Introduction

The most important clear goal of many energy–climate–environment policies in the world is to achieve net zero CO2 emissions. The imperative to limit the global average air temperature increment to 2 °C or even 1.5 °C relative to the pre-industrial period would require global yearly CO2 emissions to be net zero or net negative at the end of this century and maybe at the end of 2050. It is also known that the quantity of global warming rise is related to the total CO2 emissions. For these reasons, achieving net zero emissions as a clear and absolute goal is becoming an increasingly important goal of energy, climate, and emissions politicians worldwide [1]. However, the energy needed in buildings today covers approximately 33% of the total global energy consumption (Fig. 1), and 27% of global greenhouse gas emissions originate from building energy operations [2].

Fig. 1
figure 1

Share of the global energy demand for buildings [2]

Full size image

Energy demand studies play a crucial role in designing a cost-effective and optimum design for a building’s heating, ventilating, and air-conditioning (HVAC) units. Facades of buildings are becoming more significant than ever because of the goal of fulfilling energy demand with net zero emissions. The workload of the HVAC systems is decreased and aids in energy savings owing to the building facades created with the proper design. Determining the amount of degree days of heating and cooling (HDD and CDD) separately in the area where the building will be constructed is critical in terms of evaluating the HVAC system capacities and costs. In this sense, the CDD and HDD values are the most important impact indicators utilized to analyze the energy demand of buildings [3, 4]. Designers can integrate both active and passive cooling/heating techniques into the building faces they create while taking these indicators into account. It is essential to accurately analyze meteorological or climatic parameters in the design of indoor comfort qualities such as the cooling and heating conditions of buildings. In particular, for countries in the western hemisphere, the CDD and HDD values play a critical role in the design of HVAC units, energy demand planning, and estimation of cooling and heating loads. Among several approaches for calculating monthly or annual energy demand, the degree-day technique is the simplest. Since the degree-day technique can provide a simple assessment of yearly energy demand, it is widely utilized in HVAC applications to predict the energy needs of cooling systems in the summer season and heating systems in the winter season [5].

Literature studies show that an increase in the heating or cooling needs of consumers will affect future energy demand. Due to climate change, extreme cold in the winter and overheating in the summer increase energy consumption, and depending on this increase, the heating and cooling costs of households or users also increase. Determining how much of this increase in energy costs is due to climate can only be done with the help of degree-day calculations. The number of hot days is used when calculating CDD, while the number of cold days is considered when calculating HDD. An increase in the number of high degree days indicates more intense heat or cold. A greater number of days also require more energy consumption to heat or cool a place. In this regard, it is crucial for future energy demand management to decrease energy demand costs and achieve maximum efficiency by accurately controlling environmental concerns related to energy consumption. Estimating and meeting the future energy consumption accurately are related to the most accurate forecasting of the cooling and heating demands due to climate change [6].

Globally, a decline in the number of heating days and a growth in the number of cooling days are predicted in the future. The net effects of these two usual extreme changes caused by the increase in average temperatures need to be scrutinized, especially in predicting the expected energy demand. Determining seasonal energy demand has critical importance in computing the cooling and heating demands of any building. Today, air-conditioning experts utilize design data in building energy demand analysis; therefore, they analyze cooling and heating situations in terms of interior thermal comfort design in buildings. HDD and CDD supply significant facilities for computing the building heating and cooling energy demands, designing the building facade to be used, energy management planning, and defining the size of HVAC units [7].

Related works

The time series analysis is classified into two types based on the total number of sample parameters at the time point: univariate and multivariate. Because of some complex features, modeling of time series and hence accurate time series estimation can become extremely difficult. In recent years, time series modeling and forecasting have become significant areas of study in engineering applications. Several distinct time series methods have been suggested in the literature, such as the autoregressive model [8, 9], the autoregressive integrated moving average (ARIMA) model [10,11,12], the support vector machine model [13, 14], and neural network-based techniques [15, 16]. Apart from these basic approaches, many hybrid methods [17, 18] have also been presented. Since time series datasets such as climate, financial, and seismic values are often non-stationary and nonlinear, these current techniques cannot virtually reveal sufficiently ordered data characteristics to yield accurate time series forecast results. Therefore, the deep learning approach, which has been accepted as a new technique that can accept a multilevel model of the original input by integrating simple but nonlinear modules, has been developed as a time series model. Because of these sufficient transformations, deep learning can learn all its properties from the input data at a sufficient level.

Degree-day methods have been widely presented in the literature by researchers in the energy analysis of buildings. As given in Table 1, HDD and CDD data were analyzed for different nations in international studies such as [19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36]. From the literature study, it was revealed that a study was conducted in two directions in terms of time in determining the HDD and CDD data. These are historical CDD and HDD trends and future CDD and HDD trends. Several scholars have investigated the historical tendencies of CDD and HDD data in different locations across the world using historical meteorological records. For example, Satman and Yalcinkaya [5] computed HDD and CDD data for 77 provinces throughout Türkiye. The HDD and CDD values were determined by Büyükalaca et al. [26] utilizing daily atmospheric temperature data from 78 cities throughout Türkiye. Using hourly atmospheric temperature data from 1983 to 1992, Papakostas and Kyriakis [23] calculated CDD and HDD data for two cities in Greece. Yildiz and Sosaoglu [21] used data over 30 years from 1975 to 2004 to report yearly and monthly HDD and CDD data for 100 unique localities in Türkiye. Jiang et al. [25] used data on atmospheric temperature values recorded at 51 meteorological observations between 1959 and 2004 to calculate yearly and seasonal CDD and HDD data in Xinjiang, China. With the help of 21-year (from 1985 to 2005) daily maximum and lowest outdoor temperature data for 79 cities, Dombaycı [7] estimated the CDD and HDD data based on various base air temperatures and ultimately produced an HDD and CDD map of Türkiye. Al-Hadhrami [22] used daily average atmospheric temperatures at 38 meteorological points to report annual and seasonal CDD and HDD data in the Kingdom of Saudi Arabia. Maps of the CDD and HDD for Morocco were produced by Idchabani et al. [24] using information from 37 meteorological stations during a 10-year period from 2000 to 2009. De Rosa et al. [19] used daily averaged meteorological data from 1978 to 2013 to study the present situation and historical tendency of HDD and CDD in Italy. Using historical data from 2005 to 2014, Indraganti and Boussaa [29] presented CDD and HDD data for several cities in Saudi Arabia. Using 30-year long-term collected data, Amber et al. [30] calculated annual CDD and HDD levels in several Pakistani locations. For the years 1999 through 2019 in Australia, Livada et al. [27] examined the heating and cooling degree hours utilizing ambient air temperature and wind direction. The distribution of HDDs worldwide, and their interannual variations and significant factors during 1970–2018, was investigated by Li et al. [28]. Sadeqi et al. [20] examined the temporal trends and change points in HDD, CDD, and their simultaneous combination (HDD + CDD) in Iran, spanning a 60-year period between 1960 and 2019. Using the idea of dehumidification degree days based on moisture content, Cao et al. [37] examined the degree days over the previous 57 years for temperature decline and dehumidification in four cities belonging to key climate zones in China. To establish climate zones, Omarov et al. [38] suggested a novel method based on HDD, CDD, and building energy performance. The authors chose 90 provinces from 18 climate zones around the world for this purpose, and utilizing hourly weather data files from two separate record periods (2004–2018 and 2016–2020), they estimated and compared the DD values. Using data for the years 1979–2021, Pangsy–Kania et al. [39] examined the effects of climate change, or global warming, on the number of HDD and CDD as well as the weather-related final energy consumption of European households (EU–27 and Norway). On the basis of daily temperature measurements from 27 sites in Bangladesh, Islam et al. [40] investigated the trend and variability of CDD and HDD and their potential causes for fluctuation throughout the study period 1980–2017. Al-Saadi et al. [41] used monthly average temperature data from 2005 to 2015 to calculate the CDD and HDD for 31 meteorological stations in Oman. Using the average monthly temperature data for the years 2017–2021, Salem et al. [42] provided the CDD values for the capitals of the Gulf Cooperation Council (GCC).

Table 1 Some of the typical studies on the HDD and CDD estimations
Full size table

Several studies have predicted future CDD and HDD values utilizing future climate data in addition to analyzing historical CDD and HDD trends. For instance, OrtizBeviá et al. [32] studied CDD and HDD data for the years 2001–2050 after first calculating HDD and CDD data from 1958 to 2005 for Spain using air temperatures from 31 sites. Spinoni et al. [36] used regional climate models to investigate the spatial dispersal and quantification of possible significant differences in CDD and HDD data between 1981 and 2100. Ramon et al. [31] analyzed CDD and HDD data in Belgium utilizing hourly mean atmospheric temperatures from 1976 to 2004 for the recent past and 2070 to 2098 for the future. Ukey and Rai [34] analyzed the future and past developments of annual HDD and CDD data from 1969 to 2017 and 2018 to 2100 using weather data. Bilgili et al. [43] used SARIMA and LSTM models to predict HDD data for one month ahead and their short-term future. Bilgili [44] first created the SARIMA model to predict CDD data in different areas with a high cooling need in Türkiye and then predicted the future tendencies of the CDD data between 2022 and 2031. Karagiannidis et al. [45] used regional climate models/general circulation models (RCM/GCM) simulations under two representative concentration paths (RPC4.5 and RPC8.5) to study changes in HDD and CDD in Greece in the near future (2021–2050). Karagiannidis et al. [46] presented maps of average annual CDD and HDD deviations under RCP4.5 and RCP8.5 for the period 2023–2030 compared with the reference period 1991–2020. Mehmood et al. [47] aimed to calculate Pakistan’s cooling degree days based on the heat index for the years 2050 and 2080 as well as the present.

Artificial intelligence techniques based on time series approaches have been widely employed in the estimation of weather or climate sciences in recent years. Autoregressive (AR) models, such as the ARIMA model and its derivatives, are used in traditional time series approaches. According to the literature survey, there is a little research on the use of artificial intelligence algorithms to determine CDD and HDD values. Kuru and Calis [6], for example, forecasted the monthly mean HDD data over France using a time series technique based on the SARIMA method. Işık et al. [3] estimated HDD and CDD data for 50 provinces of Türkiye utilizing artificial neural networks (ANN) and an adaptive network-based fuzzy inference system (ANFIS) and created a Türkiye map.

In the literature, the HDD and CDD data needed by the energy sector are generally calculated analytically according to a certain base atmospheric air temperature. Recent studies have dealt with the degree-day projection of future years for different global regions [31,32,33,34]. However, in these studies, future air temperature data under predicted future global warming are needed to predict HDD and CDD data. Furthermore, the determination of these meteorological data for future years is obtained using an estimation result. However, studies on the time series approach to identify HDD and CDD data for future years are limited. To the best of our knowledge, there are no published studies on the analysis of the future tendency of CDD for Türkiye to date. This work presents a model for forecasting monthly CDD data and future prediction of a data-driven CDD based on a classical (statistical) time series forecasting approach. In this study, a time series forecasting model with a very small margin of error is developed for monthly and yearly CDD time series for Türkiye. In this approach, there is no need for extra meteorological information. The developed SARIMA model is used to estimate future CDD data for Türkiye. With the SARIMA model, the future trend of CDD is forecasted for the period 2023–2040. The Engle autoregressive conditional heteroscedasticity (ARCH) test and Augmented Dickey–Fuller (ADF) test in determining, choosing, and assessing models for CDD data are used. Finally, future CDD estimations with an approximately 95% estimation interval are produced for the 216-month period from the proposed SARIMA model. The novelty and purpose of the study are summarized as follows:

  • to analyze the historical trend and current state of CDD in Türkiye from 1991 to 2022,

  • to forecast the future trend of CDD for Türkiye from 2023 to 2040,

  • to present yearly CDD maps for Türkiye using the ArcGIS software program,

  • to develop a model based on a classical time series forecasting approach for monthly CDD data in Türkiye,

  • to classify CDD zones in Türkiye in terms of facade design and energy efficiency for cooling energy demand in buildings.

Methods

Cooling degree day

Degree day is defined as the amount of energy necessary to heat and cool buildings [48]. CDD for an n-day is expressed as [25]:

$$\text{CDD}=\sum_{\text{i}=1}^{n}{\left({T}_{\text{meani}}-{T}_{\text{bc}}\right)}^{+}$$
(1)

where Tmean and Tbc are the daily average atmospheric temperature value and the reference (base) temperature value, respectively. Tbc is considered to be 22 °C for this study. The units of CDD and temperature are the same [25, 49].

ARIMA model

ARIMA model, commonly comprehended as Box and Jenkins models, is one of the statistical approaches used for forecasting the future. Autoregressive Models (AR) are methods that represent the observation value of a time series at any period as a linear computation of the same series’ observation value for a given number of periods before it and the error term. Moving Average Models (MA) are time series models in which the observation value in each period is stated as a linear combination of the same period’s error terms and the error terms of a given number of previous periods. Autoregressive Moving Average Models (ARMA) are a blend of AR and MA models that are utilized to model stationary time series. The observation value of any period of a time series is described in these models as a linear combination of a predetermined number of prior observation values and the error term [10, 11]. An ARIMA model is defined as [50]:

$${\Delta }^{\text{D}}{y}_{\text{t}}=c+{{\Delta }^{\text{D}}\phi }_{1}{y}_{\text{t}-1}+\dots +{\phi }_{\text{p}}{\Delta }^{\text{D}}{y}_{\text{t}-\text{p}}+{\varepsilon }_{\text{t}}+{\theta }_{1}{\varepsilon }_{\text{t}-1}+\dots +{\theta }_{\text{q}}{\varepsilon }_{\text{t}-\text{q}}$$
(2)

where \({\Delta }^{\text{D}}{y}_{\text{t}}\) is a Dth differenced time series. p is the AR degree and denotes the number of past observations. q is the MA degree and expresses the number of past innovations. An ARIMA model can also be stated in lag operator notation as follows:

$${\phi }^{*}\left(L\right){y}_{\text{t}}=\phi \left(L\right){(1-L)}^{\text{D}}{y}_{\text{t}}=c+\theta (L){\varepsilon }_{\text{t}}$$
(3)

where \({\phi }^{*}\) is a polynomial with exactly D unit roots and an unstable AR operator. The nonseasonal differencing operator, \({(1-L)}^{\text{D}}\), adjusts for nonstationarity in data collected throughout time.

SARIMA model

The SARIMA model is an ARIMA extension that deals with seasonality and probable seasonal unit roots. The SARIMA model is described as multiplicative ARIMA(p,D,q)x(ps,Ds,qs) for a series with periodicity s and is alternatively defined as [50]:

$$\phi \left(L\right)\Phi \left(L\right){(1-L)}^{\text{D}}{(1-{L}^{\text{s}})}^{\text{D}_{\text{s}}}{y}_{\text{t}}=c+\theta (L)\Theta (L){\varepsilon }_{\text{t}}$$
(4)

where Ds denotes the seasonal integration degree. \({(1-{L}^{\text s})}^{\text{D}_{\text{s}}}\) is the seasonal differencing operator and expresses the nonstationarity in data measured in the same period at successive times. Φ(L) is a stable and degree ps AR operator of the \(\phi \left(L\right)\) form. Also, Θ(L) is an invertible and degree qs MA operator of the θ(L) form. The SARIMA method is described by Box and Jenkins [50] in detail.

Model evaluation metrics

To assess the prediction performance of the constructed model, the root-mean-square error (RMSE), mean absolute error (MAE), and correlation coefficient (R) were utilized. The required equations are expressed in Eqs. (5)–(7) [51]:

$$\text{RMSE}=\sqrt{\frac{1}{n}\sum_{\text{i}=1}^{n}{\left({\widehat{y}}_{\text{i}}-{y}_{\text{i}}\right)}^{2}}$$
(5)
$$\text{MAE}=\frac{1}{n}\sum_{\text{i}=1}^{n}\left|{\widehat{y}}_{\text{i}}-{y}_{\text{i}}\right|$$
(6)
$$R={\left(1-\frac{\sum_{\text{i}=1}^{n}{\left({\widehat{y}}_{\text{i}}-{y}_{\text{i}}\right)}^{2}}{\sum_{\text{i}=1}^{n}{\left({\overline{y} }_{\text{i}}-{y}_{\text{i}}\right)}^{2}}\right)}^{1/2}$$
(7)

where \({\widehat{y}}_{\text{i}}\), \({y}_{\text{i}}\), and \({\overline{y} }_{\text{i}}\) present the estimated, actual, and mean values, respectively.

Results and discussion

In this study, the SARIMA model was used to perform monthly CDD forecasting data. CDD data for 81 provinces of Turkey, calculated monthly using meteorological data covering the period 1991–2022 and available on the website in [52], were used. Table 2 gives the geographical data for the 81 measurement stations or cities considered in this study. Figure 2 presents the locations of the 81 measurement stations on a map of Türkiye. Figure 3 presents the annual average air temperature map (1927–2021) of Türkiye. As seen in the figure, while the yearly average air temperatures in the cities of Ardahan, Erzurum, Kars, and Ağrı in the Eastern Anatolian Region of Türkiye are quite low, the annual average temperatures in the cities of Şanlıurfa, Mersin, Adana, Antalya, Osmaniye, and in the Southeastern Anatolia and Mediterranean Regions are quite high. As seen in Fig. 4, the elevation values in Türkiye vary from region to region, and the elevation and rough terrain increase toward the east of the country. The mean elevation of the country is 1141 m, and the region where the elevation is maximum is the Eastern Anatolia Region. As a result, a wide variety of weather appear in Türkiye, as the annual average temperatures vary according to regions and altitudes. This means that CDD values will vary by region in Türkiye.

Table 2 Geographical data of the 81 measurement stations or cities considered in the study
Full size table
Fig. 2
figure 2

Locations of the measurement stations on a map of Türkiye

Full size image
Fig. 3
figure 3

Yearly average air temperature/°C map (1927–2021) of Türkiye

Full size image
Fig. 4
figure 4

Türkiye elevation map

Full size image

Figure 5a plots the monthly CDD values between January 1991 and December 2022 for Türkiye. CDD data show strong seasonality or periodicity throughout the year, as presented in the figure, but a linear upward trend is observed. Figure 5b presents the Box plot diagram of the CDD data according to the months of the year. It is clear from the figure that there is periodicity or seasonality over the year in the CDD data. Figure 5c, d shows the sample autocorrelation function (ACF) and partial autocorrelation function (PACF) plots of the CDD time series data. The autocorrelation results showed that lag 12 can be utilized in the forecasting model of the CDD values. This means that the seasonality period of the CDD data should be selected as s = 12. Consequently, to demonstrate the necessity of the SARIMA model, first-order seasonal differencing with period s = 12 and D = 1 was performed to convert the original data into the stationary condition.

Fig. 5
figure 5

Original data analysis for SARIMA model a Monthly average CDD historical trend in Türkiye (22 °C base temperature), b box plot of the times series data, c autocorrelation function graph of the times series data, d sample autocorrelation function graph of the times series data

Full size image

Table 3 gives the Augmented Dickey–Fuller (ADF) test results of the original and transformed CDD data. The test results show that both ADF tests for the original data and transformed data achieve to decline null hypothesis that the CDD time series is trend stationary. To make CDD time series data more stationary, they were transformed into nonseasonal, normally, and seasonally differenced series. Following the first-order difference and seasonality procedures, the transformed CDD data analysis is displayed in Fig. 6.

Table 3 Results of the ADF test
Full size table
Fig. 6
figure 6

Transformed data analysis after difference and seasonal processes a Monthly average CDD historical trend in Türkiye (22 °C base temperature), b box plot of the times series data, c autocorrelation function graph of the times series data, d sample autocorrelation function graph of the times series data

Full size image

The Akaike Information Criterion (AIC) was utilized to compare training models, and the Bayesian information criterion (BIC) was applied to identify the model with the best parsimonious sample fit. The minimum BIC and AIC values were selected as the decisive criteria to determine the most appropriate model using the Econometric Modeler application. Table 4 gives the most suitable SARIMA model and the resulting BIC and AIC values. As a result of different training applications, the SARIMA(1,0,1)x(1,1,1)12 model was determined to be the best model satisfying the given conditions in the CDD time series forecasting model for Türkiye. Table 5 gives the output parameter results of the SARIMA(1,0,1)x(1,1,1)12 model. Results show that AR, MA, SAR, and MAR coefficients are significant at the 0.05 significance level.

Table 4 Most suitable SARIMA model and the resulting BIC and AIC values
Full size table
Table 5 Output parameters for the SARIMA model
Full size table

Using the SARIMA (1,0,1)x(1,1,1)12 model, the estimation values of the monthly average CDD data for Türkiye were determined from 1991 to 2040. The estimation outputs were compared with the actual data to assess the forecast performance of the SARIMA model. Figure 7 presents the historical and future trends of the monthly average CDD for Türkiye. As seen, the CDD values estimated by the SARIMA model are generally agree well with the actual data. Furthermore, it can be seen from the figure that there is a general increase in the monthly CDD values. Yearly average CDD values for the period 1991–2040 were calculated depending on the monthly CDD values for Türkiye. Figure 8 shows the yearly average CDD historical and future trends for Türkiye. As already observed in Fig. 7, a general increase in the yearly average CDD values has also been determined.

Fig. 7
figure 7

Monthly average CDD historical and future trends for Türkiye

Full size image
Fig. 8
figure 8

Yearly average CDD historical and future trends for Türkiye

Full size image

Figure 9 shows the variation in the monthly CDD against the respective monthly average air temperatures. CDD values can be directly correlated with the monthly average air temperature, because of the large temperature difference with the selected base temperature of 22 °C. The correlation obtained in this figure can describe an average CDD for Türkiye without complex calculations.

Fig. 9
figure 9

Variation in the monthly CDD against the respective monthly average air temperature

Full size image

To determine whether the SARIMA model adequately represents the statistical properties of the real CDD time series, a diagnostic check for the residuals of the SARIMA model was performed. In this regard, Engle’s autoregressive conditional heteroskedasticity (ARCH) test was applied to the residuals of the SARIMA model. Table 6 gives Engle’s ARCH test results for the residuals of the SARIMA model. The results suggest the rejection of the null hypothesis and highlight that the residuals show no ARCH effects at the 5% significance level.

Table 6 Results of the Engle's ARCH test
Full size table

To ensure the reliability of the obtained SARIMA model, apart from Engle's ARCH test, time series, normality, autocorrelation, and heteroscedasticity of residual values were checked graphically. As seen in Fig. 10a, the residual time series plot reveals that the residuals have a mean of approximately zero with heteroscedastic properties. It is clear from Fig. 10b that the residuals are approximately reasonably normally distributed with constant mean and variance. The Q–Q plot in Fig. 10c shows that the residuals are approximately normal with slightly heavier tails. All sample autocorrelation values are roughly within the 95% lower and higher confidence ranges, as shown in Fig. 10d. The acquired results demonstrated that the residuals were homoscedastic, normally distributed, and serially uncorrelated and that they were roughly centered on zero. Figure 11 presents a scatter plot of the monthly CDD data estimated by the SARIMA model with the actual dataset. Results show that the predicted CDD values commonly agree with the actual CDD data. The MAE, RMSE, and R values of the model were obtained as 5.18 °C month−1, 9.83 °C month−1, and 0.9712, respectively. This means that the obtained SARIMA model is also suitable for future forecasting of the CDD in Türkiye.

Fig. 10
figure 10

Checking goodness of fit for residuals a time series plot, b normally distributed plot with constant mean and variance, c quantile–quantile plot, d uncorrelated sample ACF plot

Full size image
Fig. 11
figure 11

Scatter plot of the monthly CDD values predicted by the SARIMA model with the actual data set

Full size image

Yearly CDD values were determined for the 81 city centers listed in Table 2. According to data based on 81 city centers maps of Türkiye’s yearly CDD for 1991, 2006, and 2022 were drawn, as shown in Fig. 12. Then, a yearly CDD map was created on the basis of the average of the data between 1991 and 2022 (Fig. 13). As seen from the figures, the yearly CDD values vary significantly according to the region. Since the warmest provinces, in general, are found in the Southeastern Anatolia and Eastern Mediterranean regions of Türkiye, it is possible to note that the annual CDD values are highest in these regions. As expected, it is clear from the figures that the minimum cooling requirement is generally in the Central Anatolia, Eastern Anatolia, and Black Sea Regions of the country. The highest CDD values for the years 1991, 2006, and 2022 were obtained as 1023 °C year−1, 1178 °C year−1, and 1210 °C year−1, respectively, in Şanlıurfa city located in the Southeastern Anatolia Region. On the other hand, the lowest CDD values for the same years were found as 0 °C year−1, 3 °C year−1, and 1 °C year−1, respectively, in Ardahan city in the Eastern Anatolia Region. It is clearly seen on the maps in Fig. 12 that there is a general upward trend in yearly average CDD values over the years. The primary goal of this research is to simulate actual CDD data for Türkiye and estimate future values using the constructed model. Türkiye CDD time series data for the next years were revealed for this purpose utilizing the created SARIMA model. Future projections for the 216-month period (1 January 2023–31 December 2040) are shown in Fig. 7. In addition, Turkey’s annual average CDD data are calculated until 2040 and are shown in Fig. 8. Maps of Türkiye’s yearly CDD for the years 2030 and 2040 are presented in Fig. 14. The results show that CDD values for Türkiye will tend to increase in the coming years. For example, the annual average CDD value of Türkiye, which was 362.11 °C year−1 in 2022, is expected to exceed 400 °C year−1 in 2033 and reach a maximum level of 415.76 °C year−1 in 2039.

Fig. 12
figure 12

Map of yearly CDD for Turkey with a 22 °C base temperature a 1991, b 2006, c 2022

Full size image
Fig. 13
figure 13

Map of yearly CDD for Turkey based on the average of the data between 1991 and 2022 with a 22 °C base temperature

Full size image
Fig. 14
figure 14

Map of yearly CDD for Turkey with a 22 °C base temperature a 2030, b 2040

Full size image

Designers create designs with user comfort in mind. Thermal comfort is one of these comforts that allow the user to use the building comfortably. As a shell that protects the building from various factors, the facade is critical in providing thermal comfort. The cooling loads of the building can be fulfilled by a facade designed in accordance with the physical environmental conditions. Temperature increases due to climate change, as demonstrated in the study, are a factor that the designer should consider during the initial design phase. The developed facade designs fulfill the necessary cooling load based on these temperature increases, thus contributing to energy savings.

The facade of a building, which generally refers to the front or exterior of a building, is essential in architecture and engineering because of its influence on energy efficiency. Furthermore, because of its impact on energy efficiency, the facade of a building is critical. The facade frequently is the most essential design feature in architecture because it dictates the structure of the remainder of the building. In this regard, CDD zones were created for buildings in Turkey in terms of facade design and energy efficiency based on the magnitude of the CDD values acquired in this study. Zone I was designated as CDD ≤ 250 °C year−1, Zone II as 251 °C year−1 ≤ CDD ≤ 500 °C year−1, and Zone III as CDD ≥ 501 °C year−1. Figure 15 depicts the CDD regions created for the years 2022 and 2040. Zone I is the zone with the least cooling required. Zone II denotes areas that require greater cooling than Zone I. Zone III, on the other hand, depicts areas where cooling is crucial. In 2022, Zone III includes the cities of Adana, Adıyaman, Antakya, Antalya, Aydın, Batman, Denizli, Diyarbakır, Gaziantep, Iğdır, İzmir, Kahramanmaraş, Kilis, Malatya, Manisa, Mardin, Mersin, Osmaniye, Siirt, Şanlıurfa, and Şırnak. Zone III’s borders are expected to enlarge in 2040, adding the province of Tunceli to its boundaries. It is necessary to place a greater emphasis on the most appropriate facade design in terms of energy efficiency for the cooling needs of the buildings to be built for the provinces that include this region, which has grown in size since today.

Fig. 15
figure 15

CDD zones for the buildings in Turkey in terms of facade design and energy efficiency

Full size image

Conclusions

It is imperative that countries create simplified forecasting models, considering future climatic conditions, to take appropriate policy measures and develop energy consumption scenarios that guarantee the highest energy savings. In this regard, analysis of the historical evolution of CDD values in any region is critical for determining the current state and future trends of energy consumption in that region. The facade of a building can be defined as faces that establish the connection between the interior and exterior of buildings and provide control against the physical environment. In this context, facades play an important role in issues such as esthetics, structural, and protection from external influences. In this study, historical trends, current state, and future forecasts of CDD in Türkiye were performed to develop an optimum facade design for buildings. CDD values were calculated using daily average temperature data measured at meteorological stations in 81 city centers of Türkiye for the period 1991–2022. The base temperature was considered as 22 °C in the calculations. CDD data were modeled using the SARIMA approach, and a time series forecasting model was performed for Türkiye. With this time series SARIMA model, future trends in CDD were predicted for the period 2023–2040. In the developed time series model of Türkiye’s CDD data, in general, the basic time series features that characterize the long-term increase/decrease trend, periodic seasonal behavior, and stochastic behavior of the data were considered. With the help of the ArcGIS software program, yearly CDD maps of Türkiye were created for 1991, 2006, 2022, 2030, and 2040, and the historical and regional changes in CDD values throughout Turkey were visually revealed. The obtained results reveal that due to global warming, yearly CDD values in Türkiye generally show an increasing trend in the period 1991–2022, and this increasing trend is expected to continue in the 2023–2040 period. Architects and engineers can fulfill their net zero-emission building targets by designing with this information in mind.

It is widely known that heating, cooling, and lighting systems within buildings account for a significant portion of total energy consumption. The building facade is among the crucial components aimed at minimizing this energy usage. The strategic incorporation of transparent areas and shadow elements into the facade allows for controlled daylight utilization, thereby reducing the demand for artificial lighting. Furthermore, effective management of solar energy gain plays a pivotal role in achieving optimal thermal comfort and minimizing heating and cooling requirements. Contemporary advancements in facade design involve rigorous research, employing specialized software, simulation programs, and genetic algorithms to optimize various facade parameters. Given Turkey's predominantly sunny climate, designing buildings with daylight considerations is particularly pertinent. In future studies, it is aimed to use daylight and solar radiation at an optimum level in buildings in Turkey by focusing on visual and thermal comfort. Using meteorological data analysis and appropriate software, scenarios will be developed to determine the optimal facade configurations for maximizing daylight utilization.