{"id":1,"date":"2014-08-29T17:50:49","date_gmt":"2014-08-29T17:50:49","guid":{"rendered":"http:\/\/www.karlhartkopf.com\/blog\/?p=1"},"modified":"2019-03-13T02:59:05","modified_gmt":"2019-03-13T02:59:05","slug":"time-com-makes-beautiful-incorrect-maps-by-compounding-flawed-data-with-silly-mistakes","status":"publish","type":"post","link":"http:\/\/www.karlhartkopf.com\/blog\/2014\/08\/29\/time-com-makes-beautiful-incorrect-maps-by-compounding-flawed-data-with-silly-mistakes\/","title":{"rendered":"Time.com Makes Beautiful Incorrect Maps By Compounding Flawed Data With Silly Mistakes"},"content":{"rendered":"

Using big open data, Time.com has produced and published “How Safe Am I From a Natural Disaster”<\/a> in which it presented\u00a0a “disaster index” for each county in the United States based on National Weather Service (NWS) and National Oceanic and Atmospheric Administration (NOAA) data. \u00a0While the idea is exciting, the\u00a0resulting piece doesn’t recognize the\u00a0flaws in the source data and\u00a0compounds those flaws by making some mistakes of their own in the process.<\/span><\/p>\n

Issues with this interactive map and dataset\u00a0are best illustrated by examining Time.com’s data for Orleans Parish, Louisiana (LA)<\/a>, which gets you the table below *:<\/span><\/p>\n

\"New<\/a>

New Orleans – Orleans Parish Natural Disaster Data Detail From Time.com Retrieved 10:30 am EST August 29, 2014<\/p><\/div>\n

Choosing\u00a0Orleans Parish (parishes are Louisiana’s equivalent of counties) for comparison is intentional, given that over 1,500 people died during Hurricane Katrina<\/a>\u00a0in 2005, most of them residents of New Orleans, which shares\u00a0the same boundary as Orleans Parish.\u00a0 In addition to being one of the five deadliest hurricanes in history, Hurricane Katrina was the costliest ever natural disaster in the United States causing over $100 billion of\u00a0damage. <\/span><\/p>\n

Yet Orleans Parish does not appear on Time.com’s top ten list of “Most Dangerous Counties”, with Time.com ranking it as the 266th most dangerous county in the United States while ranking Ocean County, NJ the 1st most dangerous in the nation<\/a>. \u00a0Ocean County’s most dangerous ranking is assigned in\u00a0spite of the highest number of confirmed deaths from a single event over the last 65 years was 37 dead from Hurricane Sandy in 2012<\/a>\u00a0(btw, those deaths are not found this Time.com project), with\u00a0few other deaths beyond those. Furthermore, Time.com indicates only property damage on August 28, 2005 and nothing for subsequent days after Hurricane Katrina.<\/span><\/p>\n

To figure out why this seemingly inexplicable result exists, let’s start by identifying the source data from which the mapping and analysis was generated.\u00a0 Time.com claims that “The disaster index is calculated from all records in the National Oceanic and Atmospheric Administration storm events database, which includes earthquakes and tornados [sic] from 1950 and 42 other event types since 1996.”\u00a0 Here is a summary of NOAA source descriptions<\/a>:\u00a0<\/span><\/p>\n

    \n
  1. National Weather Service (NWS) Storm Prediction Center which includes Tornadoes (1950-1992), Thunderstorm Winds (1955-1992) and Hail (1955-1992)<\/span><\/li>\n
  2. National Weather Service (NWS) Weather Offices keyed Storm Data files sent to National Climatic Data Center (NCDC) in Word Perfect 5.0 format on 3.5″ floppy diskettes and imported into more unified data files.\u00a0 They note that “These data had many inconsistencies in the spelling of event types and were standardized into the 48 current event types at NCDC in 2013.”<\/span><\/li>\n
  3. National Weather Service (NWS) Paradox Database Files: I can’t do any better job describing how dirty the data was than letting NOAA’s description speak for itself; “Beginning in January 1996, the NWS began using Storm Data for forecast verification purposes and the NWS decided to collect the data at the NWS Headquarters in Silver Spring, MD. They selected Borland\/Corel Paradox format for their database and supplied the NCDC with the raw data files, which were then used for the Storm Data publication and inclusion into the Storm Events Database. From 1996-1999, the event type field was a free-text field so there were many, many variations of event types. Most of the events were standardized into the 48 current event types in 2013. In 2000 the NWS added a drop-down selector for Event Type on the data entry interface, which standardized the Event Type values sent to NCDC.”<\/span><\/li>\n
  4. National Weather Service (NWS) Windows SQL Server Database Exported Comma-Separated Text (CSV) Files: Beginning in October 2006, the NWS supplied NCDC a comma separated (CSV) text files that NCDC import into its own database for the Storm Data publication and Storm Events Database.<\/span><\/li>\n<\/ol>\n

    Time.com’s first failure was not to prominently note NOAA’s own overt recognition of messy data, data often collected and recorded without rigorous field specifications and definitions. While the NOAA data descriptions note\u00a0that the NWS and NOAA cleaned them up somewhat, the data remains messy.<\/span><\/p>\n

    Delving into how messy this data gets us answers on the faulty specifics. In the Time.com Orleans Parish data table produced by the Time.com Safest Counties tool (shown above in “New Orleans – Orleans Parish Natural Disaster Data Detail”), property damage is listed as only on August 28, 2005, with no deaths at all *** and only direct costs for August 28th and 29th, in spite of over 50 levees breaching causing billions of dollars worth of flood damage and hundreds of deaths.<\/span><\/p>\n

    Pulling New Orleans storm event data directly from the National Weather Service’s Storm Events Database<\/a>, Table 1: Hurricane Katrina Late August NWS – NOAA Storm Events Entries (abbreviated) contains two faulty entries of interest:<\/span><\/p>\n

    Table 1: Hurricane Katrina Late August NWS – NOAA Storm Events Entries (abbreviated)<\/span><\/p>\n

    \n\n\n\n\n\n
    EVENT_ID<\/td>\nCZ_NAME_STR<\/td>\nBEGIN_LOCATION<\/td>\nBEGIN_TIME<\/td>\nDEATHS_DIRECT<\/td>\n\u00a0DAMAGE_PROPERTY_NUM<\/td>\n<\/tr>\n
    5475622<\/td>\nORLEANS (ZONE)<\/td>\n8\/28\/2005<\/td>\nHurricane (Typhoon)<\/td>\n0<\/td>\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 3,560,000,000<\/td>\n<\/tr>\n
    5475650<\/td>\nORLEANS (ZONE)<\/td>\n8\/29\/2005<\/td>\nStorm Surge\/Tide<\/td>\n0<\/td>\n\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 17,900,000,000<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

    It is now clear why Time.com reported no deaths; the lack of deaths was in the Weather Service’s database! ***<\/span><\/p>\n

    The next mistake was Time.com’s and perhaps the more significant one; exclusively using the number of claims as opposed to deaths, damages, crop damage or a weighted assessment of multiple criteria. Using this single, significantly flawed database compounded the problem and led to very problematic conclusions.<\/span><\/p>\n

    There are a few lessons for data analysts and data team journalists:<\/span><\/p>\n