Amaro A., Reed D., Soares P. (editors) Modelling Forest Systems

Подождите немного. Документ загружается.

33Amaro Forests - Chap 29 25/7/03 11:09 am Page 346

346 J.J. Colbert et al.

Vanclay, J.K. (1994) Modelling For

est Growth and Yield. CAB International, Wallingford, UK, pp.

380.

von Bertalanffy, L. (1957) Quantitative laws in metabolism and growth. Quarterly Review of

Biology 32, 218–231.

34Amaro Forests - Chap 30 1/8/03 11:54 am Page 347

30 Landscape Visualization with

Three Different Forest Growth

Simulators

Falk-Juri Knauft

Abstract

Individual-based tree models are well suited to represent growth processes at scales where trees

interact through mutual shading and/or root competition. Management decisions in general are

performed at larger scales even when selective thinning is considered. Unfortunately, no satis-

factory scaling theory is available to upscale from the individual tree to the stand or even land-

scape level, but modern information technology provides several options to attempt

‘brute-force-solutions’ that circumvent the scaling problem, i.e. modelling a sufﬁcient number of

individual trees in a stand/landscape (depending on the application) to represent the spatial

distributions of properties of simulated forests with meaningful precision. This chapter will pre-

sent three different integrated systems/examples to approach this (scaling) task at various lev-

els of detail and realism. In all three examples the simulator kernel is individual tree-based; it is

linked to GIS functionality and a visualization interface. They differ in their purposes and hence

in number/detail of processes included and the resolution of the visualization interface. The

ﬁrst example, VIWA, integrates the tree-growth model SILVA with the tree visualization package

AMAP. Its main purpose is to serve as a communication interface between forest managers and

the public, requiring an almost photo-realistic visualization. The second example, VIWA2, is

aimed at forestry experts only and thus allows a more abstract presentation of trees. It combines

the forestry-oriented simulator BWIN with a landscape visualization tool ENVISION. The third

example is targeted at the communication between forest experts and ecosystem researchers. It

combines TRAGIC++ with the AMAP package and has an interactive user interface JTRAGIC.

Introduction

Communication about such complex problems as the goal-oriented management of

forest stands and landscapes is, especially in the case of discussion partners with

different backgrounds, a very difﬁcult task. Without appropriate visualization tools,

the subjective perceptions will be strongly diverse. To clarify and (potentially) solve

this problem we suggest focusing on an interactive simulation and visualization

environment.

Bayreuth Institute for Terrestrial Ecosystem Research, Germany

Correspondence to: falk-juri.knauft@bitoek.uni-bayreuth.de

34Amaro Forests - Chap 30 1/8/03 11:54 am Page 348

348 F.-J. Knauft

A simulation is a software-based implementation of models and therefore a

simpliﬁed representation of real world aspects. Common applications of computer-

ized simulation in forestry can be found in the ﬁeld of growth modelling. A number

of different approaches can be identiﬁed. Practical forestry-related groups of model

developers prefer empirical or holistic models, while groups oriented more towards

ecosystem research prefer process models.

The visualization of simulated or empirical data is a traditional method for

analysis but also for pr

esentation (e.g. of predicted scenarios). Because of the supe-

rior human ability to identify/recognize complex patterns, it is an effective interface

for the transfer of complex information. As the predominant traditional visualiza-

tion technology in forestry, the spatial context of many pieces of information (data)

led to a wide utilization of maps.

Interaction in the context of forest simulation has so far been realized inside the

model as interaction among individual tr

ees, and outside the model as interaction

between the simulator and its user. In the context of this chapter we include the inter-

action between people, who might be, for instance, different users of the simulation

tool.

The task of transferring information is supported by communication tools

(media). Maps of for

est units are one traditional example of documenting the con-

text for silvicultural decisions. Table 30.1 shows a selection of traditional communi-

cation tools in forestry and related sciences. They can be differentiated by their

partners, the type of possible interaction among them, and the corresponding tem-

poral and spatial scales.

State of the Art

The recent development in information technology has had an enormous impact on

this research-relevant technology. One of the biggest beneﬁts can be found in the

ﬁeld of simulation technologies. Just a few years ago, detailed simulation of individ-

ual-based tree growth was a difﬁcult and, in regard to computation capacity, expen-

sive task. Today, a number of different simulation tools can run on nearly every

modern PC. For the German region the most common are

BWIN (Nagel, 1997) and

SILVA (Kahn and Pretzsch, 1997; for a technical comparison see Windhager, 1999).

Both were initially developed from empirical models, but later supplemented by

Table 30.1. Traditional communication tools between foresters (F), scientists (S) and the public (P) (t →

point in time, t+dt → later point in time, ×→position, x+dx → different position) (the entry ‘no way’ marks

communication channels that face technical difﬁculties for the given task).

Communication tool Partner Space Time Direction

Forest maps F(t) → F(t+dt) Stand (ha) 10–20 years One way

F(t) → P(t)

Administrative records F (t) → F (t+dt) Stand (ha) 160 years One way

Joint visits to forest F(x) → F (x+dx) Group of trees Hours–days Interactive

Silvicultural books F (t) → F (t+dt) Variable Personal No way (?)

F(x) → F (x+dx) memory

Publications S(x) → S (x+dx) Variable Often < years One way

Political programmes, visions, goals P→ all Variable Variable Interactive ?

Predictions S(t) → all (t+dt) Variable Variable One way, No

way(?)

34Amaro Forests - Chap 30 1/8/03 11:54 am Page 349

349 Landscape Visualization with Forest Growth Simulators

process models. The opposite development can be seen for the growth simulator

TRAGIC++ (Hauhs et al., 1993, 1995, 1999; BITOEK, 2002a).

Forestry-related individual-tree-based growth models, such as

BWIN and SILVA,

differentiate the individual tree into only a few compartments, such as root, stem

and crown. Functional–structural growth models identify a much larger number of

compartments. They model speciﬁc functionalities for the compartments, and the

result of the simulation typically depends on the generated structure within the

individual tree. One approach of this type uses so-called L-systems (Lindenmayer,

1968). The idea of a rule-based behaviour of tree growth has been implemented in

the ‘growth engines’ of

GROGRA (Kurth, 1994, 1999) and AMAP (de Reffye and Blaise,

1993).

Parallel to the development of sophisticated growth simulators, visualization

technologies also advanced to new levels. T

raditional media such as graphs and

maps can now be generated with a few mouse clicks. The production of maps,

especially, gained much greater efﬁciency with the introduction of geographical

information systems (GIS). These systems even enable the user to create 3D maps

with very little effort. One step further is the creation of 3D scenes based on tech-

nology standards such as

VRML, OPENGL and DIRECTX (see Seifert, 1998). These are

used at different levels of resolution and quality as visualization tools in all of the

mentioned growth simulators. Most of the difference is caused by the high

demand of computation capacities for these visualization technologies, and their

varying availability.

The generation and processing of large quantities of data has become common

thr

ough use of the described technologies along with effective tools for data storage

and transfer. The spatial context of forestry-related data calls for GIS functionalities.

To transfer data of complex and variable structure among independent modules of

different standards, tools such as XML have been developed. The eXtensible

Markup Language is a mechanism to identify structures (content such as words, pic-

tures, etc. and some indication of what role that content plays) in a document

(ASCII ﬁle, stream, etc.).

Besides the exchange of information among the software packages, one major

task is the appr

opriate transfer to the user. For the realization of communication and

interactivity between the user and the software, the graphical user interface (GUI)

became the standard with the introduction of the Apple Macintosh. Today, conve-

nient developer kits enable the engineering of adapted GUIs.

Methodology

The three modules concept

As the general structure of our forestry-related information system we concentrate

on thr

ee main components: simulator, database and visualization tools (Knauft,

2000). This structure has been realized to various degrees of integration in profes-

sional forest information systems (

ABIES, FOGIS–INTEND Kassel), as well as in academic

systems (

VIWA, TRAGIC, FORCITE).

The availability of efﬁcient tools such as GIS, forest growth simulators and tree

or landscape visualization tools, combined with limited ﬁnancial r

esources, favours

a modular concept. Table 30.2 presents an overview of the criteria which have

inﬂuenced the selection in the presented applications.

34Amaro Forests - Chap 30 1/8/03 11:54 am Page 350

350 F.-J. Knauft

Table 30.2. Case-speciﬁc criteria for software selection.

Partner Case-speciﬁc selection criteria Selected software

Forester/general public All components already available, SILVA – ARCINFO – AMAP

high claims on realistic visualization

Forester/forester BWIN and ARCINFO already available, BWIN – ARCINFO – ENVISION

visualization software as freeware

Forester/ TRAGIC and ARCINFO already available, TRAGIC – ARCINFO – AMAP

scientiﬁc community high claims on detailed visualization

A ﬁrst design concept – VIWA

Our ﬁrst idea to develop an integrated forest simulation and visualization system

grew from the demand for a tool to support forestry management decisions by illus-

trating their impact on the forest landscape over a time scale of decades (Knauft,

2000; Knauft and Sloboda, 2000). The task deﬁnes the participants in the communi-

cation problem: foresters and the general public. It demands a reliable simulation of

forest growth and management up to a full rotation cycle, a ﬂexible and powerful

spatial database for the processing of stand and individual tree data and, ﬁnally, an

almost photo-realistic visualization of the scenery including botanically correct

three-dimensional forest tree models (in contrast to ‘billboard-icons’ as used in

ENVI-

SION (McGaughey, 2000)). The solution was found in an integration of the forest

growth simulator

SILVA, the GIS ARCINFO, and the tree architecture visualization

toolkit

AMAP.

The data processing starts in the GIS with sets of stand data and related geo-

graphical entities.

A region can be selected using a map interface. The stand data (f.i.

inventory data) of the selected region are transferred to the simulation tool, which

generates samples of the initial individual tree structure of the stands. With these

structure samples, the full stand structure is created inside the borders of the stand

unit and stored in the GIS. Based on these samples, additional growth and manage-

ment intervals can be simulated. These can be used in turn to create scenarios of for-

est landscape development. The ﬁnal step is the transfer of the individual tree data

and the digital elevation model to the visualization tool, where single images, time

series or movements through the scene in image series or animations can be ren-

dered and recorded. The implemented system

VIWA (short for ‘Virtueller Wald’ - vir-

tual forest) is costly and complicated, due to the use of commercial software

(

ARCINFO and AMAP) on three different computers with different operating systems.

The second example –

VIWA 2

These resources were not affordable for a second application. The Forest Research

Station of Lower Saxony asked for a system to support communication within the

for

estry community which should be efﬁcient to use. Therefore, we chose, as well as

the already licensed GIS

ARCINFO, the simulator BWIN (developed at the research sta-

tion) and the visualization tool

ENVISION (McGaughey, 2000). For efﬁcient use, all

these products can be run on a single computer (operating system MS Windows-NT

and upwards) and can be accessed over a GUI, which also controls the interfaces

between the software components (see Fig. 30.1). The data handling of this system is

generally the same as in the ﬁrst system, but via the GUI it is much easier to handle

and to control.

34Amaro Forests - Chap 30 1/8/03 11:54 am Page 351

351 Landscape Visualization with Forest Growth Simulators

Graphical user interface

BWINPRO

Silviculture,

growth,

mortality

ARCINFO

tree data,

stand data,

objects

ENVISION

Images,

animation of

movements

and processes

Topography,

Simulation Database Visualization

Fig. 30.1. Sc

hema of the solution for intra-forestry communication, VIWA 2.

The TRAGIC concept

Finally, at the Bayreuth Institute for Terrestrial Ecosystem Research (BITÖK) we are

developing a communication interface between ecosystem scientists and practical

experts in silvicultur

e. The interface for this speciﬁc case places much greater

requirements on the system than the previous applications. Scientists hope to

receive feedback about the quality of their models and the foresters look for a tool

that enables them to document, communicate and plan their management activities

more efﬁciently. Hence information is required to ﬂow in both directions and the

visualization tool becomes an input interface for the expert forester.

Waiving practical experience as an input and feedback to theoretical research

would be a gr

eat loss. On the opposite side, theoretical research could be interro-

gated as to whether its results would have an impact on practical forestry (see

Schanz, 1994). Therefore, we chose a highly realistic three-dimensional and interac-

tive visualization allowing interactive movements in the virtual scene and interac-

tive action such as the selection of trees for thinning. The system consists so far of

the individual-based forest growth simulator

TRAGIC++, the Internet-accessible visu-

alization front-end

JTRAGIC (Hauhs et al., 1995, 1999; Scheerer, 2000; BITOEK, 2002b)

and a database. The integration of a GIS (here also

ARCINFO) will enable the visual-

ization of landscapes and therefore allow the discussion about forest management at

this scale. The generation and simulation of the data will be realized in the same

way as in the ﬁrst two applications.

The demand for interactive access to tree and stand data and for simulated

movement thr

ough the stand does (under the constraint of the limited graphics

capabilities of the sort of computers that are typically available in forestry ofﬁces)

limit the possible detail level in the visualization. For special situations with higher

expectations, the

AMAP toolkit will be integrated at the server site to generate photo-

realistic still images and recorded animations. So far the transfer of all simulated

structural information (

TRAGIC++ even simulates branches and needles) into the vir-

tual

AMAP scene is not ﬁnished, but when completed it will enhance the discussion

between scientists and foresters.

The architecture of the

TRAGIC environment is illustrated in Fig. 30.2. While the

data-server operates the spatial database (tree and stand data) as well as a knowl-

edge repository (silvicultural activities), the simulation-server runs the

TRAGIC server

34Amaro Forests - Chap 30 1/8/03 11:54 am Page 352

352 F.-J. Knauft

TRAGIC

server

TRAGIC

AMAP

User 1

Database server

GIS

Knowledge

repository

Simulations server

Client PCs

JTRAGIC

User 2

JTRAGIC

Actions

Stand

Actions

Stand

StandStand

Stand

Actions

Image

Fig. 30.2. Schema of the fully developed TRAGIC environment as server–client design with interactive user

interface (dashed line, data before simulation; dotted line, data after simulation).

(central data interface), the simulator TRAGIC++ and the visualization tool AMAP. On

the client side, the user can download an older stand, plan its activities, and send

these back to the server, eventually with an additional call for an

AMAP visualization.

After the simulation, the stand data (as individual tree data) are stored in the data-

base along with the related activities. The resulting stand can then be downloaded

from the original user or any other user with appropriate permission.

Data sources

The following applications are based on data from three different regions in Lower

Saxony

For the case of communication with the public, we chose stand data from the

exhibition gr

ound ‘ErlebnisWald’ (‘Experience Forest’) near the town of Uslar. We

included the stand parameters of species, basal area, upper height, stand age and

regeneration type (planting or natural regeneration) in the generation of individual

tree data samples by

SILVA. Those were, as mentioned before, multiplied in the GIS

for the full stand area and stored together with a digital elevation model. For the

simulation of management and growth, the samples were again imported and

processed within

SILVA. Only tree position, individual height and species parameters

were transferred to the visualization tool.

The application for the forestry internal discussion illustrates a plot from the

Luenebur

g Heathland. For this case we were supplied with a full tree data table.

Besides position, tree height and species as well as the DBH and lower crown height

were utilized for the visualization. The last example is based on a variety of stand

and management information about the Lange Bramke catchments in the northern

parts of the Harz mountains (see also Hauhs, Chapter 5, this volume).

34Amaro Forests - Chap 30 1/8/03 11:54 am Page 353

353 Landscape Visualization with Forest Growth Simulators

Results

The VIWA example

The ﬁrst scenario (Fig. 30.3) presents a possible discussion about the impact of

changes in the for

est management objectives from an intensively managed forest to

a more extensive use with a ‘reconstruction’ of a potentially natural forest type, as is

often demanded by environmental interest groups. To support the relatively abstract

discussion, we generated a virtual scene of a speciﬁc forest landscape.

Especially for untrained persons, a photo-realistic visualization is suitable, as it

corr

esponds best with their experience. Additionally, public expectations of virtual

scenes are highly inﬂuenced by the quality of television and ﬁlm productions.

Therefore, highly detailed and botanically correct trees are demanded. Only very

few, mostly professional, rendering systems (such as

AMAP) are able to produce this

quality. They are quite demanding in terms of computational resources. Important

in the context of this chapter is the aggregation of stand structure information into a

landscape view. Less weight is given to a high degree of detail and realism, which

was not demanded for this situation (see also Knauft, 2000).

The

VIWA 2 example

The second case focused on supporting discussions of real or potential situations of

for

est development within the community of foresters and people of similar back-

ground. This group is relatively homogeneous with respect to their knowledge

about structure, growth and management of forests; they share a common expertise.

On the other hand, the communication must be effective and precise. The main top-

ics of their discussion are structure, growth rate and yield, but also stability and eco-

logical value.

In this case, silvicultural research results are presented to a professional audi-

ence. T

o illustrate the method and for better understanding of the situation, the

sampled areas are visualized using the more abstract tree models of

ENVISION (see

Fig. 30.4). Although the visualization seems to be more schematic, it contains

Fig. 30.3. Comparison of visualizations of a forest landscape under recent stand-wise forest management (a)

with the potential historical forest type (b). Both graphics generated with AMAP, reproduction from Knauft

(2000).

34Amaro Forests - Chap 30 1/8/03 11:54 am Page 354

354 F.-J. Knauft

Fig. 30.4. Visualization of a monitoring plot (graphic generated with ENVISION).

more veriﬁable information. While in the ﬁrst application only the generated

parameters of position, species and tree height were used for the visualization, in

this second example, DBH and lower crown height were also taken into account

for the visualization, and all data were actually measured for every individual

tree.

The

TRAGIC example

A combination of the visual realism as implied by the ﬁrst

VIWA system and the data

representation in the second version can be realized within the

TRAGIC environment.

It consists of the simulator

TRAGIC++, which includes a tabular interface and also a

visualization tool (Fig. 30.5a) and the visual interface

JTRAGIC. The latter does not

include its own simulation module; it is only able to visualize a stand as transferred

from a central database and record the silvicultural activities (movements or silvi-

cultural decisions such as cutting or planting of trees) of the forester. These are sent

back to the simulator and used there for the next simulation period. The example

presented here is part of the restoration of historic stand development (for more

Fig. 30.5. Visualization of a historical plot showing the situation in 1752 as part of a reconstruction from

the 18th century with TRAGIC (a) (see Hauhs, Chapter 5, this volume) and after its export to AMAP (b). The

position and size of stems and crowns are the same in both scenes.

34Amaro Forests - Chap 30 1/8/03 11:54 am Page 355

355 Landscape Visualization with Forest Growth Simulators

details please refer to Hauhs, Chapter 5, this volume). Some indication parameters

are used to evaluate the simulated stands and their structure by the researcher, but

also by foresters. A number of foresters who have been interviewed demanded an

increase in the visual realism of the trees (Schanz, 1994). We included an interface to

the

AMAP toolkit (Fig. 30.5b). For a full stand impression, the plot can than be multi-

plied in the scene (Fig. 30.6).

Discussion and Conclusions

The characteristics of the communication partners

For the ﬁeld of forest modelling we can identify three major interest groups.

1. The general public possesses little knowledge about the complex processes of for-

est ecosystems. The public is primarily inter

ested in changes at the landscape scale,

as they are interested in the aesthetics and recreational value of forest landscapes,

rather than in individual plant dynamics. Their preferences become settled in poli-

tics, media and economy.

2. Experienced foresters have an inside view about processes which inﬂuence the

dynamics of for

ests, but they are also more interested in the (economic) result on

enterprise (or landscape) level rather than individual trees. Based on practical expe-

rience, they have the competence to achieve their goals for speciﬁc situations.

3. Less homogeneous are the ecosystem-related scientists, as they focus on individ-

ual or

ganisms, populations or ecosystems. Their competence is based on theories,

experiments and observations, often at the process level.

These three partners are already engaged in several communication processes,

but most of these traditional channels ar

e technically characterized as ‘one-way’ (see

Table 30.1). If interactive communication occurs at all, then it is usually only within

the individual groups. Here the question is whether this is a technical constraint

(there being no proper interactive communication tool), a social constraint and/or

whether it is due to the organization for the tasks at hand?

In the schematic summary of established communication tools in Table 30.1, the

link between r

esearch and management is not covered (see also Schanz, 1994). It is

argued in another chapter (Hauhs, Chapter 5, this volume) that the reason for this

gap is that the relationship between empirical/heuristic models of management and

Fig. 30.6. Visualization of a full stand based on the historical plot reconstruction from the 18th century.