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

Подождите немного. Документ загружается.

35Amaro Forests - Chap 31 25/7/03 11:09 am Page 366

36Amaro Forests Part 5 25/7/03 11:10 am Page 367

Part 5

Model Archives and Metadata

Forest models are a valuable summary of the forest data from which they are derived. Such

models encapsulate the information and knowledge within the source data and, as such, are

a primary means of understanding forest structure and processes.

There are no standards for the description or presentation of modelling metadata. As a

consequence, metadata ar

e presented in varying ways; in many cases, there is little common

information available for models developed for similar systems. This makes it virtually

impossible for third-party users to adequately evaluate available models and make decisions

on model suitability for their situation.

Having considered the need for, and implementation of, a for

est model archive to conserve

and enhance the sharing of forest models of the past, it is natural to consider the environment

for the development, publication and sharing of future forest models. Workshop participants

reviewed the current state of the Forest Model Archive, which was initiated following an ear-

lier IUFRO meeting in Greenwich, UK, in 2001.

The editors thank Keith Rennolls (UK) for his coordination of the contributions on this topic.

367

36Amaro Forests Part 5 25/7/03 11:10 am Page 368

37Amaro Forests - Chap 32 25/7/03 11:10 am Page 369

32 Forest Modelling: Conserving the

Past and Building the Future

Keith Rennolls

Abstract

Forest models are a valuable summary of the forest data from which they are derived. Such

models encapsulate the information and knowledge within the source data and as such are a

primary means of understanding forest structure and processes. At a time when there is much

discussion and activity concerned with building distributed data warehouses of forest data, it

is incumbent on forest modellers to consider the conservation of forest models that have been

developed to-date. Accordingly, the establishment of a forest model archive (FMA) has been

proposed by Rennolls et al. at an IUFRO 4.11 meeting in Greenwich, June 2001, and such an ini-

tiative has been adopted by IUFRO 4.11, and subsequently also by IUFRO 4.01, at the IUFRO

4.01 Vancouver conference, 2001. The ﬁrst part of this chapter reports on the progress in this

initiative since then. In particular, some of the features of a ‘small’ prototype FMA

(www.forestmodelarchive.info) are described and discussed.

Having considered the need for, and implementation of, a forest model archive to conserve

and enhance the sharing of forest models of the past, it is natural to consider the environment

for the development, publication and sharing of future forest models. This has led to the for-

mation of a new peer-reviewed e-journal, Forest Biometry, Modelling, and Information Sciences,

(FBMIS), which was launched in the summer of 2002 (www.FBMIS.info). The aims of FBMIS

are presented and discussed.

Introduction

Forest modellers, whether forest biostatisticians, mathematical or computer mod-

ellers, or information scientists, generally content themselves with dealing with the

forest data that they have available to them, the process of storing or managing it,

building models based upon it, and making such models and their interpretations

available to forest managers. This work is crucial to the scientiﬁc process of research

in the forest sciences, because the models we produce are a valuable concise descrip-

tive summary of the patterns, information and knowledge that are embedded in the

forest data that we work with. Further than this concise capture of information and

CMS, University of Greenwich, UK

Correspondence to: k.rennolls@gre.ac.uk

37Amaro Forests - Chap 32 25/7/03 11:10 am Page 370

370 K. Rennolls

patterns in calibrated descriptive models, the forest modelling community has

developed a variety of modelling approaches that give insight into, and understand-

ing of, forest structure and process, and hence allow the possibility of predicting for-

est behaviour in contexts that go beyond the range of available forest data.

Stand-level models have been developed as appropriately parameterized non-linear

models of forest processes such that the ﬁtted parameters relate to observable physi-

cal features of the forest, and the forest environment. Spatial individual-tree models

of forests allow the processes of tree competition, growth and mortality to be mod-

elled with a framework that can be easily used to simulate forest behaviour under

novel management regimes. Process-based models allow the ﬂow of forest growth

resources through trees and the stand to be modelled, and simulations from such

models allow the impact of environmental pollution on the growth and production

of the forest to be evaluated. Clearly, forest models are of paramount importance in

the representation and understanding of forests, not only for research purposes, but

also for forest management purposes. There are also substantial modelling contribu-

tions relating to populations in forests, and populations dependent on forests.

Now, at the start of the 21st century, we live in a vast electronic community of

the W

orld Wide Web. Large international efforts are under way to set up large dis-

tributed data warehouses of forest data (e.g. the IUFRO GFIS taskforce). There is

much talk of data-mining such large data warehouses in order to increase our

knowledge and the understanding of the world’s forests, both for forest manage-

ment and for conservation purposes. However, forest modellers have been doing

such ‘data-mining’ of forest data resources for several generations, and the models

that have been obtained are the ‘discovered information and knowledge’. It is par-

ticularly apposite, therefore, that the forest modelling community take stock of what

it has achieved, and consider whether the valuable heritage that exists in forest

modelling is being satisfactorily conserved, so that it may be shared, both on a

worldwide basis currently, and with subsequent generations of forest modellers.

Such considerations led Rennolls et al. (2001) to suggest the need for a forest model

archive (FMA). Some initial considerations for an FMA, and a report on the estab-

lishment of a simple prototype FMA are described in the next section.

However, modellers are always keen to move on to the next, and better, model.

Hence, while establishment of an FMA

is an important consideration for the conser-

vation and sharing of forest modelling heritage, we should also consider the envi-

ronment of the future in which models may be developed, published and shared.

Clearly, the FMA could develop in time and grow with future modelling achieve-

ments, and in so doing would provide a valuable means of sharing best modelling

practice on a worldwide scale. However, publications about forest modelling stud-

ies are scattered across a wide variety of theoretical and applications journals; there

is no focused publishing outlet for the forest modelling disciplines. While such pub-

lication outlets have played, and will continue to play, a major role in the publica-

tion and communication of forest modelling studies, these publication outlets are

not generally available in the developing world, because of the high cost of sub-

scriptions to such journals. IUFRO, with its conferences organized throughout the

world by various IUFRO groups and working parties, has been a primary forum for

communication between forest modellers. However, such IUFRO conferences are

generally only attended by scientists from the developed world, because of the high

costs of attendance at such meetings, and consequently the proceedings of such con-

ferences usually do not reach potentially needy recipients in the developing world,

because no worldwide dissemination framework exists.

It is therefore concluded that a new and more focused publication outlet in for-

est modelling is r

equired which is freely accessible by forest researchers, and in par-

37Amaro Forests - Chap 32 25/7/03 11:10 am Page 371

371 Conserving the Past and Building the Future

ticular by forest researchers and managers in the developing world. Accordingly, a

new peer-reviewed e-journal, Forest Biometry, Modelling, and Information Sciences

(FBMIS), was launched in September 2002. FBMIS is intended to provide free access

to anyone with access to the Internet and it is hoped thereby that the journal will

enhance communication worlwide in the forest modelling sciences. Details about

FBMIS are presented here, together with a discussion of links between FBMIS and

the FMA.

A Forest Model Archive

Initial considerations

Rennolls et al.

(2001), in a session on ‘Information Sharing’ at a IUFRO 4.11 confer-

ence (Greenwich, UK), suggested the need for a forest model archive (FMA) to con-

serve the heritage of forest models that had been developed and calibrated over

several decades, and to enhance sharing on a worldwide basis. The main points of

justiﬁcation have been outlined in the Introduction of this chapter, and are covered

more fully in Rennolls et. al. (2001).

Rennolls et. al.

(2001) considered three alternative approaches to development

of such a FMA, while indicating that an FMA of the future might have features of all

of them.

The ﬁrst approach was based on state-of-the-art information systems tech-

niques, and was described as:

…a system that is a multi-tiered client/server system with clients running Web browsers

linked to Web servers and, as required, to application servers and/or the data and

metadata servers. The application software architecture adopts the software engineering

principle of separation of concerns and is thus multi-layered. Each layer is designed to

perform a coherent function or a set of closely related functions.

(Berners-Lee et al., 1999, 2000, 2001)

Such an architecture is consistent with professional IS standards within which (well-

resourced) forest information systems are currently being developed. However, such

an approach requires major inputs from a widely skilled team of information scien-

tists and programmers, experienced in distributed database technology (e.g.

CORBA,

SOAP) and a variety of programming languages (Java, C++, XML). Such develop-

ment would undoubtedly need substantial funding for development to lead to a

usable system

The second approach considered by Rennolls et al.

(2001) was based on recent

developments in the computer communications industry. Partly based upon the suc-

cess of the MP3 standard as a basis for sharing music across the Internet, it was con-

sidered that Gnutella, a newly introduced distributed communications system,

might provide the basis for a distributed network of model archive nodes, with each

node being the home location of a subscribing forest modeller.

However, as in the ﬁrst approach, development of such a system would require

high-level skills in state-of-the-art systems and pr

ogramming languages. Such skills

are not easy to come by, and are usually in high demand and therefore very costly. In

fact, it now seems that Gnutella has gone out of fashion in the communications indus-

try and that the .net environment is now proffered as the new hope for a worldwide

operating system that would allow the aims of this approach to be achieved.

The third approach considered by Rennolls et al.

(2001) was a simple website

using ASP and Microsoft as standard:

37Amaro Forests - Chap 32 25/7/03 11:10 am Page 372

372 K. Rennolls

The FMA would be a simple web site with an ASP/HTML interface that is accessed via a

standard browser, making use of a simple ACCESS database for the storage of various

registers and model metadata, with the community contributing to a single archive,

which might possibly be mirrored. The archive would be maintained by an ‘archivist’

(group). The model contributors could merely email the archivist the relevant ﬁles which

would then be made available on the FMA site.

(Paraphrased slightly from the original.)

Simplicity is, in fact, an important design criterion. The recent history of software

developments is littered with expensive and over-ambitious projects that never deliv-

ered the project’s aims. Simple prototypes usually do deliver, do so cheaply, and natu-

rally lead to realistic extensions in aims, design and implementation. Also, choice of a

software environment that has proved itself effective, even if rather outdated in com-

parison with the most recent and fashionable software offerings available, is a reason-

able prototyping approach in the rapidly changing world of software fashion.

The reader will undoubtedly see that it was almost inevitable that the only

ogress possible in the year since the Greenwich Initiative, in a totally unfunded

context, was the third approach listed above. This has indeed been the case, and a

description of the resulting prototype FMA is presented. However, the humble pro-

totype FMA presented here does not preclude future developments moving towards

the ﬁrst and second approaches described above, particularly if supporting funding

were to be available.

The Greenwich Initiative

Following the presentation of Rennolls et al.

(2001) at the Greenwich meeting, an

impromptu discussion of the ‘Trafalgar Group’ (consisting of Risto Päivinen, Tim

Richards, Keith Rennolls, Keith Reynolds, Moh Ibrahim, Alex Fedoric, Peter Smith,

Robert Muetzelfeldt and John Palmer) considered various factors and other

approaches to the development of an FMA. The result was the Greenwich Initiative

on FMA development. The following were the main factors and considerations

thought to be of particular importance in this initiative.

1. The initiative would be named Forest Model Archive (FMA).

2. That consideration would be given to logic frameworks relevant to the FMA. This

discussion would focus on use of logic models as higher

-order abstractions for speci-

fying logical relationships among potentially diverse models.

3. That consideration be given to the delivery framework of the FMA.

4. That metadata and user-orientation be given particular consideration.

5. That model speciﬁcations and implementations be considered.

6. That the Greenwich Initiative should be linked to the IUFRO GFIS project.

7. That the initiative could possibly be extended to include the co-activities of other

IUFRO gr

oups, in particular IUFRO 4.01, Mensuration, Growth and Yield.

In order to focus the actions of the initiative, a 1-year time horizon was set, with the

following two main actions determined:

A. That a Yahoo chat thread on the various aspects of the initiative be set up.

B. That a workshop be conducted in Summer 2002 to discuss any progress.

At the IUFRO 4.01 meeting in Vancouver in August 2001, an invitation was

extended to IUFRO 4.01 to join the Gr

eenwich Initiative. This was accepted and

point 7 fulﬁlled. Action B is fulﬁlled to some extent by this chapter. Action A was

implemented by K. Reynolds, but unfortunately the discussion never really took off.

37Amaro Forests - Chap 32 25/7/03 11:10 am Page 373

373 Conserving the Past and Building the Future

A simple prototype FMA

A simple FMA prototype has been developed by the author along the lines of the

thir

d approach described in the previous section. It is available online at www.

forestmodelarchcive.info. It is an open-access website which allows model submis-

sion in two forms.

First, there is a Register of Forest Models (RFM) that has been modelled on the

to pr

ovide consistency of metadata standards between the Registers. See

www.ibilio.org/london/agriculture/links/1/msg00128.html for the original

announcement of this REM service or go to eco.wiz.uni-kassel.de/ecobas.html to

see the REM.

Secondly, a full submission of forest models is possible (subject to peer review),

and this leaves the copyright and IPR of contributed models with the model

authors. User r

egistration, which assumes an acceptance of these conditions, is

required before access is granted to these full models. The ideal requirements of

such a full model submission are stated in the FMA site to be the following:

The names given in parentheses should be either the section headings in a single

submitted ﬁle, or the ﬁle names if each component is included in separate ﬁles.

1. A description of the theory of the model (THEORY)

This may be either a W

ord ﬁle, a pdf ﬁle, or a reference to a published article in a

freely accessible e-journal, such as FBMIS. Full mathematical details should be given.

2. A listing of the code for the model algorithm (MODEL)

This should be a txt ﬁle, or a W

ord rtf ﬁle with the code in Courier font. You may

include an executable form of the code if it may be run on a PC server platform.

However, if your model is very large, or has other system requirements, you may

possibly wish to reference your own web site from which the model may be

demonstrated.

3. The ﬁtting criterion used, if applicable (CRITERION)

This may not be submitted if the THEOR

Y section contains this information.

However, if the model ﬁtting process has been complex, possibly maximum

likelihood or a non-standard weighting procedure, then the contributor may wish to

put this in a separate section.

4. The ﬁtting algorithm used, if applicable (FITTING)

That is, the optimization algorithm by which the Criterion is maximized or

minimized. If a statistical package is used then this may not be known. However

, if

the ﬁtting has been achieved by use of known algorithms, such as from Numerical

Recipes, or the NAG library, then reference to the routine used would be useful.

Details of the starting parameter values used for ﬁtting the DATA to give the

RESULTS, and how they were determined, should be given.

5. Test data, with metadata (DATA)

This may be r

eal data, if the owner of the data has given permission for the data to be

made public. Full details of the nature of the data should be given: for example,

species, location, time span, environmental factors and conditions, the way the data

were collected. Graphical displays of the data would also enhance the accessibility and

comprehensibility of the data. Full metadata should be given including the variable

names and what they represent, the units used, and any approximations or rounding

used. If the data have been selected from a larger data set then the method of selection

should be documented.

If real data cannot be supplied, possibly for conﬁdentiality or ownership reasons,

then one of the following options may be followed:

(i) Anonymization of the data by withholding some of the metadata mentioned

above, e.g. location, speciﬁc dates, species.

37Amaro Forests - Chap 32 25/7/03 11:10 am Page 374

374 K. Rennolls

(ii) Fuzzing of the data, by the addition of random values to the actual data values.

(iii) Provision of simulated data. The method of simulation of the data should be

described. For example it would be nice to know if the data are simulated from

the model ﬁtted to to an earlier, raw, data set, since model mis-speciﬁcation

errors would not be expected in such a case.

6. The results of ﬁtting the model to the test data using the ﬁtting algorithm (RESULTS)

The ﬁtted parameters of the MODEL

on the DATA using FITTING on CRITERION,

together with any standard errors, etc., if available. A display of the ﬁtted model

superimposed on the DATA would be desirable.

Not all of these will be available for all models submitted to the FMA, particularly

‘legacy’ models, or ‘historical’ models. Submission is still welcomed if it is consider

they make a contribution to the overall forest modelling heritage.

The current structure of the FMA website is shown in Fig. 32.1.

It may be seen that there are a variety of user facilities, including an ability to

sear

ch for models, either by author, or by keyword. Some facilities are still under

development. The models currently populating the FMA are only provided as

examples. The ﬁrst listed in Fig. 32.1 is concerned with a model of the relationship

between the number of species and the sample plot area for a forest site in Sumatra,

a simple functional measure of species diversity. An accumulation of such data and

models in the FMA might assist a meta-analysis of such diversity relationships on a

global scale. The second example listed could include an indexed family of trans-

formed polymorphic Chapman–Richards growth functions in which the rate of

movement from one extreme curve to another varies with age (Fig. 32.2), a model of

some potential use in height–age studies. Other entries (Fig. 32.1) are an integrated

forest process model (IFPM), also a part of this author’s forest modelling legacy, and

the well-known and much-used

FVS model (information extracted from REM, and

the author’s own site).

Standards

The Models

History

News

Forest Mo...

Feedback

RFM REM

ECOBAS_...

Contribute...

Find a Model

Run a Model

Download

Species-Are...

Height Age...

IFPM

FVS

ECOBAS_M...

ECOBAS_MI...

FVS Software

User Registr...

User Serv...

Fig. 32.1. Structure of the prototype FMA.

37Amaro Forests - Chap 32 25/7/03 11:10 am Page 375

375 Conserving the Past and Building the Future

and comments, and hopefully contribute their models.

enabled by the existence of an FMA. The following is an incomplete list.

have been found to be suitable, and the limits of applicability of the models.

particular issues could easily be set up.

tant example.

These possibilities may be ambitious, but it seems that all of them would be

teams, often those that happen to be successful in competition for funding. Such

120

100

y-values

0 5 10 15 20 25 30

FMA Futures

First, all forest modellers are invited to examine the FMA site, to make suggestions

Beyond the original aims of the FMA to provide a means of conserving and

sharing forest models, we ought to consider some further possibilities that are

The FMA could become a test-bed of models. Benchmark data sets could be

included on which various models might be compared. Feedback from model users

potentially could be included in the FMA, indicating areas where particular models

The FMA could become a basis for collaboration between modellers working on

similar, related or even contrasting models. Theme-related discussion threads on

The FMA could become a means of presenting situations and scenarios where

various types of model development are needed. The need for more extensive

efforts to model tropical forests, and their impact on local communities, is an impor-

The FMA could be a starting point for international collaborative efforts to inte-

grate forest models of different forest components, possibly at differing scales.

very worthwhile. Such endeavours are normally only approached by fairly small

exclusive and focused team approaches might be best for the development of mod-

els with well-focused aims. However, for the rather wide-ranging aims listed above,

it might be that a more ‘open-club’ approach would be more appropriate, and more

effective, and that the FMA would provide an underpinning framework. Let’s try!

A New e-Journal: FBMIS

The case for a new free-access e-journal in the forest modelling sciences has been

made brieﬂy in the Introduction. The FMA as described above is in itself a new form

of publication and communication suited to the nature of the modelling product,

which makes use of the unique opportunities provided by the World Wide Web. A

Time

Fig. 32.2. Indexed family of Chapman–Richards growth curves with age-variable rates on convergence.