This readme file was generated on [2024-5-03] by [ERIK PATTON]


-------------------
GENERAL INFORMATION
-------------------

Title of Dataset:
Data sets described in the article Wet bulb globe temperature from climate model outputs: a method for projecting hourly site-specific values and trends.

Author Contact Information

	Principal Investigator: Erik Patton
	Institution: Duke University
	Email: erik.patton@duke.edu	
	ORCID: 0000-0002-0628-7706 

*Date range of data collected or projected: 19920101-21001231

*Date range of individual files (files are further described below)
      1. Historical record files (XXXX_last30_wbgt_from_obs): 19920101-20211231. This file is described as file C below.
      2. Historical climate mean files (XXXX_30.year_climate_mean): 20220101-20221231. Note this date range/use of year 2022 is a placeholder; the file represents the climate average values derived from 30 years of historical records. Year 2022 is assigned only for convenience. These files are described as file B below.
      3. All_Install_Dataset: 20071001-21001231. Note this dataset can be considered in three distinct time periods. 20071001-20141231 values are derived from GCM historical runs; 20150101-20211231 values are derived from GCM scenario runs over a time period overlapping with the historical record data set; 20250101-21001231 values are derived from future GCM scenario runs. This file is described as file A below.

*Geographic location of data collection: 
Ft Moore, GA
Ft Jackson, SC
Ft Sill, OK
Ft Leonard Wood, MO

--------------------
STUDY OVERVIEW
--------------------

The datasets provided here support an article titled "Wet bulb globe temperature from climate model outputs: a method for projecting hourly site-specific values and trends". The article has been submitted to International Journal of Biometeorology.

ARTICLE ABSTRACT:

Increasing temperature will impact future outdoor worker safety but quantifying this impact to develop local adaptations is challenging. Wet bulb globe temperature (WBGT) is the preferred thermal index for regulating outdoor activities in occupational health, athletic, and military settings, but global circulation models (GCMs) have coarse spatiotemporal resolution and do not always provide outputs required to project the full diurnal range of WBGT. This article presents a novel method to project WBGT at local spatial and hourly temporal resolutions without many assumptions inherent in previous research. We calculate sub-daily future WBGT from GCM output and then estimate hourly WBGT based on a site-specific, historical diurnal cycles. We test this method against observations at U.S. Army installations and find results match closely. We then project hourly WBGT at these locations from January 1, 2025, to December 31, 2100, to quantify trends and estimate future periods exceeding outdoor activity modification thresholds.  We find regional patterns affecting WBGT, suggesting accurately projecting WBGT demands a localized approach. Results show increased frequency of hours at high WBGT and, using U.S. military heat thresholds, we estimate impacts to future outdoor labor. By mid-century, some locations are projected to experience  an average of 20 or more days each summer when outdoor labor will be significantly impacted. The methodÕs fine spatiotemporal resolution enables detailed analysis of WBGT projections, making it useful applied at specific locations of interest.

--------------------
DATA & FILE OVERVIEW
--------------------

Files used in this study were read using RStudio version 2023.06.2+561 using the read.csv command. 

File list (A B C):

A. All_Install_Dataset. This is the primary data set. This data set is the final outcome of the methods described in the associated article. Specifically, this data set contains WBGT values estimated between 20071001 - 21001231 for six global circulation models for climate scenario SSP2 and SSP3, or (for dates prior to 20150101) the GCM historical runs for all four locations. It can be thought of as the output for the novel methodology described in the article. 

The dataset itself is somewhat unwieldy; below we attempt to clarify the variables and how they are used in the study. As the data set is large and intended to be read into the statistical programming language R (see CODE AVAILABILITY below), as such it it is in long format. There are 9 columns in this data set.

1. The column titled X is an artifact of the code use in RStudio and simply a running count of row number. It adds nothing of value.
2. datetime - is the day, month, year, and hour associated with each WBGT value.
3. ssp - There are five variables in this column. 
      3.a: WBGT from observations refers to WBGT values provided by the USAF. Date range for this variable is 20071001-20211231.
      3.b-c: ssp245.hist and ssp370.hist refer to WBGT values derived from GCMs using SSP 2-4.5 and SSP 3-7.0, respectively, which can be compared to historical observations. Date range for this variable is 20150101-20211231. This overlapping period is used to create quantile delta mapping transformation functions.
      3d-e: ssp245 and ssp370 refer to WBGT values derived from GCMs using SSP 2-4.5 and SSP 3-7.0, respectively, for future hours between 20250101-21001231,
4. model - refers to the GCM from which WBGT values are derived, except for ÔobservationsÕ which refers to values from historical observations (20071001-20211231 period only).
5. WBGTC - is the wet bulb globe temperature in degrees Celsius. 
6. year
7. date
8. Heat.Cat - refers to the heat category associated with the WBGT value. These categories are obtained from the Department of Defense Technical Bulletin (medical) 507.
9. installation - refers to the location where the WBGT value is observed or estimated. The location breakdown is:
	9a. FJSC is Ft Jackson, SC
	9b. FMGA is Ft Moore, GA
	9c. FSOK is Ft Sill, OK
	9d. FLW is Ft Leonard Wood, MO  

WBGT values titled "WBGT from observations" are provided from the 14th Weather Squadron, U.S. Air Force (USAF). Model and ssp refer to the NASA NEX-GDDP data sets. For example, the WBGTC associated with model=CANESM5 and ssp=ssp245 is derived from GCM outputs obtained from the NASA NEX-GDDP data set for GCM CANESM5 in scenario SSP 2-4.5. Describing how this WBGTC is calculated is the main theme of the article associated with this data set, and we refer the user to it.


B. Four supporting data sets are included providing the climate mean for each location. These data sets are named "XXX_30.year_climate_mean". The "XXX" is replaced in each file with the code for the location the data is from (e.g., XXXX can be replaced with FJSC). This data set describes the climate average WBGT for each hour of the year used to create the future data sets. Since the average hour of a year is described, this file is 8,760 rows long; there are 8,760 hours in a (non-leap) year. 

There are 12 columns in this dataset. Only two are of significance. 
1. WBGTC is the climatological average value derived from "observations" provided by the USAF 14th Weather Squadron. These values are used to form the 365 daily diurnal curves. As described in the article, these curves are adjusted according to GCM outputs to estimate future hourly WBGT between 2025-2100. 
2. WBGT.calc is this is the value derived from the (unadjusted) Liljegren method using meteorological values from each location. The WBGT.calc can be compared to the WBGTC to estimate how closely the Liljegren method (Liljegren 2008) matches the USAF value for the same location and hour. Although matches are very good (r^2 >0.98), some additional steps are taken when estimating WBGT in future periods (e.g., setting a floor to minimum wind speed). These steps are described in detail in the article. 

The remainder of the columns in this data set are self explanatory except for lower.ci.WBGTC and upper.ci.WBGTC. These are the upper and lower confidence intervals for each WBGT value created from the 30-year observational period (described as data set C, below). 

C. Four supporting data sets are included providing the raw data used to create files B (described above). These data sets are named "XXX_last30_wbgt_from_obs". The "XXX" is replaced in each file with the code for the location the data is from (e.g., XXXX can be replaced with FJSC). Data contained in these files are the meteorological observations over a 30-year period at each location, except WBGTC (which is the WBGT values provided by the USAF) and WBGT.calc (which is the WBGT derived from the Liljegren method), both calculated using metrological observations in this file as inputs. There are 22 columns in each data set. For the most part they are self-explanatory but some are highlighted here:

1. HR is the hour of the day (0-23)
2. WSPDKT is the windspeed, in knots, at elevation of 10 meters.
3. RELHUM is the relative humidity
4. STAPRS is the surface pressure
5. TOTRAD is the total radiation recorded at the weather station (taken to be downwelling shortwave radiation, although observations may include a longwave component).
6. WBGTC is the wet bulb globe temperature, in Celcius, provided by the USAF.
7. WBGT.calc is the wet bulb globe temperature calculated per the Liljegren method (Liljegren 2008) using these meteorological observations.

Data set C citation: 14th Weather Squadron, U.S. Air Force. Ashville, NC. https://www.557weatherwing.af.mil/Units/2d-Weather-Group/14th-Weather-Squadron/ 

----------

Data sets A and B in this repository were created in this study. The "raw" data is provided as data set C but can also be obtained from meteorological records at the weather stations described in the article supplemental or using other methods to obtain historical weather data (e.g., ERA5 reanalysis data).

We do not provide the data downloaded from the 19x NASA NEX-GDDP global circulation models. These are publicly available at: https://www.nccs.nasa.gov/services/data-collections/land-based-products/nex-gddp-cmip6. 

--------------------
CODE AVAILABILITY
--------------------

Statistical analysis for this study was performed using RStudio. RMarkdown files are included in the data repository. Two RMarkdown files are provided:

1. ERA5 data historical_data repository version: The example RMarkdown file uses Durham, NC, as an example location and ERA 5 data (not historical observations) to generate the 30-year climate mean (i.e., file B). This file is provided because it is likely more useful for a researcher than the original code using observational data provided by the USAF; ERA 5 data can be obtained at 0.1x0.1 lat/lon resolution anywhere over the EarthÕs surface. This files replaces the need for data set C, described above, so the user does not have to obtain an observational data set from the USAF or from an individual weather station. It generates data set ÔBÕ, the historical baseline for later adjustment to create future WBGTs.

2. FJSC_Create WBGT Dataset ssp370_data repository version: This is the file used to create the SSP 3-7.0 data set at the FJSC location. It is a messy file and could likely be cleaned up by a user with more efficient coding skills, but will create s section of the data set A, providing WBGT values for every hour between 2025-2100 (in this case, it creates WBGT for the FJSC location for scenario SSP 3-7.0). This is a modified version of the original file used in the study with extensive commenting in the first GCM chunk MPI_ESM1_2_HR to describe what each section is doing. The other 18 GCMs are not commented since they follow the same procedure and differ only in GCM name.

The file has three general subsections:
2.A. Global Variables - set the variables common for all climate models. It is the initial section of code. For example, date range (in years) and ssp scenario are set here.

2.B. Station Pressure Linear Models Creation - establishes the station pressure linear models which will be common for all GCMs. It also creates or establishes any adjustment or correction factors needed. For example, all minimum WBGT (i.e., night time) solar radiation values are set to Ô0Õ in this chunk.

2.C. MPI_ESM1_2_HR (or CANESM5 or ACCESS_ESM1_5 or any other climate model name) creates projected WBGT values using the climate model output corresponding to the name. 19 GCMs were used in the study, so there are 19 subsections.

Additionally, a small chunk of code at the end combines all GCM output into a single data frame for the location and ssp selected. 

To use this file at other locations or for other climate scenarios, the FJSC and ssp370 should be replaced with appropriate location name and climate scenario, respectively. The user will have to rename the file paths to load the data; any pathway is retained as used by the author.

--------------------
ADDITIONAL INFORMATION
--------------------

For Additional Information: Please contact Erik Patton at erik.patton@duke.edu or erik.m.patton.mil@army.mil

Data License: CC0 Waiver

Other Rights Information: N/A

To cite the data: Please cite the associated article, Patton et al. Wet bulb globe temperature from climate model outputs: a method for projecting hourly site-specific values and trends.