3.1 Calibration process
To validate the accuracy of the model in Designbuilder, the two sets of hourly simulated and measured air temperature data were compared. The process consisted of analyzing the differences in the mean temperatures as well as performing statistical analysis with the coefficient of determination (R2), coefficient of variance of the root mean square error (CVRMSE), and the root mean square error (RMSE). The latter has been a statistician used as a reference to validate simulation models in different studies [36, 47].
The results of this process can be observed in Table 2. This corresponds to different corridors of the house with orientations southeast (SE), northeast (NE), southwest (SW) and northwest (NW). The differences between the mean temperatures of the measured data and the simulated model were 1ºC or less in all cases. The determination coefficients (R2), depending on the corridor, vary between .94 and .95. These values are close to 1, which determines the highest accuracy. Likewise, the values for CVRMSE in all cases are no greater than 4%.
Table 2
Differences and statistical data obtained by comparing simulated and measured air temperature results.
| South-east corridor (SE) | North-east corridor (NE) | South-west corridor (SW) | North-west (NW) |
Diferences in Mean Temp. (ºC) | 1 | 0.4 | 0.9 | 0.8 |
R² (Coefficient of determination) | 0.94 | 0.95 | 0.94 | 0.94 |
CVRMSE (Coefficient of Variation of the Root Mean Square Error) | 0.04 | 0.03 | 0.04 | 0.04 |
SD (Standard Deviation) | 3.36 | 2.93 | 3.14 | 2.81 |
RMSE (Root Mean Square Error) | 1.31 | 1 | 1.22 | 1.31 |
NSE (Coefficient of efficiency) | 0.85 | 0.88 | 0.85 | 0.78 |
Performance rating (According to Ritter & Muñoz-Carpena) | Good | Good | Good | Acceptable |
Additionally, to address the evaluation of the model's performance and reduce subjectivity for a proper interpretation, the method proposed by Ritter & Muñoz-Carpena, which combines three assessment tools, was used [50]. The first tool consists of scatter plots to visually examine the performance of the model by looking at the agreement between the calculated and the observed values (Fig. 4). The second refers to the RMSE to quantify the prediction error in terms of the units of the variable calculated by the model.
Finally, the third corresponds to the coefficient of efficiency (NSE) of Nash and Sutcliffe, as an indicator of the goodness of fit obtained with the standard deviation (SD) and the root mean square error (RMSE), as follows:
\(NSE=1-{\left(\frac{RMSE}{SD}\right)}^{2}\) Ec. 3
From these tools, the authors propose four performance classes of the models as a guide on the ranges that indicated the performance as unsatisfactory, acceptable, good, and very good [50]. As is shown in Table 2, with the values calculated through this methodology, it was possible to evaluate the performance of the model in each corridor. For the corridors southeast (SE), northeast (NE) and southwest (SW), the performances were classified as good. While for the northwest (NW) corridor, the results corresponded to acceptable performance. Based on this evaluation, it was determined that the model in Designbuilder was accurate to continue with the next phase.
3.2 Parametric study
Based on the fourteen simulated models, results of solar heat gains through the openings of the courtyards, indoor ventilation, operative temperature, and percentage of discomfort hours were obtained for each case.
Among all the simulated cases, the minimum total solar heat gains in the week were received by the 5*5 N-S courtyard house (867 kW) and the maximum by the 20*20 NE-SW (6495 kW) (refer to Fig. 5). This difference between solar heat gains could be explained because the courtyards with higher dimensions of width and length had larger opening areas in the corridors. This consequently caused higher exposure to direct solar radiation in these spaces. Additionally, in the studied period, the total solar heat gains received through the openings of the corridors and rooms were mostly achieved on the internal east, southeast, west, and west-southwest facades of the courtyard. In the morning, the west side of the courtyard receives solar radiation since the sun rises from the northeast. Conversely, in the afternoon, solar radiation affects the east side when the sun sets in the northwest. Due to this sun path, the buildings with the long axis of the courtyard oriented N-S had higher solar heat gain, and the ones with the long axis oriented to the E-W had the lower.
Figure 6 shows the indoor ventilation rate of the simulated cases. The lower ventilation rates were registered in the 5*5 N-S case (28.1 ac/h) and the highest in the 20*20 NE-SW case (55.1 ac/h). The prevailing wind direction determined the differences between courtyard cases with the same proportions but different orientations. In the month simulated, the prevailing wind direction was southwest. The effect of wind direction in ventilation rates of cases with different orientations was greater in the rectangular plan courtyards. When comparing these cases rotated in different orientations, higher ventilation rates were perceived in the cases with courtyards' long axis oriented to the northwest-southeast (NE-SW). Meanwhile, results in squared plan courtyards indicated slight differences related to the orientations.
Regarding average operative temperature, the difference observed between the case with better thermal performance (5*5 N-S) and the case with worst performance (20*20 NE-SW) was 0.29ºC (Fig. 7). The operative temperature is greatly influenced by the solar heat gain of the buildings. As mentioned before, solar heat gains depend on the opening area that increases with higher dimensions of courtyards. For that reason, the case with greater solar heat gains coincides with the case with higher operative temperature. On the other hand, other parameters that influence the energy balance of the buildings and thus operative temperature are the roof area and floor area. According to the design criteria of the cases, these areas increase as the width and length of courtyards become greater. The roof area is the second parameter that affects the heat gains of the building. Larger roof areas allowed higher heat gains to the building. On the contrary, larger floor areas promote higher heat losses in the building, which has an important effect on the energy balance of the building.
In addition, the percentages of discomfort hours were calculated from the operative temperatures. The cases with higher operative temperature promote greater percentages of discomfort hours because they exceed the upper 80% acceptability limit for more hours. Considering one week period, the lower operative temperatures and, therefore, the lower percentage of discomfort hours were obtained in the 5*5 N-S case (33%). Conversely, the percentages of discomfort hours were higher in the 20*20 NE-SW case (37%) (Fig. 8).
3.2.1 Energy Balance related to proportions
The cases oriented north-south (N-S) that in most cases had better performance were selected for a better understanding of the thermal performance of the courtyard houses. For this, the total heat gains and losses of the building were compared to carry out a detailed analysis. Figure 9 shows the energy balance of the different cases where different parameters of the house determine heat losses and gains. The largest amount of heat gains in all cases was dictated by the roofs and the solar radiation through the openings. At the same time, the major losses correspond to the floors and external infiltrations.
As observed in Fig. 9, the heat gains and losses increase as the dimensions of width and length of the courtyard and the volume of the building become greater. In addition, the percentage of solar heat gains received through the openings got larger related to the total heat gains of the building. For example, in the 5*5 N-S case, the percentage of heat gains received through the solar exposure of the courtyards openings correspond to the 59% of the total amount of gains, whilst in the 20*20 N-S case, the percentage intensifies to 85%. Otherwise, the gains related to natural ventilation in all cases remained lower than 1% of the total amount.
It should be noted that varying dimensions of courtyards affect the dimensions of the building. Therefore, an important parameter is a relationship between the volume of the courtyard and the volume of the building. In the proposed cases, this relation increases from the building with the smaller courtyard (0.04) to the building with the larger one (0.35). Results obtained from the energy balance, which considers the total heat gains and losses, showed that the heat gains get higher as the ratio between the courtyard volume and the building volume grows (Fig. 10a).
Furthermore, the volume of the courtyard and aspect ratio also affects the area of the openings, which in turn influences the solar heat gains of buildings. Figure 10b shows the relation between the area of the courtyard openings and the solar heat gains that are received through them. As the area of the openings increase, the solar heat gains per square meter get higher. For instance, in the 5*5 N-S case, the solar heat gains per m2 correspond to 7.55 kW, and in the 20*20 N-S case, the gains rise to 13.82 kW.
3.2.2 Thermal performance of different spaces
Inside the courtyard houses, two different types of thermal performance were observed according to the type of space. The rooms had lower thermal swings and more hours inside the thermal comfort zone bounded by the 80% acceptability limit. These spaces also could offer the possibility for inhabitants to control the openings to enhance natural ventilation or decrease air temperature during the first hours of the day. On the other hand, corridors had higher thermal swing but offered a wider range of conditions than the rooms. In some traditional houses of Colima, these spaces also function as dining rooms and living rooms because of this. Concerning the different cases proposed, the operative temperatures of corridors increase as the length and width get greater, in some cases exceeding the outdoor air temperature.
For the analysis of the thermal performance in a 24-hour cycle of the different spaces of the house, two cases were selected. These were the cases with lower (5*5 N-S) and higher (20*20 NE-SW) results in solar heat gains, operative temperature, and ventilation rate. As for the spaces inside the house, they consisted of two types. The enclosed rooms in the periphery of the building and the corridors function as a transitional space between the courtyard and rooms. The analysis was carried out on May 30, a typical day of the month according to climatic data. Additionally, two ventilation strategies were considered. The first one, where the glass windows with the wooden frame between the rooms and the courtyard remained closed 24/7 (0% opened) and the latter opened the windows 100% all day long. The openings connecting the corridors and the courtyard are cavities that allow wind flow permanently.
Figure 11 shows the thermal performance in a 24-hour cycle of the 5*5 N-S case. The outdoor air temperature had a thermal swing of 11.49ºC with a maximum temperature of 33.7 ºC. Among the indoor spaces, higher thermal swings (7.8ºC) were observed in the corridors in comparison with the rooms (3.7ºC). Regarding the two ventilation strategies, a difference of 1.7ºC was observed in the thermal swings of the rooms. When windows were left 100% opened, the maximum and minimum got closer to outdoor temperature than when windows remained closed. The maximums and minimums differed at 0.6 ºC and 1.1 ºC, respectively. Furthermore, if windows were 100% opened, the temperature of the rooms during the day exceeded the upper 80% acceptability limit of thermal comfort for a greater number of hours. For this reason, maintaining windows opened at this time of the day would not be adequate for the inhabitants since it would affect their thermal comfort.
On the other hand, the results for the 20*20 NE-SW case proved an increase in the maximum temperatures of the corridors compared to the outdoor temperatures (Fig. 12). In this case, the differences in temperatures in the rooms were lower when the openness of the windows was modified. In maximum temperatures, the rooms with 100% opened windows were 1ºC higher, and in minimums, they remained 0.4ºC lower. The maximum temperature was 0.5 ºC higher with respect to outdoor temperature and showed a temperature delay of 1 hour.