Predicting Meteorological Variables on Local Level with SARIMA, LSTM and Hybrid Techniques

The choice of holiday destinations is highly depended on climate considerations. Nowadays, since the effects crisis are being increasingly felt, need for accurate weather and services hotels crucial. Such a service could be beneficial both future planning tourists’ activities hotel managers as it help in decision making about expansion touristic season, due to prediction higher temperatures longer time span, thus causing increased revenue companies local sector. aim this work calculate predictions meteorological variables using statistical techniques well artificial intelligence (AI) specific area interest utilising data from an situ station, produce valuable reliable localised with most cost-effective method possible. This investigation will answer question suitable series single station that deployed location; our case, northern Crete, Greece. temporal resolution measurements used was 3 h forecast horizon considered here up 2 days. As techniques, seasonal autoregressive integrated moving average (SARIMA), AI like long short-term memory (LSTM) neural network hybrid combinations two used. Multiple input LSTM methodologies, temperature, relative humidity, atmospheric pressure wind speed, unlike SARIMA has variable. Variables divided into those present seasonality patterns, such temperature more stochastic no known speed direction. Two benchmark comparison quantification added predictive ability, namely climatological persistence model, which shows considerable amount improvement over naive methods, especially 1-day forecasts. results indicate examined methodology performs best at forecasts, closely followed by SARIMA, whereas better overall humidity forecast, even after correction model. Lastly, different methodologies discussed introduced further predictions.