Robust Hydrometeor Size Distribution Measurements – Are We Getting Closer?

It has long been recognized that different instrumentation metrologies and data processing approaches yield substantial differences in resulting hydroclimatic conditions (e.g. frequency of hail, total accumulated precipitation depth, hydrometeor number size distributions and terminal fall velocities) and, further, that those differences are a strong function of the prevailing hydroclimate.

This topic has found new relevance due to the emergence of wind turbine blade leading edge erosion (LEE) as a significant source of reduced wind turbine reliability and increased cost of energy from this green energy technology. LEE refers to material loss from the blades and causes decreased aerodynamic performance leading to lower power production and revenue and increased operations and maintenance costs. It is caused primarily by materials stresses when hydrometeors impact the rapidly rotating blades where the blade tip speed can be up to 100 ms^-1. The kinetic energy transferred by these impacts is a function of the precipitation intensity, hydrometeor size distributions (HSD), hydrometeor phase and the wind turbine rotational speed which in turn depends on the wind speed at hub-height. Of specific importance in this context is appropriate characterization of heavy precipitation events with large hydrometeors or hail coupled with strong wind speeds since these have the potential to cause the largest materials stresses. Better representation of hydrometeor properties and the joint probability distributions of those properties with wind speeds is urgently needed to quantify LEE potential and the financial efficacy of LEE mitigation measures at prospective and operating wind farms.

Here we revisit questions such as:

• How should we optimally analyze data from disdrometers?
• What level of agreement is and can be achieved across disdrometers and disdrometer types?
• What are the physical causes and functional dependencies of discrepancies between hydrometeor size distributions from disdrometers?
• What are the physical causes and functional dependencies of discrepancies between precipitation intensity from disdrometers and tipping-bucket or weighing rain gauges?

We present and analyze data from four sites that are located in very different climate regimes but that all have substantial wind energy resources and/or large numbers of wind turbines operating nearby. The sites and operating technologies are as follows:

• The Risø-DTU site in Denmark, northern Europe. There are two optical disdrometers; Thies LPM and an OTT Parsivel² operating at the base of a 125 meteorological mast along with a RIMCO tipping bucket rain gauge (tip resolution of 2mm) and a sonic anemometer.
• The National Institute of Advanced Industrial Science and Technology (AIST) test facility in Japan. At this site there is an OTT Parsivel² and tipping bucket rain gauge operating with a cup and a sonic anemometer and a nacelle-mounted lidar for wind speed measurements. Additionally, the AIST team are performing experiments using an OTT Parsivel² within an R&D Rain Erosion Test system where flow conditions and droplet size distributions can be highly controlled.
• The Cornell University site in upstate New York, USA. There are four co-deployed OTT Parsivel² disdrometers and an OTT microrain radar. A further experiment is just commencing using wind shielding of one of the disdrometers and co-deployment of shielded tipping bucket rain gauge.
• The Department of Energy Atmospheric Radiation Measurement site in Lamont in the US Southern Great Plains (SGP). Measurements include an OTT Parsivel² disdrometer, a Joss-Waldvogel type impact disdrometer, and a by Joanneum Research Digital video disdrometer, along with a Pluvio-2 Weighing Bucket Precipitation Gauge and wind-speed measurements from a Doppler lidar.

We begin by showing an example of the critical importance of hydrometeor size distributions and phase to estimated wind turbine blade lifetimes using multi-year records from the SGP site. We show that screening data when the observed fall velocity (vf) from the OTT Parsivel² lies beyond ±50% (±60%) of the terminal fall velocity estimated by Gunn and Kinzer leads to exclusion of 29% or 24% of reported hydrometeors. Use of these filters versus no vf filtering changes the amount of kinetic energy transferred to the blades by an order of magnitude. Blade lifetimes calculated based on hydrometeor size distributions from the three co-located disdrometers vary by a factor of two.

We then continue by showing results from a common set of analyses applied to these data sets to develop best practice for measurements in the context of wind turbine blade leading edge erosion. Initial results from the SGP suggest discrepancies between HSD from the disdrometers and for total accumulated precipitation relative to the rain gauge are a function of wind speed and are instrument dependent. For example, the influence of wind speed appears to be most pronounced in data from the optical disdrometer. Further, discrepancies between the HSD from the disdrometers and accumulated precipitation relative to the tipping bucket rain gauge is largest during the most damaging events (i.e. high intensity rain and/or during events with hail). Using data sets from the four sites we quantify the degree to which closure in HSD from different disdrometers and accumulated precipitation relative to rain gauges can be improved. We further decompose the discrepancies, particularly during periods which significantly contribute to LEE, to quantify the relative importance of (i) the horizontal wind speed and the resulting translational component of hydrometeor fall patterns, (ii) atmospheric turbulence at the site and specifically the vertical component of the flow as it influences the fall velocity, (iii) precipitation rate as a cause of mis-counting of hydrometeors, and (iv) hydrometeor phase.