Study of the impact of climate change on the spatial distribution of dengue cases in Rio de Janeiro

Abstract

Introduction There is evidence that the distribution of dengue cases is correlated with environment variables such as temperature, precipitation and others. This study aimed to select the best variables for modelling the number of dengue cases in Rio de Janeiro city, and used data from 2015 to 2016 as comparison for results. Methods The number of cases was obtained by the Brazilian health system and the variables were mined from public satellites data, separating Rio de Janeiro city in 21-pixel blocks in image data. The data gathered from 2015 to 2016 was divided in 8-day periods, where the mean of each variable in this period was stored. The study made a comparison using the following variables and their respective lags in time: NDVI, day temperature, night temperature, precipitation, population size, latitude and longitude. The comparison between them was made using the one who fitted more a Generalized Autoregressive Log-linear Model with a negative binomial family to predict the dengue cases. Results The most significant predictors inside the model were temperature at night, precipitation three weeks before, and dengue cases from one week before. Using only these three variables and an intercept, the model showed an error mean of 0.3% and a standard deviation of 11% in a prediction against real value comparison. Conclusions This work is a first approach to understanding the determinants of dengue and only using public data. The next steps will be to extend both the temporal and spatial reach of the analysis.