The monitoring campaign adopts the same methodology developed in companion works by research group TEP-130 of Seville University (24). This task aimed to record environmental indoor data and analyze their evolution over a year (from April 1, 2013 to March 31, 2014). UNE-EN ISO 7726:2002 (25) protocol was followed in the monitoring process. Indoor air temperature and relative humidity data were collected every 30 minutes using wireless devices with sensors located in the living room and the main bedroom. Outdoors, a weather center station also recorded outdoor air temperature, relative humidity data and wind speed and direction. 563,328 data items were collected during the year of monitoring (26).

A pressurization and depressurization trial was carried out in the dwelling, using Minneapolis Blower-Door Model 4/230V System equipment, to check the envelope air-tightness levels (27), following UNE EN 13829:2000 (28).

An infrared thermography study was also carried out to study thermal homogeneity, as well as to identify air leaks and whether thermal bridges are present or not in the envelope. Thermal images were analyzed with FLIR Systems ThermaCAM reporter (29) (30).

The energy model was developed with DesignBuilder (v.2.4.2), which uses EnergyPlus (31), a simulation program recognized by the US DOE (32).

The energy model recreates the living room (day zone) and the bedroom (night zone) of the monitored residential unit in a simplified form, maintaining real orientation and weather data collected while monitoring (Figure 2b).

Location and climate conditions of Seville were considered when developing the energy model, as were the envelope’s constructive definition and the dwelling’s operating conditions: occupancy, real daily user profile, internal loads and hot and cold temperature set points.

As it is on the first floor of a nine-storey building, surrounded by other similar blocks from the same group, the residential unit is affected by the shadows cast by the blocks as well as vegetation. All rooms are exposed to outdoor conditions, with the only exception of the thermal envelope’s part that remains connected to the common areas in the block (Figure 2a-2b). The residential unit is connected above to other similar dwellings, and below to commercial premises with no thermal insulation on the ground floor between them.

As the information was recorded under real-time conditions, the user had to be considered when generating the energy model. DesignBuilder allows the residential unit energy model to be supplemented with a complete description of the user’s profile and the internal loads: occupancy data, lighting, and use of electric equipment. Table 3 shows input data schedules and distribution over the year. Furthermore, a natural ventilation schedule was established (Table 4) – as mentioned above, there is no other sort of ventilation system. Table 5 shows a schedule for the use of solar protection in the bedrooms, the common solar protection system in collective housing in southern Spain (Table 5). These Tables reflect the annual distribution of the user’s energy consumption habits, information that was obtained from the users through various surveys over the year.

Three schedules were established for natural ventilation: Winter, Daytime, Night-time (Table 4). August was considered an exception as the only month in which the dwelling is unoccupied and unventilated during vacations.

Solar protection followed an all-or-nothing (0-100 %) schedule for opening and closing the blinds, and the annual distribution is shown in Table 5. This mostly coincides with real use.

Once the user’s calendar was confirmed and usage and operating conditions were established in the energy model incorporating the various surveys provided by users over the year, uncertainties in the energy model calibration process were greatly reduced. However, the remaining doubts can be summed up as follows:

Regarding the above considerations, a sensitivity analysis was carried out, based on two prior hypotheses in order to

define the energy model, according to the different environmental conditions which affect heat transfer performance between indoors and outdoors (the outdoor vegetation and surrounding blocks, and their influence on solar radiation, as well as the contact with the uninsulated space on the ground floor mentioned above):

Furthermore, once the results of H01 and H02 were assessed, a final analysis was carried out, starting from the model in this second hypothesis, which reflected users’ calendars and energy consumption habits. A calibration process was developed in order to incorporate the effect of natural ventilation, whose presence in this particular type of residence does not only have an environmental meaning but also a cultural one, as it is a widespread method of environmental control in collective housing stock in the Mediterranean area (33).

Although a natural ventilation usage calendar was constructed using the data from the user surveys (Table 4), it was not possible to quantify intensity, measured in air change per hour rates which also vary throughout the year, depending on general climate conditions. Owing to this, this analysis used different ventilation rates for input data: 0, 0.5, 1, 2, 3 and 4 ACH, for each month analyzed, that is, May (intermediate season), June (summer) and January (winter), following the order of the monitoring campaign, studying the effects of the different rates on the model.

The results of the energy model simulation for each of the rooms were analyzed, assuming that heat transfer was only taking place through external walls (Figure 2b), as the other partitions connected to adjacent spaces with different consumption were considered adiabatic (Table 6).

Given the existence of a blind box in the openings of the spaces monitored and following correction and conversion to atmospheric pressure, the results for the in situ air-tightness tests gave an infiltration rate ranging between 0.23 and 0.54 ACH, which were acceptable values. Indoor air quality, however, was very poor given the lack of functioning ventilation systems, with the only ventilation provided by open windows.

Figure 3 represents simulated indoor temperature evolution and indoor and outdoor data monitored in the energy models analyzed (bedroom and living room, with south and southeast orientations, respectively) in August, with zero natural ventilation rate. The results show major discrepancies between simulated and monitored data, from 5 to 7 °C in the bedroom, and from 10 to 12 °C in the living room.

Given that the residential unit is located on a first floor, the lack of solar protection provided by the blocks and vegetation is the main cause of temperatures’ dispersion. Additionally, the fact that every partition except the envelope was considered adiabatic, even the first-floor framework in contact with the uninsulated commercial premises on the ground floor, has caused a greater heat accumulation indoors, with a delayed effect on indoor temperatures. Consequently, this simplified model (hypothesis 01) is not satisfactory, so that hypothesis 02, which embodies environmental conditions, is needed.

This calibration model incorporates thermal transfer through the floor of the monitored spaces in contact with the uninsulated commercial premises on the ground floor. The model has been supplemented with elements from the immediate surroundings of the dwelling, to check the effect of solar radiation on the monitored spaces, including the other blocks in the complex and vegetation as generators of shade (Table 7).

Figure 4 shows the simulated indoor temperature in each different space and indoor and outdoor monitored data, in August, with the same natural ventilation rate as in hypothesis 01.

The results obtained for this second hypothesis show a better match of the simulated energy model, with only a minimum difference of 0.5 to 1 °C between monitored and simulated indoor temperatures in both spaces. The consideration of the surrounding elements limits solar exposure in the energy models, and therefore daytime energy gains, which translates into a sharp decrease in simulated indoor temperature values and a better match to the monitored data. Moreover, considering the energy flow through the first-floor framework in contact with the commercial premises on the ground floor reduces thermal storage.

These results suggest two main ideas: the particularly significant lack of agreement when simplified models are used, which is very common when using simulation tools to predict thermal and energy performance in buildings, and the determining fact that is the first-floor location of the residential unit, surrounded by shading elements limiting solar exposure, which should be kept in mind when assessing behavior.

After the assessment of the energy model performance in H02 for August, with the residential unit unoccupied, from which it can be concluded that the energy model works for the pre-established conditions, it is time to evaluate how far user and energy consumption habits (Table 3) affect the energy model, particularly the use they make of natural ventilation to improve indoor air quality (Table 4). As natural ventilation intensity measured in air changes per hour not only depends on how far windows are open, but also on outdoor climate conditions, the main objective is to assess which air change rate best fits the real model, keeping available measured data in mind. This analysis was carried out introducing different ventilation rate values as energy model input data: 0, 0.5, 1, 2, 3 and 4 ACH, for each of the three months of the year analyzed: May (intermediate season), June (summer) and January (winter), following the order of the monitoring campaign (Table 8).

Figures 5, 6 and 7 represent simulated indoor temperature evolution in each different room and indoor and outdoor monitored data for the months of May, June and January, with various ventilation rates from 0 to 4 ACH.

As expected, in general, the results obtained in this case do not show the same fit as H02 in August (Figure 4), when the residential unit was empty. This information shows how difficult ventilation control in these months can be, when it is almost impossible to get an exact fit between monitored and simulated indoor conditions.

In May, the intermediate season, simulation results in both rooms show a clear match with a ventilation rate from 0 to 1 ACH, during the hours when the residential unit is ventilated (Table 4), fluctuating from 1 to 2 °C up or down from the other rates, essentially when windows are open (Figure 5). When ventilation comes into play, there is a sharp leap in indoor temperatures, which almost reach outdoor conditions. This is more noticeable in the bedroom than in the living room.

In June, in the bedroom, the fitting occurs with an air renovation rate of 4 ACH, with differences of about 2 °C between monitored and simulated indoor data, while in the living room the match is clearer for a 1 to 2 ACH ventilation rate, maintaining a difference of about 2-3 °C between temperatures (Figure 6). For both rooms, during the daytime, simulated and monitored indoor data are closer together, whereas at night-time, simulated indoor data are sensitively lower, regardless of the ventilation rate adopted.

In January, observing the simulated indoor temperature tendency with different ventilation rates (Figure 7), the model that fits best is that with a 0 ACH rate, considering outdoors climate conditions. Temperature tendencies almost go hand in hand with those of the monitored data, and in general, a difference of 2 °C between the simulated and monitored indoor temperature is observed in both rooms, sometimes exceeding 3 °C when heating equipment is on, as can be observed by sharp leaps in the monitored indoor temperature.

To sum up, the H02 model, with a ventilation schedule based on user surveys (Table 4) can achieve a suitable match between simulated environmental conditions and monitored ones, with different ventilation rates, according to the seasons of the year and user occupation.