MANEVU/MRPO Project:Paired Aerosol / Trajectory Database Analysis Tool Development

Combined Aerosol Trajectory Tool, CATT 

R. Husar, K. Hoijarvi, S. Falke, CAPITA, Wash. U, St. Louis

Project Officer, Serpil Kayin, MARAMA

Project Period: September 2002 – July 2003, $50K

Presented at MANE-VU Data Analysis Workshop

Windsor Locks, CT, June 18-19, 2003

Analysis Value Chain: CATT’s Habitat

Next Process

Next Process

TrajData Cube

CATT TAT CAPITA

Aggreg. Traject.

AerData Cube

CATT CAT CAPITA

Aggreg.Aerosol

CATT-In CAPITA

CATT-In CAPITA

Aerosol Data

Collection IMP. EPA

Aerosol Sensors

Integration VIEWS

Integrated AerData

A E R O S O L

Weather Data

Assimilate NWS

Gridded Meteor.

Trajectory ARL

Traject.Data

T R A N S P O R T

Why? How?

When? Where?

Background• Atmospheric aerosol system has three extra dimensions (red), compared to gases (blue):

– Spatial dimensions (X, Y, Z) – Temporal Dimensions (T)– Particle size (D)– Particle Composition ( C ) – Particle Shape (S)

• Bad news: The mere characterization of the 7D aerosol system is a challenge– Spatially dense network -X, Y, Z(??)– Continuous monitoring (T)– Size segregated sampling (D) – Speciated analysis ( C )– Shape (??)

• Good news: The aerosol system is self-describing. – Once the aerosol is characterized (Speciated monitoring) and multidimensional aerosol data are

organized, (see RPO VIEWS effort), unique opportunities exists for extracting information about the aerosol system (sources, transformations) from the data directly.

• Analysts challenge: Deciphering the handwriting contained in the data – Chemical fingerprinting/source apportionment– Meteorological back-trajectory analysis– Dynamic modeling

Project Background

• The source-receptor relationship of particulate matter can be estimated by a number of empirical observation-based techniques. Some techniques are based chemical fingerprinting others on meteorological transport techniques.

• A particularly attractive source-attribution technique, Paired Aerosol / Trajectory Analysis developed by Poirot and Wishinski. It combines the chemical and transport techniques by:– Establishing the major aerosol types at a specific receptor location and time

(PMF and UNMIX)

– Estimating the geographic transport regions for each aerosol type (Residence Time Analysis)

Biomass Smoke

Avg. Mass: 2.4 ug/m3 (32%)

Species: OC, EC, S, K

Summer Maximum

East Coast Residual Oil

Avg. Mass: 0.38 ug/m3 (5%)

Species: OC, EC, S, Si, Ni, V

Winter Maximum

Secondary Coal

Avg. Mass: 3.2 ug/m3 (42%)

Species: S, OC, EC, Na

Summer Maximum

Combining Chemical Fingerprints and Transport, Lye Brook, NH Based on Positive Matrix Factorization, PMF results from B. Coutant and ATAD trajectories from K. Gebhart

Project Goal: Develop an interactive data query and analysis tool for the paired chemical/trajectory analysis

Aerosol Source Type and Transport Origin Analysis (Wishinski and Poirot (2002)

Project Deliverables

1. Implement a relational database that incorporates both the PMF/UNMIX results and for gridded trajectory data.

2. Develop specific SQL filtering and aggregation queries for • aggregated trajectory data based on chemical conditions

• aggregated chemical data based on geographic conditions

3. Develop a graphic interface for user input (query) and for data output as renderd images or as exportable numeric data.

4. Transfer the resulting database to a designated SQL server and provide instructions for addition of chemical and trajectory data.

Relational Database

• PMF and UNMIX data transfer to SQL– The PMF and UNMIX results were provided to us as Excel spreadsheet.

– The metadata for the PMF and UNMIX were obtained verbally from R. Poirot.

– The spreadsheet data were reformatted and imported in the SQL server two tables (ChemFactTable and LocTable)

• Residence Time data transfer to SQL– The residence time data were provided by P. Wishinski on CDROM including full

metadata documentation

– The data were imported into two SQL tables (ResTimeFactTable and LocTable)

PMF/UNMIX Data

CAPITASQL

Database

ATAD Residence Time

Data Input: PMF and UNMIX Model Results

The results of the Battelle/Sonoma modeling project are source profiles and time series for each source contribution at each location

Prepared by

Battelle and Sonoma Tech. Inc.

Source attribution results (PMF and UNMIX) for 16 receptor sites between Illinois and New England using IMPROVE and CastNet data have been completed by a previous project.

Database Structure

The dTrajResTime and dSourceApp tables share the site_code and date keys thereby allowing paired queries to the SQL database.

SQL Queries

SELECT Lat as lat, Lon as lon, Loc_Code as loc_code, SUM(ResTime) AS [VALUE]FROM dTrajResTime WHERE ([Date] IN (SELECT datetime FROM dSourceApp WHERE (Loc_Code = 'loc_code') sql_filter_clause)) GROUP BY GridCode, Lat, Lon, Loc_Code ORDER BY Lon ASC

Settings that are unique to a specific query are designated by red text

Query (filter) result: List of dates the satisfy the chemical filter conditions

ATAD Trajectory and Residence Time Grid

Residence time and ATAD trajectory data superimposed for June 1, 2000.

Residence time aggregate (sum) for a range of dates

Airmass Spource Regions by Season e.g. Sum ResTime for Loc=LYBR, Date between June-Sept

Lye Brook, DJF

Gr Smoky Mtn, JJALye Brook, JJA

Gr Smoky Mtn, JJA

Source Regions by Concentrations - High and Low

ResTime for High C6 (BioSmoke?) Chemical Conditions ResTime for Low C6 (BioSmoke?)

Chemical Conditions

Incremental Transport Probalility

Seasonal Incremental Probability

Secular Differences: 1988-94; 1995-2000

1988-2000

1988-19941994-2000

Transport Probability Metrics

• The transport metric is calculated from two residence time grids, one for all trajectories and another for trajectories on selected (filtered days). Both residence time grids are normalized by the sum of all resdence times in all grid cells:

pijf=rij/ rij pij

a=rij/ rij

• pijf, is the filtered and pij

a is the unfiltered residence time probabilitiy

that an airmasses passes through a specific grid. There is a choice of transport probaility metrics:

• The Incremental Residence Time Probability (IRTP) proposed by Poirot et al., 2001 is obtained by subtracting the chemically filtered grid from the unfiltered residence time grid, IRTP = pij

f - pija

• The other metric is the Potential Source Contribution Function (PSCF) proposed by Hopke et al., 19xx which is the ratio of the filtered and unfiltered residence time probabilities, PSCF = pij

f / pija

Transport Metric Selection

• Currently, there is a choice of two different transport probability metrics:

• Incremental Residence Time Probability (IRTP) proposed by Poirot et al., 2001 is the difference between the chemically filtered and unfiltered residence time probalbilities. Positive values of IRTP in a grid indicates more than average liekihood of transport; (red); negative IRTP values (blue) represent less than average likeihood of transport.

• Potential Source Contribution Function (PSCF) proposed by Hopke et al., 19?? is computed as the ratio of the filtered and unfiltered residence time probabilities. Higher values of PSCF is indicative of inreased source contribution.

• The desired metric is selected through a dialog box invoked by clicking on the right-most button in the TRAJ_CHEM layer.

Results

Combined Aerosol

Trajectory Tool

CATT

CATT Project Status

Current effort to finalize queries‘Public’ testing and user feedback is in progress now

Develop relational database of PMF/UNMIX and trajectory data

Develop specific SQL filtering and aggregation queries - Chemical filtering/aggregation query

- Trajectory filtering/aggregation query

- Paired Chemical/Trajectory data query

Develop graphical user interface to database

Transfer the resulting database to a designated SQL server

CATT Presentation and Workgroup Discussions

Project Status/Summary

1. Relational Database of PMF/UNMIX and trajectory data: Complete

2. Develop specific SQL filtering and aggregation queries• Chemical filtering/aggregation: Developed• Trajectory filtering/aggregation: Developed• Paired Chemical/Trajectory data: Developed, needs user input, testing, feedback

3. Graphic interface for user input (query) and for data output: Developed, needs user input, testing, feedback

4. Transfer the resulting database to a designated SQL server: Not done

Project Milestones Jan-June 2003 1. Feb 1, 03: Complete initial queries, user interface and displays2. Apr 1, 03: Finalize design/implementation of queries, user interface and displays3. Apr-Jun 03: ‘Public’ testing and user feedback4. July 03: Tool delivery

Trajectory Tools Project Options

VIEWS Database Compatibility • Make the chemical-trajectory exploration tool compatible with the

evolving VIEWS database at CIRA, Colorado State U.: – insuring consistency of the data base schema– query tools compatibility– data presentation compatibility

Dynamic Trajectory Aggregation• Online filtering and aggregation of trajectory data

– ad hoc gridding, contouring at arbitrary grid resolution– alternative rendering, e.g. trajectory bundles, instead of residence time

contours

The CATT tool has two components, usable separately or linked

:

1. Chemical filter component. This component is accomplished through queries to chemical data sets. The output of this step is a list of “qualified” dates for a specific receptor location.

2. Trajectory aggregator component. This component receives the list of dates for a specific location and performs the trajectory aggregation, residence time calculation and other spatial operations to yield a transport pattern for specific receptor location and chemical conditions.

Receptor location. Single location; multiple receptors; weighed multi-site

Receptor times. Time range for each site

Temporal filter/weight conditions. Date range; specific dates; weights for each date

Trajectory input files. Pre-computed or on the fly calculated (e.g. HYSPLIT, ATAD etc)

Trajectory aggregation metrics. Endpoint counts, residence time, incr. probability

TAT Output: ASCII point, XMLGrid, GIS

• TAT requires an airmass trajectory dataset for specific locations. The trajectories can be either pre-computed or generated on the fly from meteorological fields.

• The Trajectory Aggregator Tool, TAT, will performs the residence time and other trajectory aggregations on the fly.

• For trajectory aggregation, TAT will require user selection of: