The constructed model was then validated by dividing the data file into two data sets. The data between January 2000 and December 2008 were used to develop a DLM and those between January 2009 and December 2010 were used to validate the model. The validation indicates that the model had reasonable accuracy as the observed and predicted values were mostly coincided (Supplementary Fig. 1). In addition, adequacy of the model predicting outbreak (≥168) was evaluated by sensitivity analyses. For sensitivity analyses, the monthly number of dengue cases was transformed into a categorical variable (i.e., outbreak/ non-outbreak). Then the sensitivity or true positivity rate (predicted number of months with dengue outbreak/observed number of months with dengue outbreak) and specificity or true negativity rate (predicted number of months without dengue outbreak/observed number of months without dengue outbreak) of the predictive model were calculated.