us army corps of engineers building strong ® geophysical classification at the camp sibert remedial...

31
US Army Corps of Engineers BUILDING STRONG ® Geophysical Classification at the Camp Sibert Remedial Action Amy Walker Geophysicist & Innovative Technology Advocate USACE, Huntsville November 20, 2013

Upload: domenic-chambers

Post on 15-Jan-2016

215 views

Category:

Documents


0 download

TRANSCRIPT

PRESENTATION TITLE

Geophysical Classification at the Camp Sibert Remedial Action

Amy WalkerGeophysicist & Innovative Technology AdvocateUSACE, Huntsville

November 20, 2013

US Army Corps of EngineersBUILDING STRONG1OutlineSite overviewPrep: WP, Seeding, Test StandClassification MethodResults QC/QA: IVS results, seed results, Validation digging

2BUILDING STRONG

37,000 acre FUDS near Gadsden, ALUsed from 1942-1945 as a Replacement Training Center (RTC) for the Army Chemical Warfare Service.4.2-inch mortar used in most training (primarily HE & smoke/WP/tearing agent)Current Use: mostly private ownership, used for grazing, woodlands, agriculture & residencesUSACE involvement since 1990 (records research, aerial photography analysis, site characterizations, EECAs, removal actions, RIFS)

Former Camp Sibert3BUILDING STRONG4

Remedial Action SitesBUILDING STRONG4Site 18Site 18- 400 acres (~118 open) in impact area, used for training to attack Japanese-type pillboxes; 4.2 mortars (HE & WP)Currently used as wildlife/hunting area with some agriculture2007- EECA & 1st ESTCP Munitions Classification Demo (thousands of digs- 2 UXO) Range 28A- Adjoining, same impact area, ~40 acres open pasture2010-RIFS- better define boundaries

5BUILDING STRONGHow did we decide to use Classification?Over the years, lots of discussions about classification given the ESTCP success here (Easiest site in the country)Benefits: easy site to show success (well characterized, only 4.2 mortars), minimize digs/disruption for landownersParsons under old RA contract - at award, classification capabilities not well understoodSince then- Several successful ESTCP demonstrations by Parsons; MetalMapper available as GFETeam decided to perform MM classification under existing contract (FFP/PBC) - Final WP allowed for possibility with WP Addendum

6BUILDING STRONGDefine TOIA lot of data : Only 4.2 mortars, MEC to 14 on Site 18, Site 17- one item at 40 (swampy)MD - 90% in upper foot, all >1m between picks)

Yes8BUILDING STRONGClassification PlanWork Plan AddendumBased on Parsons ESTCP Dem Plan- Using same methods for MetalMapper data collection and analysis as successfully demonstrated at other more difficult sitesPerformance metrics/success criteria established based on ESTCP demos, lessons learned and modified as needed for leaving metal Blind seedingADEM comments incorporatedNo additional large-scale demo, but dig initial 200 to verify process early on, present results and plan for additional validation diggingConservative classification decisionsLooking for additional items found elsewhere at Sibert (2.36 rockets, Livens)Looking for deep items dig decisions based on conservative thresholds

9BUILDING STRONGBlind QA SeedsInert 4.2 mortarsUnknown to ParsonsBlind QC SeedsMedium ISOs used for testing EM61 data response and positioning repeatabilityUnknown to MM analyst and EM61 analystKnown QC SeedsMedium ISOsUnknown to EM61 analystKnown to MM analyst to provide ongoing feedback of data analysisSeeding

10BUILDING STRONGTest Stand

Varieties of inert 4.2 mortars (with/without noses, base plates, flattened, etc.) All Matched Data at 36 & 42 starts to look noisy, matches Livens better than 4.2mortar, but still meets dig criteria at 48- not clearly detectable in MM data11BUILDING STRONGTest Stand/GPO

4.2 mortar base plate4.2 mortar12

4.2 mortar half shellBUILDING STRONG

Site 18pondData gaps, saturated areas, woods- Mag & Dig13BUILDING STRONGEM61: Target center- aka the perfect place for a hunting lodge high density (construction debris & MD)- polygonsMM Cued ~6000 targets13Classification Method

Dig:3-curve Library match > 0.575 2-curve Library match > 0.71-curve Library match > 0.8QC Digs:Targets added at the discretion of the analyst- noisy data with confidence metrics close to the thresholds or based on location within feature space Cant Analyze targetsAll 3 curves identified as poor by analyst/ bad fitEM61 data looks real (not noise spike, anomaly present on multiple lines)Validation digs:Selected by USACE/ADEMNo dig: Targets not meeting the above criteria14BUILDING STRONG14Site 18 Classification Results

15BUILDING STRONGCould easily have dug only 3-4%15Site 18 Dig Results

16BUILDING STRONGThe relatively high number of NC and Hot Soil digs resulted mostly from QC digs, which were added in an attempt to recover very deep mortars. As seen in the test stand data, when 4.2 mortars are deeper than ~1m, the library match degrades; however, they still have a large size. When there is no metal and/or hot soil is present, the results can appear similar to very deep items.

16ROC Curve (Site 18 & 28A)UXO (all 0.99 matches)Not dug (presumed clutter) & 500 validation digs (known clutter)17BUILDING STRONG-only 3-4 clutter digs hiding in the straight vertical lineNot dug in this order- field efficiencies, certain areas first for planting schedules-Val digs near threshold, but also throughout- will describe in more detail later17ROC Curve18BUILDING STRONGUXO ClassificationAll three recovered UXO (4.2 mortars) had Library Match of 0.99Able to inform field crew of last expected UXO - assist in scheduling

19BUILDING STRONGAfter analyzing the classification results, including seeds, USACE was able to inform the field team of a high likelihood UXO item. This information was especially useful to the team because the 4.2 mortars at Site 18 are potentially liquid filled (likely tearing agent), which requires a series of notifications and additional security measures, including scheduling EOD availability. Knowing in advance that the item might be a suspect liquid filled item, the team was able to check with EOD ahead of time and schedule the dig based on their availability. This helped to minimize exposure to the item, impacts to the property owner, and costs associated with security.

19Site 18 Intrusive Results

20BUILDING STRONGThis result confirms the conceptual site model, which was based upon previous investigations. Inert 4.2 mortar blind seeds were placed at depths ranging to 43 inches, which represented a much more difficult classification problem than the UXO, but were all classified to dig.

2021BUILDING STRONGFeature Space plot showing Size versus Decay calculated from s. Camp Sibert data shows excellent separation between TOI and clutter, with evident clusters due to common 4.2 mortar munitions debris. Targets (black dots) in 4.2 mortar cluster without corresponding mortar are due to multiple picks on the same items or large non-munitions debris (e.g. fencepost).21Nose frag22BUILDING STRONG22Cluster Nose Frag

23BUILDING STRONGCluster Small Frag

24BUILDING STRONGCluster Base Plates

25BUILDING STRONGHalf Shells/Large Frag

26BUILDING STRONG27Performance Objectives

Performance ObjectiveMetricSuccess LevelSuccess AchievedCorrectly identify seed items in Instrument Verification Strip (IVS)Percentage of IVS items identified correctly98% of IVS items identified correctly with confidence metric of > 0.90 YesCorrectly positionMetalMapper relative to sourceDistance between collection location and inverted target location100% of inverted locations within 40 cm of collection point unless re-shot also outside radius1 YesCorrectly position MetalMapper relative to EM61 targetDistance between MetalMapper collection location and EM61 target location100% of collection points within65 cm of EM61 target location2YesMaximize TOI retained on dig listPercentage of TOI identified as dig targets.100% of TOI identified as dig targets3 YesMinimize non-TOI retained on dig listPercentage of false alarms eliminated.75% of non-TOI left in groundYesCorrectly identify type of TOIPercentage of TOI correctly identified by group475% of TOI identified correctlyYesClassify type of non-TOIPercentage of non-TOI correctly classified by size and shape575% of non-TOI classified correctly with regard to size and shape6 YesCorrect estimation of target locationAccuracy of estimated target parameters for dig list targets marked as dig.X, Y < 30 cm (1s) Z < 15 cm (1s) YesBUILDING STRONG27Validation DiggingInitial 200 (early verification)- results presented to ADEM (QC metrics pass, seeds classified, dig results make sense)Remainder selected by USACE and ADEM, final list concurrently, while reviewing data in UX-Analyze. Based on: (1)proximity to contractor dig thresholds, (2)location in feature space, and (3)location in large anomalous features where classification may be more difficult. No TOI recovered, results make sense (hot soil results for potentially large/deep items in feature space, base plates correctly labeled, etc)Retrospect- too much (nothing looked suspicious)

28BUILDING STRONG28Non-TOI Results make sense

0.86 match to MK2 Hand Grenade0.87 match to 2.36 rocket

Small, symmetric, slow decay0.90 match to 60mm mortar

29BUILDING STRONGDidnt recover TOI- results make sense; expected large MD/Base plates29ConclusionsSuccess!Using very conservative dig criteria & high % validation: 84% anomalies left undugUXO ranked very high on dig listLess disruption for landowner (land & time)PDT agrees: no increased risk of UXO remaining versus the traditional approach of digging all selected EM61 anomalies30BUILDING STRONGAcknowledgmentsJohn Baptiste, Patti Berry, Greg Nivens

Sherri Anderson-Hudgins, Karl Blankinship

Julie Ange, Stephen Cobb

Jim Pastorick

Herb Nelson

31BUILDING STRONG