Economic Decision Making and the Application of Nonparametric Prediction Models

Abstract

Sustained increases in energy prices have focused attention on gas resources in low-permeability shale or in coals that were previously considered economically marginal. Daily well deliverability is often relatively small, although the estimates of the total volumes of recoverable resources in these settings are often large. Planning and development decisions for extraction of such resources must be areawide because profitable extraction requires optimization of scale economies to minimize costs and reduce risk. For an individual firm, the decision to enter such plays depends on reconnaissance-level estimates of regional recoverable resources and on cost estimates to develop untested areas.This paper shows how simple nonparametric local regression models, used to predict technically recoverable resources at untested sites, can be combined with economic models to compute regional-scale cost functions. The context of the worked example is the Devonian Antrim-shale gas play in the Michigan basin. One finding relates to selection of the resource prediction model to be used with economic models. Models chosen because they can best predict aggregate volume over larger areas (many hundreds of sites) smooth out granularity in the distribution of predicted volumes at individual sites. This loss of detail affects the representation of economic cost functions and may affect economic decisions. Second, because some analysts consider unconventional resources to be ubiquitous, the selection and order of specific drilling sites may, in practice, be determined arbitrarily by extraneous factors. The analysis shows a 15-20% gain in gas volume when these simple models are applied to order drilling prospects strategically rather than to choose drilling locations randomly.

Combining Geostatistics With Bayesian Updating To Continually Optimize Drilling Strategy in Shale-Gas Plays

“…However, the wells analyzed here represent only the eastern part of the Barnett shale, whereas Fan et al (2011) used data from horizontal wells that are across the entire Barnett shale. Baihly et al (2010) analyzed 838 horizontal wells from the core of the Barnett shale and reported an EUR of 3.0 Bcf. Ezisi et al (2012) reported EUR of 2.4 Bcf in Johnson County and 3.4 Bcf in Tarrant County.…”