Tree Pests & Diseases in the UK: Patterns, Drivers & Management

Editor’s recommendation:

This research review integrates more than 18,000 pest and disease records in the UK spanning 22 years, and uses an integrated species distribution model to reveal the spatial patterns and driving factors of pest and disease burdens on nine important host trees. The study found that broadleaf hotspots are concentrated in England, while areas of high conifer burden are more common in Scotland. Urban area, population density and local recreational activities are key factors driving broadleaf tree pests and diseases, while woodland connectivity, afforestation and coniferous forest coverage are the main driving factors for coniferous trees. The research provides a scientific basis for targeted monitoring and management of pests and diseases in British tree landscapes and decision-making on future tree species planting priorities.

introduction

Tree pests and diseases are an integral part of forest ecosystems, but their outbreaks can have significant impacts on forest biodiversity and ecosystem services. Global changes, including climate change, land use changes and increased global trade links, are significantly altering the distribution and dynamics of pests and diseases, leading to more frequent and severe outbreaks. In the UK, ash dieback, for example, caused by the fungus Hymenoscyphus fraxineus, is expected to cost the economy £14.8 billion over the next century and put 45 species that rely exclusively on ash trees at risk of extinction. Pest and disease risks are not uniformly distributed spatially, and understanding their fine-scale spatial heterogeneity, patterns, and drivers is critical to designing and implementing effective monitoring, intervention, and management strategies. Using a Bayesian integrated species distribution model, this study aimed to predict the spatial pattern of total pest and disease burden for nine important host tree species in mainland Britain and assess the effects of eight potential drivers to inform the design of plant health interventions.

method

data

The study used inspection records from the Animal and Plant Health Service (APHA) and public reporting records from the Forest Service’s Tree Health Diagnostic and Advisory Service (THDAS), totaling 18,871 pest and disease occurrence records spanning 22 years (2000-2022). APHA data are presence-absence data, recording whether pests and diseases are detected at a specific site during multiple visits; THDAS data are presence-background data. In addition, a range of potentially relevant environmental variables were collected, including area and edge length of woodland (all types, broadleaf, coniferous), urban/suburban land area, area of ​​ancient woodland, proportion of afforestation and deforestation, plant alpha diversity, canopy height, population density, road length, recreational demand (weekly and annual), distance to the nearest registered park/garden, distance to the nearest border check point (BCP), river and catchment information, vapor pressure deficit (VPD) and elevation. All data were processed into raster layers with a 1 km grid resolution and normalized. The analysis code is available at Zenodo.

Pattern of occurrence of pests and diseases

In order to predict the spatial pattern of pests and diseases, the study used the PointedSDMs package to fit Bayesian integrated species distribution models for nine host trees. ISDM combines presence-absence and presence-background data sets through a shared latent point process model (log-Gaussian Cox process). The model incorporates a Gaussian random field with a Matérn covariance function to account for spatial autocorrelation, and a second spatial random field for presence-background data to correct for sampling bias. The model was fitted by ensemble of nested Laplacian approximations, using regularized priors for fixed effects and random fields. Covariates were selected for each host tree by an information theory-based model selection method (using WAIC to compare binomial generalized linear models based on presence-absence data only). The final result is a predicted intensity map at 5 km resolution, with the intensity rescaled to the 0-1 range. In addition, we used forest subregion data to model predicted intensity as a function of host tree species cover and identify areas with higher or lower intensity than expected through residual plots.

Driving factors for pests and diseases

The study explores the impact of eight potential key drivers (recreation, urban area, population, afforestation, deforestation, distance to the nearest border checkpoint, coniferous forest area, woodland connectivity) on spatial patterns of pest and disease intensity in nine host tree species. The study adopts a graph-based causal inference method, which first encodes hypotheses about the data generation process in a directed acyclic graph that contains eight focal variables and six additional variables that may mediate or confound the focal effect. The DAG was then analyzed using the dagitty package to obtain the minimum set of adjustments required to identify the total effect of each focal variable. Subsequently, an ISDM is fitted separately for each focal variable, including the corresponding minimal adjustment set as covariates. In order to test the robustness of the inference results to the DAG structure assumption, we conducted a sensitivity analysis, constructed a series of intermediate structures from only the focal variables to the complete DAG, and ran all unique models implied by these structures. Simultaneously, the sensitivity of the model to the spatial grid resolution was also examined.

result

Pattern of occurrence of pests and diseases

Model predictions show that pest and disease intensity varies significantly spatially across mainland Britain. Several broadleaf tree species (A. pseudoplatanus, F. sylvatica, Q. robur, S. aucuparia) exhibit pest and disease hotspots in parts of England, particularly around London, Liverpool and Manchester. F. excelsior has hotspots in several areas of England, including a particularly large hotspot in East Anglia. In contrast, the predicted pest intensity of B. pendula is highest in south-west and north-east Scotland. Two conifer species (P. sitchensis and P. sylvestris) have large areas of high predicted intensity in Scotland and numerous smaller hotspots in England and Wales. Hotspots for P. abies are less obvious, with areas of highest predicted intensity in southern England, Wales and north-east and central Scotland. Forecast uncertainties are generally low but are higher in Scotland for some tree species. The sensitivity of prediction results to spatial grid resolution varied among tree species, with results for P. abies and Q. robur being the most sensitive to grid selection.

When host tree species cover is considered, predicted pest intensities for F. sylvatica, Q. robur and, to a lesser extent, F. excelsior were generally higher than expected in southern and eastern England and lower than expected in Scotland and Wales. P. abies showed a similar pattern. In contrast, P. sitchensis and P. sylvestris were present at higher than expected intensity across much of Scotland and parts of England and Wales. B. pendula has also been stronger than expected in parts of Scotland, but generally lower than expected in Wales.

Driving factors for pests and diseases

Based on the assumption of maximum DAG, the impact of each driving factor on different host trees is different. Weekly leisure activities have a relatively strong positive effect on the occurrence of pests and diseases of S. aucuparia, and a weak positive effect on F. sylvatica, F. excelsior and Q. robur. Annual leisure activities had no positive effects on any host tree species, but had negative effects on A. pseudoplatanus and P. sylvestris. Urban area has a positive effect on the occurrence of pests and diseases of five host tree species (all broadleaf trees). Population density has a clear positive effect on the occurrence of pests and diseases of B. pendula, F. sylvatica and P. sitchensis. Both afforestation and deforestation were positively correlated with the occurrence of diseases and insect pests in several host trees, and deforestation had a clear positive effect on all tree species except three host trees. Distance to border checkpoints has no clear effect on most host tree species, but the greater the distance, the lower the incidence of pests and diseases of P. abies (and to a lesser extent P. sylvestris and F. sylvatica). Coniferous forest coverage has a clear positive effect on three coniferous host tree species and one broad-leaved tree, A. pseudoplatanus. Forestland connectivity has a clear positive effect on the occurrence of pests and diseases of two host tree species, P. sitchensis and P. sylvestris.

Sensitivity analysis of the DAG structure shows that, with few exceptions, the inference results are generally insensitive to changes in modeling assumptions. However, the effect estimates of some driving factors are sensitive to the spatial grid resolution. When using coarser grids, weekly leisure activities have a strong negative effect on multiple tree species, while annual leisure activities have a clear positive effect on multiple tree species, and some other effect estimates have also undergone qualitative changes.

discuss

The study found that there are hotspots in England for predicted pest and disease intensity for several important broadleaf trees, which may coincide with major cities and is consistent with findings that urban size and population density are driving factors. Some hotspots (such as F. excelsior in East Anglia) may be the result of historical outbreak dynamics. The high intensity of the conifers P. sitchensis and P. sylvestris across large areas of Scotland may reflect the prevalence of coniferous forests in the region (including native pine forests of P. sylvestris). These findings provide a reference for understanding the overall distribution of pest and disease risk in UK tree landscapes and can be used to guide monitoring and management resource allocation for simulated host tree species, but decisions need to be made in conjunction with other factors (e.g. cost of surveys, effectiveness of different management options).

Urban area and population density are clear drivers of pest and disease intensity across multiple host tree species, related to factors such as the potential presence of more non-native hosts in urban areas, introduction of propagules via live plants, reduced natural enemies, physiological stress on trees, and management challenges. Given the potential damage of pests and diseases to urban trees and their ecosystem services, the findings highlight the importance of urban treescapes as a key target for monitoring and management.

Weekly fallowing, but not annual fallowing, was found to be a driver of pest and disease burden in multiple host tree species, suggesting that relatively frequent, local-scale activities may contribute more to total pest and disease burden than infrequent, long-distance activities. This may be due to greater local leisure activity, or to differences in gear cleaning behaviour. This finding suggests that customized biosecurity awareness campaigns targeting specific high-risk leisure activities may be an effective mitigation strategy. Note, however, that long-range propagation events are still important for invading new areas.

Woodland connectivity was positively correlated with pest and disease intensity in two coniferous species, P. sitchensis and P. sylvestris. The importance of connectivity in the spread of pests and diseases has been highlighted by previous studies. As these two tree species are mostly plantations in the UK and have historically been cultivated as monocultures, they may have only suffered the costs of increased connectivity (increased pest and disease propagule pressure) without reaping the protective effects of increased biodiversity. This suggests the need for future research to explore the interactions and potential trade-offs between forestland connectivity, biodiversity, and pest and disease outbreaks.

Afforestation and deforestation are associated with increased pest and disease burdens in several tree species, which may result from the introduction of propagules by mechanical equipment or the introduction of pathogens through live plants used in replanting and restoration activities. Although studies have accounted for sampling effort by incorporating presence-absence data and handling spatial bias in presence-background data, the possibility that effect estimates may be high due to differences in the ability of different forest users to identify pests and diseases cannot be completely ruled out. Therefore, these results need to be interpreted with caution, but investigating whether forestry activities contribute to the spread of pests and diseases in the UK remains an important future research direction.

This study demonstrates the practical benefits of integrated species distribution modeling that combines data from different sources while taking advantage of the geographic coverage of public reporting data and sampling effort information from plant health monitoring data. By using a more stringent penalized complexity prior and a regularization prior for fixed effects, the study avoids the overfitting problem that arises with spatial grid refinement reported in previous work.

In summary, the study revealed spatial variation in the total pest and disease burden across a range of important host tree species in the UK, with significant hotspots for some species that could be targeted for enhanced monitoring and management. Drivers of pest and disease intensity vary among hosts, but urban area, population density and recreational activity are important for several (mainly native broadleaf) tree species, and woodland connectivity is important for both conifers. Afforestation and deforestation are also associated with increased burden on several host trees, but these effects need to be interpreted with caution. The findings have implications for managing forests to cope with ongoing tree pest and disease threats and highlight the importance of incorporating landscape cover and connectivity metrics that reflect different pathways of propagule dispersal into predictive models. Finally, the study also illustrates the power of combining different species occurrence data using integrated species distribution models.

reward