A framework for the collection, aggregation, collation and prediction of meteorological data

Abstract

Nowadays, there are several different sources (weather stations, sites) of meteorological data, regarding a specific geographical area, however it remains uncertain and unclear how much these data converge. One approach to solve this particular problem is to collect meteorological data from different stations for the same geographical area, visualize and aggregate them in a time series analysis. The above approach was used for the city of “Tripoli” – Peloponesse – Greece, and after collecting data from seven different local meteorological stations, we concluded that there are clear differences on climate parameters (temperature, humidity, wind, precipitation, etc.) between the different stations, even though these stations are very close to each other. Αlso, using the historical data that have been collected, we train and deploy a short-term temperature prediction model, using a biredirectional LSTM models. The deployed model proves to be accurate in the prediction of temperature for the specific area, demonstrating its efficiency for short-term predictions of temperature values.