Scientists have proposed a new method to forecast precipitation over the western Himalayas by using real time satellite images, which they claim is more accurate and takes less time to prepare, Trend reports citing The Tribune.
Termed as the Artificial Neural Network (ANN) model, it extracts the pixel values corresponding to infra-red and water vapour images from the Kalpana-I satellite to generate precipitation forecast.
A geostationary satellite, Kalpana-1 satellite provides images in visible, infra-red and water vapour bands at an interval of every half an hour, which provide information about the cloud movement and water content present in the atmosphere over South Asia.
ANN has been developed by four scientists from the Defence Geoinformatics Research Establishment, Chandigarh, Defence Institute of Bio Energy Research, Haldwani and National Institute of Technology, Kurukshetra.
“The model is capable of interpreting satellite images in a generalised manner without biasing. Results show good correspondence between pixel values and observed precipitation,” the scientists have claimed in a research paper published recently by the India Meteorological Department (IMD).
Six locations, Haddantaj, Gulmarg, Dhundi, Drass Stage-II and Patsio, in various mountain ranges of the western Himalayas were considered for the study and computer software was used to develop an algorithm for extracting the pixel values. Relationship between the extracted pixel values and associated precipitation was modeled using a three-layer Artificial Neural Net.
A multiple regression model with the same input and output parameters for training and validating data was also developed for all the six locations and the results were compared with the ANN model. There were a total 1,012 data points, out of which 785 points were taken for training and 227 points were taken as independent data set for validation of the developed ANN model.
“Results show that ANN model has better skill to predict the precipitation at any specific location. Hence the model can be used as a good supporting tool for operational weather forecasts,” the paper states. The computational time is also comparatively less.
According to the researchers, accurate and timely prediction of weather has a large impact on our day to day activities, water resources, agriculture as well as the economy of the country.