Crowd-Sourced Environmental Sensing Using Mobile Devices on Land and Sea

Smartphones now include onboard sensors for direct measurement of environmental conditions such as barometric pressure, ambient temperature, and relative humidity. Other sensors such as the accelerometer, light sensor, camera, and microphone can be used to estimate additional surface weather conditions such as precipitation, wind speed, and wind direction. These sensors can also be used to estimate essential operational conditions on land and at sea such as mobility, visibility, and wave modes. Collection, assimilation, and distribution of such real-time observations from a broadly distributed and spatially dense network of mobile devices would lead to improved awareness of current weather conditions, improvement of maps of current and future conditions, data assimilation of numerical weather forecasts, and more accurate forecasting at regional, national, and global scales. However, measurement and collection of these observations should require minimal user interaction and not interfere with the primary intended uses of the mobile device.

This paper will review the preliminary development of a mobile-based software framework (WeatherCitizen) for collecting crowd-sourced environmental observations on land and sea. The application enables the collection of crowd-sourced environmental observations through different modes of operation, including automated polling of available weather sensors, manual entry of standardized observations, and semi-automated collection of audio and video. Observations are uploaded to a geospatially aware cloud database and can optionally be published to a user’s social media feed for wider dissemination. The cloud database aggregates user observations and sources ground truth observations from other open weather APIs. The cloud database serves each data set via a flexible and robust web application programming interface (API). The API is demonstrated in a queryable map-based visualization showing geographically distributed mobile device data alongside weather observations from other sources.

We will describe several methods for estimation of weather conditions based on this data, including prior published approaches by others in this area. We then assess different data reduction strategies such as crowd-sourcing, machine learning, and physics-based modeling to validate and filter the disparate data sources. We then provide a live demonstration of our prototype smartphone-based application interfaced to our prototype cloud server. Finally, we discuss our preliminary experience utilizing this information, including data assimilation for improved regional weather forecasting on land and sea, and crowd-sourced situational awareness of weather and terrain conditions in the battlespace using soldier-worn mobile devices.