Establishing next-generation pest control services in rice fields: eco-agriculture

Pesticides are commonly used in food crop production systems to control crop pests and diseases and ensure maximum yield with high market value. However, the accumulation of these chemical inputs in crop fields increases risks to biodiversity and human health. In addition, people are increasingly seeking foods in which pesticide residues are low or absent and that have been produced in a sustainable fashion. More than half of the world’s human population is dependent on rice as a staple food and chemical pesticides to control pests is the dominant paradigm in rice production. In contrast, the use of natural enemies to suppress crop pests has the potential to reduce chemical pesticide inputs in rice production systems. Currently, predators and parasitoids often do not persist in rice production landscapes due to the absence of shelter or nutritional sources. In this study, we modified the existing rice landscape through an eco-engineering technique that aims to increase natural biocontrol agents for crop protection. In this system, planting nectar-rich flowering plants on rice bunds provides food and shelter to enhance biocontrol agent activity and reduce pest numbers, while maintaining grain yield. The abundance of predators and parasitoids and parasitism rates increased significantly in the eco-engineering plots compared to the insecticide-treated and control plots. Moreover, a significantly lower number of principal insect pests and damage symptoms were found in treatments where flowering plants were grown on bunds than in plots where such plants were not grown. This study indicates that manipulating habitat for natural enemies in rice landscapes enhances pest suppression and maintains equal yields while reducing the need for insecticide use in crop fields.

Enhancing existing biocontrol agents can also reduce the need for augmenting natural enemies via mass release, reducing expense and risks associated with importing foreign species [26][27][28][29] .
Conservation biological control involves modifying the environment or existing practices to protect and enhance natural enemies and reduce pest damage 30 . Habitat management is a form of ecological engineering aimed at providing food and shelter for natural pest control agents. Carefully planned habitat manipulations can increase the abundance and effectiveness of natural enemies while promoting biodiversity and structural complexity of agroecosystems 31,32 . In particular, providing natural enemies with needed resources such as nectar 33 , pollen 34 , physical refugia 35 , alternative prey 36 , alternative hosts 37 and lekking sites 38 .
The concept of ecological engineering has been used to restore or enhance biodiversity in the rice landscape [39][40][41][42] . Unlike many flowering plants, rice lacks floral or extra-floral nectar resources that can be used by natural enemies. Planting additional nectar-rich flowering crops in rice landscapes can enhance year-round resource for natural enemies. In Bangladesh, individual rice fields are typically surrounded by an earthen ridge approximately 0.5 m in height to retain irrigation water. This ridge is commonly known as a rice bund or ail and is typically fallow throughout the year. Under eco-agricultural management, bunds are enriched with nectar producing plants and non-rice habitats used to grow perennial plants to provide additional food and shelter for natural enemies 43,44 . These improved habitat-characteristics can greatly influence natural enemy longevity, fecundity and behavior and lead to reduced pest abundance 42,[45][46][47][48] . The main goal of this study was to test the efficacy of an eco-agricultural system in promoting natural enemies, reducing crop pests and maintaining rice yield. We contrasted three treatments; growing nectar-rich plants (marigold, cosmos, sunflower and sesame) on bunds ( Fig. 1) coupled with no insecticides on rice (hereafter T 1 ), compared to fallow bunds with (T 2 ), and without prophylactic insecticide use on rice (T 3 ).

Results
Gazipur. Insect pests and natural enemies observed during the three Boro seasons (2015-16, 2016-17 and 2017-18) are presented in Fig. 2. The results show that the highest number of grasshoppers (GHs) and yellow stem borers (YSBs) were found in T 3 during Boro 2015-16 ( Fig. 2A). A significantly lower number of GHs was found in the T 1 treatment (P < 0.05). Other insect pests were not observed and are not presented here. The green mirid bug (GMB) is the most important predator of brown planthoppers in rice fields. Populations of GMBs were observed in only the T 1 treatment ( Fig. 2A). In subsequent years, significantly higher numbers of natural enemies, including spiders, damsel flies, and lady bird beetles were found in T 1 than in the other treatments ( Fig. 2B,C, P < 0.05). We also recorded parasitoids of rice insect pests using yellow sticky traps. The highest number of parasitoids were observed in the T 1 treatment plots during the Boro season in 2016-17 and traps facing east and west in each field caught similar numbers of parasitoids (Fig. 2D).
In the T. Aman 2016 season, grasshoppers were more abundant in the T 1 treatment, but remained below the economic threshold level (ETL) (Fig. 3, upper panel) and a similar pattern was observed in 2017 (Fig. 3, lower panel). However, grasshoppers generally do not hamper successful rice production and are seldom treated in the absence of other key pests. In 2017, abundance of a key economic insect pest, the green leafhopper (GLH) was lower in T 1 than the control treatment (Fig. 3, lower panel). In 2016, parasitoids, spiders and damsel flies were more abundant in T 1 (Fig. 3, upper panel) but not in 2017 (Fig. 3, lower panel). www.nature.com/scientificreports www.nature.com/scientificreports/   (Fig. 4). In all cases parasitism of pest eggs was significantly higher in T 1 than in other treatments ( Fig. 4, P < 0.01). Parasitism of planthopper eggs was higher in the T. Aman season, exceeding 80% (Fig. 4A) than in the Boro season when it never exceeded 60% (Fig. 4C,D).
The presence of flowering plants on rice bunds had a significant impact on the damage intensity caused by YSBs (P = 0.05). The highest incidence of white head per 100 hills due to YSB infestation was found in T 2 plots, where insecticide was applied three times during the rice growing period (Fig. 5). The lowest incidence of white head was in the T 1 plots, which were not sprayed with insecticide and bordered by sesame/marigold/cosmos. No significant impact of any treatment on yield per hectare of land was observed ( Rajshahi. Results for pest and natural enemy abundance as well as yield impacts were broadly similar at the Rajshai site. During Boro 2015-16, the highest number of grasshoppers (GHs), white leafhoppers (WLHs) and rice bugs (RBs) were found in T 1 (Fig. S1). However, their incidence was below the ETL. Significantly higher numbers of spiders, damsel flies, lady bird beetles and carabid beetles were also found in T 1 (Fig. S1, P < 0.05). In T 3 , insecticide was applied four times, none the less, yields were similar to that of T 1 and T 2 . In T. Aman 2015, significantly higher number of YSBs and white leafhopper (WLHs) were observed in the T 2 treatment (Fig. S2, P < 0.05). More importantly, T 1 hosted a significantly higher number of natural enemies, including spiders, lady bird beetles, carabid beetles and damsel flies (Fig. S2), while T 3 had significantly lower abundance of natural enemies, indicating a detrimental effect of insecticide (P < 0.05). Yield across the three treatments were similar ( Fig. S3, P > 0.05). In the Boro 2016-17 season, the highest number of economic pests, YSBs, GHs, WLHs and stink bugs, were found in the T 3 treatment (Fig. S1) while significantly higher numbers of lady bird beetles and damsel flies were found in T 1 (Fig. S2, P < 0.05). Again, yields across the three treatments were similar ( Fig. S3, P > 0.05).

Discussion
Two hundred thirty-two insect pest species, 183 parasitoids and 192 predators are known to occur in Bangladesh rice ecosystems 49 . However, fewer than 20 species are considered significant pests capable of causing yield losses if they infest plants in sufficiently large numbers. Predators and parasitoids often attack these pests and control them naturally in the field. Over the last three decades, the introduction of high-yielding rice varieties to feed the fast-growing human population in developing countries, such as Bangladesh, has resulted in the use of large amounts of chemical insecticides. This heavy reliance on chemical insecticides reduces natural enemy www.nature.com/scientificreports www.nature.com/scientificreports/ populations in rice landscapes, promoting pest outbreaks 15 . In this study, we show that modification of existing rice landscapes via ecological engineering techniques in which nectar-rich flowering plants are grown on rice bunds surrounding rice fields promotes biocontrol agents, reduces pests and maintains crop yields.
In Bangladesh, land for rice production is commonly fragmented into numerous small plots held by individual farmers. Each rice plot has a bund around the plot. This bund is typically fallow and serves as a boundary to separate the plots owned by other land holders. We grew nectar-rich flowering plants, including marigold, cosmos, sunflower and sesame, on the bund of rice plots, which provided food, shelter and other essential nutrients for biocontrol agent development and reproduction in the rice landscape (Fig. 1). Parasitoids regularly consume nectar from flowers for sugars and carbohydrates, which increases their fitness (reproduction rate, population size, longevity, etc.) 50 . Our results show that adding flowering plants to rice bunds resulted in a significantly higher abundance of parasitoids and predators (Figs 2B,C, 3) and increased parasitism of planthopper, rice hispa, and yellow stem borer eggs in the rice field (Fig. 4). Parasitoids observed in the field or that emerged from parasitized eggs were identified as Trichogramma zahiri, T. chilonis, Anagrus spp., tachinid flies and many other ichneumon wasps. In previous studies, sesame flowers planted on  www.nature.com/scientificreports www.nature.com/scientificreports/ bunds in the T. Aman season provided food and shelter resources to natural enemies, including hymenopteran parasitoids 40 . Sesame flowers have been found to significantly improve searching success for planthopper egg parasitoids of the genus Anagrus. Zheng et al. 50 reported that adding nectar-rich plants and withholding insecticide consistently and significantly increased the rate of parasitism of planthopper eggs and maximized the observation of planthopper and lepidopteran egg parasitoids in rice fields.
In our study, adding nectar-rich flowering plants to rice bunds primarily facilitates parasitoids, which are recognized as essential pest control agents globally 51 . Several parasitoids have been identified for C. suppressalis, C. medinalis, D. armigera, S. incertulas and planthoppers. Forty-two parasitoids of C. suppressalis, 44 parasitoids of C. medinalis and 16 parasitoids of planthoppers have been identified as effective control agents 52 . Anagrus spp. were commonly found in our experiment, and these egg parasitoids have also been found to be effective against planthoppers in other Asian rice-growing countries 40,52 . Significantly higher spider populations were observed in the rice plot where the nectar-rich flowering plants were treated with insecticides. Spiders are considered to be very effective predators of many insects worldwide as well as in rice fields [52][53][54] . Nyffeler and Birkhofer 55 reported that global spider communities might destroy 400-800 million tons of insect pest species. Tetragnathidae is the most dominant spider taxon in rice ecosystems 56 .
The lowest parasitism rate and catches of parasitoids of planthoppers, lepidopteran and coleopteran eggs and larvae were observed in the T 2 treatment, where broad spectrum insecticides were applied three to four times in the absence of nectar-rich flowering plants (Figs 2B,C, 3, 4). Insecticide use reduced the abundance of parasitoids in the field, thus inducing a lower parasitism rate of planthopper, rice hispa and yellow stem borer eggs. Application of insecticides in rice is known to have a negative impact on planthopper natural enemies [17][18][19][20][21][22][23][24]50 . The use or misuse of insecticides can induce planthopper outbreaks in rice fields by lowering the number of parasitoids and predators 2 . Therefore, unsurprisingly the highest number of natural enemies and lowest number of insect pests were observed in the rice field near nectar-rich flowering plants that were not treated with insecticides. The lowest number of natural enemies (parasitoids, predators) and parasitism rates were found in rice fields where insecticides were used in the T 2 treatment. In the control plot (T 3 treatment), withholding insecticide in the absence of nectar-rich flowering plants on the bund resulted in higher parasitism rates and catches of planthopper and yellow stem borer egg parasitoids than those in the T 2 treatment but significantly lower values than those in the T 1 treatment (Figs 2-4). This confirms that the application of insecticides in rice fields can reduce the number of parasitoids and the parasitism rate of the tested insect pest eggs. Gurr et al. 42 reported a similar result in suggested that growing nectar-rich flowering plants on rice bunds can enhanced the population size of natural enemies in rice fields.
Insect pest infestation was also observed in the T 1 plots, but the infestation levels were low and did not exceed the ETL (Figs 2A, 3) 49 . Yellow sticky traps revealed that the eco-engineering plots harbored more parasitoids than the other treatments (Fig. 2D, P = 0.05). Damage symptoms (white head) due to the feeding of YSB was significantly lower in the T 1 plots than in the other plots (T 2 and T 3 ) (Fig. 5, P = 0.05). The highest amount of white head was observed in the T 2 plots, where insecticides were applied prophylactically at 15-day intervals during the crop growing period. This indicates that insecticides cannot stop YSB infestation completely and can even induce higher infestation levels likely due to adverse impacts on natural enemies. Moreover, the application of regular insecticides in rice fields did not improve yield. In plots where insecticides were applied yields were similar to plots where nectar-rich flowering plants were grown on bunds without insecticide and in control plots (Fig. 6, P > 0.05). However, more importantly, this method reduces the application of insecticides and ultimately the amount of pesticide load in rice fields. Our experimental results showed that modification of the existing rice landscape (via eco-engineering) promotes pest control without any yield penalty. In addition, the application of insecticides increased the cost of crop production, ultimately reducing the benefit cost ratio (BCR). The economic analysis (BCR) also confirms that the system we proposed here has an economic benefit. The highest BCR was found in the T 1 treatment (supplementary information, Table S1 and S2). Growing flowering crops on rice bunds without insecticides increased the BCR. In addition to increasing the BCR, the T 1 treatment most importantly improves the environmental quality and safety for humans, animals and biodiversity. Planting nectar-rich plants on bunds may also provide additional income sources to poor rice farmers. In this study, we used marigold, cosmos, sunflower and sesame, but other suitable ail crops such as okra and common bean also have nectar-rich flowers and could provide additional income.

Conclusion
The highest number of natural enemies and percent parasitism of insect pest eggs by parasitoids/parasites in rice fields near nectar-rich flowering plants and the lowest number of natural enemies and parasitism were found in rice fields where insecticides were applied prophylactically. Moreover, there was no yield reduction observed in the rice field surrounded by flowering plants compared to those with insecticide applications. This suggests that farmers could avoid the use of toxic and hazardous insecticides to control insect pests by growing nectar-rich flowering plants on the bunds surrounding rice crops. This approach could help restore rice ecosystems, conserve natural enemies that ultimately help reduce production costs and chemical inputs, improve environmental protection, reduce pest infestations, and reduce labor in terms of pesticide spraying and thus increase net income. In Bangladesh, the land is fragmented into numerous patches of small size with boundaries that can be used to grow flowering plants. This creates more opportunities to use this technology in the rice landscape and the remaining challenge is to spread this technology among farmers in Bangladesh.

Materials and Methods
Study sites. The study was conducted at the main Bangladesh Rice Research Institute (BRRI) research facility at Gazipur (24°0′N, 90°25.5′E), and regional rice field research facility at Rajshahi (24°22′26.40″N, 88°36′4.10″E), Bangladesh. Both study sites contain rice production systems typical across the country. At each facility approximately 90% of the total area is managed for rice production and thus they are representative of the regions rice production landscapes.
www.nature.com/scientificreports www.nature.com/scientificreports/ Experimental design. The experiment was conducted over two consecutive years spanning multiple rice growing seasons. Three rice growing seasons are commonly recognized in Bangladesh termed the Aus, transplanted Aman (hereafter T. Aman) and Boro seasons. The experiments described here inculded the Boro and T. Aman seasons. Three treatments were used in the study. The first treatment had flowering plant borders established on the earthen bunds ( Fig. 1) with no insecticides used on the rice (hereafter T 1 ). In the second treatment, the bund was left fallow and insecticides were applied to rice as per routine farming practices in Bangladesh (i.e. insecticide applied prophylactically at a 15-day intervals (T 2 ). The third treatment was a control with fallow bunds and all insecticide use withheld (T 3 ). The flower plant borders consisted of sunflowers (Helianthus annus, marigolds (Tagetes ssp.), and cosmos (Cosmos spp.), planted on rice bunds in the Boro season, and sesame (Sesamum indicum), planted on the bunds in the T. Aman rice-growing season. Sunflower and marigold were planted @ 15000 plants/ha and 30000 plants/ha respectively. Sesame was broadcasted on the bund @ 2 kg/ha. At each site, the overall experiment included 12 individual plots surrounded by bunds, arranged in 3 blocks and treatments were assigned to plots in a randomized complete block design (n = 3).
Gazipur. The mega rice cultivars, BRRI dhan28, BRRI dhan58 and BRRI dhan52, were cultivated in the Boro 2015-16 and T. Aman 2016 seasons, respectively. According to the respective cultivar production package, 30-to 40-day-old rice seedlings were manually transplanted into fields. Two to three seedlings were transplanted onto a hill at 20 × 20 cm spacing. All fertilizers and irrigation were applied subsequently according to treatment. The application of insecticide in T 2 was initiated 15-days after the date of transplanting using Virtako 40WG (thiamethoxam 20% + chlorantraniliprole 20%) applied at 75 g/ha and subsequently at 15-day intervals (3 times) over the season. Bangladeshi farmers usually apply this insecticide for rice stem borer control. Sampling insect pests and damage. Sweepnet sampling. Insects were sampled in each treatment at the maximum tillering stage of rice. A sweep net (40-cm diameter) was used to sample insect pests and natural enemies in all plots. Twenty complete sweeps were made randomly at the canopy level of the plants. The collected insects were kept in labeled bags and transported to the laboratory for sorting, identification and quantification.
Yellow sticky traps. Yellow sticky traps (20 × 25 cm; Zhangzhou Enjoy Agriculture Technology Co., Ltd., Fujian, China) were also used to record insect pests and their natural enemies in the experimental plots. Individual traps were fixed to bamboo canes and placed in rice plots such that sticky boards were positioned just above the plant canopy. After 48 h, all sticky traps were removed, taken to the laboratory and stored at 4 °C until all insects were counted and identified. This sticky trap was used only in one site due to lack of sufficient number of traps.
Parasitism of key pests. Percent parasitism of key pests was monitored using sentinels eggs of the brown planthopper (BPH), white-backed planthopper (WBPH), rice hispa and yellow stem borer (YSB). To create sentinel egg baits, potted 45-day-old rice plants were exposed to the respective pests in the lab. For the BPH and WBPH, five gravid females were caged on each pot with six plants for 48 h after which adult planthoppers were removed and the plants placed in the experimental plots. Three pots with six plants/pot each were placed in random positions centrally in each plot. For YSB, moths were captured from the field and caged on greenhouse-grown potted rice plants. After 24 h, for oviposition, YSBs were removed and plants were placed in the field in a similar manner to that used for planthopper eggs. Rice hispa egg bait traps were similarly prepared and placed in the field. After 48 h of exposure to parasitoids in the field, the plants were returned to the laboratory and kept in a greenhouse at room temperature for 4 days to allow parasitoid emergence. After this time, the plants were microscopically observed to determine the number of parasitized and nonparasitized eggs.
Damage assessment. The hill counting method 57 was also used to record insect pests, their damage symptoms and the number of natural enemies per hill. During the reproductive stage of rice crops, damage symptoms, termed white head, caused by YSBs were recorded and expressed as % white head. The % white head was calculated using the formula below. Statistical analysis. The goal of this study was to evaluate the effect of flowering plants on rice bunds on pest and their natural enemies in rice fields. Treatment effects on the abundance of predators, parasitoids, insect pests, parasitism and rice yields were analyzed by one-way ANOVA, and a post hoc Tukey's honest significance test (Tukey's HSD). To improve homogeneity, where needed the data were log or arc sine transformed before analysis. In addition, when the value of the data was 0, then transformation was conducted by ln(n + 1). All data were analyzed using SPSS software version 16.0.