Leveraging “Virtual Sensing” for Real-Time Analysis and Weather Forecasting

No matter where you are around the globe, you’ll always feel the presence of the weather. Sometimes you’ll feel it directly - such as when an ill-timed shower soaks you on the way to the subway. But increasingly in our digitally connected world, you’ll feel it indirectly as the result of critical infrastructure such as roads and telecommunications networks which themselves are impacted. ClimaCell is focused on turning this intuition - that our connected world is weather sensitive - on its head. In this talk, we share our vision and approach to leveraging a world full of “virtual sensors” and how we aim to build a new framework for monitoring and forecasting the weather. With this framework, we hope to both fill gaps in traditional weather observation networks (such as large portions of the world which have poor access to real-time weather data) and resolve detail in their blind spots (such as providing details about surface conditions below where radar beams are aimed).

Since ClimaCell was founded a few years ago, we’ve wrestled with many innovative techniques for “virtual sensing”. We focus on leveraging these types of measurements for two particular applications: real-time weather analysis and weather forecasting. Each application we pursue faces a unique set of challenges, but there are some commonalities. For instance, inferring signal from noise in very big, streaming data sets can be challenging - even more-so when we must reconcile different signals coming from a diverse set of traditional and non-traditional observations! There are many non-technical challenges associated with these data, too, such as gaining access to and acquiring creative new “virtual sensors.” Realizing our vision requires solving all of these problems, and then moving one step further - understanding how these unique data sources can augment or bolster existing weather monitoring systems.

Here, we demonstrate some of our technical and scientific solutions to leveraging “virtual sensor” data in real-world applications and products - from real-time monitoring of air quality and precipitation to regional forecasting. We also demonstrate, with the help of case studies, exactly how these novel data sources can improve weather analyses, either by enhancing the spatio-temporal resolution of the weather or by providing novel insights into surface conditions where no observation existed previously.