Causes of Dust. Data Analysis

1
Causes of Dust. Data Analysis

• Ilias Kavouras, Vic Etyemezian, Dave DuBois, Mark
  Green, Marc Pitchford, Jin Xu
• Division of Atmospheric Sciences, Desert Research
  Institute

2
Scope and methodology

• Scope identify and quantify sources of airborne
  dust
• Local and regional windblown dust
• Long-range transported dust (e.g. Asia)
• Wildfire-related dust
• Other unknown sources
• Approach Analysis of IMPROVE network and
  meteorological data
• Chemical fingerprints of dust (e.g. Asian,
  wildfire-related)
• Multivariate statistical analysis of Dust
  concentrations, wind speed/direction and
  precipitation

3
Database development
RAWS
Days with precipitation for more than 12h or
precipitation occurred after 1200 p.m.
CASTNET
AZDEQ
Central Meteorological Database
Modified Central Meteorological Database
NPS
ISH
NASA
Grouped in 16 categories according to wind
speed/direction WS10-14, WS214-20, WS320-26,
WS4gt26 mph WD1A315-45, WD2A45-135,
WD3A135-225, WD4A225-315 WD1B0-90,
WD2B90-180, WD3B180-270, WD4B270-360
Dust Meteorological Database
Dust Database
IMPROVE database
4
Sensitivity analysis
Regression coefficients
Dust Database
Model Database
GPS data
Maps for each day
IMPROVE-data YES/NO
Meteo-data YES/NO
Dust event YES/NO
Precipitation YES/NO When? 0-12 or 12-24
Worst day YES/NO Worst dust day YES/NO
5
Statistical analysis Multi-linear regression
analysis

• Measurement inter-correlations Durbin-Watson
  test mostly higher than 1.4
• Tolerance higher than 0.80
• Linear regression was done using three methods
• Forward selection One component is added (if pgt
  set value, rejected)
• Backward selection One component is removed if
  pgt set value
• Stepwise selection One component is added
  those with p gt set value are eliminated

6
Statistical analysis Criteria development

• Significance level 0.100 or 0.150 or higher
• Valid prediction Cpredicted Epredicted gt 0 or
  P0.05,Measured

7
Monthly variation of model dust days
8
Dust days per site (based on regression analysis)
9
1. Salt Creek descriptive statistics
Monitoring period 01/01/01 12/31/03 IMPROVE
database completeness 93.2 Meteorological
database completeness 82.4
Measured dust mass
Predicted dust mass
10
1. Salt Creek Regression coefficients
11
1. Salt Creek Predicted vs. Measured Dust
A-groups
Worst dust days 7 / 4
B-groups
12
2. Bandelier Nat. Mon. descriptive statistics
Monitoring period 01/01/01 12/31/03 IMPROVE
database completeness 92.6 Meteorological
database completeness 76.4
Measured dust mass
Predicted dust mass
13
2. Bandelier Nat. Mon. Regression coefficients
14
2. Bandelier Nat. Predicted vs. Measured Dust
A-groups
Worst dust days 3 / 1
B-groups
15
Date May 15, 2003
X Worst day Worst dust day O Meteorological
data available