The verification of lightning location accuracy in Finland deduced from lightning strikes to trees Antti Mäkelä a, ⁎, Jakke Mäkelä a , Jussi Haapalainen a , Niko Porjo b a Finnish Meteorological Institute, Helsinki, Finland b ArcDia Oy, Turku, Finland abstract article info Article history: Received 3 July 2015 Received in revised form 16 November 2015 Accepted 14 December 2015 Available online 24 December 2015 We present a new method to determine the ground truth and accuracy of lightning location systems (LLS), using natural lightning strikes to trees. Observations of strikes to trees are being collected with a Web-based survey tool at the Finnish Meteorological Institute. Since the Finnish thunderstorms tend to have on average a low flash rate, it is often possible to identify from the LLS data unambiguously the stroke that caused damage to a given tree. The coordinates of the tree are then the ground truth for that stroke. The technique has clear advantages over other methods used to determine the ground truth. Instrumented towers and rocket launches measure upward- propagating lightning. Video and audio records, even with triangulation, are rarely capable of high accuracy. We present data for 36 quality-controlled tree strikes in the years 2007–2008. We show that the average inaccu- racy of the lightning location network for that period was 600 m. In addition, we show that the 50% confidence ellipse calculated by the lightning location network and used operationally for describing the location accuracy is physically meaningful: half of all the strikes were located within the uncertainty ellipse of the nearest recorded stroke. Using tree strike data thus allows not only the accuracy of the LLS to be estimated but also the reliability of the uncertainty ellipse. To our knowledge, this method has not been attempted before for natural lightning. © 2015 Elsevier B.V. All rights reserved. Keywords: Lightning location Thunderstorms Lightning Location accuracy Lightning location performance 1. Introduction Trees can be analyzed from several different perspectives. First, un- like a tall mast or a rocket-triggered lightning, a tree can be viewed as a natural object to study the lightning attachment process, e.g., how the soil and tree properties, such as height, affect the attachment (Mäkelä et al., 2009). Second, lightning strike to a tree can be viewed from the biological perspective, i.e., what kind of damages are typical to certain tree types, and are the damages more related to the properties of the tree and its surroundings, rather than to the stroke properties (e.g., Anderson and Anderson, 1968). Strikes to trees can also be consid- ered from the standpoint of the hazard they cause to humans and infra- structure (Das et al., 2009; McKechnie and Jandrell, 2008), or as sources of forest fires (Larjavaara et al., 2005). Taylor (1965) attempted to esti- mate the diameter of a lightning current channel by analyzing the dam- age occurring to the trees. Finally, as will be discussed in this paper, tree damages can be used to verify the lightning location data. The present lightning location system (LLS) of the Finnish Meteoro- logical Institute (FMI) has been in operation since 1998 (Tuomi and Mäkelä, 2008a; Mäkelä et al., 2010, 2014). Besides the temporal and spatial information of the located cloud-to-ground (CG) strokes, the system provides also information, e.g., about polarity, multiplicity, and peak current. Furthermore, the system provides estimate about the un- certainty related to the calculated stroke location, often termed as loca- tion accuracy. As an LLS is a remote sensing instrument, all reported values are estimates, and not directly measured values. This means, that if the performance of the LLS should be checked or verified, ground truth observations are needed. Two parameters are often reported to represent the performance of an LLS: detection efficiency (DE) and location accuracy (LA). Location accuracy studies can be arranged into three categories: (1) comparison between two or more different LLS's, (2) comparison of video and LLS observations, and (3) comparison of LLS locations to known strike points. Method 1 (e.g., Rodger et al., 2004; Lay et al., 2004; Pohjola and Mäkelä, 2013) can be done for large data sets but it gives only the information on the relative performance of the compared LLS's (i.e., not against any ground truth). Method 2 is more objective but the comparison can be usually done only for subsequent strokes occur- ring in the same lightning channel, i.e., strokes having the same ground strike point (Idone et al., 1998; Ballarotti et al., 2006; Biagi et al., 2007; Poelman et al., 2013). This makes possible the inspection of the random location bias of the LLS for subsequent strokes. Method 3 is the most ob- jective method, but it is hampered by the small number of known ground strike points, and usually needs cooperation with, for example, insurance companies. Atmospheric Research 172–173 (2016) 1–7 ⁎ Corresponding author at: Finnish Meteorological Institute, P.O. Box 503, FIN-00101, Helsinki, Finland. Tel.:+358 503011988. E-mail address: antti.makela@fmi.fi (A. Mäkelä). http://dx.doi.org/10.1016/j.atmosres.2015.12.009 0169-8095/© 2015 Elsevier B.V. All rights reserved. Contents lists available at ScienceDirect Atmospheric Research journal homepage: www.elsevier.com/locate/atmosres