Assessing impact of uncertainties in decline curve analysis through hindcasting

Research output: Contribution to journalArticlepeer-review

3 Scopus citations


This study attempts to characterize the uncertainties in production forecasting for unconventional gas wells through hindcasting. The analysis focuses on how results from decline analyses change as the well population ages. The study is based on publicly available and subscription-based production data on 34,000 shale gas wells in the Marcellus, Fayetteville, Haynesville and Barnett plays. A hindcasting analysis was performed on wells, where production has been forecast using a physics-based decline curve analysis. This allows the accuracy of decline methods to be tested and the nature of errors to be assessed. Based on this analysis, the study identified systematic and random errors that impact estimates of ultimate recovery. Fields where few wells have entered boundary-dominated flow show higher uncertainty for field-wide production forecasts. Uncertainty decreases as more production history is accumulated. The physics-based approach for decline curve analysis is shown to be robust through hindcast analysis. The uncertainty in per-well recoveries for the fields studied are in the range 4–8%. The uncertainty in field-wide average ultimate recovery is less than 4% for three of the fields and 5.6% for the Marcellus. Understanding the sources and magnitudes of errors and uncertainties in estimated ultimate recovery values for unconventional gas wells allows operators to account for these in determining economic outcomes and performing financial planning of wells.

Original languageEnglish (US)
Pages (from-to)340-348
Number of pages9
JournalJournal of Petroleum Science and Engineering
StatePublished - Jan 2019

All Science Journal Classification (ASJC) codes

  • Fuel Technology
  • Geotechnical Engineering and Engineering Geology


Dive into the research topics of 'Assessing impact of uncertainties in decline curve analysis through hindcasting'. Together they form a unique fingerprint.

Cite this