Experiences in using low-cost sensors to observe road weather

Road weather information is essential for keeping the roads well maintained and safe during wintertime. Main source of road weather observations are road weather stations, but IoT (Internet of things) sensor technology provides new ways to observe road weather. Finnish Meteorological Institute (FMI) and Fintraffic Road are studying whether such IoT technology could help increase spatial density and/or improve coverage in the observation network and whether these additional observations could also be used to improve road weather forecasts. Around 100 autonomous battery-operated low-cost IoT sensors based on LoRaWAN communication technology were installed into the roadside area of a motorway in southern Finland and at the Sodankylä airport test track during winter 2020. Most of the sensors were of the types UC11-T1 from Ursalink and ELT-2 from ELSYS AB, but there were a few MCF-LW12TERWP sensors from MCF88 as well. All sensors measure air temperature and humidity and the MCF sensors also measure air pressure. Some of the sensors were installed at a weather station and some at road weather stations to enable data comparison with reference stations. During wintertime the IoT sensors’ air temperature measurements correspond rather well to the reference measurements. However, during other times of the year the solar radiation often causes warm bias to the measurements. The bias is reduced when the sensors are installed inside radiation shields. However, the reliability of the IoT devices needs improvement, as several sensors stopped working during the measurement campaign. This was probably caused by a firmware bug, that led to excess power consumption and emptying of batteries in some of the devices.

The FMI road weather model uses surface temperature observations in the model initialization to improve the forecasts. As the model surface temperature is forced to the observed surface temperature, the air temperature measurements don’t have that much effect in the initialization. When there are no surface temperature observations available at the forecast location, the model uses values interpolated from road weather station observations. The interpolation is done with the universal kriging method, where elevation is used as an explanatory variable. In this project we studied whether air temperature observations from IoT sensors could be used as explanatory variable as well. The results thus far show that use of air temperature observations from road weather stations improves the interpolated surface temperature values at least in some situations. However, this is rather location dependent. Initial results suggest that IoT observations would be useful this way as well. According to the results, IoT observations show potential to improve road weather monitoring and forecasting, but more studies are still needed.