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

Подождите немного. Документ загружается.

38Amaro Forests - Chap 33 25/7/03 11:10 am Page 386

39Amaro Forests Part 6 25/7/03 11:10 am Page 387

Part 6

Conclusions

387

39Amaro Forests Part 6 25/7/03 11:10 am Page 388

40Amaro Forests - Chap 34 25/7/03 11:10 am Page 389

34 Emerging Trends and Future

Directions: a Workshop Synthesis

David D. Reed,

Ana Amaro,

Ralph Amateis,

Shongming Huang

and Margarida Tomé

Abstract

Workshop participants divided into four working groups to examine emerging trends and

future research directions of forest ecosystem modelling from different perspectives. Each

group reported their individual discussions in detail in a common session. Here, a synthesis of

common items resulting from the four discussions is presented. While not being all-inclusive,

seven general issues were identiﬁed in three general research areas: (i) the design or develop-

ment of forest models or development of modelling methodology; (ii) the use of forest models;

and (iii) the evaluation and documentation of forest models. Taken together, this synthesis pro-

vides a workshop consensus on the future of forest growth modelling.

Introduction

Following presentation of the technical papers, workshop participants divided into

four working gr

oups to examine the emerging trends and future research directions

of forest ecosystem modelling from different perspectives. The four working groups

were organized around the following topics:

• forest reality and modelling strategies;

• mathematical properties and reasoning;

• estimation processes; and

• validation and uncertainty.

Here, the goal is to synthesize the four discussions and identify common items that

emer

ged from the respective discussions.

School of Forest Resources and Environmental Science, Michigan Technological University, USA

Correspondence to:ddreed@mtu.edu

Department of Mathematics, Instituto Superior de Gestão, Portugal

Department of Forestry, Virginia Tech, USA

Forest Management Branch, Land and Forest Division, Alberta Sustainable Resource Development,

Canada

Department of Forestry, Instituto Superior de Agronomia, Portugal

40Amaro Forests - Chap 34 25/7/03 11:10 am Page 390

390 D.D. Reed et al.

Emerging Trends and Future Research Directions

Seven general issues emerged from the discussions. Some of these were mentioned

by all groups, some by fewer, but all surfaced in at least two of the discussions.

These seven issues loosely fall in the following three areas:

• the design or development of forest models or development of modelling

methodology;

• the use of forest models; or

• the evaluation and documentation of forest models.

The individual issues are discussed below. The discussions were wide-ranging and

these summaries cannot possibly captur

e all of the points that arose. They are

intended to capture the general essence of the discussions and serve as a summary

of the consensus of the workshop participants regarding emerging trends and future

research directions in the ﬁeld of forest modelling.

Forest model development and modelling methodology

Methods used in developing models of forest growth and development are con-

stantly evolving. W

orkshop participants identiﬁed at least two general areas that

require further development to address current issues and needs, and suggested

activities that may increase the efﬁciency of developing methodological advances in

forest modelling.

Scale

Forest modelling really began with methods to estimate individual tree volume

and stand level yield beginning in the 18th century

. Today’s information needs

require the ability to develop models that can function across temporal, spatial and

hierarchical (organizational) scales. Temporally, questions of gaseous pollutant

impact may require modelling of physiological processes at the timescale of sec-

onds to minutes, while questions of sustainability and succession require

timescales of decades to centuries. Spatially, many contemporary questions must be

addressed at landscape levels (10

–10

ha) as opposed to focusing on individual

trees or single stands (10

–10

ha). This requires understanding and modelling of

physiological processes at the cell or organ level to understand pollutant impacts

on forest health and productivity, but assessment of long-term sustainability

requires extrapolating those results to the organismal, community and population

levels. Currently available modelling methodology can only begin to address issues

crossing several orders of magnitude temporally or spatially, and integration across

hierarchical levels is still in its early stages of development, despite efforts over the

last couple of decades.

Complex systems

Forest ecosystems are complex, non-linear systems with many feedback mecha-

nisms and interactions with external factors such as weather

, as well as human

intervention. All models are simpliﬁcations that may or may not consider various

aspects of system structure and behaviour, depending on the purpose of the model-

ling exercise. Factors such as climate have often been assumed to be constant in for-

40Amaro Forests - Chap 34 25/7/03 11:10 am Page 391

391 Emerging Trends and Future Directions

est modelling exercises, but advances in understanding of climate emphasize its

variability and non-static nature. Questions requiring the use of models for predic-

tive purposes may require that models have the ability to incorporate future climate

regimes for which there are no data on ecosystem response. Utilization of modelling

methodology with the ﬂexibility to represent system behaviour under such chang-

ing conditions is becoming increasingly important. It is interesting to note the anal-

ogy to automobiles; those manufactured in 2002 are much more complex than those

manufactured in 1913, but this added complexity increases functionality, reliability

and comfort for the occupants. Technologies that do not improve performance have

been added, but they usually end up disappearing after some time. It is important

that models are not made more complex simply for the sake of adding complexity,

but that model complexity be added to improve functionality or reliability.

Interaction with other disciplines

One needs to look no further than R.A. Fisher to know that applied problems in

agricultur

e and natural resources have spurred the development of modelling and

analytical methods applicable to an extremely wide range of disciplines. Similarly,

forest modelling practitioners have always borrowed liberally from other ﬁelds such

as econometrics to develop and utilize new modelling technologies. The extreme

efﬁciency of adapting modelling technologies to new applications, as opposed to

developing similar technologies or methodologies from scratch, should drive all for-

est modelling practitioners to actively pursue interactions with theoretical and

applied scientists from other disciplines as an efﬁcient route to identifying new

approaches and methodologies for particular problems. While this may seem obvi-

ous, it is easy to become insulated and focus interactions within a narrow group

interested in the same problem. It is critical that effort be continually expended to

maintain and expand contacts with investigators in other disciplines, whether for-

mally through membership in broad professional organizations, or informally

through day-to-day interactions within an investigator’s institution.

Integration with other resources

To date, models of forest growth and production, hydrological models, wildlife pop-

ulation models and models of other associated r

esources have been developed inde-

pendently of each other, even when the intent was to apply them to the same

geographical area. Resource management is requiring knowledge of multiple

resources and their interactions, yet models combining resources are only available

in very rudimentary forms. To provide information required by managers, and ulti-

mately by society, forest modellers must increase their interactions with specialists

from other resource areas. This implies not only an increase in the level of communi-

cation, but the development of truly integrated models that will of necessity require

the cooperation of specialists from many disciplines.

Use of forest models

Many of the barriers to adoption of forest models are due to models not satisfying

the per

ceived information needs of the user. This may occur for several reasons.

Models may not provide information in a form that is of utility to users, or they may

not provide critical information about a process that managers need to manage the

40Amaro Forests - Chap 34 25/7/03 11:10 am Page 392

392 D.D. Reed et al.

system. There are also cases where model results are difﬁcult to interpret or are not

suitable to the cultural background of the ultimate user of the information.

User involvement

Model utility can be maximized if users are involved in the design of model output,

egarding both the content of the information presented and the form of presenta-

tion. User information needs must be considered in model development. There are

many cases, of course, where models are applied for uses the developer did not

intend or consider. In this case, documentation of the limitations, as well as the

capabilities, of a model is critical. Model developers may also have opportunities to

interact with people interested in using the model in situations that were not antici-

pated during model development; it is important to take advantage of such oppor-

tunities.

The value of allowing model users to interact with models during execution,

oviding feedback during interactive use of the model and direction for further

execution, is being increasingly realized. In many ways, this leads to a modeller

thinking of a user as analogous to a black-box component or subroutine – informa-

tion is sent and processed, and new information is returned for use in further model

execution by other model components. This is very different from considering a user

to be a passive recipient of model results, or little more than a provider of directions

for model execution. This viewpoint leads to dramatically different approaches to

the entering of information or display of model results that we are only just begin-

ning to explore. It also expands the possibilities of model utility far beyond cur-

rently perceived boundaries. The development of new methods and approaches to

integrate user interaction into model execution will lead to great changes in both the

development and use of forest models.

Communication of results

The presentation of model results has evolved from simple tables, to charts, to com-

plex maps, and mor

e recently to real-time, web-based dissemination. As presented

by several authors in this volume, it is now possible to provide realistic visualiza-

tion of stands and landscapes. It is critical to remember the intended audience and

their capacity and skills in interpreting such information. Scientists in a research

institute have very different capacities for interpreting different types of displays,

and associated expectations of information content, than do rural farmers. It is

important to remember that such scientists are equally unable to interpret informa-

tion as typically perceived and processed by the farmers. Cultural differences in

such things as the perceptions of shapes and colours, or methods of describing loca-

tion, can make maps perceived as excellent by some almost indecipherable by oth-

ers. As model output inevitably moves to becoming more graphical, model

developers are going to be forced to develop and utilize dramatically different

methods of communicating results to model users.

Forest model evaluation and documentation

The question of when to trust model results is not new. Model users usually have to

either accept the claims of model developers or under

go extensive (and expensive)

model evaluation efforts themselves, which is often neither feasible nor possible.

40Amaro Forests - Chap 34 25/7/03 11:10 am Page 393

393 Emerging Trends and Future Directions

While not eliminating the need for users to implement their own evaluation

processes, the establishment of commonly accepted guidelines and standards of

model evaluation, and establishment of widely accepted descriptions of model per-

formance, would greatly simplify the process of model evaluation.

Benchmark data, standards and guidelines

Model evaluation and documentation has almost always been in the hands of the

developers. V

alidation data are often unavailable, and in the rare cases when they are

available, they are often only a subset of the development data. There are no com-

monly accepted guidelines for model evaluation, and no commonly accepted perfor-

mance standards for particular model uses. It is obvious that the purposes of

developing a model drive the choice of appropriate benchmarks and performance

standards. Unfortunately, this variety has contributed to a form of paralysis, where it

is impossible to assemble consistent comparisons, even for a particular, well-deﬁned

use. The development and acceptance of benchmark data, standards of model perfor-

mance, and guidelines of model evaluation for a variety of purposes are needed to

provide potential users with unbiased assessment of model performance. Workshop

participants identiﬁed several important actions in this regard. One is the establish-

ment of model documentation standards, such as those incorporated in the proposed

Forest Model Archive. The genomics disciplines have established the submission of

raw data to public archives as a broadly accepted practice that is even considered

mandatory by publishers under certain circumstances. A similar understanding

could possibly be reached with publishers of forest modelling research and docu-

mentation. A third area concerns the establishment of benchmark data sets.

Obviously, it would not be reasonable to establish benchmark data sets of every pos-

sible species and management situation, and it would not be reasonable to evaluate a

Eucalyptus model using benchmark Picea data. It might be possible, though, to

develop comprehensive, empirical descriptions of forest system behaviour that could

be compared with model results in an evaluation process. All these efforts will

require considerable coordination among members of the forest modelling commu-

nity, but establishment of infrastructure such as the Forest Model Archive is enabling

these developments in ways that have not been possible in the past.

Summary

Workshop participants identiﬁed emerging trends and future directions of forest

growth modelling. There were seven speciﬁc trends that were identiﬁed in three

general research areas.

• The design or development of forest models or development of modelling

methodology:

� issues of scale

� complex systems

� interaction with other disciplines

� integration with other resources

• The use of forest models:

� user involvement

� communication of results

• The evaluation and documentation of forest models:

� benchmark data, standards and guidelines

40Amaro Forests - Chap 34 25/7/03 11:10 am Page 394

394 D.D. Reed et al.

Aspects of each of these issues are presented and discussed in the context of

their impact on the development and utilization of forest models. These discussions

synthesize the perspectives and thoughts of workshop participants and, as such, can

serve as guidance for those interested in the development of this ﬁeld in the near

future.

Amaro Forests Index 1/8/03 11:54 am Page 395

Index

Abies

balsamea 158

sibirica 189

Accuracy 43, 47

MSE

5–6

Age adjustment method 82

ALBA 215, 216, 217, 218

ALDO 215, 216, 217, 218

Allometric

growth 160, 164

model 158, 160

ANUSPLIN 17–20, 22–23

ARCINFO 350, 351

ArcView 17, 18, 23

Australia

Tasmania 28

ictoria 28

AVVIRK-2000 296

BALANCE 210, 211

Balsam ﬁr 158

Bavaria 57

Belgium

Meerdaalwoud 146

Best linear unbiased estimator (BLUE)

135

Betula pubescens 230

Binary variable

360

Biomass 30, 31

ﬂuxes 27

model see

CAR4D

Brazil 66

Canada

Alberta 12–13

CAPSIS 319, 320–322, 323

CAR4D 27, 28, 30–31, 35

Carbon

balance see

GORCAM

cycle 144

sequestration 149–151, 152, 153

soil 36

stocks 37

Choristoneura pinus 123, 124

Classiﬁcation see

Cluster analysis

Climatic

change 152, 190, 194

dryness 190, 195

temperatur

e sum 190

temperature sum  dryness

194

Clone

182, 185

Cluster analysis 237, 239, 246

Communication 392

Competition

asymmetric 99, 100, 105

index

100–102

one-sided 99, 100

symmetric 99, 100

two-sided 99, 100

395