Tom PuruckerUS Environmental Protection Agency

Athens, Georgia, USA

PROTECT Working GroupOslo, Norway Jan 28 2008

Spatial approaches for ecological screening and exposure

assessment of chemicals and radionuclides:

Application of Spatial Analysis and Decision Assistance (SADA) Freeware

2

SADA General InformationWindows--based freeware designed to integrate scientific models with decision and cost analysis frameworks in a user-friendly manner. Can be used to analyze spatially referenced analytical data or to post-process model output for risk assessment purposes.

SADA has been supported by DOE, EPA, and the NRC. SADA Version 3.0 had about 11000 downloads. Version 4.0/4.1 has had 10000+ since January, 2005. Version 5 due this summer.

• Visualization/GIS

• Statistical Analysis

• Geospatial Interpolation

• Geospatial Uncertainty Analysis

• Human Health Risk Assessment

• Ecological Risk Assessment

• Custom Analysis

• Area of Concern Frameworks

• Cost Benefit Analysis

• Sampling Designs

• Export to Arcview/Earthvision

• MARSSIM

3

Derivation and application of benchmarks

SADA is primarily chemical (though MARSSIM (dual level DCGL exposures) and BDAC BCGs accessible)

Not a rad model that competes with Resrad/Erica/R&D 128

Can be used to (back)calculate species/site-specific benchmarks using exposure models

Sample size and designs for collecting and comparing data against benchmarks

Find accurate percentiles of exposure when spatial correlation is present

Ecologically, SADA most often used to apply benchmarks (e.g. site pass/fail, risk quotient maps)

4

Basic GIS Capabilities

Exposure modeling (in SADA) is spatial, not dynamic

Consistent with applying numerical target values (e.g., BDAC BCGs) and (chemical) dose

Spatial interpolation methods useful for soil/sediment exposures/benchmark applications

Surface water- applications of benchmarksGIS Capabilities- (Layers, Polygons, Vertical

layers, etc.)

5

Sample DesignsSADA has a number of sample design strategies in Version 4. These strategies include initial and secondary designs. Some are based on data alone while others are based on modeling results. With the exception of a couple of exclusively 2d designs all are available in 3d dimensions.

Systematic, unaligned, & hotspot search grids

Random

Judgmental

Adaptive Fill

High Value

AOC Boundary

Initial Designs Secondary Designs

6

Sample Analytical Data

Or transport model output files

7

Data Formats

SADA can accept data in two formats: comma delimited files (csv) and Microsoft Access.

Requires the presence of certain fields in the data set.EastingNorthingDepthValueName

Can use other forms of information as wellMediaDetectionDateCAS Number

Any other form of meta data can be imported as well. User can plot and retrieve this meta data during an analysis.

SADA recognizes soil, sediment, surfacewater, groundwater, air, biota, and background, and the “basic” media type. Basic is assigned to data that have no media type.

8

Scaleable interface

Interviews

Analysis Box Data Type Box Data Name Box Labels Box Layers Box

Results WindowSteps Window Parameters Window

9

Basic Data ExplorationData/GIS Visualization

Spatial AggregationData Screening

3D visualizationStatistics

Spatial Data Query

10

Classical StatisticsEPA DQO/DQA

Numerous univariate statistics

Non-parametric hypothesis testing

Power curve based sample sizes

Histograms and CDFs

11

Calculate sample size based on Sign Test and WRS Test

Develop initial sample design incorporating DCGLS, Area Factors, Instrument sensitivity

Post sampling analysis (A site passes or fails)

Detecting and Defining Elevated Areas

MARSSIM Functionality

12

Spatial Estimation

The estimated value Vo at an unsampled location is estimated as the weighted average of nearby values.

Search Neighborhood

Unsampled Point

13

Modeling spatial auto-correlationSemi-variograms often do not conform to the well behaved monotonic increasing variogram

structures seen in text book examples.

Distance

Sem

ivar

Distance DistanceSADA provides 3 standard correlation models that provide flexibility in semi-variogram data: Spherical, Exponential, and Gaussian.

Autofit routines are available to assist in fitting correlation models. These models are then used to generate kriging maps.

14

Comparing spatial models (Cross Validation)

Cross validation provides estimates of model error based on existing data.

Process of removing one sample at a time and predicting the concentration at that location, and measuring the error.

Statistic generated for all samples that can be used to make comparative statements between different spatial models:Mean of errorsAbsolute mean errorMean squared error

15

Spatial Analysis

Spatial EstimationSpatial Uncertainty

Model Spatial Correlation

16

Kriging MapsKriging provides an estimate with an associated kriging variance at

each grid node.

Allows one to be conservative (e.g. percentiles > .5)

Allows one to spatially ascertain the difference between an “optimistic, realistic, and pessimistic” (e.g. 25th, 50th, and 75th) maps.

“optimistic”

p=0.25

“realistic”

p=0.5

“pessimistic”

p=0.75

17

Spatial Dose/Risk Assessment

Conventional Risk Assessment Limitations

Typically regulatory exposure assessment guidance recommends a summary statistic for the exposure concentration

Spatial information is lost when a summary statistic is used in the RA- exposure is assumed to be continuous in space and time

Often this lost info not recovered in the rest of the assessment/remediation process

18

Spatial Dose/Risk Assessment

Reasons for incorporating spatial statistics into dose/risk assessment

Maximize the use of limited resources

Efficiently collect data

Retain collected spatial info in the risk assessment

Use all types of available data, including expert judgment

To more adequately characterize the exposure distribution

Extrapolate from known data to cover data gaps

Account for spatial processes related to exposure

Better understand uncertainties in the exposure assessment

19

Ecological Capabilities in SADA

SADA implements EPA methods for conducting ecological risk assessments

Benchmark database for contaminant effects on ecological receptors

Exposure modeling for over 20 other terrestrial species

Contains EPA default exposure parameters for the risk models where available

Tabular screening and risk results

Point screens

Risk and dose mapping

20

Ecological Functionality Setting Up Ecological Risk

Ecological Risk Assessment Procedure

Setting Physical Parameters

Description of Ecological Benchmark Database

Histograms of Benchmark Values

Tables of Benchmark Values

Setting Screening and Exposure Statistics

Area Result Tables (Screens, Ratios)

Map Result Values (Screens, Ratios)

Rematching a Single Contaminant

Checking Ecological Version

Terrestrial Dose Modeling

21

Setup Ecological Risk

Identify source benchmarks database

Match contaminants in data to benchmark contaminants

Adds ecological information to SADA file

22

Hazard v. Risk AssessmentEcological Hazard Assessment- a comparison of an environmental

concentration to an estimated toxic threshold for a particular contaminant

-most common method for examining effects of chemicals in environment

-comparison of environmental exposure concentration to a toxic threshold (benchmark)

-iterative (or tiered) implementation

-number of toxicity data sets for soil, sediment, and surface water available for screening

Ecological Dose/Risk Assessment- explicitly attempts to estimate the probability and magnitude of the effects of exposure to contamination

23

Benchmark ScreeningMedia-specific concentration benchmarks

Choice of statistics (max, percentile, UCL95, etc.)

Hierarchy of media-specific benchmarks for screening

Spatial and tabular display of ratios

Derivation of Benchmarks:Toxicity testing (acute or chronic)

-regression of concentration-response data-hypothesis testing

Extrapolation from another benchmark

Simulation of an assessment endpoint

24

Ecological Benchmark ScreeningSuitable for screening ERAs

Compilation of ecological benchmarks for surface water (14), soil (11), sediment (17), and biota (8)

Benchmarks a function of environmental variables where appropriate

Choice of statistics (max, percentile, UCL95, etc.)

Hierarchy of media-specific benchmarks for screening

Spatial and tabular display of ratios

25

SADA Eco Rad screening/dose calculations

BDAC support, next version of eco risk library has:

BDAC BCG rad benchmarks to open up spatial screening tools for rads

Simple (DCF/BCF) aquatic/riparian/terrestrial dose calculation for rads, again to allow spatial dose calculations

We can turn this on in SADA V5 if there is interest

26

Map Result Screens and Ratios

Screens concentration against benchmarks at each sample location

Places a box around locations that exceed benchmark

Can use one benchmark source or establish a site-specific hierarchy

27

SADA Terrestrial Dose Modeling

SADA calculates dose (mg/kg BW d) from food ingestion, soil ingestion, dermal contact, and inhalation for terrestrial exposures as well as total dose summed over all pathways selected.

SSL, Female, Male, or Juvenile

Number of different speciesUse a polygon to identify

home rangeSelect species/sexClick exposure pathwaysReturns dose in mg/kg/day for

each exposure pathway

28

Ecological Exposures Inputs and Outputs

a)

d)

c)

b)

29

Calculate Exposure for Home Range

Use a polygon to identify home range

Select species/sex

Click exposure pathways

Returns dose in mg/kg/day for each exposure pathway

30

Spatial Estimation + Risk/Dose MethodsCombining spatial methods with dose/risk assessment allows for decision support applications

Search Neighborhood

Unsampled Point

31

Decision AnalysisSpatial Screens

Sampling Strategies

Cost Vs Risk Reduction

Spatial Risk

Area of Concern

Cost Benefit

32

Selective Remediation

Process that achieves a local- and/or site-specific concentration level while minimizing cleanup volume.

Implementation requires:- data- spatial interpolation model- decision-maker cleanup criteria- spatial scale inputs

Results in a spatially explicit remedial design

33

Determining Areas of Concern

Map of interpolated concentration values can be compared to ecological or human health risk criteria to develop areas of concern

SADA reports area or volume of exceedance and coordinates or areal extent

34

Cost-Risk Curves for Soil/Sediment remediation

35

3D Visualization

True 3d Views: Points, Blocks, and Isosurfaces

Source term definition, etc.

36

If you don’t like all that, at least I have a hard copy of Till & Meyer for Nick

Although this work was reviewed by EPA and approved for presentation, it may not necessarily reflect official Agency policy. Mention of trade names or commercial products does not constitute endorsement or recommendation for use.