*2.3. Future Climate Scenarios*

For future climate scenarios, the meteorological time series from the rural station Rotterdam during the period 2002–2016 were transformed into the projected climate of 2050. The Netherlands weather services (KNMI) developed four different scenarios to depict how the climate may evolve, namely the GL, GH, WL, and WH scenarios. The scenarios differ in the estimated global temperature rise (G or W) and the degree of change in air circulation patterns (L and H suffix), all of which are schematically represented in Figure 2.

**Figure 2.** Four scenarios were used to calculate future heat stress. They differ in global temperature rise and change in air circulation pattern. Reproduced with permission from KNMI, Brochure KNMI'14 climate scenarios, 2015 [34].

These climate scenarios are based on the Intergovernmental Panel on Climate Change (IPCC) global climate model calculations, on global climate model EC-Earth results, and on a downscaling step within the regional climate model, Racmo [34]. All four scenarios have an equal chance of occurrence. As indicated in Figure 2, the scenarios differ in the degree of change in air circulation patterns and the global temperature rise. The differences in global temperature rise are due to different projections of greenhouse-gas emissions adopted from the emission scenarios from the IPCC [34]. The scenarios GL and GH match the lower end of the emission reduction scenarios (RCP4.5 and RCP6), and the scenarios WL and WH match the high-emission scenario (RCP8.5) which does not include specific climate mitigation [35,36]. Within all of the model simulations, a distinction can be made between models showing a large and small precipitation response [37]. Simulations with a large precipitation response foresee wet humid winters and dry summers, and are assigned to H-scenarios. The change in precipitation response is linked to circulation change, and H-scenarios show more frequent westerly circulation in the winter and domination of high pressure in the summer. In the summer, this means an increase in solar radiation and more easterly winds, which implies higher temperatures. This weather type also favors urban heat islands.

The scenarios provide monthly temperature increments for daily percentiles [38]. These increments were relative to the climate period of 1981–2010. There is already a climate signal in the time series (2002–2016) compared to this climate period, and this bias was subtracted from the results. Using the procedure of Molenaar et al. [24], these increments were interpolated for all of the days in a month. In this study, the transformation was performed on hourly values, because the time frame in which minimum temperatures were determined differs from the standard, i.e., from 8:00 a.m. universal

coordinated time (UTC) to 8:00 a.m. UTC the next day. After the creation of a future hourly time series, a proper minimum temperature can be derived. As such, we use

$$T^f(h) = T^c(h) + \Delta T(d) + \Delta HTD(h) - \overline{\Delta HTD(h)},\tag{3}$$

where

$$
\Delta HTD(h) = \frac{\left(T^c(h) - T\_{\text{min}}^c\right)}{\left(T\_{\text{max}}^c - T\_{\text{min}}^c\right)} \times \Delta DTR. \tag{4}
$$

The symbols are defined by the following:

*Tc*(*h*) = hourly temperature current climate *Tc min* = minimum daily temperature, 8:00 a.m. to 8:00 a.m. the next day *Tc max* = maximum daily temperature, 8:00 a.m. to 8:00 a.m. the next day *Tf*(*h*) = hourly temperature for future climate Δ*T*(*d*) = daily temperature increment to future climate Δ*DTR* = change in average diurnal temperature range to future climate Δ*HTD*(*h*) = hourly temperature deviation Δ*HTD*(*h*) = 24-hour average bias of hourly temperature deviation

For a transformation to hourly values, the daily temperature increments, Δ*T*(*d*), were added to individual hours (see Equation (3)). This increment, Δ*T*(*d*), was the outcome of the procedure of Molenaar et al. [24]. A novel aspect is that the change in diurnal temperature range (DTR), Δ*DTR*, between the future and current climate was also taken into account (Equations (3) and (4)). With Equation (3), a new dataset of minimum temperatures for 2050 can be directly calculated using the hourly temperature dataset of *Tf*(*h*).

The UHImax also changes in a future climate (see Equation (1)). The scenarios generally show an increase in global radiation and generally show a decrease in DTR, which influences the UHImax. On the contrary, however, the H-scenarios (GH and WH) show a small increase in DTR for the warmest months of July and August. In all climate scenarios, these months show an increase in global radiation from 2% in the GL scenario to 7% in the WH scenario, as well as a change in DTR between −3% in the GL scenario and 2% in the WH scenario. For the transformation in global radiation, we followed the method of Bakker [39]. The climate scenarios provided only monthly changes in global radiation. In our study, the change in radiation was distributed over all days without exceeding the maximum realistic daily radiation sum, which was set at 75% of the radiation at the top of the atmosphere [39]. Changes in wind speed were not considered, because they fell within the natural variation range [34]. Transformation numbers can
