**1. Introduction**

In the agroecosystem, pollinators are a pivotal component of biodiversity that provide an important ecosystem service through crop pollination [1] and increasing fruit set [2]. Pollinators also can be used as indicators of ecosystem health because of their sensitiveness to environmental stressors [3], for instance, the negative impacts of pesticide application [4]. There is growing concern relating to declines found in pollinators around the world [5]. In Europe and the US, a decline in wild bee species richness has been recorded, where the declining trends are in the abundance of honey bees (*Apis mellifera*) and a small number of wild pollinators [6]. Although high diversity of bees is found in the tropics (e.g., [7,8]), there is a lack of information about this phenomenon. Therefore, investigation needs to be undertaken into the scale, magnitude, and causes of the decline and the effects on pollination services.

Global declines in honey bees and wild bees have been associated to habitat loss and fragmentation, pesticide application, pathogens, invasive species, and climate change [9,10]. The potential threat of insecticides, such as neonicotinoids, for honey bees and wild bees has been reported, although it is still in debate [11]. Neonicotinoids have negative impacts such as increasing the mortality of honey bees by impairing their homing ability [12] and reducing the reproductive success of bumble bees and solitary bees [13], although other studies have reported no effects [13]. There is limited information from comprehensive studies on the impact of neonicotinoids toward long-term survival of honey bee colonies [14]. Landscape-scale experiments in different geographical regions are needed to investigate the impacts of neonicotinoids on bees [13,14].

The research was conducted in various land-use types both in Bogor (West Java) and Malang (East Java), Indonesia. Bogor has unique agricultural characteristics, as it is surrounded by mountain areas, and has a seminatural habitat dominated by agricultural fields cultivated with rotations of crop plants, which are mainly rice and vegetables [15]. Similar to Bogor, Malang is also surrounded by mountain areas and consists of tropical rainforests as well as cultivated and settlement areas. Agriculture is the primary land-use on the island of Java, so its managemen<sup>t</sup> has profound consequences for the environment and for biodiversity. Agricultural intensification, especially pesticide application, is commonly used as a consequence of the green revolution [16] and has a negative effect on biodiversity, especially pollinators [17].

The objectives of this research were to investigate land-use effects and the indirect impact of insecticide application in agricultural areas on insect pollinators, particularly bees. Most evidence of the impact of pesticides on pollinators, especially honey bees, has come from laboratory-based toxicity tests. Negative effects of insecticide have been reported (e.g., [18]), but field research is still needed to understand how laboratory-derived toxicity levels effect pollinator communities in the agroecosystem, although some field- and semi-field-based studies have been conducted [19].

#### **2. Materials and Methods**

#### *2.1. Research Site and Experiment Plot Selection*

Research sites were located in West Java (Bogor and its surrounding area) and East Java (Malang and its surrounding area) to compare between two different geographical areas (Figure 1). Both areas are reported as honey producers in Indonesia [20]. Beekeepers in Bogor were characterized by breeding in a lower elevation (<300 m asl), while in Malang it was in a higher elevation (>500 m asl) (https://en.climate-data.org). To conduct the experiment, we selected two different habitat types, agriculture and forest, both in Bogor and Malang. As a comparison, we also observed the hives in selected beekeeper gardens (home garden) (Table 1). We used three different species of bees during the experiment that were commonly breed by beekeepers: *A. mellifera* and *A. cerana* in Malang and *A. cerana* and *T. laeviceps* in Bogor. In each habitat, we selected three experiment plots in different locations for replication, and the minimum distance between plots was 2 km (Figure 1).


**Table 1.** Satellite images and description of habitat types for research experiments in Bogor and Malang. Satellite images were derived from Google Maps, accessed year 2019.

**Table 1.** *Cont.*

**Figure 1.** Map of study sites for research experiments in different habitat types (Ag: agriculture, Fr: forest, Hg: home garden (beekeepers place) in Bogor (West Java) and Malang (East Java)). The numbers after the habitat code indicates plot number. Home garden with triangle symbol indicates beekeepers of *Apis cerana*.

We placed three beehives in each plot for observation. The sizes of beehives were different depending on the beekeeper's practice in rearing bees. In Malang, the hive size of *A. cerana* was 40 × 25 × 25 cm with four combs and *A. mellifera* was 50 × 40 × 25 cm with four combs. While in Bogor, the hive size of *A. cerana* was 35 × 30 × 25 cm with six combs and *T. laeviceps* was 30 × 10 ×10 cm without combs.

#### *2.2. Observation of Bees in the Hives and Residue Analysis*

To study the effect of habitat type on bees, we observed the hives on each experimental plot, measured the population growth of bees, and performed a residue analysis. Population growth of bees

was measured by counting the forager abundance (foraging activity) and colony weight (weighing full hive). The method of foraging activity monitoring was based on [21] by counting the foragers departing or returning to the colony for thirty minutes per hive. In each plot, the observation was conducted from 7 a.m. until 11 a.m. Weighing full hives was conducted to calculate the colony weight that included the summed weight of the box, combs with food stores, and the bees [22]. Both forager abundance and colony weight were observed every two weeks during two months. Observations were conducted in di fferent seasons (i.e., rainy and dry seasons). Observations in the dry season were done from March to May 2019, while for the rainy season, observations were from July to September 2019.

In addition, insecticide residue analysis was conducted by collecting honey and bee (foragers) samples in three habitat types. We collected 5 mg of bees and 5 mg of honey per plot and initially froze it before being analyzed using the QuEChERS protocol [23]. Insecticide residue analysis was conducted in the medical laboratory of Jakarta (https://labkesda.jakarta.go.id) and was focused on imidacloprid content as the representation of neonicotinoid insecticide.
