Smartphone Pressure Data: Statistical Characteristics, Bias Correction, and Applications in Weather Forecasting

Smartphones are now commonly used, with ownership ranging from 30 to 70 percent of the population in numerous countries. With GPS and pressure sensor installed, smartphones can measure ambient air pressure, and have the potential to be used as meteorological observations. Our recent studies have shown that the pressure of smartphones can be used to describe the strength and location of large-scale vortex, as well as mesoscale surface high and low during thunderstorm events. Nevertheless,  addressing the quality issues of smartphone barometric data owing to human activity and other factors remains a great challenge. Our study indicated that a machine-learning technique could significantly improve the usability of the smartphone data. Since smartphone users are mostly concentrated in urban areas, significantly higher-density pressure coverage is achieved than the conventional surface networks in large cities. However, can these high-density data improve the forecasting ability of weather-related disasters? This presentation will introduce the barometric pressure data observed by smartphones collected from Moji weather app, the bias correction process, and its applications in analysis and forecasting via case studies.

We used tropical cyclones (TCs) Lekima in 2019, Hagupit in 2020 and In-fa in 2021 as examples to conduct bias correction on labeled smartphone pressure data from the Moji Weather app. A quality control procedure was proposed utilizing random forest machine learning models. By applying this quality control approach to the selected TCs, we discovered that the performance of the method for labeled data significantly surpassed that for unlabeled data developed in a previous study, reducing the mean absolute error from 3.105 to 0.904 hPa.

The smartphone (SPO) and traditional weather stations (TWS) pressure data during a hailstorm that occurred on 30 June 2021 in Beijing, are assimilated into a one-hour frequent 3DVAR system based on WRF model, respectively. The results demonstrate that the spatial density of SPO data is tens of times larger than that of TWS. Compared with the control experiment and assimilation experiment which only use the TWS data, the assimilation experiments with SPO data show significant improvements, in terms of the cold pool and gust front. The improvement of FSS in 10 mm hourly precipitation reached to 3% to 12%. Our findings indicate that high-resolution SPO data have the potential to enhance the forecasting capabilities of mesoscale convective system.