Research Article |
Corresponding author: Manuela Branco ( mrbranco@isa.ulisboa.pt ) Academic editor: Jianghua Sun
© 2019 Manuela Branco, Pedro Nunes, Alain Roques, Maria Rosário Fernandes, Christophe Orazio, Hervé Jactel.
This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Citation:
Branco M, Nunes P, Roques A, Fernandes MR, Orazio C, Jactel H (2019) Urban trees facilitate the establishment of non-native forest insects. NeoBiota 52: 25-46. https://doi.org/10.3897/neobiota.52.36358
|
Cities, due to the presence of ports and airports and the high diversity of trees in streets, parks, and gardens, may play an important role for the introduction of invasive forest pests. We hypothesize that areas of urban forest facilitate the establishment of non-native forest pests. Based on scientific literature and a pan-European database on non-native species feeding on woody plants, we analysed where the first detections occurred in European countries. We collected site data for 137 first detections in Europe and 508 first European country-specific records. We also estimated the percentage of tree cover and suitable habitat (green areas with trees) in buffers around detection points. The large majority of first records (89% for first record in Europe and 88% for first records in a European country) were found in cities or suburban areas. Only 7% of the cases were in forests far from cities. The probability of occurrence decreased sharply with distance from the city. The probability to be detected in urban areas was higher for sap feeders, gall makers, and seed or fruit feeders (>90%) than for bark and wood borers (81%). Detection sites in cities were highly diverse, including public parks, street trees, university campus, arboreta, zoos, and botanical gardens. The average proportion of suitable habitat was less than 10% in urban areas where the species were detected. Further, more than 72% of the cases occurred in sites with less than 20% of tree cover. Hotspots of first detection were identified along the coastal regions of the Mediterranean and Atlantic, and near industrial areas of central Europe. We conclude that urban trees are main facilitators for the establishment of non-native forest pests, and that cities should thus be intensely surveyed. Moreover, as urban areas are highly populated, the involvement of citizens is highly recommended.
Cities, forest pests, invasive species, surveillance, urban trees
Forests, like other terrestrial ecosystems, are increasingly threatened by the establishment and spread of non-native pests worldwide (
The increase of forest pest invasions in recent years is mostly the consequence of increasing global trade and international travel (
The main pathways for the accidental introduction of non-native forest insects are the trade of live trees for planting, hitchhiking with containers, imports of timber, and use of wood packaging material (
There is little doubt that urbanization and population concentration in large cities are of critical importance for the arrival rate of invasive species (
Most cities usually harbor a relatively high diversity of tree species, dispersed in many urban parks and gardens. These could provide a larger number of putative hosts and, thus, increase the risk of non-native pest establishment (
To test these hypotheses, we compared the rate of first detection in urban vs non-urban areas of invasive pests feeding on woody plants in Europe. We focused on this continent because it is among the most affected by forest pest invasions in the world, and we have very good records of non-native species detection in European countries (
Data sources. We first used the pan-European database for non-native organisms, DAISIE (Delivering Non-native Invasive Species Inventories for Europe) (
Criteria for data selection. Only insect species introduced in Europe since 1950 were considered because most of the forest non-native species in Europe were introduced in the last five decades (
For each country, a unique first record was retrieved with the exception of a particular species detected in geographically well separated regions of a given country (e.g. mainland and distant islands). A case study was thus defined as a new species detected in a new country for the first time. The same introduced species could be reported several times in Europe as long as it was successively recorded for the first time in different European countries. We further distinguished between first record in Europe and first records in any of the European regions as a given species could use the first introduction as bridgehead for spreading through Europe, or be introduced several times in different European regions. Rare cases where two different European countries reported the first detection in the same year were both accounted as first records in Europe.
Complementary data. For each case study (one species × one country × one date of first detection) we documented information regarding the insect species, and the time and location of first record. For each species we retrieved from the literature the order, family, feeding guild (Bark & wood borers, Defoliators, Sap suckers, Gall makers, Root feeders, Seed and fruit feeders), host range (Broadleaves, Conifers, Palms, Polyphagous), and body length (mm). For each first detection we recorded the year of detection, the geographical coordinates of the site, the type of habitat (Urban if reported in a city or suburban urbanized areas, Forests, Rural areas, or Nurseries), and the distance in km to the border of the nearest city (with at least 10 000 inhabitants) or large city (at least 100 000 inhabitants or with an international transport system, e.g. airport, seaport, railway station).
The site coordinates were retrieved from the reporting articles. In several cases, exact site coordinates were not available. When the description of the location was reliable and narrow enough to delimitate a location area (e.g. Lisbon Zoo, Nepliget Park in Budapest), its central point was used to recalculate site coordinates. For records that had inaccurate location but with some useful geographical information, e.g. “in the town of Rome” (
The Euclidean (straight-line) distance between the detection location and the external limit of the nearest city and nearest large city were calculated using ArcGIS 10.5 (ESRI, Redlands, CA, USA). Cities were visually identified using ArcGIS Online World Imagery map (Copyright ESRI). The distances were then reclassified in distance classes (×10 km).
Hotspot analysis was performed with Getis-Ord GI* spatial statistics (
Population size of cities near detection points was retrieved from the online Wikipedia encyclopedia. For each detection point the population density, i.e. inhabitants per square kilometer, by NUTS 2 region (Nomenclature of Territorial Units for Statistics, level 2) was obtained from Eurostat (https://ec.europa.eu/eurostat/web/products-datasets/product?code=tgs00024, assessed on 2019-8-22). The year 2015 was used as reference. We then estimated the ratio between the average population density in the NUTS 2 region where the detection point was located and the population density of the respective country. For the analysis, we considered countries with two or more NUTS 2 regions.
Forest cover and suitable habitat around the point of detection. The percentage of forest cover was estimated using the Tree Cover Density (TCD) of the Copernicus Land Monitoring Service – High Resolution Layer Forest (https://land.copernicus.eu/pan-european/high-resolution-layers/forests/view, 2012). TCD provides continuous-scale information on the proportional crown coverage (0–100%) detected per pixel (20 m of spatial resolution) at the European scale, including the following Land Use – Land Cover (LULC) classes: evergreen and deciduous broadleaved, sclerophyllous and coniferous trees, orchards, olive groves, fruit and other tree plantations, agro-forestry areas, transitional woodlands, forests in regeneration, groups of trees within urban areas. The percentage of forest cover was calculated in four buffers (100, 500, 1000, and 5000 m of radius) around the detection points with precise geographical coordinates. The Set Null function was used to remove the no-data values from the databases.
Complementarily, to test the hypothesis that the percentage of tree cover within 100 m around the detection point was similar to that of any other 100 m radius buffer in the surrounding area, we randomly created three additional 100 m radius sampling areas within the 5000 m buffer area. These sampling areas were generated with the constraints of its central point being at least 200 m far from the central detection point and 200 m far from the other two random sampled areas. Additionally, any randomly created central point that was located in the ocean or inland water surface was manually removed and replaced. The tree cover around each of the three random central points was calculated in the same manner as around the central detection point.
To estimate the percentage cover of suitable habitats in large cities we used the Urban Atlas database from 2012, from the Copernicus Land Monitoring Service (https://land.copernicus.eu/local/urban-atlas). Urban Atlas delivers pan-European comparable LULC data for Functional Urban Areas, i.e. city and its commuting zone (
To estimate the probability of detection in function of the classes of distance to the nearest city we used generalized linear models (GLM) with Binomial distribution and log link function. A model was applied to each feeding guild separately. A Gaussian GLM with log link function was further used to test the effect of body size on the distance to the nearest small city and large city. We also used Gaussian GLM to test temporal trends in detection years and distance of the detection points to the nearest city and nearest large city.
Paired t-test statistic was used to compare the percentage tree cover in the 100, 500, 1000, and 5000 m radius buffer. Paired t-test statistic was also used to compare the percentage of tree cover in the buffer area (100 m radius) around the detection point and the mean percentage of tree cover in the three buffer areas (100 m radius) sampled at random within the same 5000 m buffer area. One-way ANOVA was used to compare the percentage of tree cover in the buffer area (100 m radius) per feeding guilds. The relationship between the average population density per NUTS 2 and the country population density was tested by paired t-test statistics.
We retrieved data from 133 non-native insect species, belonging to six feeding guilds. Sap feeders (order Hemiptera) were the most represented guild (40% of the cases), followed by bark beetles and woodborers (29%). Defoliators, gall makers, and seed and fruit feeders represented 14%, 10%, and 7% respectively. Only one species was a root feeder, which was not used for comparisons between feeding guilds due to its low representativeness.
In total 508 first country-specific records were retrieved, from 38 regions (including mainland and separated islands) and 25 countries (Suppl. material
From our data, 137 cases were first records for Europe (mainland and islands). Italy registered the highest number of first records in Europe (36), followed by Spain (19), France (18), and Portugal (10). Eight first detections were made in islands of the Mediterranean Sea (Sicily, Corsica, and Balearics) or the Atlantic Ocean (the Canaries, Madeira, and the Azores). UK and Germany had intermediate values of 8 and 7, respectively. All other cases were distributed among 19 other countries.
The hotspot analysis of first records in Europe revealed an uneven distribution at the European scale. Several hotspot areas with a Getis-Ord Gi* Z-score greater than 3.80 (p-value < 0.01) were identified in continental Europe (Fig.
Hotspots map of first detection points in Europe of non-native insects feeding on woody plants, recorded since 1950. The Getis-Ord GI* (GiZ Scores) are provided to indicate different levels of clustering of either high values (Z-score positive, hotspot) or low values (Z-score negative, coldspot). The respective p-values are: Z Score > 3.8, p-value < 0.01; ZScore [3.2, 3.8], p-value < 0.05; ZScore [3.0, 3.2], p-value < 0. 1).
About 64% of first records in Europe occurred in large cities and 89% in cities or their suburban areas. Similarly, 62% of country specific detections were reported in large cities and 88% in cities or their suburban areas (i.e. within 10 km distance from their limit). The probability of first detection decreased sharply with distance from the nearest city or large city (Fig.
Probability (mean ± SE) of first detection of non-native insect feeding on woody plants in Europe (estimated by GLM) in function of distance class (in 10 km) to a) the nearest city and b) the nearest large city.
In 69% of the cases the population of the nearest city to the detection point, i.e. located within a 20 km distance, was over 100 000 people, and in 35% of the cases above 500 000 people (Fig.
Percentage and number of first detections of non-native forest insect species by class of city population (up to 20 km distance).
Urban habitat was the most frequently observed land cover type around first detection points, accounting for 74% of the cases. Urban habitats reported were highly diverse, including schoolyards, university campuses and experimental stations, trees in airport and port areas, railway stations, industrial areas, urban arboreta, botanical gardens, public parks, zoos, and street trees. Arboreta, botanical gardens, gardens, and urban parks were the most often reported cases in cities (60% of the cases with site information). Only 11% of the cases were found in forest habitats. In 4% of the cases these forests were close to cities (i.e. at less than 10 km), while the other 7% were in forests far from cities. Other cases were reported in nurseries (4%) and rural landscapes (11%).
The percentage of first detection in the urban habitat significantly varied with the insect feeding guild (Chi2 = 19.519; p < 0.001). Sap suckers, gall makers, and seed and fruit seeders were more frequently found in urban habitats, 80%, 78%, and 81%, respectively, than defoliators (69%) and bark & wood borers (58%) (Fig.
Insects feeding on broadleaves were more frequently found for the first time in urban habitat (76%) than species feeding on conifers (59%). Still, the difference was not significant (chi2 = 1.130, p = 0.288). Polyphagous species, feeding on both conifers and broadleaves, were reported in nine cases only, but six of these cases (67%) were also in urban areas.
Temporal trend shows an exponential increase in the number of first records with decade, with a steep increment since the 1990s (Fig.
The mean percentage of tree cover at 100 and 500 m around the detection point was 17.1% ± 1.3 and 17.2% ± 1.1, respectively, ranging from 1 to 85%, with no differences between these two buffer sizes (t-test = 0.158, df = 307, p = 0.875). Considering these buffer radii, 55% of the detection points were in sites with only 10% or less of tree cover, and in 73% of the cases in sites with less than 20% of tree cover. However, within a buffer of 1000 m radius around the detection point, the mean proportion of tree cover was significantly higher (31.8% ± 1.9) than at 100 m (t-test = 14.6, df = 307, p < 0.001). Again, at 5000 m radius buffer size, the proportion of tree cover was higher (35.4% ± 1.0) than at 100 m (t-test = 14.5, df = 307, p < 0.001). Within the largest buffer radii, i.e. 1000 m and 5000 m around the detection point, there were no difference among feeding guilds for the percentage of tree cover (F4,300 = 2.179, p = 0.071, and F4,300 = 1.928, p = 0.106, respectively for 1 km and 5 km). However, at 100 and 500 m radius, we found differences among feeding guilds for the proportion of tree cover around the detection point (F4,300 = 3.065, p = 0.017 and F4,300 = 3.132, p = 0.015, respectively for 100 and 500 m). In both cases, defoliators tend to occur in sites with higher percentage of tree cover (which was 27% and 25%, respectively for 100 and 500) than for other feeding guilds.
A complementary analysis concerning the estimation of the percentage cover of suitable habitats (following five LULC classes: Green Urban Areas, Sports and Leisure Facilities, Orchards, Forests, and Herbaceous vegetation associations) in urban areas of large cities was conducted in 94 cases. The proportion of suitable habitat was on average 9.7% ± 1.1, 9.0% ± 0.6, and 9.0% ± 0.6, for 500 m, 1000 m and 5000 m buffer radius, respectively. There were no significant differences between buffer sizes.
The comparison with surrounding landscape showed that the percentage of tree cover within a 100 m buffer radius around the detection point (focal point) was slightly, but significantly (t-test mean difference = −4.937 ± 1.471, p = 0.001), lower than tree cover in three buffer areas of the same radius randomly sampled within a distance of 5000 m (17% ± 1.3 vs 22% ± 1.2).
Using European data on first detection records of non-native insect species feeding on woody plants since 1950, we could confirm the trend for an exponential increase with time. However, the most striking outcome of the survey is that 88% of first detections were made in cities and, for the majority, in large cities (62% in total, 70% of urban records). Moreover, the number of detections decreased dramatically in the first 10 km outside the city (Fig.
The proximity of main transport facilities (e.g. airports and ports) and the high density of people make cities under high propagule pressure, i.e. high frequencies of introductions of non-native organisms, plants or animals (
Several records specifically documented first occurrences in urban areas near transport facilities and could identify the pathways. For example, the first infestation of A. glabripennis in the Netherlands, in 2010, was found on native host plants, in an industrial area in the city of Almere, and was related to pallets used for transport of industrial machinery (
We may argue that first occurrences occurred mostly in urban areas because more researchers are living in these areas and are, thus, more likely to detect recently introduced forest pests. In some cases, researchers found new records within their own faculty campus (e.g.
Arrival does not necessarily imply successful establishment of introduced species. The establishment of a species in a new area further needs suitable habitat and resources, depending on its ecological niche. In fact, it is estimated that only a minor proportion of new arrivals results in successful establishment in a new region (
However, we did find some differences in habitat requirements according to feeding guild. Defoliators seemed to be more demanding in terms of the presence of a certain density of tree cover. Bark and wood borers were more frequently detected outside urban areas than other guilds. On the other extreme, gall makers, sap suckers, and seed and fruit feeders were mostly found (more than 90% of the cases) in urban areas. In some cases, these species occurred in circumstances in which only a small number of trees of a specific host was present. For example, the invasive gall wasp Epichrysocharis burwelli, which is known to form galls only on the lemon-scented gum, Corymbia citriodora (Myrtacea), was found in an urban park and the Zoo in Lisbon, where only a few host trees of that particular host species were present (
Cities may facilitate the establishment of tree pests because of their large diversity of tree genera and species, giving non-native pests a better chance of finding a suitable host tree (
In several cases, species were found near cities, i.e. in suburban areas. These areas are often characterized by heterogeneous landscapes, where gardens, orchards, forest fragments, and many rural habitats are present and tree abundance and diversity is greatly increased. In fact, the proportion of tree cover increased from 17% in urban areas around detection points to 32% and 35% in buffers of 1000 m and 5000 m radius, respectively. Thus, suburban areas could further facilitate the establishment of non-native forest pests. In only 4% of the cases (20 out of 508), non-native species were detected in nurseries. For these particular cases, detections occurred probably before establishment, which would also facilitate eradication attempts.
Cities may also offer better conditions for non-native species establishment due to their more suitable climate, in particular warmer temperatures resulting from the heat island effect (
At a larger spatial scale, hotspot analyses clearly showed a clustered pattern of first detection records in Europe. Most hotspots for the first detection of non-native forest pests were found along the coastal regions of Europe, from the Mediterranean coast of Italy, France, and Spain to the Atlantic coast, from Portugal to the Netherlands (Fig.
Still, two other hotspots of first detections were identified, one in Central Europe, from southern Germany, to northern Italy, and the other in Eastern Europe. These areas coincide with intense industrial regions and a number of river ports. Their proximity to Middle East and Asia, from where more than 40% of the non-native species from our study originate, may further suggest a pathway of progression from eastern regions. Some particular areas may also reflect a concentration of forest entomologists, but this is difficult to verify.
The economic impact of invasive forest insect pests is huge on both forest and urban environments (
The study was part of the PLURIFOR project, EU INTERREG SUDOE, and of the HOMED project, which received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement no. 771271. This study received backing from Forest Research Center (CEF). CEF is a research unit funded by Foundation for Science and Technology (FCT), Portugal (UID/AGR/00239/2019). M.R. Fernandes was supported by national funds via the FCT, under “Norma Transitória–DL 57/2016/CP1382/CT0019”. P. Nunes was supported by SUSFOR (PD/00157/2012) doctoral grant from FCT (PD/BD/142960/2018).
Table S1. Data on first records by species and countries
Data type: species data