Impact of building energy mitigation measures on future climate

As Cities are increasing technological e�cacy on greenhouse gas (GH) emissions reduction efforts, the surrounding urban ecosystems and natural resources maybe affected by these measures. In this research, climate indicators such as heat index, extreme hot events, intensi�ed urban heat island (UHI) and sea-breeze are projected for mid and end of 21st century to understand the climate change signal on these variables with and without building energy mitigation measures. Cities amplify extreme heat and UHI impacts by concentrating large populations and critical infrastructure in relatively small areas. Here, we evaluate the combined climate and building energy mitigation impacts on localized climate metrics throughout the 21st century across extreme emissions scenarios (RCP8.5) for the tropical coastal city of San Juan. Analysis of statistically downscaled global circulation models outputs shows underestimation for uncorrected summer daily maximum temperatures, leading to lower extreme heat intensity and duration projections from present time which are corrected using bias corrected techniques. High resolution dynamical downscaling simulations reveal strong dependency of changes in extreme heat events at urban settings, however the intensities shift to lower level grassland and cropland with energy mitigation measures (combination of white roof, tilted photovoltaic and e�cient heating ventilation and air conditioning system). The building energy mitigation measures have the potential of reducing the UHI intensities to 1 0 C and 0.5 0 C for 2050 and 2100 climate period, respectively.


Introduction
Tropical coastal areas contain almost one-third of the total world population and are highly vulnerable to global climate change (IPCC 2007b).The intra American Region (IAR) is a tropical-coastal converging zone, de ned as the area enclosed by 0°N to 30°N and 100°W to 40°.The IAR region encompasses Northern South America, Central America, Gulf of Mexico, the Caribbean region, and the Western Atlantic, where a complex interaction among synoptic atmospheric/oceanic patterns drives the rainfall activity in the region (Gamble et al. 2008;Angeles et al. 2010;Glenn et al. 2015;Hosannah et al. 2017).This tropical region is used in this study as a case study to assess its climatology and the relationship between rising temperatures and other environmental variables due to global warming climate.
The vulnerability to climate change and associated extreme weather events in tropical coastal areas are directly related to accelerated Sea Surface Temperatures (SST) and air temperatures increase (Moises et al. 2017).In addition to higher SSTs in the IAR, the Bermuda-Azores high-pressure system (one pole of the North Atlantic Oscillation Index) interacts with the Caribbean low-level jet, causing a downward dry air, hindering precipitation, and warming the surface (Gamble et al. 2008;Moran et al. 2002;Wang et al. 2007).A clear regional warming was detected for the particular case of the IAR region for the years 1982 to 2013, where the SST had an annual trend of 0.0209°C per year (Glenn et al. 2015), where the air temperature is observed to be rising at a higher rate of 0.030°C per year (Angeles et al. 2017).In addition, the Atlantic Warm Pool depicts a continuous intensi cation in the last decades, which in conjunction with a warmer atmosphere increases the risks for extreme temperature events.Excessive warmer atmosphere may lead to localized high-pressure ridges, which could produce heat waves or extreme hot events, with potentially disastrous social, health and economic consequences (González et al. 2017).Of concern are the several de nitions of a heat wave and extreme hot events have been proposed in the scienti c literature (Karl et al. 1995;Delworth et al. 1999;Deo et al. 2005;Robinson 2010; Fischer and Schär 2010).
A widely used approaches considers 97th percentiles for the entire weather station data based on either heat index or temperature for two to three consecutive days (Robinson 2010; Ramirez-Beltran et al. 2017, Angeles et al. 2018).Using this de nition for the entire IAR region between 1980-2014, it was found that 144 extreme heat events were reported for a period of 35 years out of which 11 events were recorded for San Juan, Puerto Rico (Ramirez-Beltran 2017).Local records San Juan indicates that maximum daily temperature of 30-35 C occurred over 14% of the time in the last 52 years and these very warm conditions are becoming more frequent and intense (Méndez-Lázaro et al. 2015).For a particular documented case of San Juan Puerto Rico for, a high-pressure system induced low southeasterly winds responsible for record high summer temperatures (August 2012), leading to high rates of mortality related to heat stroke and cardiovascular diseases (Méndez-Lázaro et al. 2016).Extremes heat events in general expose elderly population and people without air-conditioning unit, in high risk vulnerability.For the IAR region Heat Waves Projection for 21st century is conducted using General Circulation Models (GCM) for different climate change scenarios indicating an increase in both frequency and intensities (Angeles et al 2018).As it is useful to evaluate global and continental scale climate impacts, the coarser resolution (~ 100km) limits its potential to represent regional and local signi cant process due to complex elevation, coastlines, and heterogeneity in Land Cover Land Use (LCLU) as well as ner scale atmospheric process (e.g.anthropogenic feedback, clouds and convection).A general question for urban regions, especially in tropics is how different climate indices (especially extreme heat events, urban heat island, and heat index) are represented along with other environmental variables such as humidity and winds for an extreme climate change scenario.One can further ask the role of mitigation measures adopted to combat climate change impacts on urban and its surrounding LCLU.
LCLU in urban surfaces generates feedbacks among land-atmosphere that can exacerbate extreme heat event conditions in densely populated cities.For example, Li and Bou-Zeid 2013 found UHI intensi cation in the Baltimore, MD, while Ramamurthy et al. 2017 found the NYC UHI reaching up to 10°C during the summer of 2016, both attributing synergistic interactions between extreme heat events and urban surfaces to low evapotranspiration over cities, high anthropogenic heat, large thermal mass of building and paved materials, low wind speed and air pollution.Li et al. 2016 found enhancement of the Beijing Metropolitan Area UHI due to wind pro le changes (both speed and direction) during heat waves, similar to results from Founda and Santamouris 2017, who found UHI intensi cation in Athens, Greece to be highly dependent on wind pro les, especially low wind speed and changes in wind directions.Others have found evidence of UHI intensi cation during extreme heat in Madison, WI (reaching 1.80°C during day and 5.3 0 C during night; Schatz and Kucharik.2015).However, Scott et al. 2018 showed in total of 54 US cities, with daily observation record for 15 years (2000-2015) that in most of the cities analyzed, rural temperatures increased faster than urban temperatures, leading to lower UHI magnitudes during extreme hot days.Active responses, as a result, to the UHI phenomenon are highly and urgently required not only to improve outdoor thermal comfort, but also the indoor environment (He Bao-Jie.2019).This underscores the need for studies underlying land surface processes that could determine how UHI vary over coastal cities (Caribbean in this case) and how could it be mitigated on a future changing climate.
Consequently, policies and programs addressing mitigation due to rapidly increasing temperatures are rapidly growing in Latin America and the Caribbean (UNEP 2003, UNEP 2000).The United Nations Environmental Program (UNEP), through its Economic Commission for Latin America and the Caribbean studies (ECLAC), has demonstrated that renewable energy sources would play an essential role in these regions contributing to improving the inhabitant quality of life.UNEP and the World Bank have several projects dealing with mitigating climate change in regional areas, where 31% correspond to the e cient use of energy and promotion of renewable energy sources (UNEP 2003).Aggressive policies aimed at upgrading only heating/cooling systems and appliances could result in decreased electricity use as low as 28%, potentially avoiding the installation of new generation capacity (Reyna et al. 2017) that could have a positive impact on the quality of the urban climate.A recent study highlights that the built environment of the future would transform buildings into resource assets-fully self-aware, adaptive, and two-way communication with the electric grid (to optimize operating cost) and add market value to the assets (Wang et al. 2017).In addition to this, the stricter de-carbonization regulation would open doors to innovative designs and renewable energy integration in all building sectors of urban region.The free space available on rooftops used with the full potential for energy services to achieve the decarbonization goals and increase the building value.For city-scale deployment, different types of roof applications (cool roof, green roof, and Photo-voltaic roof) on buildings have shown to reduce air temperatures and energy consumptions (Salamanca et (Pokhrel et al. 2020).These later works show that a higher albedo of a tilted Photo-voltaic shaded roof can act as a radiant barrier that re ects heat from thermal radiation from the roof surface (Scherba 2011;Porkhrel et al. 2020).
In this work, we evaluate the environmental impacts of mitigation options (a combination of cool roof, Titled Photo-voltaic roof and e cient HVAC systems) based on the recommendations for reducing peak air conditioning demands (with a reduction potential of 33%) from earlier studies by Pokhrel et al.Evaluation of climate variables and energy demands at larger urban scales have either focused on the use of statistical (Beccali et al. 2004;Howard et al. 2012) or process based models (Ahmed et al. 2017;Olivo et al. 2017).The absence of interaction between weather and building is one of the limitations of these approaches and could amplify in the context of changing climate.An approach to resolve this limitation is by coupling weather prediction to Building Energy Models (BEM) (Vahmani et al. 2016;Tewari et al. 2017;Ortiz et al. 2018).However, the computation cost and lack of urban morphology and its corresponding parameters have limited the studies to a few events or in-cases to short periods (< 1 season).Here, we present the impacts of mitigation options on 2-m air temperature, extreme heat events and UHI reduction potential for the dense population region of SJMA for an extreme emission scenario for a multi-year period.The key science questions this study attempts to answer are; how are extreme temperature events projected under a warming climate, and what maybe the role of mitigation measures on environmental sustainability under a warming climate for a tropical coastal city?
The organization of the manuscript following this introduction focuses on methods on data used (GCM ensemble), point based statistical downscaling for SJMA and modeling with dynamic downscaling.The results are categorized with extreme heat events projections and also presents dynamic downscaling results for underlying land surface process that changes with and without mitigation measures for future climate change scenario.

GCM ensemble
This study explores impacts of building energy mitigation options in a warming climate.For this, two approaches are used; single point projections and dynamical downscaling techniques.For single point projections, ensemble members of GCM for temperature records are further bias corrected for mean and standard deviations for a reference point, in this case the San Juan international Airport (SJIA).These historical records of SJIA were in turn used to perform bias correction GCM used to develop all projections.A single point is used due to GCMs' generally coarse resolution, meaning that other grid points might be too distant to SJIA to provide relevant information.

Dynamic downscaling for future climate change
The results from statistical downscaling are used and served as a reference to carry out dynamic downscaling providing key periods to simulate, refer here as time slices, beside additional advantages of a dynamic downscaling effort.The ow chart, which describes this study's overall methodology for future climate projections (including statistical and dynamical downscaling), is illustrated by Fig.

C ESM BC = ERAI (mean) + C ESM ′
The simulation covers two sets of experiments: normal conditions and building energy mitigation alternatives.The normal condition represents normal roof conditions as in BEM and mitigation alternatives covers a combination of building-integrated active and passive systems such as cool roof (albedo changed to 0.7 from 0.15), higher Coe cient of Performance (COP increase to 3.5 from 3), and addition of tilted solar photovoltaic in roofs (PV) (50% of roof area for mid-century, and 100% of roof area for end-century) based on their reduction potential for the same region.The passive building-integrated mitigation options considered here have previously studied by Pokhrel et al. (2019b).For a roof with tilted PV, the approach follows work from a recent study by Pokhrel et al. (2020), where the building roof temperature of BEM has been adjusted using modi cations based on the energy balance of roof including titled PV.All numerical experiments assessed by maintaining all land surface morphology as of 2008-2012, which consists of World Urban Database Access Portal Tool (WUDAPT) Land Class Zones (LCZs) for urban classes and MODIS LCLU for natural classes.The domain con guration for dynamic downscaling and LCLU for ner domain is presented in Fig. 2.

Impacts of climate change: GCM ensemble of SBC (Static bias correction) projections
This section details the possible impacts of climate change on extreme heat events for SJMA.We apply the bias correction technique outlined in section 2 (Hawkins et al., 2013) to downscale each model in a 25-member ensemble which reveals signi cant cold biases in raw GCM output for SJIA daily maximum temperature records.We use Kernel density estimates of daily maximum temperatures for each ensemble member to quantify the impact of the statistical downscaling technique on mean and standard deviation statistics.As shown in Fig. 3, models without bias correction generally underestimate observations on an average of 5 0 C.This may be partly due to SJIA's proximity to the ocean, which is included in more of the GCM's grid cell area.The statistical downscaling technique modi es each ensemble member's distribution to match that of the station observations more closely.Besides, intermodel spread in mean and standard deviation is reduced during the reference period for the bias-corrected GCM.
Bias-corrected and un-corrected mean daily maximum temperature are shown in Fig. 4a and 4b show a nearly linear trend in both high emissions scenario (RCP8.5)and stabilization scenario (RCP 4.5).RCP4.5 shows a linear trend of 0.1778 0 C/decade, while RCP8.5 grows twice as faster, 0.3555 0 C/decade for raw GCM.The bias corrected GCM increases at a rate of 0.5 0 C/decade and 0.233 0 C/decade for both RCP 8.5 and RCP 4.5 scenarios, respectively.Model spread, quanti ed as 95% con dence intervals, become slightly wider towards the latter half of the century, however both scenario shows signi cant changes only after mid-of-century.
Bias-corrected mean intensity, de ned as the mean of extreme hot events that represents 95 percentiles (32.8 0 C) for three consecutive days (Fig. 4c), grows at a rate nearly 0.222 0 C (RCP8.5) per decade, while the uncorrected record grows at 0.1 0 C/decade, more than two times slower.Mean extreme hot event frequency (Fig. 4d) increases at a similar rate for both scenarios until mid of century increasing from 2 to 6 events, however, events frequency remains constant after 2050 mainly due to increase in total duration (days) of the events (Fig. 4e).The total event duration records that extreme heat events would last longer and increases at a rate of 3 days/year for RCP 8.5 higher by more than two times (1.222 days/year) as compared to RCP 4.5.Towards the end of century there are more hot days for RCP 8.5 scenario (Fig. 4e).This is due to both usage of a limited window for hot events to happen (i.e., August-November) as well as a constant temperature threshold for the event.As mean daily temperatures surpass 32.8 0 C by early 2060s, the likelihood of any given day surpassing this temperature increases.This might lead to separate events "consolidating" into longer lasting extreme hot events.It also explains why changes in this metric are much lower in uncorrected data (Fig. 4a), as temperatures are less likely to reach the extreme hot event threshold.These results compare favorably with the 26-model ensemble used in the projections of heat wave events for New York City (Ortiz et al. 2019), which corrects for the mean of the temperature distribution and standard deviations.The bias correction technique to downscale temperatures from GCM to local scale reduces differences between observation records by reducing biases in both the mean and variance of the model.Applying the technique to each model in the 26-member ensemble results in a reduction, in general, of inter-model spread, as all models are downscaled to the same historical record.One limitation of this approach is the assumption that the relationship between observations and GCM output will remain stationary for the entire projection period (Dixon et al., 2016), which might not account for feedback processes such as additional anthropogenic heat and soil desiccation or moistening.
Besides, the coarser spatial resolutions of GCM and limitations of availability of different variables from weather stations limits the application of statistical downscaling.
To partially address some of these limitations, we conducted a set of simulations using urban WRF, considered a state-of-the-art high resolution urbanized regional climate model.This dynamic downscaling effort incorporates most of the urban surface-atmosphere feedbacks that may modify extreme hot event conditions, rather than statistical relationships developed or assumptions about the stationarity of bias correction parameters.These assumptions are particularly relevant for projections in SJMA as it has been shown that the stationarity assumption may be violated in coastlines and especially in warm projections (Lanzante et al., 2018), where grid cells may contain water, which in turn modi es near-surface temperatures.The results from statistical downscaling with and without bias correction shows that the temperature starts to change from mid-century (2050) for both RCP 4.5 and RCP 8.5.Also towards the end of century RCP 8.5 presents a maximum rise in daily maximum temperature.These results guide us to choose RCP 8.5 for mid and end century for dynamic downscaling.

Dynamic downscaling model validation
Historical period simulations (in this case urbanized WRF) are evaluated against the San Juan international airport (SJIA) station for daily maximum temperature, relative humidity, and wind speed for the reference historical period of 2008-2012 (Fig. 5).Simulated results compared against Kernel density estimates (KDE), an approximation of a dataset's distribution.The airport station reported a mean daily maximum of 27.6 0 C with a standard deviation of 3.59 0 C. WRF simulations result, interpolated using nearest neighbor showed a mean daily maximum of 26.9 0 C (2.5% error) with 3.84 0 C standard deviation (7% error).These results are consistent with Ortiz et al. (2019), which found less than 1% and 10% error on the mean of daily maximum temperatures and standard deviation, respectively.Relative humidity, in general, underestimated in the WRF simulation for values less than 40% and higher than 80%.The bimodal nature of the relative humidity is observed in weather stations at 73 and 82%.The bimodal nature is captured in WRF simulation at 57 and 73%.Wind speed for both observation and simulation has a mean of 4.8m/s; however, the maximum wind speed is underestimated by the simulation for values higher than 8m/s by 1 to 1.5m/s.The simulated maximum wind speed of 15m/s compares with 23m/s with observation.

Dynamic downscaling model results for climate change impacts on extreme heat events
The impact of climate change on extreme hot events and on other environmental variables for mid of century and end of century are evaluated in this section.The simulations results are evaluated with and without building energy mitigation measures.Building energy mitigation options are those that are discussed in section 2, and include a combination of white roof, tilted PV and e cient HVAC systems (higher COP).Mean extreme heat event intensity (Fig. 6) projections show that more signi cant increases are closer to the coast and in the metropolitan region than inland locations.Dynamical downscaling without any building energy mitigation measures for mid-of-century (2048-2052; 2050-) and end-ofcentury (2092-2096; 2100) results in increases of the extreme hot events intensities to 35 0 C and 35.5 0 C, respectively, for the central urban location of SJMA similar to results in the statistically downscaled projections.The maximum intensity is simulated at the core of the city, inland to the coast at lower elevations where the dominant land category is compact low rise.At locations south-east of the SJIA where the dominant LCLU is cropland and grassland, the intensity is as high as 34 0 C higher than the threshold of 32.2 0 C by 1.8 0 C. Intensity, simulated to mitigate extreme temperature events, reduces the impact on the metropolitan region.Results for 2050 and 2100 with building energy mitigation measures reduce the event intensity by 1 0 C and 1.5 0 C at urban centers whereas, the maximum intensity simulated to shift towards low-level cropland and grassland LCLU close to south-east and south of SJIA.As cities increase e cacy on greenhouse gas emissions reduction efforts, the surrounding ecosystems and natural resources close to urbanization might be affected by climate system changes and variability.Thus, it is essential to nd a balance, so the impacts on ecosystems could be minimized.Here, the building energy mitigation options have a net result of decreasing the UHI, with a dual positive impact of mitigating climate change impacts, and reducing greenhouse gases.

Dynamic downscaled model results for climate change impacts on heat index
Relative humidity (RH), which measures the saturation of the atmosphere to water vapor, is projected to increase in all cases (or scenarios).Relative humidity in combination with air temperature provides information on Heat Index, which is an important consideration that determines how hot the body feels when exposed to ambient conditions, and this index has also been used to study extreme heat events for the Caribbean region (Angeles et al. 2017; Ramirez-Beltran et al. 2018).Studies have shown that water vapor content of air, in addition to ambient temperature, regulates the ability of humans to cool down via evaporation of sweat (Malchaire et al. 2000).In the business as usual scenario, the RH anomalies (Fig. 7; left panel) without mitigation measures for the mid and end of the century show increments.For 2050 there is no change along the coast, however, at urban centers, a positive increase of 2.5% is simulated.
For 2100 the anomalies are uniform of 2.5 to 5%.As RH follows the temperature trend, the anomalies are lower for the higher temp of 2100 compared to 2050.The mitigation measures increase RH in the metropolitan region ranging from 17.5-20% increase from historic periods.For tropical coastal city, extreme heat waves events are de ned based on heat index and is an essential consideration for human health and comfort.Projection of heat index anomalies (Fig. 8) indicate increases in all periods reaching 5% by 2050 and 7.5% by 2100 at SJMA.The mitigating measures reduce the heat index by 2.5% over urban centers by 2050; however, it increases to 7.5% by the end of the century.
The probability density function of daily maximum temperature and daily maximum heat index for the densely packed region of SJMA is presented in Fig. 8.The mean of daily maximum temperature is 31.2 0C, 32.3 0 C, and 33.3 0 C, respectively, for historic, mid-of-century, and end-of-century.However, the building energy mitigation measures reduce the mean to 31 0 C and 31.8 0 C with a reduction potential of 2.3 0 C and 1.5 0 C for 2050 and 2100, respectively.The tail of distribution for 2050 has a reduction potential of 0.8 0 C from historic periods.Comparison of the tail of distributions from 2050 and 2100 with and without mitigation measures shows a reduction potential of 1.8 0 C and 1.2 0 C simultaneously, indicating reduction in extreme hot events.The daily maximum heat index also shows the same reduction potential for the mean distribution without building energy mitigation measures.However, the distribution does not indicate any signi cant improvements when compared with and without mitigation measures.This is mainly due to the proportional increase of Relative Humidity as a decrease in temperature along SJMA.

Model results for climate change impacts on the boundary layer
In order to study the boundary layer pro le variation for mid and end century with and without building energy mitigation measures, air temperature (contour lines in 0 C) is plotted with horizontal wind vectors and shaded contours for vertical wind are shown in Fig. 9 for historic period at 18 UTC and 10 UTC.From Fig. 9 it is noticed that the temperature in the upper atmosphere for topography greater than 600 meters does not change signi cantly between both 18 and 10 UTPC.However, the horizontal wind speed near the surface (elevation < 200m) is simulated to be greater than 2-3 0 C for 18 UTC, also the horizontal wind speed is greater for this region especially due to the in uence of sea-breeze.Vertical upward convective motion is seen to be maximum during afternoon in the urban region and a circulation is simulated between urban locations and the nearby ocean.However, during the morning hour (10 UTC) vertical motion is noticed at higher elevation (> 400m) over the land with much higher wind speed as compared to the urban region and nearby ocean, and circulation is seen over this topography with higher horizontal wind speed.
In addition to the historic period analysis of the boundary layer, the in uence of climate change with and without building energy mitigation measures is studied to learn of any signi cant impacts on the vertical structure of the temperature, the wind speed and convective circulation for for 2 PM LST (Fig. 10).For this reason, Fig. 10 is shown with a temperature (contour lines) and horizontal wind vector difference for mid and end of century with the historic period in additional to the vertical wind (W) as a shaded contour.
From the gure it is noticed that the climate change impacts increase the near surface urban air temperature by 0.9 to 1.1 0 C for 2050 and 1.8-2 0 C by 2100, respectively, as compared to the historic period.The horizontal wind speed just over the urban region is seen to decrease by 0.5m/s and 1-1.5m/s for 2050 and 2100 respectively.The vertical wind circulation for 2050 and 2100 without building energy mitigation measures does not show any noticeable difference, despite the high difference on mean maximum temperature.This is mainly because of high reduction of wind speed for 2100 as compared to 2050 which fails to produce a signi cant change in vertical wind speed.Energy mitigation measures doesn't show signi cant difference in temperature over the entire region expect over urban region where, 0-0.3 0 C and 1 0 C increase is noticed for both 2050 and 2100 periods as compared to historic period.

Model results for climate change impacts on UHI
The UHI phenomenon can be interacting with the heat wave or extreme temperature event, and it is one of the most common behaviors reported of the climate change, making the local urban environment warming more lasting and devastating (He Bao-Jie.2019).For this reason, we look into peak day time temperatures to evaluate the UHI over the metropolitan region of San Juan passing through − 66.1 0 longitude (shown in Fig. 1).The maximum daily UHI intensities for historic, mid and end century are 3 0 C, 4 and 5 0 C, respectively, representing climate change signal without building energy mitigation measures (Fig. 11).The mitigation measures have the potential of reducing the UHI intensities to 1 0 C and 0.5 0 C for 2050 and 2100 climate period, respectively.

Conclusions
This study investigated the spatio-temporal impacts of climate change in the coastal tropical city of San Juan, Puerto Rico.Both statistically-bias-corrected GCM ensemble and dynamic downscaling projections were used.The statistical downscaled results were used to guide the selection of time slices for the dynamic downscaling.Historic period was compared with mid-century (2048-2052), and end of century (2092-2096).The environmental variables of interest investigated were maximum temperatures, humidity, and extreme heat events, all with direct social consequences to energy increases and human health.The dynamic downscaling used an urbanized version of WRF which enabled the exploration of building energy mitigation options (white roof, tilted PV and higher COP).The global climate change signal for San Juan International Airport indicates an increase of daily maximum temperature by 0.5 0 C per decade and dynamic downscaling results demonstrates that the extreme heat event intensities may reach 35 0 C for both mid and end of the century periods.The statistical downscaling further shows that the extreme hot events would be more long lasting for the end of century period.The dynamic downscaling results shows that the extreme hot events are more pronounced in the metropolitan region (urban centers) as a climate change signal for both mid and end century periods.However, the combination of the mitigation measures reveals that the extreme hot event frequency is shifted from urban centers to low level agriculture and grass land.
The average daily maximum temperature for San Juan indicates that the magnitude is greater than the historic threshold for 95 percentiles (32.8 0 C; for extreme hot events) for 2100 (33.The results presented here provides useful insights on the interplay between regional and local climate and the potential for localized intensi cation of extreme heat events.As cities increase e cacy on greenhouse gas emissions reduction efforts, the surrounding ecosystems and natural resources close to urbanization is simulated to be affected by climate system changes and variability.These insights may be useful for city-level stakeholders for planning of adaptation strategies for improving population health and energy usage. As there are uncertainties in the temporal change of urban parameters such as land cover, building height, or building technologies (e.g., higher air conditioning e ciency and improved thermal performance), which might modify urban-atmosphere interactions.One particular limitation is the assumption of static urban surface; no urban densi cation and no change in populations.This limitation might be addressed by the use of future land cover and population projection which could be provided by the policymakers and governmental agencies.The overall methodology adopted in this study.

(
2019b) and Pokhrel et al. (2020); for San Juan Metropolitan Area (SJMA) of Puerto Rico.This study may also support public energy policy such as the Puerto Rico Integrated Energy Resource Plan which aims to increase energy e ciency in the Island by 25% by 2030, and the integration of renewable energy resources by 100% by 2050 (IRP, PREPA 2018-2019).
For single point projections, ensemble members of GCMs are used belonging to the Fifth Climate Model Inter-comparison Project (CMIP5; Taylor et al., 2012) and are detailed in the Table 1.For each model, we considered daily maximum temperature between 2006 and 2100.Two scenarios are considered based on the representative concentration pathways (van Vuuren et al., 2011), RCP4.5 and RCP8.5, which use a combination of global policies, technologies, and demographic projections to estimate global radiative forcing paths.RCP 4.5 (Thomson et al., 2011) is considered a medium emissions scenario, withincreasing global radiative forcing that stabilizes by 2100 at 4.5 W/m 2 .RCP8.5(Riahi et al., 2011) is a high emission or "business as usual" scenario, with increasing radiative forcing reaching around 8.5 W/m 2 by end of century.
projections following the work of Piani et al. 2010 and Hawkins et al. 2013.The bias correction technique corrects for model mean and standard deviation using a linear model.Here, T refers to the temperature records and σ refers to its standard deviation.Subscripts Obs and GCM refer to observation and model data, respectively, while REF and RAW refer to the reference (2008-2017) and entire projection periods (2006-2100).The over bar ( ¯ ) marker denotes use of the average for the speci ed data set and time period.For all models, the geographically closest land grid point to SJIA was

Figure 1
Figures

Figure 7 Relative
Figure 7

Table 1
Twenty-ve-model ensemble and centre-of-origin used in single point heat wave projections based on location of SJIA (18.3522 0 lat.× -66.1186 0 lon.) (Gao et al. 2012;Ortiz et al. 2019peratures are used to forecast long term projections of these variables until 2100, which guides the more detailed projections.As indicated before, for detailed projections, this study uses a high-resolution con guration of the urbanized WRF model coupled with a modi ed multi-layer urban canopy and BEM parametrization as a tool to study changes in environmental variable under climate change conditions.The bias-corrected runs of the Community Earth Systems Model version 1 (CESM1) Antic et al. 2004; Miller et al. 2008; El-Samra et al. 2017; Hughes et al. 2017; Garuma et al. 2018), especially in locations where complex surface processes are signi cant (e.g., mountains, coasts, and cities), although some studies have found geographically inconsistent accuracy improvements(Wang and Kotamarthi, 2015).Besides, high-resolution dynamical downscaling methods are used to derive projections of extreme events, such as heat waves(Gao et al. 2012;Ortiz et al. 2019).Here, we employ advances in the representation of urban physics in the WRF model to project heat wave metrics and building energy throughout SJMA.Our simulation approach focuses on three time periods (Bruyere et al. 2014) datasets are used as initial and boundary conditions at a horizontal resolution 0.94 latitude × 1.24 longitude.The regional-scale biases due to having coarse spatial resolution and limited representation of some physical processes are corrected in CESM1 with bias the correction method developed by Bruyere et al.(2015).Their work adjusts CESM outputs by combining a25-year (1981- 2005)mean annual cycle from ERA-Interim reanalysis and a 6-hourly perturbation terms representing the climate signal.The bias correction removes the mean annual bias while retaining the day-to-day climate variability from CESM as following.Regional modeling forced with bias-corrected CESM was shown to improve results (Bruyere et al. 2014).Speci cally, the air temperatures over the Caribbean region showed a deceased cold bias when all boundary condition variables corrected with reanalysis data.Sea Surface Temperatures from biascorrected CESM are updated daily.High-resolution regional climate models have been used to improve the representation of precipitation and temperature (representing historical(2008)(2009)(2010)(2011)(2012), mid-century (2048-2052), and end of the century (2092-2096) and for the late rainfall season (LRS) only.We choose LRS (Aug-Nov) as it is a period where extreme hot events are more evident for San Juan Metropolitan Area(Pokhrel et al 2019a).
(Angeles et al. 2018;ower for 2050 (32.2 0 C).These are reduced signi cantly than the threshold with building energy mitigation measures for both periods.Heat Index is an important metric for heat waves in the Caribbean(Angeles et al. 2018; Ramirez-Beltran et al. 2018) and it is simulated to increase for both periods with and without mitigation measures, indicating that the mitigation measures do not help in the reduction potential or heat index.The UHI phenomenon can make extreme hot events and heat waves more lasting and devastating (He Bao-Jie.2019).The UHI increases to 4 and 5 0 C for mid and end century as compared with historic period.The building energy mitigation measures have the potential of reducing the UHI intensities to 1 0 C and 0.5 0 C for 2050 and 2100 climate period, respectively.