**2. Data and Methodology**

For the present study, we have collected lightning data from Tai-Power Company of Taiwan for the years 1998–2012 to determine the urban effect on lightning over Taipei and its surrounding places. The Lightning Location System (LLS) was built in 1989 with one APA (Advanced Position Analyzer), and six Direction Finders (DFs) installed at sites covering the entire area of Taiwan. The LLS was upgraded

to a Total Lightning Detection System (TLDS) in 2002. The TLDS consist of seven lightning detection sensors (SAFIR 3000), which are located at the top of Ying-tsu-ling microwave tower, Wu-shih-pi microwave tower, Ji-shan microwave tower, Nan-Ke extra voltage substation, Feng-lin microwave tower, building roof of Ming-tan power plant, and building roof of Xiao-liou-chiou. The location of these seven sensors, distributed throughout Taiwan, is shown in Figure 1. The VHF interferometric technique is the main basis for the localization principles of SAFIR network [28,29]. The seven lightning detection sensors, formed a lightning detection network, could detect cloud-to-ground (CG) lightning discharges, intra cloud (IC) lightning discharges, and breakdown events. The lightning discharges detection is accomplished through the use of multiple, remote sensors that detect signals emitted by lightning discharges, and by filtering out the signals from non-lightning sources. The long rang localization of all lightning discharges (CG and CC lightning flashes) is governed by triangulation performed on GPS time synchronized direction of arrival provided by interferometric sensor of two different detection station in a SAFIR network. Each sensor detecting a lightning event sends data about the event to a central processor (SCM) that triangulates the results from each sensor creating an optimal estimate of location of the lightning event. The lightning detection network average efficiency is greater than 90%, and the lightning detection localization accuracy is less than 1 km. However, especially near the edges of the network the assumption of more than 90% uniform flash detection efficiency may not be realistic, but because of comparatively higher average detection efficiency and localization accuracy no attempt was taken to correct the detection efficiency because previous studies (e.g., [30] for a Lightning Position and Tracking System (LPATS); [31] for the National Lightning Detection Network in the United States; [32] for a LPATS in Germany, and [33] for a LPATS in Brazil) reported an overall detection efficiency of 90% for several lightning detection networks. For the present study we have only considered only CG lightning discharges and have ignored the IC discharges. CG lightning flash density was computed from the aggregated CG lightning flash data for the period 1998–2012.

**Figure 1.** Location of the sensors used in the Total Lightning Detection System (TLDS) of Tai-Power Company, Taiwan.

Surface temperature data over Taipei are collected from Central Weather Bureau (CWB) for the years 1965–2010. With more than 400 stations, consisting of conventional surface station and Automatic Rainfall and Meteorological Telemetry System (ARMTS), established by CWB, an extremely dense surface observation network was developed over Taiwan. Hourly observations at these stations were used to obtain the daily mean surface temperature. The census record and land use data are collected from various Internet sites of the Taiwan government agencies. Landsat satellite images are employed to demonstrate the urbanization of the city of Taipei and its surrounding areas and are downloaded from United State Geological Survey. Errors for these images refer especially to acquisition loss due to clouds covering the region of interest. Thermal band of the Land-Sat 7 satellite is used to generate apparent surface temperature of Taipei and New Taipei city. The sensors of Land-sat 7 acquire temperature data and store the information as a digital number (DN) with a range between 0 and 255. These DNs are first converted to radiance values using the bias and gain values specific to the individual scene. Thereafter an atmospheric correction using appropriate local values for several parameters are performed to generate more accurate surface temperature map. Air pollutants data were collected from a well-organized air quality-monitoring network operated by Taiwan Environmental Protection Administration (EPA) of Taiwan for the period 2003–2012. Taiwan's air quality monitoring network measures PM10 concentrations by the automatic Wedding β-gauge monitors, which is one of the US EPA-designated equivalent methods (no. EQPM-0391-081). The PM10 inlet is a cyclone operated at 18.9 min–1. Particles are detected once every hour from its continuous collection on the filter tape and the daily average is computed for at least 16 effective hours every day. A comparison experiment had been made between Wedding β-gauge monitors and the manual samplers because of the frequent abundance of high humidity in the ambient air of Taipei. The results obtained from automatic Wedding β-gauge monitor and the manual samplers were very close. For the present study PM10 and SO2 over Taipei City and New Taipei City are considered.
