Finding the temperature people actually want

Before mechanical heating and ventilation the design of the thermal environment was largely decided by experience. The number of fireplaces to provide in a building, the way in which the rooms were shaded and ventilated were part of the skill which was passed from builder to builder, successful solutions becoming normal.

With the advent of modern air-handling and central heating systems the question of what conditions the system should be providing became critical. Because one of the major advances such systems afforded was control over the humidity, early attempts to lay down standards concentrated on finding combinations of temperature and humidity which best describe the feeling of warmth given by the environment.

One approach to finding what conditions are comfortable is to conduct field surveys. The experimenter then measures the physical characteristics of the environment and relates these to the subjects' feeling of warmth to find a relationship. Bedford's (1936) survey of factory workers is a classic of this type of survey.

Experimental work can also be carried out in a climate chamber. Climate chambers are laboratories which enable the experimenter to adjust the air and radiant temperature, humidity and air velocity. Such chambers have been widely used in controlled experiments investigating the effect of physical parameters on comfort.

These called the empirical approach and the analytical approach.

Empirical field surveys

In the field survey the method is to ask subjects taking part in the survey to assess their thermal sensation on a subjective scale. This assessment is generally known as the `Comfort Vote'. The environmental variables are measured at the same time as the subjective reactions are taken. The aim is to find a temperature or a range of temperatures and other environmental variables which people will find comfortable. Because the aim is to obtain a typical reaction to conditions in a particular locality there is no attempt to interfere with normal conditions or modes of dress. The full complexity of the situation is included in the responses of the subjects.

The first aim is to discover what combination of environmental variables best describes the subjective responses of the subjects. This is generally done using multiple regression analysis of comfort vote as a dependent variable against the independent variables of the physical environment. A number of such `comfort indices' have been put forward over the years. Another approach is to use probit analysis which enables the proportion of people comfortable at any particular or combination of variables to be calculated.

The underlying assumption of the field survey is that people are able to act as meters of their environment. This assumption is rooted in the findings of psychophysics. In effect the subject is used as a comfort meter, not of temperature alone but of all the environmental and social variables simultaneously. Only the effect of time is generally ignored in the analysis. The regression is performed of the comfort vote on the simultaneous environmental conditions: the possibility that the comfort relates more closely to conditions at some previous time, or on some more complex time-series of the environment is not usually taken in to account.

Problems with field surveys

The accepted method of analysis for field surveys is to use statistics. The comfort vote is taken as the dependent variable and the environmental measurements as the independent variables. At the same time people are left to suit themselves in their choice of clothing, their use of environmental controls, their posture, activity and so on. Many of these actions will have been taken in response to the comfort vote. So it is not true to say that the environment is independent of comfort vote. In addition by their changes of clothing, activity and posture people will also change their own characteristic responses to the environment - again on the basis of the comfort vote. Because of this the results obtained are very specific to the conditions measured. It also means that any formula resulting from the statistical process must be treated with extreme caution, and any such formula should be judged on physical as well as statistical grounds.

Nevertheless the field survey is the key to understanding thermal comfort. Any theoretical model which does not explain the results of measurements in the field among real people cannot be trusted to set standards which will have meaning among those same real people. Field study measurements have been criticised by laboratory workers for not reflecting theoretical findings. But surely the scientist must face the other way: the results of the field survey "are facts that need to be explained. It is no answer to say that they do cannot be true because they conflict with our theory" (Jay 1972).

Yet in order for the field results to have general applicability we have to produce general rules from the individual results. This is the basis upon which we can move forward. One such attempt was that of Humphreys (1976) of which more later.

Analytical Approaches

There is an obvious advantage to having a complete picture of the various thermal factors involved in man's interactions with the environment. A number of workers have set out to build models of the physical and physiological conditions governing thermal comfort, the best known are those of Fanger's (1970) Predicted Mean Vote (PMV) and Gagge's (1972) Standard Effective Temperature. Fanger's model forms the basis for ISO Standard 7730 which includes a computer programme for calculating PMV.

Fanger's basic premise is that a balance between the heat produced by the body and the heat lost from it is a necessary, but not a sufficient condition for thermal comfort. It is not sufficient because one can imagine situations in which a theoretical balance would occur, but which would not be considered comfortable. So the determination of comfort conditions is in two stages: first find the conditions for thermal balance and then determine which of the conditions so defined are consistent with comfort.

[refer to program for calculating PMV or to R deDear’s website]

Fanger proposed that the condition for thermal comfort is that the skin temperature and sweat secretion lies within narrow limits. Fanger obtained data from climate chamber experiments, in which sweat rate and skin temperature were measured on people who considered themselves comfortable at various metabolic rates. Fanger proposed that optimal conditions for thermal comfort were expressed by the regression line of skin temperature and sweat rate on metabolic rate in data from these experiments. In this way an expression for optimal thermal comfort can be deduced from the metabolic rate, clothing insulation and environmental conditions.

The final equation for optimal thermal comfort is fairly complex and need not concern us here. Fanger has solved the equations by computer and presented the results in the form of diagrams from which optimal comfort conditions can be read given a knowledge of metabolic rate and clothing insulation.

Predicted mean vote (PMV) and Predicted percentage dissatisfied (PPD).

Fanger extended the usefulness of his work by proposing a method by which the actual thermal sensation could be predicted. His assumption for this was that the sensation experienced by a person was a function of the physiological strain imposed on him by the environment. This he defined as "the difference between the internal heat production and the heat loss to the actual environment for a man kept at the comfort values for skin temperature and sweat production at the actual activity level" (Fanger 1970). He calculated this extra load for people involved in climate chamber experiments and plotted their comfort vote against it. Thus he was able to predict what comfort vote would arise from a given set of environmental conditions for a given clothing insulation and metabolic rate. Tables of PMV are available for different environments for given clothing and metabolic rates. Such tables form the basis of ISO standard 7730 Note however that his method for PMV is inconsistent with the basic assumptions of his equation (Humphreys and Nicol 1995).

Fanger realised that the vote predicted was only the mean value to be expected from a group of people, and he extended the PMV to predict the proportion of any population who will be dissatisfied with the environment. A person's dissatisfaction was defined in terms of their comfort vote. Those who vote outside the central three scaling points on the ASHRAE scale were counted as dissatisfied. PPD is defined in terms of the PMV, and adds no information to that already available in PMV. The distribution of PPD is based on observations from climate chamber experiments and not from field measurements.

The model for Standard Effective Temperature also uses skin temperature as part of it's limiting conditions, but uses skin wettedness (w) rather than sweat rate for the other limiting condition. The values for T_sk and w are derived from the Pierce `two-node' model of human physiology (see Nevins & Gagge (1972). SET relates the real conditions to the (effective) temperature in standard clothing and metabolic rate and 50% RH which would give the same physiological (??) response. Effective temperature can then be related to subjective response.

Problems with the analytical approach.

We have deliberately avoided a detailed description of the Fanger PMV because of a wish to keep this discussion simple. There are however a number of points which need to be noted.

1. the subjective data on which Fanger's model is based were obtained exclusively from climate chamber studies where a steady state had been reached when the subjects had been in constant conditions in the chamber for three hours.
2. prediction of conditions for optimal comfort, PMV or PPD require a knowledge of the clothing insulation and the metabolic rate
3. value of clothing insulation used is obtained by the practitioner from tables in which clothing insulation is listed against descriptions of items or ensembles of clothing. The values of clothing insulation have been determined in experiments using heated manikins (see appendix C).
4. metabolic rate is similarly obtained from tables of activities for which the appropriate metabolic rate is given (see appendix D).

For the environmental designer these characteristics of the Fanger model pose a number of problems.

• He must know what clothing the occupants of the building will wear.
• He must know what activity they will be engaged in and there is an additional problem for buildings where a number of activities are taking place in the same space.
• He must assume that conditions in the building approach those of the steady-state in the climate chamber.

All these factors will influence the designer towards a highly serviced building producing closely controlled internal conditions appropriate to some assumed clothing norm and activity.

Such considerations render the method very difficult to apply to buildings with no mechanical heating and ventilation. The temperature in a free-running building will almost certainly change continually with time, particularly if the inhabitants are able to control it to some extent. So to the difficulty in predicting clothing and metabolic rate is added the problem of applying a steady-state model to an intrinsically variable situation.

Differences between the results from empirical and analytical investigations

An additional problem has been found with the Fanger model. Some recent field surveys, in which clothing and activity descriptions were made for the subjects at the time of the survey, have shown the average values of PMV to be quite seriously different from the average comfort vote. Outside a band of temperature in the mid-twenties Celsius, PMV can be seriously in error. The slope of the line between temperature and average comfort vote is quite different from that between Temperature and PMV (see fig 3.1) and is in error in such a way as to overestimate the discomfort of the environment. This means that buildings heated according to accepted standards will be overheated, and those cooled will be overcooled. The evidence for this effect is as yet based on fairly slender evidence, but it is sufficient to cast serious doubts on the reliability of this method, which is, after all, currently used by the heating and ventilation industry internationally to set indoor temperatures.

Showing the bias in PMV in some recent field studies. The open circles are the mean subject vote in the surveys, the filled circles are the values are the mean PMV calculated from the environmental data collected at the same time.

The reason for the mismatch between empirical field studies and the Fanger model needs to be explored. It is remarkable that it has taken over twenty years for the problem to be fully recognised. The heat-balance model should be consistent with the empirical findings, which should be a subset of it where the particular social and environmental conditions embodied in it apply.

In addition to this slope effect, or following from it, there is a range effect. The range of average temperatures which people in different climates can find comfortable is much greater than could be explained in terms of differences in clothing alone.

Some possible sources of error in the assumptions underlying the analytical method suggest themselves:

• The tables of clothing insulation values assume that the insulation for a particular ensemble is independent of climate or culture. Systematic errors can occur due to

1. the interpretation of a particular description of a clothing ensemble
2. climatically-determined effects such as the wetting of the clothing by sweat from the skin and
3. the different ways in which clothing is actually used in different climates and cultures.

• Metabolic rate could also be affected by climate in a systematic way. People used to hot conditions undertaking a particular activity in a more economical way. Metabolic rates determined simply on the basis of an activity descriptor would not pick this effect up.
• Individual response to different environments depends on the order in which they happen, a mean vote predicted by a steady-state model would not reflect this.
• The way in which people react to an environment depends on their expectations of it based on experience. People presented with the same environment in a climate chamber have no such expectations or experience. As a result the people in the real situation will be comparing their actual sensations with their expected sensation, and not with some `absolute' scale as would climate chamber subjects.
• People will use environmental controls at their disposal to modify the environment to their liking. Given that there is a natural variation between people in the conditions they desirable, the conditions people choose will be closer to their desired temperature than would be some imposed standard. The overall effect of this will be to suit the conditions to the occupants, so reducing the discomfort as against prediction.

All these effect tend to reduce the variation in the comfort vote as against the PMV. All of them arise from the inability of the heat-balance model to take account of social and climatic factors that are included by field surveys. All of them work in the direction of the PMV overestimating the discomfort of any real situation. Which of the effects is most important, and whether there are other effects must be the subject of research and evaluation if comfort standards are to improve.

Humphreys and Nicol suggested a method of approach to these problems in their work in the 1970s (Humphreys and Nicol 1970, Nicol and Humphreys 1972, Humphreys 1976, 1978) which would allow these social problems to be incorporated into comfort standards. The approach they proposed has been called the adaptive approach (Humphreys and Nicol 1998)