Claudia Sahd

Brazil

Climate-driven dengue forecasting: SARIMA/SARIMAX models and web application

Claudia Stoeglehner Sahd1, Alison Henrique Marcelino1, Elisângela Aparecida da Silva Lizzi1

Mathematics Department, Graduate Program in Bioinformatics, University Tecnology Federal of Paraná – (UTFPR), Paraná, Brazil

Abstract

Background

Climate change has intensified arboviral transmission by altering temperature, humidity, and precipitation patterns that favor the reproductive cycle of Aedes aegypti. In this context, predictive modeling tools are essential for anticipating dengue outbreaks and supporting strategic decision-making in public health. This study analyzed dengue incidence in Londrina (Paraná, Brazil) from 2013 to 2024, integrating weekly and monthly epidemiological records with climate variables such as minimum temperature, precipitation, and humidity. Stochastic time-series models from the Box-Jenkins framework–specifically SARIMA and SARIMAX–were applied, considering performance, predictive capacity, and suitability to the seasonal dynamics of the disease.

Methods

Data were obtained from the InfoDengue system and official meteorological databases. For the monthly series, the SARIMA (2,1,2)(1,1,1)12 model provided the best fit, capturing seasonality, structural trends, and atypical oscillations observed during the 2020 pandemic period. For the weekly series, the SARIMAX (2,1,8)(1,1,1)52 model–incorporating precipitation from the two preceding weeks and minimum temperature from the three preceding weeks–showed superior performance, with residuals meeting theoretical assumptions and accurate out-of-sample forecasts. These results highlight the relevance of climate variables in the temporal behavior of dengue, particularly amid environmental changes.

Results

As an applied product, a user interface was developed as a Shiny-based web application, integrating the full analytical back-end of the models and enabling automated analyses through an intuitive interface. The tool allows visualization of historical series, autocorrelation plots, parameter selection, model execution, and real-time forecasting, facilitating use by health managers, researchers, and practitioners. Its open-source implementation enhances adoption and customization potential.

Conclusions

Results demonstrate that SARIMA and SARIMAX models are effective and robust approaches for forecasting dengue cases in urban contexts influenced by climatic variability. Integrating advanced statistical modeling with an accessible web application represents a significant step forward for dynamic surveillance and outbreak anticipation in the face of climate-driven impacts on arboviral diseases.