What Is the Temperature Acceptance in Home-Office Households in the Winter?

Abstract

Remote work can have many benefits when set up properly. Nevertheless, the preferences of home workers’ comfort havenot been profoundly studied yet. Therefore, this study aims to understand their accepted Indoor Environmental Quality values in winter based on self-reported comfort. In this regard, 26 households in Catalonia’s Mediterranean region were monitored and surveyed for separated periods of 15 days during from 22 November 2021 to 6 March 2022. Measurements including temperature, temperature, CO₂ and Relative Humidity levels were data logged in their most used workspace. Results show that most people work between 18 °C to 20 °C and find those temperatures acceptable. Common spaces used as homeoffices are on average 0.8 °C warmer and there are 470 ppm lower CO₂ concentrations. Families with children or teenagers and especially women tend to have a warmer mean operative temperature. The comparison between theoretical thermal comfort models (Fanger and Adaptive) to direct vote perception from a sample of spaces and conditions shows low correlation with real perception having the Adaptive model a better prediction of votes. In general, people report feeling more comfortable at the office, although a high number of participants feel indifferent or reckon that depends on other factors.

1. Introduction

In the context of climate change, efforts are focused on the avoidance of greenhouse gas emissions together with the application of energy efficiency measures. Considering its life cycle, buildings are responsible for 36% of greenhouse gas emissions and 40% of energy consumption in the EU [1]. To correctly design and run low-energy buildings, it is vital to comprehend the factors that influence human thermal responses. It is acknowledged that thermal comfort and indoor temperatures are strongly related to health [2,3,4]. Poor air quality has a negative effect on the performance of office work by adults [5,6]. Thus, reducing consumption and maintaining adequate levels of indoor environmental quality in buildings is a challenge addressed in the current days. Furthermore, human behavior and perception influence energy performance in buildings [4,7]. Human behavior is stochastic, not easily foreseen and depends on many factors, such as the use purpose of the building.

Theoretical models are used to design and simulate comfort conditions in buildings. There are two main empirical models used in buildings that can represent the human perception of thermal comfort: the Fanger model and the Adaptive model. Those models are widely used and referred to in the literature [8,9,10]. Both depend on given conditions or factors such as air and radiant temperature, clothing and metabolic rates. These parameters along with human perception in buildings are difficult to obtain. For example, they require expensive instruments to accurately measure [9]. The Fanger model proposes two main indexes: Predicted Mean Vote (PMV) and Predicted Percentage of Dissatisfied (PPD). Spanish guidelines [11] provide the values of all the objective thermal factors determining the thermal sensation to guarantee the desired level of thermal comfort in offices. In the winter period, it establishes a PPD inferior to 10% considering a clothing factor of 1, an operative temperature between 20 and 24 °C and a relative humidity between 30% and 70%. The newest regulation RD 178/2021 [12] considers the same hypothesis (PPD < 10% and clo = 1) but narrows this range to 21–23 °C and 40–50%. The Adaptive modelis described in ISO EN16798-1 [13]. It sets the temperature range and the relative humidity in offices between 20 to 24 °C and 40%, respectively. Thermal comfort has been studied in offices in the EU, revealing that office workers have worse perceptions than standards [5]. Another international study in Finland focused on offices finds temperature and relative humidity throughout the year in similar ranges to the RD 178/2021, being 22.5 °C and 27.2% [14]. As stated in standards [15], one of the most common air pollutants that contribute to poor indoor air quality is carbon dioxide (CO₂). Governments have been pointing to certain concentrations levels of CO₂ to minimize the risk of contagion in closed spaces, the maximum recommended value was set at 800–1000 ppm [16,17,18].

Lockdown and remote work policies have been applied when the risk of high SARS-CoV-2 transmission was detected. A growing number of people still use remote work as a common practice [19], leading to a new state of normality. Households offer human-control-based strategies to modulate indoor environmental quality, which are not available in offices. Factors such as safety from virus propagation, mental and physical health, comfort (light, noise, humidity, temperature, indoor air), equipment for work and internet [20] and facilities (separate from living and ergonomic working space, greenery) affect remote work satisfaction and productivity in the residential built environment, being the facilities the most influencing factor among others [21]. These studies are performed using surveys. However, none have been compared with the monitored data.

Currently, detailed knowledge on thermal comfort behavior is needed for more accurate simulation in energy building modeling. The objective of this paper is to deepen understanding on thermal comfort and indoor environmental comfort inside homes performing remote work. It aims to understand human perception in a home work environment and understand how theoretical models correlate to it by using both surveys and measurements. Hence, it monitors temperature, humidity and CO₂ ranges while working at home comparing them to standard ranges in offices. Thus, new thermal profiles are obtained, data which may be used in energy-building models.

This paper is structured as follows: Section 2 explains the methodology and main indicators used to analyze the data. Section 3 shows results divided into a general overview of the household sample, IEQ results, perception, CO₂, relative humidity monitoring, operative temperatures and temperature differences between rooms. and last but not least the comfort models the comfort models calculations and correlation between measurements and perception voting. Section 4 is booked for discussion and Section 5 contains the conclusions.

2. Materials and Methods

This paper combines quantitative measurements of temperature, CO₂ levels and relative humidity, qualitative daily surveys and a final interview to give and receive feedback from all participants. Furthermore, information about the characteristics of each household is collected. The summary of the data collected can be found in Table 1.

Table 1. Summary of the data collected and methodology.

Name of Data Collected	Scope	Data Collected
Monitored data	14-day monitoring   Every 2-min	Main home-office space: Temperature, relative humidity and CO2   In other rooms: Temperature   Using dataloggers COMET U3430 and Elitech RC-5 or RC-5+ (Figure 1)
Household characterization survey	41 questions   One answer per household.	Household surface, location, typology and equipment; age, gender and health of inhabitants and use and behavior with equipment and ventilation.   Performed with Google Forms
Point-in-time environmental comfort survey	14 questions   Answered daily	User identification and location (5 questions). If the answer is outside the home, the survey ends.   Clothing, metabolic activity, (scale acceptable 1 to 7 non-acceptable), thermal sensation (scale Very Cold −3 to 3 Very Hot), humidity (scale Dry −2 to 2 Humid), air quality, noise and light sensation (scale acceptable 1 to 5 non-acceptable). (9 questions)   Performed with Google Forms
Individual report	One personal report per participant	Summary of the household characterization   Evolution of temperatures during the monitored period, containing indoor and outdoor temperatures. When the environmental comfort survey is answered at home, a cross is plotted to identify occupancy periods   The application of two different comfort models (Fanger and the Adaptive model) compared with answers in surveys   Evolution of CO2, IEQ thresholds plotted according to [13].   Evolution of relative humidity, IEQ thresholds plotted according to [13].   Summary of point-in-time environmental comfort survey and comparison with the monitored values at the moment of answering the survey   Generated with PDF in python
Final interview	15-min interview   One interview per household	Times participant has been ill during the winter, where they feel more comfortable at the office or home and other specific questions depending on the report
Meteorological data	21-day data, matching with the 7 days previous to monitoring period and 14 to the monitoring period   Every 30-min	External temperature and relative humidity downloaded from the public database [22].   The nearest public meteorological station is found for each household.

Table 1. Summary of the data collected and methodology.

Name of Data Collected	Scope	Data Collected
Monitored data	14-day monitoring   Every 2-min	Main home-office space: Temperature, relative humidity and CO2   In other rooms: Temperature   Using dataloggers COMET U3430 and Elitech RC-5 or RC-5+ (Figure 1)
Household characterization survey	41 questions   One answer per household.	Household surface, location, typology and equipment; age, gender and health of inhabitants and use and behavior with equipment and ventilation.   Performed with Google Forms
Point-in-time environmental comfort survey	14 questions   Answered daily	User identification and location (5 questions). If the answer is outside the home, the survey ends.   Clothing, metabolic activity, (scale acceptable 1 to 7 non-acceptable), thermal sensation (scale Very Cold −3 to 3 Very Hot), humidity (scale Dry −2 to 2 Humid), air quality, noise and light sensation (scale acceptable 1 to 5 non-acceptable). (9 questions)   Performed with Google Forms
Individual report	One personal report per participant	Summary of the household characterization   Evolution of temperatures during the monitored period, containing indoor and outdoor temperatures. When the environmental comfort survey is answered at home, a cross is plotted to identify occupancy periods   The application of two different comfort models (Fanger and the Adaptive model) compared with answers in surveys   Evolution of CO2, IEQ thresholds plotted according to [13].   Evolution of relative humidity, IEQ thresholds plotted according to [13].   Summary of point-in-time environmental comfort survey and comparison with the monitored values at the moment of answering the survey   Generated with PDF in python
Final interview	15-min interview   One interview per household	Times participant has been ill during the winter, where they feel more comfortable at the office or home and other specific questions depending on the report
Meteorological data	21-day data, matching with the 7 days previous to monitoring period and 14 to the monitoring period   Every 30-min	External temperature and relative humidity downloaded from the public database [22].   The nearest public meteorological station is found for each household.

Figure 1. Equipment used and Specifications.

Figure 1. Equipment used and Specifications.

Volunteers are recruited in an office environment where working from home is common. A mass email was used to encourage them to participate. The requested condition was to work from home at least twice a week and the commitment of each participant is to:

• Monitor their home for 15 days, where they will work at least twice a week
• Answer daily point-in-time environmental comfort surveys mostly while working from home. The survey is not completed when the respondent indicates that they are away from home, such as weekends or office days
• Answer a household characterization survey, only one is necessary if participants live in the same household.
• Being interviewed in person with an individual report containing their results in hand, only one is necessary if participants live in the same household. In this final interview, participants are asked how many times they have been ill during the winter, where they feel more comfortable at the office or home and other specific aspects depending on the results.

The available equipment for this study is:

• One COMET U3430 temperature, relative humidity and CO₂ datalogger recording data every 2-min (To be placed in the home-office room).
• Two Elitech RC-5 or RC-5+ temperature dataloggers recording data every 2-min (to be placed in the bedroom and the living room).
• OR
• One Elitech RC-5 or RC-5+ temperature datalogger recording data every 2-min is delivered if the bedroom or living room is used as the home-office room.

The data loggers are set static away from direct light (windows or direct lamp radiation) or heat (radiators or electronic equipment). Figure 1 displays the sensors’ and their technical specifications.

Three different monitoring campaigns were performed:

• First period: From 22 November to 5 December 2021 (eight households—nine participants)
• Second period: From 17 to 30 January 2022 (eight households—eight participants)
• Third period: From 21 February to 6 March 2022 (ten households—thirteen participants)

2.1. Analysis

Some errors from the survey files are corrected or filtered (answers to the environmental survey outside the period scope, incorrect user location inside the home…). The dates when the environmental comfort surveys are answered at home are matched with point-in-time monitored values. Data from the equipment is filtered to the scoping period of analysis. After this initial filtering, the data shown in the results can be selected from all the sample, or with different criteria of filtering.

In order to ease the understanding of this paper, the following terms are defined:

Non-filtered:Are all values monitored during the 14 days. These can be referred to as the home-office space containing temperature, relative humidity and CO₂ or other rooms with only temperature.

Point-in-time: Is referred to a single value matched with the exact time of answering a survey. Since CO₂ and relative humidity are monitored only in one room, this value is considered the same for all the houses. Nevertheless, for temperature, it is also taken into account in which room the participant is placed inside the house.

Filtered: A filter is applied to delete periods when the home-office space has different uses from working (Considered as non-occupancy periods). This filtered data only refers to the home-office space and contains temperature, relative humidity and CO₂. The criteria when home-office space is empty are:

• On weekends.
• From 22 h to 8 h.
• When work was held in the office headquarters.

Linear fitting is performed with the point-in-time values and the different corresponding answers in the survey. The variables chosen for the fitting are:

• Predicted Mean Vote (Fanger model) vs. Thermal sensation.
• Adaptive categories vs. Thermal sensation.
• Operative Temperature vs. Thermal sensation.
• Relative humidity vs. Humidity sensation.
• CO₂ vs. Air-quality perception.
• Operative temperature vs. Clothing factor.

2.2. Indicators

The main indicators analyzed are the operative temperatures, Indoor Environmental Quality (IEQ) classification with two different thermal comfort models (Fanger and Adaptive), CO₂ concentrations (according to IEQ thresholds), temperature satisfaction and acceptance, as well as the difference in temperature between rooms.

Operative temperature (T_op, °C) is the thermal comfort index used in the Adaptive model, which relates acceptable temperature ranges to weather conditions. This Adaptive comfort method is applied in occupant-controlled naturally conditioned spaces. This temperature depends on indoor air velocity, indoor air temperature and radiant mean temperature. Indoor air temperature is the only parameter measured in the households, therefore, and according to standard ASHARE 55 [23], if air velocity is below 0.2 m/s, it is assumed that the operative temperature is equal to the air temperature monitored.

Running mean outdoor temperature (T_o,rm, °C) is an indicator of the trend of ambient temperature used in the Adaptive comfort method. An hourly value is calculated with the 24 h mean of the 7 previous days according to CEN/TR 16798-2 [24].

$$\begin{array}{c}{T}_{o,rm}{}_d^h=(T_{outdoor}{}_{d-1}^h+0.8·T_{outdoor}{}_{d-2}^h+0.6·T_{outdoor}{}_{d-3}^h+0.5·T_{outdoor}{}_{d-4}^h+0.4· T_{outdoor}{}_{d-5}^h+\\ 0.3·T_{outdoor}{}_{d-6}^h+0.2· T_{outdoor}{}_{d-7}^h)/3.8\end{array}$$

(1)

where d corresponds to the day being considered, h is the hour considered to calculate the mean of the previous 24 h and T_outdoor is the daily mean outdoor air temperature.

Table 2. Comfort range for the different comfort indexes: Operative temperature, CO₂ concentration and HR levels [13].

Comfort Range	Level of Expectation	Top [°C]	Predicted Mean Vote	CO2 1 [ppm]	HR [%]
IEQI	High. Occupants with special needs (children, elderly, persons with disabilities, etc.).	$T_{op}=0.33·T_{o,rm}+18.8+2$ 2 $T_{op}=0.33·T_{o,rm}+18.8-3$	−0.2 < PMV < +0.2	550	30–50
IEQII	Medium. Standard level.	$T_{op}=0.33·T_{o,rm}+18.8+3$ $T_{op}=0.33·T_{o,rm}+18.8-4$	−0.5 < PMV < +0.5	800	25–60
IEQIII	Moderate. It will not provide any health risk but may decrease comfort.	$T_{op}=0.33·T_{o,rm}+18.8+4$ $T_{op}=0.33·T_{o,rm}+18.8-5$	−0.7 < PMV < +0.7	1350	20–70
IEQIV	Low. Acceptable only for very short periods throughout the year.	-	−1.0 < PMV < +1.0	>1350	0–20 & 70–100

¹ Corresponding CO₂ concentration above outdoors levels. ² Upper and lower limits of comfort range levels.

details the operative comfort ranges depending on the Indoor Environmental Quality (IEQ) category, being IEQ_II for a recently constructed residential building. The equations are valid for T_o,rm (weighted running mean of the daily mean outdoor air temperature) between 10 and 30 °C, taking the T_op, (T_o,rm = 10 °C) when the T_o,rm < 10 °C and Top (T_o,rm = 30 °C) when the T_o,rm > 30 °C.

CO₂ concentration (CO₂, ppm) is used as a tracer of human occupancy and allows us to determine if the household has appropriate ventilation rates to guarantee acceptable indoor air quality. The comfort ranges are introduced in Table 2 and correspond to the CO₂ concentration above outdoors conditions. The outdoor CO₂ concentration used in the present work is 416.45 ppm, 2021 annual mean given by the ESRL’s Global Monitoring Laboratory (GML) of the National Oceanic and Atmospheric Administration in Mauna Loa [25], which is considered the least polluted location in the planet.

The Predicted Mean Vote (PMV) is an indicator calculated based on six variables: air velocity, air temperature, mean radiant temperature, relative humidity, clothing and metabolism rate). Since air velocity is not monitored, an estimated value of 0.1 m/s is used. As for the air temperature and mean radiant temperature, the same assumption as in the Adaptive model is done (they are equal). PMV is calculated using a python library called pythermalcomfort [26] which uses the same formula described in UNE-EN ISO 7730:2006 [27]. Comfort ranges are described in Table 2. The clothing and metabolism rates are considered with the clo and met factors, respectively. They are determined with the answer on the survey and quantified in

Table 3. Contribution to clo factor related to answers in the survey.

Answer in Survey	Contribution to Clo Factor
-	0.03
Short-sleeved T-shirt	0.15
Long-sleeved T-shirt	0.25
Short trousers	0.06
Long trousers	0.25
Undershirt	0.1
Sweater	0.28
Short Skirt	0.25
Long skirt	0.26
Jacket	0.35
Dress	0.4
Socks or light panties	0.1
Open shoe or sandal	0.02
Closed shoe	0.03

and

Table 4. Met factor according to answers in the survey.

Answer in Survey	Contribution to Met Factor
Low activity (resting, sitting)	1
Medium activity (cleaning, walking)	2
Intense activity (sports)	3

according to UNE-EN ISO 7730:2006 [27].

The temperature difference between the home-office room and other rooms is calculated with the mean temperature of both spaces. It is calculated in households where more than one participant is found,

$$\Delta T_{household}=\frac{{\sum }_{i=1}^n{T}_{office,i}-T_{othe{r}_{room},i}}{n}$$

(2)

where $T_{office,i}$ is the temperature monitored every 2 min inthe home-office room, $T_{otherroom,i}$ the equivalent for the other room and n is the number of monitored values in 14 days.

Table 5. Main glossary.

Term	Definition
General comfort perception	Temperature perception in winter and summer, humidity, illumination, noise and air quality in a scale from 1 to 7. One answer is collected by household
Thermal perception (point-in-time environmental comfort)	Scale perception from 1 to 7 being 1 Acceptable an 7 Not Acceptable for a precise moment while working at home (in winter season)
Thermal sensation (point-in-time environmental comfort)	Scale sensation from −3 to 3 being −3 Very Cold and 3 Very Hot for a precise moment while working at home (in winter season)
Operative temperature (Top, °C)	Uniform temperature of an imaginary black enclosure, and the air within it, in which an occupant would exchange the same amount of heat by radiation plus convection as in the actual nonuniform environment; calculated in accordance with ASHRAE 55 [23]
Running mean outdoor temperature (To,rm, °C)	Exponentially weighted, running mean of a sequence of mean daily outdoor temperatures prior to the day in question [23]
Adaptive model	Thermal model which relates acceptable temperature ranges to weather conditions (Top and To,rm,)
CO2 concentration (CO2, ppm)	Amount of CO2 in the room in parts per million, and is used as a tracer of human occupancy
Predicted Mean Vote (PMV)	Empirical fit to the human sensation of thermal comfort. Scale from −3 to 3, being 0 neutral.
Fanger model	Thermal model which uses the PMV as a key indicator
Clothing factor (clo)	Clothing is defined in terms of clo units. Clo is a unit used to express the thermal insulation provided by garments and clothing ensembles, where 1 clo = 0.155 (m2*K/W) [23]. In this study, a relation between the answer in the survey and clo units is found (Table 3)
Point-in-time	Values are monitored or surveyed, in a precise moment in time.
Filtered	Refers to the home-office space and contains temperature, relative humidity and CO2. The criteria when home-office space is empty are: On weekends From 22 h to 8 h When work is held in the office headquarters
Non-filtered	All values monitored during 14 days

sums up the crucial terms already described in this section to understand the results in Section 3.

3. Results

3.1. Households and Participants’ Characteristics

The household characterization surveys have been answered in a 100% of the cases.

Table 6. Household characterization of the sample monitored.

Household	Construction Year	Location	N° Participants/Self-Reported Gender	N° Inhabitants	Household Typology	Principal Heating System	Setpoint Temperature?
1	2008	Gavà	1 W	2	Flatmates or single-person (two adult sisters)	Natural Gas Bolier/ Radiators	No
2	1960	Barcelona	1 M	1	Flatmates or single-person	Natural Gas Bolier/ Radiators	No
3	1900	Sabadell	1 M	1	Flatmates or single-person	Air-source Heat Pump	24 °C cold season, 22 °C mild season
4	1933	Sabadell	1 M	2	Couple without children	Natural Gas Bolier/ Radiators	No
5	1965	Barcelona	2 M (5_2),W (5_1)	2	Couple without children	Electric radiators (and HP not used)	No
6	1936	Barcelona	1 W	4	Family (couple and children)	Natural Gas Bolier/ Radiators	20 °C
7	1965	Badalona	1 M	2	Couple without children	Air-source Heat Pump	24 °C
8	2006	Barberà del Vallès	1 W	3	Family (one parent and children)	Natural Gas Bolier/ Radiators	20–21 °C daytime, 19 °C nighttime
9	1910	Barcelona	1 M	4	Flatmates or single-person	Natural Gas Bolier/ Radiators	18 °C
10	2007	Granollers	1 M	1	Flatmates or single-person	Natural Gas Bolier/ Radiators	20 °C (but variable)
11	1980	Barcelona	1 W	4	Family (couple and children)	Natural Gas Bolier/ Radiators	18 °C
12	1990	Terrassa	1 M	4	Family (couple and children)	Natural Gas Bolier/ Radiators	19.5–20.5 °C
13	1969	Orrius	1 W	2	Couple without children (Woman pregnant)	Other (Fireplace, mobile gas radiator and a heat pump)	No
14	1988	Barcelona	1 W	3	Family (couple and children)	Natural Gas Bolier/ Radiators	No
15	1967	Barcelona	1 W	2	Couple without children	Air-source Heat Pump	No
16	1963	Barcelona	1 M	3	Couple + another family member	Electric radiators	No
17	1890	Barcelona	1 M	3	Flatmates or single-person	Natural Gas Bolier/ Radiators	18 °C
18	2017	Barcelona	2 W(18_1),W(18_2)	2	Flatmates or single-person	Geothermal heat pump with radiating walls	No
19	2017	Barcelona	1 M	1	Flatmates or single-person		No
20	2017	Barcelona	2 M(20_2),W(20_1)	2	Couple without children		No
21	Before 1980	Barcelona	2 M(21_1),W(21_2)	2	Couple without children	Electric radiators	No
22	1936	Barcelona	1 M	3	Flatmates or single-person	Air-source Heat Pump	No
23	1910	Barcelona	1 M	4	Flatmates or single-person	No heating system, just one small electric radiator	No
24	1929	Barcelona	1 M	2	Couple without children	Air-source Heat Pump	No
25	2001	Barcelona	1 M	2	Flatmates or single-person	Air-source Heat Pump	No
26	1978	Barcelona	1 W	4	Flatmates or single-person	Electric radiators	No

collects summarized information about the monitored households. The variety of profiles between the twenty-six monitored households represents a random sample of the stock of Mediterranean residential buildings: 62% is built previous to 1980, 19% between 1980 and 2006 and 19% after 2007; The 42% is inhabited by two occupants, mainly couples, although 27% of these households have no family nucleus. Households with one, three and four occupants, represent 16%, 19% and 23% of households in the sample, respectively. Mostly, three and four-occupant households are composed of family members. In these households, there are a total of nine inhabitants younger than seventeen years old (14% of total inhabitants in all households). Globally speaking, there are 38% of inhabitants between eighteen and twenty-nine years old, and 48% of inhabitants between thirty and sixty-four years old, the vast majority are younger than fifty. Counting all the households, there is 57% self-reported men and 43% self-reported women.

When it comes to the heating system, only one household reports not having any heating system. Among the ones that do have one, 44% uses natural gas boilers and fixed radiators, 16% uses a heat pump with air distribution, 20% uses mobile electric radiators, 12% has geothermal heat pump with radiating walls and an additional electric radiation and the rest use more than one type of heating system. Households 18, 19 and 20 are placed in the same building, which is a cooperative housingcalled La Balma built in 2017 where inhabitants have been living for less than a year.

When it comes to health, only one inhabitant in household 16 reports to have a Poor self-perceived health (being the scale Excellent/Very Good/Good/Poor/Bad). The rest answer Excellent (29 people), Very Good (23), Good (7) or blank (5). As for other health indicators, in general, participants feel happy, calm, active, wake up with energy and feel motivated by their daily life more than half of the time. In the final interview, it is reported to not have been sick, or to have had one or two mild colds with low or no medication. No trend is found between the household characteristics and the health state, due to the size of the sample.

Figure 2 shows general comfort perception at home in winter, taking into account temperature perception, humidity, illumination, noise and air quality (only one answer is collected per household). In general, there is satisfaction at all levels. Figure 2 reflects that thermal perception is clearly affected by the household’s efficiency. Newer buildings are generally perceived as more comfortable, in terms of thermal comfort, humidity and air quality. However, this does not occur with noise and light perception.

In addition, participants are asked to compare comfortability at the office or at home, taking into consideration the above-mentioned factors. Participants in the same household give one mutually consented answer, so only one answer per household is given. Eleven households report being more comfortable at the office (42%), six at home (23%), and nine report that they are indifferent or it depends on other factors such asseason or ergonomy (35%).

Point-in-time environmental comfort surveys have been answered at a mean rate of 1.6 answers per day and participant, or 1.8 answer per day and household. Household 11 counts with the largest number of surveys answered while at home (24 surveys answered), and household 16 has the lesser responses (only one). The participant in household 16 reported to always have similar perceptions in every aspect analyzed (temperature, humidity, air quality, noise and light) while working from home. Nevertheless, the monitoring manifests a notable range of variation, especially in temperature and CO₂ levels. Since not all participants work from an exclusively dedicated office space, the answer to the location within the house has led to some misunderstandings that have been corrected manually.

Figure 3 plots the votes distribution for point-in-time acceptance in terms of noise and natural illumination during remote work performance. The noise general perception answered in the household characterization average mark is 2.8 (Figure 2) while the average of the point-in-time acceptance is 1.7 (Figure 3), i.e people tend to have more acceptance to noise in a daily basis. Noise and natural illumination are qualified between 1 and 2 more than 50% of the times. Households that repeatedly answer non-acceptable noise levels (votes 4 and 5 more than 3 times) are households 7, 9, 18, 23 and 26. Ten households vote more than 3 times a poor natural illumination quality. Household 7 reports the noise to be due to the neighbors and the others because of the street noise. The votes for poor natural illumination of five out of ten households (18, 20, 21, 24 and 26) are attributable to the clouds strong irruption during the 3rd period of monitoring, when during three days solar irradiance did not even reach 6 MJ/m². Clear sky days or with few clouds exceed 15 MJ/m² during the same period, according to the ObservatoriFabra [28]. This fact is contrasted with the mean for households 9, 15 and 20, which contribute to the top ranking votes 4.4, 3.6 and 3.1, respectively, in terms of natural illumination.

3.2. CO₂ Concentrations

Figure 4 plots the CO₂ concentration levels in the home office space, which averagely are below 1770 ppm. This means that the space is found inside IEQ_III, considering occupied and non-occupied periods. When considering all the data monitored, households spending more time in IEQ_IV in terms of CO₂, are 16, 22 and 26, spending 51%, 46% and 34% of the time, respectively. Households 22 and 26 not only use the monitored space as a homeoffice but also as a bedroom. At large, when the home-office space is used as a bedroom at night, its outcomes havea wider distribution in CO₂ concentration. Even when nights are filtered (from 22 to 8 h), households 22 and 16 maintain a similar percentage of time in IEQ_IV time (54% and 40%); however, household 26 drops to 16% (Figure 4). Contrastingwith these data, household 26 reports not to be renovating the air daily and house 22 does confirm daily ventilation.

Ten households of the sample are always in IEQ_III or better, from which seven use the office-space as a living or dining room. In addition, households 3 and 25 have the best air quality of the sample, with values over 98% of the time in IEQ_I and never falling into IEQ_III. In fact, the eleven households of the sample that use a common area of the house during their working period are more than 95% of the time in IEQ_III or better. The only exception is household 13, which is heated with a fireplace and a mobile gas radiator. The comparison of average values are compared filtering nights, weekends and days in the office show that lower CO₂ concentrations are found in spaces used also as a living or dining room, having an average difference of 470 ppm approximately.

The analysis according the periods shows that the third period (21 Feb–6 Mar) presented higher air renovations, when there are milder outdoor temperatures. In this period, only two households (22 and 26), are less than 50% of the time in the IEQ_I quality range at the home-office space. As for the first and second periods of monitoring, similar levels of ventilation are detected and, in general, CO₂ levels are inside IEQ_III, according to ISO EN16798 [13]. In these periods, only households 5 and 16 are more than 30% of the time out of quality ranges, thus, they are studied in detail in this section.

No clear pattern is detected in terms of relative humidity even when separating data according to monitoring periods. The cases with highest and lowest average point-in-time relative humidity are household 24 (70%) and household 13 (13%), respectively. As for perception votes (Figure 5), the most answered options are neutral, slightly humid and slightly dry, only 3 answers out of the total surveys differ from these options. The maximum and minimum mean votes are found in households 16 (mean vote 1.0 slightly humid), 15 (0.9), 24 (0.9) and 11 (−0.8 neutral/slightly dry). Household 11 occurs to be the second driest (36% mean during point-in-time interviews which would be inside IEQ_I) and household 24 ends to be the most humid (70%, IEQ_III). In general, perception is neutral or slightly humid, summing up in an average mean of 0.

The general mean perception of air quality is rated with 2.15 out of a scale from 1 (fresh air) to 5 (dense air). The 22% of the votes is 1, 46% is 2, 25% is 3 and 7% is 4. The worstmean votes in air quality are found in households 9 (mean vote 3.6), 24 (3.5) and 18 (3.3). However, their average levels of CO₂ at the point-in-time of the surveys are 1043 (in IEQ_II), 1541 (in IEQ_III) and 720 (in IEQ_I), respectively.

Since households 16, 5 and 22 have higher levels of CO₂ they are considered singular in terms of the use of the space and the monitoring period, hence, they studied closer. Household 22 is reported to have air renovation daily during the monitoring period. However, CO₂ concentrations maintain high. The household is placed in a noisy street with dense traffic in the surrounding area. Consequently, a random week in spring is monitored and displayed to the occupant to find an explanation to this elevated levels of CO₂. The user is asked to air the vast majority of the time and in the week monitored it is found that CO₂ levels descend to 400–450 ppm, which are standard exterior CO₂ levels. It is concluded that during the winter there was less ventilation than reported, probably due to noise or cold. Household 5 uses the space exclusively as a home-office. This fact contrasts with CO₂ levels descending during weekends to values around 700–1500 ppm while nights stay in a higher range (Figure 6). The participant working in that space admits having a small radiator next to the desk, revealing that the deficient ventilation is due to thermal reasons. On the contrary, in household 16 the monitored space is also used as a home-office, but levels seem to vary considerable between the first and the second week. Participant in household 16 is asked about this fact and assures that the occupation has not varied during the monitoring period, and reckons a similar ventilation behavior. Participants in both households mean vote is between 2.4 and 1, so their perception does not indicate a discomfort in terms of air quality.

3.3. Operative Temperatures

Figure 7 plots the monitored temperatures in the home-office space for each household, filtering weekends, nights and headquarters-office days. When dividing results according to the usage of the space, it is found that home-office spaces used as living or dining room are in average 0.8 °C warmer. Families with children or teenagers tend to have a warmer mean operative temperature, being the cases of households 12, 8, 11, 14 and 6 (filled in green in Figure 7). Furthermore, they all have a fixed heating system that consists of water radiators powered by natural gas boilers, and except for 14, they all have a setpoint temperature between 18 and 22 °C. Households 11, 14 and 6 have the highest mean operative temperatures (above 20 °C) and have female participants. Households 8, 11, 12 and 13 own the house where they are living. Only two more subjects in the whole sample own the house. Male participants tend to have lower temperatures in the home-office space in comparison to female participants. There are no remarkable differences when comparing between periods in terms of the median operative temperature.

Having already removed non-occupational periods, households with wider operative temperature distributions are those who have major differences between the average of filtered and non-filtered temperatures (households 2, 3, 4 and 5).

A correlation between the filtered and non-filtered mean values is plotted in Figure 8a. In general, the mean temperature does not vary notably after deleting values within the filter criteria described in the methodology. Distributions of temperature are narrower in the third monitoring period. Some households from the first period of monitoring (22 November–5 December) are clearly separated from the trend line, especially those with a wider distribution. Three groups are distinguished: low, medium and high temperature. Low and high temperatures contain six households and are those with a filtered mean temperature below 18 °C and those above 20 °C, respectively. The medium temperature group works averagely in temperatures between 18 and 20 °C. The low-temperature group, specifically the four households with worse results, is inhabited by male participants. When interviewed, participants from households 17 and 24 assured to be used to low temperatures at home, whereas participants in households 14, 13 and 6, who are all women, reported in the interview to be sensitive to cold. Other households in the low-temperature group in the high temperature group, attribute the temperatures to the energy efficiency of the building or report having low sensitivity to cold temperatures. The middle group is larger and there is a heterogeneous variety of households and participants, so it can be considered that the work temperatures at home would normally range between 18 and 20 °C. Similar results are found when plotting only point-in-time temperature measures (Figure 8b), common values are between 18 and 20 °C (total average is 19.0 °C).

Temperature difference between rooms in the house is relevant when there is more than one participant in the household (which is the case for 5, 18, 20 and 21). It must be said that participant 2 in household 5 is working in the living room, which is around 1 °C hotter than the home-office space in average considering all data. However, in the other households, the temperature difference between rooms is less than 0.5 °C, being the home-office space usually the hotter one This value is taken into account when applying the comfort models and when analyzing the thermal comfort perception.

3.4. Temperature Perception and Sensation

Figure 9 contains the thermal perception and sensation vote acceptance while the point-in-time environmental comfort survey is answered. There are two scales for this vote: acceptable to not acceptable (scale from 1 to 7) and very cold to very hot (scale from −3 to 3). Generally speaking, participants are satisfied with an average temperature of 19 °C (the average vote is 2), and the individual mean is always inferior to 4 (middle value), i.e., all participants find their household temperature fairly acceptable while working at home. Also, the sensation is reasonably neutral, the total mean vote being −0.42 which corresponds to IEQ_II according to the PMV thresholds [13]. There are no extreme votes, the marks are always between 2 and −2. Temperature acceptance in the first monitoring period (22 November–5 December) is stricter when voting. More than 50% of the vote in that period are 3 or above, while in the other periods 50% are 2 or below.

Households with lower temperatures are found in the upper third of the most non-acceptable voting (Figure 9c). The most unsatisfied participants are found in households 5, 7, 8 and 17. The first three share a monitoring period and have a mean operative temperature in the home-office space superior to 19 °C. Most households with mean operative temperature below 17.5 °C (households 24, 17, 23 and 1) vote 3 or above and feel it is cool or cold more than 50% of the times. However, this is not the case for household 16 and participant 2 in household 21, which in average work around or below 17.5 °C and vote below 2 over 75% of the time. Households with a mean operative temperature over 20.5 °C (households 6, 7, 13 and 14), do not have in general a better thermal perception. Most satisfied participants live in households 4, 10, 12 and 18. All of those have average operative temperatures between 18 °C and 20 °C.

The mismatch between temperatures and perception can be explained by multiple factors. For example, participants in households 1, 2 and 24 reported to have direct sun radiation at their desks and participants in households 5, 9 and 13 have direct radiation from the heating system. Other relevant factors that are collected from interviews are hormone periods (households 13 and 14 have pregnant participants) and a larger acceptance to cold temperatures than the standards.

Participants in the same households do not have the same perception. However, this is not attributed to the temperature difference between the working rooms. In fact, participant 2 in household 5 finds temperature acceptable 14% of the time despite being averagely 1 °C above participant 1, who finds temperature acceptable 88% of the time. Similar to this case, in household 18, despite having similar temperature conditions, acceptance is 40% for participant 1 and 100% for participant 2.

3.5. Thermal Comfort Models

Two comfort models have been applied: Adaptive and Fanger models. Both are applied with point-in-time data, i.e.,the temperature measured while answering the environmental comfort survey and with all monitored values, without applying any filter to visualize major differences. Figure 10 plots the percentage of values in each category using the Adaptive model and PMV [13]. Results show that the third monitoring period (21 February–6 March) is much worse in the ranges and criteria established by the standard. For example, when contrasting this data with the operative temperature in Figure 7, it is found that households 18 and 19 have average temperatures between 21 and 19 °C, similar to households 8 or 7. However, household 8 is 66% in IEQ_I whereas household 18 is only 17% and household 7 is 55% in IEQ_I and household 19 is only 25%. Thus, one explanation for the lower indoor environmental quality ranges can be attributed to milder temperatures in the thirdperiod of monitoring.

Adaptive and Fanger models have differing results. With the Fanger model, no household is less than 40% of the monitored time in IEQ_IV. In contrast with the Adaptive model instead, there are households for almost 100% of the time without filtering in IEQ_I (households 6 and 11) using all data (without filtering). Therefore, the Fanger model is a much stricter criterion than the Adaptive. Figure 11 relates the percentage of time in categories IEQ_I and IEQ_II for the point-in-time models (Figure 10a,b) with the acceptable and neutral votes (Figure 9c). Since the member of household 16 answered the survey only once, the data was deleted from the sample. The difference between models is noticeable, showing that the Adaptive model is closer to real perception than the one of Fanger. In both models, it is found that participants accept worse thermal conditions than foreseen by the theory. Only four participants (5_2, 7, 8 and 11) are more dissatisfied than what supposed by the Adaptive model, and none is more dissatisfied than the results of Fanger’s model. Four other participants (6, 12, 14 and 18_1) are always in IEQ_I or IEQ_II and have a neutral or acceptable perception. Additionally, the Adaptive model is sensitive to outdoor conditions: higher outdoor temperatures lead to an increase inindoor temperature thresholds. This is why it manifests less agreement to the surveyed perception during the thirdmonitoring period (21 February–6 March).

The PMV of the Fanger model is sensitive to factors such as clothing, metabolic rate activity (collected in the survey), radiant temperature, air temperature, air velocity and relative humidity, exhibiting that the assumptions made may have affected the results. Some participants (4, 9, 15 and 26) reported (in the interview after the monitoring) wearing more clothes than those answered in the survey, such as blankets or robes, which were not available in the response options. Therefore, a closer focus onthe clo factor is taken.

3.6. Clo Factor

Clothing is introduced inside the PMV formula employing the clo factor quantified in the methodology in Table 3. Typical clothing levels for winter (1.0 clo) are compared to total mean values in surveys (0.8 clo). Figure 12 contains the clo factor for all participants and the average value for men and women. Participants with a mean clo factor lower than 0.7 (24, 20_2, 23, 5_2, 21_2 and 16) are all male participants and are all, except for one (21_2), in the group of less satisfied thermally speaking (Figure 9). Participants with a mean clo factor higher than 0.9 (18_1, 18_2, 15, 21_1, 4 and 6) are all female except for one (4). Those are spread out both in terms of temperature and perception vote. Average clothing has a 0.11clo difference between genders.

Comparing this result to another large-scale study that analyses [5] the relationship between IEQ parameters and occupants’ satisfaction in 400 offices from 20 different buildings in the US; it is found that female workers in summer were significantly less satisfied with their thermal environment than male workers. According to the authors, the different clothing insulation was the main factor for these observations. Thus, the results are consistent in terms of the clothing factor and level of satisfaction.

Another study [29] finds that clo values worn by office workers on casual days show a significant correlation with outdoor temperatures (R² = 0.44). However, in this study, no relation is found between the clo factor and the indoor temperature (R² = 0.06) and results do not depend on the monitoring period, hence, in outdoor temperatures.

3.7. Correlation between Votes and Measurements

Linear fitting is applied to correlate direct votes to point-in-time measures.

Table 7. Linear fitting results for direct votes in surveys (Y) and point-in-time measurements (X).

X	Y	Data Correlated	R2
PMV	Thermal sensation	All values	0.03
		1st period monitoring	0.07
		2nd period monitoring	0.00
		3rd period monitoring	0.08
		Natural gas boiler and water radiators	0.00
		Heat Pump	0.02
		Electric radiators	0.00
		Geothermal heat pump with radiating walls or no heating system	0.06
		Built before 2000	0.03
		Built after 2000	0.60
Adaptive model categories (IEQI, IEQII, IEQIII, IEQIV)	Thermal sensation	All values	0.03
Operative Temperature	Thermal sensation	All values	0.04
Relative humidity	Humidity sensation	All values	0.29
		Without outliers (4, 5_1, 12, 14, 15 and 20_2)	0.42
CO2	Air quality perception	All values	0.00

contains the R-squared results of the linear fitting.

According to the comfort models, participants should be highly dissatisfied with the indoor temperatures. However, this is not the case. For the correlation of PMV and direct sensation vote, three different taxonomies are chosen to visualize factors contributing to a punctuation difference: monitoring period, heating system and year of construction of household. The compared vote asks about the thermal perception on a −3 to 3 scale (Very cold/cold/Slightly cool/Neutral/Slightly warm/Hot/Very hot). In general, there is a low correlation, between those two variables, taking into account monitoring periods and the heating system. Slightly better results are found when segregating by year of construction. This is consistent since inhabitants of newer houses have less acceptance in comparison to the other participants, and except for household 1, they have mean filtered temperatures higher than 19 °C.

Operative temperature and votes are also fitted linearly. Both all monitored mean temperatures and point-in-time measures when surveying are used and related to the survey vote. R-squared values are inferior to 0.1, concluding a low correlation between temperature and thermal perception vote survey. A fair correlation is found between the mean vote and mean humidity: if six out of 30 participants considered as outliers are filtered from the sample (Participants in households 4, 5_1, 12, 14, 15 and 20_2), the R-squared of the linear fitting is 0.42. No correlation is found between CO₂ measures and perceived air quality.

4. Discussion

After analyzing the monitoring data, average temperatures while working at home range around 18 °C to 20 °C; these are found generally acceptable according to the participant’s opinion. The standard acceptable temperatures in offices [13] range from 20 to 24 °C which would indicate that people working from home accept lower temperatures in winter than those established in the regulations. It must be taken into account, that several participants reported having direct sun radiation at their desks or direct radiation from the heating system, which indicates that the local point temperature is higher than the one monitored by the sensor. The temperature range widens from 16 to 21.5 °C if all temperature working conditions are considered. In households between 19 °C and 21 °C (ninehouseholds), fiveof the households are composed by family members. People who live in households with average temperatures below 18 °C are mostly male participants and those with higher temperatures are female participants. This fact is consistent with an EU study [30], where up to 3 °C of differences between women and men were found in terms of thermal comfort. This difference is most evident in offices where women show a better cognitive performance in a warmer environment, while men do better in colder temperatures.

When a common space such as a living or dining room is used as a home-office, better results are found in terms of operative temperature and CO₂ levels. In common spaces, it is 0.8 °C warmer and there are 470 ppm lower CO₂ concentrations on average. Generally, levels of CO₂ while working at home are below 1700 ppm (IEQ_III). A study in Denmark [8], also finds that living rooms have the most homogeneous temperature distribution during a 24 h period and the bedroom hasthe most significant variations. Another study simulating window operation behavior foundbedrooms to have around 1.000 ppm higher concentration levels than living rooms when ventilation is not performed and in a lockdown scenario [31].

Theoretical models (Fanger and Adaptive) differ from real vote perception when appliedto individuals in point-in-time surveys. This is consistent with other studies, which show these models have low predictive performance [9,32], and also that perception depends on many other factors [33]. A low correlation is found between direct votes in surveys to point-in-time measures, especially comparing the PMV, temperatures and CO₂. However, for buildings constructed more recently, the similarity between the PMV and direct vote is closer and for colder periods, the Adaptive also performs better. A fair correlation between perception and relative humidity is found in most subjects of the sample.

The levels of clothing have been surveyed in working-at-home conditions and are found slightly under the standards criteria (0.8 clo vs. 1.0 clo), indicating the use of lighter clothes while working at home. However, it is found that participants with lower values of clo, who are mainly men, are more dissatisfied in terms of comfort according to the answers in surveys. Comparing this result to another study in the US [5] it is found that female workers were significantly less satisfied with their thermal environment than male workers due to clothing insulation. At universities in the UK [30], heavier clothing insulation optimum acceptable worn by women (≈0.92 clo) than men (≈0.83 clo) and the higher optimum acceptable temperature of females (23.5 °C) than males (22.0 °C) supports the warmer thermal requirements of women compared to men.

This research is limited in terms of equipment and time. On the one hand, the equipment available monitors only a few parameters to evaluate the indoor environmental quality. For example, it does not monitor radiant temperature or air velocity. In addition, until mid-February, there were eight dataloggers U3430 available and two more were bought afterwards, limiting the number of households monitored in each period. It was preferred to reach a greater number of households, rather than longer periods and the same dates between households. It was intended to find a variety of home typologies; however, participants are found mostly between 25 and 50 years old, because the scope of this research is remote working profiles. Their participation was dependent on their engagement with the project and some households have resuted to have little participation. Furthermore, some parameters such as clothing or the heating systems are asked in the surveys but not checked with specialists or devices.

5. Conclusions

In this paper, 30 participants from 26 households representing a random sample of the stock of Mediterranean residential buildings were monitored and surveyed in remote working conditions. Temperature, CO₂ and Relative Humidity levels were monitored at their main home-office space. Through point-in-time surveys, workers have reported their daily comfort in provision for its comparison with the monitored data.

In general, indoor environmental quality perception is voted as comfortable (in terms of temperature, humidity, air quality, noise and light). Thermal perception was better ranked in buildings that are more efficient. Noise was found on a daily basis acceptable, except for some households that reported street or neighbors interference. Light perception was also found acceptable, except for the last period of monitoring, when the weather was rather cloudy. Eleven households report to be more comfortable at the office (42%), six at home (23%), and nine report that it is indifferent or depends on other factors like season or ergonomics (35%). So in general, there is a tendency to feel more comfortable at the office.

CO₂ concentration levels in the home office space are averagely below 1770 ppm. Common areas utilized as home offices have lower CO₂ concentrations (in average 470 ppm less than bedrooms or other rooms used only as offices) and are typically 0.8 °C warmer. The majority of workers found temperatures between 18 and 20 °C acceptable when working from home. Families with kids or teenagers, particularly women, have a tendency to have a warmer mean operating temperature. The Adaptive model outperformed the Fanger model in terms of direct voting perception prediction when both were applied in various working environments and settings. All those results have been contrasted with other studies. In general, people claim to feel more comfortable at work, yet a sizable portion of participants express indifference or believe that it relies on other variables.

Detailed analysis has been performed concerning the clothing factor and it is found that in average women tend to wear more clothes than men, although values when working at home are slightly below the standard. Additional analysis is fulfilled to find a correlation between perceptions and monitored values, but no significant results are found, noting that perceptions depend on multiple variables.

Further research must be addressed in households of the same typology but in summer season. Also, monitoring in offices where the same people work would be ideal to complement the findings. Other profiles or periods may be considered, such as students or longer-term occupancy periods (for example a whole year). In addition, the influence of other factors on acceptable temperature, such as ergonomics or controllability, should be explored and taken into account in further studies.