Spatial multi-criteria decision analysis for modelling suitable habitats of Ornithodoros soft ticks in the Western Palearctic region.

Ticks are economically and medically important ectoparasites due to the injuries inflicted through their bite, and their ability to transmit pathogens to humans, livestock, and wildlife. Whereas hard ticks have been intensively studied, little is known about soft ticks, even though they can also transmit pathogens, including African Swine Fever Virus (ASFV) affecting domestic and wild suids or Borrelia bacteria causing tick-borne relapsing fever (TBRF) in humans. We thus developed a regional model to identify suitable spatial areas for a community of nine Ornithodoros tick species (O. erraticus, O. sonrai, O. alactagalis, O. nereensis, O. tholozani, O. papillipes, O. tartakovskyi, O. asperus, O. verrucosus), which may be of medical and veterinary importance in the Western Palearctic region. Multi-Criteria Decision Analysis was used due to the relative scarcity of high-quality occurrence data. After an in-depth literature review on the ecological requirements of the selected tick community, five climate-related factors appeared critical for feeding activity and tick development: (i) a spring temperature exceeding 10°C to induce the end of winter soft tick quiescent period, (ii) a three-months summer temperature above 20°C to allow tick physiological activities, (iii) annual precipitation ranging from 60mm to 750mm and, in very arid areas, (iv) dry seasons interrupted by small rain showers to maintain minimum moisture inside their habitat along the year or (v) residual water provided by perennial rivers near habitats. We deliberately chose not to include biological factors such as host availability or vegetation patterns. A sensitivity analysis was done by performing multiple runs of the model altering the environmental variables, their suitability function, and their attributed weights. To validate the models, we used 355 occurrence data points, complemented by random points within sampled ecoregions. All models indicated suitable areas in the Mediterranean Basin and semi-desert areas in South-West and Central Asia. Most variability between models was observed along northern and southern edges of highly suitable areas. The predictions featured a relatively good accuracy with an average Area Under Curve (AUC) of 0.779. These first models provide a useful tool for estimating the global distribution of Ornithodoros ticks and targeting their surveillance in the Western Palearctic region.

Nationwide inventory of mosquito biodiversity (Diptera: Culicidae) in Belgium, Europe.

To advance our restricted knowledge on mosquito biodiversity and distribution in Belgium, a national inventory started in 2007 (MODIRISK) based on a random selection of 936 collection points in three main environmental types: urban, rural and natural areas. Additionally, 64 sites were selected because of the risk of importing a vector or pathogen in these sites. Each site was sampled once between May and October 2007 and once in 2008 using Mosquito Magnet Liberty Plus traps. Diversity in pre-defined habitat types was calculated using three indices. The association between species and environmental types was assessed using a correspondence analysis. Twenty-three mosquito species belonging to traditionally recognized genera were found, including 21 indigenous and two exotic species. Highest species diversity (Simpson 0.765) and species richness (20 species) was observed in natural areas, although urban sites scored also well (Simpson 0.476, 16 species). Four clusters could be distinguished based on the correspondence analysis. The first one is related to human modified landscapes (such as urban, rural and industrial sites). A second is composed of species not associated with a specific habitat type, including the now widely distributed Anopheles plumbeus. A third group includes species commonly found in restored natural or bird migration areas, and a fourth cluster is composed of forest species. Outcomes of this study demonstrate the effectiveness of the designed sampling scheme and support the choice of the trap type. Obtained results of this first country-wide inventory of the Culicidae in Belgium may serve as a basis for risk assessment of emerging mosquito-borne diseases.

Absence reduction in entomological surveillance data to improve niche-based distribution models for Culicoides imicola.

Data-driven models for the prediction of bluetongue vector distributions are valuable tools for the identification of areas at risk for bluetongue outbreaks. Various models have been developed during the last decade, and the majority of them use linear discriminant analysis or logistic regression to infer vector-environment relationships. This study presents a performance assessment of two established models compared to a distribution model based on a promising ensemble learning technique called Random Forests. Additionally, the impact of false absences, i.e. data records of suitable vector habitat that are, for various reasons, incorrectly labelled as absent, on the model outcome was assessed using alternative calibration-validation schemes. Three reduction methods were applied to reduce the number of false absences in the calibration data, without loss of information on the environmental gradient of suitable vector habitat: random reduction and stratified reduction based on the distance between absence and presence records in geographical (Euclidean distance) or environmental space (Mahalanobis distance). The results indicated that the predicted vector distribution by the Random Forest model was significantly more accurate than the vector distributions predicted by the two established models (McNemar test, p<0.01) when the calibration data were not reduced with respect to false absences. The performance of the established models, however, increased considerably by application of stratified false absence reductions. Model validation revealed no significant difference between the performance of the three distinct Culicoides imicola distribution models for the majority of alternative stratified reduction schemes. The main conclusion of this study is that the application of Random Forests, or linear discriminant analysis and logistic regression on the condition that calibration data were first reduced on geographical or environmental information, potentially lead toward better vector distribution models.

Relative importance of management, meteorological and environmental factors in the spatial distribution of Fasciola hepatica in dairy cattle in a temperate climate zone.

Fasciola hepatica, a trematode parasite with a worldwide distribution, is the cause of important production losses in the dairy industry. Diagnosis is hampered by the fact that the infection is mostly subclinical. To increase awareness and develop regionally adapted control methods, knowledge on the spatial distribution of economically important infection levels is needed. Previous studies modelling the spatial distribution of F. hepatica are mostly based on single cross-sectional samplings and have focussed on climatic and environmental factors, often ignoring management factors. This study investigated the associations between management, climatic and environmental factors affecting the spatial distribution of infection with F. hepatica in dairy herds in a temperate climate zone (Flanders, Belgium) over three consecutive years. A bulk-tank milk antibody ELISA was used to measure F. hepatica infection levels in a random sample of 1762 dairy herds in the autumns of 2006, 2007 and 2008. The infection levels were included in a Geographic Information System together with meteorological, environmental and management parameters. Logistic regression models were used to determine associations between possible risk factors and infection levels. The prevalence and spatial distribution of F. hepatica was relatively stable, with small interannual differences in prevalence and location of clusters. The logistic regression model based on both management and climatic/environmental factors included the factors: annual rainfall, mowing of pastures, proportion of grazed grass in the diet and length of grazing season as significant predictors and described the spatial distribution of F. hepatica better than the model based on climatic/environmental factors only (annual rainfall, elevation and slope, soil type), with an Area Under the Curve of the Receiver Operating Characteristic of 0.68 compared with 0.62. The results indicate that in temperate climate zones without large climatic and environmental variation, management factors affect the spatial distribution of F. hepatica, and should be included in future spatial distribution models.

The use of bulk-tank milk ELISAs to assess the spatial distribution of Fasciola hepatica, Ostertagia ostertagi and Dictyocaulus viviparus in dairy cattle in Flanders (Belgium).

Fasciola hepatica, Ostertagia ostertagi and Dictyocaulus viviparus are helminth parasites with a wide distribution and an important economic impact in cattle in temperate climates. This paper describes the spatial distribution of F. hepatica, O. ostertagi and D. viviparus in dairy herds in Flanders (Belgium). One thousand eight hundred herds were selected at random from the Flemish dairy population (n=7002), stratified on community level to obtain a sample representative for the entire study area. From each herd, a bulk milk sample collected in autumn 2006 was analysed with previously described antibody-ELISAs in order to identify herds where the parasite infection level is likely to cause production loss (F. hepatica and O. ostertagi) (defined as economic infections) or where patent infections have been present over the past grazing season (D. viviparus). The herd prevalence of economic infections with F. hepatica and O. ostertagi was 37.3% (95% Confidence Interval (CI): 35.1-39.7) and 59.1% (95%CI: 56.8-61.4), respectively. The herd prevalence of D. viviparus was 19.6% (95%CI: 17.7-21.6). On 28.9% (CI 26.8-31.3) of the herds, low levels of infection were observed for all three of the helminths. The presence of clustering of (economic) infections was studied using Moran's I, whereas the location and size of the clusters were studied using the spatial scan statistic, the Local Indicator of Spatial Association and Kernel density plotting. A marked clustering in the spatial distribution of F. hepatica and a mild clustering in the spatial distribution of O. ostertagi were observed. D. viviparus infections were spread evenly over Flanders. Knowledge of locations of high risk areas can lead to increased awareness and may be the start of the development of regionally adapted control measures.

The impact of habitat fragmentation on tsetse abundance on the plateau of eastern Zambia.

Tsetse-transmitted human or livestock trypanosomiasis is one of the major constraints to rural development in sub-Saharan Africa. The epidemiology of the disease is determined largely by tsetse fly density. A major factor, contributing to tsetse population density is the availability of suitable habitat. In large parts of Africa, encroachment of people and their livestock resulted in a destruction and fragmentation of such suitable habitat. To determine the effect of habitat change on tsetse density a study was initiated in a tsetse-infested zone of eastern Zambia. The study area represents a gradient of habitat change, starting from a zone with high levels of habitat destruction and ending in an area where livestock and people are almost absent. To determine the distribution and density of the fly, tsetse surveys were conducted throughout the study area in the dry and in the rainy season. Landsat ETM+ imagery covering the study area were classified into four land cover classes (munga, miombo, agriculture and settlements) and two auxiliary spectral classes (clouds and shadow) using a Gaussian Maximum Likelihood Classifier. The classes were regrouped into natural vegetation and agricultural zone. The binary images were overlaid with hexagons to obtain the spatial spectrum of spatial pattern. Hexagonal coverage was selected because of its compact and regular form. To identify scale-specific spatial patterns and associated entomological phenomena, the size of the hexagonal coverage was varied (250 and 500 m). Per coverage, total class area, mean patch size, number of patches and patch size standard deviation were used as fragmentation indices. Based on the fragmentation index values, the study zone was classified using a Partitioning Around Mediods (PAM) method. The number of classes was determined using the Wilks' lambda coefficient. To determine the impact of habitat fragmentation on tsetse abundance, the correlation between the fragmentation indices and the index of apparent density of the flies was determined and habitat changes most affecting tsetse abundance was identified. From this it followed that there is a clear relationship between habitat fragmentation and the abundance of tsetse flies. Heavily fragmented areas have lower numbers of tsetse flies, but when the fragmentation of natural vegetation decreases, the number of tsetse flies increases following a sigmoidal-like curve.

Fragmentation analysis for prediction of suitable habitat for vectors: example of riverine tsetse flies in Burkina Faso.

Tsetse flies are the cyclic vectors of sleeping sickness and African animal trypanosomosis. The possibility to classify the natural habitat of riverine tsetse species is explored in the Mouhoun River basin, Burkina Faso: the objectives were to discriminate the riverine forests community types and their fragmentation levels by using Landsat 7 enhanced thematic mapper images, to map tsetse densities. Glossina palpalis gambiensis Vanderplank 1949 (Diptera: Glossinidae) and G. tachinoides Westwood, 1850 are the vectors of trypanosomoses in this area. After a supervised classification, the community types were discriminated using the water area in 400-m-wide polygons around the river. A fragmentation analysis of the swamp forest unit, cross-tabulated with the community types, lead to identification of the final landscapes where tsetse apparent densities (ADT) were implemented using a training data set of 608 trap locations. The predicted ADT were then compared with an independent validation data set of 78 trap locations. The correlation between the model predictions and the validation data set was high, validating this approach (P < 0.001). The riverine forest community type and fragmentation level are critical factors for riverine tsetse species, which should be taken into consideration to map their suitable habitat.

Distribution and risk factors of bovine cysticercosis in Belgian dairy and mixed herds.

Bovine cysticercosis is an important food safety issue that is of economic concern. In Belgium, in the last years an increase in the number of cases, mostly light infections, was observed. The role of contact with contaminated surface water has been hypothesized as the main route of transmission. Based on abattoir records from 2001 till 2003 the distribution and risk factors of bovine cysticercosis among dairy and mixed farms were studied in four provinces, using Geographic Information Systems (GIS) and questionnaires. The risk factors were analysed using a case-control study design. The case group consisted of herds from which homebred cattle with cysticercosis had been detected at the abattoir; the control group was composed of herds where no cases had been detected. Case herds were distributed over the study area. A logistic regression analysis revealed that the location (province), the number of slaughtered cattle, the flooding of pastures, free access of cattle to surface water and the proximity of wastewater effluent were significant explanatory variables for bovine cysticercosis to be recorded in a herd.

Quantifying the wind dispersal of Culicoides species in Greece and Bulgaria.

This paper tests the hypothesis that Culicoides (Diptera: Ceratopogonidae) species can be propagated by wind over long distances. Movement patterns of midges were inferred indirectly from patterns of the spread of bluetongue outbreaks between farms (using outbreak data from 1999-2001 for Greece, Bulgaria and Turkey) and then matched to concurrent wind patterns. The general methodology was to determine wind trajectories to and from each outbreak site based on the horizontal and vertical wind components of the European ReAnalysis-40 (ERA-40) dataset from the European centre for medium-range weather forecast (ECMWF). Forward trajectories (downwind or where the windvectors pointed to) and backward trajectories (upwind or where the wind-vectors originated from) were calculated for each outbreak for the period from one week before to one week after it had been recorded. These wind trajectories were then compared with the general outbreak patterns taking into consideration the different serotypes involved. It was found that the wind trajectories could be matched to the temporal distribution of the outbreak cases. Furthermore, the spread of the infected vector via the calculated wind trajectories was corroborated by molecular evidence. The conclusion is that the methodology presented is appropriate for quantifying the risk of spread of infected Culicoides midges by wind and that this approach could form an important component of a regional early-warning system for bluetongue.