D e p a r t m e n
t o f
Since the early 1970’s, forestry companies worldwide have been planning the management of forest estates with the assistance of linear programming (LP) based optimisation models. These forest estate management modelling systems assist forest managers with decisions such as the timing, scale and mix of what should be harvested, the area of new land to plant, what silvicultural (thinning and pruning) regimes should be used and whether or not land should be replanted following harvesting. Due to their flexibility and usefulness in assisting with these types of decisions, forest estate modelling programs have become an integral part of modern forest estate management.
Accordingly, a significant part of teaching forest estate management at the New Zealand School of Forestry involves teaching students about the principles of forest estate modelling. To assist with their understanding, students learn to use a typical forest estate management modelling program. However, the current program used at the School of Forestry is not ideally suited as a teaching tool. Having been developed in the mid 1980’s and designed primarily for use as a commercial package, the current program has certain features that result in students finding the program difficult to learn and use. Consequently, students spend considerable effort learning how to use the program instead of achieving an understanding of the important principles of forest estate management modelling.
Therefore, there existed a desire that a new program be developed that contained the strengths of traditional forest estate modelling programs while also being easier for forestry students to learn and use. LOG (Linear Optimisation Generator), was developed by Tim Dobbs and Andrew Maxwell as part of their Management Science Honours Project over a two and a half month period with the aim of providing an improved tool for assisting the teaching of forest estate modelling to forestry students.
Similar to other forest estate modelling programs, LOG models forest management by receiving data and user-defined problem specific information. It then generates a forestry model which is converted into a linear programming format and solved using a linear programming solver. Like other programs, LOG also allows users to specify complicated models in forestry terms without the need of specialist knowledge of linear programming. It similarly is able to interpret linear programming output and generate reports using the same forestry terminology.
However, LOG contains a number of features that differentiate it from other forest estate modelling systems. Rather than being a stand alone system like forest estate modelling systems, LOG incorporates a range of modern software including the mathematical modelling language AMPL, Microsoft Access, a linear programming solver and a Windows based user interface programmed using Visual Basic.
Utilising this generic software meant that LOG was able to be developed extremely quickly but also meant that particular advantages of packages such as AMPL and Microsoft Access were integrated into LOG. These and other features of LOG are discussed in the following sections.
LOG allows users to construct models in forestry terms. It then translates the forestry problem into a LP format that can be interpreted and solved by an LP solver. To convert the user specified problem into the format required by the LP software, LOG incorporates a generic mathematical programming language called AMPL. Utilising AMPL to provide this function saved a considerable amount of time, as without AMPL it would have been necessary to program a specialised matrix generator to convert forestry problems into an LP format. Developing a matrix generator is a complicated and time-consuming procedure and having AMPL provide this interface automatically allowed LOG to progress significantly further than if a matrix generator had to constructed manually.
Incorporating AMPL in LOG also provides a number of other advantages. It is able to convert forestry problems into an LP format much more efficiently and quickly than a specialised matrix generator. This is commonly the most time consuming procedure when solving a model in a similar forest estate modelling program. AMPL also performs extensive pre-processing routines that serve to simplify models making them more efficient and faster to solve.
A further advantage of AMPL is the ease of which it facilitates the addition of new model features. AMPL requires only that a new feature be added to the model in equation form. It is automatically able to create the LP representation of the new model without any additional modifications.
To provide a modern and easy to use layout, LOG has a Windows based user interface. Compatible with modern operating systems the Windows format is familiar to most users. Utilising common drop boxes, list boxes, command buttons and tick boxes, as demonstrated in Figure 1 below enables users to easily define the specifics of a particular problem.
The design of LOG’s user interface also encourages one of the major strengths of forest estate modelling programs, which is to perform ‘what-if’ analysis. This commonly involves varying constraints enforced on the model or analysing the effect of changes in parameters. To ensure that users are able to travel easily throughout various sections of LOG in order to access and change particular parameters, LOG incorporates a tree structure directory.
To enable AMPL, the program that constructs the linear program representation of the forestry model, to read the data, LOG efficiently and neatly stores all data relating to products, costs and revenues within a single Microsoft Access database. Storing information in a database also results in numerous other advantages. Data is presented in a table format that is easily understandable, readable and allows for straightforward analysis and editing of individual data values. An example of the LOG data format is shown to the right.
Other advantages of storing data within a database include the capacity to store substantial amounts of information, and the ability for data contained in other formats, for instance spreadsheets, to be converted simply into a LOG data file. For example, data contained in a spreadsheet can be easily modified to fit the structure required by LOG and copied to a database file.
LOG also stores the specifics of a problem and solutions to a particular model in a single database file. This allows for flexibility in analysing solutions. For instance, utilising the capabilities of Microsoft Access, solutions can be sorted by different characteristics for easier analysis. Solutions can also be exported to other applications such as Microsoft Excel to enable calculations and graphical analysis of solutions.
Due to time constraints on this project, LOG is currently only a prototype program and does not yet include a number of desirable features of other forestry estate management modelling programs. Such features not currently incorporated in LOG include users not being able to define customised objectives or constraints and LOG not having the capability to model the allocation of logs to processing plants. However, due to the design of LOG and particularly the incorporation of AMPL, such features will be relatively straightforward to incorporate into LOG in the future. In fact, in conjunction with the sponsor of this project, Associate Professor Bruce Manley at the New Zealand School of Forestry, it is intended that the capabilities of LOG be extended to include these and other desirable features of other forest estate management modelling programs. As well as this, it is planned that LOG will be tested extensively before being employed as a tool for teaching forest estate modelling at the New Zealand School of Forestry in 2003.
For questions regarding this project please contact any of the following people:
LOG Designers: Tim Dobbs and Andrew Maxwell
Project Supervisor: Dr John F. Raffensperger
Project Sponsor: Associate Professor Bruce Manley