Analysis of Malta meteorological data collected by the OMRG real-time stations

Abstract

Everchanging meteorological conditions have an impact on the state of the world’s ocean. There is an ongoing need to capture those small differences to provide useful information to many economic sectors, including marine navigation, offshore operations, and others. Theoretical background of the meteorological parameters is fundamental to reflect upon. It is crucial to consider the functionality of measuring devices and recognise their limitations. This dissertation aims to provide an assessment of the robustness of the meteo stations located in Cirkewwa and Qammieh in Malta, and in Mgarr in Gozo. The objectives of the research include establishing the variability between parameters and stations, quality control of the datasets, and investigating the potential for using wind speed records from one station to forecast values at other stations. The analysed datasets were recorded over around two years. Site specifications and exposures of these three stations have been compared. Quality control over the initial datasets was also performed. The following parameters have been investigated: wind speed, wind gust, wind direction, air temperature, atmospheric pressure, and humidity. Statistical measures tailored for meteorological variables to be robust and resistant were calculated to determine the nature and structure of the datasets. Moreover, the mathematical correlations between parameters and stations have been estimated. Both the geographical positioning and statistical measures of parameters showed certain differences between meteo stations. The latter data manipulations and further the sample cross-correlation function were used as a preliminary analysis for wind speed prediction models. Furthermore, the possibility of using wind speed records from one station to predict other stations values, were explored. Based on the analysis of exposure of the stations and variability between stations’ parameters, the station in Qammieh differs significantly from the ones in Cirkewwa and Mgarr. As a result, the prospect of forecasting wind speed at Qammieh does not seem feasible. Linear regression and artificial neural network model (ANN) have been employed as forecasting methods for the pair Cirkewwa and Mgarr. The ANN model results indicated a higher level of accuracy of wind speed predictions for Cirkewwa and Mgarr than the linear regression. It shows to be flexible and suitable at capturing stochastic wind speed behaviour. In conclusion, the correlations between parameters and stations have been evaluated and limitations of the prediction methods have been addressed.