008a. Habitat invasions by alien plants: a quantitative comparison between Mediterranean, subcontinental and oceanic regions of Europe. Journal of Applied Ecology 45: 448–458. Chytrý M., Pyšek P., Wild J., Maskell L. C., Pino J. & Vilà M. 2009. European map of alien plant invasions, based on the quantitative assessment across habitats. Diversity and Distributions (in press). EPPO. 2007. EPPO PRA process: definition of the "endangered area" and the "area of potential establishment". EPPO Document: 07-13572. Kriticos, D.J. & Randall, R.P. 2001. A comparison of systems to analyse potential weed distributions. Groves, R. H.; Panetta, F. D., and Virtue, J. G., Eds. Weed Risk Assessment. Melbourne, Australia: CSIRO Publishing. pp. 61-79. Magarey, R.D., Kriticos, D.J., Fowler, G.A., Kalaris, T. M., Pitt. J., Baker, R.H.A. & Koch, F. 2007. Report on the APHIS-PPQ-CPHST Workshop on Pest Risk Mapping. June 5-7, 2007 in Fort Collins, Colorado, USA http://www.nappfast.org/ASPRM%20web/ASPRM%20Overview2.doc Magarey, R.D., Fowler, G.A., Borchert, D.M., Sutton, T.B., Colunga-Garcia, M. & Simpson, J.A. NAPPFAST: An internet system for the weather-based mapping of plant pathogens. Plant Disease, 91: 336-345. Pyšek P., Chytrý M. & Jarošík V. 2008. Habitats and land-use as determinants of plant invasions in the temperate zone of Europe. In: Perrings C., Mooney H. A. & Williamson M. (eds.), Bioinvasions and globalization: Ecology, economics, management and policy, Oxford University Press, Oxford (in press). Sutherst, R.W., Maywald, G. F. & Kriticos, D.J. 2007. CLIMEX Version 3. User’s Guide. CSIRO. Hearne Scientific Software Pty Ltd Sutherst, R.W., Maywald, G. F. & Skarratt, D.B. 1995. Predicting insect distributions in a changed climate In: R. Harrington and N. E. Stork (eds), Insects in a changing environment, pp. 59-91.Academic Press, London. Annex 1 Profile of risk mapping software packages (from the APHIS-PPQ-CPHST Workshop on Pest Risk Mapping on June 5-7, 2007 in Fort Collins, Colorado, USA) Package name Objective Model style Artificial Neural Networks (ANN) ANN can identify relationships between the presence and absence of the insect species and climatic variables at different sites, To describe the climatic envelope of a species and to predict its occurrence ANNs are an alternative modeling technique based on machine learning. BioMOD BIOMOD: BIOdiversity Modeling aims to maximize the predictive accuracy of current species distributions and the reliability of future potential distributions using different types of statistical modeling methods. CLIMATE To predict the distribution of an organism based upon climate preferences – mainly weed risk assessment CLIMATE ENVELOPE To predict the potential distribution of species using point data from herbaria or museums BIOCLIM/ ANUCLIM Computer platform Proces s or regress ion Orient ed Reference Various R Gevrey and Worner (2006) Climate patternmatching with minimum bounding rectangle (MBR) PC and UNIX R Nix (1986), Busby (1991) Hutchinson et al. (1996) Biomod computes, for each species and in the same package, the four most widely used modeling techniques in species predictions, namely Generalized Linear Models (GLM), Generalized Additive Models (GAM), Classification and Regression Tree analysis (CART) and Artificial Neural Networks (ANN). Climate patternmatching with choice of several match techniques including MBR and point-topoint similarity indices (Gower 1971) Climate patternmatching using MBR Unknown R Thuiller (2003) Apple Macintosh/PC R Pheloung (1996) Web (UNIX) R Boston & Stockwell (1994) CLIMEX for Windows (Compare locations) To compare locations or match climates To predict the relative climatic suitability for a species at selected locations (Match climates) To predict the relative climatic similarity between different locations DOMAIN Conservation ecology, assessing adequacy of reserve design and designing sampling strategies The ENFA’s principle is to compare the distributions of the EGV between the presence data set (species distribution) and the whole area (global distribution). ENFA (Environmental Niche Factor Analysis) FloraMap GARP GLIM/GAM FloraMap is a specialized computer program (and associated data) that was developed to map the predicted distribution, or areas of possible climatic adaptation, of organisms in the wild. To predict the potential distribution of species using point data from herbaria or museums using climatic and nonclimatic data To predict the probability of occurrence of species on a fine scale based upon statistical regression models Process-oriented model describing species response to climatic variables, and predicting climatic suitability. Climate patternmatching procedure Windows 2000, XP P Sutherst et al 2007 R Sutherst et al 2007 Climate patternmatching using a point-to-point similarity index Windows 95/NT R Carpenter et al (1993); CIFOR (1996) The EcologicalNiche Factor Analysis (ENFA) computes suitability functions by comparing the species distribution in the ecogeographical variables (EGV) space with that of the whole set of cells using a multivariate approach. Principal components analysis of monthly climate data using multivariate and Fourier transformation techniques Part of Biomapper software, Windows (most versions) R Hirzel et al. 2002 Windows R Jones and Gladkov (1999) Generates environmentdescription rules using machinelearning techniques General statistical procedure for fitting species response functions to survey data Web (UNIX) R Boston & Stockwell (1994) Not applicable R Austin & Meyers (1996) GRASP HABITAT MaxEnt A regression modeling is used to establish relationships between a response variable and a set of spatial predictors To tightly define the environmental envelope of a species or other biotic entity and to predict the environments in which it may be present To predict species distribution NAPPFAST A tool for phytosanitary risk mapping. Regression Tree Analysis A general statistical procedure to analyse the environmental correlates of species distributions STASH To describe the present and natural distribution of northern Europe's major tree species Generalized Regression analysis and Spatial Prediction Creates a convex polytope in n-dimensional space MS Windows PC R Lehman et al. 2002 PC R Walker & Cocks (1991) Machine learning technique based on the distribution of maximum entropy On-line templates for phenology, infection and empirical models. Simple climate matching tool General statistical procedure for defining set membership based upon environmental correlates Process-oriented model describing species response to climatic variables, and predicting climatic suitability Java based R Phillips et al. 2006 Internet explorer P or R Magarey et el. 2007 Not applicable R UNIX; though could be run on any system running FORTRAN P Sykes et al. (1996) Annex 2: Mapping the potential distribution of insect species: what is the best practice? Christelle Robinet, Alain Roques & Sylvie Augustin INRA, UR633 Zoologie Forestière, F-45166 Olivet, France 1- The use of distribution models Understanding the impact of climate change on species distribution has become one of the major challenges nowadays among ecologists. Field and lab experiments can be conducted to determine the climatic range a particular species, but the ultimate approach consists in developing models and determining bioclimatic envelopes or climate response surfaces. Some studies about the effects of climate change on the species distribution analyses the projection of this envelope in the future. However, some questions should be addressed first: how reliable are these results? How can we select the best model? Some basic choices should be made to develop a model (Beaumont et al. 2007): a) choice of the bioclimatic model (for instance CLIMEX, GARP, GAM,… see Manel et al. 1999, Guisan & Zimmermann 2000, Kriticos & Randall 2001, Guisan & Thuiller 2005, Elith et al. 2006, Lawler et al. 2006, Pearson et al. 2006); b) choice of predictor variables (e.g. temperature, precipitation or non climatic factors such as land cover, see Peterson & Cohoon 1999, Beaumont et al 2005, Heikkinen et al 2006), and c) choice of the climate scenario (including the emission scenario, the climate model or even an idealized scenario, e.g.: +3°C) (Meehl et al. 2007). Bioclimatic models are the primary tools for simulating the impact of climate change on species distribution (Beaumont et al 2007). Uncertainty in the climate predictions can result from: (1) uncertainty in the climate scenario (to overcome this problem, we should test many scenarios); (2) variability in each climate scenario (we should run each climate model multiple times). Generally only the first uncertainty is considered but Beaumont et al (2007) has clearly proved that the second uncertainty could be even greater and affect considerable the predicted distribution of nine Australian butterfly species when using BIOCLIM model. Several climate scenarios and several realizations of each scenario are required to determine the range of projected distribution in the future. Statistical techniques used to determine the climate envelope tend to select arbitrary meteorological variables that are not necessarily associated with biological processes. Temporal resolution of these variables might be also arbitrary, eg. temperature of the coldest month (Zalucki & Furlong 2005). The nature of occurrence data may also be important: presence-absence models are more accurate than presence-only models (Elith et al. 2006, Beaumont et al. 2007). In case of species interactions, simple presence-absence datasets are not exhaustive enough, and occurrence data of each species under different environmental conditions are required. The best approach is generally to compare different models and test different scenarios to obtain a confidence range of the potential distribution, but it is unfortunately not often the case in most of studies and the choice for a certain model is not always justified in papers, especially about insect species. Few studies really assess the performance of the models developed and even fewer assess the difference in performance of different models. Even in that case, there is generally a strong bias because authors initially support a particular model, and thus analysis is not completely objective. Across various taxa, there are some good comparisons (e.g. Kriticos & Randall 2001, Elith et al. 2006), but across insect species or more generally terrestrial invertebrate species they are not frequent. Different indexes of model performance can be calculated for validation. First we can summarize this performance in a confusion matrix (Table 1), with a the number of true positive values, b false positives, c false negatives, and d true negatives. Based on these values, we can then derive other performance measures such as: correct classification rate, misclassification rate, sensitivity, specificity, odds ratio, positive and negative predictive power, normalized mutual information statistic, and Cohen’s kappa (Fielding & Bell 1997; Manel et al. 2001). ACTUAL + PREDICTED Table 1. Confusion matrix. + a c b d In addition to the choice of the index performance, there is also a choice about the dataset. Although the same dataset can be used for calibration and validation of the model (method called ‘resubstitution’), it is generally preferable to use independent datasets (Sutherst & Maywald 1985). Since it is not common to have several independent datasets (e.g. occurrence data in different regions), the data should usually be partitioned into a calibration dataset and a validation dataset (Fielding & Bell 1997, Araújo et al. 2005, Ulrichs & Hopper 2008). Also, one data point can be discarded in the calibration dataset and the remaining dataset sequentially used for the calibration dataset; this method is called a jackknife sampling or LeaveOne-Out. Despite all these available methods to evaluate the success of a species distribution mapping, this validation part is generally neglected and only few studies on insect species actually provide a comparison among different models (e.g. Ward 2007) because papers focus more on the interpretation of the projection results. Here we review insect case studies for which a climate response surface has been developed in order to determine the best practice(s) for mapping the potential distribution. Insect species are particularly interesting because distribution of most poikilothermic animals is determined by climate (Andrewartha & Birch 1954, 1984). Although we focussed on insect species, we sometimes refer to other studies in the discussion that we believe bring a general point of view on this subject and thus could be interesting also for mapping the potential range of insect species. In this case, we clearly mention that the reference concerns other species. In a second part, we review the models developed for mapping the distribution of the pine processionary moth. Performance of bioclimate envelope models is generally criticized because of three reasons: they ignore biotic interactions, evolutionary changes, and dispersal abilities (Davis et al. 1998; Pearson & Dawson 2003). There is also an assumption rarely verified: the species’ distribution is supposed to be relatively stable and in equilibrium with its surrounding environment (Sutherst & Maywald 2005 is one exception). In case this assumption is not verified, stochastic dynamic models should be used instead of response surfaces (Guisan & Zimmermann 2000). There is also a debate about whether spatial distribution models define a species fundamental niche or realized niche (Kriticos et al. 2007). Guisan & Thuiller (2005) state that most of the literature assumes, without proper evidence, that spatial models represent the realized niche of the species, because their observed distributions are already constrained by biotic interactions and limiting resources. 2- Review of case studies Review description We have reviewed 53 papers (grouped into 48 case studies) dealing with modelling the potential distribution of insect species, published between 1985 and 2008. Table 2 is a synthetic summary of the papers reviewed. This is not an exhaustive list of papers dealing with climate response surface. For instance, for the Climex model, you can obtain a non-exhaustive but more complete list in the 2007 user’s guide (Sutherst et al. 2007b). We focussed on the more recent studies when similar works have been published previously by the same authors because we aim to examine the performance of the current methods. Most of insects studied were biological control agents. Others were mostly invasive species (or potentially invasive). This review reports few bioclimatic model developed on endangered species. There is only one case for which the species was neither studied for biological control or invasion risk, but rather in terms of conservation because of a low vagility (Stockman et al. 2006) in order to evaluate the performance models for a species whose distribution was not completely known. Various models have been used: ANN, BIOCLIM, CLIMEX, discriminant function, DOMAIN, GAM, GARP, GLM, logistic regression, MAXENT (reviewed by Kriticos and Randall 2001), but also some specific mechanistic models: diapause model, ecophysiological model, life stage model and phenological model. Proportion of classic statistical models represents 80% against 20% for specific models. Only 19% of the papers effectively compared (at least two) different models. Surprisingly, most of the studies aim to determine the potential range distribution of an invasive species but few of them (only 35%) try to determine or include the effects of the climate change on the overall potential distribution. Analysis and interpretation Based on the papers reviewed (table 2), we aim to present a synthetic view on the performance of the main models employed. BIOCLIM is a correlative model often used to determine the effects of climate change (Beaumont & Hughes 2002, Beaumont et al. 2007). Despite its usefulness, it seems that other models such as DOMAIN and MAXENT perform better (Ward 2007). Predictions of the CLIMEX model were quite successful (except for one case, van Klinken et al. 2003), and the discrepancies could be explained by non climatic factors (e.g. cattle resistance, Sutherst & Maywald 1985). Performance is usually based on a visual inspection. See the next section for a deeper analysis of the CLIMEX predictions. GARP is a genetic algorithm for rule-set prediction, derived from ecological niches of species that has dispersal capabilities, evaluating correlations between distributional occurrences and environmental characteristics. Due to stochastic elements in this algorithm, subsequent runs using the same data will produce slightly different results. There are three problems: (1) a “black box” method: we cannot explore the role of each predictor individually, (2) goodness-of-fit is seldom checked with field samples, (3) problems with spatial resolution and selection of environmental layers (Stockman et al. 2006). Other models such as BIOCLIM and GLM seem to perform better. The logistic regression performed a little better than the linear discriminant analysis (Cumming 2000), and the discriminant analysis was as successful as the life development stage model (Hunter & Lindgren 1995). The logistic regression was particularly efficient with a predictive variable connected to the habitat suitability (Hill et al. 1999, Warren et al. 2001). Ecophysiological models, phenological models and life stage models are quite successful and probably more robust than statistical models because they reflect the underlying mechanism and they are not based on a correlation that may change with climate warming, environmental change or more generally global change. Nevertheless, rigorous comparisons between statistical and mechanistic models are needed. Sometimes, climate is not the main limiting factor and, in these cases, predicting the species potential distribution is very difficult (eg. Samways et al. 1999). Other factors can affect the species distribution such as a localized response to microclimate, host type and availability, presence of natural enemies. 3- The most frequent model The CLIMEX model appears to be an effective tool and the most popular method for predicting the potential distribution of poikilotherm species. This single model represents approximately 50% of the reviewed papers on insect species. CLIMEX is generally used to predict the suitability of a region for a species based on long-term average climate, but it is also possible to determine the suitability of a site over the years (e.g.: Zalucki & Furlong 2005). This approach is very useful to determine the effects of extreme variation of climatic variables on the species abundance and distribution. This computer program includes three separate modules: match climates, compare years and compare locations. It combines both simulation modelling and inference approach to determine the species’ response to climate. Growth and stress indices (Table 3) are derived from weekly meteorological data, and an ecoclimatic index EI is calculated ranging from 0 (if unsuitable area) to 100 (if optimal conditions). Parameter values can be estimated directly by the model based upon observations only, but they can also be estimated independently using physiological data even if the CLIMEX model is particularly valuable when too little biological data is available and the native distribution well known. This method is undoubtedly the most frequent in mapping the range of insect species. Validation of this method is however not performed in details in many cases since it is considered as one of the most efficient models for insect species. Only visual inspection is often reported (e.g. Sutherst & Maywald 2005; Poutsma et al. 2008), but an automated parameter fitting procedure have been recently developed (Sutherst et al. 2007b). In the papers reviewed here, some limitations of the CLIMEX model have been reported: (1) the distribution of weather stations greatly affects the output of the model; (2) natural weather variability and extreme climatic conditions may affect the species distribution but only mean parameters’ value over 30 years are generally considered; (3) the model ignores microclimates around rivers or irrigated areas; (4) the model assumes that the species distribution is only determined by climate; and (5) parameters are adjusted following an iteration procedure and results can be easily manipulated (Sutherst & Maywald 1985, Worner 1988, Scott 1992, Davis 1998, Baker et al. 2000, Poutsma et al. 2008). Nevertheless these drawbacks could be easily minimized when compared to other models because: (1) distribution of weather stations will affect any climate response surface and, when available, additional weather data can be included in the CLIMEX model; (2) understanding the effects of extreme climatic conditions has become one of the major challenge in the future, whatever model is used; (3) improvements have been made and, for instance, Sutherst et al. (2007a) considered irrigation as a predictor variable; (4) CLIMEX projection is a first step of a more detailed and realistic model: other layers such as resource distribution and other processes such as competition should be considered. It is always necessary to identify non climatic factors that could explain the species occurrence (Sutherst 2003). In fact, discrepancies between observed and predicted distribution can help to identify these limiting factors (Sutherst & Maywald 1985). Quite recently, Sutherst et al. (2007ab) succeeded in including species traits directly in the CLIMEX model. They have investigated the simultaneous effects of both climate warming and interaction among species, and found that effects of species interaction could even exceed effects of climate change. Menéndez et al. (2008) also gave evidence that, in the range expansion, individuals could escape natural enemies and thus the limit of the distribution could shift more rapidly than previously thought. However, climate alone is usually a significant driver of the species distribution (Hodkinson 1999) and the CLIMEX reliability seems more closely related to the data quality used in the model than in the nature of the model itself (Sutherst & Maywald 1985). (5) Objective approaches were not available to estimate the parameter values in the past, furthermore small changes in parameter values do not change considerably the model outputs (Worner 1988). In the last version of CLIMEX, an automating fitting procedure has been implemented via a genetic algorithm (Sutherst et al. 2007b). CLIMEX not only allows mapping the climatic suitability for a certain species, but also allows broader applications: Peacock & Worner (2006) have determined analogous climates of a certain location (Auckland, New-Zealand) in order to identify potential sources of new invasive insect species. 4- Consideration of other factors Dispersal There are some attempts to consider in a simplistic way the dispersal ability in addition of the response climate surface (e.g. unlimited dispersal, contiguous dispersal or no dispersal, Peterson et al. 2002). Outside insect species, Araújo et al (2006) found contrasting effects with or without dispersal: most amphibians and reptiles could extend their distribution with climate change in case of unlimited dispersal ability, but will probably loose range in case of no dispersal ability. Possibility of the insect species to track the climate change also depends on the dispersal ability of the larval host plant (Araújo & Luoto 2007). Considering even simple hypotheses can help us to understand better whether the species would be able to track the climate change or not. Based on climate envelopes, and various climate and dispersal scenarios, Thomas et al. (2004) proposed three methods to calculate the proportion of species committed to extinction as a function of estimated area lost. It seems that lifehistory traits could help us to determine whether a species is able to track the climate change. For instance, Jiguet et al. (2007) found that natal dispersal but also annual fecundity and the number of generations per year could inform about the birds’ sensitivity to a climate change. Dispersal associated with the population dynamics also might be important. If individuals are subject to Allee effects (a reduced population growth rate at low densities), then the species may not be able to track the climate change if their potential distribution is predicted to go through corridors and then enter a large suitable area (Roques et al. 2008) Interactions Importance of interactions is perhaps one of the most discussed questions (see Davis et al. 1998). For herbivore species, studying the interaction with its host tree or host plant is crucial. They should stay in synchrony but climate change may alter differently the phenology of each one and completely disrupt this relationship (van Ash & Visser 2007). Araújo & Luoto (2007) also supported the idea that biotic interactions could be important also at macroecological scales. Effects of species interaction could even exceed effects of climate change (Sutherst et al. 2007ab). However, Huntley et al. (2004) claimed that the performance of climate envelope models did not depend on the taxonomic group nor to trophic levels. Although these models are strongly based on a correlative approach (between the species distribution and climatic factors) and consider individually each species with no interaction, they are probably the best ones to study the effects of the climate change. Authors recalled that most species interactions are generalist and not specialist, and this is maybe the reason why these interactions have few effects on the species distribution. As a result, it seems that biotic interactions drive certainly the species range, but for a first assessment, assuming that climate is the main driving factor seems quite reasonable. Habitat suitability Even though the distribution of many species is likely to expand in response to climate warming, the species may not be able to track the climate change because of other important factors such as habitat suitability. Hill et al (2001) used a spatially explicit mechanistic model called MIGRATE to determine the impact of the landscape structure on the species range expansion, and they have clearly demonstrated that development of such models are important to understand in further details the response of the species to climate warming in heterogeneous habitats. Land transformation derived from the ‘Human Footprint’ can also be considered in addition to common environmental factors generally used to determine the potential distribution (Thuiller et al. 2006, for African mammals). Both climate change and land transformation could affect the current species distribution, but also the community composition (Thuiller et al. 2006, for African mammals), and thus species interaction. Climate variables and spatial scale Climate response surfaces are important to determine a potential effect of a change in climatic factors on the species distribution when such factors primarily govern the species niche. Generally this is the case at large geographical scales where habitat availability, local extinctions and colonisations, and adaptability have minor effects (Berry et al. 2002). For instance, at the European scale, land cover is mostly correlated with climate and it is interesting to include only the variables weakly explained by the climate in the model to improve its performance (Thuiller et al. 2004). When the geographical scale considered is restricted, and the landscape quite fragmented, simple logistic regression on land cover data can determine successfully the species distribution using no weather data as predictor (Cowley et al. 2000). Thus the spatial scale may also interfere with the results. The same argument is true for other factors such as species interaction: for ants, climatic drivers appear at large scale, whereas the species distribution at small scale is mainly driven by microhabitat specialisation and competition (Hölldobler & Wilson 1990). The global scale disturbances of El Niño Southern Oscillation (ENSO) influence insect migration when these disturbances result in exceptional rainfall in semi-arid regions and lead to large populations of migratory insects (Drake & Farrow 1988). Unfortunately to show any relationship between abundance and such predictors, long series abundance data for the species are required (Zalucki & Furlong 2005). Finally, it seems that more realistic dispersal abilities, species interaction and population dynamics should be included in distribution models in the future (Guisan & Thuiller 2005), and we should test whether they bring important supplementary information and increase the model performance or not. 5- Conclusion on this review In conclusion, many distribution models are available but few comparisons have been really made. Generally, modellers are very confident in their model and only show the advantages. Drawbacks are rarely discussed if ever mentioned. For insect species, the most popular model is undoubtedly the CLIMEX model, all the more that improvements have been made recently (Sutherst et al 2007b). CLIMEX was effectively developed for insect species and its drawbacks are reduced with the recent improvements. Ecophysiological models are usually better to determine the underlying mechanism, but they should be calibrated with consistent data. In contrast to statistical models, their goodness-of-fit is probably higher when environmental factors take unusual values. Models based on mechanistic understanding should be more robust than purely statistical models because correlations may vary with climate change (Pearson & Dawson 2003). Although these mechanistic models are particularly appealing for well known and well documented species, they often require too many data to be a general approach (Guisan & Thuiller 2005). The model performance seems to rely mainly on the data quality and the rigor of the method employed for any model. Observing some discrepancies can help to identify a missing factor, and then improve the model performance. 6- Case study: the pine processionary moth The pine processionary moth is an insect species native to Mediterranean countries. For a few decades, the species has expanded its distribution in higher latitudes and higher elevation (Battisti et al 2005). Model 1: exclusion map Huchon and Démolin (1970) developed the first distribution model for the pine processionary moth. Based on correlative observations not clearly determined, they defined the presence threshold by a combination of the mean minimum temperature in January (TNJ, °C) and the cumulative annual sunshine (S, hours): TNJ ≥ 0°C and S ≥ 1800 0 ≥ TNJ ≥ -4°C and S ≥ 1800 +100*(-TNJ) This means that, outside the optimum area, 100 h of sunshine could compensate 1°C below 0, but it cannot compensate more than 4°C below 0. The weather variables used for this model were averaged from 1946 to 1960. This exclusion model predicted that a large north part of France was unfavourable for the pine processionary moth, and the species occurrence has been in agreement with this exclusion map for many years. Model 2: GAM model Since the pine processionary moth invaded the south of the Paris Basin in the 1990s, it was necessary to update this historical model and introduce the climate warming. Since the cumulative annual sunshine is not a variable commonly used in climate scenarios, we replaced this variable by the annual mean solar radiation (Wh/m²) and used a generalized additive model (GAM). Correct classification rate was 83% for 1970-1980, but absence is incorrectly predicted for the ongoing years, even when considering climate warming (Robinet et al. 2007). Model 3: ecophysiological model To understand in-depth the impact of climate warming on the species survival, lab and field experiments have been conducted and we found that temperature could affect the larval feeding activity and thus the larvae survival. Indeed two conditions should be satisfied for the feeding: temperature inside the nest during the day should reach 9°C, and then during the following night, air temperature should be above 0°C (Battisti et al. 2005). If both conditions area satisfied, then larvae can go out of the nest during the night and feed on the needles. If not, larvae do not feed and, in some cases, starvation occurs. The feeding activity could explain a large part of the colony survival. Therefore, we modelled this mechanism and calculated the mean number of feeding days and the longest period of starvation (Robinet et al. 2007). There was an unfavourable area in the South of the Paris Basin in the 1992-1996, which vanished during 2001-2004. This change in the potential feeding activity can explain the spectacular expansion observed in this region since then. Model 4: diffusion model Since habitat distribution and dispersal ability was not included in the previous model, this ecophysiological model was then simplified and coupled to a growth model and a diffusion model (Robinet 2006; Robinet et al. 2008). A climate scenario (regionalized scenario B2 – scenario ARPEGE-Climat from Météo-France) was also considered. The model was validated at larger temporal and spatial scales. It predicted retraction of the distribution during cold winters in the past and a continuous expansion since the late 1990s. Based on this model and hypotheses (climate scenario, dispersal ability of 3 km/year), the pine processionary moth could reach Paris by 2025. This model describes only the natural range expansion of the population, but quite recently, some isolated colonies have been discovered far from the current distribution, probably inadvertently transported by humans (Robinet et al. 2008). Thus, longdistance dispersal models should now be developed. References Allen, JC, Foltz, JL, Dixon, WN, Liebhold, AM, Colbert, JJ, Régnière, J., Gray, DR, Wilder, JW & Christie, I. (1993) Will the gypsy moth become a pest in Florida? Florida Entomologist, 76, 102-113. Andrewartha, HG & Birch, LC (1954) The distribution and abundance of animals. Chicago Press. Andrewartha, HG & Birch, LC (1984) The ecological web: more on the distribution and abundance of animals. University of Chicago Press, Chicago, 520 pp. Araújo, M.B. and Luoto, M. (2007) The importance of biotic interactions for modelling species' responses to climate change. Global Ecology and Biogeography, 16: 743–753. Araújo, M.B., Pearson, R.G., Thuiller, W., and Erhard, M. (2005) Validation of species-climate impact models under climate change. Global Change Biology, 11: 1504-1513. Araújo, M.B., Thuiller, W & Pearson, RG (2006) Climate warming and the decline of amphibians and reptiles in Europe. Journal of Biogeography, 33, 1712-1728. van Ash, M & Visser, ME (2007) Phenology of forest caterpillars and their host trees: the importance of synchrony. Annu. Rev. Entomol., 52, 37-55. Baker, RHA, Sansford, CE, Jarvis, CH, Cannon, RJC, MacLeod, A & Walters, KFA (2000) The role of climatic mapping in predicting the potential geographical distribution of non-indigenous pests under current and future climates. Agriculture, Ecosystems and Environment, 82, 57-71. Battisti, A, Stastny, M, Netherer, S, Robinet, C, Schopf, A, Roques, A & Larsson, S (2005) Expansion of geographic range in the pine processionary moth caused by increased winter temperatures. Ecological Applications, 15, 2084-2096. Beaumont, L.J. & Hughes, L. (2002) Potential changes in the distributions of latitudinally restricted Australian butterfly species in response to climate change. Global Change Biology, 8, 954-971. Beaumont, LJ, Hughes, L & Poulsen, M (2005) Predicting species distributions: use of climatic parameters in BIOCLIM and its impact on predictions of species’ current and future distributions. Ecological Modelling, 186, 250-269. Beaumont, L.J., Pitman, A.J., Poulsen, M. & Hughes, L. (2007) Where will species go? Incorporating new advances in climate modelling into projections of species distributions. Global Change Biology, 13, 1368-1385. Berry, PM, Dawson, TP, Harrison, PA & Pearson, RG (2002) Modelling potential impacts of climate change on the bioclimatic envelope of species in Britain and Ireland. Global Ecology and Biogeography, 11, 453-462. Cowley, MJR, Wilson, RJ, León-Cortès, JL, Gutiérrez, D Bulman, CR & Thomas, CD (2000) Habitat-based statistical models for predicting the spatial distribution of butterflies and day-flying moths in a fragmented landscape. Journal of Applied Ecology, 37, 60-72. Crozier, L & Dwyer, G (2006) Combining population-dynamic and ecophysiological models to predict climate-induced insect range shifts. The American Naturalist, 167, 853-866. Cumming, GS (2000) Using between-model comparisons to fine-tune linear models of species ranges. Journal of Biogeography, 27, 441-455. D’Adamo, P., Sackmann, P. & Corley, J.C. (2002) The potential distribution of German wasps (Vespula germanica) in Argentina. New Zealand Journal of Zoology, 29, 79-85. Davis, AJ, Lawton, JH, Shorrocks, B & Jenkinson, LS (1998) Individualistic species responses invalidate simple physiological models of community dynamics under global environmental change. Journal of Animal Ecology, 67, 600-612. Drake, VA &Farrow, RA (1988) The influence of atmospheric structure and motions on insect migration. Annual Review of Entomology, 33, 183-210. Elith, J et al. (2006) Novel methods improve prediction of species’ distributions from occurrence data. Ecography, 29, 129-151. Fielding, AH & Bell, JF (1997) A review of methods for the assessment of prediction errors in conservation presence/absence models. Environmental Conservation, 24, 38-49. Fitzpatrick, MC, Weltzin, JF, Sanders, NJ & Dunn, RR (2007) The biogeography of prediction error: why does the introduced range of the fire ant over-predict its native range? Global Ecology and Biogeography, 16, 24-33. Gray, DR, Logan, JA, Ravlin, FW & Carlson, JA (1991) Toward a model of gypsy moth egg phenology: using respiration rates of individual eggs to determine temperature-time requirements of prediapause development. Environmental Entomology, 20, 1645-1652. Gray, D.R. (2004) The gypsy moth life stage model: landscape-wide estimates of gypsy moth establishment using a multi-generational model. Ecological Modelling, 176, 155-171. Gray, DR, Ravlin, FW & Braine, JA (2001) Diapause in the gypsy moth: a model of inhibition and development. Journal of Insect Physiology, 47, 173-184. Guisan, A & Zimmermann, NE (2000) Predictive habitat distribution models in ecology. Ecological Modelling, 135, 147-186. Guisan, A & Thuiller, W (2005) Predicting species distribution: offering more than simple habitat models. Ecology Letters, 8, 993-1009. Heikkinen, RK, Luoto, M & Virkkala, R (2006) Does seasonal fine-tuning of climatic variables improve the performance of bioclimatic envelope models for migratory birds? Diversity and Distributions, 12, 502-510. Heard, TA & Forno, IW (1996) Host selection and host range of the flower-feeding weevil, Coelocephalapion pigrae, a potential biological control agent of Mimosa pigra. Biological Control, 6, 83-95. Hill, JK, Thomas, CD & Huntley, B (1999) Climate and habitat availability determine 20th century changes in a butterfly’s range margin. Proc. R. Soc. Lond. B, 266, 1197-1206. Hill, JK, Collingham, YC, Thomas, CD, Blakeley, DS, Fox, R, Moss, D & Huntley, B (2001) Impacts of landscape structure on butterfly range expansion. Ecology Letters, 4, 313-321. Hodkinson, ID (1999) Species response to global environmental change or why ecophysiological models are important: a reply to Davis et al. Journal of Animal Ecology, 68, 1259-1262. Hölldobler, B & Wilson, EO (1990) The ants. Harvard University Press, Cambridge. Huchon, H & Démolin, G (1970) La bioécologie de la processionnaire du pin. Dispersion potentielle – dispersion actuelle. Revue Forestière Française, spécial ‘La lutte biologique en forêt’, 220-234. Hunter, AF & Lindgren, BS (1995) Range of gypsy moth in British Columbia: a study of climatic suitability. J Entomol Soc Brit Columbia, 92, 45-55. Huntley, B., Green, RE, Collingham, YC, Hill, JK, Willis, SG, Bartlein, PJ, Cramer, W, Hagemeijer, WJM & Thomas CJ (2004) The performance of models relating species geographical distributions to climate is independent of trophic level. Ecology Letters, 7, 417-426. Jarvis, CH & Baker, RHA (2001a) Risk assessment for nonindigeneous pests: 1. Mapping the outputs of phenology models to assess the likelihood of establishment. Diversity and Distributions, 7, 223-235. Jarvis, CH & Baker, RHA (2001b) Risk assessment for nonindigeneous pests: 2. Accounting for interyear climate variability. Diversity and Distributions, 7, 237248. Jiguet, F, Gadot, A-S, Julliard, R, Newson, SE & Couvet, D (2007) Climate envelope, life history traits and the resilience of birds facing global change. Global Change Biology, 13, 1672-1684. Johnson, PC, Mason, DP, Radke, SL & Tracewski, KT (1983) Gypsy moth, Lymantria dispar (L.)(Lepidoptera: Lymantriidae), egg eclosion: degree-day accumulation. Environ. Entomol., 12, 929-932. van Klinken, R.D., Fichera, G. & Cordo, H. (2006) Targeting biological control across diverse landscapes: the release, establishment, and early success of two insects on mesquite (Prosopis spp.) insects in Autralian rangelands. Biological Control, 26, 8-20. Korzukhin MD, Porter SD, Thompson LC & Wiley S (2001) Modeling temperaturedependent range limits for the fire ant Solenopsis invicta (Hymenoptera: Formicidae) in the United-States. Environ Entomol 30:645-655. Kriticos, DJ & Randall, RP (2001) A comparison of systems to analyse potential weed distributions. In: Weed Risk Assessment, eds RH Groves, FD Panetta & JG Panetta, pp. 61-79, CSIRO Publishing, Melbourne. Kriticos, D, Stephens, AEA & Leriche, A (2007) Effects of climate change on oriental fruit fly in New Zealand and the Pacific. New Zealand Plant Protection, 60, 271278. Lawler, JJ, White, D, Neilson, RP & Blaustein, AR (2006) Predicting climate-induced range shifts: model differences and model reliability. Global Change Biology, 12, 1568-1584. Li, H-M, Xiao, H., Peng, H., Han, H-X & Hue, D-Y (2006) Potential global range expansion of a new invasive species, the erythrina gall wap, quadrastichus erythrinae Kim (Insecta : Hymenoptera : eulophidae). The raffles bulletin of zoology, 54, 229-234. MacLeod, A., Evans, H.F. & Baker, R.H.A. (2002) An analysis of pest risk from an Asian longhorn beetle (Anaplophora glabripennis) to hardwood trees in the European community. Crop Protection, 21, 635-645. Manel, S, Dias, JM, Buckton, ST & Ormerod, SJ (1999) Alternative methods for predicting species distribution: an illustration with Himalayan river birds. Journal of Animal Ecology, 36, 734-747. Manel, S, Williams, HC & Ormerod, SJ (2001) Evaluating presence-absence models in ecology: the need to account for prevalence. Mason, PG (2003) Actual and potential distribution of an invasive canola pest, Meligethes viridescens (Coleoptera: Nitidulidae), in Canada. The Canadian Entomologist, 135, 405-413. Matsuki, M., Kay, M., Serin, J., Floyd, R. & Scott, J.K. (2001) Potential risk of accidental introduction of Asian gypsy moth (Lymantria dispar) to Autralasia: effects of climatic conditions and suitability of native plants. Agricultural and Forest Entomology, 3, 305-320. Meehl, G.A., T.F. Stocker, W.D. Collins, P. Friedlingstein, A.T. Gaye, J.M. Gregory, A. Kitoh, R. Knutti, J.M. Murphy, A. Noda, S.C.B. Raper, I.G. Watterson, A.J. Weaver and Z.-C. Zhao ( 2007) Global Climate Projections. In: Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change [Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M. Tignor and H.L. Miller (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA. Menéndez, R, González-Megías, A, Lewis, OT, Shaw, MR & Thomas, CD (2008) Escape from natural enemies during climate-driven range expansion: a case study. Ecological Entomology, 33, 413-421. Morrison, LW, Porter, SD, Daniels, E. & Korzukhin, MD (2004) Potential global range expansion of the invasive fire ant, Solenopsis invicta. Biological Invasions, 6, 183-191. Oberhauser, K. & Peterson, A.T. (2003) Modeling current and future potential wintering distributions of eastern North American monarch butterflies. PNAS, 100, 14063-04068. Olfert, O, Weiss, RM & Wood, S (2004) Potential distribution and relative abundance of an invasive cereal crop pest, Oulema melanopus (Coleoptera: Chrysomelidae), in Canada. The Canadian Entomologist, 136, 277-287. Peacock, L. & Worner, S (2006) Using analogous climates and global insect distribution data to identify potential sources of new invasive insect pests in NewZealand. New-Zealand Journal of Zoology, 33, 414-145. Pearson, RG & Dawson, TP (2003) Predicting the impacts of climate change on the distribution of species: are bioclimate envelope models useful? Global Ecology and Biogeography, 12, 361-371. Pearson, RG, Thuiller, W, Araújo, MB, Martinez-Meyer, E, Brotons, L, McClean, C, Miles, L, Segurado, P, Dawson, TP & Lees, DC (2006) Model-based uncertainty in species range prediction. Journal of Biogeography, 33, 1704-1711. Peterson, AT & Cohoon, KP (1999) Sensitivity of distributional prediction algorithms to geographic data completeness. Ecological Modelling, 117, 159-164. Peterson, AT, Ortega-Huerta, MA, Bartley, J, Sánchez-Cordero, Soberón, J, Buddemeier, RH & Stockwell, RRB (2002) Future projections for Mexican faunas under global climate change scenarios. Nature, 416, 626-629. Poutsma, J, Loomans, AJM, Aukema, B & Heijerman, T (2008) Predicting the potential geographical distribution of the harlequin ladybird, Harmonia axyridis, using the CLIMEX model. BioControl, 53, 103-125. Rafoss, T & Sæthre, M-G (2003) Spatial and temporal distribution of bioclimatic potential for the Codling moth and the Colorado potato beetle in Norway: model predictions versus climate and field data from the 1990s. Agricultural and Forest Entomology, 5, 75-85. Régnière, J & Nealis, V (2002) Modelling seasonality of gypsy moth, Lymantria dispar (Lepidoptera: Lymantriidae), to evaluate probability of its persistence in novel environments. The Canadian Entomologist, 134, 805-824. Robertson, MP, Kriticos, DJ & Zachariades, C. (2008) Climate matching techniques to narrow the search for biological control agents. Biological control, DOI: 10.1016/j.biocontrol.2008.04.002 Robinet, C (2006) Mathematical modelling of invasion processes in ecology : the pine processionary moth as a case study. PhD thesis, École des Hautes Études en Sciences Sociales, Paris (in French with English summary). Robinet, C, Baier, P, Pennerstorfer, J, Schopf, A & Roques, A (2007) Modelling the effects of climate change on the potential feeding activity of Thaumetopoea pityocampa (Den. & Schiff.)(Lep., Notodontidae) in France. Global Ecology and Biogeography, 16, 460-471. Robinet, C, Roques, A, Rousselet, J & Battisti, A (2008) Expansion of pine processionary moth in Europe: patterns and predictive modelling. XXIII International Congress of Entomology, ICE 2008, Section 16: Invasive Species, Symposium 16.3: Insect Invasions and Climate Change. Durban (South Africa), 06-12/07/2008. Roques, L, Roques, A, Berestycki, B & Kretzschmar, A (2008) A population facing climate change: joint influences of Allee effects and environmental boundary geometry. Population Ecology, 50, 215-225. Samways, MJ, Osborn, R, Hastings, H & Hattingh, V (1999) Global climate change and accuracy of prediction of species’ geographical ranges: establishment success of introduced ladybirds (Coccinellidae, Chilocorus spp.) worldwide. Journal of Biogeography, 26, 795-812. Sawyer, Aj, Tauber, MJ, Tauber, CA & Ruberson, JR (1993). Gypsy moth (Lepidoptera: Lymantriidae) egg development: a simulation analysis of laboratory and field data. Ecological Modelling, 66, 121-155. Schweiger, O, Dormann, CF, Bailey, D & Frenzel, M (2006) Occurrence pattern of Parage aegeria (Lepidoptera: Nymphalidae) with respect to local habitat suitability, climate and landscape structure. Landscape ecology, 21, 989-1001. Scott, JK (1992) Biology and climatic requirements of Perapion antiquum (Coleoptera: Apionidae) in southern Africa: implications for the biological control of Emex spp. in Australia. Bulletin of Entomological Research, 82, 399-406. Stephens, AEA, Kriticos, DJ & Leriche, A (2007) The current and future potential geographical distribution of the oriental fruit fly, Bactrocera dorsalis (Diptera: Tephritidae). Bulletin of Entomological Research, 97, 369-378. Stockman, A.K., Beamer, D.A. & Bond, J.E. (2006) An evaluation of a GARP model as an approach to predicting the spatial distribution of non-vagile invertebrate. Diversity and Distribution, 12, 81-89. Sutherst, R.W. (2003) Prediction of species geographical ranges. Journal of Biogeography, 30, 805-816. Sutherst, R.W. & Maywald, G. (1985) A computerised system for matching climates in ecology. Agriculture, Ecosystems and Environment, 13, 281-299. Sutherst, R.W. & Maywald, G. (2005) A climate model of the red imported fire ant, Solenopsis invicta Buren (Hymenoptera: Formicidae): implications for invasion of new regions, particularly oceania. Environ. Entomol., 34, 317-335. Sutherst, RW, Maywald, GF & Bourne, AS (2007a) Including species interactions in risk assessments for global change. Global Change Biology, 13, 1843-1859. Sutherst, RW, Maywald, GF & Kriticos, D (2007b) CLIMEX Version 3: User’s Guide. http://www.Hearne.com.au (last accessed 11 August 2008). Hearne Scientific Software Pty Ltd. 131 pp. Thomas, CD, Cameron, A, Green, RE, Bakkenes, M, Beaumont, LJ, Collingham, YC, Erasmus, BFN, Ferreira de Siqueira, M, Grainger, A, Hannah, L, Hughes, L, Huntley, B, van Jaarsveld, AS, Midgley, GF, Miles, L, Ortega-Huerta, MA, Peterson, AT, Phillips, OL & Williams, SE (2004) Extinction risk from climate change. Nature, 427, 145-148. Thuiller, W, Araújo, MB & Lavorel, S (2004) Do we need land-cover data to model species distributions in Europe? Journal of Biogeography, 31, 353-361. Thuiller, W, Broennimann, O, Hughes, G, Alkemade, JRM, Midgley, GF & Corsi, F (2006) Vulnerability of African mammals to anthropogenic climate change under conservative land transformation assumptions. Global Change Biology, 12, 424440. Ulrichs, C. & Hopper, K.R. (2008) Predicting insect distributions from climate and habitat data. Biocontrol, DOI: 10.1007/s10526-007-9143-8 Ungerer, M.J., Ayres, M.P. & Lombardero, M.J. (1999) Climate and the northern distribution limits of Dendroctonus frontalis Zimmermann (Coleoptera: Scolytidae). Journal of Biogeography, 26, 1133-1145. Venette, RC & Hutchison, WD (1999) Assessing the risk of establishment by pink bollworm (Lepidoptera: Gelechiidae) in the southeastern United-States. Environ. Entomol., 28, 445-455. Venette, RC & Ragsdale, DW (2004) Assessing the Invasion by Soybean Aphid (Homoptera: Aphididae): where will it end? Ann. Entomol. Soc. Am., 97, 219-226. Ward, DF (2007) Modelling the potential geographic distribution of invasive ant species in New Zealand. Biological Invasions, 9, 723-735. Warren, MS, Hill, JK, Thomas, JA, Asher, J, Fox, R, Huntley, B, Roy, DB, Telfer, MG, Jeffcoate, S, Harding, P, Jeffcoate, G, Willis, SG, Greatorex-Davies, JN, Moss, D & Thomas, CD (2001) Rapid responses of British butterflies to opposing forces of climate and habitat change. Nature, 414, 65-69. Wharton, TN & Kriticos, DJ (2004) The fundamental and realized niche of the Monterey Pine aphid, Essigella californica (Essig) (Hemiptera: Aphidae): implications for managing softwood plantations in Australia. Diversity and Distributions, 10, 253-262. Worner, SP (1988) Ecoclimatic assessment of potential establishment of exotic pests. J Econ Entomol, 81, 973-983. Xiong, Y, Chen, J-D, Gu, Z-Y, Wu, X-H, Wan, F-H & Hong, X-Y (2008) The potential suitability of Jiangsu Province, east China for the invasive red imported fire ant, Solenopsis invicta. Biological Invasions, 10, 475-481. Yonow, T & Sutherst, RW (1998) The geographical distribution of the Queensland fruit fly, Bactrocera (Dacus) tryoni, in relation to climate. Australian Journal of Agricultural Research, 49, 935-953. Zalucki, MP & Furlong, MJ (2005) Forecasting Helicoverpa populations in Australia: A comparison of regression based models and a bioclimatic based modelling approach. Insect Science, 12, 45-56.