Belgian honey bee winter mortality during 2012-2013: a case-control study and spatial analysis

• Published in: Journal of apicultural research Vol. 55; no. 1; pp. 19 - 28
• Main Authors: Roelandt, Sophie, Méroc, Estelle, Riocreux, Flavien, de Graaf, Dirk C, Nguyen, Kim Bach, Brunain, Marleen, Verhoeven, Bénédicte, Dispas, Marc, Roels, Stefan, Van der Stede, Yves
• Format: Journal Article
• Language: English
• Published: Taylor & Francis 01.01.2016
• Subjects: case-control; cluster analysis; EPILOBEE; honey bees; logistic regression; veterinary surveillance; winter mortality
• ISSN: 0021-8839; 2078-6913
• DOI: 10.1080/00218839.2016.1201945

An increase in honey bee mortality has been reported within and outside the European Union. A Belgian honey bee health monitoring was started in 2012 based on a standardized and pan-European voluntary surveillance program (EPILOBEE). The main objective was to estimate honey bee mortality during winter and the apicultural season using a harmonized survey. For the Belgian study, the guidelines were adapted according to small-scale Belgian beekeeping practices and previous pathogen prevalence studies. A two-stage sampling stratified by province resulted in 150 apiaries selected from a sampling frame of approximately 3,000 registered beekeepers. These apiaries were visited twice, questionnaires were completed, collecting information on risk factors and mortality. Samples systematically taken in autumn were screened / quantified for Varroa destructor mites. Weighted colony winter mortality rates were estimated per individual apiary and for geographical entities. We then attempted to identify risk factors in a case-control data interpretation, with observed mortality as a binary dependent variable. Questionnaire variables were evaluated in univariable logistic regression. The final multivariable model was retained: age of beekeeper, wanting to continue beekeeping, increasing number of surrounding landscapes, average colony varroa infestation level, all associated with increased mortality. Chemical acaricide treatment before 1 September and Thymol-containing acaricide were associated with lower mortality. The predictive accuracy of this final model was estimated by the area under the curve (AUC) and was 80.70%. The final predicted mortality risk was visualized on a map as an interpolated layer and compared with the observed mortality; clusters of high/low mortality were identified. This analysis has generated further hypotheses and highlighted where the case-control study could benefit from increased sample size.