Advances in Meteorology

Advances in Meteorology / 2015 / Article

Research Article | Open Access

Volume 2015 |Article ID 649614 |

Jason Vargo, Qingyang Xiao, Yang Liu, "The Performance of the National Weather Service Heat Warning System against Ground Observations and Satellite Imagery", Advances in Meteorology, vol. 2015, Article ID 649614, 15 pages, 2015.

The Performance of the National Weather Service Heat Warning System against Ground Observations and Satellite Imagery

Academic Editor: Filomena Romano
Received31 Mar 2015
Accepted26 May 2015
Published29 Jun 2015


Deadly heat waves are increasing with climate change. Public forecasts and warnings are a primary public health strategy for dealing with such extreme weather events; however, temperatures can vary widely within the administrative units used to issue warnings, particularly across urban landscapes. The emergence of more frequent and widely distributed sources of urban temperature data provide the opportunity to investigate the specificity of the current National Weather Service (NWS) warnings and to improve their accuracy and precision. In this work, temperatures from distributed public weather stations, NWS heat advisories and warnings, and land surface temperature imagery throughout two large metropolitan areas, Atlanta and Chicago, during the 2006–2012 summers are considered. We investigate the spatial variability of hazardous temperatures and their agreement against NWS advisories. Second, we examine the potential for thermal imagery to replicate National Weather Service heat warnings. Observations from weather stations exhibit varying degrees of agreement with NWS advisories. The level of agreement varied by station and was not found to be associated with the station’s proximate land cover. Air temperatures estimated from satellite imagery correspond with NWS Advisory status regionally and may enable creating more refined public warnings regarding hazardous temperatures and protective actions

1. Introduction

The potential for future extreme heat events that pose danger to human health is increasing with global climate change. As Hansen and colleagues at NASA’s Goddard Institute for Space Studies put it, we have been loading the climate dice for the last 30 years, increasing the likelihood of meteorological events that used to occur once ever hundred years [1]. Models and recent historical records echo this sentiment with data and the recent news that atmospheric CO2 concentrations passed 400 ppm for the first time in 3 million years imply that these trends are likely to continue. Early and accurate warnings are a crucial component for adapting to more frequent hazardous heat events.

The frequency, duration, areal coverage, and intensity of heat waves are expected to increase for most populated places because of global climate change [2, 3]. Such events have caused major episodic mortality, including an estimated 70,000 excess deaths in Europe in 2003 [4] and more than 55,000 in Russia during July and August of 2010 [5, 6]. Heat wave characteristics like intensity and seasonality are also changing in ways that increase their potential hazard to health [7]. In many places, early warning systems are a key, if not the primary, component of measures for avoiding heat related deaths and illness. In the United States the National Weather Service (NWS) issues these warnings. In this work we are concerned with the spatial resolution and specificity of the issued warnings. The predictions for urban areas, in particular, could be improved given that cities are typically hotter than surrounding areas, concentrate people, and demonstrate great variability in temperatures and population vulnerabilities.

(a) Adaptation Strategies. The relationship between heat and health is well understood and known to change over time as well as with location and population characteristics [8]. Future changes in climate will increase adverse health impacts [9, 10]. Adapting to changes in extreme heat events has largely followed three strategies: identifying vulnerable populations, ensuring access to mechanical cooling (air conditioning), and implementing early warning systems [11, 12].

Several studies using large population datasets and several summers of data demonstrate a positive relationship between temperatures and excess all-cause mortality on days that are abnormally warm compared to long-term summer weather for the region [1315]. Public warning systems for heat are targeted toward managing such events. Response activities for heat events have been institutionalized as part of relatively recent efforts in some cities, including Chicago, Milwaukee, and Philadelphia following deadly heat waves in the early and mid 1990s [16, 17].

Deadly heat events have led to greater attention on individual and population level factors affecting heat vulnerability [18]. These include age, social isolation, housing type, income, and ethnicity [19, 20]. Adaptive capacity of individuals is also a key determinant in whether and to what degree hazards of climate change will result in adverse health effects [21]. Understanding vulnerabilities to heat has led to efforts such as censuses of susceptible populations and door-to-door visits during heat events that have been shown to successfully reduce the related excess mortality [17, 22]. These approaches tend to be expensive to implement and rely on significant human resources for scaling up.

The most effective protection against hyperthermia in extreme weather is air conditioning (AC). The increased prevalence of this technology is one reason for declining heat-related mortality in the US despite increasing frequency of heat events [23]. However, reliance on energy-intensive adaptation strategies like AC results in greater energy use and the production of waste heat. Thus such adaptation measures produce positive feedbacks on global climate change and can exacerbate the urban heat island effect [24].

The general public are familiar with warning systems for weather-related hazards including floods, tornados, winds, and severe storms. As an adaptation strategy, warnings are important for quickly reaching large numbers of people at low cost and providing targeted messaging that can help minimize damages. Heat warning systems have been shown to save lives during extreme heat [16], but heat waves have received less attention than other natural disasters, in part, due to the fact that there is less aftermath (particularly property damage). Even when the public is aware of warnings and climatic conditions, there is often far less awareness of what protective actions should be taken [25].

(b) The National Weather Service Warnings. Currently the NWS uses four products based on the heat index (HI), a metric combining temperature and relative humidity to describe heat stress and discomfort, to issue hazard notifications. Heat index values are forecast for three to seven days out and are based on an ensemble of model and human forecasts. Day three forecasts are based on gridded model output statistics (MOS) temperature and dew point temperature forecasts. Forecasts on days four through seven rely on minimum and maximum 24-hour temperatures and dew points at 00 and 12 UTC supplied by the Weather Prediction Center’s (WPC) 5-km resolution grid data products. Probabilities for heat index values exceeding thresholds are also provided in the form of color filled displays. Probabilities are informed by uncertainty information from an ensemble of modeled forecasts and follow a normal distribution around mean MOS or WPC values assessed on a regular 20-km resolution grid [26].

The NWS produces criteria for extreme weather and potentially hazardous conditions.

Excessive Heat Outlook may be issued 3 to 7 days prior to a heat episode requiring issuance of a heat warning, provided forecaster confidence is relatively high.

Excessive Heat Watch may be issued 12 to 48 hours prior to heat episode with a 50 percent chance or greater of daytime heat indices equal to or greater than 110°F for at least two consecutive days.

Heat Advisory is issued in the first and/or second period when there is an 80 percent chance or greater of daytime heat indices equal to or greater than 105°F (40.6°C) for at least two consecutive days.

Excessive Heat Warning is issued in the first and/or second period when there is an 80 percent or greater of daytime heat indices equal to or greater than 110°F (43.3°C) for at least two consecutive days.

These criteria can be modified for specific regions and cities as the local offices see fit. For example, the city of Chicago uses additional criteria related to nighttime low temperatures to trigger advisories.

The NWS’s advisory guidelines are, by default, applied uniformly for different places across the country. In this respect they have not been representative of regional adaptation behaviors and practices. This leads to warnings and advisories in some areas that may be issued too frequently or not frequently enough and that misrepresent actual weather-related risks for people. In both cases the issuances may lead to skepticism and neglect for the NWS notifications and decrease their effectiveness. In recent years some areas, like the city of Chicago, have been more proactive about defining their own criteria not only for daytime high as well as nighttime low HI. Other cities, like Philadelphia, PA, has employed their own heat watch and warning systems [27].

Other concerns with the NWS advisories and warnings relate to urban areas where anthropogenic land cover modifications may elevate temperatures dangerously high even outside periods of regional extreme weather. Hazardous thermal conditions may also exist in small pockets of the urban landscape, which displays great variability in temperatures over space. The current NWS notification system is based on low resolution forecast model grids, relative to heterogeneity in urban landscapes and surface temperatures, and releases warnings and advisories at the county level. Because counties vary in size from place to place and may not completely contain major population centers, the NWS products could fail to provide specific warnings and meaningful guidance for large numbers of people. Even if the products are accurately predicting regional hot weather, detail to the spatial distribution of temperatures within the region could improve warnings, individual adaptation measures, and larger response efforts.

Previous work has examined whether the HI values used are indicative of health threats by comparing HI values and mortality for various locations [28]. Like the NWS products the HI values are often collected at a single site and used to represent an entire region. Regionally representative combinations of temperature and humidity have also been compared with mortality and morbidity rates in places [29]. The outcome of such studies has shown that HI can be a useful metric for informing decisions to protect public health in some locations. Alternatives based on weather classifications, like the Synoptic Spatial Classification (SSC), have also been examined for agreement with variability in mortality. The NWS system was compared against public perceptions of heat risk, a determining factor in whether precautionary actions are taken, and found to vary between regions and across age, income, gender, and ethnicities within regions [30].

(c) New Data Sources. In situ and remote sensors of temperature at or near the Earth’s surface offer the potential to improve the current warning system by increasing its specificity, particularly with regard to location. These data sources are new and numerous. They provide data with greater spatial coverage, frequent revisit times, and low cost. Distributed weather stations may have a shorter history than reliable long term stations used in climatological studies but can effectively describe spatial patterns in temperatures across the urban landscape. Satellite data products like land surface temperature (LST) from the Moderate Resolution Imaging Spectroradiometer (MODIS) sensors offer 1 km spatial resolution data for areas around the globe up to four times daily [31]. Such products are systematically processed to remove inconsistencies and distortions, as well as to calculate usable temperature outputs. While these data cannot currently offer predictive weather information, they are nonetheless useful for analyzing and describing spatial variation and patterns which, when combined with predictive models currently in use, can improve information to the public and the adaptive response.

(d) Investigations. In this work we consider temperatures from distributed public weather stations, NWS heat advisories and warnings, and land surface temperature imagery throughout two large metropolitan areas, Atlanta and Chicago, during the summers from 2006 to 2012. We first investigate the spatial variability in hazardous temperatures and agreement with advisories issued by the NWS, and second we examine the potential for a widely available thermal imagery product to replicate National Weather Service heat warnings.

Here we test two primary hypotheses. The first is that measured data from local stations will exhibit the conditions for heat warnings at the same time and place as NWS forecasted warnings. We expect to find that the NWS advisories miss some instances where hazardous heat conditions are met at locations within counties. We will examine some of the characteristics of such locations, should they exist.

Second, we examine specific days where there is some evidence of extreme heat in either the NWS or surface station records. These days help describe whether estimates of near surface air temperatures at subcounty resolutions can improve the identification of hazardous thermal conditions. We investigate estimates derived from LST collected by satellites coupled with station measurements. We expect to find that LST and station records are well correlated, particularly on NWS-determined advisory days. This would suggest that satellite data may provide a means of identifying hazardous heat conditions where stations do not exist and at scales finer that forecast models. The findings from these tests are expected to improve the existing national heat warning system by demonstrating the need for subcounty specification that the current system misses and by providing evidence for how existing data sources could be used to complement the current NWS system.

2. Methods

2.1. Study Locations

This study focuses on two metropolitan areas in different climatic zones of the US: Atlanta, GA, and Chicago, IL (see Figure 1). Atlanta is an inland city in the southeastern US (centered near 84.37 W, 33.74 N). Atlanta’s historic downtown and central city are located primarily in Fulton County. A portion of the city sits in DeKalb County east of Fulton. Three other counties (two north and one south of the city) were included in the analysis to cover the core of the metropolitan area: Clayton, Cobb, and Gwinnett. Chicago is in the midwest and sits on the large water body of Lake Michigan (centered near 87.63 W, 41.89 N). The city of Chicago sits within Cook County. Four surrounding counties (DuPage to the west, Lake, IL to the north, Will to the southwest, and Lake, IN to the southeast) were also included in the analysis of the Chicago Metro area.

2.2. Heat Advisory Data

As for the historical record of heat warnings, we used NWS-issued Heat Advisories and Excessive Heat Warnings. Text records of the statements issued by NWS stations are archived and maintained by the National Oceanic and Atmospheric Administration’s (NOAA) National Climatic Data Center (NCDC). The data are stored in the Hierarchical Data Storage System (HDSS), which includes a tape robotics system for data archived on tape. NCDC provides direct online access to these data through the HDSS Access System (HAS). Weather statements were issued from two NWS stations responsible for the Atlanta (KFFC, Peachtree City, Falcon, GA) and Chicago (KLOT, Lewis University, IL) Metropolitan Areas. Specifically, recorded Heat Advisories and Excessive Heat Warnings are contained in Non-Precipitation Watches, Warnings, Advisories Bulletins (bulletin ID WWUS7) and are accessible through the Service Records Retention System (SRRS) Text Products/Bulletin Selection interface (available at Text file compilations of bulletins were obtained for May 1 to September 30 for the years 2006–2012. Individual text files were combed to identify mention of “HEAT” and were further examined to determine the day-county to which Heat Advisories or Excessive Heat Warnings applied.

2.3. Weather Station Data

Measurements from weather stations serve as diagnostics of whether or not a hazardous heat event actually occurred for a given location and day. Two data sources were used for weather station data. First, temperature and humidity data from NCDC’s Surface Data, Hourly Global (DS3505) data were obtained for Atlanta Hartsfield International Airport (USW00013874) and Chicago O’hare International Airport (USW00094846) from the NCDC website ( Second, Weather Underground, a commercial weather service provider, established a personal weather station network, which they use to inform their BestForecast system. The Weather Underground network allows individuals to share real-time weather information recorded by personal weather stations. The measurement intervals range between 1 and 60 minutes and differ by station. Temperature (°F) and relative humidity (%) from 64 stations in five counties in the Atlanta metro area and 144 stations in five counties in the Chicago metro area (see Figure 1) were obtained from the Weather Underground website ( accessed in Sep-Oct 2013). Stations within the ten counties were queried for hourly temperature and humidity data back to May 2006. Since the Weather Underground network is continuously expanding, stations used in the analysis may differ from those assembled when this paper is being read.

To control the quality of data and eliminate outliers, for each city, we removed observations beyond the range of mean ± five standard deviations. For each station, days with less than 16 hours valid hourly data and years with less than 75% valid daily data were also excluded to eliminate possible sampling bias. The hourly average temperature and relative humidity values were processed to calculate the HI based on algorithms provided by the NWS ( To evaluate if stations experienced hazardous heat conditions, the NWS definition of Heat Advisory was used, such that if there are at least two consecutive days with at least one hourly HI greater than 105°F (40.6°C) in each day, these days were marked as heat wave days.

2.4. Land Surface Temperature Imagery

The potential for regularly collected satellite data to provide more information on within-county variability in hazardous temperatures is examined through images from May through September 2011-2012 in the Atlanta and Chicago Metropolitan cores.

The daily 1 km land surface temperature (LST) data [32] from MODIS aboard the National Aeronautics and Space Administration (NASA) Earth Observing System (EOS) Aqua and Terra satellite, labeled as MYD11_A1 and MOD11_A1, were obtained from the Goddard Space Flight Center ( We extracted the LST parameter “LST_Day_1 km” and “LST_Night_1 km” at 1 × 1 km spatial resolution over Atlanta and Chicago. Retrievals with quality flags of “fair consistency” and “good consistency” were included in the following analyses. The 1 km monthly MODIS normalized difference vegetation index (NDVI) product (MOD13 A3) was obtained from the Goddard Space Flight Center (

To integrate the air temperature measurements, spatial predictors, and LST retrievals, we projected all the data to the MODIS Sinusoidal Grid. Each Weather Underground station was assigned to the LST pixels of the same grid cell. The NDVI value, average elevation value, and percent impervious area value were constructed based on the same grid.

Following a published methodology [33] we developed a linear mixed effect model to estimate the daily average air temperature from LST retrievals. The model consisted of day-specific random intercepts and random LST slopes as follows:where (°F) is the measured air temperature in grid cell on day ; denotes the fixed intercept; denotes the day-specific random intercept; are the fixed slopes of LST, elev, impervPCT, and NDVI, respectively; is the day-specific random slope of LST; is the land surface temperature retrieval in grid cell on day ; is the percent of impervious area in grid cell ; is the monthly NDVI value in grid cell for the month in which day falls; (meters) is the mean elevation in grid cell ; is the error term in grid cell on day .

We also considered wind speed as a predictor in the model; however, preliminary results indicated that it was not significant and it was removed from the model. We used both daytime and nighttime LST retrievals from both Terra and Aqua satellite as the predictors and evaluated the model performance, respectively. The average night LST was selected as the predictor in the final model because it provides the highest accuracy and greatest coverage. The average nighttime LST is defined as the night LST from Aqua or the night LST from Terra if only one of them is available; if both Aqua and Terra night LSTs are available, the average night LST is the mean of these two values. Fivefold cross validation (CV) was used to validate our model performance. We randomly divided the data into a training dataset (80%) and a testing dataset (20%) and then made predictions for testing dataset using the model fitted from the training dataset. The model was trained and tested for five times by different training and testing datasets to ensure that each data point ends up in the testing dataset exactly once. This process was repeated for 1,000 times and the root mean-square prediction errors (RMSPE) and were reported to estimate the model prediction precision. The RMSPE were calculated as follows:where is the number of observations and and are the th predicted and observed value, respectively.

3. Results

3.1. Data Descriptions

Among records from the airport and Weather Underground weather stations from May–September 2006–2012 there are 26,943 and 70,312 valid station-day data records in total for Atlanta and Chicago, respectively, and 1,172 and 1,499 station days were marked as heat wave days. This is 4.3% of daily observations meeting Heat Advisory conditions in Atlanta and 2.14% in Chicago. In the years 2011 and 2012 there was far less missing data among stations. In these years there were 14,086 and 34,347 observations with 694 (4.9%) and 1,143 (3.3%) station days exceeded Heat Advisory criteria.

Both 2011 and 2012 featured several NWS Heat Advisories in the two metro regions. Several years prior featured no NWS-issued Heat Advisories for the core metro counties. In Atlanta 2006, 2008, and 2009 had no Advisories, and in the Chicago counties 2006–2008 were without Heat Advisories (see Figure 2). In Atlanta, Heat Advisories were mostly applied to the five counties in the core metro area uniformly. That is, on days for which an advisory was issued it was applied to all counties. In only three instances Clayton county was issued a warning when other counties were not. In 2011 there were two Heat advisories in each month of July and August, and in 2012 there were four concurrent advisories that spanned June and July. In Chicago there was more variability in issuing NWS Advisories, particularly in July of 2011. In that month both Cook and DuPage Counties were issued six Advisories, Lake, Illinois had four, and both Lake, Indiana and Will County experienced eight. All five Chicago area counties experienced the same number of advisories in other months: one in August 2011, two in June 2012, and seven in July 2012.

The majority of the Wunderground Stations are found on “developed” land cover of some type. Of Atlanta’s 57 stations, most were located on high-, medium-, and low-intensity and open space developed land. Only 9 stations were located on land covers not considered developed (four evergreen forests, four deciduous forests, and one woody wetland). In Chicago 10 of the 137 (7.3%) stations were found on land covers other than some form of “developed.” All of Cook county’s 44 stations are found on low-, medium-, or high intensity developed land (see Tables 1 and 2).

Land cover (NLCD 2011)2011 2012 False pos. False neg.Obs
Norm Heat % Norm Heat %

  KGAATLAN58 Developed high intensity 146 6 4% 2 4 152
  ATLAirport Developed low intensity 153 0 0% 151 2 1% 6 0 306
  KGAMARIE40 Deciduous forest 120 26 18% 126 14 10% 2 34 286
  KGAACWOR9 Developed open space 151 2 1% 134 18 12% 3 15 305
  KGAPOWDE3 Developed open space 142 11 7% 148 5 3% 2 10 306
  KGAMABLE1 Evergreen forest 145 8 5% 149 4 3% 3 7 306
  MD0601 Developed open space 144 7 5% 146 5 3% 2 6 302
  KGAMARIE25 Developed open space 124 7 5% 1 4 131
  KGAMABLE3 Developed open space 142 7 5% 144 3 2% 2 4 296
  KGAMARIE43 Developed low intensity 113 5 4% 150 3 2% 2 2 271
  KGAMARIE42 Developed open space 148 3 2% 2 1 151
  KGAMARIE34 Developed low intensity 121 0 0% 2 0 121
  KGAMARIE35 Developed open space 148 0 0% 134 3 2% 5 0 285
  KGAMARIE20 Developed low intensity 153 0 0% 150 3 2% 5 0 306
  KGAROSWE4 Developed open space 152 0 0% 120 2 2% 6 0 274
  KTGAMARI2 Deciduous forest 132 2 1% 2 0 134
  KGAMARIE46 Developed medium intensity 134 2 1% 2 0 136
  KGAKENNE16 Developed open space 148 2 1% 2 0 150
  KGAKENNE15 Developed low intensity 120 2 2% 149 2 1% 4 0 273
  KGAMARIE39 Woody wetlands 148 2 1% 150 2 1% 4 0 302
  KGASMYRN2 Developed low intensity 128 0 0% 145 0 0% 6 0 273
  MAT950 Developed open space 147 0 0% 132 0 0% 7 0 279
  KGADECAT8 Developed open space 128 25 16% 136 17 11% 2 36 306
  KGAAVOND3 Developed high intensity 128 24 16% 2 22 152
  KGAAVOND4 Developed low intensity 147 2 1% 2 0 149
  KGAATLAN52 Evergreen forest 149 0 0% 148 0 0% 8 0 297
  KGAATLAN41 Developed low intensity 96 53 36% 114 24 17% 2 71 287
  KGAATLAN70 Developed open space 77 39 34% 0 35 116
  KGAALPHA10 Developed low intensity 128 22 15% 135 18 12% 1 33 303
  KGAATLAN67 Deciduous forest 127 26 17% 1 23 153
  KGAROSWE5 Developed low intensity 117 9 7% 131 13 9% 1 17 270
  MD5335 Developed open space 137 16 10% 4 16 153
  KGAATLAN56 Developed open space 125 13 9% 2 12 138
  KGAJOHNS2 Developed low intensity 143 10 7% 147 5 3% 4 12 305
  KGAATLAN54 Developed medium intensity 139 11 7% 129 2 2% 4 9 281
  KGAJOHNS5 Developed open space 151 2 1% 148 5 3% 3 2 306
  KGAATLAN16 Developed low intensity 148 4 3% 148 3 2% 3 2 303
  KGAATLAN37 Developed low intensity 145 4 3% 149 2 1% 4 2 300
  MAT910 Developed medium intensity 141 3 2% 143 3 2% 3 1 290
  KGAATLAN40 Developed low intensity 147 4 3% 149 2 1% 3 1 302
  KGAATLAN57 Developed open space 149 3 2% 150 2 1% 4 1 304
  KGAALPHA16 Developed open space 119 0 0% 0 0 119
  KGAPALME2 Deciduous forest 125 2 2% 2 0 127
  KGAATLAN50 Developed open space 148 2 1% 138 2 1% 4 0 290
  KGAATLAN49 Developed medium intensity 141 0 0% 149 0 0% 8 0 290
  KGAMOUNT2 Developed open space 147 0 0% 153 0 0% 8 0 300
  KGALILBU10 Evergreen forest 120 28 19% 112 36 24% 1 57 296
  KGADULUT7 Developed open space 114 34 23% 0 30 148
  KGALAWRE13 Developed open space 131 22 14% 140 5 3% 5 24 298
  KGALOGAN12 Evergreen forest 138 13 9% 127 5 4% 2 13 283
  KGASTONE4 Developed low intensity 147 6 4% 1 3 153
  KGALILBU11 Developed low intensity 122 4 3% 112 3 3% 3 2 241
  KGALILBU4 Developed low intensity 153 0 0% 151 2 1% 6 0 306
  KGALILBU8 Developed open space 151 2 1% 151 2 1% 4 0 306
  KGALAWRE6 Developed open space 151 0 0% 152 0 0% 8 0 303
  KGANORCR12 Developed low intensity 144 2 1% 151 0 0% 6 0 297
  KGANORCR14 Developed open space 148 0 0% 121 0 0% 8 0 269

Stations within each county are sorted by number of false negatives.

Land cover (NLCD 2011)20112012False pos.False neg.Obs

  KILWESTC4Developed low intensity15200%14900%160301
  KILCHICA106Developed high intensity15300%15121%140306
  KILCHICA86Developed high intensity14921%15300%140304
  KILCHICA52Developed high intensity15100%15032%130304
  KILLINCO6Developed medium intensity13521%15300%130290
  KILCHICA37Developed high intensity15121%15121%120306
  KILORLAN3Developed low intensity14221%14921%120295
  KILWESTC7Developed low intensity13932%15121%121295
  KILARLIN3Developed medium intensity15121%15032%110306
  KILMELRO2Developed low intensity14821%15032%110303
  KILCHICA30Developed medium intensity14621%15032%110301
  KILLYONS1Developed medium intensity13321%13632%110274
  KILLAGRA2Developed low intensity14421%12800%110274
  MAT224Developed medium intensity12822%13700%110267
  KILDESPL6Developed medium intensity15121%14853%101306
  KILEVANS2Developed medium intensity14943%14943%102306
  KILELKGR4Developed low intensity13732%14943%101293
  KILDEERF3Developed low intensity14700%14332%100293
  KILOAKPA1Developed high intensity13032%13643%101273
  MAS935Developed low intensity12922%13232%100266
  KILLYONS3Developed medium intensity14853%14764%94306
  KILPALAT4Developed low intensity14432%14764%92300
  MAT062Developed low intensity14453%14364%94298
  KILCHICA69Developed low intensity14521%14521%90294
  KILPALOS2Developed low intensity14332%13064%92282
  KILNORTH9Developed medium intensity13943%13354%92281
  KILCHICA112Developed high intensity15300%90153
  KILCHICA111Developed medium intensity13100%90131
  KILORLAN4Developed medium intensity13675%14000%84283
  KILBARRI7Developed medium intensity11733%14243%81266
  KILARLIN6Developed low intensity13521%13364%71276
  KILCHICA105Developed high intensity13000%70130
  MAU210Developed low intensity14932%60152
  KILOAKLA4Developed low intensity139128%14496%511304
  KILCHICA114Developed low intensity14843%50152
  KILWINNE3Developed low intensity14843%50152
  KILPALOS4Developed medium intensity14643%50150
  KILTINLE1Developed medium intensity14721%50149
  KILCHICA115Developed medium intensity14543%50149
  KILORLAN5Developed high intensity14443%50148
  KILCHICA107Developed medium intensity12943%50133
  KILCHICA68Developed medium intensity14296%46151
  KILARLIN4Developed low intensity13164%31137
  KILROLLI4Developed low intensity12986%22137
  KILCAROL5Developed low intensity11500%15100%150266
  KILWESTC8Developed medium intensity14932%15300%141305
  KILCLARE2Developed low intensity14732%13421%121286
  CHIAirportDeveloped open space15121%15032%110306
  KILNAPER15Developed low intensity15032%14943%101306
  KILNAPER7Developed low intensity15032%14943%101306
  KILDOWNE3Deciduous forest15032%14621%91301
  KILELMHU5Developed low intensity13654%14732%93291
  KILLISLE3Developed open space14832%14464%92301
  KILLOMBA5Developed low intensity12300%14943%90276
  KILWARRE2Developed low intensity14621%14243%91294
  KILWHEAT5Developed medium intensity13354%13700%93275
  KILWHEAT8Developed medium intensity12132%13543%91263
  MC5020Developed medium intensity14832%14632%91300
  MD1973Developed low intensity14321%13654%93286
  KILNAPER9Developed low intensity14653%14585%85304
  KILWOODR2Developed low intensity13985%13164%86284
  KILWOODR4Developed low intensity14653%14585%85304
  KILCHICA63Developed low intensity14800%70148
  KILDARIE3Developed low intensity135107%14343%76292
  KILGLENE1Developed low intensity13022%14000%70272
  KILNAPER19Developed medium intensity14575%14485%76304
  KILWHEAT7Developed low intensity14532%13953%71292
  MD0023Developed open space12822%14000%70270
  KILWESTC5Developed low intensity15032%51153
  KILNAPER21Developed low intensity14764%41153
  KILGLENE5Developed low intensity12565%33131
  KILWHEAT9Developed low intensity13696%33145
 Lake, IL
  KILGURNE6Developed open space14600%15200%140298
  KILLAKEB2Developed low intensity14500%13000%140275
  MC2377Developed low intensity14500%15200%140297
  KILDEERF2Deciduous forest15121%15300%120306
  KILLAKEB4Developed low intensity13721%14200%120281
  KILLIBER6Developed low intensity15021%15300%120305
  KILANTIO1Developed open space15121%15032%90306
  KILGURNE1Developed open space14800%14132%90292
  KILGURNE8Developed low intensity14821%14932%90302
  KILMUNDE4Open water14832%14832%81302
  KILGRAYS3Developed low intensity14253%14853%73300
  KILINGLE2Developed low intensity14853%14853%73306
  KILLAKEV6Open water15121%14575%72305
  KILLAKEV7Developed medium intensity14443%14643%71298
  KILWINTH3Developed low intensity13643%13543%71279
  KILLAKEV4Developed low intensity11932%60122
  KILLIBER2Developed open space14000%1163724%630293
  KILVERNO2Developed medium intensity14743%50151
  KILZION4Developed open space14332%50146
  KILLAKEZ3Developed low intensity14496%14585%47306
  KILTHIRD3Deciduous forest13996%44148
  KILBARRI8Developed low intensity13786%32145
  KILHIGHL6Developed low intensity865238%146138
 Lake, IN
  KINHOBAR5Developed medium intensity14853%14943%101306
  KINLOWEL5Cultivated crops14675%14664%94305
  KINEASTC3Developed high intensity12922%12654%81262
  KINHAMMO3Developed low intensity13564%14385%85292
  KILCRETE1Developed low intensity12422%60126
  KINMUNST1Developed low intensity1341510%143107%613302
  KINCEDAR4Developed low intensity13964%1302315%517298
  KINHAMMO2Developed low intensity1321711%142107%514301
  KINLOWEL4Cultivated crops10787%12115
  KILROMEO4Developed open space15200%14500%180297
  KILWILMI1Developed low intensity13100%11700%180248
  KILMOKEN1Developed low intensity14700%14800%160295
  KILFRANK1Developed low intensity14932%14743%121303
  KILBOLIN10Developed low intensity14721%13343%110286
  KILGLENW1Cultivated crops12397%12122%115255
  KILJOLIE4Developed low intensity14932%14743%111303
  KILMINOO3Developed low intensity14853%12065%114279
  KILPLAIN5Developed low Intensity15032%14764%112306
  KILROMEO5Developed medium intensity13554%14464%114290
  KILLOCKP6Developed low intensity15032%14575%102305
  KILPEOTO1Developed low intensity14853%14843%101305
  KILCHANN3Developed low intensity14553%12964%94285
  KILJOLIE12Developed medium intensity14553%14585%94303
  KILPLAIN6Developed low intensity14664%14764%93305
  KILROMEO7Developed medium intensity15100%90151
  KILBOLIN8Developed open space13764%12786%75278
  KILELWOO2Woody wetlands1341912%140138%620306
  KILJOLIE15Developed low intensity1381510%143107%613306
  KILLOCKP1Developed medium intensity1223020%1361510%633303
  KILMANHA2Developed medium intensity14943%50153
  KILNEWLE3Developed low intensity14943%50153
  KILNEWLE9Developed medium intensity14843%50152
  KILPLAIN15Developed low intensity1152216%14585%517290
  KILLOCKP9Developed low intensity13464%41140
  KILNEWLE8Developed medium intensity14764%41153
  KILPLAIN19Developed low intensity14764%41153
  MD9022Developed medium intensity13764%41143
  KILBOLIN11Developed low intensity1381510%39153
  KILBRAID6Developed medium intensity14585%32153
  KILPLAIN7Developed medium intensity14296%33151
  KILAUROR20Developed low intensity964532%137141

Stations within each county are sorted by number of false positives.
3.2. Comparing NWS and Station Data

To test the agreement of hazardous heat conditions at stations with the NWS Advisories, we treated the station measurements as the diagnostic gold standard against which the NWS notifications were compared. A false negative test, thus, describes an instance in which the daily heat index measured at the station met the criteria for a Heat Advisory without an official NWS-issued Advisory. In both metro areas the average number of false negatives and false positives differed significantly from perfect agreement (see Tables 1 and 2). This was true for both 2011 and 2012 and across different land covers for which there were more than 10 examples.

The average number of false positives per station per year in Chicago (4.80) was much higher than the average number of false negatives (1.96), which was slightly less than the number of true positive days where stations and NWS notifications agreed (2.92), while in Atlanta, the average number of false positives (1.98) was much lower than the number of false negatives per station (5.28), both of which were greater than the days/year of station-NWS agreement (1.87). The results suggest that the NWS-issued Heat Advisories in Chicago covered many instances when hazardous conditions were never reached at stations, and in Atlanta the Advisories were more likely to miss instances of hazardous conditions at specific locations. Some stations in each metro area exhibited a high degree of disagreement between stations and the NWS (see Figure 3); however, no pattern related to land cover or location emerged to explain these variations.

3.3. Comparing Satellite Imagery and Temperatures

Three specific categories of summer days were investigated more closely in each metro area using satellite data on LST. These categories were identified using the NWS Heat Advisory status and general trends among the metro’s weather stations. In Atlanta, for example, on July 1, 2012 NWS issued a Heat Advisory for all five counties in the metro area and nearly all of the stations (87.7%) met the conditions for the Advisory. An NWS Advisory was also issued on June 29, 2012; however only about half (47.4%) of stations met Advisory conditions. In the third case, on July 25, 2012 just under half (48.3%) of the stations exhibited Advisory conditions but no NWS Heat Advisory was issued. In Chicago July 6, 2012 was a NWS Heat Advisory with agreement from 93.8% of stations. July 17, 2012 was another NWS Heat Advisory, but with only 42.1% of stations in agreement. Finally, July 2, 2012 was an example of several stations (49.6%) exhibiting Advisory conditions without a NWS-issued Advisory.

We examined these six days in detail using MODIS imagery to produce estimated daily mean air temperatures. Our mixed effects model with the average night LST as the predictor and average daily air temperature as the dependent variable fits the dataset well and provides high prediction precision. For the model fittings, the is 0.71 and 0.74 in Chicago and Atlanta, respectively. For the fivefold CV, the and RMSPE are 0.697 and 1.92°F in Chicago, respectively; and and RMSPE are 0.711 and 1.75°F in Atlanta, respectively. We used this model to predict the mean daily temperature over Chicago and Atlanta on three typical days, respectively.

In both metro areas, the days on which there was high agreement between stations and NWS had higher estimated air temperatures (see Figures 4(a) and 4(b)). Even when there was daily disagreement among stations ((c)–(f)), the estimated temperatures on NWS Heat Advisory days ((c) and (d)) are higher than on days when there was no Advisory ((e) and (f)). The results support the merits of each dataset: the NWS Advisories appear to correspond to higher regional temperatures as estimated from LST, and station data help to identify specific locations of hazardous heat, particularly on days when no NWS Heat Advisory is issued.

4. Discussion

This work describes some fundamental characteristics of the NWS warning system for heat. The NWS system does well capturing regional weather movements which result in hazardously warm weather, and it delivers early information to large populations. This work highlights how the NWS lacks the spatial precision to inform more carefully targeted interventions which could save lives during regional heat events. The broad applicability of the NWS Heat Advisories and Warnings leads them to warn of the existence of hazardous heat conditions for places on summer days when such conditions are not present and to miss several instances of hazardous conditions in metro areas on days for which regional weather patterns may not trigger a warning. This work used relatively new and quickly growing datasets like unregulated and noncentralized weather stations as well as daily satellite imagery to demonstrate how the coverage and resolution of heat conditions within metro areas may be better represented and studied.

Moving forward, increased attention should be devoted to uncovering the drivers of non-NWS Advisory hot spots with urban areas. The existence and mechanisms of the urban heat island are well understood but largely considered only by research and science communities, rather than local planning entities [34]. The existence of datasets, like those used in this study, present the opportunity for cities to understand and identify where and how local topography, climate, and demographics interact to create elevated climate-related health risks. One example of such practice already being adopted is with Chicago’s Sustainability Office and their use of Landsat imagery to identify hotspots, which then serve as an inventory of priority locations for new green infrastructure investments ( [35].

The opportunities to improve the existing NWS early warning system are considerable, particularly given the growing number of networked weather stations within urban areas. The Wunderground network of stations is but one of several sites aggregating information from such distributed environmental monitoring systems. Weather Bug is another network which was used in the original study using MODIS data to generate a surface of estimated air temperatures [33]. The ability of imagery to provide increased coverage and resolution for urban temperatures continues to be explored and improved [36]. New processing techniques, such as those used here, which combine remotely sensed thermal data with in situ temperature readings offer the potential for cities to obtain recurring descriptions of intraurban temperature variation. Patterns that emerge from examining the urban thermal environmental over years could help improve public expenditures for reducing the urban heat island, for example, by increasing green space. These investments can be important parts of municipal strategies to combat climate change and improve quality of life through the multiple cobenefits that such strategies offer [37].

Warning systems for natural hazards have improved most for non-heat-related threats [38]. For example, in 2007 the NWS has developed storm-based warnings for tornados as an alternative to countywide warnings. New data sources can improve precision of heat modeling and forecasting in metropolitan areas and may help with the reception of warning messages. This is particularly true as more of the public obtain weather information through interactive displays and interfaces. Compared with message delivery via radio, or static maps, meteorological information is increasingly made available on location-smart devices and with animated visualizations of weather conditions describing recent historic and near future weather conditions with improved precision [39, 40]. As users become used to this form of weather information, more specific forecasts of local heat can improve individual response and public trust of warnings [41]. The location- and time-specific information available from such networks can also be paired with mobile technology to provide targeted warnings to subscribers within a certain distance of stations when they are exhibiting hazardous conditions. Mobile weather applications are already beginning to use location-aware devices’ abilities to improve information delivery and some, such as Weather Underground’s mobile app, allow users to provide “Hazard Reports” for their current location to crowdsource information on conditions such as power outages, road closures, and flooding.

In our analysis we found the estimated air temperatures produced from combining LST with station observations to capture the general patterns also described by the NWS warning classification, we do not find it to capture all of the heterogeneity in the weather station data. This may be due to the inability of LST-produced estimates to reflect the importance of humidity in determining human health risk, as is accomplished with a heat index-like factor. In all cases (the six days examined) there was no significant difference in estimated mean temperature between pixels with stations exhibiting Advisory-like conditions and for those with stations measuring heat index values below 105°F. This perhaps suggests the importance of investigating methods for estimating heat index from satellite imagery sources, for validating the station measurements more thoroughly, and for examining the importance of local drivers of humidity.

We acknowledge that this analysis relies on measurements from weather station data that have not been as closely quality controlled as those from other long-standing weather stations that are part of large national and global networks. We limited our analysis because we believed the 2011-2012 Wunderground weather station data to be more complete than earlier years we investigated. For Atlanta, the fraction of daily observations exhibiting hazardous conditions was similar between the 7-year (4.3%) and the 2-year (4.9%) datasets. For Chicago the fractions were more divergent, 2.1% for 2006–2012 and 3.3% for only 2011 and 2012; however, we noticed that the 2006–2010 contained less hazardous heat weather in all datasets (NWS and Wunderground). Though we collected data from a number of years that were not used in the analysis, we expect that our findings are not anomalous in their description of the agreement between NWS and weather station measurements.

5. Conclusion

Records from individual weather stations show that the conditions for hazardous heat exposures exist for many locations within counties, even when the forecasted conditions do not call for issued warnings applied to the county. Further analysis of larger, long term datasets is needed to produce a more robust understanding of the interaction of regional weather and local land cover conditions that combine to produce hazardous conditions at specific locations within counties. These factors likely change with time of year and metro area being considered.

Conflict of Interests

The authors declare that there is no conflict of interests regarding the publication of this paper.


The work of Liu and Xiao was partially supported by NASA Applied Sciences Program (Grants NNX11AI53G and NNX14AG01G, PI: Liu). The authors would also like to thank government climatologists and scientists including Brad Pierce, Ed Hopkins, Barry Gooden, Jim Allsopp, and Rusty Kapela for their conversations as the authors developed this research.


  1. J. Hansen, M. Sato, and R. Ruedy, “Perception of climate change,” Proceedings of the National Academy of Sciences of the United States of America, vol. 109, no. 37, pp. E2415–E2423, 2012. View at: Publisher Site | Google Scholar
  2. G. A. Meehl and C. Tebaldi, “More intense, more frequent, and longer lasting heat waves in the 21st century,” Science, vol. 305, no. 5686, pp. 994–997, 2004. View at: Publisher Site | Google Scholar
  3. D. R. Easterling, G. A. Meehl, C. Parmesan, S. A. Changnon, T. R. Karl, and L. O. Mearns, “Climate extremes: observations, modeling, and impacts,” Science, vol. 289, no. 5487, pp. 2068–2074, 2000. View at: Publisher Site | Google Scholar
  4. J.-M. Robine, S. L. K. Cheung, S. Le Roy et al., “Death toll exceeded 70,000 in Europe during the summer of 2003,” Comptes Rendus: Biologies, vol. 331, no. 2, pp. 171–178, 2008. View at: Publisher Site | Google Scholar
  5. D. Guha-Sapir, F. Vos, R. Below, and S. Ponserre, Annual Disaster Statistical Review 2010, Centre for Research on the Epidemiology of Disasters, 2011.
  6. B. A. Revich, “Heat-wave, air quality and mortality in european russia in summer 2010: preliminary assessment,” Human Ecology, no. 7, pp. 3–9, 2011. View at: Google Scholar
  7. D. Habeeb, J. Vargo, and B. Stone, “Rising heat wave trends in large US cities,” Natural Hazards, vol. 76, no. 3, pp. 1651–1665, 2015. View at: Publisher Site | Google Scholar
  8. R. S. Kovats and S. Hajat, “Heat stress and public health: a critical review,” Annual Review of Public Health, vol. 29, pp. 41–55, 2008. View at: Publisher Site | Google Scholar
  9. J. A. Patz, D. Campbell-Lendrum, T. Holloway, and J. A. Foley, “Impact of regional climate change on human health,” Nature, vol. 438, no. 7066, pp. 310–317, 2005. View at: Publisher Site | Google Scholar
  10. G. Luber and M. McGeehin, “Climate change and extreme heat events,” American Journal of Preventive Medicine, vol. 35, no. 5, pp. 429–435, 2008. View at: Publisher Site | Google Scholar
  11. United States Environmental Protection Agency—Office of Atmospheric Programs (6207J), “Excessive heat events guidebook,” Tech. Rep. 430-b-06-005, 2006. View at: Google Scholar
  12. M. S. O'Neill, D. K. Jackman, M. Wyman et al., “US local action on heat and health: are we prepared for climate change?” International Journal of Public Health, vol. 55, no. 2, pp. 105–112, 2010. View at: Publisher Site | Google Scholar
  13. M. Medina-Ramón and J. Schwartz, “Temperature, temperature extremes, and mortality: a study of acclimatisation and effect modification in 50 US cities,” Occupational and Environmental Medicine, vol. 64, no. 12, pp. 827–833, 2007. View at: Publisher Site | Google Scholar
  14. R. Basu and J. M. Samet, “Relation between elevated ambient temperature and mortality: a review of the epidemiologic evidence,” Epidemiologic Reviews, vol. 24, no. 2, pp. 190–202, 2002. View at: Publisher Site | Google Scholar
  15. F. C. Curriero, K. S. Heiner, J. M. Samet, S. L. Zeger, L. Strug, and J. A. Patz, “Temperature and mortality in 11 cities of the eastern United States,” American Journal of Epidemiology, vol. 155, no. 1, pp. 80–87, 2002. View at: Publisher Site | Google Scholar
  16. K. L. Ebi, T. J. Teisberg, L. S. Kalkstein, L. Robinson, and R. F. Weiher, “Heat watch/warning systems save lives: estimated costs and benefits for Philadelphia 1995–98,” Bulletin of the American Meteorological Society, vol. 85, no. 8, pp. 1067–1073, 2004. View at: Publisher Site | Google Scholar
  17. M. G. Weisskopf, H. A. Anderson, S. Foldy et al., “Heat wave morbidity and mortality, Milwaukee, Wis, 1999 vs 1995: an improved response?” American Journal of Public Health, vol. 92, no. 5, pp. 830–833, 2002. View at: Publisher Site | Google Scholar
  18. M. A. Palecki, S. A. Changnon, and K. E. Kunkel, “The nature and impacts of the July 1999 heat wave in the midwestern United States: learning from the lessons of 1995,” Bulletin of the American Meteorological Society, vol. 82, no. 7, pp. 1353–1367, 2001. View at: Google Scholar
  19. C. R. Browning, D. Wallace, S. L. Feinberg, and K. A. Cagney, “Neighborhood social processes, physical conditions, and disaster-related mortality: the case of the 1995 Chicago heat wave,” American Sociological Review, vol. 71, no. 4, pp. 661–678, 2006. View at: Publisher Site | Google Scholar
  20. M. P. Naughton, A. Henderson, M. C. Mirabelli et al., “Heat-related mortality during a 1999 heat wave in Chicago,” The American Journal of Preventive Medicine, vol. 22, no. 4, pp. 221–227, 2002. View at: Publisher Site | Google Scholar
  21. O. V. Wilhelmi and M. H. Hayden, “Connecting people and place: a new framework for reducing urban vulnerability to extreme heat,” Environmental Research Letters, vol. 5, no. 1, Article ID 014021, 2010. View at: Publisher Site | Google Scholar
  22. B. Li, S. Sain, L. O. Mearns et al., “The impact of extreme heat on morbidity in Milwaukee, Wisconsin,” Climatic Change, vol. 110, no. 3-4, pp. 959–976, 2012. View at: Publisher Site | Google Scholar
  23. R. E. Davis, P. C. Knappenberger, P. J. Michaels, and W. M. Novicoff, “Changing heat-related mortality in the United States,” Environmental Health Perspectives, vol. 111, no. 14, pp. 1712–1718, 2003. View at: Publisher Site | Google Scholar
  24. F. Salamanca, M. Georgescu, A. Mahalov, M. Moustaoui, and M. Wang, “Anthropogenic heating of the urban environment due to air conditioning,” Journal of Geophysical Research: Atmospheres, vol. 119, no. 10, pp. 5949–5965, 2014. View at: Publisher Site | Google Scholar
  25. S. C. Sheridan, “A survey of public perception and response to heat warnings across four North American cities: an evaluation of municipal effectiveness,” International Journal of Biometeorology, vol. 52, no. 1, pp. 3–15, 2007. View at: Publisher Site | Google Scholar
  26. Noaa—national weather service—weather prediction center—heat index probability forecasts, 2014,
  27. L. S. Kalkstein, P. F. Jamason, J. S. Greene, J. Libby, and L. Robinson, “The Philadelphia hot weather-health watch/warning system: development and application, summer 1995,” Bulletin of the American Meteorological Society, vol. 77, no. 7, pp. 1519–1528, 1996. View at: Publisher Site | Google Scholar
  28. L. S. Kalkstein and R. E. Davis, “Weather and human mortality: an evaluation of demographic and interregional responses in the United States,” Annals of the Association of American Geographers, vol. 79, no. 1, pp. 44–64, 1989. View at: Publisher Site | Google Scholar
  29. D. J. Gaffen and R. J. Ross, “Increased summertime heat stress in the US,” Nature, vol. 396, no. 6711, pp. 529–530, 1998. View at: Publisher Site | Google Scholar
  30. A. J. Kalkstein and S. C. Sheridan, “The social impacts of the heat-health watch/warning system in Phoenix, Arizona: assessing the perceived risk and response of the public,” International Journal of Biometeorology, vol. 52, no. 1, pp. 43–55, 2007. View at: Publisher Site | Google Scholar
  31. C. Huang, X. Li, and L. Lu, “Retrieving soil temperature profile by assimilating MODIS LST products with ensemble Kalman filter,” Remote Sensing of Environment, vol. 112, no. 4, pp. 1320–1336, 2008. View at: Publisher Site | Google Scholar
  32. Z. Wan and Z.-L. Li, “Radiance-based validation of the V5 MODIS land-surface temperature product,” International Journal of Remote Sensing, vol. 29, no. 17-18, pp. 5373–5395, 2008. View at: Publisher Site | Google Scholar
  33. I. Kloog, A. Chudnovsky, P. Koutrakis, and J. Schwartz, “Temporal and spatial assessments of minimum air temperature using satellite surface temperature measurements in Massachusetts, USA,” Science of the Total Environment, vol. 432, pp. 85–92, 2012. View at: Publisher Site | Google Scholar
  34. B. Stone, J. Vargo, and D. Habeeb, “Managing climate change in cities: will climate action plans work?” Landscape and Urban Planning, vol. 107, no. 3, pp. 263–271, 2012. View at: Publisher Site | Google Scholar
  35. S. I. Corporation, “Chicago Uses Satellite Images to Determine City's True Hot Spots,” Satellite Image News, 2006, View at: Google Scholar
  36. J. A. Voogt and T. R. Oke, “Thermal remote sensing of urban climates,” Remote Sensing of Environment, vol. 86, no. 3, pp. 370–384, 2003. View at: Publisher Site | Google Scholar
  37. J. Patz, D. Campbell-Lendrum, H. Gibbs, and R. Woodruff, “Health impact assessment of global climate change: expanding on comparative risk assessment approaches for policy making,” Annual Review of Public Health, vol. 29, pp. 27–39, 2008. View at: Publisher Site | Google Scholar
  38. J. H. Sorensen, “Hazard warning systems: review of 20 years of progress,” Natural Hazards Review, vol. 1, no. 2, pp. 119–125, 2000. View at: Publisher Site | Google Scholar
  39. National Rearch Council, Public Response to Alerts and Warnings on Mobile Devices: Summary of a Workshop on Current Knowledge and Research Gaps, National Rearch Council, 2011.
  40. M. A. Casteel and J. R. Downing, “How individuals process NWS weather warning messages on their cell phones,” Weather, Climate, and Society, vol. 5, no. 3, pp. 254–265, 2013. View at: Publisher Site | Google Scholar
  41. S. Savelli and S. Joslyn, “The advantages of predictive interval forecasts for non-expert users and the impact of visualizations,” Applied Cognitive Psychology, vol. 27, no. 4, pp. 527–541, 2013. View at: Publisher Site | Google Scholar

Copyright © 2015 Jason Vargo et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

More related articles

 PDF Download Citation Citation
 Download other formatsMore
 Order printed copiesOrder

Related articles