1. Francisco X. Aguilar, Michael Goerndt, Stephen Shifley, Nianfu Song
Department of Forestry
The School of Natural Resources
University of Missouri

2.  Logging residues
 Removal of excess biomass (fuel treatments)
 Fuelwood from forestlands
 Primary and secondary wood processing mill
residues and pulping liquors
 Urban wood residues
 Dedicated energy plantations

3.  Climate change: Reduce CO2 emissions from
fossil fuels.
 Rules and regulations: E.g. Renewable Portfolio
Standards.
 Economics: Relatively low cost for conversion to
co-firing compared to other renewable energy
(e.g. wind, solar, liquid biofuels)
 Forest stewardship: e.g. Promote forest health

4.  Estimate potential for co-firing of biomass in
existing coal-fired power plants for the U.S.
Northern Region
 Use results to establish a “coarse screen” for
county-level potential of co-firing biomass for
electricity based on physical factors

5.  County-level is smallest practical scale for
estimation, given restrictions on estimation of
explanatory factors (e.g. infrastructure,
waterways, biomass resource availability).
 Potential for co-firing can be indicated by
estimated presence (probability estimate>0.5).
5

7.  Important Issues:
◦ Possible spatial interdependence
◦ Dependence of county-level co-firing on
presence of coal-fired power plant(s)

8.  Theoretical Framework
◦ Natural conditionality of co-firing on presence of
coal-fired power plants.
◦ Probability of co-firing y within the ith county is
conditional on the expected probability of a coal
power plant in the same county (E[ci]) & other
location factors captured in an information factor
matrix X.
◦ Prob(yi=1| E[ci], X) = F(E[ci|Lα], Xβ)

9.  County-level probability for placement of coal-fired
power plants was analyzed as a first stage (Model A)
 Two models created for final stage (co-firing
probability (potential))
1. Model B: Known coal power plant frequency
included as independent variable
2. Model C: First stage (Model A) estimates included
as independent variable

10.  Standard probit regression
◦ Assumes binary response (0,1)
◦ Does not account for spatial dependencies
 Bayesian spatial autoregressive probit
◦ Assumes binary response (0,1)
◦ Accounts for spatial dependencies
 Preliminary Chi-squared tests conducted on
dependent variables for spatial dependence
prior to assessing Bayesian spatial
autoregressive probit

11.  Dependent
◦ Location of coal-fired power plants & co-firing
status (EPA, DOE)
 Independent
◦ Electricity demand (EIA)
◦ Infrastructure (EPA, US Census)
◦ Coal availability and price
◦ Renewable energy policy
◦ Resource availability of biomass (TPO, NASS)
◦ Sub-regional variation

12.  Energy demand
◦ Population
◦ County area
 Infrastructure
◦ Rail presence
◦ Road presence
◦ River & stream presence
 Renewable energy policy
◦ Renewable energy portfolio standards (RPS) by 2001
 Resource availability of biomass
◦ Wood mill residues
◦ Corn yield (stover)

13.  Spatial autoregressive probit: no significant
improvement over standard probit
 Energy demand proxies such as county area &
urban percentage of county area were highly
significant
 Infrastructural proxies of road presence &
stream presence (namely road x stream
interaction) were highly significant.

14.  Known frequency of coal-fired power plant
highly significant.
 Significant proxies
◦ Electricity Price
◦ Rail Presence
◦ Road presence x stream presence
◦ Wood mill residues
◦ RPS implementation
◦ One sub-regional indicator

15.  Component from Model A not significant
 Significant proxies
◦ Rail Presence
◦ Road presence x stream presence
◦ Wood mill residues
◦ RPS implementation
◦ Two sub-regional indicators

16.  5 counties with
high potential
but no current
co-firing
facilities
 Indicated
counties have
high values for
electricity
demand,
infrastructure &
mill residues
 Model success
rate = 96%

17.  3 counties with
high potential
but no current
co-firing
facilities
 Indicated
counties have
high values for
infrastructure &
mill residues
 Model success
rate = 96%

18.  Notable positive relationship between electricity
price and probability of co-firing biomass
 Adoption of RPS was significant for both final
models, denoting a strong relationship between
energy policy and co-firing
 Counties identified by Models B & C had fairly high
values for relevant infrastructure and biomass
supply (mill residues)

19.  Inclusion of known coal-fired power plant
frequency in Model B did not decrease significance
of infrastructural proxies
 Infrastructure variables such as road presence are
vital to co-firing operations with or without current
presence of coal-fired power plants
 Sub-regional variation has a greater effect on co-
firing probability in the absence of known coal-
fired power plant frequency

20.  Physical potential of co-firing biomass is highly
influenced by variables indicating
 Supply infrastructure
 Current availability of wood mill residues
 Implementation of RPS has a significant positive
effect on co-firing
 Valuable county-level preliminary examination of
co-firing potential across the Northern region.

21. Dr. Michael Goerndt
goerndtm@missouri.edu
Department of Forestry
University of Missouri