Computational intelligence
To power the whole MELA software ecosystem, the project will develop a so-called search and inference engine which will make use of computational intelligence in order to select the best combination of components for each specific use. On one hand, the computational intelligence will help in pursuing improved management of uncertainty, and on the other hand in making the computations less time consuming. This will make it possible to do cost-effective forest planning for a large number of increasingly small forest units with the remotely sensed information available.
The target will be reached by:
- Employing a self-learning simulator
- Developing a prototype of the search and inference engine for producing calculation chains, and
- Redefining an integrated simulation-optimisation logic.
Dataset interface
In developing the dataset interface, the aim is to achieve both more versatile and completely new types of analyses, which make use of the data more efficiently than before.
The target will be reached by:
- Developing the automatisation of the pre-processing of NFI survey data in order to speed up the calculations on that data. The automatisation of the pre-processing makes it easier to correct errors in the data and to complement it with computational models, as well as to combine different data sources (information on ownership, conservation status, habitats of special importance, etc.)
- Extending the dataset interface of the simulator to cover different datasets for stand characteristics. In particular, this will make possible the processing of grid-shaped forest resource information and its use in analysis requiring data on neighbouring trees.
Model function libraries
Defining the structure and standardisation of model function libraries aims at acquiring a larger variety of functions illustrating forest resources and operations. They can be used for example in planning new forest industry products, ecosystem services (e.g. carbon sinks, groundwater) or the production and use of non-wood forest products.
The target will be reached by:
- Defining a standard for model functions and carrying out the interfaces (model function developers, simulator) in R environment, taking into account compatibility with existing model libraries as far as possible.
- Co-developing the function libraries. The library offers the model developers the possibility of testing their model functions as a part of larger computational entities and applying new research results in practice more quickly.
Optimisation
The integrated simulation-optimisation component is developed to better consider the existence of both earlier forest planning solutions as well as risks and uncertainties in the design and solution of optimisation tasks.
The target will be reached by:
- Making use of the already existing, efficient IT solutions of the MELA and EFDM software and re-programming them into forming a part of the MELA software ecosystem.
- Defining test and example applications (various degrees of problems in forest planning) and testing the usability of the search and inference engine in solving them.
Interfaces with external systems and users
Interfaces are being modernised to make new types of use and reports possible. It is expected that the user interface development will attract new user groups for the MELA software ecosystem.
The target will be reached by:
- Producing a standardised interface to external systems such as the Finnish forest information standards, other grid-shaped datasets, other model libraries (MOTTI, SIMO) as well as digital Best Practice Guidelines for Sustainable Forest Management (Tapio) and LUKE’s own analysis results and sharing platforms (MELATuPa, VMIKaaVa, Metsämittari)
- Carefully documenting the upgrading process.