An evaluation of the uncertainty of precipitation measurements from optical sensors at a Norwegian mountainous site.

In remote areas with high mountains or challenging weather conditions, ground-based precipitation measurements cannot be performed by instruments that require extensive maintenance. Still, in-situ measurements are essential to validate model predictions or remote sensing measurement methods. Optical sensors (disdrometers and present weather sensors) constitute a good alternative to traditional methods, requiring little maintenance and having no moving parts that could deteriorate. However, these devices are calibrated in laboratory conditions that can be very different from the ones they experience in the field. Therefore, it is essential to improve our understanding of the reliability of these instruments when they operate in a challenging environment.

Users who rely on optical instruments are primarily interested in high-frequency reports of precipitation type (e.g. public road management, aviation) and precipitation rate (e.g. hydropower systems, agriculture). These variables are derived from semi-empirical knowledge together with measurements of hydrometeors’ sizes and fall velocities, that can both be affected by the wind or other instrumental systematic biases. 

In this work, we analyze data collected at the former WMO-SPICE site of Haukeliseter in Telemark, Norway. This station is operated by MET Norway. Two to three models of different popular optical instruments (OTT Parsivel2, Thies Clima LPM, Vaisala PWD12 and PWD22), unevenly exposed to the wind, have been deployed there since September 2023, providing two winters of precipitation data to analyze. Haukeliseter is located in a mountainous area commonly experiencing strong winds and is covered with snow for about 6 months a year, making it an excellent location to study solid precipitation.

We perform a systematic intercomparison of these instruments to evaluate their level of agreement and, in turn, quantify their accuracy. A reference for the precipitation rates consisting of a Geonor rain gauge placed in a standard WMO-defined DFAR (double fence automated reference) setup is available. There exists no similar standard field reference for precipitation type detection. Where possible, human observations are used as a benchmark, but they are often available at a much lower time resolution than automatic measurements. To compensate for the lack of such long-term observations at Haukeliseter, a campaign of on-site high-frequency human observations of the precipitation type performed in early 2025 is used as a comparison reference. Preliminary results of the intercomparison and analysis from this winter’s measurement campaign will be presented.