This is the first study aimed to assess the risk of ASFV introduction into the EU associated to TAR. Current presence of the disease without control in areas of RF close to the EU borders, together with results of the EFSA advise , other studies  and recent published legislation , point out the importance that illegal trade and other potential pathways such as transport fomites may have in the ASFV introduction into the EU, which remarks the importance of the study here.
However, the estimation of TAR is not a simple task. Vehicles and waste from international means of transport have been frequently suggested as a potential route for disease spread, specifically for ASFV introduction into free areas [29, 30]; but no studies have quantified this risk. The lack of information for these TAR does not allow to use traditional risk assessment models making necessary to develop alternative approaches to analyze the risk of ASFV introduction by these pathways. The methodology proposed here combines methods used in the knowledge driven models used for spatial modeling of diseases  as well as methods for the risk estimation based on expert opinion elicitation [20, 21]. Moreover, we used available data on risk factors, conveniently standardized, weighted by EO and linearly combined (as done in spatial modeling), to estimate the relative risk of TAR in the different EU countries. Although results should be cautiously interpreted considering all the assumptions and uncertainties associated with the model structure and data used, the approach is believed to be useful to evaluate the TAR risk. This study was specifically performed considering the selected routes of entrance, and, importantly, the specific characteristics of the pathogen, for the risk tended to be estimated. For example, the long survival of ASFV in all kind of meat and infected products allows to measure the risk based on potential incoming volumes of infected material, without considering the survival time of the virus on it. However, when adapting the methodology presented here for other animal diseases or routes of entrance, this important feature, as well as many other specific characteristics and parameters, should be modified conveniently to incorporate the features of the disease under study.
One of the most important aspects to be considered is that the model does not provide probabilities, but compares the relative risk between the 27 EU countries based on the risk factors evaluated. Indeed, a high value on the model results does not imply an absolute high level of risk, but a higher one compared with other EU members. On the other hand, the selection of information for each of the risk factors used in the model is influenced by the quality and availability of data sources for the 27 EU countries. For example, the degree of cleaning of returning trucks is based on scenario rather than real data due to the absence of this data for each of the EU countries. Therefore the results presented here depend on the quality and reliability of this data. It is important to consider also that the model only estimates the risk of entrance/release of potential ASFV-contaminated material/transports, but does not consider the subsequent exposure of the susceptible livestock population in the destination country.
Another idiosyncrasy of the model is the use of weights obtained by EO for combination of risk factor and pathways. EO process is a valuable method widely use when no other “more objective” information is available (i.e. literature, etc.) and particularly for the estimation of complex parameters or parameters with significant uncertainties, as some of the presented in this study. Particularly, Delphi approach is one of the most frequently used methods of EO that originally does not allow for interaction between the experts . However, in this study small modifications were made by the little interaction between experts during the results presentation and the use of electronic devices for voting. This technique implies many advantages, being an adequate way to collect information for solving problems. However, the lack of universal guidelines or standardized procedures for its performance could arrive into difficulties that should be cautiously considered . Some problems of the technique could be derived from the inappropriate selection of the experts, the lack of previous information, the inadequate performance of the questionnaire or the combination of the results. Nevertheless, this method may provide a more realistic and updated view of the scenarios under evaluation, in this case related with ASF risk, based on the experts valuable experience. Moreover, and because weights used in the weighted combination of the risk factor are a critical aspect of the model, an intensive SA was performed in order to identify the impact of these estimated weights in the final results. This SA reveals that the model is robust and do not significantly change when changing the weights provided by EO. In fact, none of the countries changed more than one category in the different SA scenarios evaluated. For example, the 25% decrease on the weight of returning trucks, which is the scenario with lowest correlation coefficient (Rho = 0.97), affected categories of seven countries. Most of them (three countries) changed from low to very low risk, two decreased from medium to low risk, and two from high to medium risk. These changes result in a very similar risk map, but with a slight difference of risk category in these countries, which confirms the robustness of the model.
At the same time, the use of different scenarios in some of the measured parameters allows Animal Health (AH) Authorities in each EU country to have the possibility to select the scenario that more realistically represents their current situation based on their expert opinion. For example, we are providing three different results based on certain assumptions, but AH Authorities may consider that for their countries only the scenario one is realistic, so they will have the possibility to select it and visualize the correspondent outputs. This flexibility, as well as, the easy update and the possibility of incorporation of more detailed information (if available for some countries) instead of being considered a limitation, is considered as one of the main strengths of the model.
Model results reveal that the median of the risk values for ASFV by TAR in the 27 EU countries is low (for 16 of the 27 EU countries), although big differences were found between countries and pathways. An expected result of the model is that EU countries closer to RF and TCC borders are the ones at higher risk for ASFV introduction by TAR, being Lithuania and Poland the countries at higher risk for ASFV entrance, followed by Finland, Estonia and Germany. Returning trucks is the TAR at highest risk for ASFV introduction into the UE, being almost three and six times more important than waste from ships and planes, respectively. This result is in agreement with the EU commission risk perception which recently approved a legislation  that strengthen and remind the importance of cleaning and disinfection for returning livestock vehicles coming from affected areas. In fact, the differences found in the results of the three scenarios for the disinfection of trucks (Figure 2) highlight the importance of that measure in preventing the entrance of animal diseases into free territories. On the other hand, ships waste has two times higher risk than plane waste. Ships waste has been recently suggested as the origin of the outbreaks in the Caucasus region , which may have influenced the opinion of the experts regarding their weights. Again we should highlight that the risk associated to ships and planes waste would depend not only on the release in the EU country, but on the final exposure, or contact, with susceptible populations, and this fact has not been measured on this work.
The analysis in detail of the results obtained for the different countries and the different analyzed pathways give us a better characterization of the risk. For example, in the case of Lithuania, although it has an overall high risk of ASF introduction by TAR, this risk is mostly associated to trucks, but not to waste from international ships or planes. These results are certainly influenced by the geographic location of the country, close to the current affected area, and the intense commercial relations with RF, which has been demonstrated by the amount of pig exports to this country. The opposite case is Germany that resulted in a medium risk for the overall TAR, but only a very low level for returning trucks. In this case the presence of most of the EU airports (50%, five over ten) that receive large number of flights from ASF-affected countries  determines the high level of risk associated to waste from planes. Similarly, Germany is a very important country in maritime trade (cargo ships, SSS and cruises) which explains also the high risk associated to waste from ships. Other countries with high risk associated to waste from planes are France and United Kingdom, where the two most important airports in terms of number of flights coming from ASF-affected countries (Charles de Gaulle and Heathrow, respectively) are located (Figure 5).
Another interesting result is related with the big differences found among the different ship types. Although Finland is the unique country that concentrates an average very high risk for waste from ships pathway, other countries are only highlighted when a specific type of ship movements is analyzed. For example, Bulgaria has an estimated very high risk by SSS movements through Black sea, particularly associated with the port of Burgas, the second most important port in the Black sea . Several countries surrounding the Baltic sea (Denmark, Finland, Lithuania or Sweden among others) are also highlighted in SSS movements and cruises, mainly due to their geographical closeness and trade with RF.
However, the most interesting results are related with waste from cargo ships, for which, four countries concentrate very high risk (France, Italy, the Netherlands and Spain). The detailed analysis of these movements, considering origin and EU destination ports, reveals some interesting differences among these countries. For example, in Netherlands, the risk both from RF as well as from Africa (mainly from Angola and Nigeria) is particularly concentrated in the port of Rotterdam, the one that receives the biggest amount of potentially risky cargo ships (those coming from ASF-affected countries) in the EU. However, in Spain the risk is more distributed, with two important ports receiving high number of cargo ships from Africa and other two from RF. In this particular case, these countries with the same level of risk present different profiles with one or several important ports in terms of risk. This fact enhances the importance of the detailed analysis of these results (represented in Figures 2, 3, 4 and 5) that could be much more informative to the policy makers than the general overview of the results (Figure 1).
Authors believe that this model has an important logical and biological approach as its results reflect areas and pathways identified at high risk by experts. This kind of models built using a simple and easy to understand methodology, are faster to develop and easier to interpret compared with the quantitative ones, and are particularly suitable when few information is available. For this reason this model may be considered as an adequate alternative in data scarce situations to provide a scientific support to risk managers, and ultimately, to prevent animal diseases introduction in free territories.