Insights from a wind-risk modelling workshop

Forest Research and IUFRO joint conference and training workshop for the ForestGALES wind-risk modelling package

A blog post from our researcher, Dr Jon McCalmont, who participated in a conference and training workshop to find out more about the ForestGALES wind risk-modelling package.

Organised by Forest Research, one of our research partners and IUFRO at the Forest Research, Northern Research Station, Edinburgh, the event focused on improving our understanding of the risks posed by wind to forests and woodlands in the UK – increasingly important in a changing climate where we seem to be experiencing more extreme weather, and vital to factor in to efforts to increase tree cover as part of a solution to meeting UK net zero commitments. Find out more of Jon’s experience in the blog post below:

Day 1:

Day 1 of the meeting started on Monday afternoon, to give attendees time to arrive in Edinburgh and get settled into their accommodation. We began with an ‘ice breaker’ walking tour of the city centre, after meeting up under the Duke of Wellington statue on Prince’s Street. This gave those of us who were new to the wind-risk modelling community the opportunity to meet informally and get to know each other, as well as for old friends to catch-up. We were treated to a very interesting tour of the city centre, with those who live locally providing historical and geographic insights. Our tour included climbing Calton hill to take in the panoramic views before finishing off with an evening meal of the “best fish and chips in the world” at Bertie’s Fish and Chip bar. The ice breaker worked very well with conversations flowing freely and new acquaintances made ahead of the formal start of the meeting the following day.

Day 2:

The workshop/conference itself kicked off early on Tuesday morning at the Forest Research Northern Research Station (NRS) in Roslin, just to the south of Edinburgh. Coffee and biscuits were provided from 08:30 with talks beginning from 09:00.  After a welcome from the Chief Executive of Forest Research, James Pendlebury, and an introduction to IUFRO (the International Union of Forest Research Organizations) from Dirk Schindler, their coordinator for ‘Impact of the Wind on Trees’, we moved on to hearing about the development of forest wind risk modelling in the UK. This ranged from the Wind Throw Hazard Classification of the 1980s to the subject of this workshop, the process based ForestGALES model, now provided on request as a package within the R coding framework.

It was not until the 1950s that foresters began to appreciate how much of a problem wind was going to be for large-scale timber production in the UK, as plantations established following the depletions of WW1 matured. Perhaps not so surprising in hindsight though, since we are one of the windiest countries in the world. A major storm in 1963 reinforced these concerns when it flattened around 25% of plantations in the Northwest of the UK, spurring efforts to better understand these risks. Wind risk is now recognised as the primary limiting factor to tree growth in this country.

But of course this isn’t an issue unique to the UK; during the first day’s talks we heard how, in 1990, storm damage brought down more weight of timber in Germany than the sum of all scheduled fellings for that year and how, on Alpine slopes in Italy in 2018, Storm Vaia felled more than 15 million trees across 41,000 hectares. This international concern for wind risk to forests was reflected in the wide range of national research represented at the meeting, including the UK, Norway, France, Italy, Germany and even Madagascar.

The morning of the first day focussed on the history of wind risk model development and we heard fascinating talks about some of the practical work that went into parameterising factors like root anchorage and stem resistance to breaking. Hugely labour-intensive studies uprooting hundreds of trees (hopefully those already scheduled for felling!) under as wide a range of soil conditions as possible, coupled to laboratory testing of the efforts needed to subsequently break those stems under load.

A general discussion session to close the morning was followed by lunch, and then the afternoon was spent with our laptops in a guided practical session applying the newly released R language package of the ForestGALES model (fgr). Earlier in the day, we had been provided with a set of materials to use in running the model: including a pen-drive with some example datasets, code scripts and a very useful accompanying booklet to help guide us through the examples. The aim in the first afternoon was to work through a simple application of the fgr model in one of two primary ‘run-modes’. This first one (the canopy roughness method) is better suited to homogenous, even aged stands and struck me as perhaps being particularly suitable to the kind of scenario testing that we might find useful in the NETZERO+ project.  ForestGales calculates the wind speeds at which particular species, in different soils, are likely to sustain damage and whether that would be stem breakage or uprooting. These values are referred to as Critical Wind Speeds (CWS) with the model well parameterised for more than 20 tree species across a range of soil types and potential rooting depths. It combines aerodynamic modelling with canopy structure and even gravitational effects, a key parameter growing in significance as a tree leans further from the vertical.

The first day was rounded off in fine style with a superb conference dinner at the Twenty Prince’s Street Restaurant in the centre of Edinburgh.

Day 3:

The second day started with talks based around the integration of local wind data with critical wind speed thresholds to estimate the probability of trees being damaged in specific locations. These calculations rely on information about not only average wind speeds but the distribution of their extremes, i.e. their variability, or ‘gustiness’. In ForestGALES this wind speed distribution for a given period is described by inputting ‘Weibull’ parameters: Weibull A, which describes the total amount of wind over time, and K which indicates how much of this wind will be found at the extreme end (i.e. in more intense gusts). Previous iterations of the model have relied on observation data (a tatter flag network) to estimate a small-scale resolution of mean wind speed from larger-scale wind datasets, from which Weibull A can be derived; but these had been confined to a single generic value for variability (Weibull K). In the first of the days’ talks we heard about new developments in downscaled atmospheric modelling that have produced very high-resolution mapping (250 x 250m) for A and K across the globe, allowing much more detailed parameterisation of ForestGALES modelling. GIS maps of these across the UK were provided with the course materials for us to experiment with. Again, the potential to combine these maps of wind variability with derivations of wind magnitude spatially from climate predictions across the UK grid struck me as of great potential interest for NetZeroPlus and inspired me to start thinking about how we might code a framework to apply this within the project across our available soil and climate datasets.

After coffee, we heard from two projects utilising ForestGALES in ‘real-world’ research. The first came from Norway, where advantage was taken of a nationwide, very high-resolution LIDAR dataset across their entire forested area.  This research employed the more sophisticated TMC (turning moment coefficient) run mode in fgr to estimate wind risk across more than 200 million grid cells (16m x 16m resolution) down to the individual tree level and from four different wind directions.  This was far beyond what the R package would be able to do on its own, purely due to computing time, so it was fascinating to hear how this was overcome using a combination of R and the C++ compiled language. The second talk presented a case study example recently published by one of the original developers of ForestGALES, Prof. Barry Gardiner, where he had worked with the rail company Deutsche Bahn in Germany to calculate wind risk to trees along their railway lines. This was a particularly interesting example of combining GIS and LIDAR data to build a very high-resolution map of the distribution of trees and associated gaps between them in all cardinal directions and applying fgr across the resulting data. Such maps of critical wind speeds in different directions might be used to assess the likely impact of upcoming storm tracks from weather forecasts, and direct clean-up resources towards rail lines most at risk from falling trees. Applying the fgr model in these situations is still a very involved process combining different modelling tools (i.e. GIS information, LIDAR data, tree distribution and weather data), work is ongoing in the ForestGALES team, and the wider wind-risk modelling community, to better integrate these disparate datasets.

After another general discussion around points raised in the talks and a break for lunch, we returned to the practical training sessions. This time running an example of the much more detailed TMC mode in fgr, as demonstrated in the morning’s talks. For our example we were provided with detailed maps of the Forest of Dean and calculated wind risk across the forest. We were also provided with an example of the same model run but using the open-source GIS software (QGIS) rather than R code scripts directly, as a demonstration of how the model can be employed outside of the R programming environment for those less comfortable with coding.

Day 4:

Talks and training were now complete, and the final day was given over to a visit to the Queen Elizabeth Forest Park in the beautiful Trossachs National Park. The day started early, meeting at the Northern Research Centre for 07:30 and departing for a 90-minute coach ride to the Forest Park’s visitor centre, before splitting up into two minibuses for a tour of the forest. The focus of this visit was on a decades long project to convert around 10% of this region of Scottish forestry to a more ‘natural’ state of continuous cover through targeted thinning, with serious consideration given to the impact of this on future wind risk to the remaining trees. This visit was of particular interest to me since, in a previous career in forestry, I had worked in this very area 30 years ago felling and extracting trees. Having no real grasp of the bigger picture at the time I hadn’t realised the role I would be playing in helping to shape this region into what now has to be one of the most beautiful evergreen forests in the UK.

The remit of this area, and UK forestry plantations in general, has moved far beyond just the production of timber and now encompasses goals of amenity, biodiversity, and environmental resilience. And this specific area of the region demonstrates this wonderfully, incorporating a network of walking and cycling trails with viewpoints and hides to observe the  abundant wildlife, including red squirrels (which we saw through the mini-bus windows as we arrived) and pine martens. Deer are plentiful too, a little too much so and we heard of the damage they can present due to overabundance, this is managed here by the Park’s own rangers culling them and supplying the meat into the market. The day finished with a talk by the Forest Senior Manager, John Hair, at the visitor centre around their management goals and how they accommodate wind risk in their landscape planning. We finished the day, and the workshop, with a general discussion among the attendees around future challenges and opportunities for wind risk modelling, and forestry in general, before boarding our coach and heading back to the Research Station at Roslin.