Comparison of SA approaches and analysis of non linearities in a - - PowerPoint PPT Presentation

The European Commission’s science and knowledge service

Joint Research Centre

Comparison of SA approaches and analysis of non linearities in a real case model application

C.A. Belis, P. Thunis, C. Cuvelier EC-JRC,

• G. Pirovano RSE S.p.A,
• A. Clappier Univ. Strasbourg,

Fairmode Tech meeting Tallinn, 26-28 February 2018

2

Source Apportionment and sensitivity analysis

Strong non-linearity is associated to secondary pollutants deriving from precursors emitted by different sources. We need to understand how often (when, which ones) sources stray from linearity Clappier et al., 2017 Stein and Alpert decomposition 2 sources

∆𝐷𝑩𝑪= ∆𝐷𝑩 + ∆𝐷𝑪 + 𝐷𝑩𝑪

Stein and Alpert decomposition 3 sources ∆𝐷𝑺𝑩𝑱= ∆𝐷𝑺 + ∆𝐷𝑩 + ∆𝐷𝑱 + 𝐷𝑺𝑩 + 𝐷𝑺𝑱 + 𝐷𝑩𝑱 + 𝐷𝑺𝑩𝑱

3

Simulation with tagged species and Brute Force approaches

Experiment Design and data processing by JRC Model: CAMx run by RSE Tagged species module: PSAT Pollutant: PM2.5 Area: Po Valley Reference year: 2010 Time window: full year Domain: 580 x 400 km2 Grid step: 5 km x 5km approx. Meteorology: WRF 14 layers Emissions: EMEP (Europe), ISPRA (Italy), INEMAR (regional) processed with SMOKE

4

Simulation with tagged species and Brute Force approach

Brute force: 3 sources reduced: agriculture, industry and transport Base case 100% reduction of:

• Scenario 1: M10,
• Scenario 2: M34,
• Scenario 3 M7,
• Scenario 4 M10 and M34,
• Scenario 5 M10 and M7 ,
• Scenario 6 M7 and M34 ,
• Scenario 7 M10, M7 and M34

Same scenarios with 50% reduction

5

CAMX SIMULATION PO VALLEY WITH PSAT AND BRUTE FORCE

• Comparison PSAT and Brute Force approaches
• Check interaction terms (of the Alpert algebraic expression)

for every pair of sources

• Analyse geographical patterns
• Analyse influence of time resolution
• Examine emissions in support to data analysis
• Test behavior in different types of areas (urban, rural, etc.)

6

CAMX BASE CASE

PM10 CONCENTRATIONS

7

CAMX BASE CASE

EMISSIONS NH3 EMISSIONS SO2 EMISSIONS NOX EMISSIONS SO4-2

8

BF 100%-tagged, Agriculture

PM10 AGRICULTURE

9

BF 100%- tagged, Industry

PM10 INDUSTRY

10

BF 100%- tagged, Traffic

PM10 TRAFFIC

11

BF 100% Interaction terms Agriculture Traffic

PM10 AGRI-TRAFF

13

BF Interaction terms Agriculture Traffic

PM10 AGRI-TRAFF100% PM10 AGRI-TRAFF 50%

14

BF 100% Interaction terms Agriculture Industry

PM10 AGRI-INDU

15

BF 100% Interaction terms Traffic Industry

PM10 TRAF-INDU

16

BF 100% Interaction terms Agri. Ind. Traf.

PM10 THREE SOURCES

17

Conclusions

• Tagged species (TS) lower than Brute force (BF) especially in Agriculture

(avg. 70%). Higher differences in rural areas compared to urban ones.

• Traffic and industry differences are much lower (avg. 5-10%).
• Hourly and daily variability very noisy
• Non linearity for 50% reduction much lower (1/4) than 100% reduction
• The two and three way interactions terms have marked seasonal trends.
• The stronger non linearity is in the agriculture - traffic (A-T) interaction
• Significant differences between cities and rural (different regimes: NH3

limited vs NO2 limited?)

• A-T non linearities (negative) more relevant in summer (NO2 limited?)
• In certain cities (NH3 limited?) the seasonal trends compensate leading

to annual average close to zero.

18

Thank you for your attention