hf chemicals reporting: summary of fracfocus analysis · hf chemicals reporting: summary of...
TRANSCRIPT
HF Chemicals Reporting: Summary of FracFocus Analysis
Groundwater Protection CouncilSeptember 29, 2015 Oklahoma City
Kate Konschnik and Archana Dayalu
FracFocus Data
28 states require some HF disclosure
23 states and the federal BLM require or recommend disclosure directly to FracFocus
FracFocus provides well-by-well pdfreporting.
As of May 2015, FracFocus also provides aggregated well data for download.
FracFocus Data
FracFocus Data:RegistryUpload (Header) Data
Equivalent SQL Header Data
JobStartDate
JobEndDate
StateName
CountyName
APINumber
OperatorName
WellName
Longitude
Latitude
Projection
FederalWell
TVD
TotalBaseWaterVolume
TotalBaseNonWaterVolume
Additional Header Data
OriginalSubmitTimeStamp StateNumber
SubmitTimeStamp CountyNumber
Amended WellType
FFVersion pKey
PDF Form Header Data
* Header Fields Added During Our Analysis *
pKeyRegistryUpload search_id_form
submit_time_fracday LatitudeClean
StateNameFromCoords LongitudeClean
StateAbbFromCoords ProjectionClean
CountyNameFromCoords StateOK
CountyCodeFromCoords
StateNameFromAPI
FracFocus Data: Purpose and Ingredients Data
Equivalent SQL Purpose Data
TradeName
Supplier
Purpose
Equivalent SQL Ingredients Data
IngredientName
CASNumber
PercentHighAdditive
PercentHFJob
IngredientComment
Additional SQL Ingredients Data
IngredientMSDS MassIngredient
ThirdPartyCalc pKey* Purpose Fields Added During Our Analysis *
SupplierClean
Systems_Approach
pKeyPurpose
Additional SQL Purpose Data
pKey
* Ingredients Fields Added During Our Analysis *
CASNumberClean
IngredientNameClean
CASLabel
WithheldFlag
pKeyIngredients
FracFocus Data
Our original dataset: 97,964 forms.
After removing duplicates, earlier disclosures:
- 96,449 forms, submitted
- March 9, 2011 – April 13, 2015 in
- 23 states, including
- 21 with HF disclosure laws during the study period.
EPA FF 1.0 Study – forms submitted between January 11, 2011 – February 28, 2013, in 20 states.
FracFocus Data
After removing 149 additional forms for state location discrepancy, our data set contained:
96,300 forms, describing frac jobs at
92,844 wells, including
53,073 FF 2.0 forms, which described
1,979,128 rows of chemical information
EPA Study – 38,000 FF 1.0 forms
Snapshot of Data
Snapshot of Data:Approximate distribution of forms,
through Feb 2015*
Disclosures per county
01 - 1011 - 100101 - 500501 - 10001001 - 5857
*Note 1: Alaska counties are not displayed in this map. *Note 2: Distribution assumes user input county codes are accurate, which is not necessarily true.
Snapshot of Data – Top Inputs after Water, Sand
Inquiry 1: Location Data Entry Error
FF 1.0 Data – 22% error in State Names
FF 2.0 Data – 0% error in State Names
- Why? Field automatically populated in 2.0.
State discrepancies in well coordinates and API state code, for 0.2% of forms (149).
- All but 5 were FF 1.0 forms.
- Why? API input broken out into 3 parts in 2.0.
Inquiry 1: Location Data Entry Error
API: 05029061090000 (Colorado)Coordinates:Pacific Ocean
API: 42283334270000 (Texas)Coordinates:Mexico (close to USA border)
API: 47097037680000 (West Virginia)Coordinates:Pennsylvania
API: 04213326330000 (California)Coordinates:Texas
Location Error Examples – State API does not match user-input coordinates(Images © Google Maps 2015)
Inquiry 1: Location Data Entry Error
County discrepancies in well coordinates and API state code: 3.9%
- FF 1.0 data: 5.0%
- FF 2.0 data: 2.9%
Why? API input broken out into 3 parts in 2.0.
Inquiry 2: CAS # Data Entry Error
Almost 80% of CAS #s were valid, identifying 983 unique chemical substances.
- About 0.75% (14,881 entries) had data errors.
EPA FF 1.0 study: 65% of CAS #s were valid, identifying 692 chemical substances.
- Why? FF 2.0 warned if check digit did not pass verification.
- Some data scraping errors
Inquiries 1 and 2: Data Entry Errors Recommendations /Lessons
Automatic field entries, prompts, can flags can reduce data error
FF 2.0 began using these tools:- Separate boxes for each numeric series in an API or CAS #- Notification of a discrepancy between location data points- Warning if CAS # check digit does not pass verification
FF 3.0 will expand on these tools:- Error message if CAS# check digit does not pass verification- Drop-down menus of substances
Inquiry 3: Timeliness
Effect of reporting deadlines
- Before, operators took a weighted average of 83 days to file a form with FracFocus.
- After, operators took a weighted average of 33 days to file a form.
Generally low compliance rates until states announce intent to enforce (ex. Colorado “late rate” dropped from 36.1% to 5.2%).
Inquiry 3: Timeliness
State Laws: Companies must report within …
Day 1 16 31 46 61 76 91 106 121
Drilling HF WC
LA
AL, AK, MS, WY
AR, MT, SD, TX
NM
CA, IL, NE, NV, OK, PA, UT
60 days of WC
90 days of WC
120 days of spud - KS
20 days of WC
30 days of HF
30 days of WC
45 days of WC
60 days of HF
CO, ND, OH
WV
Inquiry 3: Timeliness
Hard to track deadlines, applicability
Date clock starts may not be reported to FF (spud date, completion end date).
FF 1.0, 2.0 forms did not report submission date.
States with pre-fracturing disclosure rules see lower compliance rates (38.3%) with post-fracturing disclosure deadlines than states without pre-fracturing rules (50.7%).
Inquiry 3: Timeliness Recommendations/Lessons
Indications a state takes a requirement or deadline seriously seems to have a strong effect on the regulated community
Education on deadlines, reasons could help.
Relevant dates for deadlines should be public. FF 3.0 will publish submission date.
Inquiry 4: Data Withholding
Median percentage of ingredients withheld per formError bars represent 25th and 75th percentiles.
Inquiry 4: Data Withholding
Median percentage of ingredients withheld for proprietary reasons per formError bars represent 25th and 75th percentiles.
Inquiry 4: Data Withholding
Proprietary
Confidential
Trade Secret
N/A
41%
28%
18%
12%
Justification of Non-Disclosure
Inquiry 4: Data Withholding
Justification of Non-Disclosure
Inquiry 4: Data Withholding
EPA: 11% withholding rate in forms filed before February 28, 2013.
GWPC: 16.7% withholding rate in forms filed June – December 31, 2013.
Harvard: 16.6% withholding rate in forms filed January 1, 2014 - April 13, 2015.
(NOTE: These rates exclude N/A designations.)
Inquiry 4: Data Withholding Recommendations /Lessons
Different state procedures do not drive differences in disclosure rates.
States that permit withholding for non-trade secret proprietary reasons do not see higher rates of withholding.
Re-allocation of justification occurs based on state law.
Education on standards needed.
Inquiry 5: Systems Approach
Company # of Regular Records
WH Rate
# of SA Records
WH Rate (Difference)
2-sided p-value
95% CI
Schlumberger 47,991 10.7% 127,799 2.9% (7.8%) <0.001 8.1% -7.6%
Baker Hughes 169,229 17.8% 12,432 3.9% (13.9%) <0.0001 14.3% -13.5%
Halliburton 232,618 12.1% 315 17.8% (5.6%) <0.0001 9.8% -1.4%
Multi-Chem 20,727 21% 845 43% (22%) <0.0001 25.2% -18.9%
Inquiry 5: Systems Approach
141,391 unique systems approach (SA) records, representing 7.1% of the data set.
Systems approach reporting reduced withholding rates by four-fold (14.4% to 3.3%)
Overall, lower rates of withholding by these firms than across all forms.
Inquiry 5: Systems ApproachRecommendations/Lessons
Systems Approach reduces withholding rates
More research needed to determine if smaller firms would use systems approach or withhold fewer records as a result
States should retain authority to ask for product-by-product chemical information, in case of release, or for other reasons.
Questions?
Kate Konschnik, Director
Harvard Environmental Policy Initiative
Archana Dayalu
Harvard SEAS