Relationship between meteorological variables/dust and the number of meningitis cases in Burkina Faso (original) (raw)

2017, Meteorological Applications

Using four meteorological variables (northeasterly surface wind (WS), relative humidity (RH), rainfall (Rain) and temperature (T 2m)) and one of four dust products (dust surface mass concentration or aerosol optical depth, D1-D4) over Burkina Faso, a differential equation for meningitis incidence (N) was applied to the multivariate log-linear regression analysis to get each contribution from the variables (WS, RH, Rain, T 2m and one of four dust products) to N. The climatological data show that dust and temperature are synchronized with meningitis incidence, but the meningitis incidence reaches a peak several months after the northeasterly wind becomes maximum and the relative humidity becomes minimum during the no-rain period. That is, meningitis incidence increases when the northeasterly wind prevails under dry and no-rain conditions and decreases when the southwesterly wind prevails under wet and rain conditions, and it has a peak under dusty and hot conditions. After performing all possible combinations of the regression analysis (but choosing only one dust dataset for each combination) using models with one to five parameters, the time derivative of the weekly meningitis incidence from 2006 to 2014 was estimated and compared with that observed. The more parameters that are included, the higher the correlation coefficients between the estimated and observed tendency. However, the northeasterly wind has a major contribution to the rate of change of the number of cases. The highest correlation coefficient was for the models with all four meteorological variables plus the dust surface mass concentration data. Even in one-or two-parameter models, a maximum correlation coefficient of 0.666 is obtained for the WS model, and the WS + RH model gives a maximum of 0.754, which indicates some forecast skill using surface wind and relative humidity data. Although the modelled derivative underestimated the outbreaks in 2006 and 2007, it correctly simulated the timing of the zero crossing of the weekly rate of change of N. Thus, this approach may be useful to identify the timing of the peak season of meningitis in Burkina Faso.