Estimating Ixodes ricinus densities on the landscape scale.
Bottom Line:
Calibrating the Poisson regression models with these nymphal densities results in an explained variance of 72 % and a prediction error of 110 nymphs/100 m(2) in 2013.The accuracy of the mapped tick densities was investigated by leave-one-out cross-validation resulting in root-mean-square-errors of 227 nymphs/100 m(2) for 2013 and 104 nymphs/100 m(2) for 2014.Finally, the study is a first step towards the spatial estimation of tick-borne diseases in Central Europe.
View Article:
PubMed Central - PubMed
Affiliation: Institute of Geography and Geoecology, Karlsruhe Institute of Technology, Reinhard-Baumeister-Platz 1, 76131, Karlsruhe, Germany. denise.boehnke@kit.edu.
ABSTRACT
Background: The study describes the estimation of the spatial distribution of questing nymphal tick densities by investigating Ixodes ricinus in Southwest Germany as an example. The production of high-resolution maps of questing tick densities is an important key to quantify the risk of tick-borne diseases. Previous I. ricinus maps were based on quantitative as well as semi-quantitative categorisations of the tick density observed at study sites with different vegetation types or indices, all compiled on local scales. Here, a quantitative approach on the landscape scale is introduced. Methods: During 2 years, 2013 and 2014, host-seeking ticks were collected each month at 25 sampling sites by flagging an area of 100 square meters. All tick stages were identified to species level to select nymphal ticks of I. ricinus, which were used to develop and calibrate Poisson regression models. The environmental variables height above sea level, temperature, relative humidity, saturation deficit and land cover classification were used as explanatory variables. Results: The number of flagged nymphal tick densities range from zero (mountain site) to more than 1,000 nymphs/100 m(2). Calibrating the Poisson regression models with these nymphal densities results in an explained variance of 72 % and a prediction error of 110 nymphs/100 m(2) in 2013. Generally, nymphal densities (maximum 374 nymphs/100 m(2)), explained variance (46 %) and prediction error (61 nymphs/100 m(2)) were lower in 2014. The models were used to compile high-resolution maps with 0.5 km(2) grid size for the study region of the German federal state Baden-Württemberg. The accuracy of the mapped tick densities was investigated by leave-one-out cross-validation resulting in root-mean-square-errors of 227 nymphs/100 m(2) for 2013 and 104 nymphs/100 m(2) for 2014. Conclusions: The methodology introduced here may be applied to further tick species or extended to other study regions. Finally, the study is a first step towards the spatial estimation of tick-borne diseases in Central Europe. No MeSH data available. Related in: MedlinePlus |
Related In:
Results -
Collection
License 1 - License 2 getmorefigures.php?uid=PMC4536605&req=5
Mentions: The results of the field study were summarised in Table 2. For each of the 25 sampling locations described in Table 1, the total number of I. ricinus nymphs flagged monthly during 1 year is given together with the air temperatures for 2013 and 2014 and the long-term mean of the relative humidity, both extracted from gridded data provided by the German weather service (Deutscher Wetterdienst, DWD). Additionally, the calculated saturation deficit and land cover classifications are listed. Table 2 shows a wide range of values for the flagged nymphs. In 2013 for example, nymphal densities range from N = 0 at the mountain site FB (Feldberg) or N = 3 at AH (Allerheiligen) to N = 1,066 at BT (Botnang). Thus, for an optimal representation nymphal densities were plotted on a logarithmic scale. Solving the Poisson regression model to predict nymphal densities for 2013 and 2014 using the gridded explanatory variables results in the scatterplots depicted in Fig. 3. The Poisson regression model performs slightly better with the explorative variables of the year 2013. Looking at McFadden’s pseudo Rp2, a total of 71.7 % of the variation in the observed nymphal density was explained by the modelled nymphal density in 2013. In contrast, only 46.1 % of the explained variance was estimated for 2014. The regression coefficients β for both models are summarised in Table 3. While for 2013 all of the explanatory variables contribute highly significantly to the performance of the model, for 2014 only three variables depict a significant contribution. Further, a collinearity of some variables was estimated. Although this collinearity does not influence the explained variances, it leads to regression parameters, which are not generally applicable (see the discussion in the next section).Fig. 3 |
View Article: PubMed Central - PubMed
Affiliation: Institute of Geography and Geoecology, Karlsruhe Institute of Technology, Reinhard-Baumeister-Platz 1, 76131, Karlsruhe, Germany. denise.boehnke@kit.edu.
Background: The study describes the estimation of the spatial distribution of questing nymphal tick densities by investigating Ixodes ricinus in Southwest Germany as an example. The production of high-resolution maps of questing tick densities is an important key to quantify the risk of tick-borne diseases. Previous I. ricinus maps were based on quantitative as well as semi-quantitative categorisations of the tick density observed at study sites with different vegetation types or indices, all compiled on local scales. Here, a quantitative approach on the landscape scale is introduced.
Methods: During 2 years, 2013 and 2014, host-seeking ticks were collected each month at 25 sampling sites by flagging an area of 100 square meters. All tick stages were identified to species level to select nymphal ticks of I. ricinus, which were used to develop and calibrate Poisson regression models. The environmental variables height above sea level, temperature, relative humidity, saturation deficit and land cover classification were used as explanatory variables.
Results: The number of flagged nymphal tick densities range from zero (mountain site) to more than 1,000 nymphs/100 m(2). Calibrating the Poisson regression models with these nymphal densities results in an explained variance of 72 % and a prediction error of 110 nymphs/100 m(2) in 2013. Generally, nymphal densities (maximum 374 nymphs/100 m(2)), explained variance (46 %) and prediction error (61 nymphs/100 m(2)) were lower in 2014. The models were used to compile high-resolution maps with 0.5 km(2) grid size for the study region of the German federal state Baden-Württemberg. The accuracy of the mapped tick densities was investigated by leave-one-out cross-validation resulting in root-mean-square-errors of 227 nymphs/100 m(2) for 2013 and 104 nymphs/100 m(2) for 2014.
Conclusions: The methodology introduced here may be applied to further tick species or extended to other study regions. Finally, the study is a first step towards the spatial estimation of tick-borne diseases in Central Europe.
No MeSH data available.