Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Downscaling Pest Risk Analyses: Identifying Current and Future Potentially Suitable Habitats for Parthenium hysterophorus with Particular Reference to Europe and North Africa

Downscaling Pest Risk Analyses: Identifying Current and Future Potentially Suitable Habitats for Parthenium hysterophorus with Particular Reference to Europe and North Africa

  • Darren J. Kriticos, 
  • Sarah Brunel, 
  • Noboru Ota, 
  • Guillaume Fried, 
  • Alfons G. J. M. Oude Lansink, 
  • F. Dane Panetta, 
  • T. V. Ramachandra Prasad, 
  • Asad Shabbir, 
  • Tuvia Yaacoby


Pest Risk Assessments (PRAs) routinely employ climatic niche models to identify endangered areas. Typically, these models consider only climatic factors, ignoring the ‘Swiss Cheese’ nature of species ranges due to the interplay of climatic and habitat factors. As part of a PRA conducted for the European and Mediterranean Plant Protection Organization, we developed a climatic niche model for Parthenium hysterophorus, explicitly including the effects of irrigation where it was known to be practiced. We then downscaled the climatic risk model using two different methods to identify the suitable habitat types: expert opinion (following the EPPO PRA guidelines) and inferred from the global spatial distribution. The PRA revealed a substantial risk to the EPPO region and Central and Western Africa, highlighting the desirability of avoiding an invasion by P. hysterophorus. We also consider the effects of climate change on the modelled risks. The climate change scenario indicated the risk of substantial further spread of P. hysterophorus in temperate northern hemisphere regions (North America, Europe and the northern Middle East), and also high elevation equatorial regions (Western Brazil, Central Africa, and South East Asia) if minimum temperatures increase substantially. Downscaling the climate model using habitat factors resulted in substantial (approximately 22–53%) reductions in the areas estimated to be endangered. Applying expert assessments as to suitable habitat classes resulted in the greatest reduction in the estimated endangered area, whereas inferring suitable habitats factors from distribution data identified more land use classes and a larger endangered area. Despite some scaling issues with using a globally conformal Land Use Systems dataset, the inferential downscaling method shows promise as a routine addition to the PRA toolkit, as either a direct model component, or simply as a means of better informing an expert assessment of the suitable habitat types.


Whilst the roots of pest risk modelling extend back to early in the 20th Century [1], modern computer-based pest risk modelling has only been practised for some 30 years [2,3]. In that time, there has been a progressive refinement of the spatial distributions of the modelled risks. In the earliest maps, risks were portrayed wherever climate stations were situated [2]. Following the development of climatic splining techniques [4], spatially interpolated results were presented e.g., [5,6]. Increased computing power, and a thirst for more detailed risk maps saw the development of finer-scaled gridded climate datasets [7,8,9], and their application to pest risk modelling problems e.g., [10,11,12].

Under the International Standards for Phytosanitary Measures (ISPM’s), Pest Risk Assessments (PRAs) need to identify the endangered area, “an area where ecological factors favour the establishment of a pest whose presence in the area will result in economically important loss” [13]. Whilst the standards define the area as “…an officially defined country, part of a country, or all or part of several countries”, the Decision-support scheme for quarantine pests of the European and Mediterranean Plant Protection Organisation [14] encourages the risk assessor to define the endangered area at a very fine ecological and geographical scale. In order to achieve this, it is not sufficient to use even finer resolution climate datasets. Ecological theory indicates that we need to consider the effects of non-climatic factors as we investigate species niches at finer geographical scales [15].

Considering the non-climatic factors affecting a species potential distribution can be a challenging prospect. Many factors could affect the potential habitat suitability for a species, and the importance and effect of these factors may often, themselves, depend on climatic factors [1,16,17]. For example, topographic features that concentrate overland flow of water may improve the suitability of habitat at the dry end of the species' potential range, helping it to avoid drought stress; conversely, at the wet end of the range, this same factor may decrease habitat suitability due to waterlogging. Whilst it is theoretically possible for correlative species distribution models to uncover such relationships, the inclusion of these variables in models may add further to the notorious problems of model over-fitting. This will have the effect of diminishing model transferability; consequently reducing even further the value of such models for pre-border pest risk applications.

Until ecological niche modelling methods improve to the point where these non-climatic factors can be better understood and incorporated into modelling frameworks appropriately, there is a need for a practical risk analysis method that can refine a climatic analysis. Baker et al. [18] is amongst the earliest attempts to incorporate non-climatic information into a PRA, combining a CLIMEX model of climate suitability with a crop host distribution map for Diabrotica virgifera virgifera. In order to assess the pest risks from invasive alien species more precisely, one prospect is to extend the method of Baker et al. [18], combining the semi-mechanistic climate modelling methods with spatial land use. In the present study, we use Parthenium hysterophorus (Asteraceae) as a case study.

Parthenium hysterophorus is an annual or short-lived perennial plant native to the subtropics of North and South America. It is a notorious invasive species which has spread to Australia, Africa, Asia, Oceania, and the Middle-east, where it has become a serious agricultural and rangeland weed affecting crop production and animal husbandry, as well as human health and biodiversity [19,20].

Within the European and Mediterranean Plant Protection Organization region, P. hysterophorus is presently officially recorded only in Israel [21]. It is recorded as naturalised in Egypt [22] and it has also been observed as casual in Belgium [23] and Poland [24]. It is thought to have been introduced in Israel in 1980, probably through the import of contaminated grains from the USA for use as fish food in ponds [25]. The species was also introduced in India and Ethiopia, possibly as a contaminant of grain from the USA. In addition, there are records of its introduction as a contaminant of pasture seed and food aid [26], and through the movement of animals and seed attached to used vehicles (harvesters, military machinery, and other vehicles) [27].

Parthenium hysterophorus reproduces by seeds and is known to be highly prolific, as a single plant may produce on average 40 000 seeds [28]. These seeds are dispersed locally by wind and water and as a contaminant of hay, seed, harvested material, soil, vehicles, machinery, or animals. Parthenium hysterophorus seeds exhibit dormancy mechanisms and can form persistent seed banks, especially where the seeds are incorporated into soil at moderate depths [29]. The species tolerates a wide variety of soils and is a pioneer that can colonise a wide range of habitats: grazing land, summer crops, disturbed and cultivated areas, roadsides, recreation areas, as well as riverbanks and floodplains. Parthenium hysterophorus matures very quickly, with flowering commencing 4–6 weeks after germination; given suitable temperatures it can establish in areas receiving very low rainfall [30].

Parthenium hysterophorus causes major negative impacts on pastures and crops. In India, it has been observed that P. hysterophorus can cause yield losses of up to 40% in several dryland crops [31] cited in [32]. In Ethiopia, the yield of Sorghum bicolor grain was reduced by between 40 and 97% when P. hysterophorus was left uncontrolled throughout the growing season [33]. In Queensland (Australia), it has invaded 170 000 km² of high quality grazing areas and losses to the cattle industry have been estimated to be AUD$22 million per year in control costs and loss of pasture [34]. Infestations of P. hysterophorus can also degrade natural ecosystems, and outcompete native plant species [35,36]. Because P. hysterophorus contains sesquiterpenes and phenolics, it is toxic to cattle, horses and other animals [30]. In addition, meat and milk produced from livestock that has eaten the plant can develop an undesirable flavour [37]. Frequent contact with P. hysterophorus or its pollen can produce serious allergic reactions such as dermatitis, hay fever and asthma in humans and livestock, especially horses [38].

The impacts of P. hysterophorus and reports of its presence in Israel and Belgium sparked concern within the EPPO region and a desire for a PRA to gauge the extent of the threat it posed [39]. A critical component of pest risk is an understanding of the potential distribution of the pest within the PRA area. McConnachie et al. [40] presents a CLIMEX model of P. hysterophorus based on its then known distribution and experimental observations drawn from the scientific literature. In the light of the present known distribution of P. hysterophorus, the CLIMEX model of McConnachie et al. appears somewhat conservative, especially with respect to the cold tolerance limits of this species.

In this paper we refit the CLIMEX model of P. hysterophorus developed by McConnachie et al. [40], and apply irrigation and climate change scenarios to inform global pest risks. We extend the methods of Baker et al. [18] using readily available habitat data, comparing two methods for downscaling the risk map, globally, and for Europe. The first method uses the standard EPPO PRA procedure involving expert assessment of the habitat types that are suitable for invasion, while the second uses an objective inferential method.

Materials and Methods

Modelling outline

The modelling scheme is presented in Fig 1. The distribution data and ecophysiological knowledge for P. hysterophorus were used to develop a CLIMEX model under natural rainfall conditions. Because some distribution records for P. hysterophorus appear to represent populations that are able to persist only due to the presence of supplementary soil moisture, the CLIMEX model is used to run a natural rainfall and an irrigation scenario. These model outputs are combined on a cell-by-cell basis using a map of the distribution of irrigation areas [41] to create composite climate risk models for transient and established populations. The suitable habitat types are used to refine the climate suitability map for establishment to create the endangered area map for the risk assessment. A climate change scenario based on a Global Climate Model is then used to create a future composite climate risk scenario as a means of better understanding the sensitivity of any policy responses to the risks posed by P. hysterophorus.

Fig 1. Modelling scheme for assessing pest risks for Parthenium hysterophorus in the EPPO region using the EPPO Decision-support scheme for quarantine pests.

Green boxes are inputs, blue boxes are models, grey is an intermediate product, and orange boxes are outputs.

Distribution data

The known distribution of P. hysterophorus was assembled from the Global Biodiversity Information Facility (, Clark & Lotter [42], Dhileepan [43], Department of Natural Resources [44], Kilian et al. [45], and Shabbir et al. [46] (Fig 2). Administrative regions that had been reported as being infested by P. hysterophorus, but had no point location records were added to the distribution map as polygons, and shaded lightly to reinforce the lack of spatial precision of these reports. The 2 536 point distribution records were transformed into shapefiles and imported into CLIMEX for overlaying results during model fitting. During model fitting for the natural rainfall scenario, records were checked to consider whether the populations were likely to be able to persist in the absence of irrigation, and whether they represented Established or Transient populations (sensu FAO [13]).

Fig 2. Known global distribution of Parthenium hysterophorus.

Red circles represent distribution points where P. hysterophorus is known to be established, blue triangles indicate outliers in apparently excessively cold locations, yellow triangles excessively dry locations, green triangles excessively wet locations. Pink areas represent national or sub-national administrative units where the species has been recorded established, blue areas indicate countries where the species has been reported as transient populations.

CLIMEX modelling

CLIMEX V3 [2,47] was used to refit the model of McConnachie et al. [40] for P. hysterophorus. CLIMEX calculates a weekly Growth Index (GIW) that describes the species population response to temperature and soil moisture through the Temperature (TI) and Soil Moisture (MI) indices respectively. GIW is integrated annually to calculate the Annual Growth Index (GIA). Stress indices (hot, cold, wet, dry) are factors that limit a species’ ability to persist at a particular location. Individual stress values are combined to create the total Stress Index (SI), and when combined with the Annual Growth Index (GIA) CLIMEX calculates the Ecoclimatic index (EI). The EI is a measure of the overall suitability of a location for species persistence year-round (the larger the value the more suitable). We classified the invasion risk as Endangered if the model indicated that P. hysterophorus was likely to be able to persist year-round (EI>0). At locations where it could grow during a favourable season, but is unlikely to persist year-round due to an inability to complete a generation, due either to stresses or an insufficient heat sum to complete reproductive development (EI = 0, GIA>0), we classified it as Transient [13] (which is synonymous with casual populations sensu Richardson et al. [48]).

The model-fitting strategy involved fitting the stresses to the distribution data in the native range in South America, and the introduced range in Africa, India, and North America. Distribution data in Australia and Eastern Asia were reserved for model validation. In fitting the stress and growth functions, consideration was given to any reported experimental data or theoretical expectations. This practice, combined with the structure of the CLIMEX Compare Locations model helps guard against over-fitting [49]. All CLIMEX model parameters for P. hysterophorus are provided in Table 1, and their derivation is detailed below.

Table 1. CLIMEX model parameters for Parthenium hysterophorus.

Parameter mnemonics follow Sutherst et al. [47].

Temperature index.

Williams and Groves [50] found an optimal temperature regime for P. hysterophorus of 25°C night/30°C day. The Temperature Index parameter values remain unchanged from McConnachie et al. [40].

Cold stress.

The cold stress threshold and rate parameters of McConnachie et al. [40] were relaxed to allow P. hysterophorus to persist in the known, northern locations in the USA and northern India. In doing so, the extreme cold records in China and northern Pakistan and India also became suitable. Williams and Groves [50] (p. 50) noted that plants that were frosted at -6°C suffered “…leaf damage, leading to complete senescence and lateral floret development ceased”. Using -7.5°C as a damaging cold stress threshold (TTCS), the stress accumulation rate of -0.01 week-1 fitted all bar two of the coldest locality records in the northern hemisphere. The outlying records in the Himalayas are found in a region of extremely dissected topography, and the altitude and temperature are so extremely different to the next closest location records that this is likely to be a case of mismatch in either geocoding precision or the climate data. In Argentina, a number of location records for P. hysterophorus in the GBIF database referred to locations that were apparently too cold or too dry for persistence, and for the dry records, did not appear to fall in irrigation areas defined in the irrigation areas database of Siebert et al. [41]. Searching Google Earth using the locality description of these records revealed that they were incorrectly geocoded, and actually referred to wetter locations found at lower elevations.

Dry stress.

In the CLIMEX framework, dry stress may not be a factor that affects annual plants directly, because these plants may be able to survive extended periods of drought in the seed life stage. In this case, Dry Stress (in concert with the GIA) acts in such a manner as to ensure that there is a sufficient period within which the soil moisture is sufficient to complete the life cycle. The dry stress accumulation rate was increased to make the westernmost record in Queensland, Australia barely climatically unsuitable. This had the consequence of making some of the records in Pakistan and Western Argentina unsuitable in the absence of irrigation, which was practised there according to the GMIA database of Portmann et al. [51]. In a small number of cases, location records in Argentina (17), Australia (1), India (1) and Pakistan (2) fell in areas that, according to the climate database were extremely xeric and which were not associated with widespread crop irrigation, at least as portrayed in the global irrigated area database we used (see Composite Risk Mapping below). Examining these locations in Google Earth revealed that these records were not able to be related logically to a long-term climatology. The Argentinian records fell in towns or roadsides where there was irrigation or a concentration of rainfall respectively within areas that were extremely sparsely vegetated. The Australian record was within a braided river channel that floods very infrequently due to rain mostly falling further up the catchment. The Indian record fell in Bikaner, a moderately large town that is in the middle of a desert. Bikaner and its surrounding cropping plots are sustained by the Ganges and Indira Ghandi Canals. The Pakistani records were located along a road through an area between the Indus and Chenab Rivers. This area is a desert, which is covered in extremely sparse vegetation, except for some scattered cropping plots.

Wet stress.

In the native range of P. hysterophorus in South America, there is an extremely large area around the Amazon Basin where the CLIMEX model indicates potential for growth and persistence, but where there are no location records. Whilst this may be due to a lack of surveying and reporting effort, we explored the possibility that P. hysterophorus is unable to persist there due to excessive cloudiness associated with high rainfall (the species is reportedly sensitive to shading [50]. It was possible to make this wet habitat unsuitable using wet stress, improving the model specificity in this area. However, when this level of wet stress was applied, all of Bangladesh, North-eastern India and parts of Central Kenya also became unsuitable; but these areas are covered in location records for P. hysterophorus (see [52] for detailed maps of P. hysterophorus in East Africa. This paradox can perhaps be explained by the fact that whilst the natural vegetation of Bangladesh, North-eastern India, and Central Kenya are similar in structure to that of the Amazon Basin, most of the vegetation in these introduced range locations has been disturbed by intensive agriculture [53]. In the absence of agricultural or pastoral disturbance regimes, we might expect that P. hysterophorus would tend to be outcompeted by the natural vegetation.

Annual heat sum threshold.

The annual heat sum threshold (PDD) of McConnachie et al. [40] was retained at 2 000°C days above 6°C (DV0), barely allowing P. hysterophorus to persist at the coldest known locations of P. hysterophorus in the Himalaya Mountains.

Climate data

The model was fitted initially using the 30’ CliMond CM30_1975H_WO_V1.1 dataset, and subsequently refined with the CM10_1975H_WO_V1.1 [9]. The CliMond 10’ results for 2070 of the A2 SRES climate change scenario run on the CSIRO Mk 3 GCM (CM10_2070_CS_A2_WO_V1.1) was chosen because it represented a reasonably extreme scenario that would highlight the sensitivity of the invasion potential for P. hysterophorus.


An irrigation scenario of 2.5 mm day-1 was applied as a top-up to natural rainfall. Under this scenario, in any week in which average daily precipitation did not meet this threshold, the difference was assumed to be added to the rainfall inputs to the soil moisture model. Actual irrigation rates depend on a variety of factors, including the crops, their stage of growth and climatic factors such as wind flux, temperature, and humidity. The selected rate accords with indicative low-end rates [54]. The irrigation scenario was run on the global CM10_1975H_WO_V1.1 dataset.

Composite soil moisture risk mapping

The irrigation area map from Siebert et al. [41] was used to select within each climate cell, which of the natural and irrigated CLIMEX model results to use in a composite risk map. For each 10’ cell, if the irrigation area was greater than 0, the irrigation scenario results were included. Otherwise the natural rainfall scenario value was used.

Habitat factors

We compared two methods for identifying habitat types that are suitable for invasion by P. hysterophorus. The first, loosely termed an expert assessment, reflects the current standard practice within the EPPO pest risk assessment framework, while the second is an objective inferential method.

In the expert assessment, the habitat types listed in the CORINE database [55] were considered by the EPPO Expert Working Group while performing the PRA for P. hysterophorus, and classified as either suitable or unsuitable for P. hysterophorus based upon consideration of the habitat types where it has been reported in the literature, and where the panel members had observed it in the field. The CORINE database was selected because it is preferred by the EPPO due to its fine spatial resolution. Notably, the spatial coverage of the CORINE database is limited to Europe. The assessors used a consensus method to decide on suitable land use factors, drawing upon published descriptions and personal observations of P. hysterophorus occupying different habitat types.

In the inferential method, the distribution points in Fig 2 were spatially intersected with a global habitat dataset; habitat types with one or more point records were listed. This list was used to identify the subset of habitat types in Europe that was considered suitable. Because the geographical coverage of the CORINE database is limited to Europe, the FAO Land Use Systems of the World version 1.1 [] was used to identify suitable habitat types. This database has a moderately coarse spatial resolution (5 arc minutes) which is equivalent to a map scale of approximately 1:10 000 000. This is coarser than the CORINE database, which summarises the spatial data at a scale of 1:100 000 (equivalent to a raster resolution of approximately 50 m). The attraction of the FAO dataset is that it has a global coverage, enabling risks to be projected globally.

For both the CORINE and FAO datasets, the suitable habitat classes were spatially intersected with the CLIMEX model of climate suitability to create composite climate and land use/habitat risk maps and statistics.


The modelled potential distribution of P. hysterophorus is very extensive, stretching from equatorial areas, through to warm temperate and Mediterranean climates (Fig 3). The effect of irrigation in extending the potential range into xeric regions is obvious in the scattered pockets of suitable locations in the western deserts of the USA (Fig 3A) and the Sahara Desert, where the Nile Valley is a particularly prominent feature (Fig 3B). The model also identifies that there is an additional, extremely large area in the northern hemisphere in which P. hysterophorus could pose a transient biosecurity risk (Fig 4). This accords with its observation in Belgium and Poland, where it was thought to be a transient. In its native range in the Americas, its modelled potential range extends into wet tropical areas, from which there are no recorded observations. Its modelled potential range for establishment in the USA is supported by a few northern location records. Extensive records in Asia in similarly cool conditions further support the conclusion that the plant can likely tolerate such cold conditions. In the wet tropics, consistent excessive soil moisture appears to prevent modelled population growth. In South America, the modelled potential range extends into colder regions than the recorded distribution (compare Figs 2 and 3).

Fig 3. Climate suitability for Parthenium hysterophorus establishment modelled using CLIMEX with the CliMond CM10_1975H_WO_V1.1 climate dataset [9], including the effect of irrigation [41].

(A) Global and (B) Europe and North Africa.

Fig 4. Combined establishment and transient invasion risks posed by Parthenium hysterophorus modelled using CLIMEX with the CliMond CM10_1975H_WO_V1.1 climate dataset [9], including the effect of irrigation [41].

(A) Global and (B) Europe and North Africa.

In Eastern Asia and Australasia, the areas reserved for model validation, the model agreed perfectly with the known distribution (a model sensitivity score of 1.0). Model specificity was also good, with relatively few areas of range underlap. However, in China in particular, there appears to be considerable opportunity for in-filling invasion within the climatically suitable range.

Within the EPPO region, the countries at risk are Albania, Algeria, Azerbaijan, Bosnia & Herzegovina, Bulgaria, Cyprus, Croatia, Former Republic of Macedonia, France, Greece, Hungary, Israel, Italy, Jordan, Kazakhstan, Kyrgyzstan, Malta, Moldova, Morocco, Portugal, Romania, Russia, Serbia, Slovakia, Slovenia, Spain, Tunisia, Turkey, Ukraine and Uzbekistan. The modelled climate suitability pattern is consistent with the reported transient nature of the plant populations in Belgium and Poland (Fig 4) [23,24]. Under the historical (current) climate scenario, more than 2 million ha of the EPPO region is apparently climatically suitable for establishment by P. hysterophorus (Table 2, Fig 5). Of this total area, less than half (approximately 946 000 ha) consists of habitat types considered suitable under the expert model (Table 2). The habitat classes considered at greatest risk (by area) are disturbed (urban, cropping and pastures). Perhaps also of cultural and economic significance is the threat to olive groves (100% of the plantations are at risk), vineyards (90%) and fruit and berry plantations (77%) may be threatened.

Fig 5. Endangered area considering climate (EI ≥ 1) and suitable habitat types in the CORINE database (

Table 2. Areal summary of composite invasion risk to Europe from Parthenium hysterophorus by habitat class according to the CORINE environmental database, considering climate with irrigation scenarios applied according to the GMIAV5 database [41].

Habitat classes are listed in descending order of area at risk under the current climate scenario. Land use is assumed to remain static under the future climate scenario.

Under the inferential FAO habitat model 29 land use classes were identified as being at risk in Europe, including cropping and pasture areas (Table 3, Fig 6). However, grazed forests and shrublands were also identified as being at risk (Table 3). The total area of suitable habitat in Europe modelled as at risk using the FAO dataset and the inferred habitat suitability classes was 1.6 million ha, nearly twice that from the CORINE dataset based on the expert opinion.

Fig 6. The relative frequency of land use systems in the FAO Land Use database overlain by location records for Parthenium hysterophorus from Fig 2.

Table 3. Areal summary of composite invasion risk to Europe from Parthenium hysterophorus by land use system class according to the FAO Land Use Systems of the World database, considering climate with irrigation scenarios applied according to the GMIAV5 database [41].

Habitat classes are listed in descending order of area at risk under the historical (1975H) climate scenario.

The global risk patterns based on the inferential FAO model are similar to those for the expert-based system applied to Europe (Table 4, Fig 7B). However, there are some interesting differences: there was a significant number of records collected from areas classed as open water or wetlands. The likely causes are discussed below.

Fig 7. Endangered area considering climate (EI ≥ 1) and suitable habitat types in the FAO Land Use Systems database, A) Globally, and B) for Europe and North Africa.

Table 4. A real summary of composite global invasion risk from Parthenium hysterophorus by land use system class according to the FAO Land Use Systems of the World database, considering climate with irrigation scenarios applied according to the GMIAV5 database [41].

Habitat classes are listed in descending order of area at risk under the current climate scenario.

Climate change impacts on pest risk

Under the climate change scenario explored here, in the Northern Hemisphere, the modelled pest risks from P. hysterophorus extend further poleward compared with the current climate risks (Fig 8A, see Table 5 for legend description). The USA, continental Europe and northern Middle East are particularly sensitive to this scenario, with the risks changing from transient to endangered over huge areas. There is also a marked band along the equator where decreasing rainfall conditions could allow highland areas of western South America, Central Africa and South East Asia to become endangered by P. hysterophorus (Fig 8A).

Fig 8. Change in climatic establishment risk for Parthenium hysterophorus comparing the CM10_1975H_V1.1 historical climatology and the CliMond.

CM10_2070_CS_A2_V1.1 climate scenario. (A) Global and (B) Europe and North Africa.

Table 5. Summary of modelled pest risk change classes under the 2080 climate scenario.

Within the EPPO region, many countries that appear presently to face only transient risks from P. hysterophorus may become endangered in the future, due primarily to rising temperatures (Austria, Belarus, Belgium, Czech Republic, Germany, Estonia, Latvia, Lithuania, the Netherlands, Poland, Slovenia, the United Kingdom, as well as larger parts of Bosnia and Herzegovina, Hungary, Kazakhstan, Moldova, Russia, Slovakia, Switzerland, Turkey, Ukraine, the southern coast of Sweden) (Fig 8B). The modelled change in climate suitability represents a near doubling of the endangered area (Fig 8B, Table 4).


Despite its extensive present known distribution (Fig 2), the modelled global potential distribution of P. hysterophorus greatly exceeds this, particularly in Africa, Asia, Australia, and Europe. Within its native range, the climate in the Amazon basin appears suitable for P. hysterophorus, but possibly only in the presence of frequent disturbance that reduces competition from other vegetation. If human disturbance patterns are extended into this region, we may find that P. hysterophorus also extends its range there.

Whilst P. hysterophorus is present in Israel within the EPPO region, it is thought to be absent from Europe per se. There is clearly an opportunity to prevent, or at least slow the spread of P. hysterophorus into Europe through vigilant phytosanitary measures. The requirement for free trade pathways between member states means that Israeli exports to Europe may pose a significant threat to the other EPPO member states, and special phytosanitary measures may be worth considering. The movement of people and material from Africa and the Middle East are also dispersal pathways that should be of concern to European biosecurity managers.

Within Africa, Asia and Australia, biosecurity measures to slow the spread of P. hysterophorus may still be worthwhile. Careful consideration of the present and potential distributions in these regions may assist with targeting education material and regulatory measures aimed at minimising impacts and reducing the rate of spread of this damaging invasive alien plant.

Extending the biological control programme against P. hysterophorus to Israel and other invaded countries is worthy of consideration. It may also be economically attractive for European states at risk of invasion by P. hysterophorus to co-invest in biological control measures in Israel and other places that pose a source threat.

Habitat factors

Irrigation has an important effect on extending the range of P. hysterophorus, particularly in Saharan Africa, the Middle East and Central Australia. Conversely, within Europe, restricting the endangered area by using habitat types refines the area at risk considerably within the climatic range. These analytical elements could aid in refining economic impact analyses, and also perhaps in informing surveillance and rapid responses to incursion detections.

The spatial analysis of the distribution data for P. hysterophorus using the FAO dataset was revealing; expanding the range of habitat types beyond those identified by the expert assessment process. The association between the open water and wetland land use classes and P. hysterophorus was surprising given that P. hysterophorus does not grow in waterlogged situations. However, P. hysterophorus does grow on floodplains [56], so it is likely that the location records fall within riparian zones within the coarse open water and wetland land use classes. Similarly, during the expert deliberations, forested areas were discounted as suitable habitat on the grounds that P. hysterophorus reportedly grows poorly under shaded conditions, and would therefore be unable to persist. The FAO dataset comparison underscores the fact that forests (particularly those that are actively managed) are frequently a mosaic of different seral stages, and that ruderals such as P. hysterophorus can persist either through recolonisation or the maintenance of seed banks [57]. The more granular spatial resolution of the CORINE database is reflected in a larger set of habitat classes than the FAO dataset. Both of these factors make the CORINE database inherently less likely to create confusing interpretation problems with spatial intersections, as happened with the FAO dataset. However, the limitation usually lies in the spatial resolution of the location records for invasive alien species, rather than the habitat/land use data. This is especially marked for species location data collected prior to the widespread availability of GPS units. Hence, it is unclear whether chasing a finer-scale, globally-conformal, land use/habitat type classification would result in a more accurate assessment of the non-climatic habitat risk factors.

Whilst the fine spatial resolution of the CORINE database may be highly valued for risk assessment in the EPPO region, the lack of conformal global coverage is clearly a drawback for estimating non-climatic habitat risk factors for invasive alien species that have little or no history in the risk assessment area. The large size of the CORINE database also created practical challenges for spatial analyses in geographical information systems, sometimes requiring the dataset to be split in two for spatial intersections. One option for pest risk analysts is to sacrifice some precision for potentially greater accuracy, employing the FAO method and dataset as we have demonstrated here. Another option is to use a hybrid two-phase method combining the insights gained through the FAO dataset analysis with expert opinion to select classes from the CORINE database.

Responding to climate change impacts on invasion risks

As the rate of change and the extent of future climatic changes are unknown (and largely unknowable), it is impossible and imprudent to use climate change scenarios such as the one presented here to inform future biosecurity policies and plans directly. Rather, the risk exposure revealed here should be used as the basis for understanding the nature of biosecurity decisions and their consequences under an inherently uncertain pattern of changing risks. In those areas where the future climate scenario risk maps indicate a risk of transient populations of P. hysterophorus, less effort may be placed on prevention, detection, and rapid response to this weed. However, if the risks might change in the future due to potential climate changes, several adaptation options present themselves (Table 6). It is imprudent to invest in expensive measures to address a problem that may not eventuate. The fact that the climate change scenario indicates that the risks for Europe are likely to increase in the future adds further weight to the conclusion that the present invasion risks by P. hysterophorus, based on historical climate, are significant. In the case of P. hysterophorus in the EPPO region, the climate change analysis adds little to the conclusion that there is a significant area at risk. The most cost-effective response may therefore be to consider what measures can be undertaken to stop the spread of P. hysterophorus out of Israel, or from other countries into the EPPO region, as well as to prevent its entry in EPPO countries at risk.

Table 6. Possible responses to potentially emerging pest risks under a rapidly changing climate.

Model limitations

The CLIMEX model was fitted using the best available data and understanding available at the time of the analysis. However, we should be mindful that climate and distribution data are imperfect. The spatial resolution of the distribution data varied, and the estimated precision was not always reported. The mismatch between the resolution of the land use dataset and the species distribution data had the potential to pick up spurious habitat associations; hence we were careful to scrutinise low frequency associations. We should also be mindful that the CLIMEX Compare Locations model is a simplification of the complex ecological processes that define a species niche. The land use classification in the FAO dataset and the identification of the irrigated areas will doubtless contain minor spatial and classification errors. The mis-fitting points at the dry end of P. hysterophorus’ range indicate a limit to the spatial precision in the global irrigated area database. However, despite these sources of potential errors, the analysis appears suitable for its intended purpose–to provide an indication of areas at risk of invasion should P. hysterophorus be introduced. Each of the mis-fitting points was in close spatial association with areas that were indicated as being suitable, and for which there were location records. This underscores the notion that the resulting maps should be used in aggregate to inform regional risk patterns, rather than being scrutinised at the level of an individual cell. In the extreme xeric and cold limits habitat suitability will be more subject to unusual micro-habitat variations that cannot be accounted for with global datasets and modelling.

With the climate change scenario it is important to remember that we are not applying observation data about the future. We have selected a single plausible scenario with which to stress-test the biosecurity conclusions of our niche modelling. Biosecurity managers should not make plans on the basis that the climate change scenario results presented here will eventuate. This could lead to an expensive waste of resources. Rather, managers should seek to understand firstly whether the scenario changes the invasion risks significantly within their jurisdiction. If so, they should consider what adaptive management processes they might prudently implement to monitor and manage that potential emerging threat, taking into account lead times for any adaptation measures.

Advancing pest risk modelling

In this paper we applied two advances in pest risk modelling: spatially-explicit irrigation scenarios, and the inferential derivation of non-climatic habitat classes. Both methods are relatively easy to apply using a GIS with the freely available irrigation and land use datasets. The explicit irrigation scenario method allows the niche model to describe the species niche using biologically realistic parameters. In the absence of this method, the model would be unable to identify correctly the habitats at risk in xeric environments, either under-predicting (biologically realistic parameters), or over-predicting (using biologically unrealistic parameters that allow persistence in xeric environments).

The inferential method of identifying suitable land use classes can clearly provide a degree of rigour to the downscaling process. However, it does not abrogate the responsibility of the modeller or risk assessor to evaluate the resulting list of habitats critically and sceptically. Low frequency or unexpected habitat types should serve as a warning sign of a potential error. Whilst the impact of the downscaling process on the estimated endangered area is substantial, it may have minimal implications for analyses of the economic impacts of invasive alien species where the impacts apply to industries with well-defined spatially-explicit production characteristics. However, for species whose impacts are related to the area occupied, and affect natural environments, these downscaling methods could make a substantial difference to the results.


This paper is dedicated to the memory of Robert (Bob) Sutherst, who developed the CLIMEX modelling system, and who was a pioneer in the field of computer-based pest risk modelling. Sadly, Bob passed away the week before the work for this paper commenced.

Author Contributions

Conceived and designed the experiments: DJK SB NO. Performed the experiments: DJK SB NO. Analyzed the data: DJK SB NO. Wrote the paper: DJK SB NO GF AOL FDP RP AS TY.


  1. 1. Sutherst RW. Pest species distribution modelling: origins and lessons from history. Biological Invasions. 2013; 16: 239–256.
  2. 2. Sutherst RW, Maywald GF. A computerised system for matching climates in ecology. Agriculture, Ecosystems and Environment. 1985; 13: 281–299.
  3. 3. Williams JD, Sutherst RW, Maywald GF, Petherbridge CT. The southward spread of buffalo fly (Haematobia irritans) in Eastern Australia and its survival through a severe winter. Australian Veterinary Journal. 1985; 62: 367–369. pmid:3834900
  4. 4. Hutchinson MF, Gessler PE. Splines—more than just a smooth interpolator. Geoderma. 1994; 62: 45–67.
  5. 5. Yonow T, Sutherst RW. The geographical distribution of the Queensland fruit fly, Bactrocera (Dacus) tryoni, in relation to climate. Australian Journal of Agricultural Research. 1998; 49: 935–953.
  6. 6. Booth TH, Nix HA, Hutchinson MF, Busby JR. Grid Matching: a new method for homoclime analysis. Agricultural and Forest Meteorology. 1987; 39: 241–255.
  7. 7. New M, Hulme M, Jones P. Representing twentieth-century space-time climate variability. Part I: Development of a 1961–90 mean monthly terrestrial climatology. Journal of Climate. 1999; 12: 829–856.
  8. 8. Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A. Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology. 2005; 25: 1965–1978.
  9. 9. Kriticos DJ, Webber BL, Leriche A, Ota N, Bathols J, Macadam I, et al. CliMond: global high resolution historical and future scenario climate surfaces for bioclimatic modelling. Methods in Ecology and Evolution. 2012; 3: 53–64.
  10. 10. Webber BL, Yates CJ, Le Maitre DC, Scott JK, Kriticos DJ, Ota N, et al. Modelling horses for novel climate courses: insights from projecting potential distributions of native and alien Australian acacias with correlative and mechanistic models. Diversity and Distributions. 2011; 17: 978–1000.
  11. 11. Watt MS, Kriticos DJ, Lamoureaux SL, Bourdot GW. Climate change and the potential global distribution of the invasive weed Nassella trichotoma. Weed Science. 2011; 59: 538–545.
  12. 12. Kriticos DJ, Leriche A. The effects of spatial data precision on fitting and projecting species niche models. Ecography. 2009; 33: 115–127.
  13. 13. FAO. International standards for phytosanitary measures: 1 to 24. Rome: Secretariat of the International Plant Protection Convention; 2006.
  14. 14. EPPO. PM 5/3 (5) Guidelines on Pest Risk Analysis. Decision-support scheme for quarantine pests. 2011: 1–44.
  15. 15. Brown JH, Stevens GC, Kaufman DM. The geographic range: size, shape, boundaries, and internal structure. Annual Review of Ecology and Systematics. 1996; 27: 597–623.
  16. 16. Andrewartha HG, Birch LC. The ecological web: more on the distribution and abundance of animals. Chicago: University of Chicago Press; 1984.
  17. 17. Woodward FI. Climate and plant distribution: Cambridge University Press; 1987.
  18. 18. Baker RHA, Cannon RJC, MacLeod A. Predicting the potential distribution of alien pests in the UK under global climate change: Diabrotica virgifera virgifera. Proceedings of the BCPC International Congress—Crop science and technology. Alton, UK: BCPC Publications; 2003. pp. 1201–1208.
  19. 19. CABI. Parthenium hysterophorus. 2014 Available: Accessed 18 May 2014.
  20. 20. Adkins S, Shabbir A. Biology, ecology and management of the invasive parthenium weed (Parthenium hysterophorus L.). Pest Management Science. 2014; 70: 1023–1029. pmid:24430973
  21. 21. EPPO. PQR—EPPO database on quarantine pests; 2013. Available:
  22. 22. Boulos L, El-Hadidi MN. The Weed Flora of Egypt. Cairo: American University of Cairo Press; 1984.
  23. 23. Verloove F. Catalogue of neophytes in Belgium (1800–2005). Meise: National Botanic Garden of Belgium; 2006.
  24. 24. Mirek Z, Piękoś-Mirkowa H, Zając A, Zając M. Flowering plants and pteridophytes of Poland. A Checklist. Biodiversity Poland. 2002; 1: 9–442.
  25. 25. Dafni A, Heller D. Adventive flora of Israel–phytogeographical, ecological and agricultural aspects. Plant Systematics and Evolution. 1982; 140: 1–18.
  26. 26. Murphy ST, Cheesman OD. The Aid Trade–International Assistance Programmes as Pathways for the Introduction of Invasive Alien Species (a preliminary report). A paper prepared by CABI Bioscience, UK Centre.; 2006. 38 p.
  27. 27. Auld BA, Hosking J, McFadyen R. Analysis of the spread of tiger pear and parthenium weed in Australia. Australian Weeds. 1982; 2: 56–60.
  28. 28. Dhileepan K. Reproductive variation in naturally occurring populations of the weed Parthenium hysterophorus (Asteraceae) in Australia. Weed Science. 2012; 60: 571–576.
  29. 29. Navie Panetta, McFadyen Adkins. Behaviour of buried and surface-sown seeds of Parthenium hysterophorus. Weed Research. 1998; 38: 335–341.
  30. 30. Navie S, McFadyen R, Panetta F, Adkins S. The biology of Australian weeds. 27. Parthenium hysterophorus L. Plant Protection Quarterly. 1996; 11: 76–88.
  31. 31. Khosla SN, Sobit SN. Effective control of Parthenium hysterophorus Linn. Pesticides (India). 1981; 15: 18–19.
  32. 32. Kandasamy OS. Parthenium weed: status and prospects of chemical control in India. In: Ramachandra Prasad TV, Nanjappa HV, Devendra R, Manjunath Subramanya A, Chandrashekar S et al., editors. Proceedings of the Second International Conference on Parthenium Management: University of Agricultural Sciences, Bangalore, India; 2005. pp. 134–142.
  33. 33. Tamado T, Ohlander L, Milberg P. Interference by the weed Parthenium hysterophorus L. with grain sorghum: influence of weed density and duration of competition. International Journal of Pest Management. 2002; 48: 183–188.
  34. 34. Chippendale JF, Panetta FD. The cost of Parthenium weed to the Queensland cattle industry. Plant Protection Quarterly. 1994; 9: 73–76.
  35. 35. Ramachandra Prasad T, Denesh G, Kiran Kumar VK, Sanjay MT. Impact of Parthenium hysterophorus L. on bioidiversity, ill effects and integrated approaches to manage in Southern Karanataka. International Conference on Biodiversity; 2010. pp. 206–211.
  36. 36. Ramachandra Prasad TV, Denesh GR, Ananda N, Sushilkumar, Varsheny JG. Parthenium hysterophorus L.- a national weed, its menace and integrated management strategies in India. In: Gautam RD, Mahapatro GK, Bhalla S, Shankarganehs K, Gautam S et al., editors. 3rd International Conference on Parthenium: Indian Agricultural Research Institute, New Delhi, India; 2010. pp. 13–20.
  37. 37. Towers G, Subba Rao P. Impact of the pan-tropical weed, Parthenium hysterophorus L. on human affairs; 1992; 1992. pp. 134–138.
  38. 38. McFadyen R. Parthenium weed and human health in Queensland. Australian Family Physician. 1995; 24: 1455–1459. pmid:7677614
  39. 39. EPPO. Pest risk analysis for Parthenium hysterophorus. Paris: EPPO; 2014.
  40. 40. McConnachie AJ, Strathie LW, Mersie W, Gebrehiwot L, Zewdie K, Abdurehim A, et al. Current and potential geographical distribution of the invasive plant Parthenium hysterophorus (Asteraceae) in eastern and southern Africa. Weed Research. 2011; 51: 71–84.
  41. 41. Siebert S, Doll P, Hoogeveen J, Faures JM, Frenken K, Feick S. Development and validation of the global map of irrigation areas. Hydrology and Earth System Sciences. 2005; 9: 535–547.
  42. 42. Clark K, Lotter W. What is parthenium weed up to in Tanzania? International Parthenium News. 2011; 3: 1–2.
  43. 43. Dhileepan K. Managing parthenium weed across diverse landscapes: Prospects and limitations. In: Inderjit , editor. Management of Invasive Weeds: Springer; 2009. pp. 227–259.
  44. 44. Department of Natural Resources the Arts and Sport. NT Weed Risk Assessment: Species information for Parthenium hysterophorous (Parthenium). 2010. 25 pp. p.
  45. 45. Kilian N, Hein P, Hubaishan MA. New and noteworthy records for the flora of Yemen, chiefly of Hadhramout and Al-Mahra. Willdenowia. 2002: 239–269.
  46. 46. Shabbir A, Dhileepan K, Adkins SW. Spread of parthenium weed and its biological control agent in the Punjab, Pakistan. Pakistan Journal of Weed Science Research. 2012; 18: 581–588.
  47. 47. Sutherst RW, Maywald GF, Kriticos DJ. CLIMEX Version 3: User's Guide: Hearne Scientific Software Pty Ltd; 2007.
  48. 48. Richardson DM, Pysek P, Rejmanek M, Barbour MG, Panetta FD, West CJ. Naturalization and invasion of alien plants: concepts and definitions. Diversity and Distributions. 2000; 6: 93–107.
  49. 49. Sutherst RW, Bourne AS. Modelling non-equilibrium distributions of invasive species: a tale of two modelling paradigms. Biological Invasions. 2009; 11: 1231–1237.
  50. 50. Wiliams JD, Groves RH. The influence of temperature and photoperiod on growth and development of Parthenium hysterophorous L. Weed Research. 1980; 20: 47–52.
  51. 51. Portmann FT, Siebert S, Doll P. MIRCA2000-Global monthly irrigated and rainfed crop areas around the year 2000: A new high-resolution data set for agricultural and hydrological modeling. Global Biogeochemical Cycles. 2010; 24: Gb1011.
  52. 52. Wabuyele E, Lusweti A, Bisikwa J, Kyenune G, Clark K, Lotter WD, et al. A Roadside Survey of the Invasive Weed Parthenium hysterophorus (Asteraceae) in East Africa. Journal of East African Natural History. 2014; 103: 49–57.
  53. 53. Ramankutty N, Foley JA. Estimating historical changes in global land cover: Croplands from 1700 to 1992. Global Biogeochemical Cycles. 1999; 13: 997–1027.
  54. 54. Brouwer C, Heibloem M. Irrigation Water Management: Irrigation Water Needs. Rome, Italy: FAO; 1986.
  55. 55. Bossard M, Feranec J, Otahel J. CORINE land cover technical guide–Addendum 2000. Copenhagen: European Environment Agency; 2000.
  56. 56. Cruttwell McFadyen R. Biological control against parthenium weed in Australia. Crop Protection. 1992; 11: 400–407.
  57. 57. Grime JP. Primary strategies in the established phase. In: Grime JP, editor. Plant strategies and vegetation processes Chichester New York: John Wiley & Sons Ltd; 1979. pp. 7–55.