Source: FOREST AND WILDLIFE RES CENTER submitted to
COMPUTER APPLICATIONS IN FORESTRY AND TECHNOLOGY TRANSFER
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
TERMINATED
Funding Source
Reporting Frequency
Annual
Accession No.
0174133
Grant No.
(N/A)
Project No.
MISZ-069090
Proposal No.
(N/A)
Multistate No.
(N/A)
Program Code
(N/A)
Project Start Date
Jul 1, 2003
Project End Date
Jun 30, 2009
Grant Year
(N/A)
Project Director
Schultz, E. B.
Recipient Organization
FOREST AND WILDLIFE RES CENTER
(N/A)
MISSISSIPPI STATE,MS 39762
Performing Department
FORESTRY
Non Technical Summary
As the need for better natural resource utilization increases and greater pressure is placed on state and federal dollars, computer-aided forest technology will provide an economical alternative to conventional methods of research and technology transfer. New techniques in artificial intelligence are being used to model forest systems where traditional modeling techniques may be inadequate because of complicated interactions or lack of suitable mathematical models.
Animal Health Component
90%
Research Effort Categories
Basic
10%
Applied
90%
Developmental
(N/A)
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
1230611208020%
1230612208040%
1230613208040%
Goals / Objectives
The overall objective is to provide Mississippi forest landowners, forestry professionals, public officials, and researchers with readily available and easy to use computerized systems for forest technology. Emphasis will be placed on applications in remote sensing and growth and yield predictions.
Project Methods
Artificially intelligent computer modeling techniques will provide innovative solutions or substantially enhance current methods. Expert and decision support systems will be developed to represent expert knowledge to assist landowners, professional foresters, researchers, and public officials in management practices and decision-making.

Progress 07/01/03 to 06/30/09

Outputs
OUTPUTS: In a cooperative effort between Mississippi State University and the Mississippi Institute for Forest Inventory (MIFI), a spatially-based, large-scale forest inventory data collection system and associated reporting software have been developed. The Web-based software allows the user to estimate volume and biomass by county or polygon with precision statistics appropriate to +- 15% precision at the 95% confidence level. Growth, damage and other reports are also available. New inventory procedures have been adopted and implemented under this project to allow for storm and other damage assessments, locating and quantifying invasive species, and the calculation of biomass and carbon sequestration. A new model for statistically efficient strata has been developed and is ready for implementation in the second round of the MIFI inventory. The statically efficient strata allow precision gains due to stratification and result in decreased sample size and cost savings. To make timber growth and yield models applicable to projecting growth and drain from large area inventories, procedures are needed to not only project stand growth from current volumes but allocate them spatially as well. A procedure was developed for using Landsat Thematic Mapper (TM) imagery to predict the spatial distribution of clear-cut timber harvests and for improving methods for allocating harvests and their associated volumes back to inventory information. The resulting harvested and forested area predictions can then be used to address specific problems such as the availability of forest feedstocks for existing and future mills, creation of transportation networks needed to supply raw materials, and environmental and aesthetic issues such as the assessment of the quality and spatial components of wildlife habitat concerning corridor continuity and fragmentation, and adjacency standards and clear-cut size limits for sustainable forest certification programs. A Web-based growth and yield system capable of accurately predicting the yields and value by species groups, size class, product, and grade for the Southern red oak-sweetgum bottomland hardwood forest has been completed. This type of tool provides landowners and forest product industries with the appropriate information upon which to base management and economic decisions necessary to achieve their objectives. Program models are the result of a 28-year growth and yield research effort between MSU and the US Forest Service. Web-based tree biomass and carbon estimation system (Carbon Calculator) software has been developed providing southeastern forest landowners with the biometrical tools necessary to help them evaluate their forest inventories for sequestered carbon. The software not only estimates carbon but also provides biomass, volume, and dry and green weight estimates of individual trees to any top diameter limit. Users may select biomass components for 347 species/region combinations and obtain information on: 1) tree profiles, 2) biomass/carbon/volume tables, and 3) equations to approximate the tables. PARTICIPANTS: Individuals: Emily B. Schultz, Principal Investigator - Inventory, Biometrics, and Computer Applications. Partner Organizations: Mississippi Institute for Forest Inventory, US Forest Service, Stephen F. Austin State University, Louisiana State University. Collaborators: Thomas G. Matney, Co-Investigator - Inventory, Biometrics, and Computer Applications; Robert C. Parker, Co-Investigator - Inventory, Spatial Technologies; David L. Evans, Co-Investigator - Spatial Technologies; Donald L. Grebner, Co-Investigator - Forest Economics; H. Alexis Londo, Research Associate - Biometrics, Spatial Technologies; Curtis A. Collins, Research Associate Biometrics, Spatial Technologies; J. Paul Jeffreys, Research Associate - Spatial Technologies, Growth and Yield; W. Cade Booth, Research Associate - Growth and Yield; Gustavo Perez-Verdin, Post Doctorate, Mississippi State University Department of Forestry - Forest Management/Biomass; Patrick A. Glass, Director of Operations, Mississippi Institute for Forest Inventory - Operations, Biometrics; Wayne Tucker, Executive Director, Mississippi Institute for Forest Inventory - Operations and Program Coordination; Quang Cao, Professor, School of Renewable Natural Resources, Louisiana State University - Biometrics. Training or Professional Development: Graduate Students: Garrett Nowell - Spatial Technologies; Curt Collins Modeling/Spatial Technologies; Mike Jackson - Modeling/Inventory; Luke Jones - Inventory/Economic Development; Wesley Howard - Biometrics/Modeling; Kit Posadas - Modeling; Michael Crosby - Spatial Technologies/Biometrics; Teresa Arnold - Remote Sensing; Tabatha Kelly - Remote Sensing; David Wilkinson - Remote Sensing/Inventory; Patrick Glass - Biometrics/Remote Sensing; Alexis Londo - Biometrics/Remote Sensing; Clint Iles - Biometrics; George Banzhaf - Biometrics. Post Doctorates: Li Gan, Post Doctorate - Computer Applications; Gustavo Perez-Verdin, - Forest Management/Biomass TARGET AUDIENCES: The target audiences include landowners, state planners, forest industry, researchers, and other local, state, and federal agencies concerned with forest management and utilization. The Web-based Mississippi Institute for Forest Inventory (MIFI) interface, MIFI annual reports, growth and yield software, and publications are the primary instruments that deliver the science-based knowledge to the target audiences. The delivered information will be helpful in economic development, national biomass/energy assessments, change detection, disaster and invasive species assessments, reforestation and land use planning, hardwood management decision making, landowner education, policy decisions, and wildlife habitat assessments. The MIFI tools also allow the state to entice forest products industry investors to build or expand existing mills while monitoring the sustainability of the resource. Carbon/biomass and growth and yield software will aid landowners in participating in the bioenergy and carbon markets and in hardwood management decision making. PROJECT MODIFICATIONS: Not relevant to this project.

Impacts
MIFI Dynamic Inventory Reporter software demonstrated that for much of the State more wood is being produced than consumed and this surplus can add significantly to the existing $14 billion total output of Mississippi's forest and forest products industry. Good growth and drain estimates guarantee the stability and sustainability of the forest products industry that accounts for 6.6% of the State's total industrial output. The MIFI Dynamic Inventory Reporter system is an effective tool for economic development and resource sustainability and has been used by MIFI in responding to 1) 150 requests for installation or analytical support, 2) 93 detailed analyses including 28 analyses for very specific combinations of species product availability including diameter distributions, growth-to-drain ratios and spatial distribution, 3) 25 reports for cellulosic facilities and 23 pellet facilities pursuing capital funding for locating mills, 4) decision-making analyses by two traditional forest products companies poised to locate in MS, and 5) requests for 9 electrical generating biomass facilities for determining location. Methods for producing statistically efficient strata for the MIFI inventory have reduced the number of plots required to meet the required precision (+- 15%) and confidence (95%) levels by 20% and will create a cost savings of approximately $30,000/year during the second round of the MIFI inventory. Software developed for a bottomland hardwood stand level growth and yield application and a volume by grade calculator may be downloaded from www.timbercruise.com (Download Center; Growth and Yield Models and Tree Volume Calculators). This Excel spreadsheet based software produces tables and graphs for three growth and yield scenarios (bare land, basal area, and inventory) and will aid landowners in hardwood management decision making. There was previously no information of this kind available for this forest type. Appropriate forest management decisions for the developing woody biofuel and carbon credit markets may now be made based on MIFI inventory data, growth and yield systems, and individual trees/tree lists. Carbon calculator software, www.timbercruise.com (Download Center, Tree Volume Table and Equation Generator), reports component tree dry weight biomass estimates. This integrated biomass/carbon calculator will provide Mississippi's state inventory system with bioenergy economic development tools and forest landowners trying to qualify for carbon credit programs with carbon estimates.

Publications

  • Banzhaf, G.M. 2009. Log grade volume distribution model for tree species in red oak-sweetgum forests in southern bottomlands [thesis]. Mississippi State University. Mississippi State, MS. 87p.
  • Grebner, D.L., G. Perez-Verdin, C. Sun, I.A. Munn, E.B. Schultz, T.G. Matney. 2009. An approach for estimating the availability, production costs, and implications of bioenergy development in the United States Mid-South. International Symposium on Emerging needs of society from forest ecosystems: towards opportunities and dilemmas in forest managerial economics, IUFRO Unit 4.05.00 Managerial Economics and Accounting,. Ljubljana, Slovenia.
  • Jones, T.L. 2009. A forest product/bioenergy mill location and decision support system based on a county-level forest inventory and geo-spatial information [thesis] Mississippi State University. Mississippi State, MS. 44p.


Progress 01/01/08 to 12/31/08

Outputs
OUTPUTS: The core code for the MIFI GIS based Mill Study and Growth and Drain modules has been finalized, and the MIFI user interface and server side components have been updated to implement the new modules. These components are a unique "first of a kind" achievement representing a major advancement in the application of forest inventory methods and theory, computer science and programming, and remote sensing and GIS technologies to automating extraction of decision support information from a spatial inventory database. One PhD student is working on procedures to improve image pixel volume estimates using change detection and spectral band data. A master's student is completing the delivered cost of raw material to proposed mill location linear programming optimization model for the transportation network. Once this model is finalized it will be added to the MIFI Decision Support Interface and System to report minimum delivered cost. Two masters students have built the primary component equations for the USFS Bottomland Hardwood Growth and Yield project. The equation system for developing a diameter distribution computer model for estimating the diameter distribution of grade volume has been completed and tested. These models have been incorporated into an Excel spread sheet interface that will serve as our preliminary technology delivery system. The final delivery vehicle will be a Microsoft Windows Application. The models for predicting individual tree grade volume distributions have been completed These models provide the basis for estimating volumes inside the hardwood growth and yield model, and they are also useful for providing clients with a Web based tool for obtaining grade volumes for individual trees. The Web based tool will be put on the CFR Web site. A new urban forest growth and yield project was funded through the Mississippi Forestry Commission Urban and Community Forestry Assistance Grant. PARTICIPANTS: Individuals: Emily B. Schultz, Principal Investigator - Inventory, Biometrics, and Computer Applications. Partner Organizations: Mississippi Institute for Forest Inventory, US Forest Service, Stephen F. Austin State University, Louisiana State University. Collaborators: Thomas G. Matney, Co-Investigator - Inventory, Biometrics, and Computer Applications; Robert C. Parker, Co-Investigator - Inventory, Spatial Technologies; David L. Evans, Co-Investigator - Spatial Technologies; Donald L. Grebner, Co-Investigator - Forest Economics; H. Alexis Londo, Research Associate - Biometrics, Spatial Technologies; Curtis A. Collins, Research Associate Biometrics, Spatial Technologies; J. Paul Jeffreys, Research Associate - Spatial Technologies, Growth and Yield; W. Cade Booth, Research Associate - Growth and Yield; Gustavo Perez-Verdin, Post Doctorate, Mississippi State University Department of Forestry - Forest Management/Biomass; Patrick A. Glass, Director of Operations, Mississippi Institute for Forest Inventory - Operations, Biometrics; Wayne Tucker, Executive Director, Mississippi Institute for Forest Inventory - Operations and Program Coordination; Quang Cao, Professor, School of Renewable Natural Resources, Louisiana State University - Biometrics. Training or Professional Development: Graduate Students: Garrett Nowell - Spatial Technologies; Curt Collins Modeling/Spatial Technologies; Mike Jackson - Modeling/Inventory; Luke Jones - Inventory/Economic Development; Wesley Howard - Biometrics/Modeling; Kit Posadas - Modeling; Michael Crosby - Spatial Technologies/Biometrics; Teresa Arnold - Remote Sensing; Tabatha Kelly - Remote Sensing; David Wilkinson - Remote Sensing/Inventory; Patrick Glass - Biometrics/Remote Sensing; Alexis Londo - Biometrics/Remote Sensing; Clint Iles - Biometrics; George Banzhaf - Biometrics. Post Doctorates: Li Gan, Post Doctorate - Computer Applications; Gustavo Perez-Verdin, - Forest Management/Biomass TARGET AUDIENCES: The target audiences include landowners, state planners, forest industry, researchers, and other local, state, and federal agencies concerned with forest management and utilization. The Web-based Mississippi Institute for Forest Inventory (MIFI) interface, MIFI annual reports, growth and yield software, and publications are the primary instruments that deliver the science-based knowledge to the target audiences. The delivered information will be helpful in economic development, national biomass/energy assessments, change detection, disaster and invasive species assessments, reforestation and land use planning, hardwood management decision making, landowner education, policy decisions, and wildlife habitat assessments. The MIFI tools also allow the state to entice forest products industry investors to build or expand existing mills while monitoring the sustainability of the resource. PROJECT MODIFICATIONS: Not relevant to this project.

Impacts
MIFI Inventory results demonstrate, that for much of the State, approximately 75% more wood is being produced than consumed and this surplus can add significantly to the existing $14 billion total output of Mississippi's forest and forest products industry. Good growth and drain estimates guarantee the stability and sustainability of the forest products industry that accounts for 6.6% of the State's total industrial output. In the 44 months since the MIFI decision support system has been operational, there have been 126 requests for installation or analytical support. From these requests, 89 cursory analyses were conducted to assess the availability of various product species combinations within the currently inventoried MIFI regions. The 37 remaining requests were directed toward intensive and detailed analyses for very specific combinations of species product availability including diameter distributions, growth-to-drain ratios and spatial distribution. Currently, two cellulosic ethanol facilities are actively pursuing capital funding for locating mills within Mississippi and several pellet manufacturing plants are analyzing the feasibility of establishing facilities; and two traditional timber product companies are poised to start construction of a facility pending a renewed vigor in their portion of the market. Software has been developed for a bottomland hardwood stand level growth and yield application and a volume by grade calculator. This software will aid landowners in hardwood management decision making. Appropriate forest management decisions for the developing woody biofuel and carbon credit markets require inventory and growth and yield systems reporting component tree dry weight biomass estimates. An integrated growth and yield and biomass/carbon calculator was developed. The objective is to provide Mississippi's state inventory system with bioenergy economic development tools and forest landowners trying to qualify for carbon credit programs with carbon estimates.

Publications

  • Perez-Verdin, G., D.L. Grebner, C. Sun, I.A. Munn, E.B. Schultz, T.G. Matney. 2008. Woody biomass feedstock supplies and management for bioenergy in Mississippi. SOFEW. Savannah, GA.
  • Perez-Verdin, G., D.L. Grebner, C. Sun, I.A. Munn, E.B. Schultz, T.G. Matney. 2008. Woody biomass feedstock supplies and management for bioenergy in Mississippi. Pages 121 - 130 in 2007 Southern Forest Economics Workshop (SOFEW) Annual Meeting. San Antonio, TX.
  • Posadas, B. K. 2008. An Application of Artificial Intelligence Techniques in Classifying Tree Species with Lidar and Multi-spectral Scanner Data [Thesis]. Mississippi State University. Mississippi State, MS. 79p.
  • Schultz, E.B., T.G. Matney, S.C. Grado, W.T. Jones, D.L. Grebner, P. Glass. 2008. A carbon estimation system: rural and urban forest case studies. 2008 Society of American Foresters National Convention. Reno, NV.
  • Crosby, M., T.G. Matney, E.B. Schultz. 2008. Multi-state forest carbon storage estimates: baselines, balances, and catastrophic losses. 2008 Society of American Foresters National Convention. Reno, NV.
  • Iles, J.C. 2008. A Stand Level Growth and Yield Model for Red Oak/sweetgum Forests in Southern Bottomlands. [Thesis]. Mississippi State University. Mississippi State, MS. 65p.
  • Jones, T.L., E.B. Schultz, T.G. Matney, D.L. Grebner, D.L. Evans, C.A. Collins. 2008. A forest product/bioenergy mill locations and decision support system based on a county-level forest inventory and geo-spatial information system. Pages 131-138 in 2007 Southern Forest Economics Workshop (SOFEW) Annual Meeting. San Antonio, TX.
  • Matney, T.G., E.B. Schultz. 2008. Deriving tree diameter growth and probability of survival equations from successive diameter distributions. Forest Science. 54(1):31-35.


Progress 01/01/07 to 12/31/07

Outputs
OUTPUTS: A doctoral student is developing GIS models to improve our ability to create statistically efficient sampling strata from GIS images in conjunction with the Mississippi Institute for Forest Inventory (MIFI). The preliminary results show that at a minimum we can increase our statistical efficiency by 25%. An operational geo-spatial database has been developed and satellite age and type classifications from 1972 to 2006 have been generated. These will provide for more accurate growth and yield projections and sampling frame construction. Change detection methods have been employed to create growth and drain layers for the 2004, 2005, and 2006 inventories in support of the addition of a mill location decision support module in the MIFI user interface. The MIFI interface and data processing dlls have been upgraded to allow the user to obtain volume information within an industrial working circle. This interface enhancement is designed to link to recently completed GIS-based growth and drain and transportation network models to generate a cost sustainability analysis report for a proposed mill location. A master's student has completed the present and future volume GIS layers and GIS age layers for implementing the growth and drain and mill location reports. Another master's student is analyzing the spatial distribution of harvest volume. A biomass knowledge base is being utilized to estimate the potential woody biomass supply for biofuel generation in Mississippi and is being adapted for use in small-scale inventories to estimate carbon credits. The inventory interface/decision support software is currently available to the public for downloading on the MIFI Web site, http://www.mifi.ms.gov. All data collection in a red oak-sweetgum growth and yield study funded by the USFS has been completed. One hundred thirty five (135) of the original 150 plots have been remeasured, and 35 new plots have been established and measured bringing the total number of plots available for remeasurement in 10 years to 170. Four hundred fifty six (456) additional trees have been graded bringing the total number of trees graded since 1982 to 2103. These data have been edited and prepared for analysis. A computer program has been written to coalesce all of various plot and tree grade measurements into a unified format and to produce the analysis-ready files required to build stand and tree level growth and yield models and individual tree grade distribution models. A master's student is developing the stand and tree level component equations for the growth and yield computer model. Another master's student has begun investigating the individual tree grade volume distributions and is assessing possible model forms and procedures that show promise for modeling the grade data. Four master's students have presented preliminary results at the 2007 Southern Forest Economics Workshop (SOFEW) Annual Meeting and the Joint Mississippi & Louisiana Society of American Foresters Annual Meeting. PARTICIPANTS: Individuals: Emily B. Schultz, Principal Investigator - Inventory, Biometrics, and Computer Applications. Partner Organizations: Mississippi Institute for Forest Inventory, US Forest Service, Stephen F. Austin State University, Louisiana State University. Collaborators: Thomas G. Matney, Co-Investigator - Inventory, Biometrics, and Computer Applications; Robert C. Parker, Co-Investigator - Inventory, Spatial Technologies; David L. Evans, Co-Investigator - Spatial Technologies; Donald L. Grebner, Co-Investigator - Forest Economics; H. Alexis Londo, Research Associate - Biometrics, Spatial Technologies; Curtis A. Collins, Research Associate - Biometrics, Spatial Technologies; J. Paul Jeffreys, Research Associate - Spatial Technologies, Growth and Yield; W. Cade Booth, Research Associate - Growth and Yield; Gustavo Perez-Verdin, Post Doctorate, Mississippi State University Department of Forestry - Forest Management/Biomass; Patrick A. Glass, Director of Operations, Mississippi Institute for Forest Inventory - Operations, Biometrics; Wayne Tucker, Executive Director, Mississippi Institute for Forest Inventory - Operations and Program Coordination; Ikuko Fujisaki, Department of Ecosystem Science and Management, Texas A & M University - Spatial Technologies; Quang Cao, Professor, School of Renewable Natural Resources, Louisiana State University - Biometrics. Training or Professional Development: Graduate Students: Garrett Nowell - Spatial Technologies; Curt Collins - Modeling/Spatial Technologies; Mike Jackson - Modeling/Inventory; Luke Jones - Inventory/Economic Development; Michael Crosby - Spatial Technologies/Biometrics; Teresa Arnold - Remote Sensing; Tabatha Kelly - Remote Sensing; David Wilkinson - Remote Sensing/Inventory; Patrick Glass - Biometrics/Remote Sensing; Alexis Londo - Biometrics/Remote Sensing; Clint Iles - Biometrics; George Banzhaf - Biometrics. Post Doctorates: Li Gan, Post Doctorate - Computer Applications; Gustavo Perez-Verdin, - Forest Management/Biomass TARGET AUDIENCES: The target audiences include landowners, state planners, forest industry, researchers, and other local, state, and federal agencies concerned with forest management and utilization. The Web-based Mississippi Institute for Forest Inventory (MIFI) interface, MIFI annual reports, growth and yield software, and publications are the primary instruments that deliver the science-based knowledge to the target audiences. The delivered information will be helpful in economic development, national biomass/energy assessments, change detection, disaster and invasive species assessments, reforestation and land use planning, hardwood management decision making, landowner education, policy decisions, and wildlife habitat assessments. The MIFI tools also allow the state to entice forest products industry investors to build or expand existing mills while monitoring the sustainability of the resource. PROJECT MODIFICATIONS: Not relevant to this project.

Impacts
Software for a bottomland hardwood volume by grade calculator has been developed, and a Web interface for its implementation is near completion. It will be hosted on the Forest and Wildlife Research Center server and available to landowners and others for calculating the volume by grade within a tree. This software will aid landowners in hardwood management decision making. Biomass component estimators have been added to the Mississippi Institute for Forest Inventory (MIFI) decision support interface and are being operationally utilized to estimate the potential woody biomass supply for biofuel generation in Mississippi. MIFI inventory results demonstrate that for much of the State, approximately 75% more wood is being produced than consumed, and this surplus can add significantly to the existing $14 billion total output of Mississippi's forest and forest products industry. Good growth and drain estimates guarantee the stability and sustainability of the forest products industry that accounts for 6.6% of the State's total industrial output. MIFI interface software tools will allow the state to entice forest products industry investors to build or expand existing facilities. In the 32 months since the on-line MIFI Decision Support System has been operational, there have been 95 requests for installation or analytical support. From these requests, 67 cursory analyses were conducted to assess the availability of various product species combinations within the currently inventoried MIFI regions. The 28 remaining requests were directed toward intensive and detailed analyses for very specific combinations of species product availability including diameter distributions, growth-to-drain ratios and spatial distribution. Currently, two bio-energy facilities are pursuing capital funding for locating mills within Mississippi, and one traditional timber product company is poised to start construction of a facility pending a renewed vigor in their portion of the market.

Publications

  • Fujisaki, I., M. Mohammadi-Aragh, D.L. Evans, R.J. Moorhead, D.W. Irby, S.D. Roberts, S.C. Grado, E.B. Schultz. 2007. Comparing forest assessment based on computer visualization versus videography. Landscape and Urban Planning. 81(2007):146-154.
  • Grebner, D.L., G. Perez-Verdin, C. Sun, I.A. Munn, E.B. Schultz, T.G. Matney. 2007. Current status of woody biomass feedstock supply and availability in Mississippi. http://www.sebioenergy.org/ConferencePDF/July31/330-500/A/Don%20Grebn er.pdf
  • Matney, T.G., E.B. Schultz. 2007. Mesavage and Girard Form Class Taper Functions Derived from Profile Equations. Pages 77-85 in D.S. Buckley, W.K. Clatterbuck, editors. Proceedings, 15th Central Hardwood Forest Conference. e-Gen. Tech. Rep. SRS-101. U.S. Department of Agriculture, Forest Service, Southern Research Station.
  • Perez-Verdin, G., D.L. Grebner, C. Sun, I.A. Munn, E.B. Schultz, T.G. Matney. 2007. The potential woody biomass supply for biofuel generation in Mississippi. Society of American Foresters National Convention. Portland, OR.


Progress 01/01/06 to 12/31/06

Outputs
Issues: A reliable and timely forest inventory and the availability of decision support tools for assessing current and future forest resources are of primary importance to attracting and sustaining forest industry and ensuring the future of our forest resource through good management decisions. Decision support tools based on Mississippi Institute for Forest Inventory (MIFI) data are being developed that will 1) assist forest land owners in projecting growth and yields for evaluating management alternatives, 2) assist industry in planning future mill locations and sustaining existing mills, and 3) develop geospatial modeling techniques that will provide more timely, and possibly less expensive, evaluation of forest resources. Response: 1) A spatially-based decision support system for locating forest products/bioenergy based plant facilities in Mississippi is being constructed as a cooperative effort between MSU and MIFI. The project is coalescing geo-spatial information, forest inventory, transportation costs, and socio-economic factors to allow forest-based bio-energy investors to rapidly assess the economic viability and sustainability of proposed plant locations and capacities. These same tools can also be used to evaluate proposed plant expansions/modifications and predict future raw material availability and to evaluate the impact of competing resource consumers (e.g. bioenergy particle-based panels and lumber, plywood, sawtimber, poles, veneer, pulpwood, reconstituted solid wood products like TimTek, and exports). MIFI's current county-level inventory statistics, a transportation network, and spatially allocated growth and drain estimates are being incorporated into a highly evolved database structure and spatially oriented automated MIFI inventory reporter. 2) Traditional and neural network models for predicting loblolly pine dry weight pulp yields, lumber by size class, and peeler production are being developed and incorporated into a growth and yield simulator to estimate final product yields and values. A neural network, as opposed to model-based, growth and yield prediction system is being populated with data to make projections for six forest types. The growth and yield system will be incorporated into a statewide Web-based inventory system being implemented by MIFI. 3) The lack of growth and yield information for bottomland hardwoods is a major impediment to determining optimum species mix, management regimes, economic value, and expected rotation ages. A project has been initiated in cooperation with the US Forest Service to gather growth and yield information for a major bottomland forest type. 4) Two artificially intelligent (AI) classification techniques are being used to identify species from a combination of LIDAR and multispectral data. Neural network techniques produce a 70%-80% classification accuracy compared to 30%-40% accuracy for the knowledge-based tools available from geospatial software packages. The addition of LiDAR data to multispectral data improves the knowledge-based classification but not the more accurate neural network classification system.

Impacts
Growth and yield models facilitate better biological and economic planning and management for sustained production. Without growth and yield information to convince landowners, government, and industry leaders of the economic benefits of managing and sustaining the forest resource, they will be reluctant to invest the time and money required to bring the forests into maximum production with the desired ecological health. Successful development of geospatial tools have further potential impact in determining yields and habitat types from spatial information. Decision support tools for determining optimal mill locations will aid economic development decisions for state planners and forest industry. The component growth and drain modules will also be important in reforestation planning, landowner education, policy decisions, and wildlife habitat assessments.

Publications

  • Schultz, E.B. and T.G. Matney. 2006. An integrated growth and yield simulator for predicting loblolly pine dry weight pulp yields. Wood and Fiber Science 38(4):672-681.
  • Schultz, E.B., T.G. Matney. 2006. A growth and yield model for predicting both forest stumpage and mill side manufactured product yields and economics. Pp. 305-309 in 2006 Pan Pacific Conference, Advances in Pulp & Paper Sciences and Technologies, Seoul, Korea.
  • Schultz, E., T. Matney, D. Evans, I. Fujisaki. 2006. A Landsat stand basal area classification suitable for automating stratification of forest into statistically efficient strata. Six pages in S. Lang, T. Blaschke, E. Schopfer, editors, First International Conference on Object based Image Analysis. Salzburg, Austria.


Progress 01/01/05 to 12/31/05

Outputs
Issues: A reliable and timely forest inventory and the availability of decision support tools for assessing current and future forest resources are of primary importance to attracting and sustaining forest industry and ensuring the future of our forest resource through good management decisions. The Mississippi Institute for Forest Inventory (MIFI) is meeting the need for an accurate and timely inventory of the State's forest resources and now tools based on this inventory are being developed that will 1) assist forest land owners in projecting growth and yields for evaluating management alternatives, 2) assist industry in planning future mill locations and sustaining existing mills, and 3) develop geospatial modeling techniques that will provide more timely, and possibly less expensive, evaluation of forest resources. Response: Traditional and neural network models for predicting loblolly pine dry weight pulp yields, lumber by size class, and peeler production are being developed and incorporated into a growth and yield simulator to estimate final product yields and values. User software 'CLoblolly' for the growth and yield model has been developed and is available over the Internet at http://www.cfr.msstate.edu/fwrc/software.htm. A knowledge-based, as opposed to model-based, growth and yield prediction system is under development. The system is currently being populated with data to make projections for six forest types. The growth and yield system will be incorporated into a statewide Web-based inventory system being implemented by MIFI. The lack of growth and yield information for bottomland hardwoods is a major impediment to determining optimum species mix, management regimes, economic value, and expected rotation ages. A project has been initiated in cooperation with the US Forest Service to gather growth and yield information for a major bottomland forest type. A system of neural network models has been developed to predict growth and yield for repeatedly-thinned longleaf pine plantations. Pine growth and yield plots that have been repeatedly thinned are often so sparsely occupied that empirical distributions cannot be defined with smooth functional forms. Therefore, traditional parametric models dependent upon predefined functional forms may not perform well. Results show that the neural network models predict growth with low bias and high index of fit providing a tool for forest managers and restoration ecologists to compare management strategies. Two artificially intelligent (AI) classification techniques are being used to identify species from a combination of LIDAR and multispectral data. Initial data indicates that neural network techniques produce a 70%-80% classification accuracy compared to 30%-40% accuracy for the knowledge-based tools available from geospatial software packages. The addition of LiDAR data to multispectral data improves the knowledge-based classification but not the more accurate neural network classification system. Four peer reviewed journal articles and two funding proposals were submitted based on project research.

Impacts
Growth and yield models facilitate better biological and economic planning and management for sustained production. Without growth and yield information to convince landowners, government, and industry leaders of the economic benefits of managing and sustaining the forest resource, they will be reluctant to invest the time and money required to bring the forests into maximum production with the desired ecological health. Successful development of geospatial tools have further potential impact in determining yields and habitat types from spatial information. Decision support tools for determining optimal mill locations will aid economic development decisions for state planners and forest industry. The component growth and drain modules will also be important in reforestation planning, landowner education, policy decisions, and wildlife habitat assessments.

Publications

  • Parker, R.C., P.A. Glass, H.A. Londo, D.L. Evans, K.L. Belli, T.G. Matney, and E.B. Schultz. 2005. Mississippi's Forest Inventory Pilot Program: Use of Computer and Spatial Technologies in Large Area Inventories. FWRC Research Bulletin FO274, Mississippi State University. 43 p.


Progress 01/01/04 to 12/31/04

Outputs
Study 1: Two artificially intelligent (AI) classification techniques are being used to identify species from a combination of LIDAR and multispectral data. Neural networks and a ruled-based knowledge system are being developed and will be compared to ground data to determine the ability of the AI systems to identify species from remotely sensed data. The study is based on plots in the lower mountains regions near McCall, Idaho. If these techniques are successful, they could be used with other developing technology in determining yields and wildlife habitat types from spatial information. Study 2: Models and software were developed for predicting pulp yields from loblolly pine tree and stand characteristics. An artificial neural network model for predicting wood chip thickness distributions from stand and tree characteristics of loblolly pine and a single tree dry weight pulp yield model were incorporated into a growth and yield simulator. Single tree dry weight pulp yields were estimated from relationships between moisture content and specific gravity along tree boles for selected kappa numbers. Data were derived from 11,771 individual loblolly pine chip thickness measurements. Four stand ages, five dbh classes, and three stem positions were used to predict the cumulative proportion of chip weight per chip thickness class. Tables were constructed to give projected pulp yields for different levels of age, dbh, stand density, site index, and kappa numbers. Publication efforts are in progress. Study 3: A knowledge-based, as opposed to model-based, growth and yield prediction system is under development. The system is currently being expanded to make projections for six forest types but will continue to be expanded to include many forest types and management regimes. Results from the knowledge-based system will be compared to traditional systems for assessment purposes. The growth and yield system will be incorporated into a statewide Web-based inventory system being implemented by the Mississippi Institute for Forest Inventory. Study 4: A system of neural network models has been developed to predict growth and yield for repeatedly thinned longleaf pine plantations. Pine growth and yield plots that have been repeatedly thinned are often so sparsely occupied that empirical distributions cannot be defined with smooth functional forms. Therefore, traditional parametric models dependent upon predefined functional forms may not perform well. Neural networks are not constrained to predefined functional forms and proved to be better predictors of longleaf diameter distributions after multiple thinnings. The construction of the neural network models utilized 209 longleaf, even-aged, permanent 1/5 and 1/10 acre plots established in the 1960's. A feed forward, back propagation learning system was selected to produce the model. Calculations of bias, root mean square error, and index of fit were used to compare the neural network model to two deterministic models that are in current use. The neural network models performed better than the traditional models. Publication efforts are in progress. (Graduate Students = 1)

Impacts
New techniques in artificial intelligence are being used to model forest systems where traditional modeling techniques may be inadequate because of complicated interactions or lack of suitable mathematical models. These new models will facilitate better biological and economic planning and management for sustained production and maintenance of ecosystems.

Publications

  • No publications reported this period


Progress 01/01/03 to 12/31/03

Outputs
Two artificially intelligent (AI) classification techniques are being used to identify species from a combination of LiDAR and multispectral data. Neural networks and a ruled-based knowledge system are being developed and will be compared to ground data to determine the ability of the AI systems to identify species from remotely sensed data. The study is based on forty-nine plots in the lower mountain regions near McCall, Idaho. If these techniques are successful, they could be used with other developing technology in determining yields and wildlife habitat types from spatial information. (Graduate Students = 1). Models and software were developed for predicting pulp yields from loblolly pine tree and stand characteristics. An artificial neural network model for predicting wood chip thickness distributions from stand and tree characteristics of loblolly pine and a single tree dry weight pulp yield model were incorporated into a growth and yield simulator. Single tree dry weight pulp yields were estimated from relationships between moisture content and specific gravity along tree boles for selected kappa numbers. Data were derived from 11,771 individual loblolly pine chip thickness measurements. Four stand ages, five dbh classes, and three stem positions were used to predict the cumulative proportion of chip weight per chip thickness class. Tables were constructed to give projected pulp yields for different levels of age, dbh, stand density, site index, and kappa numbers. Publication efforts are in progress. A knowledge-based, as opposed to model-based, growth and yield prediction system is under development. The system currently makes projections for one forest type but will continue to be expanded to include many forest types and management regimes. Results from the knowledge-based system will be compared to traditional systems for assessment purposes. The growth and yield system will be incorporated into a statewide Web-based inventory system being implemented by the Mississippi Institute for Forest Inventory. A system of neural network models has been developed to predict growth and yield for repeatedly thinned longleaf pine plantations. Pine growth and yield plots that have been repeatedly thinned are often so sparsely occupied that empirical distributions cannot be defined with smooth functional forms. Therefore, traditional parametric models dependent upon predefined functional forms may not perform well. Neural networks are not constrained to predefined functional forms and proved to be better predictors of longleaf diameter distributions after multiple thinnings. The construction of the neural network models utilized 209 longleaf, even-aged, permanent 1/5 and 1/10 acre plots established in the 1960's. A feed forward, back propagation learning system was selected to produce the model. Calculations of bias, root mean square error, and index of fit were used to compare the neural network model to two deterministic models that are in current use. The neural network models performed better than the traditional models. Publication efforts are in progress.

Impacts
New techniques in artificial intelligence are being used to model forest systems where traditional modeling techniques may be inadequate because of complicated interactions or lack of suitable mathematical models. These new models will facilitate better biological and economic planning and management for sustained production and maintenance of ecosystems.

Publications

  • No publications reported this period


Progress 01/01/02 to 12/31/02

Outputs
A system of neural network models has been developed to predict growth and yield for repeatedly thinned longleaf pine plantations. Pine growth and yield plots that have been repeatedly thinned are often so sparsely occupied that empirical distributions cannot be defined with smooth functional forms. Therefore, traditional parametric models dependent upon predefined functional forms may not perform well. Neural networks are not constrained to predefined functional forms and proved to be better predictors of longleaf diameter distributions after multiple thinnings. The construction of the neural network models utilized 209 longleaf, even-aged, permanent 1/5 and 1/10 acre plots established in the 1960's. A feed forward, back propagation learning system was selected to produce the model. Calculations of bias, root mean square error, and index of fit were used to compare the neural network model to two deterministic models that are in current use. The neural network models performed better than the traditional models. A master's thesis was produced. (Graduate Students = 1) A working prototype for a knowledge-based, as opposed to model-based, growth and yield prediction system was completed. The prototype currently makes projections for only one forest type but will continue to be expanded to include many forest types and management regimes. A final version of the growth and yield prediction software will be delivered over the Internet. A MS Windows user interface is already in place. Results from the knowledge-based system will be compared to traditional systems. Two artificially intelligent (AI) classification techniques are being used to identify species from a combination of LIDAR and multispectral data. Neural networks and a ruled-based knowledge system are being developed and will be compared to ground data to determine the ability of the AI systems to identify species from remotely sensed data. The study is based on forty-nine plots in the lower mountainous regions near McCall, Idaho. If these techniques are successful, they could be used with other developing technology in determining yields and wildlife habitat types from spatial information. (Graduate Students = 1) Models and software were developed for predicting pulp yields from loblolly pine tree and stand characteristics. An artificial neural network model for predicting wood chip thickness distributions from stand and tree characteristics of loblolly pine and a single tree dry weight pulp yield model were incorporated into a growth and yield simulator. Single tree dry weight pulp yields were estimated from relationships between moisture content and specific gravity along tree boles for selected kappa numbers. Data were derived from 11,771 individual loblolly pine chip thickness measurements. Four stand ages, five dbh classes, and three stem positions were used to predict the cumulative proportion of chip weight per chip thickness class. Tables were constructed to give projected pulp yields for different levels of age, dbh, stand density, site index, and kappa numbers. Total graduate students = 2.

Impacts
New techniques in artificial intelligence are being used to model forest systems where traditional modeling techniques may be inadequate because of complicated interactions or lack of suitable mathematical models. These new models will facilitate better biological and economic planning and management for sustained production and maintenance of ecosystems.

Publications

  • Schultz, Emily B. and Thomas G. Matney. 2002. Prediction of Pulp Yields from Loblolly Pine Stand and Tree Characteristics. 2002 TAPPI Proceedings. 2002 Fall Technical Conference. September 8-11, San Diego, CA. CD-ROM.


Progress 01/01/01 to 12/31/01

Outputs
A system of neural network models has been developed to predict growth and yield for repeatedly thinned longleaf pine plantations. Pine growth and yield plots that have been repeatedly thinned are often so sparsely occupied that empirical distributions cannot be defined with smooth functional forms. Therefore, traditional parametric models dependent upon predefined functional forms may not perform well. Neural networks are not constrained to predefined functional forms and proved to be better predictors of longleaf diameter distributions after multiple thinnings. The construction of the neural network models utilized 209 longleaf, even-aged, permanent 1/5 and 1/10 acre plots established in the 1960's. A feed forward, back propagation learning system was selected to produce the model. Calculations of bias, root mean square error, and index of fit were used to compare the neural network model to two deterministic models that are in current use. The neural network models performed better than the traditional models. Publication efforts are underway. (Graduate Students = 1) A working prototype for a knowledge-based, as opposed to model-based, growth and yield prediction system was completed. The prototype currently makes projections for only one forest type but will continue to be expanded to include many forest types and management regimes. A final version of the growth and yield prediction software will be delivered over the Internet. A MS Windows user interface is already in place. Results from the knowledge-based system will be compared to traditional systems. Two artificially intelligent (AI) classification techniques are being used to identify species from a combination of LIDAR and multispectral data. Neural networks and a ruled-based knowledge system are being developed and will be compared to ground data to determine the ability of the AI systems to identify species from remotely sensed data. The study is based on forty-nine plots in the lower mountains regions near McCall, Idaho. If these techniques are successful, they could be used with other developing technology in determining yields and wildlife habitat types from spatial information. (Graduate Students = 1)

Impacts
New techniques in artificial intelligence are being used to model forest systems where traditional modeling techniques may be inadequate because of complicated interactions or lack of suitable mathematical models. These new models will facilitate better biological and economic planning and management for sustained production and maintenance of ecosystems.

Publications

  • Meador, Andrew J. Applications in growth and yield of longleaf pine (Pinus palustris MILL.): A comparison of artificial neural networks and other traditional modeling methodologies [thesis]. Mississippi State, MS: Mississippi State University; 2002. 49 p.


Progress 01/01/00 to 12/31/00

Outputs
A neural network model is being developed to predict growth and yield for repeatedly thinned longleaf pine plantations. Pine growth and yield plots that have been repeatedly thinned are often so sparsely occupied that empirical distributions cannot be defined with smooth functional forms. Therefore, traditional parametric models dependent upon predefined functional forms may not perform well. Neural networks are not constrained to predefined functional forms and may be better predictors of diameter distributions after multiple thinnings. Construction of a neural network model is currently underway using 209 longleaf, even-aged, permanent 1/5 and 1/10 acre plots established in the 1960's. A feed forward, back propagation learning system has been selected to construct the network model. Calculations of bias, root mean square error, and index of fit will be used to compare the neural network model to two deterministic models that are in current use. A working prototype for a knowledge-based (as opposed to model-based) growth and yield prediction system was completed. The prototype currently makes projections for only one forest type but will continue to be expanded to include many forest types and management regimes. A final version of the growth and yield prediction software will be delivered over the Internet. A MS Windows user interface is already in place. Results from the knowledge-based system will be compared to traditional systems. (Graduate students = 1)

Impacts
New techniques in artificial intelligence are being used to model forest systems where traditional modeling techniques may be inadequate because of complicated interactions or lack of suitable mathematical models. These new models will facilitate better biological and economic planning and management for sustained production and maintenance of ecosystems.

Publications

  • Londo, H.A.,P.A. Glass, D.L. EVans, K.L. Belli, R.C. Parker, T.G. Matney, E.B. Schultz. 2000. Integration of Remote Sensing GPS with Traditional Forest Inventory Procedures. In: Proceedings of the Third Southern Forestry GIS Conference. October 10-12,2000. University of Georgia, Athens, GA.


Progress 01/01/99 to 12/31/99

Outputs
Work was completed on neural network modeling applications used to improve the ability of mill managers and procurement foresters to predict chip thickness distributions based on tree and stand characteristics. Other studies involve neural network modeling of yields from repeatedly thinned longleaf pine stands and a knowledge-based expert system for estimating future productivity of forest stands. The expert system is coordinated with a pilot inventory project for implementation. Data collection for both projects is currently underway. (Graduate Students = 1)

Impacts
New techniques in artificial intelligence are being used to model forest systems where traditional modeling techniques may be inadequate because of complicated interactions or lack of suitable mathematical models. These new models will facilitate better biological and economic planning and management for sustained production and maintenance of ecosystems.

Publications

  • Schultz, Emily B., Thomas G. Matney, and Jerry L. Koger. 1999. A neural network model for wood chip thickness distributions. Wood and Fiber Science, 31(1), 1999. pp. 2-14.


Progress 01/01/98 to 12/31/98

Outputs
A manuscript was prepared and submitted to Wood and Fiber Science on neural network modeling applications used to improve the ability of mill managers and procurement foresters to predict chip thickness distributions based on tree and stand characteristics. Another study was initiated that will focus on using a knowledge-based expert system for estimating the productivity of forest stands. Expert system tools were evaluated from case studies for comparison and selection. Data flow and data structure designs have been developed. The expert system is being coordinated with a pilot inventory project for implementation.

Impacts
(N/A)

Publications

  • No publications reported this period


Progress 01/01/97 to 12/31/97

Outputs
Neural network modeling applications were used to improve the ability of mill managers to predict chip thickness distributions based on tree and stand characteristics and to supply wood yard managers with models that predict the drying of roundwood sprinkled during long-term storage. The wood chip thickness model was combined with a growth and yield simulator to allow procurement foresters to value stands according to their expected chip yields or to manage stands in such a way as to increase chip yields. Results of modeling roundwood moisture content contradicted the popular belief that the majority of drying occurs through log ends.

Impacts
(N/A)

Publications

  • Schultz, Emily B. and Thomas G. Matney. 1997. Prediction of Moisture Contents for Sprinkled and Un-Sprinkled Stacked Roundwood Over Time. TAPPI Proceedings, 1997 Pulping Conference, October, 19-23, San Francisco, CA. pp. 289-299.