Environmental factors affecting honey bees (Apis cerana) and cabbage white butterflies (Pieris rapae) at urban farmlands

Introduction

Urban expansion in many countries is often accompanied by the loss of agricultural lands (Bren d’Amour et al., 2017). The world’s urban areas are a quarter of total agricultural lands; however, urban expansion has occurred faster than urban population growth and takes place on productive agricultural lands, leading to reduction in global crop production (Seto & Ramankutty, 2016; Bren d’Amour et al., 2017). Moreover, the proportion of global population in urban areas is forecast to increase from 55% in 2018 to 68% by 2050 (United Nations, 2019). Approximately 90% of the increase would occur in Asia and Africa, which are also the hotspot of future urban expansion. These changes, i.e., growing urban population and agricultural land loss, raise concerns for not only food security but also for the livelihoods of smallholders, especially in developing countries (IFPRI, 2017; Huang et al., 2020).

Urban agriculture has received increasing attention as one of the practices that may improve the accessibility of food, the livelihood of local people such as farmers and rural migrants, and human well-being (Deelstra & Girardet, 2000; De Bon, Parrot & Moustier, 2010; Mok et al., 2013; Martellozzo et al., 2014). One recent study also shows that small-scale urban agriculture can bring high yields (McDougall, Kristiansen & Rader, 2018). Urban agriculture can be broadly defined as agricultural production, such as vegetables, fruits, ornamental plants, and other dietary products, occurring in urban environments that include both inner city and city fringe areas (Mougeot, 2000). In China, small-scale vegetable farming is common in cities, especially at vacant lots and city fringe areas. Urban agriculture, also called peri-urban agriculture in some cases, has been supported by the Chinese government policy “Vegetable Basket” for decades (Zhong et al., 2021). Like other countries, urban expansion has been a major driver causing farmland loss in China (Tu et al., 2021). Although farmland protection policy started in mid-1990s has alleviated the loss to some extent, urban development frequently occurs on highly productive farmlands near urban edge areas, often leaves reclaimed but less productive farmlands as an offset for the development, and increases farmland fragmentation (Liang et al., 2015; Huang, Du & Castillo, 2019; Tu et al., 2021). In this situation, understanding ecological factors that could affect agricultural production at urban farmlands is crucial to promote the sustainability of these farmlands.

Insects are well known for their role in ecosystem service (e.g., pollination and pest control) and disservice (e.g., crop damage) in agricultural landscapes (Losey & Vaughan, 2006; Kremen & Chaplin-Kramer, 2007; Oliveira et al., 2014; Omkar, 2018). Their effects on crop yields have been widely studied. Over 70% of main global crops benefit from animal-mediated pollination (Klein et al., 2007) and 9–25% of staple food crops are estimated to be lost due to animal pests (Oerke, 2006), in which insects are the major group in both cases. Bees are considered the most important pollinators. For example, in the USA, 11% of the agricultural gross domestic product in 2009 depended on pollination, mainly contributed by bees (Lautenbach et al., 2012). Bee abundance affects yields of certain crops such as oilseed rape more than pesticides do (Catarino et al., 2019 and references therein). In particular, it is widely recognized that honey bees provide important pollination services for a variety of crops although pollination efficiency of the Western honey bees (Apis mellifera) in a natural habitat is somewhat debatable (Hung et al., 2018). A number of factors, e.g., quantity and quality of floral resources, semi-natural vegetation, crop diversity such as crop compositional or configurational heterogeneity and landscape complexity, influence the abundance and diversity of bees (Klein et al., 2012; Scheper et al., 2013; Potts et al., 2003; Priyadarshana et al., 2021; Raderschall et al., 2021).

In contrast to bees, cabbage white butterflies (e.g., Pieris rapae and P. brassicae) are recognized as agricultural pests because their larvae, cabbageworms, feed on the family Brassicaceae crops (cruciferous crops) and can severely damage these crops. For example, cabbage white butterflies caused about 70% yield loss in cruciferous crop in Meghalaya, the northeastern state of India (Singh, Satyanarayana & Peshin, 2014). While presence or abundance of Brassicaceae crops is positively associated with cabbageworm density, this effect may be amplified if the host plants are congregated as a large pure stand (resource concentration hypothesis; Root, 1973) or sparsely distributed (resource diffusion hypothesis; Yamamura, 1999). The type of surrounding crop has an impact on cabbageworm density (Maguire, 1984) and distance between cruciferous patches influences egg density of cabbage white butterfly (Fahrig & Paloheimo, 1988). Floral resource availability can also be important to the adult butterflies (Curtis et al., 2015).

Here, using the Asian honey bee (Apis cerana) and cabbage white butterfly (Pieris rapae), I investigated how environmental features within and surrounding urban farmlands in a megacity of China influenced the abundance of these two species, during winter in which Brassicaceae crops are of particular prevalent. A. cerana is small native species in southern and southeastern Asia. While the hive of A. cerana produces lower amount of honey than that of A. mellifera, A. cerana is more efficient in pollinating various fruits and vegetables (Partap, 2011). I focused on farmlands largely cultivating vegetables, which are dominant agricultural products of urban farming. I expected that the percentage of flowering plants and Brassicaceae crops would positively affect A. cerana and P. rapae, respectively. Given that these farmlands are located in urban areas mostly filled with buildings and houses, landscape weedy vegetation (spontaneous vegetation in vacant lots or parks) as well as available farmlands could also be critical to both species. Crop diversity could negatively affect P. rapae as found in the effect of crop diversification on insect pests in general (Hooks & Johnson, 2003; Beillouin et al., 2021). Alternatively, the effect may not be significant because urban farmlands in this study region often contain more diverse crops than in rural agricultural lands, and predator/parasitoid abundance is low in winter.

Materials & Methods

Study area

Guangzhou is the capital city of Guangdong Province, China (Fig. 1) and has a population around18 million. It also lies within the Indo-Burma Biodiversity hotspot region, the most urbanized area among biodiversity hotspot regions in China (Güneralp & Seto, 2013). Guangzhou has a subtropical monsoon climate: warm and dry winter and hot and humid summer, with a mean annual temperature of 22.2 °C (https://en.climate-data.org). A warm winter (mean temperature in January = 14.2 °C) allows for farmers to grow crops year-round. Although winter is relatively drier than other seasons, humidity is not low (60–71% between December and February). During the past two decades, farmlands between the inner city and the outskirts have drastically decreased due to urban expansion and economic development (Shi & Shi, 2020).

I selected 33 farmlands in 2020 and 57 farmlands (23 from previous year) in 2021 (Fig. 1). Several farmlands chosen in 2020 could not be surveyed in 2021 due to development, abandonment, and too-early growth stage of crops (very small seedlings). The size of all farmlands was ≥ 7,850 m². Within a farmland, I established a 50 m-radius area (site, hereafter) that was also used for bird surveys in another study (Lee, Chen & Zou, 2022). All samplings were performed within the 50 m-radius area during January each year. P. rapae was surveyed at 90 sites (32 in 2020 + 58 in 2021) across two winters: the distance between the closest two sites was 1.87 ± 1.09 km (mean ±  standard deviation) in 2020 and 0.82 ± 1.18 km in 2021. Bee sampling was conducted at 58 sites (25 sites in 2020 and 33 sites in 2021).

I did not distinguish between urban and peri-urban agriculture, including farmlands located in the inner city as well as close to the city boundary (Mougeot, 2000). However, I emphasize that all sites were in urban areas; the proportion of built area (mostly buildings and houses) within a 500 m-radius area of sample sites averaged 61 ± 13.8%, consistent with the criterion of urban areas containing >50% built area (Marzluff, 2001).

Results

A total of 1,189 P. rapae was observed across all sites. Mean abundance was 13.21 butterflies per site with a SD of 10.16, ranging from 1 to 39. The abundance of P. rapae was strongly associated with a combination of local and landscape variables: the full model showed the lowest AICc, and Δ AICc of other models was >9 (Table 1).

Local crop diversity had a positive effect on the abundance of P. rapae, whereas landscape-level farmland and weedy vegetation covers affected P. rapae abundance negatively (Fig. 2 and Table S4). The fragmentation of farmlands, i.e., patch density, was not associated with P. rapae abundance given wide 95% confidence intervals across 0 and very low estimate (Table S4). Although percent cover of cruciferous crops tended to have a positive impact on P. rapae abundance, the effect was not as strong as other significant variables (Table S4).

For two winters, 265 A. cerana individuals were captured. While more individuals were caught in 2021 than 2020, A. cerana abundance varied by sample stations: 1.15 ± 2.11 (mean abundance ± SD), ranging from 0 to 14 in 2020, and 1.82 ±  2.03, ranging from 0 to 11 in 2021. Of 8 GLM models, 4 models were selected as plausible models in 2021 data analysis and 2 models in combined years data analysis (Table 2).

Effects of weedy vegetation on A. cerana abundance were consistent across two winters given that the 95% confidence intervals did not include 0 or slightly overlapped with 0. The effects also depended on scale: the abundance of A. cerana was affected by weedy vegetation cover negatively at the landscape scale but positively at the local scale (Fig. 3 and Table S4). In combined years data analysis, increasing percent cover of farmland at the landscape scale was positively associated with increasing A. cerana abundance. A. cerana abundance was also low in a landscape with more fragmented farmlands. The percentage of flowering plants had a strong positive effect on A. cerana abundance in 2021: its parameter estimate (i.e., effect) was ≥ 1.5 higher than the estimates of other variables associated with A. cerana, suggesting the amount of floral resource at the fine scale could be critical to A. cerana (Fig. 3).

Discussion

The results of this study reveal variation in environmental factors at relevant spatial scales associated with a host-specific pest, i.e., Pieris rapae, and a pollinator, i.e., Apis cerana. P. rapae was more related to crop cover at the landscape scale (negatively), whereas A. cerana was strongly related to flower cover at the fine scale and weedy vegetation at the local scale (positively). Both species also responded negatively to weedy vegetation cover at the landscape scale. These patterns can provide insight into how small urban farmlands can be managed to improve ecosystem service, i.e., pollination by A. cerana, while minimizing ecosystem disservice, i.e., crop damage by P. rapae.

The resource concentration hypothesis predicts high abundance of host-specific herbivorous insects at their host plant rich patches, i.e., dense, large, or pure stands of host plants (Root, 1973). I expected a positive relationship between P. rapae abundance and percent cover of Brassicaceae, and I found a tendency for this relationship. However, landscape-scale crop cover was the environmental variable most strongly associated with the abundance of P. rapae. Certainly the presence or absence of Brassicaceae in surrounding farmlands could influence P. rapae distribution. If Brassicaceae cover is low in a landscape with otherwise high crop cover, we might expect a negative association. Although I lack data on Brassicaceae cover at the landscape scale, local Brassicaceae cover was not correlated with landscape crop cover. Also, almost all farmlands surveyed contained Brassicaceae: of 90 sites, 44 included >20% Brassicaceae at the local scale and only seven had <5% Brassicaceae. Thus, the negative response of P. rapae is unlikely related to the amount of landscape Brassicaceae cover. Matteson & Langellotto (2012) found that P. rapae spends less time in urban gardens in a landscape with more green spaces, which could lead to low detection of cabbage butterfly in these gardens. I often observed more P. rapae individuals and their mating, landing, and oviposition behaviors at farmlands isolated in a landscape dominated by built structure. Considering that crop and weedy vegetation covers are part of green spaces, the responses of P. rapae to both covers parallels the previous finding. Increasing crop covers in a landscape likely diffuses P. rapae, lowering abundance per farmland. It may also interrupt visual cues and consequently have a negative impact on their ability to search host and floral resources because P. rapae depends on vision to locate these resources (Hern, Edwards-Jones & McKinlay, 1996).

The positive effect of crop diversity is somewhat unexpected. Polyculture stands often show lower density or higher mortality of pest insects compared to monoculture stands (Altieri et al., 1978; Russell, 1989; Letourneau et al., 2011; Iverson et al., 2014; among others). The diversification of cruciferous crops through mixing or intercropping can effectively control pest insects due to higher abundance and richness of parasitoids/predators, visual camouflage effect, and masking effect of host plant odors, among other mechanisms (Finch & Collier, 2000; Hooks & Johnson, 2003). The positive effect I observed may be affected by the season of survey. Although winter is relatively warm and dry in southern China, the abundance and richness of predator and parasitoid insects are lower than other seasons such as Spring (M-B Lee, 2020-2021, pers. obs.). Lowenstein & Minor (2018) reported different trends in abundance between herbivores including P. rapae and predator/parasitoid insects at urban gardens and farms in Chicago, IL, USA; areas with high predators and parasitoids tended to have low herbivores. Compared to large-scale rural agricultural lands, urban farmlands have also more diverse crops despite their small size. Brassicaceae crops are cultivated at most of these farmlands in the study area. This creates a spatially and temporally heterogeneous environment due to variations in harvest time and growth stage between crops, even between Brassicaceae crops. These conditions may benefit P. rapae by providing host and floral resources, shelters, or mating chances for a longer period during the winter season. I also noticed seasonal variations in crop composition, although crop diversity at each study site did not change significantly between seasons. For example, in spring, the proportion of Brassicaceae crops declines but other crops (especially vine crops, e.g., bean and gourd family crops) increase and flowering weeds are abundant. P. rapae may respond differently to crop diversity in spring, but the extent of possible seasonal variations in the responses of P. rapae requires further research.

Bee populations are strongly regulated by food resource availability (Roulston & Goodell, 2011). The establishment of semi-natural vegetation, especially planting flowering herbaceous plants at field edges (i.e., flower strips) is often recommended to promote crop pollination and yield by insect pollinators (Blaauw & Isaac, 2014; Williams et al., 2015; Shutter, Albrecht & Jeanneret, 2018; but see Nicholson et al., 2019). One recent meta-analysis shows that crop pollination decreases with increasing distance to floral resources, indicating the importance of floral resources to pollinators (Albrecht et al., 2020). Crop diversity and semi-natural vegetation at landscape scales also increase bumble bee density in wheat-dominant agricultural areas (Raderschall et al., 2021). The result of the current study, i.e., the strong effect of percent cover of flowering plants on bee abundance, partly supports the general trend. While I do not have data on the diversity and abundance of flowering plants at local and landscape scales, the result highlights the benefit of even small amount of flowering crop and weed to honey bees in urban environments, especially during the winter in which natural floral resources may be scarce. This could also explain the different effects of weedy vegetation cover at local and landscape scales. The subtropical climate in southern China enables framers to cultivate vegetable crops year-round. While some weedy vegetation showed withering in winter, the condition was less severe at urban farmlands than some vacant lots (e.g., construction sites and abandoned lands), which is likely affected by differences in water availability. There were also more weed plants blooming at urban farmlands based on my observation.

One potential concern is the quality of floral resources in these urban farmlands because most common weed plants flowering in the study area are non-native plants, particularly Bidens species. There are few studies comparing nutritional values of nectars to bees between crops and weed plants in this region. However, honey bees are generalists, visiting a wide range of flowering plants, and require different diets to maintain colony health (Requier et al., 2015). During times of food shortage, weed plants and non-native plants can be an important component of the diets of A. mellifera and A. cerana, respectively (Requier et al., 2015; Koyama et al., 2018). A. mellifera also forages on flowers of Bidens species (Kajobe , 2007). Floral resource use and preference by bees are significantly associated with dominant plants and affected by season as floral resource availability changes throughout the year (Lowe et al., 2021). Most flowering plants at sample sites in winter are weed plants and several crops that belong to the genera Brassica, Chrysanthemum, and Allium. Weed plants can be one of major floral resources to A. cerana in winter regardless of their nutritional values. However, with seasonal variations in the diversity and abundance of floral resources, A. cerana may prefer certain crops or weed plants in other seasons such as spring, when trees and shrubs as well as more crops bloom. The degree of effects of flower strips on local bee abundance and diversity can also depend on the interaction between characteristics of flower strips and floral resources available in surrounding landscape (Scheper et al., 2015). Similarly, it is possible that the strong response of A. cerana to fine-scale flower cover could be influenced by the total amount of floral resources within farmland and landscape, which will vary by season.

It is noteworthy that patchy density of farmland had a negative impact on A. cerana abundance weakly or significantly. This suggests that A. cerana population may be susceptible to farmland fragmentation to some degree. While the honey bees’ response to habitat fragmentation is rarely explored, one recent study shows that density of green patch in urban areas can have an indirect negative effect on total abundance of bees by negatively influencing flowering plant richness (Theodorou et al., 2020). In the current study, farmland fragmentation, i.e., patch density of farmland is highly positively correlated with total edge length of farmlands. Banana plants are common along the boundaries of some farmlands and used as a fence. Banana flowers may be an important nectar and pollen source for some managed honey bee (e.g., hybrid Carniolan honey bee, A. mellifera carnica Pollmann) when they bloom (Taha, Taha & AL-Kahtani, 2019). However, winter is not a bloom season of banana plants in the Guangdong province. There are also a variety of flowering trees, weeds, and crops that can be used by bees in other seasons. Thus, banana-dominant vegetation at the edge of farmland likely reduce the space that could be covered with weedy plants and consequently floral resources available to A. cerana, especially in winter. Fragmented farmlands may be also managed intensively: for example, frequent weed removal at the farmlands can decrease local weedy vegetation cover. However, these explanations remain speculative until we have multiple studies performed in other regions and sufficient data to compare them for better understanding of the relationship between A. cerana and farmland fragmentation.

Conclusions

Overall, the findings of current study can provide a scientific basis for urban planners, policy makers, and farmers to consider in the management of urban farmlands across scales to enhance the sustainability of these farmlands, particularly in large subtropical city like Guangzhou. At the farmland (local) scale, cultivating too diverse crops may not be recommended as it can increase the risk of crop damage by P. rapae. However, the risk can be alleviated at farmlands embedded in a landscape with more overall crop cover, which may partly benefit A. cerana as well. Retaining non-crop area such as weedy vegetation patches or small portion of flowering plants including crops within a farmland can be an effective practice to maintain A. cerana population and promote pollination. It also benefits birds as local weedy vegetation has a positive effect on winter bird diversity at urban farmlands (Lee, Chen & Zou, 2022). With growing urban expansion, farmlands in China have been significantly converted into built area, leading to a decline in net primary productivity of cropland (He et al., 2017) despite farmland protection policy restrictions on development. Farmlands in a city are not exceptional. In major cities in China, new development largely spreads from the edge of city or occurs at old villages within a city (Tu et al., 2021), lowering crop cover in surrounding landscapes and isolating farmlands (Liang et al., 2015). Policy makers and urban planners need to consider preserving farmlands, especially at the city fringe or suburb for permanent agricultural activities. Priority may be given to farmlands embedded in landscapes with relatively high crop covers. It is also critical that city government monitors potential environmental problems associated with urban farmlands. For example, most of these farmlands, especially ones next to main districts where development and population are concentrated, use waste water. The average contents of heavy metals in agricultural soil in Guangzhou and Foshan cities are higher than background values of Guangdong province (Xiao et al., 2020). While the levels of metals are still lower than national standard values, the ecological risk of heavy metal pollution in these cities is not negligible. Combined with further research on seasonal variations in the patterns found in current study, the assessment of environmental quality of urban farmlands can inform efforts to improve the sustainability of urban agroecosystem. At the same time, it is important to note that the findings of current study are based on a single study conducted in one megacity. Although the findings offer valuable insights on environmental factors associated with insect pollinators and pests at urban farmlands, more studies are needed in other tropical/subtropical cities to draw general conclusions.

Supplemental Information