A Fundamental Study of Neural Network System for Conjecturing Weather in Local Area Using GPS

The GPS is used in surveying recently. The results of GPS surveying contain errors due to the delay of GPS radio waves caused by vapor in the atmosphere. Developed in this study is a system which forecasts local weather by making use of GPS data under the influence of the vapor. The neural network technology was used to relate GPS data to weather conditions. In a preliminary experiment, the simulation of weather forecasting was carried out to select GPS data items to be used in the main experiment. A pilot system was constructed on the GPS data items of the preliminary experiment. The pilot system forecast sunny and rainy weathers correctly. Then the main experiment was carried out, the results of the preliminary one taken into account. In the main experiment, meteorological factors in addition to GPS data were used to construct a forecasting system, which forecast all weathers correctly except a mixture of sunny and cloudy weathers.