Forecasting the Water Vapour Distribution Over India Using Artificial Neural Networks

Atmospheric water vapour plays an important role in radio communications for both terrestrial and earth space communication systems as well as for global climate change studies. Hence the water vapour data and its information plays vital role for scientists and researchers to evaluate the hidden patterns, trends for analysis and forecasting. In general radiosonde observations are the primary source of upper air water vapour data and are being used for estimation of different atmospheric parameters; on the other hand, different statistical and scientific methods are using to process the meteorological datasets in water vapour studies. However data assessment and usage is critical and challenging with many data level limitations to mitigate the accurate results. Currently there are extraordinary prospects in the Information Systems to process these data and explore. The data mining applications are the most promising features for radiosonde water vapour data forecasting easily and efficiently. As a part of this study, radiosonde observations over 34 Indian stations pertaining to a period of 17 years from 1997 to 2014 obtained from the British Atmospheric Data Centre(BADC) are pre-processed from the different met parameters of air temperature, dew point temperature and pressure levels for water vapour concentrations estimation at different height (pressure levels). This paper converse various analytics of surface level water vapour distribution patterns and trends with validation of the forecasting model based on the current trends by using improved version of artificial neural network (ANN) method. The proposed ANN model has been trained with training data and tested with test data for accuracy and better performance. The forecasted results are presented comparatively with the actual trends in the form of Physiographic Divisions, annual and seasonal charts water vapour distribution over India for better decision supporting systems