Air temperature and thermal comfort data measured and modelled for 121 workplaces in the Upper Rhine Valley

The uploaded files contain the measured and modelled indoor data at different workplaces in the Upper Rhine Valley between August 1, 2021, and July 31, 2022 presented in the article "Predicting Indoor Air Temperature and Thermal Comfort in Occupational Settings Using Weather Forecasts, Indoor Sensors, and Artificial Neural Networks" by Sulzer et al. (2023), doi.org/10.1016/j.buildenv.2023.110077. Information about the characteristics of each workplace can be found in the appendix of the article. For every workplace two files are uploaded, one for the data of the indoor air temperature (Ta) and one for indoor physiological equivalent temperature (PET). The workplace ID and variable are stated in the filename. The columns in the files contain the following data:

• "Datetime (UTC)": This column contains the timestamp in UTC of the starting point of the interval used for the one-hour mean values .
• "MoBiMet data": This column contains the one-hour mean values of Ta or PET of the data derived every five minutes by the low-cost Mobile Biometeorology System (MoBiMet) at the corresponding workplace in °C. The MoBiMet are presented in detail in doi.org/10.3390/s22051828.
• "Used for": This column contains the information if the data point was used for training of the models (t), evaluation of the models (e), or not used for model training or evaluation due to missing indoor observation data (n).
• "Product 0 (ICON_outdoor)": This column, in the files for the indoor air temperature, contains the ICON-D2 data of the air temperature 2m a.g.l. in °C of the grid cell in which the associated work station is located.
• "Product 0 (ICON_outdoor) air temperatur 2m (C) input for PET calculation using RayMan","Product 1 (ICON_outdoor) vapor pressure 2m (hPa) input for PET calculation using RayMan","Product 1 (ICON_outdoor) wind speed 10m (m/s) input for PET calculation using RayMan", and "Product 1 (ICON_outdoor) global radiation surface (W/m²) input for PET calculation using RayMan": This columns contain the data of the outdoor air temperature 2m a.g.l. in °C, the vapor pressure 2m a.g.l., derived from the ICON-D2 weather forecast data of the grid cell in which the associated work station is located, which were used in RayMan Pro to calculate the PET for outdoors.
• "Product 2 (ANN_Generic)": This column contains the indoor data of PET or air temperature in °C modelled by an artificial neural network using generic data as input. The generic data contain hourly solar altitude and azimuth at each location, the weekday, and a sine and cosine function of the daily and yearly cycle.
• "Product 3 (ANN_AWS) without past data": This column contains the indoor data of PET or air temperature in °C modelled by an artificial neural network using generic data and the meteorological data of air temperature, vapor pressure, mean sea level pressure, global radiation, longwave downwelling radiation, and wind speed of an automated weather station in Freiburg (Station FRCHEM; 48°00’04’’ N; 7°50’55’’ E).
• "Product 3 (ANN_AWS) with past data": This column contains similar data than the column before but the artificial neural network models used "past data" of the automated weather station as additional input variables to model indoor air temperature and PET in °C. Additional to the hourly average of the meteorological data for each actual time (t), hourly averages for t-1 h, t-3 h, t-6 h, t-12 h, and t-24 h of air temperature, global radiation, and Longwave downwelling radiation are used as so called "past data".
• "Product 4 (ANN_ICON) without past data": This column contains the indoor data of PET or air temperature in °C modelled by an artificial neural network using generic data and the meteorological data of air temperature, vapor pressure, mean sea level pressure, global radiation, longwave downwelling radiation, and wind speed derived from the ICON-D2 weather forecast data of the grid cell in which the associated work station is located.
• "Product 4 (ANN_ICON) with past data": This column contains similar data than the column before but the artificial neural network models used "past data" of the ICON-D2 weather forecast data as additional input variables to model indoor air temperature and PET in °C. Additional to the hourly average of the meteorological data for each actual time (t), hourly averages for t-1 h, t-3 h, t-6 h, t-12 h, and t-24 h of air temperature, global radiation, and Longwave downwelling radiation are used as so called "past data".
• "Product 5 (ANN_Mixed) without past data": This column contains the indoor data of PET or air temperature in °C modelled by the same artificial neural network models as in Product 3 but applied for the same input data of Product 4.
• "Product 5 (ANN_Mixed) with past data": This column contains similar data than the column before but also takes into account the "past data".

The data of Product 3 (ANN_AWS) and Product 5 (ANN_Mixed) are only available for locations in Freiburg, because the data of an automated weather station in Freiburg was used.
Data which is not available is stated as NA.