Assessing the Role of Meteorological Factors on Human Health in India and its Extended Range Prediction

There are various complex ways in which the climatic factors (e.g. temperature, precipitation, humidity, extreme weather events and sea-level rise) can have direct or indirect impacts on human health. Changes in local climatic factors alter the ecological conditions which can affect vector ecology and the appearance of various infectious diseases, namely malaria and diarrhoea. According to the World Health Organization (WHO), the worldwide number of deaths due to climate change is rapidly increasing, either indirectly through diseases like malaria, diarrhoea, malnutrition and respiratory diseases, or directly through extreme weather conditions like droughts, heat waves and floods. Better monitoring and long-term planning for health systems and communities to become more climate resilient as well as to react appropriately and in a timely manner to the negative effects of climate change on health are required. To inform policies, an estimation of the approximate magnitude of the health impacts of climate change is needed through an appropriate early warning system. For this purpose, the extended-range climatic forecasts will be very helpful to reduce health impacts through the development of early disease warning systems in 2-3 weeks in advance.

In the present study we have tried to build up suitable relationships between the diseases, malaria (MAL) and acute diarrheal diseases (ADD), for two districts in the state of Maharashtra, India and the meteorological parameters from the weather monitoring stations present in the region. A suite of weather metrics have been constructed using the standard meteorological factors (e.g. maximum/minimum temperatures and rainfall) based on available weather monitoring data based on Self-Organizing Map (SOM) technique. SOM is basically a pattern recognition technique based on unsupervised learning neural networks (i.e. without prior knowledge of the data domain). Giving an N-dimensional data space (i.e. input variables), the SOM algorithm distributes an arbitrary number of nodes in the form of a 1-D or 2-D regular lattice. Each node is uniquely defined by a reference vector consisting of weighing coefficient and therefore different datasets are normalized before using in SOM as they have very different mean and distribution. The adaptation of the reference vector in accordance with the input vector is done through minimization of Euclidean distance between the reference vector for any particular node and the input data vector. Two separate SOM classifications (one for ADD and other for MAL) have been done in 3x3 lattice (total 9 classes). Total 10 variables (4W, 2W, 1W values of R/F, TMx, TMn and 1W value of ADD or MAL) as input to SOM have been standardized. The weekly total ADD and MAL data is available from 2012 to 2016 with some missing data. Total 465 data points (205 data points are available for Pune and 260 for Nagpur) are given for classification in the case of Malaria and 516 data points (256 data points for Pune and 260 for Nagpur) are given for classification in the case of ADD.

The Multi-Model Ensemble (MME) Extended Range Prediction (ERP) system (based on Climate Forecast System (CFS) version 2) of Indian Institute of Tropical Meteorology (IITM), which is now operational in India Meteorological Department (IMD), is skilful up to 4^th pentad lead (each pentad equals to 5 days) in predicting rainfall/temperatures during the relevant seasons and hence can be used for operational purposes. The ERP system consists of different variants of the Climate Forecasting System (CFS), such as: (i) CFSv2 at T382 (ii) CFSv2 at T126 (iii) GFSbc (the stand-alone atmospheric component - GFSv2, forced with bias corrected SST from CFSv2) at T382 and (iv) GFSbc at T126, all having 4 ensemble members each.

So, after developing the suitable relationship between weather and climatic factors and their impacts on human health, an attempt is made for the real time monitoring and prediction of such conditions with a lead time of 2-3 weeks. A deterministic as well as probabilistic forecast approach for the prediction of different categories (% probability of Below Normal, Near Normal, Above Normal and Extreme occurrences) of infectious diseases (like malaria and diarrheal diseases) in sufficient lead time is taken up using ERP system developed at IITM. The SOM clustering technique and the design of the prediction system and strategy adopted for ERP of such infectious disease outbreaks will be discussed in details. Some of the important events of occurring of Malaria and ADD for these two different districts will also be highlighted.

References:

Luber G et al. 2014, Ch. 9: Human Health. Climate Change Impacts in the United States: The Third National Climate Assessment.

Kohenen T. 1990 and 1997. The Self Organizing Maps

Joseph, S. et al. 2017, IITM Res. Rep., RR-137

Abhilash et al. 2014, Atmos. Sci. Lett.

Abhilash et al. 2014, Int. J. Climatol.

Abhilash et al., 2015, JAMC

Sahai et al. 2013, Current Science

Sahai et al. 2015, Climate Dynamics