Current Research Projects
Forest Transportation Planning
Transportation of forest products from harvesting sites to mill destinations is one of the largest cost components of forest operations. Typically, the goal of forest transportation planning problems (FTPP) has been to find the set of routes that minimize total variable (hauling) and fixed (road construction and/or maintenance) costs. As these problems are NP-hard, exact algorithms such as mixed-integer programming can only be applied to solve small- and medium-scale problems. Approximation algorithms are usually applied to solve large-scale, real-world problems.
Ant colony optimization (ACO) has been successfully applied to solve numerous optimization problems and because FTPP can be naturally modeled using networks, ACO is a promising technique. We have developed ant colony optimization (ACO) algorithms to solve medium-scale FTPP that provide near-optimal solutions. These ACO algorithms have also been applied to solve large problems with thousands of road segments, hundreds of timber sale locations as well as multiple destinations, products, and time periods.
One of the largest limitations of ACO, and other approximations algorithms is that solution quality is heavily dependent of its parameter values. Thus, we are currently developing cutting-edge techniques to develop parameter-free ACO algorithms. We expect these new algorithms to serve a general approach to solve very large-scale, real-world problems while ensuring high-quality solutions and fast-computation time.
* Lin P., J. Zhang, M. Contreras. 2014. Automatically configuring ACO using multilevel ParamILS to solve transportation planning with underlying weighted networks. To appear in Swarm and Evolutionary Computation.
* Contreras M., W. Chung, G. Jones. 2008. Applying ant colony optimization meta-heuristic to solve forest transportation planning problems with side constraints. Canadian Journal of Forest Research 38(11):2896-2910.
Skid-trail Network Design
Skid-trail locations directly influence the economics and environmental impacts of ground-based harvesting operations. Typically, field managers design skid-trail networks manually based on field observations of vegetation and terrain conditions. We developed a model to automatically design skid-trail networks to reduce skidding costs and soil disturbances. The model simulates log-piles locations, creates a feasible skid-trail network across the harvest unit, estimates skidding cost and soil recovery cost for each feasible skid-trail segment, and finds the network design that connects each log-pile to extraction areas while reducing skidding and soil recovery costs.
We have succefully applied the model to find the optimal skid-trail network on several hypothetical harvest units representing conditions commonly found in timber harvesting operations (e.g. skidding pattern, uneven volume distribution, skidding obstacles, and different weights given to skidding and soil recovery costs). We have also applied the model to an actual harvest unit to evaluate its ability to design skid-trail networks considering more realistic conditions with multiple design factors.
We are conducting research to compare operator-designed and computer-generated skid-trail networks to evaluate the potential benefits of using the model.
* Contreras M., W. Chung. 2011. A modeling approach to estimating skidding costs for individual trees for thinning operations. Western Journal of Applied Forestry 26(3):133-146.
Dedicated Energy Woody Crops
Biomass has gained considerable attention because of its potential to partially replace fossil fuels and develop a sustainable bioenergy industry. Dedicated energy crops could offer a reliable and sustainable biomass supply, but there is limited research identifying economically suitable sites to establish these crops. We developed a spatially-explicit model to identify such sites based on biomass yield and break-even biomass amount, derived from production costs and delivered biomass prices. Sites where biomass yields exceed break-even biomass amounts are considered economically suitable.
We have applied the model to a four-county study area in northern Kentucky with a diverse land cover and ownership, relatively extensive transportation network, and presence of existing conversion facilities, conditions which are common in the Ohio River Valley region and much of the southern US.
We are currently conducting research to use the model to identify market scenarios and policy incentives that can most effectively promote bioenergy production, in terms of the area economically suitable to establish these crops.
* Nepal S., M. Contreras, J.M. Lhotka, G.A. Stainback. 2014. A spatially explicit model to identify suitable sites to establish dedicated woody energy crops. Biomass and Bioenergy, 71:245-255.
Forest Operations Safety
The forest industry is based on timber harvested by thousands of relatively small family-owned, local contractors (loggers) encompassing a wide range of size and business prowess. Unfortunately, the forestry workers carrying out these harvesting operations experience rates of occupational injury, illness, and death over six times higher than workers in other sectors. Several factors contribute to these high occupational injury rates, but the isolated and remote working conditions is one of the most important because it typically leads to long response times. We developed web-based application to automatically identify least-travel-time routes from emergency service locations to timber harvesting sites. This goal is driven by the need to reduce critical response time to isolated working sites, which can help reduce the fatal and serious occupational injury rates in the farming, fishing, and forestry occupational sector.
We have developed an application for a pilot area covering 53 counties in eastern Kentucky and are planning to extend the application to the twelve southeastern US.
* Emergency Personnel Location Route Founder http://www.ca.uky.edu/forestry/LoggingEPLroutes/