D.P. Mukhopadhyay

1
UNCERTAINTY AND ROLE OF STATISTICS IN CHEMICAL
CHARACTERISATION OF PARTICULATE
in a workshop on
SOME RECENT ADVANCES IN THE FIELD OF AIR QUALITY
MANAGEMENT
organised by IAAPC
September 23, 2007
D.P. Mukhopadhyay Dpmcpcb_at_yahoo.com Central
Pollution Control Board ,Zonal Office, Kolkata
2
Studies carried out in Eastern Part of India on
Chemical Characterisation of Particulates

• Characterisation of PM10 and PM2.5 at traffic
  intersection in Kolkata and assessment of their
  impact on human health
• Characterization of PM10 during Diwali
• Characterisation of emission and dust form
  different sources
• Ambient Air Quality Monitoring at Port Canning,
  West Bengal border in Bangladesh (MALE
  DECLARATION)
• Impact of sponge iron plant on environment and
  efficiency of ESP in pollution control
• Ambient air quality at Jatin Das Park, Kolkata
• Identification of pollutant type, size and
  concentration in the genesis of severe
  thunderstorm

Contd2.
3
Studies carried out in Eastern Part of India on
Chemical Characterisation of Particulates

• Performance evaluation of RDS of Envirotech
  with respect to Anderson Sampler
• Assessment of SO2 and NO2 by Active and
  Passive method at Canning, West Bengal
• Assessment of BTX in major Traffic intersection
  in Kolkata
• Performance of Air Pollution control devices
  in Foundries in Howrah, West Bengal
• Role of Statistics in interpretation of Air
  Quality data
• Evaluation of QC/QA status in ambient air
  quality monitoring projects

4
Important prerequisite for Air Quality
Management Monitoring
Basic Needs
1. Finance 2. Selection of site 3. Selection
of sampler and their performance 4. Operational
Discipline
1. Installation of sampler 2. Collection of
sample and their analysis
5. Data Management
1. Reporting of the data with level of
uncertainty 2. Interpretation
5
Potential Uncertainty Sources

• Sampling (cleaning procedure, operation of
  filter paper, flow etc.)
• Transport and storage of sample
• Blank sample
• Analysts
• Quantification of detection limit
• Calibration
• Environmental influence parameters

6
Important Steps Of Uncertainty

• Identification of sources of error
• Minimization of error
• Identification of sources of
  uncertainty and estimation of uncertainty
• Reporting of data alongwith level of
  uncertainty and sensitivity analysis of
  uncertainty

7
Selection of Sampler and its Performance

• Availability of sophistical sampling equipment
  cannot ensure reliable data
• Only operation discipline can ensure the
  reliability of data
• Performance of sampling equipment depends on
  controlling of error from different sources not
  only on selection of equipment
• Major sources in Indian context are

• Cleanliness
• Flow Accuracy
• Quality of Filter paper and mode of preparation
• Voltage fluctuation
• Design of Impingers
• High mass loading causing poor accuracy of flow
• Attrition of mass
• Artifact
• Sampling duration
• Moisture content (internal/external) in filter
  paper

8
Experimental evidence on Magnitude of error on
Operational Discipline
1. Cyclone were cleaned after 24 hours after
using it on completion of three shifts leading
to error of 2.5 to 6 (variation due to
season) 2. Cyclone were cleaned on completion
of 8-hourly shift leading to reduction of error
to lt 1 3. Black fine mass accumulated on the
wall of the cyclone particularly in Kolkata 4.
This source of error was frequently observed in
different organisation in India 5. Now error
is minimised significantly but uncertainty
prevails at the level of lt 1 which is
negligible.
9
Identification of sources of error and their
magnitude

• Several experiments were carried to estimate
  the uncertainty budget on the sampling
• Outcome of these experiments revealed the
  following major sources

• Flow accuracy
• Duration of sample
• Quality and preparation of filter paper
• Flow rate for SO2 and NO2
• Moisture content
• Design of Impingers

• Other sources may be considered as minor

10
Traceability in Air Quality Monitoring

• Measurement of the amount in sample taken for
  analysis
• Preparation of the sample according to field
  and defined experimental condition
• Calibration of an instrument with a standard
  solution of known concentration
• Measurement of the instrument response
• Calculation of the concentration of the analyte
  in original sample

11
Estimation of Precision, Accuracy and Bias
Basic components and their criteria
12
Measurement of Uncertainty of PM10

• The Uncertainty of PM10 measurement was
  quantified from precision data and collocated
  parallel measurements
• Uncertainty was 4 (95 confidence interval)
  at concentration range of 50 - 500 µg/m3
• The quantitation limit was determined from
  Standard Deviation of field blanks

13
Advantages over error minimization
Comparative study was carried by Anderson Sampler
and Envirotech RDS Sampler
14
COMPARISON OF RESULTS OBTAINRD THROUGH ACTIVE AND
PASSIVE SAMPLER
15
Concentration and ratio of PM2.5 with PM10 (12
hrs. mean value)
Since consistent ratio of PM10/PM2.5 are being
maintained and present monitoring network are
mainly covered by PM10 sampler. Chemical
characterization of PM10 can easily be
extrapolated to PM2.5 based on earlier
observation. Strong data base can be maintained
particularly where PM2.5 sampler is not
available. But one time study is required.
16
Chemical Behaviour of Particulates in Environment

• Several studies were carried out to evaluate
  chemical character in Particulates in different
  area

- Kolkata (Metropolitan) - Bhubaneswar, Orissa
(City) - Asansol, West Bengal (City) -
Dhanbad, Bihar (City) - Durgapur, West Bengal
(City) - Canning, Sundarban (Remote
Area),Transboundary station - Moutorh, West
Bengal (Rural Area)
Contd.2.
17
Chemical Behaviour of Particulates in Environment
Association of Chemicals with Particulate
-
18
Behaviour of Particulates in Environment
Data were statistically processed to evaluate
distribution of total PAH in PM10 and PM2.5 and
their variability

• Correlation coefficient was found highly
  significant(r0.99) in both the cases
• PAHs have consistent affinity toward PM2.5 on
  normalising with Particulate

19
Chemical Behaviour of Particulates in Environment

• Metals

• Except Fe, there is a tendency of adherence of
  metals in fine particulate.
• Metals are originated from the same sources
  contributing Particulate.

20
Chemical Behaviour of Particulates in Environment
Ions

• Almost same trend was observed in case of ions

21
Chemical Behaviour of Particulates in Environment
EC/IC
22
Impact of PAH on Human Health
Quantification of 1-Hydroxypyrene from the urine
23
Particulates and its Chemical Nature

• Environmental behaviour of chemicals are mainly
  governed by concentration of Particulates and
  their sizes
• Affinity of chemicals to adhere on the
  particulate depends on the sizes
• Nature of distribution of chemicals between PM10
  and PM2.5 is almost consistent
• Ratio of PM10 / PM2.5 is also consistent
• Interrelation among the parameters and very good
  correlation of these parameters with PM10 with
  few exception in particular area clearly revealed
  the adherence of chemicals to particulate.
  Increase particulate enhances of chemicals which
  is a cause of concern.

These observations prompted to study in details
the following aspects using RSPM, SO2 and NOx
data generated by WBPCB through AAQMS.

• Reliability of data and measurement of
  uncertainty
• Distribution pattern
• Behaviour of PM10 among the hours, shifts (8
  hours), days, months
• Influence of meteorological condition
• Calibration function
• Influence of different activities
  leading to air pollution.

24
Mode of Processing of Data
Mode of Processing of Data

• Mean
• Median
• Coefficient of variation
• Auto correlation
• Multiple Regression Method
• Analysis of Variance
• Principal Component Analysis

25
Distribution Pattern of Pollutants

• Distribution of 15 minutes peak
• Distribution of 1-hour mean data

Based on the following parameters Midpoints of
Groups - Frequency - Cumulative Frequency -
Percent
Observations

• Distribution shows that the large majority of
  1-hour concentration are relatively small
  compared to 15 - minutes values found to occur
  extremely high at few occasions
• Such variability may be attributed to nearby
  sources (point non-point) having emission of
  random nature influenced
  by meterological condition
• The level of sudden exposure to high
  concentration may be quantified

Cont2
26
Distribution Pattern of Pollutants

• Excedence of permissible limit of interest can
  be clearly depicted
• Influence of different activities (Traffic,
  Industry, burning landscape, meteorological
  condition responsible for air pollution directly
  or indirectly can be easily ascertained if date
  are processed accordingly.
• Consistency in concentration of Particulates
  can be evaluated based on the ratio of
  15-minutes / hourly / daily data

27
Variation among the shifts and days of each

• No systematic trend was observed
• Little variation among the days was observed
• Emission of pollutant to atmosphere and their
  retention are of random nature

28
Influence of Meteorological Parameter

• Influence of Meteorological condition was
  evaluated using pearson correlation coefficient
• These coefficients clearly indicated the strong
  association of prevalent Particulate with wind
  speed and temperature

Critical wind speed causes collapsing of
assimilation capacity is less than 1.2m/s.
Lowering of temperature aggravate the problem.
PCA has established this observation.
29
Principal Component Analysis

• SPM 2. RPM 3. SO2 4. NO2 5. NO 6. O3 7.CO
  8. Temperature 9.NMHC 10. Wind Speed 11.
  Relative Humidity.
• Wind speed (WS) is mainly responsible for such
  variability of pollutants as it figures in first
  component and Temperature is next to the WS as it
  figures in second component

30
On thorough scrutinizing and processing the
Automatic AAQ data, the salient observations are
given below

• The excedence of NAAQS stipulated for
  residential area is restricted to five months
  (Jan, Feb, Sep, Nov and Dec). But sudden increase
  of the level of concentrations occurred at
  certain point of time
• The variation of concentration of PM10 was
  significantly high among the shifts as reflected
  from CV which even reached to 50. But no
  definite trend
• Though variation of emission from different
  sources are expected during day and night but
  that is not reflected in variation of PM10. This
  aspect is more prominent during winter
• The strong association of PM10 with wind speed
  and temperature revealed that variability of PM10
  are mainly governed by wind speed and
  temperature.

31

• The prevailing wind speed in the winter in the
  range of 1.0 to 1.3 meter/sec is not adequate for
  dispersion of the pollutants. Whereas, little
  increase of this range may capable to disperse
  the pollutant in the later month, namely March
  and April.
• The problems of accumulating RPM in surface layer
  has started below 25 oC and further aggravated
  with decrease of temperature. As a result,
  scenario of ambient air quality is worst during
  January and improves with definite trend from
  the end of January except in few (not exceeded
  2 of the all raw data) occasions. This may be
  attributed to local effect such as traffic
  congestion, open burning, etc.
• The frequency of exceeding the limits shows good
  agreement with the degree of wind speed and level
  of ambient temperature. .

32
The ratio of SPM RPM was lowest in the range
of 0.32 to 0.63 during 06-14 hrs and 0.60 to 0.73
during 14-00 hrs to 22-00 hrs and 0.51 to 0.86
during 22-00 hrs to 6-00 hrs. Increase in ratio
by 0.35 to 0.86 is mainly due to more
contribution of the RSPM. Quantity of emission
both from industries and vehicles remains more or
less uniform over the years but assimilation as
well as dispersion of the pollutants is
completely dependant on meteorological factors.
Land use pattern may aggravate valley effect that
would accelerate inversion.
33
Distribution of ozone and O3 depleting substances
in the atmosphere and observed changes
Seasonal and day to day changes associated with
meteorological condition
Influence of dynamical processes on ozone
abundance
? Transport of pollutants by Brewer-Dobson
Circulation from the main source to the region
of highest abundance ? Chemistry of ozone by
causing changes in temperature leading to
chemical reaction
Ozone Chemistry
Simulation and Depletion
34
Distribution, Climatology and Natural variation
Forecasting day to day changes

• Total ozone variation are highly correlated
  with meteorological disturbances
• Statistical model has already been developed
  based on multiple regression of past values of
  AAQD and meteorological variable
• Distribution pattern based on possibility
  theory and using AAQD generated in few sites
  adopting QA programme, initial forecast produced
  by regression equation is correlated
• Outlier data were detected. These were mainly
  due to problem in calibration of analyser
• Monthly average was verified with CV. If CV gt
  10 percent, values greater 2 ? were eliminated
  with understanding the logic

35
Atmospheric CO2 variability (hourly) at city
Monthly average was verified with CV. If CV gt 10
percent, values greater 2 ? were eliminated with
understanding the logic for such high. Then ANOVA
technique was applied to evaluate the level of
variation among the three months. Then
variability of diurnal cycle on month to month
basis normalising with synoptic weather events.
The harmonic regression was applied for this.
This observation clearly reflect rate of photo
synthesis, man made activities, say condition.
Windrose clearly indicate the source contribution
limitation.

• Significant gaps in the data set as
  calibration failure
• Non functioning of sensor
• Localised effect

36
Data collection and processing
The data file in Excel and software in GW basic
developed for processing of AAQD
File - 1
Name Monitoring function CP1 PM10 PM10
Measurement system CP2 MET Meteorological
Measurements
File - 2
Data files are ASCII files with comma separated
field and text surrounded with double quotes
Year, Month, Day, Time, Parameter-1,
Parameter-2, Parametern, Flag-1, Flag-2, Comment,
ExNainf
These are created defining the field and
limitations
Contd2
37
Data collection and processing
File - 3
Weekly Data Created from the time checked raw
data files with understanding of flagging events
fresh in mind and stored in the same format
File - 4
Monthly data
File - 5
Yearly data
File - 6
Overall interpretation of data giving emphasis on
distribution pattern (spatial / temporal).
Chemical behaviour, trend, uncertainty
Software in GW Basic developed for transformation
of data file and processing of data using
different statistical technique were developed in
as Monthly.Bas, Weekly.Bas, Regyr.Bas etc.
38
Flag Definition (Calibration gas, System check)

Z - Zero Check V - Data passes all
test I - Intake sample line problem
Q - Unspecified questionable data W1 - First
working gas W2 - Second Working Gas SF - Sample
Flow fluctuation
39
THANK YOU