The development of complex models for plant growth becomes more important, as the needs of forestry and agriculture grow. Environment changes and the development of modern culture and management methods make it necessary to reconsider old ideas, schemes and models. Sustainable management and ecological forestry build the principles of modern forestry. However, the models used in the praxis till now are no more suitable to explain the changes occurring in our forests and fields. World-wide researchers develop new and accurate models for plant growth or for some other relevant physiological processes.
While in forestry stand oriented models have been principally used as support tools in the management process, individual-oriented models have been developed in biology, agriculture and also in forestry. The development of process and structure models set the basis for modern modelling techniques. With the emergence of modern technologies, high-capacity computers and more accurate measurement tools and methods, also new questions have emerged. The process of modelling itself has developed to a science of its own. With the development of structural-functional models new possibilities to increase the knowledge and understanding of the most basic processes responsible for growth have been provided.
Unfortunately, most of the modelling and research work in this field is still a domain of few specialists. The complexity of models is a cause for the prominent individualism characteristic for most models. In most cases only a few researchers have access to a given model. This way the same model is developed at the same time by different people, possibly without knowing of each other and surely without the possibility of learning mutually from errors and insights.
With a possibility of sharing knowledge and combining it, the development of more powerful models would be possible. The research process itself would be less ambivalent. There exist many models describing important processes, and others providing a framework for them by describing physical structures. It is not only important to create new and better models, but also to learn if and how they could work together. This thesis is an attempt to do that:
Combining and interconnecting process and structure models and even functional-structural models.
The programs and models presented in this thesis, particularly GROGRA, AMAP and LIGNUM are very complex programs that provide a good starting point to begin with the interconnection of procedural and structural models. GROGRA and AMAP provide the framework necessary to build other models, the Growth Engine. The growth engines are not models by themselves but a modelling tool.
The development of single models and their integration occurs often at different scales, both in time and in space. So there arises a question that is often posed. Is it more important to know what happens in one leaf during a few minutes or to know what happens in a forest stand during many years? There is no general answer to this question as it always depends on the point of view. But the integration of models can help to answer this question: both things are important. We intend to know and understand the details to be able to handle better in the whole. In this thesis a series of models were presented that are suitable for the interconnection. Most of these models have been compared often, but each was run in a specific framework, so that the results are not always compatible or usable. The interconnection of models can help us to compare them, because they will now work in a known common environment.
The information transfer is one of the principal problems that must be solved. In this thesis this problem was handled by using common files formats that should become standard for the integration. The idea of the standard format is based on two very generic file formats: MTG from the AMAP Programming Language AML and a structure description format from GROGRA, dtd.
The development of a generic data interface relies on the acceptance of the proposed file formats and on the utility of the interface. Both formats presented here have assets and drawbacks. The dtd format is more simple but also more limited in the amount of information it can contain. However, due to its simplicity, this format is perfectly suited for the information transfer with many models. Furthermore, GROGRA already provides a wide range of interfaces for other applications, which facilitates the integration process. The MTG format is more complex and it cannot be easily handled by most researchers and programs, but it allows to include more information referring to different spatial scales. Also the increasing importance of this format within the AMAP modelling team makes it essential for future development.
Beyond the acceptance of the formats it is important that the respective researchers are willing and/or able to undertake the modifications needed for the interfaces. This stage of the interconnection is the most critical phase. The interconnection of models developed at different sites and by different workgroups requires a great amount of co-ordination and logistics. The modules presented in this thesis demonstrate how difficult it can be. For the development of the Windows version of GROGRA, the original author, Prof. Dr. Kurth, was always present and ready to answer any questions about the program. Furthermore, he was always prepared to make any changes and expand the original software and to help solve any complications.
For the other two programs developed at the Institute for Forest Biometry and Applied Computer Science of the University of Göttingen the situation was more complicated. The program AIR, which is the predecessor of NEXUS, was developed by Dirk Lanwert. The software HYDRA was developed by Dr. Thomas Früh. Both were co-workers of the Institute that left by the time this project started. The possibilities for communication and feedback were, and still are very constricted. This has made the interconnection with HYDRA, which is being completed by Dr. Michael Schulte, more difficult than originally expected.
The last three modules that were incorporated in the project were MIR, MuSc and HYDRO, which were developed by Dr. Jean Dauzat at the CIRAD in Montpellier, France. He was willing to provide any information about the modules to facilitate the creation of the interface.
However, due to the pre-requisites of an object oriented approach, the software had to be
modified. For different reasons Dr. Dauzat could not make such modifications and he agreed to provide the code so that the programs could be modified. During a stage of three months at the CIRAD, Dr. Dauzat answered any question about the program, as long as it was possible.
The collaboration with other programmers of the work team AMAP made it possible to access vital information for understanding the software AMAP and its functionality. It was however often difficult to keep continuous contact between Göttingen and Montpellier so that some questions remained unanswered.
The functionality of the original models presented here has already been tested by the respective authors and other researchers. RAPIDEL (1995) contributed to the validation of MIR, MuSc and HYDRO. He compared the model output with other standard models by calculating the output generated by those models separately. He concludes that it is necessary to integrate the models with the growth simulation programs from AMAP. In this context, a modular model shell would have facilitated the model comparison.
The main objective of this thesis was not to re-validate the models taken, but to demonstrate that this kind of model can (with some effort) be integrated in this project. The parameterisation of the models for oak will be finished by Dr. Schulte in the framework of his project, so that the data from HYDRO can be compared and expanded with the data from HYDRA. However, the basic results obtained here confirm the results from Dauzat and Rapidel for Coffea arabica, and provide an expandable basis for the modelling of oak hydrology and growth.
The development of the NEXUS software is not yet finalised as it will continue beyond the scope of this thesis. The actual version provides some basic tools and the possibility to combine a few programs. A basic collection of sub-models has been created, but the models have not yet been tested within other major models. This includes different formulations for the multiplicative approach of JARVIS (1976) and two wind models.
The software also has some remaining bugs that should be removed later. Some errors are due to the numerics in the programs included. Other bugs were caused by the difference in windows management and event handling on different platforms.
An important part of the future development of the software is the compatibility with the latest version of AMAP, so that it can work with its Growth Engine. But this depends on the willingness of the managers of the AMAP software to allow insight in the code. This would permit to use NEXUS also with the NT version of AMAP. But this cannot be warranted by the author.
A complete software documentation for users and programmers will be posted on-line in small steps (see section 4). Most of the classes and procedures created will be posted as Open Source with exception of the procedures that handle the actual interface with AMAP. This is protected and Copyright of the CIRAD.