Executive Summary

Main Findings - May 2022

All trials have now had at least 10 growing seasons and have recently been assessed for height, survival and Diameter at Breast Height (DBH). In several cases the year 10 assessment was not possible due to COVID related issues. The height of species within a site were compared by modelling growth in years since initial planting date. This approach was valid as all trees within a site were planted at the same time. To compare the height of species between sites growth was modelled according to tree age (time in weeks since planting). By examining the height of trees according to time since planting we were able to align the species at the same point in time across the sites to make more accurate comparisons of growth

Either aspen, hybrid larch or red alder are the two tallest species trees across all sites, though significance between them and other species is variable. Ash, sweet chestnut and sycamore are frequently the smallest species. Ash often has the lowest survival, most likely due to ash dieback. Low survivals for red and Italian alder such as at East Grange in Fife are an artefact of those species only being present in one of the four blocks as a result of supply issues when the sites were beaten up following heavy losses during the extremely cold winters of 2010 and 2011.

Diameter data has not yet been analysed but will be used to produce volume estimates thereby allowing an early comparison of productivity between species and sites. The next assessment for all of the SRF species trial sites will be after 15 growing seasons (2024-2026). Destructive sampling at that point will allow direct calculation of volume, wood density and calorific values.

The aspen clone trials have now had 6 growing seasons and will next be assessed at year 10 after the 2029 growing season. Clones 1, 2, 3, 5 and 6 have grown best at Fife whereas clones 1, 2, 3, 4 and 5 have grown least well at the Mull site. Clone 4 grew best at Clydesdale and clone 6 least well on the same site. Average survival was significantly lower for clones C1 and C3 and highest for the remaining clones with no significant difference between them. The low survival of clones 1 and 3 appears to be driven by particularly poor survival at Clydesdale and Angus.

No formal assessments are due in 2022-23 but each site will be inspected at least once to identify any necessary maintenance such as vegetation management or fence repairs.

Planning and Establishment

Planning and Establishment

Six trial sites have now been established in Scotland. These are all ex-agricultural sites with the exception of Aros on Mull which is a restock site, previously a Sitka spruce crop. At each of the six sites a fully replicated randomised block experiment (4 Blocks) was established trialling species likely to have fast early growth of high-density timber suitable for use in SRF. Ten species were originally planted (see Species code).

The experiment sites were fenced and ground preparation and weed control were carried out prior to planting. 4 blocks containing a plot of each species were planted giving a total of 40 plots. Species plots were 20 m x 20 m, planted at 1 m spacing along the rows and 2 m spacing between rows, giving 200 trees per plot. Assessments were carried out in the central 12 m x 15 m area containing 96 trees.

Note that following some issues at the Rumster site eight species were replanted at the site in May 2015.

Assessments

At the end of growing seasons 1-6 (November-December) assess Height in cm above ground level to the nearest 1.0 cm, and stem diameter. Assess again at the end of growing season 10 but measure DbH of all surviving trees in the central assessment plot and the top height of the 4 fattest trees. COVID and associated travel and staffing issues meant that these assessments were not always possible but all were completed in the winter of 2021-22.

Species code

• AH = Ash
• ASP = Aspen
• CAR = Common alder
• HL = Hybrid larch
• IAR = Italian alder
• RAR = Red alder
• SBI = Silver birch
• VPSS = Sitka spruce
• SC = Sweet chestnut
• SYC = Sycamore

Species code - Rumster

• Acps = Sycamore
• Alco = Italian alder
• Algl = Common alder
• Bepe = Silver birch
• Bepu = Downy birch
• Pisi = Sitka spruce
• Ptre1 = Aspen - Rackwick
• Ptre2 = Aspen - 403

Site Assessment Year

Statistical analyses

Methodology

Analysis was conducted in R (version 4.0.2, R Core Team 2020), with graphics produced using ggplot2 in R (Wickham, 2016) as follows:

Tree height was analysed across years 1-10 for each site separately. As trees were not individually tracked across assessments the mean height of each species in each block was calculated and used to analyse growth. Linear regression models were used with log transformed height as the outcome variable and block, species and year (numeric, 1-10) as fixed effect covariates. Two degree natural cubic splines were fit to year to allow for non-linear growth over time. An interaction between year and species was found to improve model fit (using AIC values) and was retained in the model.

As the sites had different planting dates the trees ranged in age from nearly 10 years (Clydesdale) to nearly 12 years (Angus) when height and survival was assessed at the beginning of 2022. To compare the height of trees between sites mixed linear models were implemented in the ‘nlme’ package (Pinheiro et al, 2020) which used tree age in weeks to align measurements across site. This model had log transformed height as the outcome variable and species, site and age in weeks, and their interaction, as fixed effect covariates. BlockSite was fit as a random effect. Natural cubic spines with three degrees of freedom were fit to allow for non-linear growth over time. A corARMA autocorrelation-moving average structure was fitted to the model (order p=1, q=0) to correct for autocorrelation in the residuals.

Survival was analysed at each site at year 10 using a generalized linear binomial model with a logit link function. Dead or missing trees were coded as ‘0’ and alive as ‘1’. Species and block were included as fixed effect covariates. Model fit was tested using the DHARMa package (Hartig, 2020) and if overdispersion was detected the glmmTMB package (Brooks, 2017) was used to model survival and observation level random effect added to the model.

Type-II Anova tests were used to evaluate the significance of main effects in all models. The ‘emmeans’ package (Lenth, 2021) was used to extract adjusted marginal means (back-transformed height and survival proportions). The cld function in the ‘multcomp’ package allowed pairwise comparisons of height to be made between all species and the Tukey method used to correct for multiple comparisons (Hothorn et al, 2008).

References

References

• R Core Team (2021). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

• H. Wickham. ggplot2: Elegant Graphics for Data Analysis. Springer-Verlag New York, 2016.

• Douglas Bates, Martin Maechler, Ben Bolker, Steve Walker (2015). Fitting Linear Mixed-Effects Models Using lme4. Journal of Statistical Software, 67(1), 1-48.doi:10.18637/jss.v067.i01.

• Pinheiro J, Bates D, DebRoy S, Sarkar D, R Core Team (2020). nlme: Linear and Nonlinear Mixed Effects Models_. R package version 3.1-150, <URL: https://CRAN.R-project.org/package=nlme>.

• Russell Lenth (2020). emmeans: Estimated Marginal Means, aka Least-Squares Means. R package version 1.5.1. https://CRAN.R-project.org/package=emmeans

• Torsten Hothorn, Frank Bretz and Peter Westfall (2008). Simultaneous Inference in General Parametric Models. Biometrical Journal 50(3), 346–363.

• Mollie E. Brooks, Kasper Kristensen, Koen J. van Benthem, Arni Magnusson, Casper W. Berg, Anders Nielsen, Hans J. Skaug, Martin Maechler and Benjamin M. Bolker (2017). glmmTMB Balances Speed and Flexibility Among Packages for Zero-inflated Generalized Linear Mixed Modeling. The R Journal, 9(2), 378-400.

• Florian Hartig (2020). DHARMa: Residual Diagnostics for Hierarchical (Multi-Level / Mixed) Regression Models. R package version 0.3.3.0. https://CRAN.R-project.org/package=DHARMa

Height - all years

Summary

• Growth at Mull is shown for all 10 species from year 1 to year 10

• Error bars represent 95% confidence intervals

• See ‘Height - Year 10’ for a statistical comparison of all species at the most recent time-point.

• Note that the year 10 assessment was delayed and that trees at Mull were closer to age 11 when the final assessment was made.

Height - Year 10

Summary

• Estimated heights are shown in cm and error bars represent 95% confidence intervals. The Anova table for the statistical model is shown in the supplemental material.

• Aspen (ASP) and larch (HL) were the tallest species at Mull and were significantly taller than silver birch (SBI), sycamore (SYC), sweet chestnut (SC) and ash (AH)

Survival - all years

Survival - Year 10

Summary

• Survival is shown as the estimated proportion of trees alive for each species. Error bars represent 95% confidence intervals. The Anova table for the statistical model is shown in the supplemental material.

• Dead and missing trees were grouped together and compared to the proportion of trees alive.

• Aspen (ASP) had the best survival and ash (AH) had the worst survival. Aspen survival was significantly better than ash (AH), larch (HL) and sitka spruce (VPSS).

Height - all years

Summary

• Growth at Fife is shown for all 10 species from year 1 to year 10

• Error bars represent 95% confidence intervals

• See ‘Height - Year 10’ for a statistical comparison of all species at the most recent time-point.

• Note that Year 1 height was not recorded at Fife and so the results shown are predictions from the statistical model based on growth across years 2-10.

• The year 10 assessment was delayed and therefore trees were closer to 12 years of age (~11.6) when the final assessment was made.

Height - Year 10

Summary

• Estimated heights are shown in cm and error bars represent 95% confidence intervals. The Anova table for the statistical model is shown in the supplemental material.

• Aspen (ASP) and Red Alder (RAR) were the tallest trees at Fife in Year 10 and were significantly taller than the majority of other species. Ash (AH), Sycamore (SYC) and Sweet Chestnut (SC) had the worst growth and were significantly smaller than most other species.

Survival - all years

Survival - Year 10

Summary

• Survival is shown as the estimated proportion of trees alive for each species. Error bars represent 95% confidence intervals. The Anova table for the statistical model is shown in the supplemental material.

• Dead and missing trees were grouped together and compared to the proportion of trees alive. Survival was not recorded for Year 1.

• Survival was good for the majority of species by year 10. The low survival of Italian (IAR) and Red (RAR) Alder is driven by the high levels of missingness recorded; however, these species were showing high levels of missingness from Year 2 onwards.

• By year 10 the majority of Ash (AH) trees had died and Sweet chestnut (SC) trees were also beginning to show signs of poor survival.

Height - all years

Summary

• Growth at Huntly is shown for all 10 species from year 1 to year 10

• Error bars represent 95% confidence intervals

• See ‘Height - Year 10’ for a statistical comparison of all species at the most recent time-point.

• Alder species were not planted until year 2 and so the height estimates shown for year 1 are projections from the statistical model based on the growth of Alder species from years 2-10.

• Note that the year 10 assessment was delayed and therefore trees were closer to 12 years of age (~11.8) at the final assessment.

Height - Year 10

Summary

• Estimated heights are shown in cm and error bars represent 95% confidence intervals. The Anova table for the statistical model is shown in the supplemental material.

• Height was similar for the tallest species (Red Alder (RAR) to Silver Birch (SBI)) and these species were significantly taller than the remaining species. Ash (AH) and Sweet chestnut (SC) were significantly smaller than all other species.

Survival - all years

Survival - Year 10

Summary

• Survival is shown as the estimated proportion of trees alive for each species. Error bars represent 95% confidence intervals. The Anova table for the statistical model is shown in the supplemental material.

• Dead and missing trees were grouped together and compared to the proportion of trees alive. Note that all Alder species were not available for planting in year 1 and are recorded as 100% missing.

• Sitka spruce (VPSS), Larch (HL) and Silver Birch (SBI) had significantly better survival at Huntly by year 10. Ash (AH) had significantly worse survival than all other species.

Height - all years

Summary

• Growth at Clydesdale is shown for all 10 species from year 0 to year 10.Year 0 is the planting year at Clydesdale and the time point when tree height was first recorded.

• Error bars represent 95% confidence intervals

• See ‘Height - Year 10’ for a statistical comparison of all species at the most recent time-point.

Height - Year 10

Summary

• Estimated heights are shown in cm and error bars represent 95% confidence intervals. The Anova table for the statistical model is shown in the supplemental material.

• Aspen (ASP) and Red Alder (RAR) were the tallest species at year 10, although not significantly taller than the majority of other species. Sweet chestnut (SC), Ash (AH) and Italian Alder (IAR) were the smallest species at Clydesdale.

Survival - all years

Survival - Year 10

Summary

• Survival is shown as the estimated proportion of trees alive for each species. Error bars represent 95% confidence intervals. The Anova table for the statistical model is shown in the supplemental material.

• Dead and missing trees were grouped together and compared to the proportion of trees alive.

• Sycamore (SYC), Larch (HL) and Sweet chestnut (SC) had the best survival at Clydesdale at year 10. Ash (AH) and Italian Alder (IAR) had the worst survival

Height - all years

Summary

• Growth at Angus is shown for all 10 species from year 1 to year 10.

• Error bars represent 95% confidence intervals

• See ‘Height - Year 10’ for a statistical comparison of all species at the most recent time-point.

• Note that the year 10 assessment was delayed and therefore trees were closer to 12 years of age when the final assessment was made (~11.9).

Height - Year 10

Summary

• Estimated heights are shown in cm and error bars represent 95% confidence intervals. The Anova table for the statistical model is shown in the supplemental material.

• Aspen (ASP) and Larch (HL) were the tallest species at year 10, although not significantly taller than the majority of other species. Sweet chestnut (SC) and Ash (AH) were the smallest species at Angus.

Survival - all years

Survival - Year 10

Summary

• Survival is shown as the estimated proportion of trees alive for each species. Error bars represent 95% confidence intervals. The Anova table for the statistical model is shown in the supplemental material.

• Dead and missing trees were grouped together and compared to the proportion of trees alive.

• Sycamore (SYC), Larch (HL) and Sitka (VPSS) had the best survival at Angus by year 10. Ash (AH) had the worst survival.

Height - all years

Summary

• Growth at Sibster is shown for all 8 species from planting year (0) to year 7.

• Error bars represent 95% confidence intervals

• See ‘Height - Year 7’ for a statistical comparison of all species at the most recent time-point.

Height - Year 7

Summary

• Estimated heights are shown in cm and error bars represent 95% confidence intervals. The Anova table for the statistical model is shown in the supplemental material.

• Common alder (Algl) was the tallest species at Rumster in year 7. The two aspen species (Ptre1 - Aspen Rackwich and Ptre2 - Aspen 403) were the smallest species at Rumster in year 7.

Survival all years

Survival - Year 7

Summary

• Survival is shown as the estimated proportion of trees alive for each species. Error bars represent 95% confidence intervals. The Anova table for the statistical model is shown in the supplemental material.

• Dead and missing trees were grouped together and compared to the proportion of trees alive.

• Sycamore (Acps), downy birch (Bepu) and sitka spruce (Pisi) were the species with the best survival at Rumster.

Height - all species

Summary

• Between-site differences in height were analysed for all species.

• As trees at different sites had different planting years they were not all the same age at the final time point (~10-12 years old). Tree age was converted to weeks (converted to years on the x-axis) and used to align tree heights across time between the sites.

• For sites with earlier planting dates (Angus, Huntly, Fife) the height at year 10 is a prediction from the statistical model based on what their height would have been 1-2 years prior to the final measurement.

Height - Tallest species age 10

Summary

• The height of the 5 tallest species are shown at each site. Error bars represent 95% confidence intervals and the lettering represents groupings based on statistical significance.

• There were no statistical differences between sites in height at age 10 for red alder (RAR).

• Aspen (ASP) grew better at Fife and Clydesdale compared to Huntly and Angus.

• The worst growth for sitka spruce (VPSS) was observed at Mull.

• Although Italian red alder was nominally tallest at Fife there were large confidence intervals around this estimate and there were no statistical differences between sites except when comparing Fife and Huntly.

• Common alder grew better at Fife and Clydesdale when compared to Huntly and Angus

Height - Smallest species age 10

Summary

• The height of the 5 smallest species are shown at each site. Error bars represent 95% confidence intervals and the lettering represents groupings based on statistical significance.

• There were few significant between site differences between larch (HL) although trees at Fife were significantly larger than those at Huntly.

• Silver birch (SBI) grew better at Clydesdale compared to Angus and Mull.

• Sycamore grew better at Clydesdale compared to Huntly and Angus.

• Ash (AH) grew better at Clydesdale and Angus compared to all other sites and poor growth was observed for sweet chestnut (SC) at Huntly.

Height - all years

Summary

• The growth of the aspen clones is shown across all years for each site. Error bars represent 95% confidence intervals. Estimated heights are shown in cm and the Anova table for the statistical model is shown in the supplemental material.

Height - clone differences

Summary

• The estimated height for each clone averaged across all sites is shown. Error bars represent 95% confidence intervals and lettering denotes statistical groupings.

• Clone 6 was the tallest clone across all sites and clones 1 and 3 were the smallest by year 6.

Height - between site differences

Summary

• There were few robust between site differences in growth.

• Clone 6 trees were taller at Fife and clones 1,2 and 4 appeared to be smaller at Mull compared to other sites.

• Whilst clone 6 grew well at Fife height was significantly smaller at Clydesdale for this clone compared to the majority of other sites.

Survival - all years

Survival - Year 6

Summary

• Average survival across all plots is shown in the left hand figure. Clone 3 and clone 1 had significantly worse survival but there were no statstically significant differences in survival between the remaining clones.

• The right hand figure shows survival for each clone at each site. The poor survival for clone 3 and clone 1 appear to be driven by particularly low survival at Clydesdale and Angus.

Data Tables