Keywords: short-range weather forecast, local area, GPS, neural network, identification problem.

The GPS (Global Positioning System) is a satellite navigation system for ships and airplanes to find their own locations. It has recently come to be used in surveying in particular[1]. Errors due to the delay of GPS radio waves are inherent in the results of GPS surveying. Factors of observation errors are the locations of satellites, the errors of receivers and those of clocks on satellites, and the delay of radio waves due to the ionosphere, multipaths, vapor in the atmosphere, etc. All the factors but the vapor in the atmosphere can be removed to a certain degree. Because the vapor in the atmosphere is in a phantasmagoric dynamic state which cannot completely be grasped with the present technology, it is difficult to remove the errors due to the vapor-caused delay of radio waves[2]. On the other hand, research is being made to observe the vapor in the atmosphere by using the GPS[3]. The object of the present study is the development of a local short-range weather forecast system which makes use of the vapor-caused errors inherent in observation data by the GPS[4] and the neural network technology (Figure 23.1).

Firstly, in a preliminary examination, we selected the data items, where GPS data identified the weather. As a result of constructing a system using the data items, we have obtained from the preliminary examination, the "fine weather" and "rainy weather" items have been almost completely indentified. Secondly, we conducted the final examination considering with their results which we have obtained from the preliminary examination. In the final examination, the weather was indentified by the GPS data and weather factors. Measures were taken to grasp the maximum delay of GPS radio waves caused in the troposphere, and the system was constructed. The weather forecast system constructed in this study consists of four subsystems as shown in Figure 23.1. The first subsystem was for sunny weather; second, sunny/cloudy; third, cloudy; and fourth, cloudy/rainy. Each data set was first evaluated in the first subsystem. If it was not for sunny weather, it was forwarded to the second subsystem. If it was not for the sunny/cloudy weather, it was passed to the third subsystem. If it was not for cloudy weather, it was passed to the fourth subsystem. If it was not for the cloudy weather, it was considered to be for rainy weather. As a result, other than the "fine/cloudy weather" item we have been able to completely indentify all of the weather items.

It was ascertained in this study that local weather can be forecast in a short range with GPS data. To make the system developed in this study practical, however, its forecasting accuracy has to be improved further. The authors will raise the accuracy by using the data of the Geographical Survey Institute of the Ministry of Construction in Japan. Besides the authors will study further to develop a method of forecasting weather over a longer range.

1

T. Atsushi, H, Tsuji, "Easy, Simple GPS Surveying", Surveying Association of Japan, 1997.

2

G. Thierry, G. Blewitt, "The Effect of Weather Fronts on GPS Measurements" GPSworld, 1998.

3

N. Kunio, "GPS Meteorology; Measurement of Water Vapor", Parity, Maruzen Co.Ltd.,1998.

4

O. Ryu, "Application to GPS Meteorology", Navigation Society of Japan, 1999.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £78 +P&P)