About

NSF Grant Proposal

An Inquiry-Based Simulation Learning Environment for the
Ecology of Forest Growth

A Proposal to the Course, Curriculum, and
Laboratory Improvement Program
National Science Foundation
(DUE-CCLI Track: Educational Materials Development, Fall 1998)

Tom Murray
Lawrence Winship
Neil Stillings

Hampshire College
Amherst, MA

Abstract:

Computer-based simulations of natural phenomena are particularly effective educational tools, especially when integrated into field and laboratory-based experiences. The investigators propose to build and evaluate educational software that simulates tree and forest growth and the effects of natural and human-created environmental disturbances on forest growth. We hope to further advance inquiry-based science teaching practices at Hampshire College and more broadly in Western Massachusetts. An inquiry-based educational philosophy is already deeply embedded into Hampshire College's academic structure, as well as in the faculty's teaching methods. To this institutional foundation we add an extensive collaboration with the NSF-funded Five-College PALMS and STEMTEC programs. These programs have built relationships between researchers, teacher education programs, and practicing teachers at local K-12 schools, will allow us to have significant impact in local undergraduate as well as high school education. An innovative contribution in the area of evaluation will be to combine evaluation methodology and teacher education. Pre-service and in-service science teachers will carry out some of the classroom-based testing. The student teachers will serve as the vectors for introducing both the simulation technology and the inquiry-based teaching methods into existing classrooms. The results of the research, including software, student and teacher support materials, and evaluation results, will be distributed via a dedicated web site. One of the contributions of this work will be in developing an approach to representing simulation formulas that is applicable to educational simulations in any subject area. While most educational simulations are "black boxes" that do not allow students to inspect or manipulate the underlying formulas, our software will include a Formula Inspector that will allow students to more deeply address four types of inquiry questions: "what if," "relationship," "why," and "modeling." DUE themes addressed: teacher preparation, integration of technology with education, faculty development.

Project Description

1. Project Overview and Goals
The Presidential Committee of Advisors on Science and Technology's Panel on Education (March 1997) states that citizens in the next century will "require not just a larger set of facts or a larger repertoire of specific skills, but the capacity to readily acquire new knowledge, to solve new problems, and to employ creative thinking in the design of new approaches to existing problems" (p. 5). Modern educational theory stresses the importance of student-active learning and inquiry-based science education to address these educational goals (McNeal & D'Avanzo, 1996; National Research Council, 1996; AAAS, 1993). It has been documented that students often develop a view that science is a method for discovering static facts about the world, and see learning science as learning those facts (Lederman, 1992). In contrast, we wish to foster a view of science as an active process of discovering relationships between observed phenomena and being able to generate predictions, models, and explanations using these discoveries. Inquiry-based science experiences conducted in relevant, meaningful contexts have been shown to develop higher order thinking skills in students (Roth & Roychoudhury, 1993).

Hampshire College, since its inception, has affirmed that meaningful education engages learners in increasingly sophisticated, student-driven, and realistic problem solving (Prince & Kelley 1996). An inquiry-based educational philosophy is deeply embedded in the college's academic structure and undergraduate course offerings, as well as in the faculty's teaching methods. Science education in particular involves a research-based approach that is innovative for undergraduate educational institutions. Science educators at Hampshire are actively pursuing the use of technological enhancements to improve inquiry learning and to prepare students for the increasingly technology-rich nature of the work force. The Hampshire experience brings a unique focus to the design and classroom use of educational technology.
Recent research has found (Fatemi & Trotter, 1998) that computer-based education does not produce the desired results in all cases. In fact, students who use drill-and-practice style programs can perform worse on tests, while students using more sophisticated and engaging software show marked skill improvements. Computer-based simulations of natural phenomena can be particularly effective educational tools, especially when integrated into field and laboratory-based experiences that ensure students understand the relevance of the computer-based experience to actual phenomena and scientific practices.

The investigators propose to build the Forest Growth Simulator, an educational software package that simulates tree and forest growth, the succession of tree species over time, and the effects of environmental disturbances on forest growth. Our software will draw on methods found in existing simulations used in professional forestry, but add several features that address pedagogical shortcomings in existing software. The focus of existing software is on the simulation itself, and inadequate attention is paid to pedagogical aspects and to supporting the inquiry process. This is true of simulation-based educational software in general, as well as existing forest simulations. More specifically, existing simulations are "black boxes" that do not allow students to inspect or manipulate the underlying formulas. The Forest Growth simulator will include a Formula Inspector that will allow students to gain a deeper understanding of the relationships among the observed phenomena. Thus the products of the proposed research will have implications for all simulation-based science education software. Students should engage in hypothesis formation, experimental design, carrying out experiments, gathering data, analyzing data, and re-evaluating their hypotheses. The Forest simulation will support each of these activities, and act as a virtual laboratory for exploring tree growth and forest community dynamics (such as succession, gap phase dynamics, and the effect of natural and human disturbance).

The proposed research and evaluation has two main threads, which will be pursued throughout the project. First, the major research hypothesis of the project is that simulation-based software can be developed that will enhance students' active understanding of the use models in science, of the assumptions underlying scientific models, and of the emergent phenomena that arise, often unexpectedly, from quantitative models. We expect that this understanding will be enhanced not only for the course content, in this case forest ecology, but also for scientific models and theories generally. Second, we will conduct a broader program of evaluation research to assess the usability and effectiveness of the software and associated curricular materials.

Our overarching goals are to further advance inquiry-based science teaching practices at Hampshire College and more broadly in Western Massachusetts, in an effort to make science education more effective and meaningful to students and more relevant to the needs of the modern work force for flexible, pro-active problem solving.
Specific goals are: 1) To build a forest growth simulator; 2) to add pedagogically powerful features such as the Formula Inspector (described later); 3) to develop a series of activities that encourage students to use the simulation to answer various types of inquiry questions and relate their explorations to actual forests; 5) to develop curriculum materials such as student and teacher manuals that support the use of the software in classrooms; 6) to evaluate the effectiveness of the software and supporting materials in diverse classroom contexts; 7) to implement, evaluate and report on an innovative method for representing formulas in educational simulations; and 8) to implement and report on an innovative evaluation method that combines teacher training with evaluation and dissemination.

2. Motivating Opportunities and Objectives

2.1 Forest Ecology, Environmental Science, and Hampshire's Undergraduate Education

Hampshire College, since its inception, has affirmed that meaningful education engages learners in increasingly sophisticated, student-driven, and realistic problem solving. An inquiry-based educational philosophy is deeply embedded in the college's academic structure and undergraduate course offerings, as well as in the faculty's teaching methods. Science education in particular involves a research-based approach that is innovative for undergraduate education. Our programs are particularly successful at encouraging students who had not come to college to become scientists and to go on to careers in science and science teaching. While only 7% enter the college to major in science, 15-20% actually fulfill their degree requirements in the sciences, a remarkable reversal of national trends. Many of our students choose to work in the environmental sciences, a rapidly growing area at most colleges. One of the most successful introductory courses at Hampshire, The Ecology of Old Growth Forests, builds on the concern many students have about our ever increasing exploitation of natural resources to engage them in debates and investigations about the broader issues of Forest Ecology. Students in this class quickly discover that our forests are always changing, subject to disturbances from human and natural forces. By collecting field data, such as survey plots, tree cores, and soil cores, as well as reading historical records and maps, our students come face to face with the difficulty of managing and preserving a complex, hard to understand ecosystem. Forest ecology provides an approachable, relevant venue for students to struggle with real scientific issues and to make real contributions to a growing discipline. We also use forest survey methods in our other ecology and plant biology classes. The methods we use to collect data are easy to learn and rapidly lead to exciting questions from students, and they have worked well with students in secondary schools.

2.2 Limitations in existing classroom and field-based methods and the need for a simulation-based learning environment.

While data collection in forest ecology is relatively straightforward, data analysis can be an overwhelming task. Students love being out in the woods in survey teams making observations and collecting measurements, but in all but the most simple investigations, the resulting data set is large and many layered, and patterns are not immediately obvious. In some cases students want to investigate forests that are not readily accessible. They often pose questions that would best be answered using a time machine! They need help visualizing both the forests and the trees as they change over time and with the effects of disturbance. Forest systems are inherently complex so we need methods that can summarize and visualize without oversimplifying or where the level of simplification can be modified. Changes in our forests may best be described as long decades of boring tranquility punctuated by intense hours of terrifying energy - fires, tornadoes, hurricanes, clearcutting. It is hard for students walking through a stand to visualize the effect of changes on such disparate time scales.

An effective, easy-to use forest growth simulator could help with many of these problems. Although we consider field trips to actual forest plots to be an essential component to engaging and learning this material, a computer simulation can allow students to enhance their learning in several ways. First, they can simulate the passage of time over many years. Second, they can simulate a practically countless variation of environmental and forest management conditions. Third, the process of data gathering and analysis, which can become prohibitively lengthy and tedious for many students, can be automated. Students need to learn how to actually measure tree diameters and generate bar charts, but after having done this a few times they can be freed of having to repeat the process hundred of times. Thus they can more quickly and easily generate and test their hypotheses, and spend more time grappling with important conceptual aspects of the domain.

2.3 Problems with Existing Forest Growth Simulators

Forest growth is a particularly good domain for educational simulations because the formulas that determine forest growth are few (around 50 for the standard model), easy to understand, and readily available. Virtually all forest growth simulation programs are in some way derivatives of the pioneering work by Botkin, Janak, and Wallis in 1972(a,b), initially called JABOWA (Dale and Shugart, 1985). We have used a more recent improved program, JABOWA II, (Botkin and Nesbit, 1992; Botkin, 1993) in a few of our classes and in a senior honors thesis (Mills, 1993). Even the relatively unsophisticated graphic display used in JABOWA II elicits cheers and shouts from the students as they urge on the pines and anticipate the death of the large overshadowing red maple - but that's as far as it goes. The interface shows its age and its heritage of supporting professional forestry by not taking into account important pedagogical and interface design principles. Access to output data and to input parameters is difficult. But the underlying dynamics are rich and in may cases do a very good job of indicating the trends forests take in specific forest regions.

Forest growth models continue to be an area of active scientific research, and we are familiar with a number of forest growth simulators, including FORET (Shugart and West, 1977), FORSKA (Prentice & Leemans 1990; Prentice et al. 1993), ZELIG (Urban, 1990; Urban et al 1991; Urban and Shugart, 1992), SIMA (Kellomäki et al, 1992, 1994; Kellomäki, 1995), SORTIE (Pacala, 1993). These simulation models have various differences, such as the number of tree species allowed, additional environmental constraints and chemical conditions. But all are similar in that they are based on the common gap model structure (Shugart 1984), and also in that they are geared more to professional forestry and graduate level study. Each of these programs has similar drawbacks for undergraduate research and classroom learning. The underlying equations are not readily inspectable. Parameters are hard to change and to relate to the outcomes of simulations. Graphical displays are rudimentary.

We propose to rely on the visualizations in our program to facilitate the process of discovery as well as to communicate the results. We will use improvements in technology to allow us to make a previously black-box forest simulator transparent.

2.4 Issues in Existing Simulation-Based Science Learning

Educational software for the sciences tends to fall into three categories. First are traditional multimedia titles. These tend to have limited degrees of interaction and incorporate shallow pedagogical models (for example learning by being told rather than learning by doing). Second is microcomputer-based laboratory software, which allows students to plug probes, meters, and other measuring devices directly into the computer for data collection and analysis. Although this type of software is beneficial, it is only indirectly related to the goals of this proposal. The third category, the one germane to this work, is simulation-based educational software that allows students to observe dynamic phenomena, gather and analyze data, and pose and test hypotheses using the experimental method. Simulations allow students to experience and manipulate phenomena that might be too dangerous, messy, expensive, inconveniently located; too fast, slow, big, or small for students to experiment with.

Existing simulation-based educational software has two almost universal shortcomings, which we will call the "black box problem" and the "meta-model problem." First, educational simulations are usually "black boxes" that do not give the student access to the underlying formulas or models that run the simulation. Second, simulations of natural phenomena incorporate a particular mathematical model of those phenomena. The simulation is used to help students understand that model. However, there are also underlying assumptions and emergent phenomena that are outside the scope of the simulation, yet are important educational topics. In the section titled "Glass Box Simulations and Emergent Properties" we describe these shortcomings in more detail, and describe our solutions to the black box problem and the meta-model problem.

Thus, our contribution will not only be the production of a particular piece of software and associated curriculum, but we will demonstrate and evaluate new methods that will have general applicability to all simulation based educational software.

3. Scope and Audience

We propose to build and evaluate a simulation-based inquiry learning environment in forest ecology. Deliverables will include student and teacher manuals, all available on the world wide web. The software will be used to supplement undergraduate instruction in Forest Ecology and Environmental Science and can also be used as stand-alone software in nature centers and other schools to allow users to experiment with and learn about forest growth ecology. The software support materials will encourage use of the software by pairs or groups of students, but individual use will also be possible.

We expect the software to be usable in high school, undergraduate, and graduate education classes. In addition to Evolution of the Landscape, The Ecology of Old Growth Forests and Ecology, we plan to introduce our software into courses in the Five Colleges and school districts in fields such as Ecology, Biology, Forestry, Land Use Planning and Local History. The software will allow a number of types of inquiry activities which could take from one classroom session up to two months of a course's scope, if desired. Also, as mentioned, we will report on several aspects of the project that have general application to simulation-based science education in any subject area.

We also include a novel evaluation methodology that combines evaluation with teacher training and teacher enhancement efforts, as described later.

4. Learning Environment Components and Features
The software will contain: 1) A mathematical simulation of tree and forest growth and an interface for visualizing forest growth and species succession over time; 2) Tools for setting simulation parameters, gathering data, and analyzing the data; 3) A Formula Inspector tool for inspecting and understanding the formulas and relationships underlying the simulation; and 4) on-line help and activity management features.
Figures 1 and 2 show screen mockups for the software. (Note that these are non-functional design pictures.)

Figure 1. 2-D Forest Site Map, Nursery Pallet, and Analysis window

Figure 1 left shows the 2-D Forest site window. The main simulation window, not shown here, will be a three-dimensional rendition of the forest plot, which the user will be able to zoom into and move around, and click on trees to view information about them. (By three-dimensional we do not mean fully immersive "virtual reality," but simply a rendering of the image that can be viewed from various locations and angles. The degree of visual realism will depend on the state of the art the software.) Users will be able to visually distinguish tree species and tree size. An average plot will be 30 meters in diameter, and contain from 20 to 100 trees. Figure 1 shows the birds-eye view of the forest plot. This two-dimensional view may be more easily understood in some situations. Though the trees in this figure look identical, in the actual software the varying sizes and tree types will be visually evident.

To the left of Figure 1 is the Nursery Pallet. It will show thumbnail images of the approximately 50 available tree species. Students can set up the initial conditions of a forest plot by starting with a blank plot and dragging new trees from the Nursery Pallet and "planting" them on the birds-eye view of the forest plot. Users will also be able to ask the software to create a random initial tree distribution with certain properties (e.g. 75% birch and 25% pine), or to load in a tree distribution that approximates particular global environments (e.g. eastern mixed hardwood forest). We expect that for some classroom activities students will map out an actual forest plot and then recreate this plot as the starting point of the simulation.

Figure 2: Pallets of the Forest simulation

Figure 2 shows a number of pallets that students will use to view information, take measurements, and perform actions. Figure 2 includes the Run Pallet. After initializing the forest with trees of various species, and setting the forest properties, the user can run the simulation and observe as trees grow, die, and become replaced with new trees. Usually as time progresses the percentages of tree species will change. For example, poplars and pin cherry give way to maple and yellow birch in northern New England forests. The simulation can be paused, stepped, and run forward and backward. Students can set the update increment and total time (e.g. update every 2 years and run for 100 years). Particular time slices can be frozen and saved for later analysis.

Figure 2 includes the Forest Management Tools pallet. This allows students to cut trees individually as in 'selective cutting" forestry practices, cut swaths of trees as in "clear cutting" practices, and introduce other "stress factors" such as pest insects.

Figure 2 includes the Tree Measurement pallet. Students have access to tools that measure tree properties and simulate actual forestry tools. These include the diameter tape and an increment borer. The increment borer is used to obtain a 1/4 inch diameter cylindrical cross sectional sample of the tree, showing the annual growth rings.

Figure 1 right shows the Site Properties and Analysis window. It is used to both view and set various properties of the forest. These properties include soil depth, soil texture, soil fertility; latitude, longitude, and slope of the plot; and various weather related factors such as rainfall and average temperature. Students will be able to load in pre-defined parameter sets that simulate particular locations on the earth. Students can graph trends in various properties, and calculate various emergent properties of the plot such as total basal area, the average tree density, and tree diversity. Users will be able to create bar charts to compare various properties as a function of tree type, over time.

Two additional software components are not shown. The first is the ability to introduce natural disasters and disturbances such as forest fires, floods, droughts, global warming, toxic spills, and hurricanes. These will amount to automatic changes in the environmental parameters and thinning of certain tree species. The second component is the ability to create longitudinal scenarios. With this tool the student or teacher can specify a scenario. For example, normal conditions might exist for ten years, followed by a flood, then 20 years of relative inactivity, followed by a forest fire. The student could then run this scenario and observe the results.
Museum Mode. The user interface will be made highly usable via progressive user tests and software modifications. Still, the software in its primary form will be most suited to use in classroom situations with the support of teacher and student manuals. For stand-alone use in nature centers and museums we will include a "museum mode" that is much easier to use and has fewer features available to the student.

5. Glass Box Simulations and Emergent Properties

Simulation-based educational software has two almost universal shortcomings which we will call the "black box problem" and the "meta-model problem." Educational simulations are usually "black boxes" that do not give the student access to the underlying formulas or models that run the simulation. In some cases this is purposeful. For example, in a physics simulation of the trajectories of falling bodies, the student's task is to measure the trajectories of various falling bodies and from this infer the underlying mathematical formulas; thus these underlying formulas are not accessible. However, in most situations, for example a simulation of weather systems or genetic recombination, it would be beneficial for the student to be able to inspect the form and function of the equations or rules that drive the simulation (de Jong & van Joolingen 1998). In many simulation models, such as those involving differential mathematics or feedback equations, the complexity of the calculations dictate that the student can't realistically see a derivational trace of how the formulas were combined and "run" to produce the observed behavior. Also, the way formulas are implemented for efficient computer simulation may not resemble methods described in textbooks on the subject. However, even if students can not see derivational traces, there is still great pedagogical value in being able to inspect the formulas as static entities.

Students engaged in inquiry learning about natural phenomena ask a variety of questions and the curriculum encourages them to ask such questions (Collins & Stevens 1983). We will describe these types of questions and then discus how they effect how formulas are represented in our educational simulation. Four types of inquiry questions are listed below, in order of increasing intellectual sophistication:

What if? For example: "what would happen if I started a forest with almost all birches and just two maples?" What-if questions can form the basis for goal-oriented hypothesis testing, or they can result from more open-ended trial and error "fiddling" with simulation parameters.
Relationship. Relationship questions focus on the relationship between parameters of the system. These questions are key to a conceptual understanding of the system as a whole. For example: "How does soil quality affect species diversity?" "What is the relationship between soil nitrogen and leaf size?"
Why? For example "Why does increased soil quality decrease tree diversity?" These questions delve deeper into the causal relationships and underlying assumptions.
Modeling. Modeling questions deal with creating new models or critiquing existing models of natural phenomena. They require an understanding that a model, formula, or simulation is an imperfect and/or approximate representation of the world (Soloway et al. 1998). Inquiry is opened to a meta level of analysis and creativity. Examples: "What would happen if we replaced the Basal Area formula with a more complicated one that takes tree density into account?" "Can I build a model that causes birches to out-compete maples instead of the other way around as happens in nature?"

Students working on answering all of these question types use and improve their inquiry skills. They must pose questions that can actually be answered. They must decide when they have run enough tests and have collected enough data to be able to make an inference. They must be able to analyze data to infer trends, patterns, or rules. They must be willing to modify their hypotheses and preconceptions in the light of new data.

The Formula Inspector: Using knowledge-based simulation models. Our solution to the concerns mentioned above is to open up the black box of the simulation and create a "glass box" simulation in which students can inspect the formulas that constitute the model of natural phenomena. Simply put, rather than encoding simulation formulas in raw "code" we will represent them as inspectable and manipulable "objects" in the simulation. Each formula object will have a number of properties, some which are used to run the simulation, and some which serve purely pedagogical functions. Individual variables, such as Tree Diameter, will also be "objects" in the system, and will have pedagogical information associated with them. The student will use a Formula Inspector, an interactive graphical interface, to access information about formulas. The Inspector provides a consistent framework for accessing multiple representations of formulaic relationships (similar to the epistemic forms described in Collins & Ferguson, 1993). It will enable students to observe the results of formulas, compare formulas, and explain phenomena in terms of formula (essential to inquiry investigation, Tabak et al. 1996). For example imagine that the student asks the educational simulation for what equations refer to total basal area. A list of formulas is shown, from which she picks the formula "SQI = (1-BAR)/BAMAX" for further inspection. The formula object for this formula would contain information that would allow the student to use the Formula Inspector to see the following information:

Feature Example or Description

Textual representation of the formula Soil Quality Index = (1 - Total Basal Area)/Maximum Plot Basal Area

Click on a variable or parameter and get a description of it. SQI is soil quality index, which determines how the intrinsic fertility of the site limits the growth of trees

Explanation of Units Explains the units in which each variable and parameter are measured

Explanation of the purpose entire formula This formula shows how the soil quality index is a measure of how close the soil is to the maximum possible growth capacity for a given plot.

Graphical representation A picture showing the qualitative relationship between the variables involved. For example a family of curves or an exponential relationship.

Show where the results are used Once SQI is calculated, it can be used in these equations: …

Describe the underlying theory behind the formula Text that describes how the formula was derived or discovered, for example, how a formula is due to the nature of chemical bonding in photosynthesis.

Assumptions, simplifications, and limitations to the formula The formula assumes that tree circumferences are perfect circles.

Alternative formulas For a more complex formula that takes into account circumferences that are not perfect circles, see …

Other features We have considered other information to include, such as the equation in differential and integral forms

Table 1: Features of the Formula Inspector

We call this a "knowledge-based" approach because the simulation contains not only formulas but also various sorts of pedagogically relevant knowledge about the formulas. It gives students multi-modal access to various perspectives for learning about equations and their relationship to observed phenomena. The forest growth simulation makes use of data tables listing growth parameters for each tree species. Students will be able to inspect these tables as well.

Next we will describe how the knowledge-based approach is used for each of the inquiry question types mentioned above.

1. What if? To answer 'What if" questions, students simply run the simulation and see what happens (they may also have to carefully organize their data collection and analysis). The Formula Inspector is not needed.

2. Relationship. Students can run experiments without inspecting formulas to infer relationships between variables, but they can also learn much about these relationships through the Formula Inspector.

3. Why? Students can not inquire at the "Why?" level with traditional educational simulations. This is due to the "meta model problem" which we will discuss below. With our knowledge based approach, students can get information about why a formula is what it is, and what the underlying assumptions are.

4. Modeling. A model is a set of formulas that describe a phenomena. The knowledge based approach allows students to create their own models by turning on or off certain formulas, or changing their parameters. We can also store alternative formulas that calculate a given variable. Though allowing students to create their own formulas from scratch is impractical, we can allow them to change key parameters and substitute alternatives for some formulas, thus allowing them to perform experiments in "model space." They can also perform formula or model verification by running a model (the original one, one they create, or an alternate one based on a competing theory) and comparing the results with what they observe in real world forests.

The knowledge-based approach allows us to implement alternative formulations, some more sophisticated than others, and facilitates a sequencing of curricular activities that progresses students toward increasingly sophisticated models, as in the Model Evolution approach (White & Frederiksen, 1995).

The Meta-Model Problem

All mathematical models have unavoidable limitations. Some formulas are like definitions, and are tautological. But some formulas represent empirically derived laws, or underlying scientific assumptions. The formula itself does not tell you anything about where it came from or what assumptions are built into it. For any formula or model one can always go one level "deeper" and ask "why?" For example, models for biology assume (or are based on) chemical phenomena, but this chemistry knowledge is usually not explicitly included in the model. Students studying biology need to be quite facile with biological models, and need to have just some familiarity with the chemistry that underlies them. We can go one step "deeper" still, for every formula or principle in chemistry has an underlying basis in atomic physics. Again, students studying chemistry need to be facile with the chemistry formulas but need only an introductory familiarity with the underlying physics. Clearly there is no end to the depth to which a student can ask "why." Yet for every discipline it suffices to have introductory information about the next deeper level of causality. This is provided in the textual information associated with formulas using the knowledge-based representation method.

In a complimentary fashion, every model has "emergent properties" that deal with a causal level "higher" than the formulas specified in the model. For example, we want students to be able to observe or infer global trends such as "species diversity decreases with the age of the stand." This relationship is not to be found explicitly in the model, yet it can be observed by running a number of simulations trials. Figure 3 illustrates the meta-model problem: that for any mathematical model of a natural phenomena, there will always be deeper fundamental assumptions, and "higher" emergent properties that are not represented explicitly in the model's equations, yet are important for students to become familiar with.

Fundamental assumptions à SIMULATION MODEL à Emergent properties
( à increasing model granularity à )

Figure 3: The Meta-Model Problem

We are not sure whether our knowledge-based representation of equations will be as useful in addressing the emergent properties as it will be in addressing the fundamental assumptions for equations. At the very least, information about emergent properties will be included in the exercises given to students, and in the curriculum materials provided with the software. As part of the proposed work we will investigate other methods for addressing emergent properties within the simulation itself.

6. Software Implementation

The software will be prototyped in Director, which has excellent support for 3D modeling of this type, but has limitations as an end-product software development tool. It is also cross platform, and can be delivered over the World Wide Web. Associates of our team have already created a number of sophisticated learning environments using Director and have ported some of them to Java.

After developing the prototype we plan to port the software to Java, which is a more robust and full featured programming environment. The software must run on both Macs and PCs.

7. Curriculum development and integration
The first year of the project will be primarily spent developing the software and running formative tests. The second year will primarily be spent developing curriculum and support materials to allow the software to be used in classroom contexts. We will develop a suite of activities for both field-based work outdoors and simulation-based work indoors, that address each of the major question types described above: What if, Relationship, Why, and Modeling. We will develop activity sheets and manuals/workbooks for both students and teachers. We will develop suggestions for how to introduce and use the software, how to integrate the field-based and simulation-based activities, and how teachers can evaluate student performance. We will be particularly concerned with how the software and the activities can be integrated into various types of existing classroom environments. We will develop separate manuals or sections of manuals the address using the software in different grade levels, different courses, and different classroom and teaching environments (small vs. large, student-centered, standard, nature centers, etc.).

Prototypes of program modules will be tested in The Ecology of Old Growth Forests, a 100-level course taught at Hampshire College each Spring semester, and in The Evolution of the Hampshire College landscape, another 100-level course taught each Fall. As the occasion arises will use parts of the program in upper level courses such as Ecology and Tropical Ecology. These courses, as well as the various evaluation sites mentioned in later sections, will provide a diverse array of settings, from formal classrooms to independent group projects.

8. Teacher Training and Enhancement Component

This project includes what we believe to be a novel combination of evaluation methodology and teacher education. After formative evaluations and the resulting improvements to the software to ensure it is robust and user friendly, it will be brought into both undergraduate and high school classrooms for testing. Science teachers in training, or practicing teachers undergoing teacher enhancement education, will carry out some of this classroom-based testing. Teams of two such student teachers will perform the evaluation. One of them will be trained in the use of the simulation software and in inquiry-based activities for integrating it into classrooms. The other will be trained in protocols and methods for observing and data collection. In this way student teachers will gain experience in inquiry based science education and will be learning something about the design, use and evaluation of educational technology. In addition, the student teachers will serve as the vectors for introducing both the simulation technology and the inquiry-based teaching methods into existing classrooms. Students will switch their roles as observer/evaluator and instructor/facilitator, to experience both roles over several trials. The design of the lesson plans, the design of the observation protocols, and the analysis of the data will be done by research scientists, who will also sit in as a third participant in some of the classroom trials. Although the primary intended audience of the software is undergraduates, we intend to make it applicable to high school levels, and even lower grades, as we will include a "museum mode" that will be appropriate for public use at nature centers. This is important to note because the student teachers will be teachers in training for K-12, and will sometimes be testing the software out in high school classes.

As this combination of teacher education with evaluation methodology represents an innovation, one of the things we will be reporting on is the success of this method. We will summarize what we have learned about the method so that others can try it, avoiding whatever pitfalls we encounter.

Because introducing new technology into existing classrooms is not easy, we will hold a summer institute during year 2 to prepare the teachers in training, in service teachers, and the teachers whose classrooms will serve as testing grounds. In this one-week institute we will familiarize participants with the software and the inquiry exercise. The pairs who will be bringing the technology into classrooms will receive substantial additional training. Teachers whose classrooms will be visited will have this week to prepare them for the software trials in their classrooms. It will allow them to better understand the nature of these trials, and be better able to help their students integrate the experience after the software trial is complete.

We have established a collaboration with the Five College/Partnership program to find suitable test teachers and classrooms. The Five College/Public School Partnership, created in 1984, serves an average of 800 school teachers and administrators a year from throughout western Massachusetts. The Partnership sponsors summer institutes (40 over 12 years) and academic-year seminar series (122 over 12 years), each of which is planned by teams of school and college faculty who share a common discipline or area of interest. The Partnership publishes a quarterly newsletter, the Partnership Calendar, which is mailed to over 5,000 school faculty and administrators in western Massachusetts, approximately 1,000 five college faculty, and an additional 500 educational leaders throughout Massachusetts and around the country. The Calendar is the major vehicle for recruiting teachers for Partnership- sponsored seminar series. It also provides information on events sponsored by other organizations including environmental education organizations and carries update columns from a number of groups including PALMS (see below).

Susan Thrasher, coordinator the Five-College Partnership, will work closely with us to set up workshops and test settings, using the Calendar to announce our work and announce workshops. We will interface with two programs in particular: STEMTEC and PALMS. Partnerships Advancing the Learning of Mathematics and Science (PALMS) is Massachusetts' statewide systemic initiative to improve the learning and teaching of mathematics and science. It bridges the gap between higher education institutions and school districts by bringing pre-service and in-service teachers together with researchers in collaborative projects. The Science, Technology, Engineering, and Mathematics Teacher Education Collaborative (STEMTEC) links the members of the Five Colleges, Incorporated consortium - the University of Massachusetts Amherst, and Amherst, Hampshire, Mount Holyoke and Smith Colleges - with the three area community colleges (Springfield Technical, Holyoke and Greenfield) and the neighboring school districts in a collaborative effort to promote reform in the science and mathematics preparation of teachers.

9. Research Questions
We are confident that we can build the forest simulation and the Formula Inspector without complications. We are also confident that students and teachers will find it beneficial and motivating to use the Forest Simulation software in classrooms, especially when combined with actual field work in forests. However, there are a number of unanswered questions that we hope to answer during this research. We will work with students and teachers to test various aspects of the software for usability and effectiveness, and make appropriate modifications. We will also work with students and teachers to develop one or more sets of activities and curriculum support structures that make best use of the simulation. More specifically, our questions are:

· What aspects of the software (both the simulation and the Formula Inspector) will students and teachers find most useful and educational?
· What aspects of the software (both the simulation and the Formula Inspector) will students find most complex and least "usable"?
· How should activities be sequenced? For example, should the Formula Inspector be turned off for the initial activities?
· What classroom and field-related activities are possible using the simulation? (We have thought of many but expect to discover others as we work with teachers and students.)
· How well do students learn about forest growth and forest ecology, compared with students who do not use the software?
· How well do students learn about scientific inquiry and the modeling of phenomena compared with students who do not use the software?
· How are students attitudes about the domain, and about science inquiry, effected by extensive (several weeks) use of the software?
· How are teacher's attitudes affected by having the software used in their classes?
There are some design features of the software that have not been worked out in this proposal, and remain to be done as part of the research:
· How can information about emergent properties of a model be built into the simulation?
· What is the best software interface, and the best activity/curriculum support to utilize the capability for students to create their own

Forest models?

Our model for knowledge-based representation of formulas, and whatever we discover in relation to the research questions above will not be limited to the domain of forest ecology, but will have wide application to all educational simulations of natural phenomena.
In addition, we will be using a novel evaluation methodology that combines teacher education with evaluation, as described in section 5. In addition to the above research questions for this project we will be asking about the effectiveness of this method.

10. Impact on students

Each semester about 25 Hampshire College students will work directly with Dr. Winship on this project. The 10 pairs of student teachers/evaluators will probably encounter up to 25 students per class, allowing us to have potential direct effect on over 250 undergraduate and secondary school students. The Environmental Studies Program at Hampshire College has a mailing list of over 100 students who attend seminars and workshops. We will give a few of these workshops, further expanding our test base to many more students with interest but without intensive training - a good test of the Museum or stand-alone version of the package. There should be a substantial ripple effect as the pre-service and in-service teachers enhance their general knowledge and skills in the areas of inquiry learning methods and the appropriate use of educational simulations. Many more students could benefit as the software becomes available over the world wide web.

Experience and Capability of the Investigators
Tom Murray will serve as the principal investigator for the proposed NSF work. Dr. Murray has managed a number of educational software projects in industry and in university research contexts. He is most known in the research community for his work (through the University of Massachusetts) developing authoring tools for advanced technology instructional systems. He teaches courses at the University of Massachusetts and at Hampshire College in instructional technology. (For more information see descriptions of research projects in the Results from Prior Research section, and the Biographical sketch.)

Larry Winship will serve as the subject matter expert for the project. Dr. Winship teaches courses in plant biology, forest ecology, sustainable agriculture and sustainable technology. His teaching methodology is primarily inquiry based and project-centered, and he will provide crucial pedagogical insights in the project. Much of his scientific research (biophysics of gas diffusion) has involved computer modeling. His recent research has included detailed surveys and analyses of the forests on the Quabbin Reservation in central Massachusetts. His undergraduate training included many Forestry courses and two summers as a research assistant for the U. S Forest Service, surveying gypsy moth impact damage in the N. E. forest.

Neil Stillings will design and help manage the evaluation component of the project. Dr. Stillings is a cognitive psychologist who is nationally known for his work in undergraduate education. He has organized and run national workshops on teaching cognitive science for the Sloan foundation and the National Science Foundation. He has received several grants from the NSF to develop laboratories and materials for inquiry-oriented instruction in cognitive science. He is currently a co-principal investigator and the director of a three-year project funded by the NSF Learning & Intelligent Systems program entitled Inquiry-based Science Education: Cognitive Measures and Systems Support. The research protocols and measurement instruments developed in that project will be applied and developed further in this project.

Evaluation and Work Plan

The software and curriculum development will be integrated throughout the project with cognitive and educational research. The research will address significant theoretically-based issues in cognition and learning in addition to more standard questions concerning content mastery and student-faculty satisfaction. An outside, independent expert will consult on the design of all research and will collaborate on the analysis of summative data.

Our evaluation plan includes both traditional and innovative methods. The innovative aspect is in the integration of teacher training with evaluation, as described in the previous Section "Teacher Training Component." The questions we wish to answer through evaluation were listed in the Research Questions section. The software and the associated curriculum materials (teacher and student manuals, etc.) will undergo evaluation in several stages, which are described in more detail in the Summary Work Plan section. These evaluation stages include:

· Formative laboratory-based evaluation is used to verify the usability of the user interface and give us confidence about its potential effectiveness.
· Pilot testing in realistic settings is used to verify the experimental data collection and evaluation methods. This will also verify the robustness of the software to work outside of laboratory conditions.
· Formative classroom-based testing will be used to gather data on the effectiveness of both the software and the curriculum materials, with the purpose of improving both.
· Summative classroom-based testing will involve final testing with pre and post-tests to gather data on the effectiveness of the software and curriculum materials.

Research will be conducted in at least two institutions at both the college and pre-college level. The software use and evaluation will predominantly involve pairs or groups of students. The evaluation will include analysis of: 1) Automatic program traces of student behavior, 2) A few audio or video taped observations of students using the software in pairs, 3) Pre-and post software use mastery tests; 4) Attitude and opinion questionnaires students; and 5) Attitude and opinion questionnaires teachers.

Summary Work-Plan

At the most generally level, our plan is to develop and formatively test prototype software in the first year; develop a complete package of activities and curriculum support during the second year, and evaluate both the software and the curriculum package in the third year. However, formative evaluation is integrated into all phases, as shown in the table below.

Fall Spring Summer

Year 1 Design simulation engineDesign simulation user interfaceBegin implementation of simulation software Complete simulation prototypeLaboratory formative tests of simulation prototypeImprovements of simulation prototypeBegin implementation of Formula Inspector Complete Formula Inspector prototypeLaboratory formative tests of Inspector prototypeImprovements of Inspector prototypeConference reports.

Year 2 Begin development of curricular materials.Laboratory-based formative evaluation of curriculum materials.Revision on curriculum materials. Develop evaluation protocols for classroom-based trials.Pilot test curriculum in two classrooms. Improvements to software and curriculum materials based on pilot tests.Conference reports.One week summer institute Train student teachers in software, curriculum evaluation protocols.

Year 3 Formative classroom-based trials in 2 undergrad and 2 high school classes.Analysis of data from formative classroom trials.Improvement of curriculum based on formative testing.Begin creation of Nature Center stand-alone version of the software. Summative classroom-based trials in three to four undergraduate classes and two high school classes. Formative tests on "Museum mode" in the lab and in two museum or nature center contexts. Complete Museum mode Spring trails data anal.Produce CD-ROMs for dissemination.Create web-based version of manuals.Create web-based system for distribution & support. Conference reports.Write final NSF project report.

Table 2: Work Plan and Evaluation Protocol

Dissemination
The inquiry-based software and associated student and teacher manuals, worksheets, etc. will be distributed via the Work Wide Web site dedicated to this project. This distribution method will allow easy updates and improvements to these materials, and will allow us to facilitate building an interacting community of educators using the software.

The research results will be reported via this web site and in at least two conferences each year, a biology education conference and an educational technology conference.

Participants in the project, including the approximately 10 teachers whose classes form test sites, and the approximately six student teachers who assist with the evaluation, will be vectors for carrying what is learned out into school systems.

Nedah Rose, Executive Editor of the Life Sciences Division at Saunders Publishing has expressed interest in marketing and productizing the software. She indicated that it could be used as a supplementary application in Ecology, Environmental Science, and the system-oriented Earth Science publications. The PI has worked with this publisher in the past in the area of instructional technology for introductory Geology.

REFFERENCES CITED

Educational Technology and Inquiry Learning References:

American Association for the Advancement of Science (1993). Benchmarks for Science Literacy: Project 2061. Oxford Press: New York.

Collins, A. & Furguson, W. (1993). Epistemic Forms and Epistemic Games: Structures and Strategies to Guide Inquiry. Educational Psychologist, Vol. 28, pp. 25-42.

Collins, A. & Stevens , A. (1993). A Cognitive Theory of Inquiry Teaching. In Reigeluth (Ed.) Instructional Design Theories and Models, pp 2470278. Erlbaum: Hillsdale, NJ.

de Jong, T. & van Joolingen, W. (1998). Scientific Discovery Learning with Computer Simulations of Conceptual Domains. Review of Educational Research, Vol. 68, No. 2. pp. 179-201.

Fatemi, Erik & Trotter, Andrew (1998). High-Tech Pathways to Better Schools. In Technology Counts, a special issue of Education Week on educational technology. pp. 23-68.

Ledeman, N.G. (1992). Students' and Teachers' Conceptions of the Nature of Science: A review of the Research. Journal of Reseach in Science Teaching, Vol. 29, pp. 331-359.

McNeal, A. & D'Avanzo, C. (Eds) (1996). Student-Active Science: Models of Innovation in College Science Teaching. Saunders Publishing, Philadelphia.

National Research Council (1996). National Science Education Standards. National Academy Press: New York

President's Committee of Advisors on Science and Technology: Panel on Education (March 1997). Report to the President on the Use of Technology to Strengthen K-12 Education in the United States.

Prince, G. & Kelly, N. (1996). Hampshire College as a Model for Progressive Science Education. In McNeal, A. & D'Avanzo, C. (Eds) Student-Active Science: Models of Innovation in College Science Teaching. Saunders Publishing, Philadelphia.

Roth, W. & Roychoudhury, A. (1993). The Development of Science Process Skills in Authentic Contexts. J. of Research in Science Teaching, Vol. 30, No 2. pp. 127-152.

Soloway, Pryor, Krajik, Jackson, Straaford, Wisnudel, & Klein (1997). ScienceWare Model-It: Technology to Support Authentic Science Inqiry. T.H.E. Journal. pp. 54-56.

Tabak, I., Smith, B. Sandoval, W., & Reiser, B. (1996). Combining General and Domain-Specific Strategic Support for Biological Inquiry. In Frasson & Gauthier (Eds). Proceedings of the Third International conference on Intelligent Tutoring Systems. Springer: New York.

White, B. & Frederiksen, J. (1995). Developing Metacognitive Knowledge and Processes: The Key to Making Scientific Inquiry and Modeling Accessible to All Students. Technical Report No CM-95-04. Berkeley, CA: School of Education, University of California at Berkeley.

Forestry and Ecology References:

Botkin, D.B. (1993). Forest Dynamics. Oxford Univ. Press: Oxford.

Botkin, D.B., J.F. Janak, and JR. Wallis. 1972.Some ecological consequences of a computer model of forest growth. J. Ecol., 60: 849-872

Botkin, D.B., J.F. Janak, and JR. Wallis. 1972.Rationale, limitations and assumtions of a Northeastern Forest Growth Simulator. IBM J. Research Development 16: 101-116

Botkin, D. B. and R. A. Nisbet. 1992 Forest response to climatic change: effects of parameter estimation and choice of weather patterns on the reliability of projections. Climatic Change 20:80-111

Dale, V.H, Doyle, T.W. and Shugart, H.H., (1985). A comparison of tree growth models. Ecological Modelling, 29: 145-169.

Kellomäki, S., Väisänen, H., Hänninen, H., Kolström, T., Lauhanen, R., Mattila, U. and Pajari, B. 1992. Sima: A model for forest succession based on the carbon and nitrogen cycles with application to silvicultural management of the forest ecosystem. Silva Carelica 22. 85.p.

University of Joensuu, Faculty of Forestry. ISBN 951-708-060-3. ISSN 0780-822232. UDK 630.182.2, 630.2, 574.4. Gummerus Kirjapaino Oy, Jyväskylä 1992.

Kellomäki, S. 1995. Computations on the influence of changing climate on the soil moisture and productivity in Scots pine stands in Southern and Northern Finland. Climatic Change 29: 35-51.

Kellomäki, S. and Kolström, M. 1994. The influence of climatic change on the productivity of Scots pine, Norway Spruce, Pendula birch and Pubescent birch in southern and northern Finland. 1993. Forest Ecology and Management 65 (1994) 201-217.

Kellomäki, S. and Kolström, M. 1992. Simulation of tree species composition and organic matter accumulation in Finnish boreal forests under changing climatic conditions. Vegetatio 102: 47-68.

Kellomäki, S., Väisänen, H., Hänninen, H., Kolström, T., Lauhanen, R., Mattila, U. and Pajari, B. 1992. A simulation model for the succession of boreal forest ecosystem.. Silva Fennica 26: 1-18.

Mills, A. V. (1993).Predicting forest growth and composition - a test of the JABOWA model using data from Earl Stephens' study in the Tom Swamp tract.
Senior Thesis, Hampshire College, Amherst, MA.

Pacala, S.W., Canham, C.D., and Silander, J.A., JR. (1993).Forest models defined by field measurements: 1. The design of a northeastern forest simulator. Can. J. For. Res. 23: 1980-1988.

Shugart, H.H. (1984). A Theory of Forest Dynamics. Springer Verlag, NY.

Shugart, Jr. H. H. and D. C. West. 1977. Development of an Appalachian deciduous forest succession model and its application to assessment of the impact of Chestnut Blight. Journal of Environmental management 5:161-179

Urban, D.L., 1990. A versatile model to simulate forest pattern: a user's guide toZELIG version 1.0. Department of environmental sciences, University of Viginia, Charlottesville,VA, 108 pp.

Urban, D.L., G.B. Bonan, T.M. Smith, H.H. Shugart, 1991. Spatial applications of gap models. For. Ecol. Manage., 42, 95-110.

Urban, D.L., H.H. Shugart, 1992. Individual based models of forest succession. In D.C. Glenn-Lewin,R.K. Peet, T.T. Veblen (eds.): Plant Succession: Theory and Prediction.Chapman and Hall, London, pp. 249-286.