Technical Details of Network Assessment Methodology: Concentration Estimation Uncertainty Area of Station Sampling Zone Population in Station Sampling.

Slides:



Advertisements
Similar presentations
Spatial point patterns and Geostatistics an introduction
Advertisements

NASA AQAST 6th Biannual Meeting January 15-17, 2014 Heather Simon Changes in Spatial and Temporal Ozone Patterns Resulting from Emissions Reductions: Implications.
Adam Pacsi GISWR 2011, 11/17/2011 Fine Particulate Matter Concentrations in the Dallas-Fort Worth Area Adam Pacsi.
Representativity of the Iowa Environmental Mesonet Daryl Herzmann and Jeff Wolt, Department of Agronomy, Iowa State University The Iowa Environmental Mesonet.
University of Wisconsin-Milwaukee Geographic Information Science Geography 625 Intermediate Geographic Information Science Instructor: Changshan Wu Department.
Correlation and Autocorrelation
QUANTITATIVE DATA ANALYSIS
More Raster and Surface Analysis in Spatial Analyst
Ka-fu Wong © 2007 ECON1003: Analysis of Economic Data Lesson2-1 Lesson 2: Descriptive Statistics.
Applied Geostatistics Geostatistical techniques are designed to evaluate the spatial structure of a variable, or the relationship between a value measured.
Spatial Analysis Longley et al., Ch 14,15. Transformations Buffering (Point, Line, Area) Point-in-polygon Polygon Overlay Spatial Interpolation –Theissen.
Spatial Interpolation
Business Statistics: A Decision-Making Approach, 7e © 2008 Prentice-Hall, Inc. Chap 3-1 Business Statistics: A Decision-Making Approach 7 th Edition Chapter.
Subcenters in the Los Angeles region Genevieve Giuliano & Kenneth Small Presented by Kemeng Li.
Measures of Variability. Why are measures of variability important? Why not just stick with the mean?  Ratings of attractiveness (out of 10) – Mean =
Chap 3-1 Statistics for Business and Economics, 6e © 2007 Pearson Education, Inc. Chapter 3 Describing Data: Numerical Statistics for Business and Economics.
Correction of Particulate Matter Concentrations to Reference Temperature and Pressure Conditions Stefan R. Falke and Rudolf B. Husar Center for Air Pollution.
1 Spatial Statistics and Analysis Methods (for GEOG 104 class). Provided by Dr. An Li, San Diego State University.
Measures of Variability: Range, Variance, and Standard Deviation
Maps of PM2.5 over the U.S. Derived from Regional PM2.5 and Surrogate Visibility and PM10 Monitoring Data Stefan R. Falke and Rudolf B. Husar Center for.
CAPITA CAPITA PM and Ozone Analysis A. PM2.5 National Maps B. Visibility (PM2.5) trends C. Natural (out of EPA jurisdiction) Events D. US-Canada Ozone.
Spatial Statistics Applied to point data.
Types of data and how to present them 47:269: Research Methods I Dr. Leonard March 31, :269: Research Methods I Dr. Leonard March 31, 2010.
Applied Cartography and Introduction to GIS GEOG 2017 EL
Alternative Approaches for PM2.5 Mapping: Visibility as a Surrogate Stefan Falke AAAS Science and Engineering Fellow U.S. EPA - Office of Environmental.
GIS Modeling for Primary Stroke Center Development Anna Kate Sokol, M.U.P. Sr. GIS Specialist City of South Bend, IN.
1 A Bayesian Method for Guessing the Extreme Values in a Data Set Mingxi Wu, Chris Jermaine University of Florida September 2007.
Variability The goal for variability is to obtain a measure of how spread out the scores are in a distribution. A measure of variability usually accompanies.
MODELS3 – IMPROVE – PM/FRM: Comparison of Time-Averaged Concentrations R. B. Husar S. R. Falke 1 and B. S. Schichtel 2 Center for Air Pollution Impact.
Measures of Central Tendency and Dispersion Preferred measures of central location & dispersion DispersionCentral locationType of Distribution SDMeanNormal.
Psyc 235: Introduction to Statistics Lecture Format New Content/Conceptual Info Questions & Work through problems.
Spatial Pattern of 1-hour and 8-hour Daily Maximum Ozone over the OTAG Region Rudolf B. Husar CAPITA, Center for Air Pollution Impact and Trend Analysis.
© 2006 McGraw-Hill Higher Education. All rights reserved. Numbers Numbers mean different things in different situations. Consider three answers that appear.
CMAS special session Oct 13, 2010 Air pollution exposure estimation: 1.what’s been done? 2.what’s wrong with that? 3.what can be done? 4.how and what to.
Declustering in the Spatial Interpolation of Air Quality Data Stefan R. Falke and Rudolf B. Husar.
PM Network Assessment: Speciated Network Planning Prepared for EPA OAQPS Richard Scheffe by Rudolf B. Husar Center for Air Pollution Impact and Trend Analysis,
Ideas on a Network Evaluation and Design System Prepared for EPA OAQPS Richard Scheffe by Rudolf B. Husar and Stefan R. Falke Center for Air Pollution.
Data Types Entities and fields can be transformed to the other type Vectors compared to rasters.
National PM2.5 Concentration Spatial Pattern Contact: Stefan Falke, Background and Rationale Annual Average.
Spatial Interpolation III
Spatial Pattern of PM2.5 over the US PM2.5 FRM Network Analysis for the First Year: July 1999-June 2000 Prepared for EPA OAQPS Richard Scheffe by Rudolf.
Extent and Mask Extent of original data Extent of analysis area Mask – areas of interest Remember all rasters are rectangles.
Quantitative Network Assessment Methodology: Illustration for PM Data Prepared for EPA OAQPS Richard Scheffe by Rudolf B. Husar and Stefan R. Falke Center.
Measures of Dispersion. Introduction Measures of central tendency are incomplete and need to be paired with measures of dispersion Measures of dispersion.
CHAPTER 3  Descriptive Statistics Measures of Central Tendency 1.
PM Network Assessment: Speciated Network Planning Prepared for EPA OAQPS Richard Scheffe by Rudolf B. Husar Center for Air Pollution Impact and Trend Analysis,
Air Quality Monitoring Network Assessment: Illustration of the Assessment and Planning Methodology Speciated PM2.5 Prepared for EPA OAQPS Richard Scheffe.
Visual Correlation between Air Pollution and Population Density in Major Metropolitan Areas Texas A&M University, Department of Civil Engineering, Applications.
August 1999PM Data Analysis Workbook: Characterizing PM23 Spatial Patterns Urban spatial patterns: explore PM concentrations in urban settings. Urban/Rural.
So, what’s the “point” to all of this?….
Final Project : 460 VALLEY CRIMES. Chontanat Suwan Geography 460 : Spatial Analysis Prof. Steven Graves, Ph.D.
Network Assessment by Station Rankings: Description of Methodology Network Assessment Technical Support Group June 2001.
Ozone Data Integration for OTAG Quality Analysis and Evaluation Model Janja D. Husar and Rudolf B. Husar CAPITA, Center for Air Pollution Impact and trend.
L15 – Spatial Interpolation – Part 1 Chapter 12. INTERPOLATION Procedure to predict values of attributes at unsampled points Why? Can’t measure all locations:
Network Assessment Based on Daily Max Ozone Concentration Prepared for EPA OAQPS Richard Scheffe by Rudolf B. Husar and Stefan R. Falke Center for Air.
Spatial Statistics and Analysis Methods (for GEOG 104 class).
Network Assessment Based on Compliance Monitoring (Deviation from NAAQS) Prepared for EPA OAQPS Richard Scheffe by Rudolf B. Husar and Stefan R. Falke.
Introduction to statistics I Sophia King Rm. P24 HWB
Variability Introduction to Statistics Chapter 4 Jan 22, 2009 Class #4.
Alternative Approaches for PM2.5 Mapping: Visibility as a Surrogate Stefan Falke AAAS Science and Engineering Fellow U.S. EPA - Office of Environmental.
August 1999PM Data Analysis Workbook: Characterizing PM23 Spatial Patterns Urban spatial patterns: explore PM concentrations in urban settings. Urban/Rural.
Assessment of the Speciated PM Network (Initial Draft, November 2004 ) Washington University, St. Louis CIRA/NPS VIEWS Team.
Technical Details of Network Assessment Methodology: Concentration Estimation Uncertainty Area of Station Sampling Zone Population in Station Sampling.
INTERPOLATION Procedure to predict values of attributes at unsampled points within the region sampled Why?Examples: -Can not measure all locations: - temperature.
Exposure Prediction and Measurement Error in Air Pollution and Health Studies Lianne Sheppard Adam A. Szpiro, Sun-Young Kim University of Washington CMAS.
Skudnik M. 1*, Jeran Z. 2, Batič F. 3 & Kastelec D. 3 1 Slovenian Forestry Institute, Ljubljana, Slovenia 2 Jožef Stefan Institute, Ljubljana, Slovenia.
Introduction to Spatial Statistical Analysis
Statistics: The Z score and the normal distribution
Chapter 3 Describing Data Using Numerical Measures
Descriptive Statistics
Presentation transcript:

Technical Details of Network Assessment Methodology: Concentration Estimation Uncertainty Area of Station Sampling Zone Population in Station Sampling Zone Prepared for EPA OAQPS under Cooperative Agreement Richard Scheffe by Stefan R. Falke and Rudolf B. Husar Center for Air Pollution Impact and Trend Analysis, CAPITA Washington University, St. Louis December 2000

Network Assessment Measures AQ Management ActivityGeographic Info. Need Station Measure Risk assessmentPollutant concentration 4 th highest O3 Risk AssessmentPersons/Station Compliance evaluationConc. vicinity to NAAQS Deviation from NAAQS Reg./local source attribution & trackingSpatial coverage Area of Sampling Zone All aboveEstimation uncertainty Conc. & Est. Difference Persons in sampling zone The general framework for Air Quality Network Assessment is presented elsewhere. There are at least five different measures that represent the information need for (1) risk assessment, (2) compliance monitoring and (3) tracking are listed below. This report describes three of the measures by which the networks can be evaluated: Concentration Estimation Uncertainty (Information Value of Station) Area of Sampling Zone Persons in Station Sampling Zone The computational detains of each measure are presented along with illustrative examples.

Error Estimation by Cross-Validatation Cross-validation is applied to obtain an estimation error. This involves removing a monitor site from the data base and using the remaining sites to calculate an estimated concentration at the removed monitor location. The estimation error is calculated as the difference between the Estimated - Measured Concentration. The estimate is determined using declustered inverse squared distance weighted (1/r2) interpolation. The nearest 5 sites within a 750 kilometer radius of the estimation location are used in the estimation calculation. Declustering reduces the relative weight of spatially clustered monitor sites during spatial interpolation. It is accomplished by the introduction of two characteristic distances 1) the distance from the monitoring station to the estimation point (R ij ) and 2) the average of the distances between the monitoring site and its surrounding sites. A cluster weight (CW) is defined as the ratio of to R ij. i is the estimation location j is the monitoring site being declustered k is an index of the sites surrounding the monitoring site within a distance R ij of the estimation point r is the average distance between the monitoring site and the sites surrounding it R ij is the distance between the monitoring site and estimation location n is the number of sites within a distance R ij of the monitoring site (including the monitoring site being declustered). A site is clustered if the distances between the monitoring site and its neighboring sites are small (/R ij << 1) compared to the distance between the monitoring site and the estimation point. A feature of this definition is that the degree to which the station is clustered changes depending on the location of the estimation point. For example, a group of sites in a city is considered unclustered when estimating the concentration within the city but is clustered when estimating concentrations in locations substantially outside of the city.

Declustering Configurations The sites X 1, X 2, and X 3 are equidistant from the estimation point i and are unclustered Declustered weighing shows the proper allocation of the 1/3 weight to the cluster of sites. There is a cluster of four sites. When applying standard distance weighted interpolation, the cluster will account for 2/3 of estimated value at i while the two single sites each only account for 1/6 of the total weight Standard interpolation applies equal weight; each site has 1/3 of the weight on the estimate at i.

Concentration Error, E The concentration error is determined by –selectively removing each site from the database –estimating the concentration at that site by spatial interpolation –setting the error as the difference between the estimated and measured values, E = Est.-Meas. The error estimates in both metric of ozone concentration over the Eastern US ranges between 0-15 %. High estimation error is generally observed over areas with low station density. Low estimation error generally occurs over areas with high station density

Concentration Estimation Error (~ 5-6 ppb) For the entire dataset, the measured and estimated data have the same avg. and a slope of 1. The standard deviation of the meas.-est. difference is about 5 ppb, no bias. The estimation error is also random in space. The CastNet network characterizes the non- urban concentrations. The estimation error for CastNet data is the same as for the entire network. This implies that the errors are random.

Example: O 3 Station Ranking by Estimation Uncertainty The uncertainty measures the ability to estimate the concentration from other data. The highest uncertainty (red) is found urban stations where the concentrations are highly variable in space and time. The lowest uncertainty (blue) is at remote sites where the concentrations are more homogeneous in space and time From the perspective of estimation uncertainty, the blue stations have the lowest rank.

Technical Details of Network Assessment Methodology: Concentration Uncertainty Area of Station Sampling Zone Population in Station Sampling Zone Prepared for EPA OAQPS Richard Scheffe by Stefan R. Falke and Rudolf B. Husar Center for Air Pollution Impact and Trend Analysis, CAPITA Washington University, St. Louis December 2000

Station Sampling Zones Every location on the map is assigned to the closest monitoring station. At the boundaries the distance to two stations is equal. Following the above rules, the ‘sampling zone’ surrounding each site is a polygon. The polygons are converted to an ESRI ArcView ‘shapefile’ The the area (km 2 ) of each polygon is calculated ArcView ‘calcarea’ function.

Sampling Zone Illustrations (St. Louis, MO) There are three different types of sampling zones: Single monitoring stations that are far from other stations (light blue) have large and symmetric sampling zones. Stations inside clusters (red) in urban areas have small but symmetric sampling zones. Stations on the edge of clusters (yellow) have larger asymmetric, elongated sampling zones.

Station Sampling Zones in Different Parts of EUS

Sampling Area Distribution Function Over the Eastern US, the average O3 sampling zone is 5900 km 2 or 77x77 km. The distribution of sampling areas is very broad, ranging from 44 km 2 to 98,000 km 2. The 25% of the stations with the smallest area covers only 2.5% of the total EUS area while the upper station quartile covers 70%. In other words, eliminating 25% the ‘smallest’ stations would increase the area of the remaining stations only by 2.5%.

Example: Ranking by Area of Sampling Zone The area of the sampling zone is measure of the spatial coverage and uniformity The stations with large sampling areas (red dots) are unclustered remote sites outside of urban areas Conversely, the stations with small sampling areas (blue dots) are in clusters, mostly in urban regions. The clusters with small station areas are located over the NE megalopolis, Chicago, Pittsburgh, St. Louis, etc. They rank lowest in area coverage.

Network Assessment Based on Estimation Uncertainty Prepared for EPA OAQPS Richard Scheffe by Stefan R. Falke and Rudolf B. Husar Center for Air Pollution Impact and Trend Analysis, CAPITA Washington University, St. Louis December 2000

Population Density The population data are available on census track resolution. The population density is highly textured; it varies by three orders of magnitude over the EUS.

Population in Each Station Sampling Zone The population for each monitor was calculated from census tract population data (1990) from available from ESRI.  The station population was calculated by summing the persons in each census tracts located within the sampling zone polygon of the station.  Following the spatial join operation, the population was added as an attribute of each station.

Population Distribution Function Over the Eastern US, the average number of persons in an O3 station sampling zone is 220,000. The distribution of population/station varies from virtually zero (hard to measure) to 2.8 million. The 25% of the stations with the smallest population accounts for 4% of the total EUS population while the upper station quartile includes 60% of the total population. In other words, eliminating 25% stations with the lowest population would be affecting only 4% of the EUS population. The population in each sampling zone is weakly correlated with the sampling zone area since the population density varies greatly.

Ranking by Population in the Sampling Zone The number of persons in a station’s sampling zone is a scaling factor for the overall health risk. Areas of large population per station (red) are found over the NE megalopolis but also over more remote areas. Small population/station (blue) is generally found remote sites but also in some urban clusters, e.g. Chicago, New Orleans, St. Louis. From the perspective of population coverage, the blue stations have the lowest rank.