Developing Cloud Computing's novel Computational methods for improving long-term weather global forecast

Abstract

Weather data mining methods and forecast algorithms have been of long standing interest. Recent research based on the global satellite data and special synergetic methods showed possibility of the long-term (up to half a year ahead) forecast with up to 10 % average mistake (standard is 20 %). Particularly, the average daily air temperature forecast’s mistake is up to 6.5 % for Skopje Airport (half a year ahead). This approach is characterized by the final linear difference equations’ simplicity and the high computational complexity of the above equations reasoning. The cloud computing web-site’s prototype was developed (weatherforecast.tk). Main research proposals: improving the user interface based on 3D or/and ubiquitous computing technologies; developing new synergetic methods for the appropriate realization in the multithread cloud application, including the code and data parallelization; increase of the forecast parameters’ quantity (e.g., precipitation). This paper main results are: precipitation’s long-term (up to half a year ahead) forecast has very low quality now, and, therefore, it is not recommended for practice; the forecasting places’ quantity is changed modifying the text file in the cloud application’s package; the web-site http://weatherforecast.tk user interface was enhanced using 3D Chart diagram.