Forecasting Of Ground Level-Ozone Exceedences in

runmidgeAI and Robotics

Oct 20, 2013 (3 years and 10 months ago)

78 views

LOGO

2011

By

Omar M.Hussein
1

and Bassam Tawabini
2


Forecasting Of Ground Level
-
Ozone Exceedences in
Eastern Province of Saudi Arabia Using Zaitun Time
Series Model


Company Logo

Contents;
-



o
Introduction

o
Problem

statement

o
Research

objectives

o
Review


Stratospheric/

topospheric



ozone

Standards(Globally

&

KSA)


Methods

of

forecast

o
Methodology

(Zaitun

Time

series)

o
Results

and

discussion


o
Conclusion

and

recommendation


Introduction


Ground

level

ozone

is

a

secondary

pollutant

that

is

produced

from

photochemical

reactions

between

VOCs

and

NOX

under

the

existence

of

sunlight
.


Ozone

control

has

became

a

challenge

to

urban

air

quality

management
.


One

of

the

most

important

effort

in

ozone

control

is

to

develop

a

sophisticated

ozone
-
level

forecasting

system
.


There

are

a

number

of

successful

air

quality

models

for

forecasting

severe

ozone

events,

including

phenomenological

models

and

statistical

models
.





Research objectives

Company Logo


To

search

and

source

out

for

potential

forecasting

tool

for

ozone

ground

level
.


To

apply

the

forecasting

model

tool

to

predict/forecast

exceedences

of

ground

level

ozone

concentration

in

eastern

province

of

KSA

using

available

data


Problem statement


The

transport

sector,

through

emissions

of

exhaust

gases,

and

emissions

due

to

refilling

of

fuels

is

the

major

contributor

to

the

total

emission

of

precursors

(NO
x
,

VOC’
s

and

CO),especially

in

KSA
.


The

combination

of

the

low

wind

and

high

solar

radiation

effects(eastern

province)

may

lead

to

the

ozone

episodes

in

the

summer

time
.


This

leads

to

higher

ozone

level,

which

necessitates

a

proper

and

good

tool

to

forecast

in

relation

to

data

available
.


Review;

basic ozone chemistry


The reaction that produces ozone in the atmosphere


O + O
2

+ M O
3

+ M


Ozone forms readily in the stratosphere as incoming ultraviolet
radiation breaks molecular oxygen (two atoms) into atomic
oxygen (a single atom).


Difference between stratospheric and tropospheric ozone
generation is in the source of atomic O.


In troposphere;
-


NO
2

+ Sunlight NO + O


O +
O
2
O
3


Also;


CO+ OH CO
2

+H


H+ O
2

HO
2


HO
2

+

NO
2

NO + O


O + O
2

O
3



Company Logo

Primary standards

Secondary
standards

level

Average
time

level

Average
time

Ozone

0.075 ppm (2008
std)

8
-
hour

Same as Primary


0.08 ppm (1997
std)

8
-
hour

Same as Primary


0.12 ppm

1
-
hour

Same as Primary


Standards

KSA Standards of Topospheric Ozone level


Presidency of Meteorology and Environment (PME) in
Saudi Arabia has the most flexible Ozone standard of 150
ppb as compared to the Royal Commissions of Jubail and
Yanbu.



Company Logo

Location


Regulatory
Agency


1
-
hour (ppb)


8
-
hour (ppb)



Saudi Arabia



PME



150



NA


Saudi Arabia



RCJY



120



NA


United States



US EPA



120



80


Methods of forecasting


common

methods

used

to

forecast

ozone

concentrations
;


Persistence
;

Today's

(or

yesterday's)

observed

ozone

concentration

is

tomorrow's

forecasted

ozone

concentration


Climatology
;

Historical

frequency

of

ozone

events

help

guide

and

bound

ozone

forecast
.


Criteria
;

When

parameters

that

influence

ozone

are

forecasted

to

reach

a

pre
-
determined

level

(criteria),

high

ozone

concentrations

are

forecasted
.


CART
;

A

decision

tree

predicts

ozone

based

on

values

of

various

meteorological

and

air

quality

parameters
.


Regression
;

A

regression

equation

predicts

ozone

concentrations

using

observed

and

forecasted

meteorological

and

air

quality

variables
.


Company Logo



Neural

Networks
;

A

non
-
linear

set

of

equations

and

weighting

factors

predicts

ozone

concentrations

using

observed

and

forecasted

meteorological

and

air

quality

variables
.


Phenomenological/Intuition
;

A

person

synthesizes

meteorological

and

air

quality

information

including

ozone

predictions

from

other

methods

to

produce

a

final

ozone

forecast
.


3
-
D

Air

Quality

Models
;

A

three
-
dimensional

prognostic

model

replicates

the

meteorological

and

air

quality

processes

that

create

ozone
.


Each

type

has

advantages

and

disadvantages



Company Logo

Methodology;
Introduction



Data of ozone exceedences were obtained from a research
thesis conducted by a former student from Earth science
department.


Which was subsequently obtained before by KFUPM
Earth Science department for the last ten (10) years (1997
-
2006) from PME, and Royal Commission for jubail and
yanbu.


Data was hourly average of ozone concentration in Jubail,
dammam, and Hofuf cities for the last 10 years.


Zaitun Time Series
-

Time Series Analysis and Forecasting
Software.


Company Logo

Company Logo

Methodology;

Company Logo


Used for statistical analysis, series modeling and
forecasting of time series data.



It provides several statistics and neural networks
models, and graphical tools that will make work on
time series analysis easier.



Statistics and Neural Networks Analysis include;
Trend Analysis, Decomposition, Moving Average,
Exponential Smoothing, Linear Regression,
Correlogram,and Neural Networks.


Features


Creating a time series data in


certain frequency.


Variable and Group




Spreadsheet View, Graphic View, and Statistics View.



Time series analysis model; Trend Analysis,
Decomposition, Moving Average, Exponential Smoothing,
Neural Networks etc


Company Logo

Company Logo

Trend Analysis


Zaitun Time Series provides a feature to analyze
trend component of a time series.


There are several trend types available e.g. linear,
quadratic, cubic , and exponential.


Company Logo

Exponential Smoothing Analysis with
Zaitun Time Series

Company Logo

Decomposition Analysis

Company Logo

Neural network (Artificial Neural Networks)


Neural networks mechanisms imitate biological
neural network mechanisms.


Like biological neural networks, neural networks
consists of neurons which are connected to each
other and operate in parallel.

Company Logo

Company Logo

Company Logo

Results & Discussion


Verification

; is the process of evaluating the
quality of a forecasting system by comparing the
predicted values with the observed(actual) ones.


Verification was done with trend analysis,
decomposition, exponential smoothing and neural
network methods and actually comparison showed
that neural network produced the closest to the
actual values of our data.


Forecasting
; the process of forecasting was done
also with all the mention techniques available in
Zaitun time series software.



Company Logo

Company Logo

Jubail site 2

Month

Ozone conc.
(ppb)

Oct 2005

122.5

April 2005

123

June 2005

123

Ozone Exceedance Events in Jubail Site 2 from 1994 to 2006

Company Logo

Results (Trend analysis)


Actual vs. forecasted

month

forecasted

Oct 2005

122.0731

Nov 2005

122.0808

Dec 2005

122.0885

Jan 2006

122.0961

Feb 2006

122.1036

Marc 200

122.1112

April 2006

122.1187

May 2006

122.1261

June 2006

122.1335

Company Logo

Decomposition


Actual vs. forecasted

month

forecasted

Oct 2005

122.0731

Nov 2005

122.0808

Dec 2005

122.0885

Jan 2006

122.0961

Feb 2006

122.1036

Marc 200

122.1112

April 2006

122.1187

May 2006

122.1261

June 2006

122.1335

Company Logo

Exponential smoothing

Actual vs. forecasted

month

forecasted

Oct 2005

126.6450

Nov 2005

126.6450

Dec 2005

126.6450

Jan 2006

126.6450

Feb 2006

126.6450

Marc 200

126.6450

April 2006

1
26.6450

May 2006

126.6450

June 2006

126.6450

Company Logo

Neural network

(Actual vs. forecasted)

month

forecasted

Oct 2005

119.5763

Nov 2005

128.1507

Dec 2005

118.4750

Jan 2006

118.4723

Feb 2006

118.8998

Marc 200

118.6361

April 2006

118.9360

May 2006

119.1792

June 2006

130.1853

Company Logo

Dammam Industrial Site

Month

Ozone
conc.(ppb)

Jan 2005

153

Nov 2005

152

June 2006

151

Ozone Exceedance Events in Dammam Industrial Site from 1997 to 2006

Company Logo

Results (Trend analysis)


Actual vs. forecasted

month

forecasted

Jan 2005

149.3525

Feb 2005

149.2776

March2005

149.2017

April 2005

149.1247

May 2005

149.0467

June2005

148.9676

July 2005

148.8875

Aug 2005

148.8063

Sept 2005

148.7241

Oct 2005

148.6408

Nov 2005

148.5565

Dec 2005

148.4712

Jan 2006

148.3847

Feb 2006

148.2973

March2006

148.2088

April 2006

148.1192

May 2006

148.0286

June2006

147.9369

Company Logo

Decomposition

Actual vs. forecasted

month

forecasted

Jan 2005

149.4814

Feb 2005

149.1566

March2005

149.3308

April 2005

149.0039

May 2005

149.1760

June2005

148.8470

July 2005

149.0169

Aug 2005

148.6858

Sept 2005

148.8537

Oct 2005

148.5205

Nov 2005

148.6863

Dec 2005

148.3510

Jan 2006

148.5147

Feb 2006

148.1773

March2006

148.3389

April 2006

147.9994

May 2006

148.1589

June2006

147.8173

Company Logo

Exponential smoothing

Actual vs. forecasted

month

forecasted

Jan 2005

150.1888

Feb 2005

150.1888

March2005

150.1888

April 2005

150.1888

May 2005

150.1888

June2005

150.1888

July 2005

150.1888

Aug 2005

150.1888

Sept 2005

150.1888

Oct 2005

150.1888

Nov 2005

150.1888

Dec 2005

150.1888

Jan 2006

150.1888

Feb 2006

150.1888

March2006

150.1888

April 2006

150.1888

May 2006

150.1888

June2006

150.1888

Company Logo

Neural network

Actual vs. forecasted

Jan 2005

forecasted

Feb 2005

150.6706

March2005

150.5700

April 2005

150.7185

May 2005

150.9838

June2005

151.0662

July 2005

151.1120

Aug 2005

152.7844

Sept 2005

150.7134

Oct 2005

150.7171

Nov 2005

151.8163

Dec 2005

150.9418

Jan 2006

150.6492

Feb 2006

151.0419

March2006

150.3383

April 2006

149.9304

May 2006

149.7535

June2006

149.7566

Company Logo

Discussion


ground

level

ozone

concentration

are

uncertain,

dynamic

and

non
-
linear,

therefore

normal

linear

statitistical

models

inability

to

be

used

for

forecast
.



Zaitun

time

series

has

all

it

takes

to

forecast

while

having

a

lot

of

options

in

its

disposal

for

forecast

application
.


It

is

also

noted

that

in

the

analysis

some

features

are

more

accurate

and

better

forecasting

depending

on

the

nature

of

the

data

available

i
.
e
.

linear,

quadratic,

cubic,

exponential

etc


Neural

network

looks

certainly

the

best

of

all

in

current

and

future

application
.



Company Logo

Conclusion &recommendation


The

study

presented

a

forecast

of

ground

level

ozone

concentration

of

eastern

province

KSA

(
1994
-
2006
),using

zeitun

time

series

model
.


Substantial

results

was

obtained

after

applying

zaitun

time

series

to

the

available

data
.


More

data

is

needed

for

more

accurate

forecast

while

studying

and

incorporating

parameters

like

regional

growth,

meteorological

conditions,

tranportation,and

precursor

emissions
.


References;
-



http://en.wikipedia.org/wiki/Tropospheric_ozone



www.rcjy.com.sa



www.pme.org.sa



http://www.epa.gov/NE/airquality/pdfs/2007
-
2009_dv.pdf



http://env.uregina.ca/publications/thesis/Phunsak/ThesisM
-
TheramongkolPhunsak.pdf



http://www.zaitunsoftware.com/



www.wmo.int/pages/prog/arep/gaw/.../train_what_forecasting.ppt


www.wmo.int/pages/prog/arep/gaw/.../train_12_aq_forecasting_tools.p
pt



www.epa.gov/airnow/.../
monday
.../miller
-
forecasting
-
training.ppt


www.atmosp.physics.utoronto.ca/MAM/plummer.ppt

Company Logo