SILVICULTURAL STRATEGIES FOR PREDICTING DAMAGE TO FORESTS FROM WIND, FIRE AND SNOW: INTEGRATING TREE, SITE AND STAND PROPERTIES WITH GEOGRAPHICAL INFORMATION SYSTEMS AND REGIONAL ENVIRONMENTAL MODELS TO EVALUATE OPTIONS FOR FOREST MANAGEMENT


CONTRACT NUMBER - AIR3-CT94/2392

Participants:

P01 Faculty of Forestry, University of Joensuu, Joensuu, Finland

P02 Department of Forestry, The University of Aberdeen, Aberdeen, UK

P03 Centro Nacional de Informacao Geografica, Lisbon, Portugal

P04 Forestry Commission, Northern Research Station, UK

P05 Macaulay Land Use Research Institute, Aberdeen, UK

P06 Instituto Florestal, Ministerio da Agricultura, Lisbon, Portugal

P07 Faculty of Forestry, The Swedish University of Agricultural Sciences, Umeå, Sweden

P08 School of Engineering, University College, Galway, Ireland

Period of report: January 1995 - December 1996

SUMMARY

WIND, FIRE AND SNOW: INTEGRATING TREE, SITE AND STAND PROPERTIES WITH GEOGRAPHICAL INFORMATION SYSTEMS AND REGIONAL ENVIRONMENTAL MODELS TO EVALUATE OPTIONS FOR FOREST MANAGEMENT

Participants

P01 Faculty of Forestry, University of Joensuu, Joensuu, Finland

P02 Department of Forestry, The University of Aberdeen, Aberdeen, UK

P03 Centro Nacional de Informacao Geografica, Lisbon, Portugal

P04 The Forestry Authority, Northern Research Station, UK

P05 Macaulay Land Use Research Institute, Aberdeen, UK

P06 Instituto Florestal, Ministerio da Agricultura, Lisbon, Portugal

P07 Faculty of Forestry, The Swedish University of Agricultural Sciences, Umeå, Sweden

P08 School of Engineering, University College, Galway, Ireland

Scientific objectives

This project aims at producing generic models which use the factors common to wind, snow and fire damage; to underpin, those models with an understanding of the forces and site factors acting on single trees or the fuel hazard risks associated with site factors to produce risk assessments to forestry to minor or catastrophic damage; to test these models as a way to derive long term strategies of silvicultural methods for managing forests against wind, snow and fire damage, that optimize wood production while appraised of the risk of forest damage. This project utilizes methodology where models for estimating the breakage and overturning due to wind and snow load and models to estimate fuel are integrated with the data bases for properties of climate and vegetation controlling the snow, wind and fire damages to forests in order to optimize wood production considering the risks to forests.

The progress within the period of report

Task 1.

Sub-tasks 1.1 and 1.2. The planning report has been undergoing continual update throughout the period of report. However, the complete report is now available on the web and P05 ftp sites. Making the document "live" in this way has meant it is continuously available for update and perusal by all participants, without the need for temporally discrete stages of update and dissemination which leaves participants guessing as to the most recent updates. This is particularly crucial for such a large and geographically distributed group who are working in close collaboration. Metamodel and metadata forms, as detailed in Task 6 are also available on the web, and participants have been actively completing these forms in remote locations to provide instantly available documentation to the rest of the group. This has greatly enhanced the planning and management of the project, and facilitated a process of documentation for models and data which can be used as input to Task 6, and to increase inter-participant understanding of other participant's work which is vital when models developed by one participant are being tested on data supplied by another.

Task 2.

Sub-task 2.1. Two severely wind damaged Sitka spruce stands have been studied in 1995 and wood samples have been taken from a total of 72 of snapped, over-turned, and undamaged trees by P02. Laboratory analysis of these samples has now been completed and preliminary results have been obtained for ring width, density, and compression wood characteristics of wind damaged and undamaged trees. In 1996 two further badly wind/snow damaged sites, one Sitka spruce, the other Scots pine, have been sampled and some results are also available of the laboratory analysis of these samples. To date, no predictive correlations have yet been produced due to the generally non-significant differences between the data for the different damage types. The results from the work to date are provisional and further analysis is required before final conclusions can be drawn. Within this sub-task, P02 has also tested root samples provided by P01 to allow correlations between maximum turning moment and root strength and elasticity to be assessed.

Sub-task 2.2. Critical tree parameters controlling snow and wind damage for a sub-sample of sample plots from the existing Swedish National Forest Inventory (NFI) have been determined and a database from the same sub-samples made available for P02. Logistic models have been developed for Scots pine, which show that it is possible to predict future damage from snow and wind by using single tree characteristics as indicators of site risk. Dissemination of some research findings have also been carried out within the period of report (see Valinger and Fridman, 1996).

Sub-task 2.3. The extensive tree pulling database of total number of over 1800 trees has been constructed by Participant 04, and linked to this work Participant 01 has constructed Finnish tree pulling database of 115 trees of Scots pine, Norway spruce and birch spp on podzolic soils (see Peltola et al. 1996f), i.e. as was aimed within the period of report in technical annex. Regressions have also been calculated between various tree and site characteristics and critical turning moments needed to cause uprooting of single trees or stem breakage based on these tree pulling databases by P01 and P04 for determination of values for the critical parameters required to support the development of the models to be done in Sub­tasks 3.1 and 3.3 (see Granander 1996; Peltola et al. 1996f). P01 has also made wind and tree swaying measurements at the edge of a Scots pine stand and within the stand especially two tree heights from the edge prior and after two thinnings (see Peltola 1995, 1996a, Hassinen et al. 1996). To date, mean wind profiles and stem bending of trees at the stand edge and within the stand have been analysed in a Scots pine stand having stand density of 2700 and 1500 stems/ha, i.e. prior and after first thinning (see Peltola 1995, 1996a). The wind and tree swaying experiment carried out by P01 will provide further information on mean windspeed profiles and tree stem bending for varying stand densities, and give thus essential information for subsequent modelling work and models testing to be carried out in Sub­tasks 3.1 and 3.3.

Furthermore, P08 has developed a new video analysis technique to be used to estimate the overturning moments experienced by a tree during a storm event. P08 has also selected test sites with test trees (three site preparations on surface water gley) in Ireland and currently a field testing programme is ready for implementation. This field testing will consist of both monotonic and dynamic loading tests of trees on the various site preparations as well as a long term tensiometry study of the characteristics of the three site preparations. Dissemination of some research findings is currently under work in Sub-task 2.3 (see Granander 1996; Peltola 1995, 1996a, Peltola et al. 1996f, Hassinen et al. 1996).

Sub-task 2.4. The spectral and biometric data was collected in two field campaigns in 1995 and 1996 for pine stands and shrublands for supporting the regional classification of fuel types using satellite imagery. In addition, a controlled experiment to develop methodologies for unmixing narrow band-spectra of tree-soil-understorey composites was made. Objectives for this period of report have been fully achieved and the work developed is within the pre-defined time frames (no deliverables or milestones to be achieved).

Task 3.

Sub-task 3.1. P04 has developed and made available empirical model predicting the windspeeds required to break or overturn trees based on a knowledge of the tree species, tree height, tree stem diameter, inter­tree spacing and soil type, and using the fundamental information on wood properties and resistance to overturning obtained in Sub-tasks 2.1, 2.2 and 2.3 (see Gardiner and Peltola 1996b).

Sub-task 3.2. A database on snow and wind damage have been made available by P07 in Sub­task 3.2 (as was aimed in technical annex within the period of report). Within Sub-task 3.2, logistic models calculating the predictability of snow and wind damage to sites have also been developed using tree, stand, and site data characteristics from permanent sample plots in Västerbotten in northern Sweden as input variables (Fridman and Valinger 1996). The fitness of the developed models have been evaluated using i) temporay sample plot data from the county of Västerbotten, and ii) permanent sample plot data from other Swedish counties.

Sub-task 3.3. The updated mechanistic model of wind and snow damage of single trees (HWIND) developed under Sub­task 3.3 by P01 will be capable of calculating the mean windspeed to uproot or break a tree along with critical snow loading for various distances from the stand edge at the margins of clear-felled areas. To date the model covers especially Scots pine, Norway spruce and birch sp. growing on podzolic soils in Finnish conditions. However, other tree species and soil types (for various geographical locations) can be used by changing the controlling parameters and equations for species (and soils).

To date, HWIND­model has also been tested/improved based on Finnish tree pulling database constructed in Sub­task 2.3 for Scots pine, Norway spruce and birch spp. Whereas, validation of HWIND-model with measurements of windspeed and tree movement prior and after thinnings by P01 (from Sub-task 2.3) will be finished within couple of furthcoming months. On the other hand, sensitivity tests have already been conducted for HWIND-model by P01 to determine the critical parameters involved and the reliability that can be placed on the critical windspeed predictions.

HWIND­model predictions by P01 have also been compared with the predictions based on empirical wind damage model developed by P04 in Sub­task 3.1 to improve the methods for estimating critical windspeed (see Gardiner and Peltola 1996a, 1996b). In addition, a literature review has been finished on the factors affecting the snow damage of trees to support the mechanistic model development and its validation within Sub­task 3.3, in a co­operation by P01, P04 and P05 (see Nykänen et al. 1996). Dissimination of research findings within this Sub­task 3.3 is currently under work (Gardiner and Peltola 1996a, 1996b; Nykänen et al. 1996; Peltola 1995, 1996c, 1996d; Peltola et al. 1995, 1996a, 1996b, 1996c, 1996d; see also Metamodel form for HWIND, 1996).

Sub-task 3.4. Two indirect methods to estimate the LAI have been developed by P03 to ensure a precise characterisation of maritime pine stands LAI by ground measurements. During the summer period of 1996 an extensive data gathering campaign was undertaken in which 30 trees were studied by direct measurements in order to develop an allometric model to predict single trees Leaf Area. Two hand held instruments (Sunfleck Ceptometer and DEMON) designed to estimate the LAI from light measurements were tested on 25 field plots. This task had the collaboration of the P06, who provided logistic and field support.

Task 4.

Sub-task 4.1. Preliminary efforts have been made by P01 to calculate the wind loading on individual Scots pines as a function to distance from stand edge, stand density (in terms of leaf area index) and tree height and width of upwind gaps using airflow model developed by Miller at the University of Connecticut (see Peltola 1996b). In addition, because there has been some initial delay in getting the more sophisticated airflow modelling efforts available by P04, empirical relationships have been derived from existing wind tunnel studies to relate mean and extreme wind loading to distance from stand edge, stand density and tree height. These empirical relationships derived from existing wind tunnel studies by P04, and used in the HWIND-model (Sub-task 3.3) has allowed progress within this Sub-task in simulation of the wind loading and critical windspeed to uprooting and stem breakage of individual Scots pines, Norway spruces and birch spp as a function to distance from stand edge, stand density (stems/ha) and tree height (see Gardiner and Peltola 1996b; Peltola et al. 1996e), until the more improved numerical airflow models become available by P04.

Sub-task 4.2. This sub-task task involves an integrated approach to investigating the effects of variability by using existing data sets, making new measurements and developing mathematical models. Efforts have been made to identify suitable data sets for analysis, testing methodologies for measuring stand variability, investigating the importance of variability in stand growth model. Good progress has been made particularly with regard to the remote sensing aspects.

Sub­Tasks 4.3. Also within this Sub-task good progress has been made with regard to the remote sensing aspects.

Sub-task 4.4. Data collection took place during the summer 1996 and will continue in March and April 1997 in order to fine tune the Knowledge Based System.

Task 5.

Sub-task 5.1. A snow review paper has been prepared in collaboration with 3 participants (P01, P04 and P05) within Sub-task 3.3 (see Nykänen et al. 1996) which summarises also the current state of knowledge of snow damage to trees for Sub-task 5.1. This information provides a means of identifying the key site and climatological elements which require to be modelled. A strategy has been identified for developing a snow model for the UK and it is currently being implemented. Further work is also being developed to look at the joint occurrence of wind and snow, so that the correct interactions are included in the climatological modelling. Within the period of report P01 has collected also wind data for generating the surface of windspeed for Finnish test area. Also a landuse classification for calculating the regional wind pattern for the Finnish test area has been made by P01.

The statistical model (Jackson, 1977) has been implemented using GIS by P05, and layers of snow incidence statistics have been calculated for the UK. However, there are some difficulties with model validation in that the available Meteorological Office Data is unsuitable because of the way in which it has been measured. Modelling snow incidence is a persistent problem and a special meeting was held to discuss the issues at the last EC meeting in Umea. The issues experienced are similar for Sweden, Finland and the UK, and generally stem from inadequate meteorological information. Several additional approaches are currently being considered with respect to the combination of snow and wind since the interaction of these two damage agents is critical for assessing damage risk, particularly in the UK. Within this Sub-task, P07 has also made snow and wind atlases ready to be digitized during next six months period of report and to be used as GIS-layers for the continuous work within Task 6.

Sub-task 5.2. The compilation of geographic database for the study areas has been continued and spatial data has been converted to agreed format. For Finnish test areas, P01 has completed all the preprocessing of satellite images and field measurements of occurred wind damage. The data has been stored to the geographic database. Since the beginning of the project we are making the compilation of all the relevant analogic and digital available datasets, for the defined study area. Some layers are ready to use. The layers that are actually present in the geographic digital database are listed in this report. Spatial database for Portugal, will be ready in January 1997 according to the plan (deliverable and milestone by P03). The metadata forms, organised by the P05 are being used to collect and collate information about the datasets through the project World Wide Web (WWW) site. This is an efficient, simple and timely mode of gathering the information and storing it in a central location for viewing by all participants. The form ensures complete documentation of the information about the data in a standard format, and thus facilitates effective communication between participants, and provides documentation for present and future use.

Sub-task 5.3. Although this task comprises a live document which will be continuously updated, the groundwork for the majority of the information has been completed, and further updates will occur using input from the completed metadata forms on the web.

Sub-task 5.4. The data collected in Sub-tasks 2.4, 3.4 and 4.3, by P03 and P06 during the summer of 1995 and 1996 are being integrated in the geographic database and used for processing and analysis of the Landsat5-TM image acquired during this year. Using the field information and image processing techniques we will make a regional map of fuel types. The bulk of this task will be done during the next period report. There were no deliverables or milestones to be achieved within this sub-task during this period of report.

Sub-task 5.5. The work developed so far is conceptual and the Knowledge Based System conceived for this project (P03 and P05) is based on induction methods that seek to derive the rules underlying fuel distribution from the collected data set (point sample and maps) and then use the learned rules to map fuels for large areas, using the information of the geographic database. P03 and P05 are developing a Knowledge Based System based on the sample points collected this summer and on the products resulting from Sub-tasks 2.4, 3.4, 5.2 and 5.4. The methods are under evaluation and the implementation stage started. This task is slightly delayed due to the delay verified in the data collection (Sub-task 4.4). However, this fact does not compromise the proposed products and respective dates of delivery.

Task 6.

Sub-task 6.1. The demonstration which was set up in 1996 allowed several issues to be identified and explored by the project group. This demonstration involved the formal documentation of selected models through the web metadata forms, and trial exchanges of data and models between participants in different countries. The issues arising from this exchange, together with the success of the web, have permitted the formulation of a prototype Integration Framework during the October meeting in Umea. These plans will be implemented on the WWW over the next six month period so that a working prototype can be discussed at the next project meeting.

Sub-task 6.2. Scale and error studies explored using the models will be an important component in the prototype.

Task 7.

Sub-task 7.1. The change detection method was developed by P01 and tested using Landsat TM images. The key results was that drastic damages could be classified correctly, but partial damages were classified half correctly and half wrong. The testing of the mechanistic wind damage model has been started at the end of 1996 and some preliminary computations are available by P01.

Task 8. The World Wide Web (WWW) pages are continuing as a valuable communication medium to communicate the aims, methods and findings of the project. The web site is also being used as a means of formally documenting the models and data for use in Task 6. The project publicity leaflets have been printed and were distributed to all participants at the last project meeting and also sent to DGXII. A poster for the project was presented at GIS Research UK Conference (STORMS, 1996) and was well received. Dissemination of many research findings is also currently under work in many Sub-tasks.


CHAPTER 1 INTRODUCTION

Scientific objectives of the project

This project aims at producing generic models which use the factors common to wind, snow and fire damage; to underpin, those models with an understanding of the forces and site factors acting on single trees or the fuel hazard risks associated with site factors to produce risk assessments to forestry to minor or catastrophic damage; to test these models as a way to derive long term strategies of silvicultural methods for managing forests against wind, snow and fire damage, that optimize wood production while appraised of the risk of forest damage. This project utilizes methodology where models for estimating the breakage and overturning due to wind and snow load and models to estimate fuel are integrated with the data bases for properties of climate and vegetation controlling the snow, wind and fire damages to forests in order to optimize wood production considering the risks to forests.

Scientific objectives during the period under report

Task 1. Whilst most of the work associated with the Task 1 planning stage was completed in the first six months of the project, this task is ongoing. Further development of the planning report will be pursued using Task 6 as a focus. As individual tasks evolve and the development of the work becomes clearer, Task 1 will be updated to reflect the increasing detail of plans and the solution of issues. Scale issues cannot be addressed until other parts of the project are better developed. The links identified in Task 6 will determine the details of the scale issues.

Task 2. The objective of this task is to determine the main factors which control snow, wind and fire damage for determination of values for the critical parameters required to support the development and the validation of the models to be done in subsequent Tasks.

Task 3. This task aims at developing empirical models of breakage and overturning of a single tree by wind (in Sub­task 3.1), logistic models of breakage and overturning of a single tree by snow and wind (in sub­task 3.2), mechanistic wind and snow damage model (HWIND) using single tree data (in sub-task 3.3), and models for prediction of Leaf Area Index for (Pinus pinaster) (in Sub­task 3.4).

Task 4. The aim of Task 4 is to determine how the relationships and models developed in Task 3 can be applied at the stand level. There are two main approaches being adopted to help answer this question. Firstly, measurement techniques are being developed to enable remote sensing of within stand variability (part of Sub­Task 4.2) and the variation in under­storey fuel capacity (Sub­Task 4.3). Secondly, modelling techniques are being used to calculate the impact of natural variability on the within stand risk (part of Sub­Task 4.2), to predict the variation in wind loading within a stand (Sub­Task 4.1) and to determine the best method for estimating fuel hazard (Sub-task 4.4). It is envisaged that a combination of data layers from satellite and aircraft images, together with model predictions, will be used to determine the within forest variability in the final risk model.

Task 5. Within Sub-task 5.1 has been worked on the development and applicability of a method for calculating the effect to topography and surface roughness on windspeed. The objective of Sub-task 5.2 is the compilation of a geographic database, aquisition and image pre-processing. The aim is the construction of a database for U.K, Finland, and Portugal for modelling and test the risks models. The work that is outcome in this Sub-task is the progress achived during the buiding up of geographic database from the site scale to regional level to more than 4 study areas (country scale). Furthermore, metadata forms to be used in the compilation of this data will be developed. Sub-task 5.3 aims at identifying and listing the data available which may be relevant to the modelling of abiotic forest damage, specifically by means of wind, snow and fire within the geographic range for which damage may be modelled. Constructing the models, deriving suitable data and assessing the validity of such modelling at different scales or resolutions is the key outcome of this project. Sub-task 5.4 aims at regional mapping of fuel types using mixture models and functional/structural vegetation classification. The aim of Sub-task 5.5 is to produce a fuel hazard map and to develop a knowledge based system (NBS) for fuel hazard mapping using stand and regional level information. Sub-task 5.6 aims at constructing a model to estimate spatially distributed fire risk.

Task 6. To initiate the integration of component models and datasets from the project, it was agreed that a demonstration should be undertaken, to be completed as a first stage. The purposes of the demonstration were to test inter-participant links and the method of information elicitation proposed as part of the integration framework. The demonstration served several further purposes by addressing the conceptual and practical issues of applying models to different data in other countries at varying scales, and of developing understanding between participants as to the data and models available in the different countries. Furthermore the results provided a means of discussing, as a group, the performance of the models and the quantification of risk assessment.

Task 7. The aim of this task was to test the models developed against independent data. Sub-task 7.1 has already been started by developing and testing the change detection method based on Landsat TM images. Antoher aim is to test the mechanistic wind damage model against field inspected forest stand data in the Finnish test area.

Task 8. Information continues to be widely documented and disseminated at regular intervals via electronic and paper media, and through inter institutional visits.

More detailed discussion of the work and achievements is given below Task by Task. Sub-tasks active during the period under report were marked with **.

Being done

in the year

Task 1: Project Planning

Sub-task 1.1: Planning Design ** 1

Sub-task 1.2: Scale ** 1

Task 2: Quantification of Component Factors Controlling Snow, Wind and Fire Damage

Sub-task 2.1: Wood Property Factors Controlling Wind Damage ** 1,2,3

Sub-task 2.2: Wood Property Factors Controlling Snow Damage ** 2,3

Sub-task 2.3: Tree and Soil Factors Controlling Wind Damage ** 1,2,3

Sub-task 2.4: Vegetation Characteristics Controlling Fire Damage ** 1,2

Task 3: Tree-level Models to Predict Circumstances for Damage

Sub-task 3.1: Empirical models of breakage and overturning of single tree by wind ** 1,2

Sub-task 3.2: Empirical models of breakage of single tree by snow ** 1,2

Sub-task 3.3: Mechanistic model for wind and snow damage of single tree ** 1,2

Sub-task 3.4: Leaf area index (LAI) estimation ** 1,2,3

Task 4: Defining stand-level Variability to Permit Application of Tree-level Models

Sub-task 4.1: Wind Flow across and within Stands Using Airflow Models ** 1,2

Sub-task 4.2: Variation in Measured Tree Characteristics within Stands

Using Various Methods Including Remote Sensing ** 1,2

Sub-task 4.3: Plot-level Mixture Modelling ** 1,2,3

Sub-task 4.4: Preparation of Knowledge Base for Fuel Hazard Mapping ** 1,2,3

Task 5: Regional-level Snow, Wind and Fire Risk Models

Sub-task 5.1: Climatology of Wind and Snow in Relation to Topographic Variability

and Temporal Incidence ** 1,2

Sub-task 5.2: Construction of Geographical Database and Use of Image Processing ** 1,2

Sub-task 5.3: Explore Relevance of Pan-European Datasets of Eurostat and CORINE ** 1

Sub-task 5.4: Regional Classification of Fuel Types Using Mixture Models and

Functional Vegetation Classification ** 1,2,3

Sub-task 5.5: Knowledge Base System Construction for Fuel Hazard Mapping ** 2,3

Sub-task 5.6: Fire Risk Model Construction 3

Task 6: Integrating Components from Tree/Stand Regional-level to Produce an Unified Risk Model

Sub-task 6.1: Integrating the Component Models ** 1,2,3

Sub-task 6.2: Scale Issues and Error Propagation 2,3

Task 7: Testing Models against Independent Data and Outlining Implications for Silvicultural Strategies

Sub-task: 7.1: Model Testing by Application of Change Detection System to Finnish Test Area.

Detecting Wind Damages on Stand Level by Landsat TM Images ** 2,3

Sub-task: 7.2: Model Testing Through Portuguese Forest Service 3

Sub-task: 7.3: Model Testing through Forestry Commission Windthrow Monitoring Areas

and Forest Districts 2,3

Task 8. Final Products, Documentation and Identification of New Opportunities

Sub-task 8.1: Dissemination of Project Information and Results ** 3

Sub-task 8.2: Documentation of Project Achievements ** 3


CHAPTER 2 MATERIALS AND METHODS

Task 1: Project planning

All participants have a copy of the Task 1 documentation, and this document is undergoing continual update as issues are identified and resolved with Task 6 acting as a focus. Previously the planning report has been disseminated as a paper document, also available on the ftp site. However, within the period of report, the document has been translated into hypertext macro language (html) and is accessible through the world wide web. Updates are made at the request of participants by MLURI and can be achieved within a matter of minutes using this system.

Task 2: Quantification of component factors controlling snow, wind and fire damage

Sub-task 2.1 Wood property factors controlling wind damage

Two severely wind damaged Sitka spruce stands growing in the North of England have been sampled in 1995 and wood samples of twelve trees were sampled at each site from each damage category (snapped, overturned and standing trees). Samples were, thus, taken alltogether from a total of 72 trees and these sampled trees formed triplets of similar diameter trees (snapped, over-turned, undamaged). Various tree characteristics were studied of sample trees (as height, dbh, crown depth and width) and sample discs were removed for assessment of stem shape, estimation of stresses within the living tree and wood characteristics. Laboratory analysis of these disc samples has been carried out for ring width, earlywood/latewood proportion, basic density, and compression wood characteristics of wind-damaged and undamaged trees. A subsample of disc samples were also used for investigation of modulus of rupture (MOR) and modulus of elasticity (MOE) of small clear specimens containing no serious strength reducing defects (appr. 30 specimens from each damage type).

In addition, in 1996 two badly wind/snow damaged sites, one Sitka spruce, the other Scots pine, were investigated, and all trees within the stand were measured for dbh, and categorised according to damage type. A total of 20 Scots pine and 24 Sitka were sampled with various crown, wood and tree stem characteristics. However, since no overturned trees were present, pairs of trees were sampled, with each pair having the same diameter. Influence of height/diameter and slenderness ration on damage type has also been studied by P02 in a separate experiment of badly wind damaged Sitka spruce stand, where more than 100 trees was measured for dbh and height, and damage status. The average height, diameter and slenderness ratio of each type of tree was calculated to reveal whether these parameters had an influence with the type of damage which occurred. Within this Sub-task, also root strength have been tested by P02 using root samples provided by P01 from Sub-task 2.3 to allow correlations between maximum turning moment and root strength and elasticity to be assessed.

Sub-task 2.2 Wood property factors controlling snow damage

Critical tree parameters for Norway spruce, Scots pine and birch controlling snow and wind damage for a sub-sample of sample plots from the existing Swedish National Forest Inventory (NFI) have been determined and a data base from the same sub-samples constructed to Participant 02. Logistic models have also been developed for Scots pine, which show that it is possible to predict future damage from snow and wind by using single tree characteristics as indicators of site risk.

Sub-task 2.3 Tree and soil factors controlling wind damage

Tree pullings. The extensive tree pulling database of total number of over 1800 trees has been constructed by Participant 04. Linked to this work Participant 01 has pulled over by a winch 115 trees in Finland on podzolic soils, and constructed Finnish tree pulling database of 51 Scots pines, 33 Norway spruces and 11 Birch sp. during unfrozen soil and of 20 Scots pines also during frozen soil conditions (Granander 1996; Peltola et al. 1996f). In Finnish tree pullings, a winch system was used to pull trees over and turning moments needed to uproot a tree or break the stem were measured at the base of the stem, and various kind of stem, crown and rooting characteristics of trees recorded along with stand and site information as described in details by Peltola and Kellomäki (1995). Most of the trees during unfrozen soil conditions were pulled over in autumn 1995, however tree pullings were continued for Norway spruce also in autumn 1996 in order to add more spruces to the database. Furthermore, additional tree pullings of Scots pines, although not mentioned in technical annex, have been conducted during frozen soil conditions in winter 1995/1996 in one of those sites also used during unfrozen soil conditions to find out if any difference exists in critical turning moment needed to cause stem breakage during frozen and unfrozen soil conditions.

To date, regressions have also been calculated between various tree and site characteristics and critical turning moments needed to cause uprooting of single trees or stem breakage based on these tree pulling databases by P01 and P04 for determination of critical parameters required to support the development and the validation of the models to be done in Sub­tasks 3.1 and 3.3. The integrity of data of extensive tree pulling database is also currently being checked and quality assurance carried out on all the data by P04. Furthermore, additional data have been made available by P04 and other scientists (from Canada and New Zealand) to be added to the existing database.

Wind and tree swaying measurements. P01 has made wind and tree swaying measurements at the edge of a Scots pine stand and within the stand especially two tree heights from the edge prior and after two thinnings (see Peltola 1995, 1996a, Hassinen et al. 1996). These measurements will further support the development and the validation of the models to be done in Sub­tasks 3.1 and 3.3. Windspeed measurements by cup anemometers and tree swaying measurements by both accelerometers and video measurements have already been analyzed in a study stand for stand densities of 2700 stems/ha and 1500 stems/ha (1991­1993), i.e. first without thinning and then two years later after first thinning (see Peltola 1995, 1996a). The wind and tree swaying measurements have also been continued in the same stand after second thinning (about 1100 stems/ha) and windspeed has been measured both by propeller anemometers and cup anemometers, whereas stem displacement measurements have been obtained using accelerometers site by site with a new prism based technology. The latter one has been used instead of video based technique (unlike planned in technical annex), because it has been proved to be much easier to use and very accurate (tested in laboratory conditions, see Hassinen et al. 1996). Wind direction has been measured using directional vanes as done earlier. Some measurements based on displacement transducers have also been obtained site by site with the prism based technique and accelerometers at the stand edge for the comparison of various methods available. In addition, some measurements of tree swaying has also been done within the stand approximately four tree heights from the edge after second thinning. This tree swaying and wind profile measuring experiment will be continued also in 1997.

Monotonic and dynamic tests. Furthermore, P08 has developed a new video analysis technique to be used to estimate the overturning moments experienced by a tree during a storm event. P08 has also selected test sites with test trees (three site preparations on surface water gley) in Ireland and currently a field testing programme is ready for implementation. This field testing will consist of both monotonic and dynamic loading tests of trees on the various site preparations as well as a long term tensiometry study of the characteristics of the three site preparations. The trees will be pulled monotonically using a load cell and winch system to enable the turning moments involved to be obtained. The displaced shapes of the trees will be recorded on video from a fixed point for calibration purposes. Video images will again be taken of the trees during storms, and by superimposing the storm images on the original images, the turning moments to which the trees are subjected during storm conditions will be estimated.

Sub-task 2.4 Vegetation characteristics controlling fire damage

The data gathering was organised in two field campaigns: 1995 (August) and 1996 (from July to September). Field protocols were developed for collecting the following data: (1) spectral data of pine and understorey mixed targets (controlled experiment); (2) spectral data of landscape pure components, (3) biometric data of different types of shrubs, (4) structural characteristics of large patches of shrublands and (5) structural characteristics of pine stands, including the understorey. Spectral and biometric data were collected for pine stands and shrublands for supporting the regional classification of fuel types using satellite imagery. In addition, a controlled experiment to develop methodologies for unmixing narrow band-spectra of tree-soil-understorey composites was made.

Task 3: Tree - level models to predict circumstances for damage

Sub-task 3.1. Computer based mathematical models are developed by P04, these models being capable of calculating the windspeeds required to break or overturn trees based on a knowledge of the tree species, tree height, tree stem diameter, inter­tree spacing and soil type. The models use the fundamental information on wood properties and resistance to overturning being obtained in Sub-tasks 2.1, 2.2 and 2.3. Initially model development is carried out using Mathcad 6.0 for ease of construction and is then converted into a Turbo Pascal programme for integration with the GIS and for use by other participants. The model takes the tree characteristics and calculates the aerodynamic roughness and zero plane displacement of the forest. From this information the drag of the forest can be calculated which represents a force due to the wind on each tree. This force can then be converted into a bending moment which the tree must be able to resist to avoid breaking or overturning. The windspeed at which the applied bending moment is greater than the resistive moment offered by the soil and roots is the critical windspeed for overturning. The windspeed at which this bending moment induces stresses in the stem greater than the Modulus of Rupture of the wood is the critical windspeed for breakage (see e.g. Gardiner and Peltola 1996b).

Sub-task 3.2. To apply logistic statistical analysis technique permanent sample plots (25 000) within the existing Swedish National Forest Inventory (NFI) database have been used. The NFI database used for the development of logistic models developed under sub­task 2.2 have been continuously utilized by the inclusion of site and stand characteristics for different counties of Sweden. Permanent sample plots are used so that the progression of damage can be monitored. Data will be separated according to species (Pinus sylvestris, Picea Abies, and Betula spp.), tree, stand, and site characteristics. Data inputs used from the Swedish National Forest Inventory (NFI) are tree species, various tree, stand, and site characteristics (i.e. height, diameters at stump height, breast height, and at 5m, stand density, standing volume, latitude, height above sea level, etc.). The database on snow and wind damage including tree, site and stand characteristics from the county of Västerbotten is available at our ftp­site (see Sub­task 2.2).

Sub-task 3.3. The mechanistic model of wind and snow damage of single trees (HWIND), capable of calculating the windspeed to uproot or break a tree, attempts to fully describe the mechanistic behaviour of trees under wind and snow loading (Peltola et al. 1996c). The critical wind and snow loading can be predicted both at stand edge adjacent to clear cuts and within the stand conditions for various distances from the stand edge. To date the model covers especially Scots pine, Norway spruce and birch spp. growing on podzolic soils in Finnish conditions. However, other tree species and soil types (for various geographical locations) can be used by changing the controlling parameters and equations for species (and soils). HWIND­model calculates the mean wind load (static loading) at stand edge for each height in the canopy using a predicted mean wind profile at stand edge. Whereas calculation of extreme loading due to gusts, which is capable to uproot a tree or break a stem, will be based on existing relationships between extreme and mean wind loading to trees (as a function of tree height, stand density and distance from the stand edge) provided by P04 based on wind tunnel studies. In addition, within the stand critical windspeeds to damage trees are also based on the relationship developed by P04 between mean loading at the stand edge and within the stand conditions. To calculate the resistance to overturning, HWIND­model uses a prediction of root plate mass to derive a resistive moment. Whereas approach used to predicting when a tree will break relies on the assumption that the stress in the outer fibres of the tree stem are constant at all heights and values of Modulus of Rupture determined for different timbers. The snow damage risk of trees is modelled in terms of maximum snowload as the product of the projected area of the crown and a snow load, corresponding to snowfall accumulated in the crown.

To date, HWIND­model has been tested/improved based on Finnish tree pulling database constructed in Sub­task 2.3 for Scots pine, Norway spruce and Birch sp. In addition, validation of the HWIND-model using available measurements of windspeed and tree movement within Sub-task 2.3 have been started. Also sensitivity tests have been conducted for HWIND-model by P01 to determine the critical parameters involved and the reliability that can be placed on the critical windspeed predictions.

HWIND­model predictions have also been compared with the predictions based on empirical wind damage model developed by P04 in Sub­task 3.1 to improve the methods for estimating critical windspeed and snow loading (see Gardiner and Peltola 1996a, 1996b). These models compared have been developed completely independently and designed to take site and stand information and predict the canopy top windspeeds at which the trees will break and overturn. They have also adopted different approaches to both the method for calculating the mean wind loading on the trees and the resistance to overturning provided by the trees. However, both models use a very similar approach to predicting when a tree will break. To date the models have been compared for Scots pine and Norway spruce growing on podzolic soils which are two of the few situations common to Great Britain and Finland. Tests have been carried out for a variety of tree heights, stem diameters and inter­tree spacing.

In addition, within this Sub-task a literature review has been under work on the factors affecting the snow damage of trees to support the mechanistic model development and its validation within Sub­task 3.3, in a co­operation by P01, P04 and P05 (see Nykänen et al. 1996).

Sub­task 3.4. For the development of an allometric model to predict single trees leaf area, two sets of 15 trees were measured by direct methods in two different forest areas within the study area (all needles collected, weighted and then converted to leaf area units by a previously determined specific leaf area value). A sampling procedure for leaf area determination was also tested in 1996 early spring, however its estimates were not sufficiently precise, and it had to be abandoned. Thus, the non-linear least squares estimation was used with the algorithms of the Genstat Statistical Software (Genstat5 Committee, 1987) to fit the allometric models. Model evaluation and selection was performed, using goodness-of-fit statistics, predictive ability statistics, and diagnostic statistics of multicollinearity.

In addition, a method for LAI estimation based on measurements of the canopy light transmittance was developed for pine stands. This method is easier to implement than the extensive use of allometric models since, when properly calibrated it only requires light measurements taken along transepts. The calibration of this method was done in field plots where the LAI was determined with the allometric model and the fotossinteticaly active radiation (PAR) was measured, above and below the canopy, with a Sunfleck Ceptometer© and with a DEMON©. It were measured a total of 25 field plots (20x20 meters) covering a wide range of stand ages and cover levels. The performance of these two devices was evaluated, and empirical and deterministic models to predict LAI from light measurements were tested.

Task 4: Defining stand-level variability to permit application of tree-level models

Sub­Task 4.1. The effect of position within the stand on the wind loading experienced by a tree is being explored within this Sub­task using especially HWIND-model developed by P01 within Sub-task 3.3 and models of airflow across forest edges being developed by P04. The results of these models will be compared against measurements of windspeeds and wind loading back from forest edges from field and wind tunnel experiments.

Currently, model simulations have been made by P01 of within stand wind loading in respect to distance from stand edge, stand density (in terms of leaf area index) and tree height and width of upwind gaps for Scots pine, based on airflow model developed by Miller at the University of Connecticut (see Peltola 1996b). Simulation of wind loading against tree position in a stand was aimed to be continued already within the period of report using more sophisticated air-flow modelling efforts by P04 making also possible to study the effect of stand density in terms of tree spacing to wind loading to trees instead of leaf area index. However, because improved air-flow model by P04 is not yet available, P01 has simulated wind loading against tree position in a stand using HWIND-model with mathematical relationships derived from existing wind tunnel studies by P04 to relate mean and extreme wind loading to distance from stand edge, stand density and tree height. This allows progress within this sub-task, before improved air-flow modelling efforts are available by P04. Currently, these improved air-flow models are being developed by P04 as part of an Open University PhD thesis and in collaboration with Leeds University as another PhD project.

Sub-task 4.2. This Sub­Task involves the use of existing data, measurement of stand variability and modelling of the effect of variability. A great deal of data exist from the wind damage monitoring areas within the United Kingdom and the extensive sample plots maintained in Sweden. Statistical analysis is being used by P07 to define the variation in stand vulnerability from the detailed set of sample plots in Sweden to find out if it is possible to identify the risk for a particular stand from the physical characteristics of a specified tree, for example the oldest tree within the stand.

Within this Sub­task remote sensing techniques are being tested by P04 and P05 to determine their potential for rapidly mapping within forest variability. An area in Cwm Berwyn forest in Wales, which is to be used in the validation tests in Task 7, has been identified as a test area. Computer comparison of derived elevations from photographs taken prior to tree planting and a few years ago have been compared in order to extract tree heights. The derived measurements are being compared to existing ground survey measurements and specific transects made using a Criterion laser relascope. The next stage will be to identify whether the derived tree characteristics are related to more easily measurable parameters such as soils, distance from stand edge, slope angle, exposure and aspect. Within this Sub-task, also stand growth models using both tree and stand characteristics developed by Soederberg and Ekoe have been made available by P07 and they can be used to determine how much variation in tree characteristics exists within stands and how this changes with species and climatic conditions.

Sub-task 4.3. The spectral characteristics of pure pine canopies and different types of understorey are obtained using a spectroradiometer in the field as part of Sub­Task 2.4. This information is then used to simulate several mixed targets with different backgrounds using the SAIL model (Scattering from Arbitrarily Inclined Leaves) to compute the canopy reflectance. Models are then developed to determine whether it is possible to derive the understorey from the spectral signal received by satellites. Initially these models have used traditional techniques based on broad band and narrow band vegetation. Recently new methodologies based on spectral unmixing, red edge detection and spectral derivatives have been under development. Development of the methodologies is the responsibility of P03 and P06 has provided assistance in the collection of the field data in Sub­Task 2.4.

Sub-task 4.4. P03, P05 and P06 are developing a knowledge based method for determining the fuel conditions within a stand based on site conditions. For obtaining information on fuel loading over a wide area it is clear that satellite remote sensing provides the best opportunity, particularly because there is a large temporal variability during the year. The method is to test the applicability of NDVI sensing from the Landsat and AVRIS satellites for determining the canopy structure of forests. Initial trials have simulated the NDVI spectral signal of canopies with different Leaf Area Indices and sub­canopies in order to determine the feasibility of the technique. The next stage will be to measure the reflectance of actual canopies and understories with a ground based spectrometer and to determine how well the measurements are related to measured canopy and sub­canopy leaf areas. The fuel sample points were selected by random sampling over aerial photography (1:25000) using 5 points per photograph. A minimum and maximum distance was kept between sample points to avoid spatial autocorrelation problems and duplication of selection likelihood in the areas of photo overlap. Each sample point position was recorded using a GPS, and a photograph was taken according to a pre­defined scheme and then used for deriving the fuel loading.

Task 5: Regional-level snow, wind and fire risk models

Sub-task 5.1. Within this Sub-task, long-term windspeed data (1971-1990) has been provided by P01 for generating the surface of windspeed for Finnish test area (from the Finnish Meteorological Institute, FMI). In addition, the satellite image (preprocessed and classified in Sub-task 5.2) has also been provided for calculating the regional wind pattern for the Finnish test area by P01. Within this Sub-task, P04 is also comparing several methods that treat topographic effects on windspeed, and these are further compared to the existing field measurements to assess which method is suitable for recommendation for wider use in the project.

Furthermore, P05 has been encoding Jackson's statistical models for the occurrence of falling snow in the UK (using AML in the ARC/INFO GIS). The data required are a digital elevation model for the UK, statistical relationships, and a contour map of snow days for the UK at 100m elevation produced by Jackson. Meteorological Office snow reports are being used to validate this model where possible, and further refinements will be made using other site data, for example TOPEX. It is also proposed that some contacts should be made with meteorologists to explore the possibilities of studying the combined probabilities of wind and snow in the UK. In addition, digitised snow and wind atlases for Sweden will be available by P07 during spring of 1997, and will be used as GIS-layers for the continuous work within Task 6.

Sub-task 5.2. P01 has continued the compilation of information for the geographic database, and stored it. The test area database has been completed with a digital elevation model covering the whole test area. The field measurements of occurred wind damage in the test areas has been completed with new measurements during February 1996. Wind damage occurred during 1990-1995 has been measured in total of 37 stands in the test area.

Within this Sub-task, P03 has continued the compilation of all the relevant analogical and digital datasets available for P03 study area. The software used is mainly INTERGRAPH and IDRISI. The coordinate system adopted is the military one, based on the Hayford ellipsoid (International) with a Gauss-Kruger projection. Within this Sub-task, P04 will be assemble the data within a GIS, representing the necessary locational, site, and tree data required to run the damage probability models. These datasets will then be used for validation of generic models. Respectively, P05 has been converted the metadata forms to html (hypertext macro language) and the pages connected to the project web pages. Forms can be filled in by accessing the pages using a web browser and the contents are automatically captured using a PERL program and written to a file, which can then be viewed on the web. These forms are, however, protected and secure, such that only participants can access this information. Due to the success of the metadata forms, the documentation technique has been extended to the models by the production of metamodel forms (see Task 6 for details).

Sub-task 5.3. Pan-European datasets report has been completed using input from FIRS project, other EC documentation (CORINE, EUROSTAT) and information from project participants. It will be updated using information collected and collated via the web site. The report is set out in the same style as the metadata forms and shows clearly how little crucial metadata (particularly error and accuracy estimates) are associated with any of the datasets detailed.

Sub-task 5.4. P03 will use the data collected in the Sub-task 2.4 during summer 1996, as reference end-members for the unmixing of the Landsat5-TM image acquired. In summer 1996, P03 gathered training areas (using a GPS), which will be used as potential image end-members during the unmixing process. The data collected in Sub-tasks 3.4 and 4.3, i.e. the Leaf Area Index (LAI) of pine forest, will work as ground-true data for satellite imagery calibration, and further to generalise LAI estimations to a regional level.

Sub-task 5.5. The Kowledge Based System conceived for this project (P03 and P05) is based on induction methods that seek to derive the rules underlying fuel distribution from the collected data set (point sample and maps) and then use the learned rules to map fuels for large areas. The approaches to be used are based on methods of pattern reccognition and induction of rules based on examples. The methods are applied in a geographic information system. The Knowledge Based System is also developed based on the sample points collected in summer 1996 and on the products resulting from Sub-tasks 2.4, 3.4, 5.2 and 5.4 within a couple of forthcoming months.

Task 6: Integrating components from tree/stand regional-level to produce an unified risk model

Sub-task 6.1. A formal documentation process was implemented by P05 to provide information and guidelines for valid model application and data use. Information was elicited from participants via metadata/model forms which can be accessed and completed interactively through the World Wide Web Site. Both forms have been carefully prepared so that they are comprehensive but straighforward to complete. These forms emphasise important information for the integration such as scale and resolution, model scope and limitations, and copyright details. Information from these forms is being used to assess the compatibility of scales between models and data, and between data for different countries. They also provides information to explore the tolerance of models to different types of data input, the suitability of data from different countries as input to models, the scope of the models, and the validity of model output results. In the short-term, the metadata/model forms have provided an effective means of exchanging information between participants for the purpose of the project demonstration. In the long-term they will facilitate complete documentation of the project.

A subset of models and data has also been selected for the short time period demonstration and a matrix of model runners and data providers derived to spread the work as evenly as possible so that no one group was over-burdened. The links addressed the three damage types equally, involved most of the participants, and capitalised on existing data and models which were sufficiently developed and immediately available. Discussion of these preliminary demonstrator results was extremely profitable for the whole group and increased understanding of the models and allowed the issues to be identified and explored. These results formed a basis for planning a prototype, web-based integration framework which was formulated prior to the project meeting in Sweden, 1996. Plans for the prototype web interface were finalised at the last project meeting in Sweden 1996 and will be implemented over the next six months. The prototype will serve to demonstrate how the project components can be integrated and presented, and will facilitate further, more detailed discussions during the next project meeting in Aberdeen 1997. Contributions from various participants have also been timetabled over the next few months. These contributions comprise maps, the results of model runs, and silvicultural information, which are being sent to P05, who will develope the prototype using web interface tools.

Sub-task 6.2. An important component of the web interface is the presentation of error and scale issues and their implications to potential model users. P05 is implementing several approaches to investigating error and accuracy issues together with other project participants. The means by which this information should be imparted to those accessing the integrated framework will be discussed in some detail in next project meeting in Aberdeen 1997. This work will use the information from the completed metamodel forms (model sensitivities) and metadata forms (scale and error).

Task 7: Testing models against independent data and outlining implications for silvicultural strategies

Sub-task 7.1. The change detection method was developed and tested by P01 using multitemporal Landsat TM images, forest stand data and field inventory data for occured wind damage. Also the testing of the predictions by HWIND-model developed within Sub-task 3.3 by P01 has been started against dertected wind damage in Finnish test area using the geographic database constructed in Sub-tas 5.2.

Task 8. Final products, documentation and identification of new opportunities

The WWW site is serving as a valuable means of information collation and dissemination both between project participants and to the wider scientific community. The WWW site currently facilitates participant inputs via web forms, which have been carefully devised to elicit clear and concise information about the data and models being used within the project. These are proving successful, both in terms of collaborative communication, and as input to the integration of component data and models in Task 6. Dissemination of scientific results is being effectively achieved via journals and conference papers, presentations made to institute visitors and by distribution of the project leaflets.


CHAPTER 3 RESULTS

Task 1: Project planning

The planning documents have been invaluable reference sources for data exchange, and the numerous visits by participants to the web site is evidence of this. Participants are referring to the web documents to ascertain the appropriate formats for exchanging information with other groups which has saved much long-winded communication, and therefore time.

Task 2: Quantification of component factors controlling snow, wind and fire damage

Sub-task 2.1. P02 has finished data collection and analyzed the effect of individual tree height, diameter and slenderness ratio, crown characteristics, stem shape, wood density, latewood and compression wood proportions, and strength properties of small clear specimens of wood (MOR, MOE) of overturned, snapped or undamaged trees (wind and wind/snow damage). To date, data sets and correlations on crown architecture and taper indices linked to wind (and wind/snow) damage are available for three stands of Sitka spruce and for one stand of Scots pine. The data sets of basic wood properties of trees which have broken, uprooted and undamaged are also proceeding well. However, no predictive correlations have yet been produced due to the generally non-significant differences between the data for the different damage types. P02 has also calculated correlations between root properties and turning moments required to overturn or break trees by static loading using Finnish tree pulling data. The results from the work to date within this sub-task are provisional and further analysis is thus required before final conclusions can be drawn. Within the period of report, one PhD-thesis has accepted within this Sub-task (Dunham 1996).

Sub-task 2.2. During this period of work critical parameters controlling snow and wind damage for a sub-sample of sample plots have been determined and a data base from the same sub-samples made available to P02. Although the above damage frequency for Västerbotten was low, the preliminary logistic model developed show that it is possible to predict future damage from snow and wind by using single tree characteristics as indicators of site risk. In this work P07 has identified the following characteristics; for spruce diameter at breast height (1.3m) and height, for Scots pine upper diameter (i.e. diameter at 3 or 5 metre) and ratio of height/diameter at breast height, and for birch height and whole quota (i.e. diameter at 5m/diameter at stump height [1% of tree height]). For Scots pine the predicted damage probability to a site has been found higher if the largest tree has low ratio of height/diameter at breast height. Within this Sub-task, P07 has also started dissemination of some research findings within the period of report (see Valinger and Fridman 1996). Further information from this Sub-task will be delivered for task 3, 4, 5, and 6 during forthcoming months of project work.

Sub-task 2.3. Tree pulling experiments. The extensive tree pulling database of total number of over 1800 trees has been constructed by P04, and linked to this work P01 has constructed Finnish tree pulling database of 115 trees (Scots pine, Norway spruce and birch spp.) to be added to the extensive tree pulling database. Regressions have also been calculated between various tree and site characteristics and critical turning moments needed to cause uprooting of single trees or stem breakage based on these tree pulling databases by P01 and P04 for determination of critical parameters required to support the development and the validation of the models to be done in Sub­tasks 3.1 and 3.3 (see e.g. Granander 1996; Peltola et al. 1996f). Additional tree pulling data have also been made available by P4 and by scientists in the United States, Canada, Germany and New Zealand to the database. The integrity of the recently added data is being checked and quality assurance carried out on all the data by P04.

Wind and tree swaying measurements. P01 has also made wind and tree swaying measurements at the edge of a Scots pine stand and within the stand especially two tree heights from the edge prior and after two thinnings (see Peltola 1995, 1996a, Hassinen et al. 1996). To date, mean wind profiles and stem bending of trees at the stand edge and within the stand have been analysed in a Scots pine stand having stand density of 2700 and 1500 stems/ha, i.e. prior and after first thinning (see Peltola 1995, 1996a). Whereas, analysis of measurements after second thinning will be finished within couple of furthcoming months. The wind and tree swaying experiment carried out by P01 will provide further information on mean windspeed profiles and tree stem bending for varying stand densities, and give thus essential information for subsequent modelling work and model testing to be carried out in Sub­tasks 3.1 and 3.3.

Monotonic and dynamic tests. P08 has developed a new video analysis technique to be used to estimate the overturning moments experienced by a tree during a storm event. In addition, P08 has selected test sites with test trees (three site preparations on surface water gley) in Ireland and currently a field testing programme of both monotonic and dynamic loading tests of trees on the various site preparations as well as a long term tensiometry study of the characteristics of the three site preparations will be ready for implementation.

Dissemination of some research findings is currently under work in Sub-task 2.3 (see Granander 1996; Hassinen et al. 1996; Peltola 1995, 1996a; Peltola et al. 1996f).

Sub-task 2.4. The different types of data collected in this Sub-task has been digitised and compiled in Excel databases to be integrated in the respective Sub-tasks (4.3 and 5.4).

Task 3: Tree-level models to predict circumstances for damage

Sub-task 3.1. A series of models for calculating critical windspeed for damage have now been constructed based on the theoretical arguments set out in the Materials and Methods section: 1) Break.exe: Pascal programme designed to take information from the GIS and produce a file of critical windspeeds which can be read back into GIS for subsequent display or conversion into a probability map (information provided from the GIS includes tree height, tree dbh, tree species, tree spacing and soil type. Integration of this model with the GIS has already been successfully tested in one of the forests being used to validate the models in Task 7.3; 2) Treesnap.exe: Pascal programme designed to directly accept information from the user. Species and soil are selected from menus and tree height, tree dbh and tree spacing are manually input. The model outputs the critical wind speeds for breakage and overturning at both canopy top and at 10 metres above the canopy and the base bending moment at failure; 3) Treebrk1.exe: A Windows based programme written in Delphi software which has the same function as Treesnap.exe. Programming in Delphi will allow much easier integration with other Windows computer packages; and 4) Gales.exe: A modular version of Break.exe and Treesnap.exe above which breaks all the different components of the model into separate units. The aim is to allow easy modification or change to any part of the model and to make the structure of the model much easier to comprehend by other Participants in the project. The interface of this version of the model allows either data input from a GIS or directly by the user.

The latest version of the model incorporates modifications which allow the bending of the tree to be calculated for a particular wind load. This removes the need to use an empirical correction to allow for the overhanging weight of the tree and allows the effect of snow loading in the canopy to be accounted for in addition. Streamlining of the canopy is now incorporated in the model based on the wind tunnel tests of Mayhead (1973). The effect of distance from a new edge, such as that created by clear-felling upwind, has been incorporated using the empirical relationships based on previous wind tunnel studies available by P04, and these discussed in Task 4.1.

Sub­task 3.2. The logistic models developed under Sub­task 2.2. have been further developed, not only for separate species, but also for tree, stand, and site characteristics. A database on snow and wind damage has also been prepared and is available. During this period of work the developed logistic model for prediction of damage, and a separate dataset in Excel­format have been made available to all other Participants through P07 www­pages. Within this sub-task models predicting snow and wind damage to sites have been developed using 1) tree-, 2) stand-, 3) site-, and 4) tree, stand, and site characteristics as input variables. The developed models have been evaluated by using temporay sample plots from the county of Västerbotten in the NFI database and permanent sample plot data from other counties. The fitness of the developed models show that the methodology used can be possible to be included into forestry management practicies when information on the susceptibility of snow and wind damage to sites are needed.

Sub­task 3.3. The updated mechanistic model of wind and snow damage of single trees (HWIND) developed under Sub­task 3.3 by P01 will be capable of calculating the mean windspeed to uproot or break a tree along with critical snow loading for various distances from the stand edge at the margins of clear-felled areas. To date, the model covers especially Scots pine, Norway spruce and birch spp. growing on podzolic soils in Finnish conditions. However, other tree species and soil types (for various geographical locations) can be used by changing the controlling parameters and equations for species (and soils). The HWIND-model can be run in PC­computer both within Windows having figures and tables drawn by specific graphical application made by Visual Basic or without Windows (tables available). Updated version of HWIND-model is also available for GIS (UNIX-version).

To date, HWIND­model has been tested/improved based on Finnish tree pulling database constructed in Sub­task 2.3 for Scots pine, Norway spruce and Birch sp. Especially, the critical parameters of HWIND­model for rooting characteristics, as depth and width of root-soil plate and the significance of root-soil plate weight of the total anchorage have been improved, and values for modulus of rupture of various tree species modified. Whereas, validation of HWIND-model with measurements of windspeed and tree movement prior and after thinnings by P01 (data from Sub-task 2.3) will be finished within couple of furthcoming months. On the other hand, sensitivity tests have already been conducted for HWIND-model by P01 to determine the critical parameters involved and the reliability that can be placed on the critical windspeed predictions.

HWIND­model predictions by P01 have also been compared with the predictions based on empirical wind damage model developed by P04 in Sub­task 3.1 to improve the methods for estimating critical windspeed (see Gardiner and Peltola 1996a, 1996b). These models adopt different approaches to both the method for calculating the mean wind loading on the trees and the resistance to overturning provided by the trees. However, both models use a very similar approach to predicting when a tree will break. To date, these models have been compared for Scots pine and Norway spruce growing on podzolic soils which are two of the few situations common to Great Britain and Finland. Good agreement has been obtained between the models in tests carried out for a variety of tree heights, stem diameters and inter­tree spacing. However, some discrepancies have arisen which have necessitated further investigations into the assumptions on which the models are based.

In addition, a literature review has been finished on the factors affecting the snow damage of trees to support the mechanistic model development and its validation within Sub­task 3.3, in a co­operation by P01, P04 and P05 (see Nykänen et al. 1996). Dissimination of some research findings within this Sub­task 3.3 is currently under work (Gardiner and Peltola 1996a, 1996b; Nykänen et al. 1996; Peltola 1995, 1996c, 1996d; Peltola et al. 1995, 1996a, 1996b, 1996c, 1996d; see also Metamodel form for HWIND, 1996).

Sub­task 3.4. Based on the evaluation of allometric models, it was found that the diameter at the base of the live crown (DBC) was a single variable most strongly related to the leaf area. However, the measurements of this variable on a large amount of trees is impractical and very time consuming. Therefore, it were tested some DBC surrogated variables to be included in the model. The selected model is a generalisation of the allometric equation using two independent variables, i.e. the height to the base of the live crown and a surrogated variable of the DBC of which value is computed from the diameter at breast height and the crown depth measurements. The developed model has a regional applicability and for its application only the DBH, the tree total height and the height to the base of the live crown has to be measured, what can easily be accomplished with a calliper and a hypsometer. Furthermore, based on the canopy transmittance measurements from 25 field plots, it was concluded that the LAI can be efficiently determined with the Sunfleck Ceptometer. The determination of the LAI using this device can be done with a single set of measurements taken during a limited range of sun zenith angles. On the other hand, the DEMON requires several measurements at different solar angles to produce a precise measurement of the LAI, and therefore this device was considered impractical for a systematic use in this Sub-task. Within the period of report, also a linear model to predict the LAI as a function of the logarithm of the transmittance has been fitted to the available data with good predictive abilities. A more detailed analysis based on radiation transfer models is still ongoing.

Task 4: Defining stand-level variability to permit application of tree-level models

Task 4.1. Preliminary efforts have been made by P01 to calculate the wind loading on individual Scots pines as a function to distance from stand edge, stand density (in terms of leaf area index) and tree height and width of upwind gaps using airflow model developed by Miller at the University of Connecticut (see Peltola 1996b). In addition, because there has been some initial delay in getting the more sophisticated airflow modelling efforts available by P04, empirical relationships have been derived from existing wind tunnel studies to relate mean and extreme wind loading to distance from stand edge, stand density and tree height (in dissemination: BENDMOM.WP6). These empirical relationships derived from existing wind tunnel studies by P04, and used in the HWIND-model (Sub-task 3.3) has allowed progress within this Sub-task in simulation of the wind loading and critical windspeed to uprooting and stem breakage of individual Scots pines, Norway spruces and birch spp as a function to distance from stand edge, stand density (stems/ha) and tree height (see Gardiner and Peltola 1996b; Peltola et al. 1996e), until the more improved numerical airflow models become available by P04.

Sub-task 4.2. Within the period of report, progress has been devoted to determining the best approaches, the identification of existing data sets and models, and the development of existing methodologies. A three dimensional model of the forest stand in the Cwm Berwyn forest, Wales has been obtained from aerial photographs in 1957 and 1995 by P04. Experiments have been carried out on the level of spatial resolution that may be derived, while not exceeding the integrity of its quality based upon the source documents and processing. Estimates have been made of the accuracy of the derived X, Y and Z values of the canopy surface based on comparison with data from 169 field observations using a GPS. The results show root mean square estimates of X = 0.79m, Y = 0.86m and Z = 0.85m. These figures suggest that the height at a point in the forest canopy may be estimated to within 1m. Further, some of the physical conditions in the vicinity of a pocket of wind damage may be interpretable from visual observations of the ortho­corrected photography for 1957 and 1995 and the 1995 elevation model. Within this Sub-task, P07 has also provided information on the calculated wind and snow risk values for the sample plots within the County of Vasterbotten in Sweden based on single tree characteristics (in dissemination: ACSCOTP.XLS).

Sub-task 4.3. The simulations of the spectral signals from different simulated forests has identified the problems of using analysis techniques derived for multispectral analysis in the analysis of hyperspectral data. For low canopy LAI, the ground cover dominates the spectral signal so that the same signal can be obtained from forests with different LAI values which also have different ground cover. For LAI values between 2­3.5, the type of background is less important and the NDVI technique may be of value in determining canopy LAI. At higher values of LAI, the spectral signal is little affected by changes in LAI and the technique is no longer of use. Future work will concentrate on developing new analysis tools for hyperspectral data and for obtaining good quality field data for testing the applicability of these tools.

Sub-task 4.4. Forty five points have had an inventory completed while preparing the fuel hazard knowledge base by P03. The coordinates of each sample point were collected with a GPS and a layer with the position of the points has been created and included in the geographic data base.

Task 5: Regional-level snow, wind and fire risk models

Sub-task 5.1. Within the period of report, P01 has provided long-term annual hourly maximum windspeed data (for years of 1971-1990) based on measured mean windspeeds of 10 minute periods at 10 m above ground, along with windspeeds at 40 m above ground for wind atlas classes 2 (above open field) and 4 (above forest) calculated by FMI for the Finnish test area.

Within this Sub-task, the main effort made by P04 has been to test various airflow models against a set of wind data obtained from the Cowal peninsula (between 1/9/90-1/2/92), which is much more highly incised (slopes are typically 1 in 4) than the area for which the previously reported model comparisons were tested (the Kintyre peninsula). Six anemometer masts has been operated during this period and although the spatial coverage has been poorer than for Kintyre the masts has been operated for much longer and allowed P04 to test the ability of the models to recreate the wind climate for an area. In this experiment, the most Westerly mast on Sron Cruaich is used to initialise the models. An attempt has been made to obtain regressions between this site and the other sites so that periods of missing data can be filled in. However, to date these regressions are poor except for the hill top site at Cruach Buidhe, illustrating the difficulty of modelling windspeed in this kind of terrain.

A further airflow model (MSFD), which has shown some promise in simulating the patterns of wind damage in New Zealand and some improvement over previous models in predicting the windspeed fluctuations across the Kintyre peninsula, has also been included in the comparisons by P04. This model does not, however, linearise the fluid flow equations, which have to be solved numerically, and consequently it takes substantially longer to run than the previous models discussed. The models require surfaces of topography and roughness to initialise them. Links have been established with P05 to derive the roughness surface from land cover datasets held by them; the data has been transferred by FTP but there have been a number of format problems to overcome before the roughness layer is ready. Within this Sub-task, efforts continue to test also another model, i.e. RAMS, against data from the Kintyre peninsula to determine whether this highly sophisticated model is able to perform better predictions of windspeed variation in complex terrain than the linear models P04 have tested up to this point. Efforts continue also to refine the DAMS windiness scoring system (Bell, Quine and Wright, 1995; Quine and White, 1994) by considering the effects of local shelter through distance limited topographic indices. The relationship would be explored by P04 between mode/range variables of the extreme value distributions and windiness score using data from a number of Meteorological Office sites in Britain. P04 compares also the method on an independent dataset produced by a Wind Energy company as part of a renewable energy project.

Climatological, site, silvicultural and tree factors which affect the risk of snow damage have also already been identified in a snow review paper produced by P01, P04, and P05 within Sub-task 3.3, and are considered as important inputs for the tree and climatological modelling within this Sub-task. It is envisaged that the model will be encoded on a GIS and will be able to calculate several outputs relating to snow incidence in the UK (number of days with falling snow, probabilities of return periods of snow days which exceed the average winter, the duration of falling snow and snow accumulation) with the risk of damage enhanced by site factors such as exposure and aspect.

Sub-task 5.2. For the Finnish test areas, P01 have stored in the geographic database the forest stand boundary layers together with the forest stand attribute data. Furthermore, a metadata form of the Finnish test area has been filled for integration (Task 6). A digital elevation model covering the whole test area has also been added to the geographical database. Also the wind damage measurements have been completed during February 1996 and the wind damage data of total of 37 stands have been stored in the geographic database and added to the existing data. The satellite images have also been preprocessed for further use (Sub-task 7.1), and a land use classification has been made available for the Finnish test area further to be used in calculation of regional wind pattern in Sub-task 5.1.

Within the period of report, P03 have made the digital terrain model ready to use, i.e. 45 (each with 16x10 Km wide) military contour maps (1:25000) were assembled in a vectorial format and converted to a single raster file. Slope and aspect maps are also ready. To date P03 is waiting for the availability of the aerial photos of 1995 and for the digital files with the landuse extracted from the aerial survey of 1990. For burned areas, the digital maps for 1990, 1991, 1992, 1993 and 1994 have been provided by P06, and this information merged and co-registered in the digital database (GIS). Field plots boundaries are also stored in the database using a GPS.

For the UK test sites, P04 has continued the construction of datasets for windthrow monitoring areas, and aerial photography obtained during summer 1995 has now been interpreted and added to the GIS data layers. Further photography has been commissioned for 1996. GPS survey is being extended to add to the number of precise locations of basal area plots, windthrown gaps and boundaries. The method of collating metadata information via the web form has already proved highly successful with participants completing forms as data sets become available. P05 have made available on the web several datasets (that were provided by the others participants) and the method has been particularly useful for the initial demonstration work in Task 6.

Sub-task 5.3. The results of the data investigations was a report documenting on what data was available for where and from what organisation the data could be obtained. It also includes other information such as data format, coverage, scale and error. This information will feed into a task 6 component to assess data availability for areas where model output might be required. The Pan-European datasets report has been completed, distributed and is now available on the ftp site and will eventually be available on the web site.

Sub-task 5.4. The results achieved so far shows that linear mixture modelling Software works well when using image end-members.

Sub-task 5.5. No result to present yet.

Task 6: Integrating components from tree/stand regional-level to produce an unified risk model

Sub-task 6.1. The completed metadata and metamodel forms are on the WWW site and have been accessed by all of the groups involved via a matrix interface. All forms have been completed for the demonstrator and the meta information is available for use by participants on the project WWW pages. Forms have also been completed for datasets and models which are not involved in the demonstration. The prototype system plan for an integrated unified risk model was conceived using the information from the forms and the results from the demonstrator. This plan was modified and approved during the last meeting in 1996 after consultation with the project group. The prototype comprises several components. Firstly there is a demonstrator unit which incorporates worked examples of risk assessment using a suite of models for each damage type run on selected areas. The user will be introduced to the characteristics of the particular area and then taken through the stages involved in deriving the risk assessment. This will involve illustrating the error and validity aspects of model application, tree and climate model integration, and the consequences of using inappropriate spatial scales or poor quality data. The demonstration will be made as interactive as possible to give the user a "feel" for the model capabilities and the use of models as an experience to help understand the system for decision support, rather than a set of calculations which give definitive results. This section will also introduce the concepts of silviculture and its effect on risk for different damage types.

The user will also be able to access more detailed information about the models and the different strategies which have been used to address risk assessment within the project. Access will be possible through a variety of routes. Current plans propose that the geographic and data space of interest be defined so that the system can guide the user to particular model suites, which are of most relevance. The user may also wish to access models according to the modelling approaches which have been used (e.g. mechanistic versus logistic). In addition, the user will be presented with a complete description of the model and appropriate data. Information from the metamodel forms will be presented in a way that is more digestible than the forms themselves, and links will be made to the metadata forms and the European data sets report where relevant. It was felt particularly important to portray the issues of data quality and the limitations of model applicability to reduce the possibility of model misuse. It was thought that this might be best achieved through worked examples. Parts of the interface will facilitate simulated interactions with the models to allow users to get a better feel for how they operate.

Sub-task 6.2. There are no results from the data-specific scale investigations yet.

Task 7: Testing models against independent data and outlining implications for silvicultural strategies

Sub-task 7.1. The change detection method was developed and tested using Landsat TM images by P01. The key results was that drastic damages could be classified correctly, but minor damages were classified only 50% correctly. The testing of the predictions by HWIND-model developed within Sub-task 3.3 by P01 has also been started at the end of 1996 and some preliminary computations has already been made. These computations showed that the HWIND-model could identify the wind damaged stands. However, it predicts wind damage also to some undamaged stands because the position of the stand in respect to topography of the test area, and thus actual wind climate, was not taken into account in these computations.

Task 8. Final products, documentation and identification of new opportunities

To date, major datasets and models from all partners have been collated and the information is available for participant perusal on the web. This not only allows the information to be standardized and communicated clearly, but limits the capacity for misunderstandings between participants and therefore the number of unnecessary queries which would otherwise be required for the interaction and integration of data and models. It also provides a detailed documented record of data and models. The public pages on the web site are being further developed to explain, in more detail, the objectives of the project and eventually to present some of the project findings and provide visitors with references to papers and documents which have been produced by the project. Dissemination of scientific results is being currently effectively achieved via journal and conference papers, presentations made to visitors and by distribution of the project leaflets.


CHAPTER 4 DISCUSSION

Task 1: Project planning

The increasing use of the metadata and metamodel forms to document the models and datasets used in the course of the project have proved an invaluable means of communication and provision of information for input to the Task 6 framework development. The rapid change in software packages and the availability of data (particularly digital versions of datasets) is now reflected in the rapidity with which the information can be updated and accessed throughout the project.

Task 2: Quantification of component factors controlling snow, wind and fire damage

new:

Sub-task 2.1. Data sets and correlations on crown architecture and taper indices linked to wind (and wind/snow) damage have been compiled for three stands of Sitka spruce and for one stand of Scots pine. The data sets of basic wood properties of broken, uprooted and undamaged trees is proceeding well. However, no simple relationships between damage type and stem form, canopy shape and wood properties has been found by P02. These results suggest that defects (knots, resin pockets, grazing damage, earlier damage) are important in modifying tree strenght. Also because strenght is an important modifier to wind and snow risk models, further work is therefore needed to better determine the role of knots and other defects in terms of their effect on the modulus of rupture of standing trees and their consequent modification on damage risk. This addition work needed to support the models development/validation in Sub-tasks 3.1 and 3.3 has been emphasized also in ACTION-section (along the proposal for additional work to be needed to carried out under the STORMS project).

Sub-task 2.2. All data needed is already available and are possible to use by other participants. The parameters controlling snow and wind will be further developed for the whole country during the following months of project work, and included into model development. To date, dissemination of some research findings is under work within this Sub-task (see Valinger and Fridman 1996).

Sub-task 2.3. The extensive tree pulling database of total number of over 1800 trees has been constructed by P04, and linked to this work P01 has constructed Finnish tree pulling database of 115 trees (Scots pine, Norway spruce and birch spp.) as was aimed in technical annex within the period of report. To date, regressions have also been calculated between various tree and site characteristics and critical turning moments needed to cause uprooting of single trees or stem breakage based on these tree pulling databases by P01 and P04 for determination of critical parameters required to support the development and the validation of the models to be done in Sub­tasks 3.1 and 3.3. However, additional tree pullings should be considered especially for birch spp. (small data) to support the development/validation of its tree species parameters for the models to be developed within Sub-tasks 3.1 and 3.3. To date, P01 has made available also the values for the critical parameters for the HWIND-model based on Finnish tree pullings as aimed in technical annex within period of report. However, the analysis of measurements of windspeed and tree movement at the edge of a Scots pine stand and within the stand prior and after two thinnings (see Hassinen et al. 1996, Peltola 1996a) by P01 is not yet completed, but it will be done in forthcoming months for further model testing and validation to be carried out within Sub-tasks 3.1 and 3.3, in addition to data already available from previous studies by P04 with Sitka spruce stands of differend densities and at different distances from the edges.

Furthermore, in dynamic loading tests by P08, it remains to see how appropriate for a large­scale tree a rocker system in the field will be (tests made using a small scale prototype rocker and a young tree). It is hoped that if the system proves successful it will enable to make one of the rather cumbersome deflection measuring systems redundant and will make the study of trees under natural storm conditions much simpler. The aim of finding a factor relating monotonic overturning moments to that obtained at presumed failure in the dynamic tests may prove to be soil and/or site specific but this remains to be determined. Dissemination of some research findings within Sub-task 2.3 has already been carried out, but is also currently under work in Sub-task 2.3 (see Granander 1996; Peltola 1995, 1996a, Peltola et al. 1996f, Hassinen et al. 1996).

Sub-task 2.4. The bad weather conditions during the 3 months of field work in 1996, namely large cloud cover and high levels of diffuse radiation, were a restraint for the planned spectral data gathering and LAI evaluation by light measurements. The need for gathering more spectral and LAI data is evaluated in the methodologies under development within Sub-task 5.4.

Task 3: Tree-level models to predict circumstances for damage

Sub-task 3.1. During the next 6 months P04 will continue the work with P01 to make more detailed comparisons to decide on the best overall approach to be taken in the development of the models. This may possibly involve the construction of a single combined model to make use of the best aspects of each model. As part of this exercise results with the exponential profile and 1D canopy flow models will be tested against results obtained with the existing model. Also dynamical tree model developed by P04 (Kerzenmacher and Gardiner 1996) will be used to test whether the gust factor expression used in the damage models is valid over all windspeeds. Dissimination of some research findings within this Sub­task 3.1 has already been carried out, but will also continue within the next period of report (e.g. Kerzenmacher and Gardiner 1996, Gardiner and Peltola 1996a, 1996b).

Sub-task 3.2. The data needed for the work is already available at this state of the work by P07. The preliminary results obtained, show that characteristics of single trees, stands, and sites are possible to use as measures of future risk of damage to sites from snow and wind. The work so far have been concentrated to tests of the initial hypotheses and evaluation of the technique used. During next 6 months risk assessment models for Scots pine, Norway spruce, and birch for the whole country will be developed and the developed logistic models will be evaluated using available data from other participants within this project. Dissimination of some research findings within this Sub­task 3.2 has already been carried out (Fridman and Valinger 1996), but will also continue within the next period of report.

Sub-task 3.3. The mechanistic model of wind and snow damage of single Scots pine, Norway spruce and Birch spp growing on podzolic soils (HWIND) has now been improved/tested using especially information obtained from Finnish tree pullings within Sub-task 2.3. There has been existing an initial delay in finishing the model testing/validation using also wind and tree swaying measurements from sub-task 2.3 as was aimed to be made in technical annex within the period of report). This is because the analysis of these measurements has not yet been finished in Sub-task 2.3. However, within a couple of forthcoming months, the HWIND-model will be further tested/validated against data from Sub-task 2.3. In addition, more detailed comparisons will be made between HWIND-model (by P01) and the Empirical Wind Damage Model (by P04) to improve the methods for estimating critical windspeed. P01 will also try to improve the HWIND-model to be able to use also some other tree species and possibly some other soil types as inputs in addition to the current tree species and podzolic soil conditions. Furthermore, also the effect of the width of the upwind gap on wind loading is aimed to be included to the HWIND-model in forthcoming months, allowing to account for management operations upwind.

Furthermore, based on discussions during the mid-term meeting hold in Bryssels, within the next period of report P01 will also try to simulate with stand growth model available by P01 various stand management alternatives (thinning regimes) and study the forest production (in a stand level) in relation to critical windspeed / snow loading to trees. Dissimination of some research findings within this Sub­task 3.3 has already been carried out (Nykänen et al. 1996; Peltola 1995, 1996c; Peltola et al. 1995, 1996c, Metamodel form for HWIND, 1996), but is also currently under work (e.g. Gardiner and Peltola 1996b, Peltola et al. 1996a; 1996b, 1996d). Within the period of report, P01 has also participated to the Second international workshop on disturbance dynamics in boreal forests, in Quebec, Canada, 1996, and presented there some preliminary results of this sub-task (see Peltola 1996d; Gardiner and Peltola 1996a). However, the participation to that workshop was payed by other financial sources than the Storms project.

Sub-task 3.4. At the end of this 24 months period two indirect methods to estimate the pine stands LAI have been established and have already been used on vegetation characterisations under Sub-task 2.4 by P03. Within next 6 months, if possible, some more data should be collected to ensure a deeper knowledge on the variables most significant to predict the LAI of maritime pine stands.

Task 4: Defining stand-level variability to permit application of tree-level models

Sub-task 4.1. There has been some delay in getting sub­task 4.1 up and running but alternative approaches have enabled to provide the necessary information from this part of the project (see Gardiner and Stacey 1995; Gardiner et al. 1996; Peltola 1996b; Peltola et al. 1996e). Within next period of report, the mathematical relationships relating wind loading to trees (as a function of height, stand density, trees position within a stand, and the effect of the width of the upwind gap) will be aimed to be made available.

Sub-task 4.2. Provision of test data continues. Good progress has been made during this reporting period, particularly with regard to the remote sensing aspects of the work. The results of the derivation of a model of the forest canopy from aerial photographs shows good potential for analysis with respect to changes in height, openings in the canopy due to damage and variability in the surface of the canopy. The geometric accuracy of the model derived from photography taken in 1995 was higher in quality to that derived from the historical photography of 1957. This was largely due to a lack of adequate information about the lens distortions in the older camera. However, the central area of the older model was sufficiently good to provide a model of the terrain prior to afforestation with which to compare the, currently, forested landscape. From this model initial studies are being carried out on the variability in the canopy height surface and on tree height. The quality of the orthophotographs, in terms of the features that can be identified when overlaying the two dates of co­registered imagery, has provided an additional opportunity to interpret the conditions that underlay the forest canopy and, in particular, in those areas where trees have been blown down. The ability to visualise the landscape from different perspectives, in some detail, has been an additional output from this Sub­Task.

Sub-task 4.3. Good progress has been made particularly with regard to the remote sensing aspects of the work.

Sub-task 4.4. Within the period of report, also within this Sub-task good progress has been made particularly with regard to the remote sensing aspects of the work.

Task 5: Regional-level snow, wind and fire risk models

Sub-task 5.1. This Sub-task is being developed well according to the plans. Digitised snow and wind atlases for Sweden will be available in GIS during the next 6 months by P07. Furthermore, P04 will shortly be able to compare wind climate estimates using weighted model outputs, with DAMS predictions and with actual field measurement. This will give a clearer idea of which method can be recommended, or whether a further hybrid method must be derived to cope with extreme topography. Encoding Jackson's statistical model on a GIS by P05, will also allow a small scale risk assessment to be calculated for a particular location in the UK. Regional site information for that particular area can then be used to modify and refine this crude risk estimate to a value which accounts for the individual site factors in a particular location. Thus small (national) scale climatological information can be used together with site factors to define the climatological site risk for any area in the UK. Using a GIS will make the model interactive and user-friendly and allow spatial manipulation and integration with other spatial information.

Considering the difficulties associated with modelling the incidence of damaging snow, the approach identified appears to encompass the essential climatological elements. It is hoped that some further investigations will result in satisfactory progress in this area which is more complex and much less well researched than originally anticipated by participants.

Sub-task 5.2. The aims identified by P01 for the period of report has been achieved and the geographic database of Finnish test areas is available for further use in various tasks. Closer collaboration by P01 and P07 in tackling the problems in contructing database is being pursued, since many of the problems and possible approaches seem to be comparable. Further deployment of the GPS and acquisition of aerial photography is also being undertaken during next period of report by P04, i.e. a simple vector-based GIS (ARCVIEW2) will be acquired to complement the existing raster-based system, this enables P04 also to view the datasets of other participants produced by ARCINFO. The potential to acquire a vector-based GIS to complement the existing raster-based system is currently being investigated. Within this Sub-task, all the aims identified by P03 for the period of report have also been achieved.

The metadata forms are proving an extremely useful method of eliciting detailed information about the project data sets. Having the forms on the web has allowed efficient collection, collation and dissemination of this material, which is fundamental to the success of the project. Furthermore, the metadata forms were kept simple for ease of use but are comprehensive enough to ensure that all relevant criteria are covered. Completion of these forms will save time in the future by minimising the possibility of important information being omitted during the collection process which might be difficult to recover later in the project. The forms constitute a summary of important metadata necessary for data use. Minor changes were made to the forms in response to comments from P05 data capture personnel and project participants who have some experience in the area. A system of metadata documentation exists whereby participants can complete forms remotely, in their own time, which will be automatically stored, in a central location, for viewing by all participants. This maintains clarity of communication with minimal effort. All this information will be used in the Task 6 integration of components, as reference material for participants who are exchanging data and models, and as a general form of documentation to collate what data is available in a European context.

Sub-task 5.3. The report is complete but "live" and so additions will continue throughout the project, and will serve as a useful reference document for this project and future European work. The information contained in the report will be important for use in Task 6 to create a system framework which has valid links between models and data. There also exists the possibility of graphically incorporating this information into the system to allow easy identification of relevant data sets for particular model runs and to enhance its user-friendliness. The report revealed some major gaps in European coverage and the inadequacy of data sets currently available when compared to the metadata form resulting from Sub-task 5.1. A significant contribution of this project to the whole subject of European data collection, dissemination and use is that these issues are being addressed in the context of metadata provision. The report has been a useful basis for organising Pan-European and participants data and identifying gaps in data sets available and gaps in the metadata associated with the data which is available.

Sub-task 5.4. The Sub-task is being developed according to the plans and there were no deliverables or milestones to be achieved during this period.

Sub-task 5.5. This Sub-task is also being developed according to the plans quite well, although there exits a slight delay in the execution of this task that is a consequence of the delay in the data collection work of Sub-task 4.4.

Task 6: Integrating components from tree/stand regional-level to produce an unified risk model

Sub-task 6.1. It was important to avoid compressing all the integration issues and software development into the last few months of the project. Many of the issues can be identified, and skeletal framework plans developed, at an early stage. Prototyping and framework development can progress in stages through consultation with all participants. Documentation of the models and data is particularly important to provide structured information to participants which will ensure effective integration, mutual understanding between participants of different national and regional datasets, models and damage issues. This strategy can then evolve to impart such knowledge to potential model users. Prototyping is a proven method of system development and has been adopted for the development of the unified risk model interface.

The demonstrator results and comments from participants about the various integration issues were used to develop the framework further into plans for a prototype system. These plans were presented to participants before the last project meeting 1996, and were critically discussed during this meeting. Participants presented their view of integration and particularly means of presenting the error and scale issues associated with their models and their integration with other models. Thus the framework plans were modified according to these comments and a timetable agreed for implementation of a prototype. This prototype would be completed for the next project meeting in 1997 and would provide a basis for further structured discussion of model integration. Furthermore the results would provide a means of discussing, as a group, the performance of the models and the quantification of risk assessment. The demonstration was conceived during the last year project meeting in Scotland. Following participant presentations, and from discussions throughout the year via email, it was obvious that there were potentially a large number of links between participants in terms of models and data. Several hitherto unforeseen additional links had developed where the expertise of one participant could be employed to explore issues involving another participant's work. Thus any problems or avenues of exploration could be identified at an early stage.

Because the aspect of integration and communication of results through the Web has

gained a lot more importance than had been anticipated at the beginning, P03 finds difficult to keep same standards as the other partners while integrating their work in the final framework designed during the project unless any additional funding for that work will not be available (see ACTION- section).

Sub-task 6.2. Preliminary findings from these scale investigations (currently in progress) for a single dataset will be used to determine the method of investigation for subsequent data sets, since the scale structure of data will differ depending on the type of information and method of recording. Initial scale studies will consider datasets which are already available at several scales (Pan-European - identified in Sub-task 5.2, regional and local) to look at the effect of generalisation and sampling strategies on the representation of different datasets. The results of these studies can then be used as a guide to the limits of resolution of different data at different scales for different areas, and the effect of changing scales. Using these results and information from the metamodel forms, it will be possible to explore the validity of model output compared to the scale structure of input data.

Task 7: Testing models against independent data and outlining implications for silvicultural strategies

Sub-task 7.1. First part of this Sub-task, i.e. the development of the change detection method, was as a independent part of this task achieved well in advance. The results were encouraging, but the lack of adequate test data prevented the futher testing. Another part of this task, the testing of the predictions by HWIND-model developed within Sub-task 3.3. by P01 has been started at the end of 1996 and some preliminary results are available. However, the testing will be continued within next period of report and the effect of topography on local wind climate will also be included to the analysis in a co-operation with P04, who will simulate wind climate to the Finnish test area.

Task 8. Final products, documentation and identification of new opportunities

A number of papers has been published or submitted to scientific journals. The public pages are being further developed to provide more detailed information about the project and include images of damage being studied. In addition, a significant expansion of the private pages is permitting exchange and development of research ideas, and the display of preliminary results to facilitate communication and mutual understanding which enhances project progress. Full participant details have also been added to the web, and participant web sites, personal pages and email addresses are available for direct access through the pages.


CHAPTER 5 DISSEMINATION

The following publications and scientific manuscripts have been published, are in press, accepted or submitted within the first 24 months of Project by Participants:

Bell, P. D., Quine, C. P. and Wright, J. A. 1995. The use of digital terrain models to calculate windiness scores for the windthrow hazard classification. Scottish Forestry 49(4): 217-225.

Caetano, M.S.; Mertes, L.A.K.; Cadete; L.and Pereira J.M.S. 1996. Assessment of AVHRR data for characterizing burned areas and post­fire vegetation recovery. EARSEL International Workshop: Remote Sensing and GIS Applications to Forest Fire Management. Alcalá de Henares, Spain Sept 7­9 . 49­52pp.

Chalmers, S. 1996. The use of the Criterion Survey Laser and ARC/INFO Geographic Information System (GIS) to Survey, Analyse and Predict Windthrow. BSc thesis. University of Aberdeen (unpubl.).

Dunham, R.A. 1996. The influence of growth rate on the wood and stem

properties of silver birch (Betula pendula Roth)" PhD thesis, Aberdeen University, Scotland.

Gardiner, B.A. and Stacey, G.R. 1995. Designing Forest Edges to Improve Wind Stability. Forestry Commission Technical Paper 16, Forestry Commission, Edinburgh.

Gardiner, B.A., Stacey, G.R., Belcher, R.E. and Wood, C.J. 1996. Field and wind tunnel assessments of the implications of respacing and thinning for tree stability. Forestry 70:(3), in press.

Granander, M. 1996. Vääntömomentti koivun (Betula pendula), kuusen (Picea abies) ja männyn (Pinus sylvestris) kaatamiseksi ja katkaisemiseksi staattisella voimalla kuormitettaessa. Metsäteknologian ja puutalouden pro-gradu. Joensuun yliopisto, Metsätieteellinen tiedekunta (thesis for M.Sc., in Finnish).

Lundqvist, L. and Valinger, E. 1996. Stem diameter growth of Scots pine trees after increased mechanical load in the crown during dormancy and (or) growth. Annals of Botany 77, 59-62.

Mackie, A., Bell,P.D., and Quine, C.P. 1996. Where on earth are we with GPS? (Use of Global

positioning systems in forests). Forestry and British Timber, October 1996, 39­41.

Nykänen, M-L, Peltola, H., Quine, C., Kellomäki, S., Broadgate, M. 1996. Factors affecting snow damage of trees: A literature review. Silva Fennica (Accepted).

Peltola, H. 1995. Studies on the mechanism of wind-induced damage of Scots pine. Research Notes 32, University of Joensuu, Faculty of Forestry. D.Sc. (Agr. and For.) thesis. 28 p.

Peltola, H. 1996a. Swaying of trees in response to wind and thinning in a stand of Scots pine. Boundary-Layer Meteorology 77:285-304.

Peltola, H. 1996b. Model computations on the wind flow and turning moment by wind for Scots pine along the margins of clear-cut areas. Forest Ecology and Management 83:203-215.

Peltola, H. 1996c. Modelling the mechanism of wind-induced damage of Scots pine. In: The Finnish Research Programme on Climatic Change. Final Report. (Ed. Jaana Roos). Publications of the Academy of Finland 4/96. pp. 260-263.

Peltola, H., Kellomäki, S. and Väisänen, H. 1996a. Model computations on the impacts of climatic change on soil frost with implications for windthrow risk of trees. Climatic Change (Submitted).

Peltola, H., Nykänen, M­L. and Kellomäki, S. 1996b. Model computations on the critical combination of snow loading and windspeed for snow damage of Scots pine, Norway spruce and birch sp at stand edge. Forest Ecology and Management (Submitted).

Quine, C.P. 1996. The chance of a windfall. Forestry and British Timber, August 1996, 17­19.

Suarez, J. 1995. The Use of WaSP Airflow Model to Estimate the Wind speed in Kintyre. Unpublished Forestry Commission Report.

Suarez, J. (1996). A comparison of methods of estimating the windspeed in Cowal.

Unpublished Forestry Commission Report.

Valinger, E. and Fridman, J. 1996. Modelling probability of snow and wind damage at sites using Scots pine tree characteristics. Forest Ecology and Management (Submitted).

Valinger, E. and Pettersson, N. 1996. Wind and snow damage in a thinning and fertilization experiment in Picea abies in southern Sweden. Forestry 69, 25-33.

Valinger, E., Lundqvist, L. and Sundberg, B. 1995. Mechanical bending stress applied during dormancy and (or) growth stimulates stem diameter growth of Scots pine seedlings. Canadian Journal of Forest Research 25, 886-890.

Varjo, J. 1996. Controlling continuously updated forest data by satellite remote sensing. International Journal of Remote Sensing 17(1):43-67.

Vasconcelos, Maria J. Perestrelo. 1995. Integration of remote sensing and geographic information systems for fire risk management. EARSEL International Workshop: Remote Sensing and GIS Applications to Forest Fire Management. Alcalá de Henares, Spain Sept 7-9 129-149 pp.

Wade, N. 1996. A comparison of the strength properties of windthrown, windsnapped and standing timber, derived from testing of small clear specimens of Picea sitchensis. BSc thesis. University of Aberdeen (unpubl.).

Poster and oral presentations

Broadgate, M. 1996. STORMS Project Group 1996 Integrating tree and environmental models using GIS to develop silvicultural strategies for minimising forest damage. GIS Research UK Conference Proceedings, April 1996 (Poster presentation).

Caetano, M.S.; and Pereira, J.M.S. 1996. The effect of the understorey on the estimation of coniferous forest leaf area index (LAI) based on remotely sensed data. Presented on the conference on Signal and Image Processing for Remote Sensing, Taormina, Italy, 23-27 September, 1996.

Gardiner, B.A. and Peltola, H. 1996. The development and testing of models for predicting the critical windspeed to damage trees. In Conference Abstracts: Second International workshop on disturbance dynamics in boreal forests, in Quebec, Canada, in August 26-30, 1996. Poster presentation.

Miller, D., Quine, C. and Broadgate, M. 1996. The application of digital

photogrammetry for monitoring forest stands. In: The Proceedings of Application of Remote Sensing in European Forest Monitoring, Joint Research Centre, in Vienna, October 1996.

Peltola, H. 1996d. Model computations by the mechanistic model for wind and snow damage of single tree. In Conference Abstracts: Second International workshop on disturbance dynamics in boreal forests, in Quebec, Canada, in August 26-30, 1996. Oral presentation.

Peltola, H., Kellomäki, S., and Väisänen, H. 1995. Model computations on the impacts of the climatic change on the soil frost and the risk of trees for windthrows. In: Climate Change, Biodiversity and Boreal Forest Ecosystems -Conference Abstracts, in Joensuu, in Finland, 30th July to 5th August, 1995. (IBFRA). Poster presentation.

Peltola, H., Kellomäki, S., Väisänen, H. and Nykänen, M-L. 1996e. Impacts of climatic change on soil frost and the risk of trees to windthrows. In: Finnish Climate Change- Conference in Tampere 4-5, June 1996. Poster presentation.

Manuscripts in preparation

Fridman, J. and Valinger, E. 1996. Modelling probability of snow and wind damage at sites using Scots pine sample plot tree, stand, and site characteristics. In prep.

Gardiner, B.A. and Peltola, H. 1996b. The development and testing of models for predicting the critical windspeed to damage trees. To be submitted to Ecological Modelling.

Hassinen, A., Lemettinen, M., Peltola, H., Kellomäki, S., and Gardiner, B.A. 1996. Application of prism based techniques to measurements of tree swaying under dynamic wind loading. Manuscript to be submitted to Agriculture and Forest Meteorology.

Kerzenmacher T. and Gardiner, B. A. 1996. A mathematical model to describe the dynamic behaviour of a spruce tree to the wind. To be submitted to Trees.

Peltola, H., Kellomäki, S., Väisänen, H., Ikonen, V-P. 1996d. HWIND: A mechanistic model for wind and snow damage of Scots pine, Norway spruce and birch sp. Manuscript in preparation.

Peltola, H., Kellomäki, S. and Nykänen, M-L. 1996e. Model computations on the critical windspeed / snow loading for wind / snow damage of Scots pine, Norway spruce and birch sp along the margins of clear-cut areas for various distances to stand edge. Manuscript in preparation.

Peltola, H., Kellomäki, S., Hassinen, A., and Granander, M. 1996f. Mechanical stability of Scots pine, Norway spruce and Birch spp. for wind-induced damage: analysis of tree pulling experiment in Finland. Manuscript in preparation.

Talkkari, A. and Peltola, H. 1997. Testing the mechanistic wind damage model against independent data: A GIS based application. Manuscript in preparation.

Other related papers to Storms project finished within the period of report by participants:

Gardiner, B.A. 1995. The interaction of wind and tree movement in forest canopies. In: Wind and trees. (Eds. Coutts, M. P. and Grace, J.). Cambridge University press, Cambridge. pp. 41-59.

Hannah, P., Palutikof, J. P. and Quine, C. P. 1995. predicting windspeeds for forest areas in complex terrain. In: Wind and trees. (Eds. Coutts, M. P. and Grace, J.). Cambridge University press, Cambridge. pp. 113-129.

Inglis, D. W. F., Choularton, T. W., Stromberg, I. M. F., Gardiner, B. A. and Hill, M. 1995. Testing of a linear airflow model for flow over complex terrain and subject to stable, structured stratification. In: Wind and trees. (Eds. Coutts, M. P. and Grace, J.). Cambridge University press, Cambridge. pp. 88-112.

Quine, C. P. 1995. Assessing the risk of wind damage to forests: practice and pitfalls. In: Wind and trees. (Eds. Coutts, M. P. and Grace, J.). Cambridge University press, Cambridge. pp. 379-403.

Rodgers, M., Casey, A., McMenamin, C. and Hendrick, E. 1995. An experimental investigation of the effects of dynamic loading on coniferous trees planted on wet mineral soils. In: Wind and trees. (Eds. Coutts, M. P. and Grace, J.). Cambridge University press, Cambridge. pp. 204-219.


Meetings:

Project planning meeting was arranged on 24-25th February 1995 in Lisbon, Portugal.

Second project meeting was arranged on 10-13th, August 1995 in Tampere and Joensuu, Finland.

Third project meeting was arranged on 13-17th, March 1996 in Inverary, Scotland.

Project mid-term meeting was arranged on 4-5th, July 1996 in Brussels, Belgium.

Fifth project meeting was arranged on 23-27th, October in Umeå, Sweden.

Models developed:

A self extracting copy HWIND-model for Windows 3.1 is available on Joensuu University ftp-site (gis.joensuu.fi /hwind/windows/hwind09.exe).

The first version of the stand alone empirical model (treesnap.exe) to calculate critical wind speeds for damage and base bending moments has been made available to participants on the Edinburgh University ftp site (ftp.ed.ac.uk/pub/risk).

Growth models developed by Soederberg and Ekoe at Swedish University of Agricultural Sciences in Umea, Sweden using single tree and stand characteristics were made available to other participants.

Databases constructed:

Findata.xls (Finnish tree pulling database during unfrozen soil conditions for Scots pine, Norway spruce and birch spp) is available on Joensuu University ftp-site.

Finfrost.xls (Finnish tree pulling database during frozen soil conditions for Scots pine) is available on Joensuu University ftp-site.

EDGE.XLS: Data on bending moment as function of distance from forest edge and stand density (Excel 5.0 format) has been placed on the Joensuu ftp site (gis.joensuu.fi/metsa3/users/damage/incoming/fc/).

ACSCOTP.XLS: Spreadsheet which provides the calculated wind and snow risk values for the sample plots within the County of Vasterbotten in Sweden based on single tree characteristics (Excel Format)

Other documentation:

The results of each task are being made available on the World Wide Web site for dissemination.

The updated Task 1 document pages and summary diagrams can be found on the ftp site in the directory air/task1/.

A report has been produced as a result of the snow modelling workshop and has been distributed to both workshop and project participants. A subsequent workshop is planned to discuss the success of the adopted strategies to the modelling approach (Sub-task 5.1).

Metamodel and metadata forms are all currently available on the web. As the project progresses more of this will be transferred to the public pages. Eventually the developed Integrated Risk Model interface will be publicly available over the web. The information added using the form will be stored automatically on the ftp site, and will thus be available for perusal by all participants (Sub-task 5.2).

The report detailing information about Pan-European data sets will be available on the ftp site and accessible via the WWW pages (Sub-task 5.3).

The summary diagrams are currently available on the ftp site and can be accessed by all participants there. They are also to be added to the Task 1 planning document (Task 6).

BENDMOM.WP6: Document discussing the derivation of empirical relationships to describe the change in wind loading as a function of stand spacing, tree height and distance from the forest edge has been placed on the Joensuu ftp site (gis.joensuu.fi/metsa3/users/damage/incoming/fc/).


CHAPTER 6 CONCLUSIONS

Coordination and management of the project is crucial to the successful integration of components in Task 6. The web site has proved an effective medium for display and update of the planning document, through which project planning and management, and data and model integration can be achieved. The interrelations of Tasks, Sub-tasks and deliverables, and the participants involved in each stage of the project is fairly complex, and this fast and visual means of communication is now considered an essential ingredient to the success of the project integration. The Web pages, the ftp site and the mailing list have afforded efficient and timely communication and made coordination and management much easier, providing not only mass communication of inter-participant messages, but the exchange of images, documents, diagrams and forms which can be instantly updated with no time delay.

The integration of such a wide variety of models (statistical, mechanistic, logistic, regional, tree scale) and data (Pan-European to wood property scale from many different countries) on such a large collaborative is a major task which requires a simple yet robust framework. The methods described above appear to be working effectively towards facilitating this integration, but many of the details still need to be resolved. Many of the issues should be identified when the demonstration is complete, and discussions at the next project meeting should resolve many of the problems and identify the next stages towards integration.

The web site, the leaflet and the verbal and written presentations have been used at frequent intervals, ensuring that the project is receiving good exposure at major fora and instigating keen interest, useful links and contacts. Communication between participants is extremely good, shown particularly by the development of a demonstrator and the acceptation of a snow damage -review which was the result of inputs given by three different participating institutes, and a poster which was the work of the entire project group.


CHAPTER 7 ACTION

The most of the activities planned to be executed during the the next period January 1997 - July 1997 are well in line the experiences during the current period. However, no simple relationships between damage type and various wood properties has been found within Sub-task 2.1. as was assumed. It seems therefore that defects (as knots) seem to be important in modifying tree strenght. Strenght is also an important modifier to wind and snow risk models, and even crucial to breakage models. Thus, further work deepening the understanding of the role of knots and other defects in terms of their effect on the modulus of rupture of standing trees and their consequent modification on damage risk is therefore needed. This addition work needed to support the models development/validation in Sub-tasks 3.1 and 3.3 has been described more in details in Appendix: Proposals for additional work to be needed to carried out under the STORMS project.

In addition, because the aspect of integration and communication of results through the Web has

gained a lot more importance than had been anticipated at the beginning, P03 has asked additional support to be able to keep same standards as the other partners while integrating their work in the final framework designed during the project. This addition work needed to keep same standards as the other partners in integration stage by P03 has been described more in details in Appendix: Proposals for additional work to be needed to carried out under the STORMS project.