Technical ReportPDF Available

Nairobi Air Quality Monitoring Sensor Network Report - April 2017

Abstract and Figures

Improving air quality is essential for achieving the Sustainable Development Goals (SDGs), particularly goals related to health (Goal 3) and cities and human settlements (Goal 11). UNEA-1 resolution 1/7 strengthens the role of UN Environment in promoting and improving air quality. In this regard, affordable air quality monitoring based on the use of novel sensor technology can be used for cost-effective evaluation of air quality and related health impacts. All major air pollutants can be measured in near real-time, making such monitoring affordable for countries that have limited or no air quality monitoring networks in place. In April 2016, a demonstration network was deployed at six sites in Nairobi, Kenya, to pilot the affordable network approach. While a cluster of affordable monitoring devices are a fraction of the cost of a single reference unit, they are designed for indicative measurements and are not a replacement for reference units.
Content may be subject to copyright.
I
II
TABLE OF CONTENTS
TABLE OF CONTENTS ............................................................................................................ II
LIST OF TABLES ......................................................................................................................III
LIST OF FIGURES ................................................................................................................... IV
LIST OF ABBREVIATIONS ....................................................................................................... V
1. INTRODUCTION ......................................................................................................... 1
1.1 Background ................................................................................................................. 1
1.2 National Context .......................................................................................................... 1
Air Quality situation in Kenya ....................................................................................... 1
1.3 Context of Nairobi County ............................................................................................ 3
1.4 Project Objectives ........................................................................................................ 3
2. ASSESSMENT METHODOLOGY ............................................................................... 4
The Nairobi air quality monitoring network ................................................................... 4
2.1 Air Quality Monitoring Unit specifications ..................................................................... 4
Limitations of sensor based approach ......................................................................... 5
Findings from Nairobi demonstration network .............................................................. 7
2.2 Network deployment .................................................................................................... 8
2.3 Results ........................................................................................................................ 9
2.4 Case Study: Nairobi Placemaking Week.....................................................................14
3. EXPOSURE AND HEALTH ASSESSMENT ..............................................................19
3.1 Pollution episodes ......................................................................................................19
3.2 Health impacts of air pollution in Nairobi from long-term exposure..............................23
Uncertainties ..............................................................................................................25
Quantifying the health impact in Nairobi .....................................................................25
REFERENCES .........................................................................................................................27
III
LIST OF TABLES
Table 1: Comparison of Air Quality standards .............................................................................................. 2
Table 2: Benefits and drawbacks of sensor based solutions for air quality monitoring ................................ 7
Table 3: Site details ....................................................................................................................................... 8
Table 4: Summary of the results from sites in the Nairobi network for the time period June to November
2016 .................................................................................................................................................... 13
Table 5: Air pollution health impact assessment tools (Adapted from Anenberg et al. 2015) .................... 24
Table 6: Mortality data from Nairobi County ............................................................................................... 26
IV
LIST OF FIGURES
Figure 1: Map of sites showing administrative boundary of Nairobi County (map data © 2016 Google
Earth) ..................................................................................................................................................... 5
Figure 2: Air quality monitoring units deployed on site at (a) All Saints Cathedral Primary, (b) St.
Scholastica Catholic School and (c) Kibera Girls Soccer Academy ..................................................... 6
Figure 3: Bivariate polar plot for PM2.5 and time series showing daily means for PM2.5 and PM10
measurements obtained during the 6 month deployment at Kibera Girls Soccer Academy ................ 9
Figure 4: Bivariate polar plot for PM2.5 and time series showing daily means for PM2.5 and PM10
measurements obtained during the 6 month deployment at Viwandani Informal Settlements ........... 10
Figure 5: Bivariate polar plot for PM2.5 and time series showing daily means for PM2.5 and PM10
measurements obtained during the 6 month deployment at St. Scholastica Catholic School. .......... 10
Figure 6: Time variation plot, averaged over different time scales, of hourly PM2.5 concentrations at St.
Scholastica Catholic School. ............................................................................................................... 11
Figure 7: Bivariate polar plot for PM2.5 and time series showing daily means for PM2.5 and PM10
measurements obtained during the 6 month deployment at UN Environment Headquarters in Gigiri
............................................................................................................................................................ 12
Figure 8: Bivariate polar plot and time series showing daily means for PM2.5 and PM10 measurements
obtained during the 6 month deployment at All Saints Cathedral Primary School ............................. 12
Figure 9: Bivariate polar plot for PM2.5 and time series showing daily means for PM2.5 and PM10
measurements obtained during the 6 month deployment at Alliance Girls High School .................... 13
Figure 10: (a) Moktar Daddah street during the placemaking week, (b) and (c) Muindi Mbingu street
during the placemaking week .............................................................................................................. 14
Figure 11: Bivariate polar plot of PM2.5 (Top) and PM10 (Bottom) showing hourly mean concentrations
before, during and after the intervention ............................................................................................. 16
Figure 12: Diurnal patterns for PM2.5 (Top) and PM10 (Bottom) ............................................................... 17
Figure 13: Bivariate polar plot comparison between average conditions (left) and pollution episodes
conditions (right) for PM2.5 at Kibera ................................................................................................... 20
Figure 14: CBPF plots for 10 intervals of PM2.5 hourly average concentrations in Kibera for June
November 2016 ................................................................................................................................... 20
Figure 15: Bivariate polar plot comparison between average conditions (left) and pollution episodes
conditions (right) for PM2.5 at Viwandani ............................................................................................. 21
Figure 16: CBPF plots for 10 intervals of PM2.5 hourly average concentrations in Viwandani for June
November 2016 ................................................................................................................................... 22
Figure 17: Bivariate polar plot comparison between average conditions (left) and pollution episodes
conditions (right) for PM2.5 at St. Scholastica ...................................................................................... 22
Figure 18: Bivariate polar plot comparison between average conditions (left) and pollution episodes
conditions (right) for PM2.5 at Alliance Girls High School. ................................................................... 23
Figure 19: AirQ+ results for PM2.5 long-term adult mortality ....................................................................... 26
Figure 20: Relative risk of long-term exposure to PM2.5 in Nairobi ............................................................. 26
V
LIST OF ABBREVIATIONS
EMCA
Environmental Management and Coordination Act
EU
European Union
GDP
Gross Domestic Product
GoK
Government of Kenya
GPRS
General Packet Radio Service
ITCZ
Inter-Tropical Convergence Zone
NEMA
National Environment Management Authority
NIUPLAN
Nairobi Integrated Urban Development Plan
SDGs
Sustainable Development Goals
UN
The United Nations
UNEA
United Nations Environment Assembly
USEPA
United States Environment Protection Agency
WHO
World Health Organisation
1
1. INTRODUCTION
1.1 Background
Improving air quality is essential for achieving the Sustainable Development Goals (SDGs),
particularly the goals related to health (Goal 3) and cities and human settlements (Goal 11).
UNEA-1 resolution 1/7 strengthens the role of UN Environment in promoting and improving air
quality.
In this regard, UN Environment has promoted the use of affordable air quality monitoring
technology aimed at cost-effective evaluation of air quality and related health impacts, based on
the use of novel sensor technology. All major air pollutants can be measured in real time,
making such monitoring affordable for countries that have limited or no air quality monitoring
networks in place.
In April 2016, a demonstration network was deployed at six sites in Nairobi to pilot the
affordable network approach.
1.2 National Context
The republic of Kenya lies on the eastern part of Africa within latitudes North and South
and longitudes 34º and 41º East. National climate and weather conditions are dominantly
controlled by the Inter-Tropical Convergence Zone (ITCZ), monsoon wind systems, and the
Inland lakes which provide local sources of moisture.
Air Quality situation in Kenya
The Constitution of Kenya, Article 42, Chapter 4 - the Bill of Rights - confers to every person
resident in Kenya, the right to a clean and healthy environment (GoK, 2010). The Environmental
Management and Coordination Acts (EMCA) is the main law guiding pollution prevention. The
National Environment Management Authority (NEMA), established under EMCA, is the
government organ that exercises general supervision and co-ordination over all matters relating
to the environment.
National Ambient air quality regulations exist and are referred to as “The Environmental
Management and Coordination (Air Quality) Regulations, 2014”. The original Air Quality
Regulations, 2008, did not set any air quality guidelines but instead proposed their formulation.
This was updated in 2014 to include limit values. Guidelines for Assessment of Air Quality in the
2014 regulations seek to, among other things, ‘establish source contributions to ambient
concentrations of pollutants’ and ‘assess the environmental benefit of measures to reduce and
maintain air quality within limit values’. Whereas there has been significant progress in
formulating guidelines, no national air quality monitoring program is in place.
Air Quality monitoring is limited to urban areas where ad hoc monitoring is done in response to
air pollution complaints and short-time research campaigns initiated by academic institutions.
2
Pollutant
Averaging
Time
EMCA (Air Quality) Regulations 2014
WHO
Guidelines
2005
EU
2008/50/EC
Residential, Rural &
Other Area
Controlled
Areas
Oxides of Sulphur
Annual
60
15
-
-
SOx (μg/m3)
24-hour
80
30
20
125
Oxides of Nitrogen
Annual
60
15
-
-
NOx (μg/m3)
24-hour
80
30
-
-
Nitrogen Dioxide
Annual
-
-
40
40
NO2 (μg/m3)
24-hour
-
-
-
-
Carbon Monoxide CO
(mg/m3)
8-hour
2
1
10
10
Ozone O3 (μg/m3)
1-hour
-
-
-
-
8-hour
-
-
100
120
Lead (μg/m3)
Annual
0.75
0.5
0.5
0.5
24-hour
1.0
0.75
-
-
PM10 (μg/m3)
Annual
50
50
20
40
24-hour
100
75
50
50
PM2.5 (μg/m3)
Annual
-
-
10
25
24-hour
-
-
25
-
Table 1: Comparison of Air Quality standards
3
1.3 Context of Nairobi County
The city of Nairobi is the capital and serves as a center of administration, politics, economy and
culture within its administrative area of approximately 692 km2. Nairobi city accounts for 50 % of
formal employment in Kenya and generates over 50 % of GDP (NIUPLAN, 2014). According to
Kenya Population and Housing Census conducted in 2009, the total population of Nairobi City
was approximately 3,138,386, accounting for 8.1% of the national population. It is estimated that
Nairobi has around 4 million inhabitants today.
The quality of ambient air in the City of Nairobi has deteriorated over the last decade. The city’s
population has continued to increase at the rate of ten per cent per annum for the last five
decades - growing from one million in 1970. More than 60% of city’s population lives in informal
settlements, a source of cheap labor for factories, commercial and domestic services.
The main sources of atmospheric pollution are vehicles, industries, emissions from use of
charcoal and firewood, and other municipal sources such as open burning of waste. Increasing
number of cars in the city intensifies traffic and pollution problems. Exhaust emission from these
vehicles is a cocktail of noxious gases carbon monoxide (CO), nitrogen oxides (NOx), sulphur
oxides (SOx) and particulate matter, including black carbon and resuspended dust.
This report presents an assessment of urban air quality in Nairobi and estimates of the health
effects of short term exposure to elevated pollutant levels, indicative of the increasing
incidences of respiratory diseases recorded in the various health facilities.
1.4 Project Objectives
At the 2014 inaugural UN Environment Assembly (UNEA), air pollution was identified to be a top
priority to be addressed by the international community. UN Environment was mandated to help
governments set standards and policies across multiple sectors to reduce emissions, and
manage the negative impacts of air pollution on health, the economy and sustainable
development. More specifically, UNEA resolution 1/7 is about strengthening the role of UN
Environment in promoting air quality.
Reliable data that can enable the identification of pollution hot spots and local source
apportionment of air pollution is required in order to set air quality standards and policies which
enable action to be taken to tackle the root causes of bad air quality.
The main objective of this demonstration project was to show how countries can be supported
to create the necessary enabling environment for initiating and thereafter sustaining affordable
air quality monitoring and reporting.
4
2. ASSESSMENT METHODOLOGY
Two main methods can be used for air quality assessment:
Ambient and source air quality measurements;
Air quality modelling combined with emission inventories;
Assessment of air quality based on fixed measurements is the recommended assessment
technique for areas where concentrations exceed the set limit values as is the case in cities
such as Nairobi. Data from fixed measurements can be supplemented by results from emission
inventories combined with modelling techniques and/or indicative measurements.
Continuous ambient air quality monitoring provides information regarding the status of air quality
and helps evaluate existing policies - their effective implementation, or lack thereof.
For this assessment air quality data from six months of continuous measurements were used.
The major uncertainties in ambient concentration measurements presented in this report are
influenced by lack of an Air Quality reference monitoring station in Kenya for field calibration, as
discussed below.
The Nairobi air quality monitoring network
In May 2016, a network of 6 affordable air quality monitoring units was deployed in the city of
Nairobi (Figure 1). The network provided measurements of nitric oxide (NO), nitrogen dioxide
(NO2), sulphur dioxide (SO2), particulate matter (PM1, PM2.5 and PM10) as well as temperature
and relative humidity at high temporal resolution (1 minute).
To account for stabilisation of the electrochemical sensors within the surrounding ambient
conditions, the initial 25 days of the measurement campaign were excluded from further
analysis. This assessment covers a six month period from June to November 2016.
2.1 Air Quality Monitoring Unit specifications
Air Quality monitoring units, hereafter referred to as nodes, were deployed as a network at
designated sites (Figure 2) with each measuring NO, NO2, SO2, PM1, PM2.5, PM10, temperature,
relative humidity and ambient noise.
Node specifications are:
Sensors - A laser Optical Particle Counter (Alphasense OPC-N2) to provide PM1, PM2.5
and PM10 measurements, three amperometric electrochemical gas sensors to measure
NO, NO2 and SO2 (Alphasense Ltd UK models NO-A4, NO2-A43F and SO2-A4),
temperature and relative humidity sensors;
Communication - linked to a remote database via GPRS. Node uses a data-capable
SIM card service to send data;
Data Storage - onboard 16GB SD Card to act as a backup of results sent over the
GPRS link;
GPS module capable of accurate location and time measurement and a Microphone to
characterise ambient noise;
Li-ion 12 volt 7Ah back-up battery and weather screen primarily for rain protection.
5
Figure 1: Map of sites showing administrative boundary of Nairobi County (map data © 2016 Google Earth)
The nodes are able to ‘store and forward’ data in times when the local cellular network is down
or temporarily overloaded. The current of each electrochemical gas sensor was measured every
20 seconds and converted to a voltage (via an Analogue-To-Digital-Converter). The raw data
were then transmitted every hour. OPC data was sampled every minute to maximise battery life
during power outages.
The transmission process induces electrical interference on the sensor signals, thus the initial
recordings were filtered out prior to analysis. The data were then converted into mixing ratios for
gas phase measurements using pre-defined, sensor-specific sensitivity (laboratory calibration)
factors provided by the sensor manufacturer.
Limitations of sensor based approach
There are no internationally agreed procedures available to quantify the performance of sensors
or sensor systems. The European Standardisation Committee (CEN) and the United States
Environmental Protection Agency (USEPA) are sharing knowledge to devise a series of
practical performance tests which are a few years from being adopted in a technical
specification or standard method. As a result, it is not yet possible to follow agreed procedures
to identify measurement accuracy, precision or uncertainty for sensors and sensor systems.
6
Figure 2: Air quality monitoring units deployed on site at (a) All Saints Cathedral Primary, (b) St. Scholastica Catholic
School and (c) Kibera Girls Soccer Academy
Despite this, as an emerging technology, sensor systems have enormous potential. If initial
performance and ongoing quality control procedures can be reliably quantified, there are a vast
number of possible applications in (for example) epidemiology, health, population exposure,
emissions management and modelling. It is therefore extremely important to devise suitably
robust quality assurance and quality control regimes to get the most value from this technology.
A number of key benefits and limitations are summarized below.
7
Benefits
Limitations
+ Low cost
+ Low power
+ Fast response time (t90 = 20s)
+ Multi-species analysis
+ Portable
+ High temporal resolution
+ Ease of use, no need for technical
operator
+ Ability to cost-effectively deploy sensor
networks over a large area
- Sensor life expectancy
- Accuracy and precision of sensors
- Calibration, characterization and co-
location requirements
- Interference with environmental artifacts
(e.g temperature, humidity)
- Cross-sensitivity to other gases
- Sensor stability over time (seasonal
variation)
Table 2: Benefits and drawbacks of sensor based solutions for air quality monitoring
Findings from Nairobi demonstration network
Of the 6 units deployed, 1 had technical problems that resulted in reduced data coverage (<2
months). To ensure that the network data are fit for purpose, the following tasks need to be
considered.
In quality assurance:
Real world performance evaluation of sensors, using reference measurements for
comparison. These are tests that would allow sensor responses to be quantified at a
local level, but feasibility will depend on local expertise and availability of a reference
station. At the very least, a network of sensor nodes should be operated at a single
location (co-location study) prior to deployment to assess sensor performance in a true
ambient monitoring environment and determine accuracy and precision of processed
measurements.
Identification of the precise monitoring locations. Careful selection to ensure station
representativeness is important. Ideally, a mixture of traffic, industrial and urban
background locations should be chosen to provide data that can be used to estimate
population exposure.
Procedures for knowledge transfer. Local expertise should be used to organise and
implement the co-location of nodes for network validation. Training of a local academic
institution should be considered, to take on the roles of data management and data
processing. A process of certification and/or ongoing accreditation and assessment
could be devised to provide confidence that the institution is following the required
procedures.
8
In quality control:
Adaptive frequency of comparisons, based on developing knowledge. Due to the nature
of measurements with sensors, operation of a network of nodes with no degree of cross-
referencing or intercomparison is not recommended. Performance evaluations should be
undertaken every three months, to quantify and correct for any seasonal variations in
temperature and relative humidity as well as possible decay in sensor response. This
frequency of comparison is set to ensure a wide range of meteorological conditions is
satisfied and any drift in sensor responses is promptly identified. Once a history of
comparisons is available, the frequency can be reviewed to see if it is possible to extend
the time between comparison exercises.
Control of system calculation algorithms. The process of turning sensor electrode
outputs into usable datasets is very complex and time consuming. This expertise, or at
least the data management aspect in a more automated process, needs to be cascaded
in a controlled manner to trained and accredited academic institutions in-country.
As familiarity and improved international guidance emerges, protocols for cross-
referencing and intercomparison of sensor nodes will be valuable in determining data
quality.
Additional nodes used as co-location transfer standards have significant limitations.
They use the same technology as the nodes, so bias error is likely. However, the co-
location of nodes is essential if precision of network measurements is to be determined.
2.2 Network deployment
The nodes were wall mounted, 3 m (the breathing zone) above the ground. Five of the nodes
were deployed within the urban environment where higher variability of pollution levels was
expected. Site names and characteristics are summarized below.
Site
No.
Name
Type
Data
Capture
Description
1
Kibera Girls Soccer
Academy
Informal
settlement
100%
Site located to measure typical
concentrations in an area of high
population density
2
Viwandani Informal
Settlements
Industrial/
Informal
settlement
32%
Site located to determine the impact of
a significant source (industry) on air
quality
3
St. Scholastica Catholic
School
Traffic
(Highway)
100%
Site located to determine the impact of
a significant source (traffic) on air
quality
4
UNEP Headquarters,
Gigiri
Sub urban
100%
‘Clean’ site
5
All Saints Cathedral
Primary School
Urban
100%
Site located to determine the impact of
a significant sources on air quality
6
Alliance Girls High
School
Urban
background
83%
Sites located to determine general
background concentration levels
Table 3: Site details
9
2.3 Results
The capability of the nodes for continuous monitoring over several months as parts of a network
in an urban environment was successfully demonstrated. The sensor network was used to
illustrate the range of conditions encountered in the urban setting associated with varying
emissions. Distinct local emission features were evident across the city.
The bivariate polar plot was used to show local source signals of pollutants and indicate long-
range transport. These plots show the relationship between three variables: pollution data
(particulate matter measurements), wind speed and wind direction.
To the furthest extent possible, pollution sources were identified. However, the details apparent
to these sources are beyond the scope of this assessment. Data analysis was done using the
OpenAir open source air quality analysis tool (OpenAir version 1.8-2 on R version 3.3.0).
For purposes of health assessment measurements of PM2.5 and PM10 are reviewed. The optical
particle counter (OPC-N2) calculates the respective concentrations according to the method
defined by European Standard EN 481 (AlphaSense OPC-N2 Manual), converting particle size
count to particle number concentrations. It should be noted that environmental temperature and
relative humidity (T/RH) corrections on particle count to particle concentration conversions have
not been applied.
In Kibera (informal settlement) data capture was 100% for the six month period. The polar plot
(Figure 3) shows evidence of a source to the south-south-east.
Figure 3: Bivariate polar plot for PM2.5 and time series showing daily means for PM2.5 and PM10 measurements
obtained during the 6 month deployment at Kibera Girls Soccer Academy
Daily means were significantly higher compared to the 24-hour WHO air quality guidelines
(black dotted line) for both PM2.5 and PM10 in the first four months (June, July, August and
September). Lower daily mean concentrations were recorded in October and November which
coincides with the short rains season, a trend observed as well in other locations across the
network. Kibera was the only site where on certain days, the 24-hour limit value for PM10 (100
μg/m3) set out in EMCA (Air Quality) Regulations, 2014 (green dotted line) was exceeded.
10
At Viwandani data capture was low (less than 2 months) occasioned by node technical issues.
The polar plot (Figure 4) shows significant sources to the north-west, west and south-west
associated with emissions from industries nearby.
Figure 4: Bivariate polar plot for PM2.5 and time series showing daily means for PM2.5 and PM10 measurements
obtained during the 6 month deployment at Viwandani Informal Settlements
Daily means were within 24-hour WHO air quality guidelines on most days (black dotted line) for
both PM2.5 and PM10.
Apparent at St. Scholastica Catholic School were high concentrations occurring when the
wind direction was oriented parallel to Thika super highway located to the south-east of the
school (Figure 5). Data capture was 100% for the six month period.
Figure 5: Bivariate polar plot for PM2.5 and time series showing daily means for PM2.5 and PM10 measurements
obtained during the 6 month deployment at St. Scholastica Catholic School.
Daily means were mostly within 24-hour WHO air quality guidelines (black dotted line) for both
PM2.5 and PM10 with occasional peaks, possibly associated with pollution events.
High daily mean concentrations associated with proximity to the highway source were observed.
Statistics derived from the data give diurnal variations (Figure 6) consistent with this analysis.
11
Figure 6: Time variation plot, averaged over different time scales, of hourly PM2.5 concentrations at St. Scholastica
Catholic School.
Figure 6 shows time series PM2.5 data averaged over different time scales and was used to
assess trends for the deployment period. Time averages used are: (A) day of the week hourly
averages for the deployment period, (B) overall hourly average 24-hour variation for the
deployment period, (C) monthly averages for whole time period and (D) mean daily
concentrations for the whole time series. The shading in the plots shows the 95% confidence
interval in the mean.
Monday mornings and Thursday evenings are observed to be most polluted. The diurnal pattern
(B) is similar to diurnal profiles described for anthropogenic urban pollutants due to traffic with
morning and evening rush hours clearly visible. The ‘Weekend effect’, often related to reduced
traffic volume on weekends compared to weekdays, was observed (D). Significant reduction in
mean PM2.5 concentration (from around 18 μg/m3 on working days down to 14 μg/m3 as an
average) is observed on Sunday.
At UN Environment Headquarters in Gigiri data capture was 100%. The polar plot shows
evidence of a source contribution from the north-west. Daily means were below 24-hour WHO
air quality guidelines for the 6 month period (black dotted line) for both PM2.5 and PM10.
12
Figure 7: Bivariate polar plot for PM2.5 and time series showing daily means for PM2.5 and PM10 measurements
obtained during the 6 month deployment at UN Environment Headquarters in Gigiri
Similarly, at All Saints Cathedral Primary School data capture was 100%. The polar plot
shows evidence of a source to the north. Daily means were below 24-hour WHO air quality
guidelines for the 6 month period (black dotted line) for both PM2.5 and PM10.
Figure 8: Bivariate polar plot and time series showing daily means for PM2.5 and PM10 measurements obtained during
the 6 month deployment at All Saints Cathedral Primary School
Alliance Girls High School was included in the network to determine general urban
background concentrations. Data capture was 83% - the loss of data being caused by an
electrical malfunction.
13
Figure 9: Bivariate polar plot for PM2.5 and time series showing daily means for PM2.5 and PM10 measurements
obtained during the 6 month deployment at Alliance Girls High School
Daily means occasionally topped the 24-hour WHO air quality guidelines (black dotted line) for
both PM2.5 and PM10 in the first three months (June, July and August). Lower daily mean
concentrations were recorded in October and November coinciding with the rainy season. On
23rd June, the daily PM10 mean was 102.0 μg/m3 and exceeded the 24-hour limit value set out in
EMCA (Air Quality) Regulations, 2014 (green dotted line). Such high concentration levels are
associated with pollution events that are further investigated in the next section.
Table 4 shows results of average PM2.5 and PM10 concentrations for the six month deployment
period. Also included are average concentrations during pollution events, discussed in the next
chapter.
Site No.
Name
Average concentration
(μg/m3)
Average concentration
during pollution episodes
(μg/m3)
PM2.5
PM10
PM2.5
PM10
1
Kibera Girls Soccer Academy
24
63
51
116
2
Viwandani Informal
Settlements
18*
40*
35
68
3
St. Scholastica Catholic
School
17
32
36
59
4
UNEP Headquarters, Gigiri
12
30
**
**
5
All Saints Cathedral Primary
School
11
27
**
**
6
Alliance Girls High School
17*
33*
42
61
Legend - * Data capture less than 100% (Table 3),
** Excluded from analysis since mean values were below WHO guidelines
Table 4: Summary of the results from sites in the Nairobi network for the time period June to November 2016
14
2.4 Case Study: Nairobi Placemaking Week
Background
There are few opportunities to quantify traffic emissions and air quality improvements resulting
from reduced traffic in Nairobi’s central business district, although the relationship often seems
self-evident. The inaugural Nairobi Placemaking Week, held from 28th November to 4th
December 2016 presented one such opportunity. This intervention focused on Jevanjee
Gardens and adjoining streets with full closure of Monrovia and Moktar Daddah streets, and
partial closure of the busier Muindi Mbingu street.
Figure 10: (a) Moktar Daddah street during the placemaking week, (b) and (c) Muindi Mbingu street during the
placemaking week
15
Placemaking is committed to “strengthening the connection between people and the places they
share” and begins with citizens working together to improve their local environment. One of the
objectives was to advocate for a healthier, inclusive and vibrant city through safe and
pedestrian-oriented streets and was achieved through temporary interventions (Figure 10).
Study Design
A fixed-site affordable air quality monitor, model AS510 from Atmospheric Sensors Ltd UK
hereafter referred to as node, was set up along Muindi Mbingu street. The node provided
continuous measurements of nitric oxide (NO), nitrogen dioxide (NO2), sulphur dioxide (SO2),
particulate matter (PM1, PM2.5 and PM10) as well as temperature and relative humidity for three
months (October, November and December) at high temporal resolution (every minute). The
analysis presented focusses on ambient particulate matter (PM10 and PM2.5) for the pre-
intervention, intervention and post-intervention periods. The first few weeks of measurements
(5th to 30th October) are excluded from the analysis to account for sensor stabilization within the
surrounding ambient environment.
Findings
Data analysis was done using the OpenAir open source air quality analysis tool (OpenAir
version 1.8-2 on R version 3.3.0). Bivariate polar plots show local source signals of pollutants
and indicate long-range transport. These plots show the relationship between three variables:
pollution data (particulate matter measurements), wind speed and wind direction.
There was negligible reduction (< 1%) in average PM2.5 concentrations during the placemaking
week with a higher reduction of 8% recorded the week after (5th to 11th December). On the
contrary, there was a high reduction (14%) in average PM10 concentrations in the
intervention week with a lower reduction of 5% recorded the week after. Sources of ambient
particulate matter vary and include, for the purposes of this intervention, combustion sources
from vehicular emissions and resuspended dust from paved and unpaved roads. Generally
combustion processes form fine particles (PM2.5 fraction) while resuspended dust sources are
predominantly coarse particles which are part of the PM10 fraction. It is important to note that
this study was performed in the short rains season when concentrations are observed to
decrease in the greater Nairobi area (5 other units deployed across the city).
The bivariate plot (Figure 11) was used to explore source locations and corresponding
atmospheric transport directions of particulates. For intervention analysis, observed
measurements were split to three different time intervals (before, during and after) using the
openair function. PM2.5 concentrations increase with winds blowing from the north-
east at low speeds (Figure 11 - Top) associated with local emissions from vehicles along Muindi
Mbingu street. Contribution from long-range transport is observed at high wind speeds. During
the placemaking week concentrations are observed to decrease with little contribution from
long-range transport.
16
Figure 11: Bivariate polar plot of PM2.5 (Top) and PM10 (Bottom) showing hourly mean concentrations before, during and after the intervention
17
Figure 12: Diurnal patterns for PM2.5 (Top) and PM10 (Bottom)
18
PM10 concentrations increase with winds blowing from the north-east and south-east directions
at high speeds (Figure 11 - Bottom) associated with polluted air masses from Moi avenue and
beyond.
High daily mean concentrations associated with proximity to the street road source were
observed. Statistics derived from the data give diurnal variations (Figure 3) consistent with this
analysis. The diurnal pattern for PM2.5 (Figure 12 - Top) is similar to diurnal profiles described
for anthropogenic urban pollutants due to traffic with morning and evening rush hours clearly
visible. The peaks are less pronounced during the intervention as concentrations decrease. In
general, the analysis reveals a bimodal pattern with two peaks during morning, ~6:00 am local
time, and evening, ~6:00 pm local time, that follow times of peak traffic.
Air Quality and Policy implications
As cities increasingly urbanise industrial and transport activities concentrate and affect urban
ambient air quality. Air pollution is a priority area of action contributing significantly to the global
burden of disease. Interventions aimed at improving health by reducing emissions and personal
exposure can help assess the impact of policies, whether positive or negative.
Study findings suggest that it would be beneficial to focus on actions that limit ambient pollutant
concentration and spatial distribution. Emitted pollutants disperse in the atmosphere at different
rates often influenced by meteorology. Green spaces are able to trap and absorb suspended
particulate matter. Density of green spaces in a city therefore affects concentration of these
pollutants and fosters social cohesion.
One strategy to limit spatial distribution would be traffic routing particularly in the city center
where traffic density is not evenly distributed resulting in higher emissions in certain zones. This
would present city authorities with the option to offer alternative forms of non-motorised
transport to city dwellers.
19
3. EXPOSURE AND HEALTH ASSESSMENT
Urban air quality management should be aimed at minimizing health impacts of air pollution.
Associations between ambient particulate matter and adverse health effects are focused on
either short-term (acute) or long-term (chronic) exposure. Increasing concentrations of
particulate matter have been linked with deteriorating human health by affecting respiratory and
cardiovascular systems.
Lack of spatially resolved daily mean concentration data restricts most health impact
assessments of ambient air pollution to urban areas where monitoring sites are located and may
not be representative of the population as a whole. Two sites in the Nairobi network are located
in informal settlements with high population density.
This initial exposure and health assessment is focused on PM2.5.
3.1 Pollution episodes
For health impact studies it is often necessary to select periods when ambient pollutant
concentrations remain above limit values. This selection takes into account both the threshold
and persistence. For analyzing pollution episodes we are interested in selecting PM2.5
concentrations above the 24-hour WHO air quality guideline, 25 μg/m3, for at least 8
consecutive hours. These pollution episodes are important from a health perspective and the
conditions under which they occur have been investigated in this assessment.
Two sites, UN Environment and All Saints, where daily means were below WHO air quality
guidelines were omitted from this analysis. The primary focus is on Kibera and Viwandani where
population density is very high (CBS 2001).
OpenAir polar plots were used to investigate local source signals of pollution events. Episodes
were determined using the OpenAir utility function as follows:
In Kibera derived episode conditions data shows evidence of pollution from almost every
direction (right plot of Figure 13). This is observed to occur at low wind speeds suggesting a
local source such as irregular open burning of solid waste. Open burning of waste has
increasingly become a preferred waste disposal option for Nairobi residents, especially in the
informal settlements.
20
Figure 13: Bivariate polar plot comparison between average conditions (left) and pollution episodes conditions (right)
for PM2.5 at Kibera
Depending on the nature of the waste, uncontrolled refuse disposal is likely to result in
incomplete combustion, releasing harmful air pollutants, such as fine particulates, polycyclic
aromatic hydrocarbons, heavy metals and dioxins.
We isolate local source signals for additional source characterization using the method
described by Uria-Tellaetxe and Carslaw, 2014. For this technique, Conditional Bivariate
Probability Function (CBPF), data is conditionally-selected to exclude factors that obscure
individual signals such as dispersion conditions occasioned by variations in meteorology.
Figure 14: CBPF plots for 10 intervals of PM2.5 hourly average concentrations in Kibera for June November 2016
21
Using several CBPF plots for a range of concentrations gives the best results for source
apportionment. For this analysis, we use ten PM2.5 intervals for each quartile representing 0
10%, 10 20%, 20 30%, 30 40%, 40 50%, 50 60%, 60 70%, 70 80%, 80 90% and
90 100%. It is evident that the source to the north east is only important between the 10 to 40th
percentiles corresponding to average hourly concentrations of 9.3 17 μg/m3. Strong winds (6
8 m s-1) associated with this finding could mean that it is a distant source that has diluted over
long distances. Local sources have their maximum influence at high concentrations of 23 242
μg/m3 corresponding to the high percentiles, 60 100th. We attribute this to open burning for
three reasons: (1) open burning is a non-point source over the area of interest seen in the CBPF
plot (80 90th percentile bottom left, row 3 column 2) as coming from all directions; (2) open
burning emissions are released near ground level and do not disperse as effectively as stack
emissions; (3) open burning emissions are episodic in time and localized. This makes such
emissions a public health concern.
In Viwandani it is evident that the highest concentrations were dominated by north-westerly
conditions (right plot of Figure 15) corresponding to flow from industrial emissions. Similar to
Kibera, there is also evidence of pollution from all directions under pollution episode conditions,
albeit exerting a lesser effect (lower concentrations) compared to the more significant industrial
source.
Figure 15: Bivariate polar plot comparison between average conditions (left) and pollution episodes conditions (right)
for PM2.5 at Viwandani
Similar to Kibera, we perform CBPF analysis for additional source characterization. For
Viwandani, burning at a distant waste disposal site to the east has influence at low
concentrations of 6.1 8.4 μg/m3. Flow from industrial emissions from the north west is
important between the 90 100th percentile and accounts for high concentrations of 37 175
μg/m3.
22
Figure 16: CBPF plots for 10 intervals of PM2.5 hourly average concentrations in Viwandani for June November
2016
At St. Sholastica Catholic School pollution episodes are dominated by north-westerly
conditions (right plot of Figure 17). South-westerly conditions corresponding to flow from
vehicular emissions from the nearby highway are also evident. Measurements from the node
were therefore representative of traffic pollution for this site.
Figure 17: Bivariate polar plot comparison between average conditions (left) and pollution episodes conditions (right)
for PM2.5 at St. Scholastica
23
At Alliance Girls High School the highest concentrations were dominated by south-westerly
conditions (right plot of Figure 18) corresponding to flow from traffic emissions from the nearby
Southern bypass occurring at wind speeds of 4-5 m s-1. The bypass, officially opened in 2012,
was meant to reduce congestion in the city by providing an alternative route for motorists,
mostly long-distance trucks. Diesel engines used in most long-distance trucks are known to
produce more particulates.
Figure 18: Bivariate polar plot comparison between average conditions (left) and pollution episodes conditions (right)
for PM2.5 at Alliance Girls High School.
3.2 Health impacts of air pollution in Nairobi from long-term
exposure.
The main purpose of an air pollution health impact assessment is to estimate the risks of
exposure to air pollution and of changes in exposure that may result from planned policies or
other modifications of air quality (HIP, 2014). Many countries require that health impact
assessments be undertaken for purposes of informing the decision-making process for air
quality management.
Part XII of the Fifth Schedule in the EMCA (Air Quality) Regulations, 2014, provides guidelines
for assessment of air quality in Kenya. These guidelines are however solely focused on
establishing ambient pollutant concentrations, source contributions and identifying mitigation
measures. The Global Burden of Disease 2013 study estimates that 24,500 deaths annually are
attributable to air pollution in Kenya, both indoor and outdoor.
Health effects of exposure to particulate matter should ideally be quantified using several years
of data. We present here early estimates from six months of monitoring covering two seasons in
Nairobi. Measurements from all six sites were averaged to give aggregate hourly mean
concentrations. Concentration-response functions used in health impact assessments are
derived from epidemiological studies using fixed-site population-oriented monitors.
24
Characteristic
AirQ2.2 (Now AirQ+)
BenMAP-CE
LEAP-IBC
Spatial resolution:
Regional
National
City-level
Pollutants:
Any grid
PM2.5
PM10
Ozone
NO2
SO2
CO
Health outcome:
Mortality
Disability-adjusted life years (DALYS) or
years of life lost (YLL)
Morbidity
Peer reviewed / Policy applications:
Peer reviewed
Expert
In preparation
Used for policy applications
User input for population exposure
characterization:
Concentrations
Concentrations
Emissions
(Translates emissions
to concentrations)
Table 5: Air pollution health impact assessment tools (Adapted from Anenberg et al. 2015)
25
Several computer-based tools are available for air pollution health impact assessment. Table 5
summarizes three commonly used tools. We chose AirQ+ as it matches the spatial resolution
(city-level) of this assessment context and is able to use measurements from deployed
monitors.
Uncertainties
The estimates generated by AirQ+ carry some uncertainties as they rely on epidemiologically-
derived concentration-response associations. The key sources of uncertainty are:
1. Concentration-response functions used come mainly from studies conducted in
Europe and North America and are based on the systematic review of all studies
available until 2013 and their meta-analysis. As a result, assessments carried outside
these regions can be associated with additional uncertainties. In recent times, studies of
short-time exposure-mortality estimates have been replicated in several cities in
developing countries. Therefore, application of mortality effects is reasonable.
2. Effect of co-pollutants is not accounted for in the estimates. Exposure to other
pollutants, such as NO2, has known health effects.
3. Threshold concentrations have not been explicitly examined in epidemiological studies
informing derived concentration-response functions. The relationship between relative
risk and ambient concentration is often perceived to be linear. However, a log-linear
function is more plausible for cities with high pollution levels.
4. Population baseline rates for health outcomes will change over time as health habits,
income and other factors change.
In establishing the relationship between ambient concentrations and health outcomes,
information from different sources is required thereby propagating these inherent uncertainties.
Quantifying the health impact in Nairobi
The Policy Question:
How many deaths (out of the number of total deaths) are attributable to long-term
exposure to PM2.5 exceeding WHO Air Quality Guidelines level (10 μg/m3)?
For ‘Impact Evaluation’ the following data was used:
Mortality (all non-external causes, per 100 000 population): 651.2 (Source: GBD 2015
estimates for Kenya);
The total number of adults (≥ 30) affected by the pollutant: estimated as 60% of 4 million
- projected population of Nairobi in 2016;
the Relative Risk values, suggested values for all-cause mortality are: 1.062 (CI 1.040-
1.083);
Impact of concentration: 10 μg/m3 as suggested by WHO Air Quality Guidelines (2005);
Aggregated air pollution data from the Nairobi network. The mean value is a six month
average of available daily values. For this analysis we consider average concentrations
from two population-oriented sites Kibera and Viwandani.
26
Figure 19: AirQ+ results for PM2.5 long-term adult mortality
The results indicate that around 1,200 premature deaths (Figure 13) are caused by long term
exposure to PM2.5 and can be avoided by limiting concentrations to 10 μg/m3. The interval
between the lower and upper values, 812 and 1,606, represents a part of the uncertainty
associated with the estimation. Nairobi county recorded 22,500 deaths in 2016 (Table 6). This is
likely to be an underestimate considering that significant portions of the populace are exposed
while walking or stuck in traffic jams.
Nairobi County Mortality Data
2014
2015
2016
Volumes
91
94
90
Total number of deaths
22,750
23,500
22,500
Source: Nairobi County
Table 6: Mortality data from Nairobi County
Figure 20: Relative risk of long-term exposure to PM2.5 in Nairobi
The relative risk estimate describes the likelihood of an adverse health outcome (e.g. premature
death, heart attack, asthma attack, emergency room visit, hospital admission) occurring in a
population exposed to a higher level of air pollution relative to that in a population with a lower
exposure level. Figure 20 shows that the relative risk of exposure to PM2.5 increases to 1.13 for
EMCA (Air Quality) regulations limit value 35 μg/m3 (red line) and even further to 1.26 during
pollution episodes (yellow line). This would translate to an increase in the premature deaths
attributable to PM2.5 exposure.
27
REFERENCES
Carslaw, D.C. and K. Ropkins, (2012). OpenAir an R package for air quality data analysis.
Environmental Modelling & Software. Volume 27-28, pp. 5261.
Carslaw, D.C. (2015). The OpenAir manual open-source tools for analyzing air pollution data.
Manual for version 1.1-4, King’s College London.
CBS 2001. Population and Housing Census. Counting our People for Development Vol.1.
Central Bureau of Statistics, Nairobi.
DIRECTIVE 2008/50/EC OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL. Official
Journal of the European Union, 2008.
Global Burden of Disease Study 2015. Global Burden of Disease Study 2015 (GBD 2015)
Reference Life Table. Seattle, United States: Institute for Health Metrics and Evaluation (IHME),
2016.
Government of Kenya (2010). The Constitution of Kenya, Government Printing Press, Nairobi,
Kenya.
HIP (2014) Frequently asked questions about integrating health impact assessment into
environmental impact assessment [online]. Human Impact Partners
(http://www.epa.gov/region9/nepa/PortsHIA/pdfs/FAQIntegratingHIA-EIA.pdf).
USEPA, National Ambient Air Quality Standards (NAAQS). 2011, United States Environmental
Protection Agency (USEPA).
WHO Air quality guidelines for particulate matter, ozone, nitrogen dioxide and sulfur dioxide.
Global update 2005. Summary of risk assessment,. 2006b, World Health Organisation.
... However, majority of the LMICs are still in the process of formulating air quality management plans (AQMP) (Naiker et al., 2012;Gulia et al., 2015). The success of any AQMP depends upon the integration and linkage between its key components i.e., policy objectives, monitoring network, emission inventory, source apportionment, air quality modelling, health exposure assessment, control strategies and public participation (Longhurst et al., 1996;Gokhale and Khare, 2007;Gulia et al., 2015;2017). The role and importance of above key components depend upon the pollution problem, its sources and climatic condition. ...
Article
Rapid urbanization along with industrial growth is one of the major causes of elevated air pollution levels in urban areas of low and middle income countries (LMICs). They are further associated with adverse health impacts within urban ecosystems. In order to manage and control deteriorating urban air quality, an efficient and effective urban air quality management plan is required consisting of systematic sampling, monitoring and analysis; modelling; and control protocols. Air quality monitoring is the essential and basic step that develops foundation of any management plan. The present research article describes a comprehensive methodology for establishing a systematic and robust air quality monitoring network in LMICs and strengthening the effectiveness and efficiency of urban air quality management frameworks. It also describes step-by-step procedures for chemical characterization of both organic and inorganic constituents of ambient particulate matter along with molecular markers, which are essential to identify the corresponding sources of particulate matter, an integral part of air pollution monitoring protocol. Additionally, it discusses the need for coupling low cost wireless sensor-based stations with a limited number of manual and conventional real time ambient air monitoring stations in order to make it cost effective, yet robust. The article demonstrates that satellite-based remote sensing monitoring calibrated with ground level measurement has the potential for regional scale air quality monitoring that captures transport of transboundary pollution.
Article
Full-text available
openair is an R package primarily developed for the analysis of air pollution measurement data but which is also of more general use in the atmospheric sciences. The package consists of many tools for importing and manipulating data, and undertaking a wide range of analyses to enhance understanding of air pollution data. In this paper we consider the development of the package with the purpose of showing how air pollution data can be analysed in more insightful ways. Examples are provided of importing data from UK air pollution networks, source identification and characterisation using bivariate polar plots, quantitative trend estimates and the use of functions for model evaluation purposes. We demonstrate how air pollution data can be analysed quickly and efficiently and in an interactive way, freeing time to consider the problem at hand. One of the central themes of openair is the use of conditioning plots and analyses, which greatly enhance inference possibilities. Finally, some consideration is given to future developments.
The OpenAir manual -open-source tools for analyzing air pollution data. Manual for version 1.1-4
  • D C Carslaw
Carslaw, D.C. (2015). The OpenAir manual -open-source tools for analyzing air pollution data. Manual for version 1.1-4, King's College London.
/50/EC OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL
CBS 2001. Population and Housing Census. Counting our People for Development Vol.1. Central Bureau of Statistics, Nairobi. DIRECTIVE 2008/50/EC OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL. Official Journal of the European Union, 2008.
The Constitution of Kenya
  • Kenya Government
Government of Kenya (2010). The Constitution of Kenya, Government Printing Press, Nairobi, Kenya.
Frequently asked questions about integrating health impact assessment into environmental impact assessment
HIP (2014) Frequently asked questions about integrating health impact assessment into environmental impact assessment [online].
WHO Air quality guidelines for particulate matter, ozone, nitrogen dioxide and sulfur dioxide. Global update 2005. Summary of risk assessment
WHO Air quality guidelines for particulate matter, ozone, nitrogen dioxide and sulfur dioxide. Global update 2005. Summary of risk assessment,. 2006b, World Health Organisation.