«Sef van den Elshout October 2014 AQI & AQ Communication D2.3/v07 PartII Work Package 2, Task 2.2 Deliverable D2.3 Version number V07 Ver Date Author ...»
2.3. The Dutch AQI revision project In the Netherlands the existing smog warning is being overhauled to be replaced by a modern AQI (Dusseldorp et al. forthcoming). Health effects of individual pollutants are chosen as the basis. The AQI is constructed in such a way that it also takes into account the three issues mentioned in 2.1.
Instead of using a local study (as in Canada, Hong Kong) a WHO study for the whole of Europe was used to select the relative risks (WHO, 2013). For each pollutant the relative risk is used to align the pollutant classes (point c. in 2.1). In this approach, each AQI band, whether it is for
2.4. Common characteristics of recently revised AQI-s that are in use In this section we specifically look at indices that are actually used, assuming that the authorities that use them have made a deliberate choice between practical, communicational and conceptual characteristics of an AQI.
Short averaging time (1h, 3h): CAQI – EU, Canada, Hong Kong, the Netherlands and the UK introduced hourly ‘triggers’ for their moving average pollutants. (See annex A.1).
Health based, all the above mentioned except the CAQI;
All except the CAQI have paid attention to carefully worded behavioral advice, often even tested on the public if it is well understood.
Pollutant interaction Canada, Hong Kong, the Netherlands (though in a simplified way);
not in the UK and the CAQI. The ones that do use interaction acknowledge that due to lack of independence between the pollutants, this could be an overestimation of the risk. This is particularly true if both PM species are included (as in the Netherlands).
Highest class shouldn’t occur too often: all the above.
Explicit mention of which pollutants should be available to be able to calculate an index:
this is very important for the additive health indices but also for the UK index and the CAQI. Hong Kong decided to use PM10 in its index instead of PM2.5 as the two show high correlation but PM10 data is more available than PM2.5.
Consistency between the PM species: explicitly implemented in the grids in the UK and the CAQI, implicitly implemented by using the relative risk in the health based AQI-s.
Should the Chinese AQI get a stronger foundation in health as this seems to be a trend? In my opinion this is, for the time being not a good idea. Technically it is unnecessarily complicated.
Practically the current concentration levels are very high (unhealthy by most international standards - see comparison in document Part III) and even without a conceptually correct link to health effects, the public can easily be warned with an AQI with a more policy based grid.
Other lessons to be learned are more important: improve and test the wording of the recommendations; make the PM grids more consistent; avoid message fatigue by avoiding that the highest class occurs too often and use short averaging times. Considering the latter an important first step was taken in 2014.
3. AQI properties & communication objectives In chapter 1 the reasons to make an AQI and several design consequences were discussed. In section3.1 we briefly touch on some other design issues and in 3.2 two summary tables are presented linking AQI properties to communication objectives.
3.1. How to condense information AQI-s are about condensing complex concentration information of multiple pollutants in awkward units of measure, sometimes different averaging times, into (usually) one number on a relative scale. The scale is often accompanied by qualifying statements (good, bad, high, low, etc.). Usually this is visually reinforced with colours and or smiley-s.
Some AQI-s are only qualitative and numbers are not even mentioned (they are needed in the background however to decide to which category the measurements belong). Most have a limited number of broad classes (3 to 5) and subcategories in each class. Sometimes a numerical value is attached and the scale can run from 1 to 5 up to 1 to 500. The conversion of actual measurements to AQI numbers can be linear, non-linear or linear in each class (often the case). Garcia et al (2002) provide a non-exhaustive overview of various conversion methods.
One method is not necessarily better than the other. A point to consider is how the classes relate to the wording used and the health advice given. Remember that close to class borders a minor change of 1 µg/m3 can cause a change of class and hence another message. The risk perception of a situation can therefore dramatically change at an insignificant change of the air quality. That is the disadvantage of a discrete indexing system. The US EPA commissioned a study to analyse the risk perception of the messages when they changed their index (Johnson, 2003). This a good paper on risk communication and indices. The UK also studied the message in the course of their recent index review (COMEAP, 2011).
On the length of the scale (e.g. 1 to 5 or 1 to 500) different views exist. Some say the AQI is an enormous generalisation of information and in addition, the spatial coverage is not well defined so anything more than a few broad categories is overly pretentious and not backed by serious science. Elshout et al (2008) argue that to raise awareness and entice repeated visits to a website the information should be dynamic. This calls for a long scale that favours differentiation and changes in the course of the day and from one site to another.
The point to be stressed here is that one should be aware of the implicit message that colours 6 or numbers may send. This largely depends on local context and culture. For example the French ATMO runs from 1 to 10 with 10 designating the worst air pollution. However in school 10 is the highest grade one can obtain and there a 10 refers to a job very well done. Likewise in many countries red is associated with alerts and hence the colour is often used for the Colours also present a challenge for colour blind people. The (culturally) best colours for those with complete colour vision might not be the best for those with a degree of colour blindness. The use of numbers and smiley-s is therefore recommended. On index maps (see section 3.3) this is not possible. If these are envisaged as well, the colour coding merits attention right from the start.
highest AQI band referring to poor air quality. However, in China red symbolises good fortune and joy.
3.1. Linking AQI properties & communication objectives Many AQI properties have been discussed and this section aims to put them in the perspective of the communication objectives mentioned in chapter 3 of the part I document. Table 5 summarizes the points discussed in chapter 1 in relation to the communication objectives. The table shows that the choice for one type of index doesn’t mean that other uses are no longer possible (see for example figure 2). It is meant to underline that careful thinking about the main communication issues might lead to different design choices.
Figure2: Canadian messaging: an AQI aimed at risk communication combining awareness messages (what can you do yourself); Source: Stieb et al (2008).
3.2. Adapting behaviour?
Shooter and Brimblecombe (2009) mention ‘risk communication’ as the principal reason to have an air quality index. However they note, citing the study by Johnson (2003) on the US AQI, that AQI information rarely succeeds in changing people’s behaviour in face of the oncoming poor air quality. They argue that better forecasting, making a timelier alert might improve people’s response in the case that the index is signalling adverse air quality. This is obviously an improvement.
On the other hand, I doubt that this would considerably change the general public’s response to an alert. In large parts of Europe and the US the air quality has to be extremely poor before it poses an acute danger to the population in general. With busy agenda’s full of obligations it is unlikely that the average person will cancel his sports game, that a school outdoor activity that was planned long ago (involving many volunteers) is postponed at short notice to avoid exposure, that work is suspended, etc. Whilst in China the pollution levels are occasionally substantially higher, the possibilities to adjust one’s plans, even if warned a day or so in advance, is probably equally limited. This seriously questions the rationale of providing health and behavioural advice using an AQI via media with a broad audience like tv or the internet.
People/patients that are truly affected by air pollution that can and do take measures (reducing their exposure, taking additional medication) because they have to, could be warned in specific ways. By particularly addressing this more limited target group using specific communication Page 24 of 43 AQ Communication D2.3/V7 part II means, the general communication via websites and the AQI to a broader audience could be framed in the context of raising awareness/being accountable.
Current day technology facilitates these targeted ways to inform the select group of people concerned. For example AirAlert provides a messaging system that those in need of this kind of information can subscribe to (see www.airalert.info/Sussex/Default.aspx). This exists already several years and a very nice feature is that those without mobile phones (e.g. some of the elderly likely to be in the target group for this services) can get a voice message on their ordinary phone. The latest development is made possible by smartphone apps. Instead of a government deciding which pollution level is harmful or not, everyone can set his/her own alarm level based on experience or advice from the own physician. The recently launched Dutch air quality app (www.luchtkwaliteitmetingen.nl/ - in Dutch language only) is, as far as I know, the first to include this facility to personalise alarm levels in addition to general AQI based information.
4. Other AQI-products (year average, maps, etc)
4.1. Year average AQI-s.
All AQI-s discussed in this document so far refer to AQI-s with an hourly or daily averaging time and communication messages aimed at short term exposure situations. Year average AQI-s are rare though quite regularly the occurrence of the daily AQI is summarized and reported at the end of the year. This is a useful way of providing an annual overview, particularly if one or more AQI bands are related to standards. In this way one can indicate which percentage of the time the air quality met the set standards. For AQI-s aimed at health and behavioural advice this summary at the end of the year is less obvious (the AQI was made specifically for short-term exposure) but it is nevertheless informative to see which percentage of the time certain conditions occurred.
The main advantage of this approach (summarizing daily data) is that it ties short- and longterm exposure together in one presentation. The disadvantage is that the results are hard to interpret for lay people: instead of having one figure that indicates the status of the air quality one gets a percentage of occurrence per index class. See box 5 for an example. A possible solution to further simplify the message is to include in the presentation, the percentage of time the standards were met. In that case the main information collapses into a single figure.
A different approach is to develop a separate year average AQI, using for example the ‘distance to target’ principle, where the target is an air quality standard (local legislation, WHO recommendation, etc.). For each pollutant the year average concentration is divided by the standard or target value. A value 1 means that the standards are not yet met. The CITEAIR project proposed such an index (see box 5). The advantage is that it is very easy to see if the air quality is improving over time and if standards are met. The disadvantage is that due to the averaging a pollutant that is doing well compensates pollutants that still exceed the standard.
A careful selection of critical pollutants is needed! A second disadvantage are standards that are based on discrete occurrences rather than year average concentrations: e.g. an ozone concentration that should not occur more than x times a year (as in the EU legislation). These are computationally more complicated. Solutions exist of course but the easy to understand transparency – the very purpose of making an index – can easily be compromised.
The US EPA website (www.epa.gov/airdata/ad_rep_aqi.html) lets you generate year average AQI summaries. Below an excerpt from the 2012 report. Often the year average summaries contain the statistics on the occurrence of each band (the grey columns). The US EPA provides several additional statistics such as the median AQI, and the number of days each pollutant dominated the AQI. The interpretation difficulties of the core table start when one wants to compare the situation in two cities, or in the same city over time.
Look at the top 2 cities: the first has more good days, but also 10 unhealthy days: which city is better? The AQI median value can provide some guidance: it is the same for both cities. Similar interpretation problems occur with the other three stations: without the median AQI it would be difficult to judge whether one city is better than another or not.
The Year Average CAQI (Elshout, 2012, Elshout and Léger 2007) uses the distance to target approach. Each pollutant is divided by the year average standard and when the result is = 1 the standard is met. For the core pollutants the average is calculated (“city index”). The presentation provides both an easy average ranking (city index) as well as the details per pollutant. It is easy to see if air quality is improving, which pollutants are doing well, and which not. As with the hour by hour CAQI, the calculation is done for traffic and city background monitoring sites.