Primal path algorithm for compositional data analysisstat.snu.ac.kr/idea/seminar/20170703/plume...
Transcript of Primal path algorithm for compositional data analysisstat.snu.ac.kr/idea/seminar/20170703/plume...
Primal path algorithm for compositional dataanalysis
Jong-June Jeon1
1University of Seoul
Idea seminar 2017 Summer
Jeon, Jong-June Idea seminar 2017 Summer
Atmospheric air pollutant dispersion
Goal: estimate air pollutant concentrations downwind of emissionpoint sources
Downwind air pollution concentrations are a function of
dispersion from point emission sources
dispersion mixing height
atmospheric stability
dispersion coefficients.
....
Note that there are three types of pollutant emission: point, line, plane.
Jeon, Jong-June Idea seminar 2017 Summer
Atmospheric air pollutant dispersion
dispersiontypes
longitudinal: mixing along direction of flowtransverse: mixing perpendicular to the flow
causesgradient in flow velocitysplitting of flow paths
First we consider a dispersion model with mixing direction of flows
Jeon, Jong-June Idea seminar 2017 Summer
Simplified Steady-State Plume Model
View from below of “Coning” plume under neutral atmosphericconditions
Figure: left: snapshot; right: time average
Jeon, Jong-June Idea seminar 2017 Summer
Simplified Steady-State Plume Model
Figure: h: effective height of plume, x: distance
Concentration of a pollutant is proportional to the number of theparticles in the unit volume.
Jeon, Jong-June Idea seminar 2017 Summer
Simplified Steady-State Plume Model
Assume that all particles lie within the cone.
Figure: Ck is a concentration at k
Jeon, Jong-June Idea seminar 2017 Summer
Simplified Steady-State Plume Model
Consider the number of particles pass through the circle in the coneper unit time.
Mass emission rate is fixed.
The particle move with a fixed wind speed.
Then,
Total number of particles = concentration × wind speed× the area of circle.
Note that the total number of particles per unit time is emissionrate
Jeon, Jong-June Idea seminar 2017 Summer
Simplified Steady-State Plume Model
It is assumed that density of pollutant at x far away from the origin isuniform.
Jeon, Jong-June Idea seminar 2017 Summer
Simplified Steady-State Plume Model
Jeon, Jong-June Idea seminar 2017 Summer
Simplified Steady-State Plume Model
C: concentration
u: wind speed
Q: emission rate
C =Qu× density of circle
Jeon, Jong-June Idea seminar 2017 Summer
Simplified Steady-State Plume Model
Figure: H: effective height; h: actual height of plume
Jeon, Jong-June Idea seminar 2017 Summer
Simplified Steady-State Plume Model
C(x, y, z): concentration at (x, y, z)
u: wind speed
Q: emission rate
C(x, y, z) =Qu× density of plane (y-z)
Jeon, Jong-June Idea seminar 2017 Summer
Simplified Steady-State Plume Model
Figure: dispersion along y axis depends on the distance x
Jeon, Jong-June Idea seminar 2017 Summer
Simplified Steady-State Plume Model
Refection
Figure: dispersion along z axis depends on the distance x
Jeon, Jong-June Idea seminar 2017 Summer
Simplified Steady-State Plume Model
Gaussian dispersion equation
C(x, y, z; H) =Qu
f (y, z; Σ(x),H)
=Qu
12πσy(x)σz(x)
exp(−(z− H)2
2σ2z (x)
) + exp(−(z + H)2
2σ2z (x)
)︸ ︷︷ ︸reflection
× exp(− y2
2σ2y
)
Jeon, Jong-June Idea seminar 2017 Summer
Simplified Steady-State Plume Model
The Gaussian dispersion model is theoretically derived by advectiondiffusion equation, which assumes
Steady state conditions: ∂C/∂t = 0
Constant wind speed with height (u does not depend on z).
Diffusion constant in σy and σz does not depend on (y, z).
Mass is conserved:∫ ∫Cdydz = Q for all x > 0
Jeon, Jong-June Idea seminar 2017 Summer
Simplified Steady-State Plume Model
Atmospheric stabilityDry adiabatic lapse rate: 0.976C /100mStability condition by actual lapse rate
super-adiabaticsub-adiabaticinversion
Jeon, Jong-June Idea seminar 2017 Summer
Simplified Steady-State Plume Model
Figure: Stability conditions
Jeon, Jong-June Idea seminar 2017 Summer
Simplified Steady-State Plume Model
Pasquill stability categorycategory: A(very unstable), B(moderately unstable), C(slightlyunstable), D(neutral), E(moderately stable), F(very stable)The category is determined by
day incoming solar radiation (strong, moderate, slight)night cloudiness (cloudy and clear)
Jeon, Jong-June Idea seminar 2017 Summer
Simplified Steady-State Plume Model
Estimation of σy(x) and σz(x)
Figure: left: σy(x); right : σz(x)
Jeon, Jong-June Idea seminar 2017 Summer
Simplified Steady-State Plume Model
EPA’s ISC ModelVertical distribution:
σz = axb
(x: km, σz: m)
Cross-wind distribution:
σy = 465.1168x tan(θ)
where θ = 0.017453293(c− d log(x)). (x: km, σy: m)
Here, the constant c and d is determined by the stability conditions.
Jeon, Jong-June Idea seminar 2017 Summer
Simplified Steady-State Plume Model
Jeon, Jong-June Idea seminar 2017 Summer
Simplified Steady-State Plume Model
Plume rise estimation Required data:
wind speed
stack exit velocity
top inside stack diameter
stack gas temperature
gravity
Jeon, Jong-June Idea seminar 2017 Summer
Simplified Steady-State Plume Model
Gaussian dispersion model
C(x, y, z; H) is computed by
location: (x, y, z);
wind speeddispersion parameter:
stability conditionssolar radiationnight cloudiness
Effective plume height (H)physical plume heightplume rise
Jeon, Jong-June Idea seminar 2017 Summer
AERMOD model
Gaussian dispersion model is used in EPA Industrial SourceComplex - Short Term (ISCST3) dispersion air quality model.ISCST3 assumes
constant wind speed with height (u does not depend on z).diffusion constant in σy and σz does not depend on (y, z).simple topographic model
AERMOD (AMS/ EPA Regularatory model) modify Gaussiandispersion model.
Jeon, Jong-June Idea seminar 2017 Summer
AERMOD model
Wind speed for height
Jeon, Jong-June Idea seminar 2017 Summer
AERMOD model
Terrain effect
Figure: Terrain correction
Jeon, Jong-June Idea seminar 2017 Summer
AERMOD model
Building downwash
Consider downwash of the building within 5 times height of theplume.
Jeon, Jong-June Idea seminar 2017 Summer
AERMOD model
SoftwareAERMOD model (diffusion model)
BPIPPRM (Building downwash pre-processing)AERMET (Meteorological pre-processing: eg wind)AERMAP (Terrain pre-processing)
Surfer 8.0 (visualization)
Jeon, Jong-June Idea seminar 2017 Summer
BPIPPRM
Inputbuilding height
building width
Building length
coordinate distance (x,y) between plume and building
Jeon, Jong-June Idea seminar 2017 Summer
BPIPPRM
Jeon, Jong-June Idea seminar 2017 Summer
AERMET
InputMeteorological file ( date, wind speed, flow vector, temperature,mixing height, stability.)
Mixing data from surface observatory and aerologicalobservatory.
Jeon, Jong-June Idea seminar 2017 Summer
Non gaussian plume
Jeon, Jong-June Idea seminar 2017 Summer
Non gaussian plume
Jeon, Jong-June Idea seminar 2017 Summer
Non gaussian plume
Jeon, Jong-June Idea seminar 2017 Summer
Non gaussian plume
Jeon, Jong-June Idea seminar 2017 Summer
CALPUFF model
CALPUFF model assumes the pollutant consists of puffs.Modeling
the movement of puffs in the 3 dimensional wind field;the contribution of a puff to receptor.
Hence, the model can reflect the change of wind direction andspeed immediately. (Unsteady model)
This model can be used for predicting diffusions of massivepollutant point source near the coastal line
Jeon, Jong-June Idea seminar 2017 Summer
Non gaussian plume
Figure: Comparison of AERMOD and CALPUFF
Jeon, Jong-June Idea seminar 2017 Summer
CALPUFF model
InputMM5 model (3 dimensional wind field model on grid system→location of the center of a puff)
Overwater and coastal interaction effects
Puff formulation
diffusion parameter (σy and σz)
vertical wind shear
plume rise
building downwash
terrain
chemical transformation option
Jeon, Jong-June Idea seminar 2017 Summer
CALPUFF model
Contribution of puffs at ground level (receptor)
C =Q
2πσxσyg exp
(− d2
x
2σ2x
)exp
(−
d2y
2σ2x
)
where
g =2√
2πσz
∞∑n=−∞
exp(−(H + 2nh)2
2σ2z
)
dx: x-coordinate distance between puffs and receptor.
H: height of puff, h: mixing height.
Jeon, Jong-June Idea seminar 2017 Summer