With rain-related hazards, it is difficult to forecast and prepare for it due to the decline in the availability and reliability of global daily weather reports. Therefore, we need to make use of available unconventional commercial weather instruments to provide supplementary information to existing systems. A more accurate and reliable weather forecast can then be made due to the prompt availability of information from ubiquitous devices. As smartphones become a widely used device, we propose a conceptual design of a multi-device ground weather observation network using smartphones and other sensors. In this paper, we first investigate the differences between smartphone-based sensors and other sensors to determine issues to address for big data on global weather information. In our experiments, we found that the data quality differs among devices in a very small area of 100 m grid.