Near- and medium-term hourly morphed mean and extreme future temperature datasets for Jyväskylä, Finland, for building thermal energy demand simulations

Near- and medium-term hourly morphed outdoor temperature files were created for Jyväskylä, Finland, to be used in building energy simulation. These future outdoor temperature files were created according to a statistical down-scaling method, morphing, which utilizes both hourly baseline data, and monthly and daily future climate projections. The used baseline data included hourly test reference year and typical meteorological year data to represent a “typical” climate year, and were appended with weather files created based on the coldest and warmest near Januaries to represent extreme weather files. Climate change data included climate change projection data from 2 different data repositories for either results from Global Climate Model or Regional Climate Model simulations for RCP2.6, RCP4.5 and RCP8.5 climate change scenarios. Morphing all 5 different baseline scenarios with each of the available climate scenarios creates 25 future outdoor temperature files for 2030 (near-term) and 25 files for 2050 (medium-term). The created files were used in [1] to simulate future thermal energy demand in buildings.


Data accessibility
Data is available at the following data repository:

Value of the Data
• Assessment of future thermal energy demand for buildings requires hourly outdoor temperature data to be used on building energy simulations. • This data can benefit anyone interested in simulating future thermal demand of buildings on Jyväskylä, Finland or Nordic area. • This data can be directly used as an input data to building energy simulations, which require future hourly outdoor temperature files. • This data can help in assessing future energy demands in local, regional or country level on typical and extreme weathers through utilization in building energy simulation. • Additionally, the data can be used to assess any future goals that require energy demand or other outputs from building energy simulations for mean and extreme weather years. • Inclusion of extreme weather years allows studying the system level design for energy systems dependable on them, e.g. on district heating network design.

Data for current climate
There are 3 datasets containing averaged outdoor temperature ( T ) data to represent averaged weather conditions in different time periods for the Jyväskylä, Finland area. These datasets are presented in Table 1 showing the averaged time period and the source of original data. Additionally, to introduce the extreme heating season weather conditions, a representative cold and warm January weather data are included and presented in Table 1 .
All the data represented in Table 1 [3] . To improve the validity of future simulations, representative cold (Cold2010) and warm (Warm2008) Januaries were added to the dataset to present extreme weather conditions. The selection of these weather data was based on the lowest and highest average temperatures in January in Jyväskylä Airport (Jyväskylä AP) observation station from Finnish Meteorological Institute [4] . The characteristics of these weather datasets are presented in Table 2 including monthly average temperatures, mean differences between daily maximum and minimum temperatures on every month and the standard deviation of daily mean temperatures in each month.
The data from Table 2 shows similarities in average outdoor temperatures on TMYx and TRY2012 weather files, but some differences on the temperature variation. The Cold2010 weather file clearly has the lowest average temperature during the heating season, while the Warm2008 file has the highest. The average heating season (October-April) temperatures of each of the weather file are presented in Table 3 .

Data for climate change
Applying data from climate change projections, 2 main aspects to acknowledge are the used Climate models, which are either Global Climate Models (GCMs) or Regional Climate Models (RCMs), and the climate change scenarios describing the future Greenhouse gas (GHG) emissions levels, both which combined will give the projected future weather variables. The current climate change data projections are conducted under Representative Concentration Pathway (RCP) scenarios, presenting different GHG emission pathways for the future based on their ending radiate forcing value for 2100. The scenarios currently include RCP2.6, RCP4.5, RCP6.0 and RCP8.5 scenarios, presented from lowest to highest radiate forcing value for 2100 [5] . The second main aspect was the used climate model, which depend on the availability and accuracy of data as well as the climate scenario. For Jyväskylä, Finland, 2 data sources were selected: -Results from Global Climate Models (GCMs) conducted originally under Coupled Model Intercomparison Project 5 (CMIP5) [6] , from which results for Finland are gathered and presented in [7] . Modified results for 10 x 10 km spacial resolution based on a baseline from [8] are available for RCP2.6, RCP4.5 and RCP8.5 scenarios on monthly mean outdoor temperature values in [9] , daily mean outdoor temperature values in [10] , daily maximum outdoor temperature value in [11] and daily minimum outdoor temperature value in [12] . -Regional Climate Models (RCMs) are dynamically down-scaled data from GCMs, presenting simulation results on finer spacial and time resolutions [13] . European Climatic Energy Mixes (ECEM) demonstrator from Copernicus Climate Change Service (C3S) 1 provide ensemble and individual results from RCMs on country and cluster level. The data for outdoor temperature projections are available on daily, monthly, seasonal and yearly level on RCP4.5 and RCP8.5 scenarios.
A summary of the available climate change projection data is presented on Table 4 . In Table 4 T daily,max and T daily,min represent the availability of daily maximum and minimum temperature data, respectively, T daily,mean represents availability of daily mean temperature data, and T monthly,mean represents the availability of monthly average temperature data.
The climate change projection data from both GCMs and RCMs for Jyväskylä is presented in Table 5 for 2030 and in Table 6 for 2050. The GCM results are ensemble on country level, and RCM results are ensemble for South-Finland cluster. The average monthly temperature T mean is a 30-year average value from monthly mean datasets, whereas the mean daily max min change  Table 5 Projected mean monthly temperatures T mean , changes between daily max and min temperatures T max,min and standard deviation of monthly daily mean temperatures σ T in RCP2.6, RCP4.5 and RCP8.5 climate scenarios from RCM (ECEM) and GCM results [9][10][11][12] in 2030.  T max,min and daily mean temperature standard deviations σ T are as well averaged over 30-year period, but use daily datasets.

Morphed outdoor temperatures
The future outdoor temperature is created with morphing method [14] from the baseline weather data and the climate change projections as described in Section 2.1 . This results in   Figs. 2 and 3 respectively. Furthermore, the average statistical results for the mean monthly temperatures and associated standard deviations are provide for each weather scenario (S1-S5, as listed in Table 1 ),

Table 7
Morphed mean monthly temperatures T mean , and standard deviation of monthly daily mean temperatures σ T in RCP2.6 GCM (Paituli) in 2050. Complete hourly dataset available in [15] . RCPs and climatic projection RCM and GCM in Tables 7-11 . A comparison with the reference  tables on climate change Tables 5 and 6 is presented in Fig. 1 . The morphed outdoor temperature files are stored in Zenodo Repository [15] under Creative Commons Licence 4.0 (CC4.0). These morphed outdoor temperatures were used in [1] to simulate future thermal energy demand in buildings. Table 8 Morphed mean monthly temperatures T mean , and standard deviation of monthly daily mean temperatures σ T in RCP4.5 GCM (Paituli) in 2050. Complete hourly dataset available in [15] .  Table 9 Morphed mean monthly temperatures T mean , and standard deviation of monthly daily mean temperatures σ T in RCP8.5 GCM (Paituli) in 2050. Complete hourly dataset available in [15] .

Methodology: morphing
The current weather data is transformed to represent a future climate with the help of a statistical down-scaling method called morphing [14] . The simulated future daily and monthly outdoor temperatures from GCMs and RCMs are statistically down-scaled to hourly level either with shifting, stretching or a combination of shifting and stretching by using hourly data from existing climate. To include both the changes in monthly average temperature and the daily temperature variation, the combination of shifting and stretching method for morphing the outdoor temperature is used: (1) where T is the morphed temperature [ °C], T 0 is the hourly baseline temperature [ °C], T m is the change in monthly average temperature [ °C], α m is the fractional change in monthly temperature Table 10 Morphed mean monthly temperatures T mean , and standard deviation of monthly daily mean temperatures σ T in RCP4.5 RCM (ECEM) in 2050. Complete hourly dataset available in [15] .

Table 11
Morphed mean monthly temperatures T mean , and standard deviation of monthly daily mean temperatures σ T in RCP8.5 RCM (ECEM) in 2050. Complete hourly dataset available in [15] . [-] and < T 0 > m is the monthly average temperature in baseline scenario [ °C] [14] . An averaging period of 30 years is used in all T m and α m calculations to reduce the impact of a yearly variations in the used datasets.
The fractional change α m is used to asses the change in daily temperature variation, and it can be calculated in 2 ways depending on the available data. The first method is to utilize the changes in daily maximum and minimum temperatures if this data is available:  [14] .
The second method to calculate α m follows method M2 from [16] . Here it is assumed that the daily temperature variation is relative to the daily mean temperature variation on a certain Fig. 4. The procedure for creating future outdoor temperature data, showing the input data sources, the methodology, scenarios and the created data (modified from [1] ). month in case no data on daily maximum and minimum temperatures is available: where σ T,new is the new daily variance on month m [-] and σ T, 0 is the daily mean temperature variance on the baseline scenario on month m [-] [17] . The whole procedure of creating the hourly future outdoor temperature data is presented in Fig. 4 . It shows the 2 sources of input data, from which the hourly data is used as the baseline data, and the climate change data is used to calculate the change in the monthly average temperature T m from monthly data and the fractional change α m from the daily data depending on the type of climate change data that is available. The morphing procedure is then used according to Equation 1 -Equation 3 . This results in the morphed future outdoor temperatures on hourly-scale, which can represent either future extreme or mean weather scenario depending on the used baseline data.
The morphing was conducted with a weather morphing application that was created in MAT-LAB environment to calculate the described variables from existing weather data files and from climate change projection results from the GCM and RCM simulations. These variables are then used to create new morphed outdoor temperatures based on the desired baseline scenario, morphed year and morphing period. The created weather morphing application allows utilizing both daily and monthly change data and supports both calculation methods for the calcula-