Lightning Reporting at 45Th Weather Squadron: Recent Improvements

Lightning Reporting At 45th Weather Squadron: Recent Improvements

Frank C. Flinn William P. Roeder Michael D. Buchanan Todd M. McNamara Michael McAleenan Katherine A. Winters Michael E. Fitzpatrick Lisa L. Huddleston*

45th Weather Squadron *National Aeronautics and Space Administration Patrick AFB, FL Kennedy Space Center, FL

1. INTRODUCTION upgraded the Cloud to Ground Lightning The 45th Weather Squadron (45 WS) Surveillance System (CGLSS) (Boyd is the U.S. Air Force unit that provides et al., 2005) and integrated it into the weather support to America's space 4DLSS. A map of the cloud to ground program at Cape Canaveral Air Force lightning sensors of the 4DLSS is in Figure-2. The 45 WS also uses the Station (CCAFS), National Aeronautics Launch Pad Lightning Warning System and Space Administration (NASA) Kennedy Space Center (KSC), and (LPLWS) (Eastern Range Instrumentation Patrick AFB (PAFB). The weather Handbook, 2009), a network of 31 surface electric field mills that has a limited total requirements of the space program are very stringent (Harms et al., 1999). In lightning detection capability. A map of addition, the weather in east central the field mills in LPLWS is in Figure-3. Florida is very complex. This is especially The last lightning detection system used true of summer thunderstorms and by 45 WS is a direct connection to the associated hazards. Central Florida is National Lightning Detection Network `Lightning Alley', the area of highest (NLDN) (Orville et al., 2002). lightning activity in the United States (U.S.) (Huffines and Orville, 1999). The 45 WS uses a dense network of various weather sensors to meet the operational requirements in this environment (Roeder et al., 2003). The 45 WS is especially well instrumented with lightning detection sensors. The Four Dimensional Lightning Surveillance System (4DLSS) (Murphy et al., 2008) included a major upgrade to the total lightning detection Lightning Detection And Ranging (LDAR) system (Boccippio et al., 2001). The 4DLSS was implemented operationally in April 2008. Figure 1. Map of the nine total lightning A map of the total lightning sensors of the sensors in the Four Dimensional Lightning 4DLSS is in Figure-1. The 4DLSS also Surveillance System.

Corresponding Author: William P. Roeder, 1201 Edward H. White II St., MS 7302, Patrick AFB, FL 32925; (321) 853-8419; [email protected] lightning watches and warnings for the safety of over 25,000 personnel and protection of over $20 billion of facilities (Weems et al., 2001). Another application is daily lightning reports to customers. Other applications include incident investigations, climatological studies for mission planning, and development of new or improved forecast tools. The daily lightning reports include the distance and peak current of cloud-to- ground (CG) lightning strokes in the vicinity of key facilities. Nearby lightning strokes can induce potentially damaging electric currents in the electronics in Figure 2. Map of the six cloud to ground satellite payloads, space launch vehicles, lightning sensors in the Four Dimensional ground support equipment, or key Lightning Surveillance System. facilities. The daily lightning reports are used to decide if the various electronics should be inspected and, if so, what level of inspection is required. If damage occurred, it is essential to conduct those inspections to identify and implement required fixes to avoid potential post- launch problems or early failure of the electronics that could result in degradation or mission failure, or in extreme cases, even destruction of the space launch vehicle. However, it is also important to avoid unnecessary inspections due to their financial cost and delays to space launch schedule. The lightning reports assist mission planners in weighing the risks of conducting, or not conducting, Figure 3. Map of the 31 surface electric further inspections. This paper will field mills in the Launch Pad Lightning describe the significant improvements to Warning System. the 45 WS lightning reports since April 2008.

The 45 WS uses these lightning 2. Recent Improvements to the 45 WS sensors for several applications (Roeder Lightning Reports et al., 2005). One of these applications is the evaluation of lightning launch commit The 45 WS and their mission partners made five major improvements to their criteria, the weather rules to avoid natural and rocket triggered lightning strikes to in- lightning reports from April 2008 through flight space launch vehicles (Roeder and 2009: 1) reporting of all strokes, 2) inclusion of lightning location error ellipses McNamara, 2006). Another application is tailored to each stroke, 3) on-demand network and a 50th percentile confidence 24/7 availability of lightning reports, for the lightning location was provided. 4) correction of a truncation error of the The 45 WS discovered that this location peak current in the lightning database, accuracy assumed all six CG lightning and 5) KSC automated e-mail alerts and sensors were used in the solution. posting of lightning events at a website. However, 4DLSS often has fewer than all six sensors per solution, even if all six 2.1 Reporting All Lightning Strokes sensors are operating, resulting in larger The recent improvements to the 45 WS location errors and more eccentric error lightning reports began when the 4DLSS ellipses than previously believed. The became operational in April 2008 (Murphy median number of sensors per solution is et al., 2008). One of the many benefits of 4.80 for local lightning strokes detected by 4DLSS is that all return strokes per flash 4DLSS with a distribution shown in are reported. The previous CGLSS Figure-4. Reporting a single constant lightning system used by 45 WS only location accuracy implied circularity of the reported one stroke per flash. Detecting error, which was misleading. In addition, and reporting all return strokes is the customers had requested a 50th important since CG lightning has an percentile location accuracy, but more average of 3.5 strokes per flash and 50% recent discussion showed this to be or more of these strokes have multiple inadequate for space launch applications. ground strike locations (Cummins et al., Frequency of Number of Sensors in Lightning Solutions 1998). These multiple ground strike in CG4DLSS for Strokes within 5 nmi of Launch Pad40 (4770 Strokes. 1 Apr-25 Jul 2009, 6 Sensors in CG-4DLSS) locations have an average spacing of 50 3 km and can extend up to 12 km (Valine Median = 4.80 sensors 40 and Krider, 2002). Reporting only one X Mean 4.99 sensorsJ stroke per flash, as done by the former °. 30 CGLSS, meant these other return strokes A would not be reported and potentially necessary inspections would be missed. 10

n n

2.2 Lightning Location Error Ellipses 2 3 4 5 6 (poor, (good) irr.c 11W, (oufstandngl

Tailored To Each Stroke Number of Sensors (1-sensor solutio —pos bb) Another significant improvement in the Figure 4. Distribution of number of 45 WS lightning reporting procedures was sensors used in location solutions by the inclusion of stroke location error 4DLSS for nearby strokes. ellipses tailored to each individual stroke. This provided finer details for the location reporting uncertainty and overcame 2.2.2 Location Error Ellipses Tailored To several previous shortfalls. Each Stroke With 99% and 95% 2.2.1 Previous Shortfalls: single best- Percentiles case 50% confidence location accuracy To address this shortfall, the location The previous 45 WS lightning location accuracy and detection efficiency across accuracy had several shortfalls. A single the local area was obtained for all location accuracy was used for all possible combinations of sensors in the lightning strokes in the center of the solutions. As expected from a similar study on the older 5-sensor configuration stroke based on the number of sensors of CGLSS, the performance of the new used in the solution, the distance from the 4DLSS is still sensitive to the central sensors to the stroke, and the geometry of sensor on CCAFS being used in the the sensors relative to the stroke. solution (Table-1). The new CGLSS is Kennedy Space Center requires 99th relatively insensitive to one, or to a lesser percentile error ellipses while the rest of degree even two sensors, not being used the space launch customers use 95th in the solution, unless one of those percentile error ellipses. In 99%/95% of sensors is the central sensor on CCAFS. the events, the best location of the stroke However, even with these performance will be inside the error ellipse and in plots, the error characteristics for each 1%/5% of the events the best location of individual return 'stroke were still needed the stroke will be outside the ellipse, since these error characteristics varied respectively. The product includes the based upon the geometry of the stroke distance from each key facility to the best location relative to the sensors used in the location of nearby lightning strokes, i.e. solution of that stroke. the center of the error ellipse, the distance to the closest edge of the ellipse, and the Table 1. peak current of the stroke, etc. A 4DLSS performance near the launch pads schematic diagram of these distances is for all combinations of one sensor in Figure-5 and a copy of the product is in excluded from the liahtnina solution. Figure-6. For each key facility, the Sensor Detection Location Accuracy customer specifies a critical distance Missing Rate (%) (median) (m) within which lightning strikes merit None 94% 300 m additional investigation. A Google-Earth Melbourne 93% 350 m visualization is also available on request Deseret 93% 350 m (Figure 7). Customer points of interest include Tosohatchee 92% 350 m launch pads, payload/space launch Seminole 91% 400 m vehicle processing facilities, and other Shilo 91% 450 m facilities. The 29 key facilities supported Cape 89% 500 m by the 45 WS daily lightning reports are Performance is most degraded by the listed in Table-2. If the distance to the absence of the central Cape sensor in the most likely stroke location is larger than lightning solution. Similar sensitivity to the the inspection threshold for a stroke of Cape sensor absence is seen when that intensity, then the customer can be higher number of sensors are excluded. confident that inspection of the electronics See Figure-2 for the CG-lightning sensor is not needed. The inspection thresholds sites. Performance estimated from allow for the uncertainty in the peak isopleths of modeled performance current estimate. If the distance to the provided by Vaisala, Inc. edge of the closest point of error ellipse is also larger than the inspection threshold for a stroke of that peak current, then the The 45 WS and their mission partners customer can be very confident that then developed an interim procedure inspection of the electronics is not where the raw data from 41DLSS were needed. used to estimate the lightning error ellipses for nearby strokes tailored to each TABLE-2. Lightning Location Error Ellipse Key Facilities for which 45 WS issues daily liahtnina reDorts. 99th Percentile Error Ellipse • 99% probability that actual stroke Key Facility Primary Customer was anywhere inside the ellipse • Most likely location is ellipse center Atlas Space Atlas Operations Center Area 59 Delta-IV Astrotech Commercial Satellite Processing Facility Launch Complex-17A Delta-II Facility Lightning Stroke Launch Complex-17B Delta-II Launch Complex-36 Space Florida Launch Complex-3713 Delta-IV Launch Complex-39A Kennedy Space Center Figure 5. Schematic diagram of a Launch Complex-39B Kennedy Space Center lightning location error ellipse. Range is Launch Complex-40 Falcon-9 the distance from the facility of interest to Launch Complex-41 Atlas-V the best location of the lightning stroke, Delta-IV Ops Center Delta-IV i.e. the center of the error ellipse. The Falcon Launch Falcon range to the closest point of the error Control Center ellipse is also provided. KSC uses 99th Horizontal Integration Delta-IV percentile error ellipses, while the rest of Facility the space launch customers use 95th Joint Surveillance U.S. Air Force percentile. and Target Attack Radar System Ops & Checkout Kennedy Space Center Patrick AFB 45th Space Wing Payload Hazardous Kennedy Space Center Servicing Facility Port Navy Range Control Cntr 45th Space Wing Skid Strip 45th Space Wing Shuttle Landing Kennedy Space Center Facility-Runway N. Shuttle Landing Kennedy Space Center Facility-Runway S. Shuttle Landing Kennedy Space Center Facility-Mate/Demate Facility Solid Motor Assembly Kennedy Space Center Building Shuttle Payload Kennedy Space Center Integration Facility Figure 7. Example of the Google Maps Space Station Kennedy Space Center Processing Facility visualization provided when the lightning Vehicle Assembly Bid Kennedy Space Center strike is within the critical distance. The Vertical Integration Atlas-V 13 points used to approximate the Facility location error ellipse are shown. Confidence Ellipse Data

Number of strokes in which the center of the lnmich complex is inside the 99% confidence Ellipse: 3

Shuttle Complex 39A, 3 August 2009 Data from the 46th Weather Squadron Cloud-to-Ground Lightning Surveillance System II (CGLSS 11) 99% Confidence Ellipse Data

Azimuth Range Target Sensors Time Magnitude Range Magnitude Date Azimuth (N M) (to (to Inside in (UTC) (kA) (kAmps) Ellipse *) Ellipse*) Ellipse*? Solution

3-Aug-09 21 10 40 527 -398 2884 046 263 0 11 -398 Yes 3 3-Au -09 21.10:40.601 -322 296 4 533 0.42 32.2 Yes 2 3-Au -09 21 10:41.240 1 -49 1 2929 1 061 1 101.9 0.09 1 -49.1 Yes 4

Figure 6. Example of the daily lightning report provided by 45 WS. For KSC, lightning strokes that are inside the operationally critical radius of 0.45 nmi are highlighted in green. Points of interest that lie inside the error ellipse are highlighted in magenta to aid interpretation of the range to ellipse.

2.3 On-demand 2417 Lightning Reports workstation was installed, the situation it The use of error ellipses was a was designed to mitigate occurred. significant improvement over the previous During the nighttime countdown for the method. However, due to computer Space Shuttle STS-128 mission, lightning security requirements, the initial process struck near the launch pad on 27 Aug 09. required a system administrator to copy The on-demand lightning report showed the data from the 4DLSS workstation and inspection of the electronics was required. hand-carry it to 45 WS. Due to the By providing the report immediately, system administrator's work schedule, this rather than waiting for the next morning as meant lightning reports could not be done previously, only a 24-hour slip of the generated nights, weekends, or holidays if launch occurred, rather than a 48-hour a nearby lightning strike occurred during a slip, providing a cost avoidance of over $1 major non-launch operation and million. sometimes early during a launch countdown. A workstation was installed in 2.4 Fixed Truncation Error Of Peak the 45 WS operations area with a Current communications link to the 4DLSS Another improvement to the 45 WS workstation on 17 Aug 09. This allowed lightning reports was the discovery and on-demand lightning reports in near real- correction of a truncation of the lightning time without system administrator support. stroke peak currents in the computer Within minutes, the data are now database used to generate the reports. available to 45 WS for calculation of the The peak current was truncated, rather error ellipses. Just 10 days after the than rounded to the nearest integer kilo Amp (kA). However, with an average technique will be much more useful to the peak current of 20 kA, this truncation was space launch customers and may causing up to a 4% underestimate of the supersede the lightning error ellipse peak current of each stroke for average approach discussed above. The technical lightning. details of this new technique are available in Huddleston (2010) and will be 2.5 KSC Automatic E-mail Alerts And presented to the 21 st International Website Lightning Detection Conference, On their own initiative, KSC provided 21-22 Apr 10 (Huddleston et al., 2010). 24/7 automatic e-mail notification in near The KSC was considering adding error real-time to customers whenever a ellipses to their website displaying nearby lightning strokes in near real-time and to lightning stroke exceeded that customer's distance and/or intensity threshold. their automatic e-mail notifications. That These e-mail use a 30 minute cycle time, effort may be superseded by this new so customers are notified of important technique. lightning with an average lag time of 15 min. KSC also displays stroke locations and distance/intensity data in near real-time at a customer publically 99°0 accessible website.

3. On-going Improvements to the 45 WS Lightning Reports Further improvements to the 45 WS 1)7 ObUbd irl . cif lightning reporting process are being pursued or considered as future projects. this .ti't1 'C)ke, 3.1 Probability That Any Nearby Lightning 117 this Circle Stroke Is Within Any Radius Of Any Point Of Interest Figure 8. Schematic of the new facility- centric process of calculating the A technique has been developed to probability of any stroke within any radius calculate the probability that any nearby of any point. lightning stroke is within any radius of any point of interest (Figure-8). In practice, this provides the probability that a nearby 3.2 Distance To The Closest Point Of lightning stroke was within a key distance Lightning Location Error Ellipse of a facility, rather than the error ellipses centered on the stroke. This process The closest point on an ellipse to any takes the current bivariate Gaussian arbitrary point cannot be found distribution of probability density provided analytically. This complicates calculating by the current lightning location error the distance from a key facility to the ellipse for the most likely location of a closest point of lightning location error lightning stroke and integrates it to get the ellipse. This distance was initially probability that the stroke is inside any approximated by the distance to the specified circle. This new facility-centric closest of 13 evenly spaced points on the error ellipse. Only 13 points were used approximating the error ellipse with due to limitations of the Microsoft° EXCEL 65,000 points and choosing the closest of spreadsheet software. Unfortunately, this the points in just 50 iterations, which was method can grossly overestimate the a vast improvement over the previous distance when the point of interest is near method of choosing the closest of 13 the ellipse and about equidistant between evenly spaced points approximating the the 13 points (Figure-9). Under worst error ellipse. Even in a very unfavorable case conditions, the error can be up to scenario, this method provides a location 1.5 nmi, which is very significant when the error no larger than 2.5 m, two orders of radius for an area of interest may be only magnitude less than the best location a few tenths of a nmi. error possible from the 4DLSS. A visual basic program was developed that iterates to a much more accurate solution for the closest point on the error ellipse to the point of interest. A schematic of this iteration process is shown in Figure-10. The process begins by calculating the distance from the point of interest to eight points on the ellipse, spaced every Tu/4 radians (45°) around the center of the error ellipse starting due west of the center of the ellipse. The closest of these eight points is chosen as the starting point. On the first iteration, three subsequent candidate closest points are selected on the error ellipse, the current chosen point and points spaced 7r/8 radians (22.5°) (half the previous Figure 9. The previous method of angle) to either side of that point relative estimating the distance from the facility to to the center of the ellipse. The closest of the closest point on the error ellipse was these three points is chosen. On the next done using the distance to the closest of iteration three subsequent candidate the 13 evenly spaced points used to closest points are selected on the error approximate the ellipse. This method ellipse, the point chosen in the precious would grossly over-estimate the distance iteration and points spaced half the angle when the facility was close to the ellipse in the previous iteration angle to either and about equidistant from the two closest side. The iteration is continued until the 13 points. This method was replaced by a iteration angle is n/216 radians far superior method as discussed in the (0.002727°). The closest of the three text and shown in Figure-10. candidate points in that last iteration is selected as the final closest point and the iteration is ended. This is equivalent to Distance From Point Of Interest To Closest Point On Uncertainty Ellipse

1) Pick Starting Point 2) Iterate To Pick Next Closest Point. Using'/ of Previous Angle a) Pick 8 points every a 4 radians Uncertainty on ellipse Ellipse b) Calculate distanceto 3 4 each of8 points d= cl Point with shortest distance Point of New Pointyn^' Is starting point Z Interest d i niH dy < di . d2 ..... d, nf4 r closestpoint Therefore, de is starting point 1 X74 5` d - New Point-1 W4 ' (formerly Point-81 6 f

Point of d , closest point < ' a New Point-3 Interest 7, Uncertainty Therefore. d. Is next closest point l Ellipse

3) Continue Iterating Using Y-- of Previous 4) Continue Iterating Until Angle = 71/216 Angle Each Time <,-New Point-2 n/2,e d, n116

d' — - __ ------Point of `New Point-1 d Interest (formerly Now P o_ipt-2) Point of closest point Interest a i?sest point d3 d, < d,. d3 d3 < d, , d2 and iteration angle = a^216, Therefore, d, is Therefore d3 is the approximated dist nce next closest point -New Point-3 from point of interest to closest point n ellipse

Figure 10. The new method estimates the distance from a facility to the closest point of the error ellipse. The new method is far superior to the previous method since it overcomes the problem of grossly overestimating the distance under some scenarios as discussed in the text and shown in Figure-8.

3.3 Strong Local Lightning Strokes sensors into 4DLSS in real-time to Sometimes Not Detected improve the detection of strong local Recent research has shown that strokes. The nine NLDN sensors being 4DLSS can be saturated by strong local considered are based on those closest to strokes and fail to detect them, especially CCAFS/KSC and those with the best those with peak currents of 50 KA or complementary geometry relative to greater (Ward et al., 2008). However, CCAFS/KSC. The nine NLDN sensors 4DLSS excels at detecting weak local being considered for incorporation into strokes. On the other hand, the wider 4DLSS are all seven of the sensors in FL, spacing of the NLDN sensors excels at one just across the state line in GA, and detecting those strong strokes, but loses one in the Bahamas Islands. detection efficiency for weaker strokes Incorporation of the sensor data from near CCAFS/KSC, especially those with those NLDN sensors into 4DLSS in real- peak currents of 7 KA or less. This time will also improve the location suggests that combining the sensor data accuracy, detection efficiency, and from both systems in real-time could lead provide smaller and less eccentric error to improved performance. The 45 WS is ellipses when only a few of the 4DLSS pursuing integrating data from nine NLDN sensors are used in the solution. The performance of 4DLSS will not be compromised when most of the 4DLSS 3.5 KSC Automatic E-mail And Website sensors are used in the lightning solution. KSC is considering adding the location As an interim measure, KSC is error ellipses to their automatic 24/7 e- purchasing StrikeNet reports from mail alerts and display error ellipses at Vaisala, Inc. when lightning strokes are their website so the customers can see detected or suspected near KSC points of this important data in near real time. This interest. The StrikeNet reports include all effort may be superseded by the new the strokes detected by NLDN, as probability of any lightning stroke being opposed to the more routinely available inside any radius of any location, as flash-only data, and so should include the discussed in section-3.1. strong local strokes missed by 4DLSS. The StrikeNet reports also allow cross- 4. Possible Future Improvements to comparison with the 4DLSS lightning the 45 WS Lightning Reports reports to identify strokes 4DLSS may have missed and to check for consistency There are two main avenues to in lightning locations and peak current. A improving the 45 WS lightning reports sample StrikeNet report is in Figure-11. even further in the future: 1) improved The StrikeNet solution is not as good as peak current estimates and improved integrating the nearby NLDN sensors into error estimates of the peak current 4DLSS since the stroke location, error accuracy, and 2) 4DLSS upgrades. ellipses, and peak current solutions are not optimized with all the sensor data from 4.1 Improved Peak Current Estimates both systems. In addition, inconsistencies And Peak Current Errors between the two reports may occur, There are five main factors in requiring manual analysis to reconcile. assessing the induced current hazard However, the StrikeNet reports are presented by nearby lightning: 1) the available now, while the integration of the detection rate of the lightning detection NLDN sensors into 4DLSS is still being system being used, 2) the distance to the developed. The 45 WS may acquire stroke, 3) the error in the location, 4) the StrikeNet reports to support their DoD, peak current of the stroke, and 5) the NASA unmanned, and commercial launch error in the peak current. Considerable customers. work has been done over the years in improving and understanding the 3.4 Fault Analysis Lightning Location detection rate and location accuracy of System lightning detection systems, including The 45 WS considered acquiring the 4DLSS. However, more work is, needed Fault Analysis Lightning Location System to improve the estimate of peak current as (FALLS) (Vaisala, 2009). The FALLS well as the error in the estimate of the would have provided advanced analysis peak current. and display capabilities of error ellipses. Anyone interested in helping conduct However, given the in-house lightning these peak current studies is encouraged reporting improvements discussed above, to contact the corresponding author. the 45 WS decided FALLS was not cost- effective for their mission. STRIKEnot Report 23&<404 Re Tae S—t"LCTM6309 T— L.V-" Sla/.e{ rMrebl 439 Ly]]vy 59a1e{ Oee![e.l r!].• S 2il Lbb].q .esyw !]ee![l.a Mven0 5 mJt ).,, ^1e1! eoe,Ow,[e !br oNVrp{ M rdu. 1M T.we S[Nr. Au0 ].:009 06 W W OYT b lul {. b09 C5 S3 W GY! L0 SW- PON"& Fa LOtn+]rq Sbd.es

f+rra.LYMfwa] <.!I / S^.sro•JO .N.N ew i

Ke^w.Ni ^.el{aJAnCIC OceOn .••. •• ^i

oza^ra.low.v llc. ^^ aoo»rrllviF+O ._-. w.n.w ^. err.r]e M.l,ewe^f y V r.r. Y [1+A w.v [ r..^ ..« c v.. ^. RCM

;;]ro 6.4 VAISALA ..... u e]nl inl

e..s..ea.r+ryw.a. a.l. on:mru oy.. ye =a W. fw+e .1m u .W. wN.ra '^N wra s wrra sw Nr r r w. awu.r. â^s^û^,.^ y.r vâM..r.^.•^â^^.^ wa^+êT V..îM wM1.

STRIKF_ne7 Re pon 238EKA rell„1 re! srRacv.^e[ Le 31w e ^ oe tore LQmp style. aelectaa {]a LyNel 5lalw [e)e[eea w^ s....,1 ..^ .loan na L clw^.v s.alw o.r.ra ee,Arw s r+re LT., ...>.. cam.... nln.e u..r.w e^ ,w... • r:. T—e S pRlp.]. 5 r ire T.ne Sin A.q 3,11109 Ofi W W fJR b Aoq 1, X09 Cb S? W frYr t ghbwg Sboll Table (Note Ea4Ne4 50 —1s cMa r. f .on]s pd—d try lenl 1

y] •ytu Barr r 0 xe{ ae.q ae edr y]:ac gNae .ri] xeao] {O GV: yl'x Been 1}i •^♦ lG^ aC C.^e y>.^ geJm .]Or ylxo. NeS ee .NS )}e0 X{«• Xe..: y],ryl OeIN .rIJ »4A ar 6Yl y]xw g9Ti IM,»G. aam ie]a0 y].•]W gelH .aeZ )Me XeN] JC ear y] .l]OG qY» Je] {i.'ie at]"] of eTq yl xcc NM» li) {i_•re xM]1 {e eT]l .V]R N'J.V -bl {eri{ Xs.w aclrei •.v 1:CW geex .]ei eslJ »el•e xe{x wyl ^ gelX .te• ILrJ xMH . «l arr ae.« .yl y].m. xa .s9lxZ -A .q{ .ilte »Ga• ie elrir y1:x4 q)e aJ -qe M2 ae.L' a]NOD yl ]mD NNi: )JlJ xM]I kNM ylxm geeR ^I9 *-I No— yl:p(,y qNS •• i-1 Mr -.1.r ylas qee+: rie IfI1 {.[p y].^lw q>1C -a) )Yea ae]:• ZOAS xm +s ]M JNe A— JD6 y y xOD xmD .]e• )J.1J xW) ai lira y]]m0 xmD q! if0{ Xele{ be01e y]xx xmD -1JJ )Ni OeO ae ll•6 y] ^D xsN .JJe {Iw} XeeD. ae N+! yi ^Yl .12 JMl aMq y]:AU xvn: nl .n9 sees +o en. y]xoc xWN .u. sill x{i» to eeo{ y) W xlne -aT 3Ser aa]D a]oOn 000G yl i® M N .a. )MJ aC o4N y) ]C[% x0.1. ar ) >ll . x eMr K ai» y]. a xn[D ]9e i.ee aera a007 grad xo.D .ni lVr. ser]J a:aro —`—o O+v VAISALA

Figure 11. Sample output from a StrikeNet report, which provides stroke data from NLDN. 4.1.1 Improved Peak Current Estimates For example, Tower-313 is the tallest The 45 WS is interested in improving tower in the network (500 ft) but is located the estimates of the peak currents from near the coast. Shorter towers farther 4DLSS. At present, the peak current inland might be more likely to be struck by estimate is calculated from the peak lightning since the climatological lightning magnetic field at each sensor. The peak flash density increases in-land. Funding magnetic field is normalized to a range of for this project was not available at the 100 km and corrected for attenuation from time this paper was written (Jan 2010). ground propagation effects. The mean of There may be ways to improve the the attenuation-corrected range- range-normalized attenuation-corrected normalized peak magnetic fields is regression equation approach used at converted to peak current via a regression present. For example, using an average equation (Cummins et al., 1998). That peak magnetic field weighted by distance regression equation was based primarily to the stroke for each sensor, rather than on data from rocket-triggered lightning. a simple mean, may yield some As a result, it is less representative for performance improvement. Sensors first strokes from natural lightning. This is farther from the stroke would receive less important to operations since the first weight in the distance weighted average. stroke in a flash tends to have the highest Another possible improvement could peak current. Thus, the first stroke can be separate regression equations based generally cause more induced current on stroke polarity. Likewise, different damage at the same distance or the same regression equations for varying peak induced current damage at farther current should also be considered, e.g. distances than subsequent return strokes. perhaps an iterative process where the Perhaps the best way to improve peak regression coefficients are modified based current estimates is to create a new on the peak current from the previous regression equation based on iteration, or a simpler approach of observations of natural lightning. stratified regression equations for weak, Unfortunately, there have been few direct moderate, and strong peak current. peak current measurements of natural Finally, entirely new approaches should lightning. An appropriately instrumented be explored to avoid the additional tall tower in a wide open flat area with uncertainties introduced by the range- frequent lightning and subsequent normalization and the regression analysis of that data should allow equation. significantly improved peak current 4.1.2 Improved Peak Current Accuracy estimates, especially for the operationally The estimated error associated with the, more important first strokes. The peak current estimates for cloud to ground CCAFS/KSC has a network of weather lightning strokes from 4DLSS has not towers that would be a natural candidate been as well studied as location accuracy for such an instrumented tower given the and detection rate, especially for various lightning frequency and terrain in that combinations of sensors used in the area. An analysis of tower height versus solution for each stroke. At present, a climatological flash density, along with single error estimate of ±20% is used for surrounding terrain and logistical all strokes, regardless of number of accessibility, should be conducted to sensors used in the solution and distance identify the best tower to be instrumented. of those sensors to the lightning stroke. This is the vendor's recommendation and them periodically. Second, replace the is based on the performance of the NLDN, 4DLSS sensors with the new model for which itself appears to be based on some long-term maintenance sustainability. old studies of relatively small sample size. Third, integrate any new nearby NLDN It appears that most customers are more sensors into 4DLSS. Fourth, add a new interested in detection rate and location seventh sensor to 4DLSS. Unfortunately, accuracy than in peak current accuracy. none of these activities is currently As a result, more effort has been invested funded. to quantify and improve the performance 4.2.1 New And Periodic Network of the former, rather than the latter. Some Performance Evaluation Program lightning detection experts have suggested that the actual errors in peak A Network Performance Evaluation current are larger than ±20% (Mata, Program (NPEP) was last accomplished 2009). for 4DLSS in summer of 2008, shortly The 45 WS is interested in improved after the system was installed. No major error estimates for peak current provided problems were found, but a minor radio by 4DLSS. One possible approach might noise problem was detected at one of the be using the variability of the peak current sites. A new NPEP should be conducted, estimated from each sensor for a better since one is recommended every measure of the peak current error. This 1.5 years. If the previous minor radio could also allow a statistical estimate of noise problem still exists, a remediation the confidence intervals and/or a high may be worthwhile. Also the NPEP would percentile, e.g. inter-quartile range, 95th check for any new problems. The NPEP or 99th percentile. A best-fit Gaussian should be repeated every 1.5 years for distribution might also be applicable. The stable lightning detection systems that are standard deviation of that best-fit performing well, as recommended by the Gaussian distribution could be used to vendor (Vaisala, 2008). generate probabilistic confidence 4.2.2 Replace 4DLSS Sensors With New intervals. The space launch customers Model could then factor the uncertainty of peak current more effectively into their decision The current CG-lightning IMPACT models for inspecting mission essential Model 141-T sensors are no longer electronics, just as they do now with supported by Vaisala, Inc. This is already location accuracy, i.e. determine the causing maintenance problems. For probability of exceeding their combined example, the Melbourne sensor was threshold of distance from their facility and damaged by a lightning strike on 26 Jul peak current. 2009 and a replacement sensor was not available, so 4DLSS is in a temporary 5- 4.2 4DLSS Upgrades sensor configuration at the time this paper is being written (Dec 09), rather than the The 45 WS is interested in upgrading nominal 6-sensor configuration. . The 4DLSS since improved lightning detection Tosohatchee sensor was moved to the will provide improved lightning reports. Melbourne location to replace the sole line Four main possible approaches to of sight to CCAFS/KSC from the south upgrade CGLSS are available. First, (see Figure-2 for sensor locations). The conduct a new Network Performance line of sight from the west provided by Evaluation Program (NPEP) and schedule Tosohatchee is duplicated in part by the Seminole sensor. Fortunately, Vaisala is in performance of the temporary 5-sensor currently manufacturing the LS7001 configuration of the cloud-to-ground sensor, which they plan to support for 4DLSS is being investigated. Once many years. This new sensor should be a diagnosed, a corrective fix will be simple plug-in replacement of the current implemented. sensors with no loss of performance and requiring no modification to the rest of ¢SeMrCG 4Dt55 Frequency of Number of Sensors in Lightning Solutions for Strokes within Snmi of Space launch ComplexAC 4DLSS. While a test of the new sensor (4ROStrokes, 1Apr-25,W 2OD9) model in 4DLSS is funded and being 60 Meoian = 4.80 sensors scheduled, the follow-on replacement of x 50 Mean = 4.99 sensors all the current sensors is not yet funded, 40 pending results of that test. However, 30 — replacing the sensors is mission-essential 20 to ensure sustainability of 4DLSS. The testing of the new LS7001 sensor, 0 2 3 4 5 6 ro^on (goo f) ,^KC and subsequent replacement of the Number of Sensors current 4DLSS sensors, may be taking on )1-semor solutions impossible) heightened urgency. Preliminary analysis indicates that the performance loss to rSe^or CG^LSS Frequency of Number of Sensors in lightning Solutions for Strokes within 5 nmi of Launch Pail-40 cloud-to-ground lightning detection was (2003 Strokes, 12 Aug-3 Dec 20091 larger than expected after the loss of the 60 x Median -2.48 sensors Melbourne sensor on 26 Jul 09 and so Mean = 3.17 sensors relocation of the Tosohatchee sensor to 40 30 - the Melbourne site (see Figure-2 for site 20 - locations). Before the reconfiguration, the median number of sensors per lightning 0 solution under the nominal 6-sensor 2 3 4 5 (poor) .m Igo o (^vcsAenn configuration was 4.80 sensors for Number of Sensors lightning near the launch pads, close to an )1-sersorsolub m impossible) Figure 12. Frequency of number of overall excellent performance for space sensors per lightning solution for strokes launch applications. Only 14.1 % of near the CCAFS/KSC launch pads under lightning solutions used only 3 or 2 the nominal 6-sensor configuration (upper sensors, corresponding to marginal or figure), and the current temporary poor performance, respectively. 5-sensor configuration (lower figure). The However, under the current temporary Melbourne sensor was damaged on 26 5-sensor configuration, the median Jul 09 and the Tosohatchee sensor was number of sensors per solution for the moved to the Melbourne site and same area is 2.58 sensors, for an overall activated on 11 Aug 09. performance just below marginal, although the mean is 3.29 sensors. A large 69.9% of lightning solutions used 4.2.3 Integrate Any New Nearby NLDN only 3 or 2 sensors for marginal to poor Sensors performance, respectively. The distribution of number of sensors per In section-3, the on-going effort to lightning solution is shown in Figure-12. inject data from nine surrounding NLDN The cause of this unexpectedly large drop sensor data to 4DLSS was discussed. When that effort began in early 2009, Vaisala, Inc. was considering adding sensor has not yet been formally another NLDN sensor in central Florida, recommended by 45 WS yet so funding perhaps near Daytona Beach. If that has not been considered. sensor is added to NLDN, then it should also be incorporated into 4DLSS. 5. Summary Likewise, any other new NLDN sensors The 45 WS provides daily lightning added in Florida, southern Georgia, or the reports to space launch customers at Bahamas Islands should be considered CCAFS/KSC. These reports are provided for integration into 4DLSS. to assess the need to inspect the If the new NLDN sensor is not added, electronics of satellite payloads, space the 45 WS may consider adding a new launch vehicles, and ground support eighth 4DLSS sensor to 4DLSS at a equipment for induced current damage distance of about 60 nmi from from nearby lightning strokes. KSC/CCAFS. This new eighth sensor The 45 WS has made several would be in addition to the new seventh improvements to the lightning reports sensor discussed in section 4.2.4. Or it during 2008-2009. The 4DLSS, may be more cost-effective to fund implemented in April 2008, provides all Vaisala to install and maintain such a lightning strokes as opposed to just one sensor for 45 WS to ingest its data into stroke per flash as done by the previous 4DLSS. Either approach should help system. reduce the problem of strong local strokes The 45 WS discovered that the peak sometimes not being detected by 4DLSS. current was being truncated to the nearest 4.2.4 Add A New Seventh Sensor To kilo amp in the database used to generate 4DLSS the daily lightning reports, which led to an The performance of 4DLSS could be up to 4% underestimate in the peak current for average lightning. This error made more robust if a new seventh sensor was added. If this new seventh was corrected and led to elimination of sensor is sited at a near center location, it this underestimate. should , reduce the sensitivity to The 45 WS and their mission partners developed lightning location error ellipses performance if the Cape sensor is not used in the solution (Table-1). In addition, for 99% and 95% location accuracies the preliminary analysis of performance tailored to each individual stroke and loss under the current temporary 5-sensor began providing them in the spring of configuration suggests the gain in 2009. The new procedure provides the robustness with a new seventh sensor distance from the point of interest to the may be worthwhile. If this new seventh best location of the stroke (the center of sensor is added, a location at the the error ellipse) and the distance to the closest edge of the ellipse. This southwest edge of KSC should provide increased performance for lightning near information is now included in the lightning reports, along with the peak the launch pads if the Cape sensor is used in the solution. If this new seventh current of the stroke. The initial method of sensor is added, moving the Seminole calculating the error ellipses could only be sensor a few miles to the northeast would used during normal duty hours, i.e. not optimize the performance of 4DLSS during nights, weekends, or holidays. slightly, but this change may not be cost- This method was improved later to effective. The addition of a new seventh provide lightning reports in near real-time, 24/7. The calculation of the distance to operations, developed the software to the closest point on the ellipse was also calculate the distance from the point of significantly improved later. Other interest to the closest point on the error improvements were also implemented. ellipse. They also developed the software A new method to calculate the that accelerated the search routine to filter probability of any nearby lightning stroke. out lightning strokes too far to be an being within any radius of any point of induced current threat, which saved time interest was developed and is being processing the remaining strokes for implemented. This may supersede the threat assessment. use of location error ellipses. A previous version of this paper was The 45 WS is pursuing adding data reviewed by Mr. John Madura, Chief of from nine NLDN sensors into 4DLSS in the KSC Weather Office. real-time. This will overcome the problem of 4DLSS missing some of the strong 6. References local strokes. This will also improve the Boccippio, D. J., S. J. Heckman, and S. J. location accuracy, reduce the size and Goodman, (2001), A diagnostic eccentricity of the location error ellipses, analysis of the Kennedy Space Center and reduce the probability of nearby LDAR network. 1. Data characteristics. strokes being inside the areas of interest Journal of Geophysical Research, 106, when few of the 4DLSS sensors are used 4769-4786. in the stroke solution. This will not reduce 4DLSS performance when most of the Boyd, B. F., W. P. Roeder, D. Hajek, and 4DLSS sensors are used in the stroke M. B. Wilson, 2005: Installation, solution. upgrade, and evaluation a short Finally, several possible future baseline cloud-to-ground lightning improvements were discussed, especially surveillance system in support of space for improving the peak current estimate launch operations, 1st Conference on and the error estimate for peak current, Meteorological Applications of and upgrading the 4DLSS. Some Lightning Data, 9-13 Jan 05, 4 pp. possible approaches for both of these Cummins, K. L., M. J. Murphy, E. A. goals were discussed. Bardo, W. L. Hiscox, R. B. Pyle, and A. E. Pifer (1998), A combined TOA/MDF 6. Acknowledgements technology upgrade of the U.S. National Lightning Detection Network, The work by Dr. Huddleston of the KSC Journal of Geophysical Research, 103, KSC Orbiter Mechanical Systems office 9035-9044. was done under the KSC Employee Development Program. This included Eastern Range Instrumentation developing the program that calculates Handbook, 2009: LPLWS, Eastern the probability of any nearby lightning Range Instrumentation Handbook stroke being within a radius of any point of (CDRL B312), Systems Engineering interest and improving the calculation of and Analysis, Computer Sciences the distance to the closest point of the Raytheon, Patrick AFB, FL 32925, error ellipse. Contract FA2521-07-C-0011, 15 Oct09, Mr. Jeremy Hinkley and Mr. Pete 17 pp. Hopman of United Space Alliance, the main contractor for Space Shuttle Harms, D. E., A. A. Guiffrida, B. F. Boyd, Orville, R. E., G. R. Huffines, W. R. L. H. Gross, G. D. Strohm, R. M. Lucci, Burrows, R. L. Holle, and K. L. J. W. Weems, E. D. Priselac, K. Cummins (2002), The North American Lammers, H. C. Herring and F. J. Lightning Detection Network Merceret, 1999: The many lives of a (NALDN)—first results: 1998-2002, meteorologist in support of space Monthly Weather Review, 130, launch, 8th Conference On Aviation, 2098-2109. Range, and Aerospace Meteorology, Roeder, W. P., and T. M. McNamara, 10-15 Jan 99, 5-9 2006: A Survey Of The Lightning Huffines, G. R., and R. E. Orville, 1999: Launch Commit Criteria, 2nd Lightning ground flash density and Conference on Meteorological thunderstorm duration in the Applications of Lightning Data, continental United States: 1989-96. 29 Jan-2 Feb 06, 18 pp. Journal of Applied Meteorology, 38, Roeder, W. P., J. W. Weems, and P. B. 1013-1019 Wahner, 2005: Applications Of The Huddleston, L. L., W. P. Roeder, and F. J. Cloud-To-Ground-Lightning- Merceret, 2010: A method to estimate Surveillance-System Database, 1St the probability that any individual Conference on Meteorological lightning stroke contacted the surface Applications of Lightning Data, 9-13 within any radius of any point, 21st Jan 05, 5 pp. International Lightning Detection Roeder, W. P., D. L. Hajek, F. C. Flinn, Conference, 21-22 Apr 10, 14 pp. G. A. Maul, and M. E. Fitzpatrick, 2003: ,Huddleston, L. L., 2010: Probability Meteorological And Oceanic. computations and enhancements for Instrumentation At Spaceport Florida – the 45th Weather Squadron lightning Opportunities For Coastal Research, spreadsheet, draft NASA report, 5th Conference on Coastal available from corresponding author Atmospheric and Oceanic Prediction ([email protected] ), 49 pp. and Processes, 6-8 Aug 03, 132-137 Mata, C. T., 2009: Personal Vaisala, 2009: Vaisala FALLS 5.0 Fault Communication, ASRC Aerospace Analysis and Lightning Location Corp., Kennedy Space Center, M/S: System datasheet, Vaisala, Inc., ASRC-10, FL 32899, http://www.vaisala.com/files/Falls–Data [email protected], (321) 867- sheet.pdf, 2008, 2 pp. 6964 Vaisala, 2008: Vaisala Thunderstorm Murphy, M. J., K. L. Cummins, N. W. S. Lightning Network Performance. Demetriades, and W. P. Roeder, 2008: Evaluation Program (NPEP) datasheet, Performance Of The New Four- Vaisala, Inc., www.vaisala.com/files/ Dimensional Lightning Surveillance NPEPDataSheet.pdf, 2008, 2 pp. System (4DLSS) At The Kennedy Valine, W. C. and E. P. Krider, 2002: Space Center/Cape Canaveral Air Statistics and characteristics of cloud- Force Station Complex, 13th to-ground lightning with multiple ground Conference on Aviation, Range, and contacts, Journal of Geophysical Aerospace Meteorology, Research, 107, D20, 4441, 20-24 Jan 2007, 18 pp. doi:10.1029/2001 J D001360. Weems, J. W., C. S. Pinder, W. P. Roeder, and B. F. Boyd, 2001: Lightning Watch And Warning Support to Spacelift Operations, 18th Conference on Weather Analysis and Forecasting, 30 Jul-2 Aug 01, 301-305 Ward, J.G., K.L. Cummins, E.P. Krider, 2008: Comparison of the KSC-ER cloud-to-ground lightning surveillance system (CGLSS) and the U.S. National Lightning Detection Network TM (NLDN), 20th International Lightning Detection Conference, 22-23 April 2008', 7 pp.