Chapter 2 Person based evaluation of quality of life, exposure to noise and air pollution
2.3 Personal activity indicator framework
2.3.2 Framework components
2.3.2.1 Spatiotemporal objects and the exposome
In the introduction (1.1.2 and 1.1.5), it is shown that many solutions focus on the activity as the key object in a personal time-activity pattern. The basic reporting level of a methodology to calculate personal exposure is the person object. A person’s time activity pattern is a se-quence of activities. The person’s whereabouts can be described as a so-called “spatiotem-poral object”, a changing location in time with a set of activity specific features. The instan-taneous position of the subject is a coordinate and a time (position, time). The daily personal time-activity can be visualized as a line in a three dimensional space with coordinates (x,y,t).
The temporal resolution used to evaluate the activity can change according to the needs and requirements of the research question at hand. Within this PhD, activities at fixed loca-tion are evaluated at a temporal resoluloca-tion of 5 minutes. The in-traffic activities are evaluat-ed at a temporal resolution of 10 seconds. The environmental burdens themselves also change in space and time and are spatiotemporal objects as well. They can be visualized as two-dimensional surfaces changing in time. Those surfaces do not have to exist as such. The only requirement is that they can be calculated for each relevant time and position when required in the personal spatiotemporal object.
The indicator is the evaluation of the environmental burdens for a specific time activity pattern. The indicator calculation is based on the intersection of the two spatiotemporal objects for person and burden. The eco-exposome for a personal activity pattern is in es-sence the evaluation of the person’s activity pattern for all environmental burdens (see Fig-ure 2.3.1).
Figure 2.3.1: Exposure, dose and indicators are a combination of two spatiotemporal object, the first describ-ing the behaviour of the person, the second the behaviour of the environmental burdens. The result is the personal exposure, in general terms referred to as the ‘exposome’.
In an object oriented development, all components will become objects, attributes or methods. An object represents a specific entity of information and has attributes and meth-ods. Attributes organize the relevant information about the object. Methods are actions that can be performed on the objects and alter objects and their attributes. In the next sections, the objects for the exposome compatible indicator model are presented. The dataflow framework will be referred to as MEX, modelling the exposure/exposome.
2.3.2.2 Person object structure
A person is a multi-layered object. The top-level person object includes all personal at-tributes and contains one or more activity patterns objects. An activity pattern includes matching attributes and a set of activity objects. The activity is the central object, includes matching attributes but has four standard attributes: the time/episode, the purpose, the micro-environment and the location or route. The standard attributes contain the basic in-formation to evaluate the activity for the environmental burdens. The combination of the purpose and micro-environment is referred to as the ‘activity typology’ and will act as the link between the activity object and the indicator object (see 2.3.2.4). For all person related objects, project specific attributes can be defined, including any relevant information from the personal input data useful for future indicator calculations. At the level of the activity, attributes can be single values or time series objects. When an activity object is created, the location attribute is by default a time series containing a sequence of (x,y,t), defining the position of the person in time. Other attributes with time series include any relevant ubiqui-tous information. This will be explained in detail in 2.3.3.2.
Figure 2.3.2: Persona objects structure and examples of the objects attribution.
When an activity object is created and evaluated for a specific indicator (see 2.3.2.6), it results in a temporary spatiotemporal activity object (STA) which contains all information gathered on the person, activity pattern and activity with a temporal resolution matching the type of the activity and the project specifications.
2.3.2.3 The activity purpose and micro-environment
The activity typology organizes the indicator calculations and is a crucial concept. The purpose of the activity is a basic attribute of an activity in mobility research and activity
modelling. It defines why the activity takes place. The set of values can be defined project specific, but it is good practice to use a generalized set of purposes. The typical set is:
“work”, “school”, “shop”, “recreation”, “social”, “service”, “bring and get” (Beckx et al., 2009, Beckx et al., 2013).
The micro-environment of the activity is a basic attribute in exposure modelling. The ex-posure dynamics depend strongly on the micro-environment (1.1.5.2). The set of values can be defined by project. An attribute of the environment object groups the micro-environments into indoor and outdoor microenvironment. Similar to the purpose, a general-ized but customizable set is defined: “Home-indoor”, “Home-outdoor“, “Office”, “School”,
“Work” (non-office work locations), “Walk”, “Bike”, “Car”, “Rail”, “Light Rail”, “Bus”, “Met-ro”, “In-transit” (in between modal choices), etc.
2.3.2.4 Indicator object structure
The indicator object contains all relevant definitions to calculate the indicator and is re-ferred to as the “Indicator definition”. The evaluation of an indicator is by design sensitive to the properties of the activity. The indicator definition contains a set of objects each defining the behaviour of the indicator for a specific type of activity, referred to as an ‘Activity Specif-ic Model’ (ASM). The activity is not only the central object in the description of the person, but it defines the behaviour of the indicator definition. The first action when evaluating an activity is dispatching the STA to the matching ASM (Figure 2.3.3). For example, if the indica-tor is the exposure to BC, the ASMs will describe the specific exposure dynamics for the dif-ferent micro-environments. The activity will in this case, be dispatched to the ASM matching its micro-environment attribute. Dispatching can be organized by any ‘activity typology’.
More sensitive subdivisions can be organized within the ASMs.
Figure 2.3.3: Step 1 in the indicator definition is dispatching the spatiotemporal activity to the matching ac-tivity specific model.
The ASM object organizes all information necessary to actually calculate the indicator.
Gathering the information is a stepwise process, matching the logical flow of the infor-mation. It defines the necessary external data sources where information has to be retrieved from and defines the type of data retrieval. It also defines any function necessary to calcu-late any feature of the activity necessary to perform the next step in the stepwise process.
Each step is referred to as an ‘Activity Calculation’ (AC). An AC can use any piece of infor-mation available in the STA at that moment. This includes all personal parameters, time-activity pattern parameters, all time-activity parameters and all new attributes gathered in the stepwise process so far, at the predefined activity specific temporal resolution. The AC re-turns an extended STA for the following step. An important reason why the attribution of the activity is organized stepwise is the speed of the calculations. Getting a thousand points from a specific data source is much faster than triggering a thousand times the retrieval of a single point, which would be the case if the calculations are organized by the time step matching the temporal resolution. The stepwise process is illustrated in Figure 2.3.4.
The core of the ASM is the ‘activity specific function’ (ASF), the function that actually cal-culates the indicator. The ASF is triggered at the moment when all necessary attributes are available. The ASF is technically also an activity calculation but is added as a specific compo-nent for different reasons. First of all, it does not have to be a ‘classical’ analytical function. It can be any external function or procedure. The only technical limitation is the possibility to execute the function from within Python, the chosen implementation technology. In the ASMs for BC exposure presented in this PhD, the ASF is the prediction function of a “Gener-alized Additive Model” (GAM), triggered externally in the statistical open source R environ-ment. The ASF can also be a validated function from a scientific publication or external re-port. In that case, the scientific reference supports the ASM at full and the ACs are only gathering the input data. The outcome of the ASF is an exposure, a dose or an indicator. The three outcomes are available in parallel if relevant. More details will be given in 2.3.2.7. Ad-ditional ACs can organize the conversion from exposure to dose and/or indicator, again ca-pable of using any available data of the STA (not show in figure). These ACs can have scien-tific references as well and can bring additional parameters into the STA in the process.
Figure 2.3.4: Data workflow in activity specific model (ASM) and the Activity Specific Function (ASF).
2.3.2.5 Activity calculation types
The activity calculations gather the relevant external data. Within the spatiotemporal model-ling, a large variety of data sources are available. A short list of the functionalities is summa-rized. The STA has the position and time as basic features. The position is used to retrieve data from spatial grids, search for closest points or polylines, retrieve attributes, calculate distances and bearings and several other geographic operations. Keys and attributes can be retrieved from GIS layers. The time stamp of the STA can retrieve information from external time series. More complex spatiotemporal operations are available to retrieve data by com-bining time and position information: multiple GIS layers with time annotations, time series with position annotations. The implementation is extendible to include new data types if necessary.
2.3.2.6 MEX processing and low level reporting
The processing is visualized for a single indicator with only one activity specific model in Figure 2.3.5. The ‘person factory’ builds the personal objects (2.3.2.2) from external popula-tion data. The different oppopula-tions for the populapopula-tion data will be presented in 2.3.2.8.
Figure 2.3.5: The person factory creates the person, time-activity pattern and activity objects from the exter-nal population data. The Spatiotemporal Activity (STA) assembles all persoexter-nal data and triggers the indicator calculation. The applicable ASM contains a sequence of Activity Calculation retrieving the external data and performs other relevant actions on the data. This results in an attributed STA. The Activity Specific Function (ASF) is applied and results in an indicator. The red stars identify the stages in the process where data can be retrieved and reporting is initiated.
The calculation of the indicator is triggered at the level of the activity object since the in-dicator evaluation is activity based. The STA is created and the sequence of activity calcula-tions is applied to the STA. The outcome values (exposure, dose, and indicator) are calculat-ed in the ASF. The results exist at this stage with the same temporal resolution as the defini-tion in the applied ASM. The data is available for detailed analysis and contains all temporary results of all activity calculations (the full STA, including the outcomes of the ASF).
2.3.2.7 Personal indicator reporting
The indicator calculation is available at the temporal resolution of the ASM. At this stage the detailed results are aggregated to the different relevant levels of the personal objects. By default, the indicator is aggregated by activity, diurnal pattern and person. By default, the flagged outcomes (exposure, dose, and indicator) will be aggregated. The aggregation func-tion can be a customized funcfunc-tion matching the properties of the calculated indicator. Every other customized reporting level can be applied in a post-processing and reporting phase on the resulting data for the full population.
2.3.2.8 MEX projects and scenarios
A scenario is the combination of a person factory and a set of indicator definitions. Different scenarios can be compared and evaluate the effect of the changes in the person factory and/or the indicator definitions. A project is a set of scenarios built to answer a specific re-search question.
Feedback loops between changing population behaviour and the external data sources for the indicator calculations are not explicitly covered but can be achieved by implemented sequences of scenarios, combined with customized reporting. The population simulation results in a set of individuals travelling the network. Aggregating the data by network seg-ment results in an estimate of traffic on all segseg-ments. This is an example of a reporting ac-tion orthogonal to the person data for that specific scenario. This technique is used to im-prove the local traffic estimates in the Traffic Liveability model (see 3.2.3 and 3.2.4). Custom-ized reporting or expert decisions can be used to adjust or recalculate the external data sources and enable the implementation of feedbacks.
2.3.2.9 Applying indicator definition on other populations
The processing of the person data and the calculation of the indicator are at this point only connected by the temporary SPA object. The only restriction to apply an indicator definition on another population is the availability of the relevant input data:
Can the person factory deliver the SPA and the relevant personal attributes to match the indicator definition to be applied?
Is the external data to calculate the indicator available in the project area?
Is the available Activity Specific Function valid for the investigated episode, region or subpopulation?
If all answers are positive, the indicator can be calculated at once. If one of the answers is negative, additional expert decisions can result in an adjusted indicator definition or per-son factory to calculate a result with a reaper-sonable error and/or acceptable reduced sensi-tivity.