Although nuclear power plants are often believed to provide an attractive option for mitigating climate change [1, 2], accidents at these plants can result in the long-term evacuation of residents over a wide area. A serious nuclear accident at the Fukushima Daiichi power plants in 2011, which occurred in an area with a higher human population density than that surrounding Chernobyl [3, 4], was followed by the evacuation of about 81,000 people [5]. According to estimates made by the national government of Japan, it will take more than 20 years of natural decay for the radiation to reach levels that will permit a complete return of the residents [6]. Decontamination and reconstruction of the evacuation zone are necessary as soon as possible in such cases, and the direction of such reconstruction must be carefully considered. Historically, most of the evacuation zone in Fukushima was covered by forests and farmlands that were traditionally managed [7], and it was assumed that the residents benefitted from biodiversity and ecosystem services in a sustainable manner. In such situations, healthy biodiversity and ecosystem services are highly important components of successful reconstruction and should be appropriately monitored and evaluated.

It has been hypothesised that not only radiation, which has often been the focus of previous studies in Fukushima and Chernobyl [8–11], but also the cessation of the disturbances caused by anthropogenic activities such as farming and gardening seriously affect the biodiversity and ecosystem services in evacuation zones. A meta-analysis of farmland abandonment suggests that this phenomenon has frequently had a negative effect on biodiversity in Africa, Asia, and Europe [12], where the history of anthropogenic land use is relatively long. In Japan, traditional farming that is subject to intermediate disturbances has formed highly heterogeneous landscapes; these have been reported to support high levels of biodiversity and sustainable ecosystem services [13, 14].

Among the components of biodiversity, some taxonomic groups, such as insects, are relatively tolerant of radiation exposure [15–17], and serious effects of exposure to radiation within the Fukushima evacuation zone are unlikely to be found at the population level [11]. A previous study that surveyed the northwestern area of the Fukushima nuclear power plants immediately (6 months) after the accident suggested that the reduction in the abundance of insects exposed to radiation was less obvious than that it was at Chernobyl [10]. However, studies in farmland areas have suggested that insects are sensitive to changes in land use [14, 18] and that these effects can become more obvious with time. Considering the fact that insects play an important role in ecosystem functions and serve as pollinators, pests, and prey for upper trophic levels, such as birds and mammals, it is critical to effectively monitor and evaluate their populations in the Fukushima evacuation area to guide effective reconstruction.

In this study, we investigated differences in the abundances of insects inside and outside the Fukushima evacuation zone (Fig 1) 3 years after the nuclear accident. The data from this study are expected to provide a valuable contribution to efforts to monitor and evaluate biodiversity and ecosystem services in the evacuation zones resulting from nuclear power plant disasters. Specifically, we focused on flying insects, including important pollinators such as Apidae, and medically important pests, such as flies and hoverflies. We covered broad areas and examined taxonomic groups that had not been previously investigated in this region but that are nonetheless important for the provision of ecosystem services.

Through sampling insects inside and outside the Fukushima evacuation zone in early summer, we found that the number of individuals of Xylocopa appendiculata was remarkably lower inside the evacuation zone, whereas those of other insects were not. In more detail, a total of 48,081 insects and spiders in 47 sampling sites were sampled using Malaise traps set from mid-May to mid-July 2014. Most were Hymenoptera and Diptera (16,583 and 20,082, respectively). The mean±SD per site of recorded individuals in each taxonomic and caste group are shown in S1 Table. In total, 46 taxonomic and caste groups (6 orders, 10 families, butterflies, moths, 12 species and 5 genera of bees and wasps, total wing ants, total workers of ants, workers of 8 ant species including a polygynous species, and queens of 1 ant species) and the species richness of Apidae were analysed (Tables 1–3; Figs 2–5; S2 Table, S1 Data).

Bayesian statistical analysis was used to assess the effect of evacuation on each taxonomic group. In the analysis of Xylocopa appendiculata, a species of carpenter bee of the Apidae family, the upper limit of the 95% credible interval (CI) regression coefficient for the evacuation effect, b₀, was <0 because few individuals were captured in the evacuation zone (Table 1; Fig 2D). Although the European honey bee Apis mellifica was captured in the evacuation zone, the number of sites in which they were present was too small (three points) to construct a statistical model. In contrast, Ceratina flavipes and Nomada spp. showed remarkable positive effects from the evacuation (the 95% lower limits of CI of their coefficients for b₀ exceeded 0) (Table 1; Fig 2B and 2E). The species richness of Apidae showed no remarkable effects from the evacuation when both cases of the Ceratina sp. were counted as one species and when they were counted as multiple species (Table 1).

No remarkable negative effects from the evacuation were shown for the other taxonomic groups (Tables 2 and 3; Figs 3–5), whereas Coleoptera, Hymenoptera, winged ants, and workers of Camponotus japonicus and Formica japonica showed remarkable positive effects (Tables 2 and 3; Fig 4B and 4G; Fig 5C and 5D).

We found several notable differences inside and outside the evacuation zone with respect to various flying insects, including pollinators, 3 years after the power plant accident, even when we incorporated spatial autocorrelation and the effects of surrounding landscapes before the disaster. Statistical modelling at the species level in Apidae suggested that the density of Xylocopa appendiculata, the largest Apidae species in the region, was significantly low in the evacuation zone, whereas the densities of small species, such as Ceratina flavipes and Nomada spp, were higher. Bumble bees were not affected by the evacuation, which is consistent with the rapid census conducted by Møller et al. [10] 6 months after the disaster. In addition, modelling at a higher taxonomic level in other insects suggested that the majority of insects did not significantly differ by evacuation (or were present at a higher density). However, it should be noted that we did not differentiate species composition.

Considering previous studies on insects and radiation levels (of less than several dozen micro greys per hour) in the evacuation zone [11, 16], it is unlikely that the radiation directly damaged the population of bees. In particular, the life history of the carpenter bee suggests that, unlike bumble bees, which often nest underground, it is not susceptible to exposure to high levels of radiation [19, 20]. Alternatively, the potential cause of the lower density of the carpenter bee in the evacuation zone might be the reduction in the flowers of garden plants following the evacuation. Xylocopa are generally found in tropical zones [21], and the bees also favour large and conspicuous flowers such as wisterias, orchids and passion fruits [22–24] as well as gardenias.

The higher density of small bees and other insect groups may be due to “release of resources” by the cessation of human activities such as farming. However, caution is required in interpreting the results because our sampling method, which relied on Malaise traps, was not optimal for worker and wingless ant species. It is worth noting that Ceratina flavipes, which prefers lowland areas [25], were found at higher density in the evacuation zone, which is in contrast to the distributions of other congeneric species. The cessation of farming, which had generally been dominant in the lowland areas of the evacuation zone, may have increased the wild plants that provide food and nesting materials for this species. The positive effects on Nomada spp. should also be interpreted with caution because they occurred at a low density and as multiple species. A potential ecological cause of the observed pattern of Nomada spp. might be that, as one of the smallest species groups in Apidae in Japan, Nomada spp. could make use of a broader range of flower species, including Compositae weeds, with relatively small flowers [25], which are often found in abandoned lands [26, 27].

Fortunately, there appeared to be no marked loss of ecosystem functions and services by flying insects in the evacuation zone. Pollinators (bees, wasps, butterflies, and hoverflies) were shown to be abundant, with the exception of the absence of carpenter bees, whose importance for ecosystem services is relatively low in temperate zones [21, 22]. In addition, re-colonisation of the species after reconstruction might not be difficult because large bees, such as bumble bees, have been observed to show high dispersal ability and rapid range expansion [28, 29]. The abundance of medically important pests, such as flies and horseflies, tended to be higher, but not significantly so, differing from the higher rates observed in the coastal areas of northern Japan immediately after flooding following the tsunami [30, 31].

However, the long-term cessation of anthropogenic activities and the resulting disturbances may negatively affect certain insect groups, or groups not evaluated in this study (e.g., aquatic insects that depend on rice paddies), in the future. Moderate disturbances from farming have been reported to have a widespread positive effect on biodiversity in Japan [13, 14, 32]. Therefore, continuous monitoring and complementary studies will be needed to effectively plan the return of residents to the evacuation zone in a way that incorporates biodiversity and ecosystem services. Future monitoring of not only insects but also changes in land use (and radiation levels) following reconstruction will also be able to demonstrate if our results, which were derived from snapshot data, were merely historical effects existing before the evacuation.

Nevertheless, our sampling method, which was relatively labour-saving but covered broader areas using socially-stabilised sampling locations (i.e., schools and community centres), favoured indicator species that commonly occurred around the disaster area and are sensitive and/or related to human activities (e.g., carpenter bees). Thus, they provide a good basis for a biodiversity monitoring system in the evacuation zone. Considering the fact that the number of nuclear power plants will increase in Asia [2], where the history of land use is relatively long [7, 33] and the population is still growing [34], our research will be of benefit efforts not only to reconstruct the Fukushima site but also to effectively respond to future potential disasters.

The local governments in Fukushima prefecture, Japan (Date city, Iitate village, Iwaki city, Minami souma city, Namie town, Naraha town, Nihonmatsu city, Souma city, and Tamura city) granted permission for the Field study.