README.TXT file for SPC TCTOR DATABASE V2.0 [2012 Version, Data thru 2011] Roger Edwards Storm Prediction Center roger.edwards@noaa.gov 1. BACKGROUND (adapted from Edwards 2010a SLS Conference Preprint -- TCTOR paper): a. Justification and Tornado-record Characteristics Given the concerns with the tornado data overall, and by extension, with TC tornado records, a more focused, updated and consistent basis of analysis should be used than those available in existing literature. This is in order to: a) ameliorate impacts of systematic “shocks” to the data record (Thorne and Vose 2010), such as that resulting from NWS modernization in the early 1990s, and b) still offer a large sample (see Doswell 2007 for thorough discussion on sample size issues with tornado data). To those ends, this study analyzes TC tornado records spanning 1995–2009 (mapped in Fig. 2). The chosen period corresponds essentially to the full national deployment of WSR-88D units, and as such, is entirely within the framework of modern warning and verification practices based thereon. Each potential TC tornado record in the conterminous U.S. was examined individually, by comparison with surface and upper air maps, archived satellite photos and/or imagery derived from archived NEXRAD Level II data (Kelleher et al. 2007), to determine its presence within the circulation envelope of either a classified or remnant tropical cyclone. Qualifying events were segregated from the nationwide Storm Prediction Center (SPC) one-tornado (ONETOR) database and assigned to their respective TCs by name. TC tornado records initially retained all information from the parent ONETOR database (e.g., identification number, time, date, states, latitude/longitudes for path ends, EF rating, casualties, monetary damage estimates, etc.). A “smooth” TC-tornado dataset (TCTOR) was compiled, expunging categorical redundancies (e.g., multiple columns specifying the same date and time), any ONETOR categories impertinent to this study (e.g., county FIPS numbers and a useless constant “3” that denotes the CST time zone already given), as well as incorporating metric and UTC unit equivalents. During the conversion of NCDC segmented data to ONETOR, intrastate county segments are stitched together to form whole-tornado tracks. However, a state border-crossing tornado in ONETOR still is parsed into one segment per state, albeit with a duplicate entry number so the state-segments still can be plotted as a single track with mapping software. Seven such events (0.06%) populated the 1995-2008 TC tornado record, each of which was combined into a single TCTOR path entry with two states listed (e.g., GA-SC in the “State” column). b. Incorporation of TC Information National Hurricane Center (NHC) best-track records (Hurricane Data, a.k.a. HURDAT) then were examined for each tornado event, from which the most recent 6-hourly central pressure and wind intensity were applied to each tornado. Using those wind maxima (for classified systems), tornadoes were binned according to their correspondence with a tropical depression (TD), tropical storm (TS), hurricane (H), or a combination of all non-tropical and post-classification categories (N). An N may include either official extratropical classifications, a change to subtropical (as with Allison of 2001), or as with TC Erin in 2007, a remnant low (Brennan et al. 2009, Monteverdi and Edwards 2010). Linear interpolation between 6-hourly HURDAT center positions yielded a cyclone-center estimate at tornado time. The distance D from this interpolated TC center to the starting position of each tornado then was computed across a great-circle surface arc, in a variation of the spherical law of cosines (e.g., Sinnott 1984) that uses latitude (lat) and longitude (lon) in radians, as follows: D = re * ( cos-1 { [sin(latTOR) * sin(latTC)] + [cos (latTOR) * cos(latTC) * (1) cos(lonTOR-lonTC)] } ) where re is the mean radius of earth. The subscripts TOR and TC signify tornado and TC-center positions respectively, neglecting any error at such relatively small radial angles that might arise from the centrifugal difference between polar and equatorial re. Cartesian bearing of reports from TC center was included in TCTOR, to foment analyses of cyclone-relative tornadic traits. The most recent 6-hourly TC classification (hurricane, tropical storm or tropical depression), central pressure and max wind also were logged for each tornado. For TC-remnant tornadoes occurring post-HURDAT, the low location, pressure, and max wind were estimated subjectively from surface analyses, while the system also was assigned a post-classification category N (not classified as tropical cyclone). For numerical sorting and ranking purposes, TC classifications were assigned as in the column-by-column table below. Records in TCTOR do include inland TC remnants interacting with low-level baroclinic zones, as long as: a) a closed surface low can be identified; and b) upper air data at the nearest 12-hourly synoptic times (0000 and 1200 UTC) indicate warm-core characteristics in the mid-troposphere (i.e., 700-500 hPa). Records also are included from any tornadic TCs that failed to make U.S. landfall (e.g., recurving just offshore from the Carolinas or Florida Keys, or entering northern Mexico with tornadoes in Texas). Where NHC classifications had been discontinued, yet TC remnants still were apparent, the nearest hour’s surface data were analyzed to estimate location of the pressure low--which, for this dataset, corresponded within analytic scale to the cyclone center derived from drawing streamlines except in the late stages of inland frontal interaction. Each TC-related datum was incorporated into TCTOR alongside the corresponding tornado information. c. Caveats and Sources of Error Several sources of potential error or uncertainty exist in TCTOR, as in ONETOR at large, that likely are nonlinear and certainly are not readily quantifiable, with the addition of some from HURDAT as well. Position errors of tornadoes, in an absolute sense, may arise from either: a) uncorroborated location estimates provided to NWS, especially for non-damaging events or weak tornadoes not causing damage detectable above that from the TC; or b) the inherent imprecision of the location reporting and translation convention in Storm Data, whereby azimuth and range (in miles) from a town typically is logged, then converted to latitude and longitude out to two decimal places. Linear interpolation of 6-hourly center fixes becomes less reliable where rapid changes in translational motion occur between them. Sharp accelerations or decelerations within temporal bins are possible, but uncommon, and introduce some potential error on calculations of tornado distance and direction from center. So do any temporal imprecisions in tornado reporting. In fact, a marked tendency exists ? e.g., 64% of TCTOR records from 1995-2008 ? for times in whole minutes to end in the digits 0 or 5. In the real atmosphere, no physical basis is apparent for any amount >20% of such timing. HURDAT records, meanwhile, truncate TC center fixes to 10**-1 degree of latitude and longitude. Inland center-fixes for decaying systems, particularly those of less than TD strength or located between relatively sparse surface observations, also may be subject to the same precision uncertainties that afflict any subjective analysis of a low’s location. Therefore, TC center-relative tornado positions given in TCTOR should not be interpreted too precisely on an individual basis, spatially or temporally, but instead assessed with respect to relative characteristics and broader, TC-scale tendencies. Previous studies incorporating tornado positions relative to TC positions also are encumbered with such uncertainties, whether or not explicitly mentioned therein, simply by virtue of the imprecisions intrinsic to the ONETOR and HURDAT data. The primary difference with TCTOR, aside from the decadal domain, is in the individually-assessed, "manual" technique for selecting TCTOR events. While time-intensive, this method is believed to offer the most accurate possible record as compared to existing TC-tornado climatologies. This is because TCTOR logs events without regard to fixed radii from TC center, inland extent, temporal cutoffs before or after landfall, or other such arbitrary and readily automatable thresholds that either may: a) exclude bonafide TC tornadoes outside the spatial cut-offs, or b) include somewhat proximal but non-TC tornadoes unnecessarily. Subjective analysis is subjective by definition, and a few fringe events for inland-decay stages may be open to argument as to whether they fall under the cyclone envelope. Such events constitute <<1% of the data. TCTOR is flexible, in that it will grow with time and be open to evidence-based revision. The author intends to update TCTOR on an annual basis as HURDAT and ONETOR data become available for the previous hurricane season’s activity. Additionally, TCTOR will be made available online for research and independent analysis, and may be amended on a post-facto basis as any errors are discovered, new information arrives and/or additional analyses are performed by other researchers on any historical TC tornadoes therein. Such adjustments made to variables common to the two datasets (i.e., occurrence time, location, path width, etc.) will be passed to corresponding ONETOR entries for consistency. Since any entry in TCTOR is open to revision, given sufficient evidence, any analyses derived from the dataset (including in this study) should be considered “best available” at the time, and potentially subject to change as well. Some changes may occur in the future in the way that tornado data are recorded overall, such as addition of decimal places to latitudes and longitudes for greater spatial precision, or greater texturing of path and damage information (Edwards et al. 2010b -- EF Scale paper at SLS Conference). 3. FORMAT Data is in MS Excel format, which can be saved as CSV from within MS Word for ready ingestion and analysis by other software. Columns (L-R) are as follows: A. Yearly tornado tracking number (from ONETOR). Quasi-sequential integer denotign the order a tornado arrived in Storm Data records. B. Year C. Month (UTC) D. Date (UTC) E. Time (UTC) F. Month (CST -- the time zone for SPC ONETOR source data) G. Date (CST) H. CST Time I. State(s). Another difference with TCTOR: Sttate-crossing tornadoes are combined into one row here. J. Yearly tornado tracking number by state. Discontinued 2008. K. F or EF Scale damage rating. EF took effect in Feb. 2007. L. Number Injured M. Number Killed N. Damage code (estimated property loss information). Prior to 1996 this is a categorization of tornado damage by dollar amount (0 or blank-unknown; 1<$50, 2=$50-$500, 3=$500-$5,000, 4=$5,000-$50,000; 5=$50,000-$500,000, 6=$500,000-$5,000,000, 7=$5,000,000-$50,000,000, 8=$50,000,000-$500,000,000, 9=$5000,000,000.) From 1996, this is tornado property damage in millions of dollars. Note: this may change to whole dollar amounts in the future. Entry of 0 does not mean $0. O. Crop code: estimated crop loss in millions of dollars (started in 2007). Entry of 0 does not mean $0. P. Starting latitude of tornado, decima degrees. Q. Starting longitude of tornado. R. Ending latitude of tornado. S. Ending Latitude of tornado. T. Path Length (mi) U. Path Length (km) V. Path width (yd) W. Path width (m) X. Number of states affected. Shading in segment columns denotes state border crossers combined from ONETOR Y. TC Name and Year Z. Last 6-hourly official "OFCL" TC latitude (from HURDAT if classified therein) AA. Last 6-hourly official TC longitude AB. Last 6-hourly official max TC wind in kt AC. Last 6-hourly official minimum central pressure AD. Last 6-hourly TC category (Saffir-Simpson value) AE. Category Code as follows: 5 = Major Hurricane, Category 5; 4 = Major Hurricane, Category 4; 3 = Major Hurricane, Category 3; 2 = Hurricane, Category 2; 1 = Hurricane, Category 1; 0 = Tropical (or Subtropical) Storm; -1 = Tropical Depression; -2 covers every other post-landfall category (extratropical, remnant low, no longer classified, etc.) AF. Next 6-hourly TC latitude (from HURDAT if classified therein, otherwise from surface map) AG. Next 6-hourly TC longitude (from HURDAT if classified therein, otherwise from surface map) AH. Linearly interpolated TC position latitude (based on time elapsed since last 6-hourly) AI. Linearly interpolated TC position longitude (based on time elapsed since last 6-hourly) AJ. Cartesian (relative to true N) tornado bearing from center, in deg. AK. Tornado distance from center (km) AL. Tornadoes for each TC (column should be removed for re-sorting spreadsheet) AM-AN. Year and tornadoes for each year (columns should be removed for re-sorting spreadsheet) 4. NEWS (most recent first) 31 Mar 12: Updated for 2011 data. 11 Sep 11: Updated for 2010 data. 16 Sep 10: First pre-release version (v0.1) available for limited preliminary research (requests for corrections are out for obviously suspicious data imported from ONETOR, such as path widths in 100s of miles).