Over the last century, global biodiversity loss and species extinction have occurred at an unprecedented rate (Barlow et al., 2016), and agricultural intensification has been one of the major drivers (Tscharntke et al., 2005; Gonthier et al., 2014). One of the features of agricultural intensification is the striking change in land use, in which complex natural ecosystems have been converted to monocultural crop production ecosystems (Tscharntke et al., 2005). This agricultural intensification has multiple consequences for ecosystems, including a decline in natural biocontrol services (Symondson et al., 2002), disruption of crop pollination (Kremen et al., 2012; Kovács-Hostyánszki et al., 2017) and extensive damage to naturally-occurring species and the environment from the heavy use of agrochemicals (Tscharntke et al., 2005; Gill et al., 2012; Stehle and Schulz, 2015).

Ecological intensification is a new paradigm for agriculture (Gaba et al., 2014; Simons and Weisser, 2017), based on intensifying ecological processes in cultivated areas (Caron et al., 2014; Bowles et al., 2017) to replace synthetic, non-renewable and often toxic anthropogenic inputs with biologically- and environmentally-friendly ecosystem services (Tscharntke et al., 2005; Geertsema et al., 2016). The goals are to achieve more sustainable crop production (Bommarco et al., 2013), increase food security, and improve the quality of agricultural products (Martin-Guay et al., 2018).

Biologically-diversified farming can contribute to ecological intensification of agriculture by providing multiple ecosystem services, such as the promotion of biocontrol services (e.g., a decrease in pest abundance and an increase in the abundance of natural enemies of pests) (Harvey et al., 2014; Gurr et al., 2016), a reduction in negative environmental impacts (Tscharntke et al., 2005; Tittonell, 2014; Zhao et al., 2016; Barot et al., 2017), and an increase in crop yields (Cassman, 1999; Tittonell and Giller, 2013; Wittwer et al., 2017). Most previous studies on biodiversity and ecological intensification have focused on smallholder farms (Tittonell and Giller, 2013; Zimmerer, 2013; Zimmerer et al., 2015), while urban agriculture has received little attention.

Starting in the 1950s, urban agriculture, defined as the production of crop and livestock goods within cities and towns, has played an increasing role in ensuring food security for growing urban populations (Zezza and Tasciotti, 2010). Urban agricultural systems have many forms, such as community farms, allotment gardens, rooftop gardens, and edible landscaping (Lin et al., 2015). Confronted with population growth and a shortage of land and natural resources in urban ecosystems, urban agriculture has depended heavily on intensive use of pesticides and fertilizers (Altieri et al., 1999). It has been hypothesized that increased biodiversity in urban agricultural systems has the potential to provide important ecosystem services such as pollination, pest control, and climate resilience, thus reducing the need for chemical inputs (Lin et al., 2015). Replicated, multi-site and long-term studies evaluating the effects of plant-diversified farming on ecosystem services have been lacking.

To reduce urban poverty, enhance food security, improve environmental management and stimulate participatory city governance, the Resource Centre on Urban Agriculture and Food Security (RUAF Foundation) has established pilot demonstration cities for modern urban agriculture worldwide. One of these cities is Shanghai, China. Shanghai has a 30 year history as a pioneer in urban agriculture, which has played an important role in the urban economy, social stability and public security (Kiminami et al., 2006). This mega-urban municipality includes nine suburbs with agriculture, covering an estimated area of 93,333 hectares of rice, accounting for 14.7% of the total land area in Shanghai. Although agriculture contributes only 0.33% of the city’s Gross Domestic Product and increases in productivity have been slower than in other economic sectors, the city authorities have been paying considerable attention to agricultural production to ensure a stable food supply for the huge urban population. In addition, the annual profit from agricultural sightseeing and tourism in Shanghai is 2.4 billion RMB, further increasing the value of the city’s agriculture. Community farms account for 80% of the total agricultural land in Shanghai. Despite strict laws and regulations, these farms have been experiencing increased intensification and mechanization, resulting in biodiversity loss, food safety problems and land degradation.

Although there is evidence that diversifying farming promotes ecological intensification in smallholder agricultural systems (Gurr et al., 2016; Rolando et al., 2017), the ecological and agronomic consequences of increased crop diversity have not been studied in depth in urban agricultural environments. The effects of biodiversity on ecosystem services could be different in an urban setting, because urban farms are not surrounded by other farms or semi-natural vegetation, which can provide organisms that deliver ecosystem services. We hypothesize that diversifying farming in an urban setting will host more insect predators, reducing the densities of herbivorous insects, and allowing reduced input of insecticides without considerable yield loss.

To test our hypothesis, we analyzed monitoring data collected from 34 community farms in nine districts in suburbs of Shanghai, and we conducted two controlled experiments for further verification. In the early 1990s, Shanghai Agricultural Technology Extension and Service Center (SATESC) established 28 community farms in 8 districts of Shanghai to monitor pest occurrence in rice fields. With the help of Chongming Agricultural Technology Extension and Service Center (CATESC), another six community farms in Chongming district were established for monitoring pest occurrence in rice fields (Figure 1; Figure 1—source data 1). Through our own investigation, we found that the six community farms in Chongming district all included diversified plants, and the 28 community farms in the other 8 districts of Shanghai consisted of rice monocultures. The farmers of the six community farms in Chongming district planted soybeans as a border crop around the periphery of each rice field and common vegetable crops (maize, eggplant, Chinese cabbage, etc.) around the whole rice-growing area.

Figure 1

Download asset Open asset

Figure 1—source data 1

Site, rice pest and predator populations, insecticide use, and yield data for comparison of plant-diversified farms (treatment) and mono-rice farms (control) in Shanghai, China.

https://doi.org/10.7554/eLife.35103.003

Download elife-35103-fig1-data1-v1.docx

Figure 1—source data 2

Chemical insecticides applied to control the main insect pests on plant-diversified farms and mono-rice farms in Shanghai, China from 2001 to 2015.

The crosses indicate, for each insecticide, when they were used.

https://doi.org/10.7554/eLife.35103.004

Download elife-35103-fig1-data2-v1.docx

The abundances of the three main rice pests, which have resulted in major crop losses since the late 1990 s, were monitored. The abundance of pink rice borer (Sesamia inferens) and rice brown planthopper (Nilaparvata lugens) in rice fields were evaluated with trapping lamps, and abundance of rice leaf roller (Cnaphalocrocis medinalis) was estimated with field surveys. We also estimated the densities of the main predators of these pests in the rice fields and in the border and neighboring crops of the diversified farms, monitored pesticide use and grain yield, and conducted a continuous economic cost–benefit analysis on both farm types. Complete monitoring data for rice pests, pesticide use and grain yield on these farms started in 2001, and monitoring of predators began at the beginning of 2007. Since the diverse and mono-rice farms were located in different areas of the city, group members from the Shanghai Academy of Agricultural Sciences supplemented the monitoring data with two controlled experiments, in which both treatments were performed at two locations over two years in a complete random design.

The initial establishment of monitoring sites of the community farms was initiated by farmers without knowledge of statistical design, so the establishment of 34 community farms was done without appropriate randomization in the 15 year study. While non-random distribution of the two farming types might have affected the results, it is unlikely that this could account for the main results. All 34 farms belong to Shanghai suburbs and have similar weather conditions and soil types. Because of the lack of randomization in the monitoring study, we performed the two common-location experiments, which verified the results of the monitoring study. Both the monitoring data on community farms and the results of the common-location experiments clearly indicated that increased predator abundance was concomitant with reduced pest abundance, resulting in reduced insecticide application on plant-diversified fields. In all likelihood, the reduced pest abundance contributed to rice growth, development and reproduction, and thus increased rice yield. These results were consistent across years in the monitoring study and in the common-location experiments. The reduction in insecticide application and decreased labor cost of spraying insecticides, together with the slightly higher grain yield on the plant-diversified farms, resulted in an increased economic benefit.

The increased abundance of predators in more diverse fields likely occurred because the higher diversity of plants provided ‘resource pools’ for the predators. The additional plants provided food sources (nectar, pollen, honey dew, etc.), as well as breeding habitats and refuges for these predators, allowing them to increase their longevity and fecundity (Mason et al., 2014; Wan et al., 2016a; Woodcock and Heard, 2011). A previous study indicated that these predators can disperse up to 40 meters, and can therefore move into rice fields from adjacent habitats and crops (Yu et al., 2002). Thus, the most likely explanation is that the increase in the abundance of invertebrate predators spread from border and neighboring crop areas onto the rice fields, increasing predation on pest populations. A similar effect due to neighboring crops was observed in Bt cotton fields (Lu et al., 2012). Diverse crops and weeds may also provide similar advantages for parasitoids, which have been shown to exhibit improved biocontrol services, reducing pests feeding on plant tissues and plant products, such as pollen, nectar, and sap (Gurr et al., 2016; Wäckers et al., 2007).

Ecosystem service providers and service-providing units are concepts that have been used to study ecological intensification (Gurr et al., 2016; Luck et al., 2009). In our study, diverse plants surrounding rice fields can be considered ecosystem service-providing organisms, supporting predators that reduced the populations of pest herbivores. The diverse plant communities initiated a trophic cascade in rice-herbivore-predator interactions, in which increased abundance of predators reduced their herbivores, which in turn resulted in a positive effect on the primary productivity and rice yields, permitting changes in management: a reduction of insecticide inputs and labor. These changes resulted in a small but immediate increase in economic performance (Figure 11). The benefits of plant-diversified farming have not previously been assessed on such a long-term basis. Some studies have reported that plant-diversified farming resulted in lower yields and higher pest densities, leading to criticism of the approach by some scientists and rejection by some farmers (Dassou and Tixier, 2016; Letourneau et al., 2011). Maintaining high crop yields is important if ecological intensification is to be widely accepted (Cassman, 1999; Geertsema et al., 2016; Pywell et al., 2015; Zimmerer, 2013). In our study, we actually observed a slight increase in crop yield and economic performance, while the pesticide use was reduced, enhancing agricultural sustainability.

Figure 11

Download asset Open asset

Organic agriculture has the potential to promote ecological intensification (Crowder et al., 2010; Winqvist et al., 2012), but its adoption has been limited, in large part because organic farming usually produces lower short-term yields (Seufert et al., 2012), although this can be compensated for by higher prices for the products in some markets. Conventional farming has been based on crop monocultures and large inputs of mineral fertilizers and pesticides, making it difficult to provide farmers with feasible strategies for incorporating ecological intensification. Biological diversification can be a first step in promoting ecological intensification of conventional agriculture (Gurr et al., 2016), as it often shows an immediate positive effect through the provisioning of ecosystem services (Bommarco et al., 2013), as demonstrated in our study.

Unlike smallholder farmers, urban farmers tend to be better educated and have a broader vision of farming. They have more contact with and obtain more information from policy makers and scientists, and are therefore more open to new ideas. Thus, ecological intensification of agriculture can be more easily implemented in urban than rural agriculture in China. Support from the government, including new laws and regulations, can facilitate ecological intensification. It is encouraging that 93.8% of urban farmers would like to adopt plant-diversified farming to promote ecological intensification, according to a questionnaire survey (Wan et al., 2016b). There is reason to expect that plant-diversified farming will be practiced extensively in urban agriculture in the near future. One major challenge is the ongoing conversion of agricultural land, sometimes very high quality agricultural land, to housing, industry and commercial use throughout the world (Singh and Swami, 2015).

Our results on pests and pesticide application resulting from the introduction of border and neighboring crops, without a highly-developed theory or a vast amount of quantitative data on the relationship between plant diversity and ecosystem services, suggest that the potential for ecological intensification through increased plant diversity is much greater than what we have shown here. A major research effort in ecological intensification could result in revolutionary changes in agriculture, in which pesticide application can be reduced drastically or even eliminated, improving food safety and environmental quality off as well as on farms. Urban agriculture can serve a laboratory for such research, while providing an important source of food for city dwellers. We urge researchers to focus on forms of biotechnology that can contribute to ecological intensification, which has a great potential to conserve and promote biodiversity, while contributing to the provision of food to a growing human population (Pywell et al., 2015; Zhao et al., 2016; Weiner, 2017), rather than on forms of biotechnology that increase unsustainable intensification.

All data generated or analysed during this study are included in the manuscript and supporting files.

1. 1. Altieri MA
   2. Companioni N
   3. Cañizares K
   4. Murphy C
   5. Rosset P
   6. Bourque M
   7. Nicholls CI

   (1999) The greening of the “barrios”: urban agriculture for food security in Cuba

   Agriculture and Human Values 16:131–140.

   https://doi.org/10.1023/A:1007545304561

   • Google Scholar
2. 1. Barlow J
   2. Lennox GD
   3. Ferreira J
   4. Berenguer E
   5. Lees AC
   6. Mac Nally R
   7. Thomson JR
   8. Ferraz SF
   9. Louzada J
   10. Oliveira VH
   11. Parry L
   12. Solar RR
   13. Vieira IC
   14. Aragão LE
   15. Begotti RA
   16. Braga RF
   17. Cardoso TM
   18. de Oliveira RC
   19. Souza CM
   20. Moura NG
   21. Nunes SS
   22. Siqueira JV
   23. Pardini R
   24. Silveira JM
   25. Vaz-de-Mello FZ
   26. Veiga RC
   27. Venturieri A
   28. Gardner TA

   (2016) Anthropogenic disturbance in tropical forests can double biodiversity loss from deforestation

   Nature 535:144–147.

   https://doi.org/10.1038/nature18326

   • PubMed
   • Google Scholar
3. 1. Barot S
   2. Allard V
   3. Cantarel A
   4. Enjalbert J
   5. Gauffreteau A
   6. Goldringer I
   7. Lata J-C
   8. Le Roux X
   9. Niboyet A
   10. Porcher E

   (2017) Designing mixtures of varieties for multifunctional agriculture with the help of ecology. A review

   Agronomy for Sustainable Development 37:13.

   https://doi.org/10.1007/s13593-017-0418-x

   • Google Scholar
4. 1. Bolker BM
   2. Brooks ME
   3. Clark CJ
   4. Geange SW
   5. Poulsen JR
   6. Stevens MH
   7. White JS

   (2009) Generalized linear mixed models: a practical guide for ecology and evolution

   Trends in Ecology & Evolution 24:127–135.

   https://doi.org/10.1016/j.tree.2008.10.008

   • PubMed
   • Google Scholar
5. 1. Bommarco R
   2. Kleijn D
   3. Potts SG

   (2013) Ecological intensification: harnessing ecosystem services for food security

   Trends in Ecology & Evolution 28:230–238.

   https://doi.org/10.1016/j.tree.2012.10.012

   • PubMed
   • Google Scholar
6. 1. Bowles TM
   2. Jackson LE
   3. Loeher M
   4. Cavagnaro TR

   (2017) Ecological intensification and arbuscular mycorrhizas: a meta-analysis of tillage and cover crop effects

   Journal of Applied Ecology 54:1785–1793.

   https://doi.org/10.1111/1365-2664.12815

   • Google Scholar
7. 1. Caron P
   2. Biénabe E
   3. Hainzelin E

   (2014) Making transition towards ecological intensification of agriculture a reality: the gaps in and the role of scientific knowledge

   Current Opinion in Environmental Sustainability 8:44–52.

   https://doi.org/10.1016/j.cosust.2014.08.004

   • Google Scholar
8. 1. Cassman KG

   (1999) Ecological intensification of cereal production systems: yield potential, soil quality, and precision agriculture

   PNAS 96:5952–5959.

   https://doi.org/10.1073/pnas.96.11.5952

   • PubMed
   • Google Scholar
9. 1. Cheng J

   (2001) The seasonal occurrence, forecasting and management strategy for Brown planthopper in Zhejiang, China

   Journal of Asia-Pacific Entomology 4:209–215.

   https://doi.org/10.1016/S1226-8615(08)60125-9

   • Google Scholar
10. 1. Crowder DW
    2. Northfield TD
    3. Strand MR
    4. Snyder WE

    (2010) Organic agriculture promotes evenness and natural pest control

    Nature 466:109–112.

    https://doi.org/10.1038/nature09183

    • PubMed
    • Google Scholar
11. 1. Dassou AG
    2. Tixier P

    (2016) Response of pest control by generalist predators to local-scale plant diversity: a meta-analysis

    Ecology and Evolution 6:1143–1153.

    https://doi.org/10.1002/ece3.1917

    • PubMed
    • Google Scholar
12. 1. Gaba S
    2. Bretagnolle François
    3. Rigaud T
    4. Philippot L

    (2014) Managing biotic interactions for ecological intensification of agroecosystems

    Frontiers in Ecology and Evolution 2:29.

    https://doi.org/10.3389/fevo.2014.00029

    • Google Scholar
13. 1. Geertsema W
    2. Rossing WAH
    3. Landis DA
    4. Bianchi FJJA
    5. van Rijn PCJ
    6. Schaminée JHJ
    7. Tscharntke T
    8. van der Werf W

    (2016) Actionable knowledge for ecological intensification of agriculture

    Frontiers in Ecology and the Environment 14:209–216.

    https://doi.org/10.1002/fee.1258

    • Google Scholar
14. 1. Gill RJ
    2. Ramos-Rodriguez O
    3. Raine NE

    (2012) Combined pesticide exposure severely affects individual- and colony-level traits in bees

    Nature 491:105–108.

    https://doi.org/10.1038/nature11585

    • PubMed
    • Google Scholar
15. 1. Gonthier DJ
    2. Ennis KK
    3. Farinas S
    4. Hsieh H-Y
    5. Iverson AL
    6. Batary P
    7. Rudolphi J
    8. Tscharntke T
    9. Cardinale BJ
    10. Perfecto I

    (2014) Biodiversity conservation in agriculture requires a multi-scale approach

    Proceedings of the Royal Society B: Biological Sciences 281:20141358.

    https://doi.org/10.1098/rspb.2014.1358

    • Google Scholar
16. 1. Green P
    2. MacLeod CJ

    (2016) SIMR : an R package for power analysis of generalized linear mixed models by simulation

    Methods in Ecology and Evolution 7:493–498.

    https://doi.org/10.1111/2041-210X.12504

    • Google Scholar
17. 1. Gurr GM
    2. Lu Z
    3. Zheng X
    4. Xu H
    5. Zhu P
    6. Chen G
    7. Yao X
    8. Cheng J
    9. Zhu Z
    10. Catindig JL
    11. Villareal S
    12. Van Chien H
    13. Cuong leQ
    14. Channoo C
    15. Chengwattana N
    16. Lan LP
    17. Hai leH
    18. Chaiwong J
    19. Nicol HI
    20. Perovic DJ
    21. Wratten SD
    22. Heong KL

    (2016) Multi-country evidence that crop diversification promotes ecological intensification of agriculture

    Nature Plants 2:16014.

    https://doi.org/10.1038/nplants.2016.14

    • PubMed
    • Google Scholar
18. 1. Halekoh U
    2. Højsgaard S

    (2014) A Kenward-Roger approximation and parametric bootstrap methods for tests in linear mixed models - The R Package pbkrtest

    Journal of Statistical Software 59:1–32.

    https://doi.org/10.18637/jss.v059.i09

    • Google Scholar
19. 1. Harvey CA
    2. Chacón M
    3. Donatti CI
    4. Garen E
    5. Hannah L
    6. Andrade A
    7. Bede L
    8. Brown D
    9. Calle A
    10. Chará J
    11. Clement C
    12. Gray E
    13. Hoang MH
    14. Minang P
    15. Rodríguez AM
    16. Seeberg-Elverfeldt C
    17. Semroc B
    18. Shames S
    19. Smukler S
    20. Somarriba E
    21. Torquebiau E
    22. van Etten J
    23. Wollenberg E

    (2014) Climate-Smart landscapes: opportunities and challenges for integrating adaptation and mitigation in tropical agriculture

    Conservation Letters 7:77–90.

    https://doi.org/10.1111/conl.12066

    • Google Scholar
20. 1. Higley LG
    2. Pedigo LP

    (1993) Economic injury level concepts and their use in sustaining environmental quality

    Agriculture, Ecosystems & Environment 46:233–243.

    https://doi.org/10.1016/0167-8809(93)90027-M

    • Google Scholar
21. 1. Kiminami LY
    2. Kiminami A
    3. Zhu M

    (2006) Sustainability of urban agriculture in Shanghai, China

    Chiikigaku Kenkyu 36:725–739.

    https://doi.org/10.2457/srs.36.725

    • Google Scholar
22. 1. Kovács-Hostyánszki A
    2. Espíndola A
    3. Vanbergen AJ
    4. Settele J
    5. Kremen C
    6. Dicks LV

    (2017) Ecological intensification to mitigate impacts of conventional intensive land use on pollinators and pollination

    Ecology Letters 20:673–689.

    https://doi.org/10.1111/ele.12762

    • PubMed
    • Google Scholar
23. 1. Kremen C
    2. Iles A
    3. Bacon C

    (2012) Diversified farming systems: an agroecological, Systems-based alternative to modern industrial agriculture

    Ecology and Society 17:44.

    https://doi.org/10.5751/ES-05103-170444

    • Google Scholar
24. 1. Letourneau DK
    2. Armbrecht I
    3. Rivera BS
    4. Lerma JM
    5. Carmona EJ
    6. Daza MC
    7. Escobar S
    8. Galindo V
    9. Gutiérrez C
    10. López SD
    11. Mejía JL
    12. Rangel AM
    13. Rangel JH
    14. Rivera L
    15. Saavedra CA
    16. Torres AM
    17. Trujillo AR

    (2011) Does plant diversity benefit agroecosystems? A synthetic review

    Ecological Applications 21:9–21.

    https://doi.org/10.1890/09-2026.1

    • PubMed
    • Google Scholar
25. 1. Lin BB
    2. Philpott SM
    3. Jha S

    (2015) The future of urban agriculture and biodiversity-ecosystem services: Challenges and next steps

    Basic and Applied Ecology 16:189–201.

    https://doi.org/10.1016/j.baae.2015.01.005

    • Google Scholar
26. 1. Lu Y
    2. Wu K
    3. Jiang Y
    4. Guo Y
    5. Desneux N

    (2012) Widespread adoption of Bt cotton and insecticide decrease promotes biocontrol services

    Nature 487:362–365.

    https://doi.org/10.1038/nature11153

    • PubMed
    • Google Scholar
27. 1. Luck GW
    2. Harrington R
    3. Harrison PA
    4. Kremen C
    5. Berry PM
    6. Bugter R
    7. Dawson TP
    8. de Bello F
    9. Díaz S
    10. Feld CK
    11. Haslett JR
    12. Hering D
    13. Kontogianni A
    14. Lavorel S
    15. Rounsevell M
    16. Samways MJ
    17. Sandin L
    18. Settele J
    19. Sykes MT
    20. van den Hove S
    21. Vandewalle M
    22. Zobel M

    (2009) Quantifying the contribution of organisms to the provision of ecosystem services

    BioScience 59:223–235.

    https://doi.org/10.1525/bio.2009.59.3.7

    • Google Scholar
28. 1. Martin-Guay MO
    2. Paquette A
    3. Dupras J
    4. Rivest D

    (2018) The new green revolution: sustainable intensification of agriculture by intercropping

    Science of the Total Environment 615:767–772.

    https://doi.org/10.1016/j.scitotenv.2017.10.024

    • PubMed
    • Google Scholar
29. 1. Mason PA
    2. Bernardo MA
    3. Singer MS

    (2014) A mixed diet of toxic plants enables increased feeding and anti-predator defense by an insect herbivore

    Oecologia 176:477–486.

    https://doi.org/10.1007/s00442-014-3029-0

    • PubMed
    • Google Scholar
30. 1. Pywell RF
    2. Heard MS
    3. Woodcock BA
    4. Hinsley S
    5. Ridding L
    6. Nowakowski M
    7. Bullock JM

    (2015) Wildlife-friendly farming increases crop yield: evidence for ecological intensification

    Proceedings of the Royal Society B: Biological Sciences 282:20151740.

    https://doi.org/10.1098/rspb.2015.1740

    • Google Scholar
31. 1. Rolando JL
    2. Turin C
    3. Ramírez DA
    4. Mares V
    5. Monerris J
    6. Quiroz R

    (2017) Key ecosystem services and ecological intensification of agriculture in the tropical high-Andean Puna as affected by land-use and climate changes

    Agriculture, Ecosystems & Environment 236:221–233.

    https://doi.org/10.1016/j.agee.2016.12.010

    • Google Scholar
32. 1. Seufert V
    2. Ramankutty N
    3. Foley JA

    (2012) Comparing the yields of organic and conventional agriculture

    Nature 485:229–232.

    https://doi.org/10.1038/nature11069

    • PubMed
    • Google Scholar
33. 1. Simons NK
    2. Weisser WW

    (2017) Agricultural intensification without biodiversity loss is possible in grassland landscapes

    Nature Ecology & Evolution 1:1136–1145.

    https://doi.org/10.1038/s41559-017-0227-2

    • PubMed
    • Google Scholar
34. 1. Singh K
    2. Swami MS

    (2015)

    Vision of urban agriculture towards biodiversity richness

    Journal of Agroecology and Natural Resource Management 2:290–295.

    • Google Scholar
35. 1. Stehle S
    2. Schulz R

    (2015) Agricultural insecticides threaten surface waters at the global scale

    PNAS 112:5750–5755.

    https://doi.org/10.1073/pnas.1500232112

    • PubMed
    • Google Scholar
36. 1. Symondson WO
    2. Sunderland KD
    3. Greenstone MH

    (2002) Can generalist predators be effective biocontrol agents?

    Annual Review of Entomology 47:561–594.

    https://doi.org/10.1146/annurev.ento.47.091201.145240

    • PubMed
    • Google Scholar
37. 1. Tittonell P
    2. Giller KE

    (2013) When yield gaps are poverty traps: the paradigm of ecological intensification in African smallholder agriculture

    Field Crops Research 143:76–90.

    https://doi.org/10.1016/j.fcr.2012.10.007

    • Google Scholar
38. 1. Tittonell P

    (2014) Ecological intensification of agriculture—sustainable by nature

    Current Opinion in Environmental Sustainability 8:53–61.

    https://doi.org/10.1016/j.cosust.2014.08.006

    • Google Scholar
39. 1. Tscharntke T
    2. Klein AM
    3. Kruess A
    4. Steffan-Dewenter I
    5. Thies C

    (2005) Landscape perspectives on agricultural intensification and biodiversity-ecosystem service management

    Ecology Letters 8:857–874.

    https://doi.org/10.1111/j.1461-0248.2005.00782.x

    • Google Scholar
40. 1. Wan N-F
    2. Ji X-Y
    3. Cao L-M
    4. Jiang J-X

    (2015) The occurrence of rice leaf roller, Cnaphalocrocis medinalis Guenée in the large-scale agricultural production on Chongming Eco-island in China

    Ecological Engineering 77:37–39.

    https://doi.org/10.1016/j.ecoleng.2015.01.006

    • Google Scholar
41. 1. Wan N-F
    2. Zhang Y-M
    3. Yao Y-J
    4. Ji X-Y
    5. Zhang H
    6. Jiang J-X

    (2016a) The occurrence of rice planthoppers in diversified farming on Chongming Eco-Island in China

    Ecological Engineering 95:50–53.

    https://doi.org/10.1016/j.ecoleng.2016.06.012

    • Google Scholar
42. 1. Wan NF
    2. Zhang MY
    3. Jiang JX
    4. Ji XY
    5. Hao-Zhang

    (2016b) An ecological method to understand agricultural standardization in peach orchard ecosystems

    Scientific Reports 6:21675.

    https://doi.org/10.1038/srep21675

    • PubMed
    • Google Scholar
43. 1. Wäckers FL
    2. Romeis J
    3. van Rijn P

    (2007) Nectar and pollen feeding by insect herbivores and implications for multitrophic interactions

    Annual Review of Entomology 52:301–323.

    https://doi.org/10.1146/annurev.ento.52.110405.091352

    • PubMed
    • Google Scholar
44. 1. Weiner J

    (2017) Applying plant ecological knowledge to increase agricultural sustainability

    Journal of Ecology 105:865–870.

    https://doi.org/10.1111/1365-2745.12792

    • Google Scholar
45. 1. Winqvist C
    2. Ahnström J
    3. Bengtsson J

    (2012) Effects of organic farming on biodiversity and ecosystem services: taking landscape complexity into account

    Annals of the New York Academy of Sciences 1249:191–203.

    https://doi.org/10.1111/j.1749-6632.2011.06413.x

    • PubMed
    • Google Scholar
46. 1. Wittwer RA
    2. Dorn B
    3. Jossi W
    4. van der Heijden MG

    (2017) Cover crops support ecological intensification of arable cropping systems

    Scientific Reports 7:41911.

    https://doi.org/10.1038/srep41911

    • PubMed
    • Google Scholar
47. 1. Woodcock BA
    2. Heard MS

    (2011) Disentangling the effects of predator hunting mode and habitat domain on the top-down control of insect herbivores

    Journal of Animal Ecology 80:495–503.

    https://doi.org/10.1111/j.1365-2656.2010.01790.x

    • PubMed
    • Google Scholar
48. 1. Yu X
    2. Zheng X
    3. Xu H
    4. Lu Z
    5. Chen J
    6. Tao L

    (2002)

    A study on the dispersal of lycosid spider, Pirata subpiraticus between rice and Zizania fields

    Acta Entomologica Sinica 45:636–640.

    • Google Scholar
49. 1. Zezza A
    2. Tasciotti L

    (2010) Urban agriculture, poverty, and food security: empirical evidence from a sample of developing countries

    Food Policy 35:265–273.

    https://doi.org/10.1016/j.foodpol.2010.04.007

    • Google Scholar
50. 1. Zhao R
    2. He P
    3. Xie J
    4. Johnston AM
    5. Xu X
    6. Qiu S
    7. Zhao S

    (2016) Ecological intensification management of maize in northeast China: agronomic and environmental response

    Agriculture, Ecosystems & Environment 224:123–130.

    https://doi.org/10.1016/j.agee.2016.03.038

    • Google Scholar
51. 1. Zimmerer KS
    2. Carney JA
    3. Vanek SJ

    (2015) Sustainable smallholder intensification in global change? Pivotal spatial interactions, gendered livelihoods, and agrobiodiversity

    Current Opinion in Environmental Sustainability 14:49–60.

    https://doi.org/10.1016/j.cosust.2015.03.004

    • Google Scholar
52. 1. Zimmerer KS

    (2013) The compatibility of agricultural intensification in a global hotspot of smallholder agrobiodiversity (Bolivia)

    PNAS 110:2769–2774.

    https://doi.org/10.1073/pnas.1216294110

    • PubMed
    • Google Scholar

[Editors’ note: a previous version of this study was rejected after peer review, but the authors submitted for reconsideration. The first decision letter after peer review is shown below.]

Thank you for submitting your work entitled "Plant diversification contributes to ecological intensification of mega-urban agriculture" for consideration by eLife. Your article has been evaluated by a Senior Editor and three reviewers, one of whom, Bernhard Schmid (Reviewer #1), is a member of our Board of Reviewing Editors.

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife.

As mentioned at the pre-submission stage, we found your study potentially very relevant and interesting for a broad readership because it seems to show that farm-level plant diversity may increase crop protection by encouraging predators of pests such that less chemical pest control will be necessary without yield losses.

However, our problems with the full submission are still the same as with the pre-submission, namely that you do not clearly enough explain the lack of randomization in the 15-year study and you continue to use inappropriate, simple statistics. These problems unfortunately were seen by all reviewers, who in addition point out further issues, in part related to the lack of appropriate analysis. In particular, one reviewer mentions the possibility that the causality between reduced insecticide use and increased predator level could go either direction.

I hope that with the extensive comments of the reviewers you can further improve your manuscript for submission to another journal.

Reviewer #1:

As mentioned at the pre-submission stage, I find this study potentially very relevant and interesting for a broad readership because it shows that farm-level plant diversity may increase crop protection by encouraging predators of pests such that less chemical pest control will be necessary without yield losses.

My problems with the full submission remain, namely that the authors do not clearly enough explain the lack of randomization in the 15-year study (was it their choice not to randomize or was it because they met a "natural experiment" that had been started by others with lack of knowledge about statistical design) and use inappropriate, simple statistics. The first problem may be resolved by explaining in more detail who prepared the study and why it was not randomized. The second problem can be easily resolved by requesting the help of someone who knows how to properly analyze data from such experiments as the authors carried out.

One possible statistical model for the 15-year study is:

y ~ farm type + farm identity within farm type

+ year linear contrast + year as factor

+ farm type x linear year + farm identity x linear year

+ farm type x factorial year

Here "farm identity within farm type" is a random-effects term and must be included in the model to serve as an error term for "farm type" (which is now tested wrongly against the residual). Similarly, "farm identity x linear year" can be used as error term to test "farm type x linear year" if the authors decide to make this probably useful linear contrast.

It seems the authors also used t-tests in which they used the year-averages of y for the two farm types. However, this is not allowed because those are repeated measures of the same set of control and treatment farms. It is like comparing the height of a single tree given nutrients with that of a single control tree, using 15 years of growth data as replicates to test if there are significant effects of nutrients on tree height in more than the two particular tree individuals.

The same problem occurs with the randomized experiment, where the paired-sample t-test is again wrongly applied and the 2-way Anova too simplistic because it ignores "hierarchical structure", i.e. farm identity as random-effects factor. Here a possible statistical model is:

y ~ site + farm type + farm identity within farm type

+ year + farm type x year

Also, several speculative statements in the Discussion could be supported with more sophisticated statistical analysis of the data. In particular, the authors should use the fact that they have evaluated different variables at the same farms to calculate multiple linear regressions or include co-variates in the above models and perhaps integrate these analyses all into a structural equation model, that will allow the authors to explore their hypotheses how the different variables affected each other in a path-analytic diagram. At present these are speculations, even though the authors often present them as facts in the Discussion.

Reviewer #2:

Nian-Feng Wan and colleagues studied the impact of diversified rice cropping on pests, their predators and yield (including economic benefit) in 34 community farms in Shanghai, China. They found that rice fields with ridges planted by soybean and guards surrounding the crop area planted by maize, eggplant and Chinese cabbage, i.e. diversified crop fields, showed reduced pest infestation, higher predator abundance, less pesticide use and occasionally even higher yield. Economically, the diversified crop fields where more beneficial than conventional ones.

The study is based on an impressive dataset, including over 10 years of data from all the community farms, with high temporal and spatial resolution. Furthermore, due to the non-random distribution of the two farming types in Shanghai, the authors also performed a well-designed experiment to test if the results obtained from the long-term farm observations hold under experimentally controlled conditions. I therefore think that this is a very well conducted and extremely well-written manuscript. I enjoyed reading and I think this could be of interest to a wide audience of researchers and stakeholders, and therefore be potentially of high impact.

My main concerns are related to (1) the selection of the farms, (2) the set-up of the diversified farms, and (3) the insecticide use.

1) It is not clear to me how the farms were selected for this study. Some more information on the selection procedure would be highly appreciated. The sampling design is very unbalanced (28 conventional vs. 6 diversified farms) and not well distributed across Shanghai (the 6 diversified farms were all located on Chongming Island whereas none of the 28 conventional farms was located in this suburb). I guess there must be a good reason for this, which would clarify this slightly awkward sampling design.

2) It is unclear to me, how the 6 diversified farms were selected, in particular if these farms were set-up as such through an initiative by those farmers before the observations started or if they were set up specifically for this study through initiative of the researchers. In the latter case, the selection would be highly questionable and could potentially confound most of the results.

3) The results of the insecticide use (in particular the number of sprays) seem awkward, given that farmers "applied pesticides according to pest forecast information offered by the Plant Protection Station". I guess this forecast is the same for all farmers and would therefore suggest the same application regime for all farms. I therefore wonder to what extent the differences in pesticide use are simply caused by the contrasting interests of the different farmer communities. A proper pesticide use study would rather apply similar amounts of pesticide and assess their impact on the pest or reduce the pest to similar amounts and quantify pesticide use. However, in this study it seems that both pesticide use and pest abundance varied among farming types, resulting in difficulties to infer sound conclusions.

Reviewer #3:

The study analyses a long-term dataset of functional arthropod groups of herbivores and predators (herbivores: three different species of specialist herbivores on rice crop; predators: three generalist predators groups, namely ladybird beetles, lacewings and spiders), replicated over multiple farms and several years. In addition, they have data on insecticide use and yield. The core idea is to test whether a diversified farming system leads to an increased top-down control compared to a monoculture situation, followed by a reduced insecticide use and increased yield. They analysed a total of 6 diversified farms and compared it to 28 control farms. One farm was approximately 2 ha in size. The diversified farm consisted of 9 rice fields surrounded by a stripe of soybeans and an adjacent stripe of another crop. Accordingly, the monoculture farms had 9 rice fields which were surrounded by bare ground. I agree with the authors that we have a lack of knowledge on factors driving ecological intensification in urban systems. However, I have a couple of major concerns with the study.

First, the authors claim that the reduced insecticide use on diversified farms was a consequence of the increased top-down control due to increased predator densities. I wonder how they can rule out that it is not the opposite mechanism, namely that there was a reduced insecticide use on the diversified farms and as a consequence there was an increased predator density which then led to increased top-down control? In other words, how can they rule out that the result they find is not simply a consequence of reduced insecticide use but an effect of having multiple crops on a farm (i.e. diversified farm)?

A second major concern is the statistical approach used. First, the data are clearly time-series with nicely oscillating densities indicating lag-effects (Figure 2). In addition to lag-effects, the models should include more variables. For example, for explaining predator densities it would be important to incorporate the prey densities in the model. Second, the model (no matter whether it is a time-series analysis or an ANOVA) needs to incorporate the fact that repeated measures were obtained from the same location. Also, to analyse long-term data with multiple t-tests (for each year one test) is not appropriate. Third, the description of the statistical methods is confusing and incomplete.

A third major concern is that the main conclusion of higher predator control is not inferred from an experiment (for example, by exposing a standardized number of prey to predators) but based on a correlation.

Finally, the Introduction should be more put into a theoretical context, i.e. why and how do diversified farming systems drive biodiversity and ecosystem functioning. Further, the quality of the literature cited should be improved and all literature cited should be included in the reference list, e.g. Tscharntke et al., 2005.

[Editors’ note: what now follows is the decision letter after the authors submitted for further consideration.]

Thank you for resubmitting your work entitled "Plant diversity promotes biocontrol services, reduces insecticide use and increases rice yield in urban agriculture" for further consideration at eLife. Your revised article has been evaluated by Ian Baldwin (Senior editor), a Reviewing editor, and two reviewers.

The Reviewing editor and the reviewer who have seen the previous manuscript agree that the new submission has been improved. It was difficult to find a replacement for one of the previous reviewers but we did get a review of an expert with a similar research area, i.e. with knowledge about the more zoological aspects of your study. This new reviewer has similar concerns to the previous zoological reviewer about lack of detail in the methods description. We hope that you can add all the requested detail in a revision. Please also check the further comments by the Reviewing Editor and the previous reviewer as outlined below.

Reviewing Editor (Reviewer #1):

This resubmission of a previously rejected manuscript has been drastically improved by considering the reviewer comments and obtaining professional statistical help from two additional co-authors. Thus, the main message of the paper, that diverse margins around rice fields can reduce pests, increase predators, reduce insecticide use, marginally increase rice yield and increase economic benefit is now well supported by data and analysis. This is a very important contribution to the new topic of "ecological intensification" in the new field of "urban agriculture".

I have no further major comments. It would be good if co-author Jacob Weiner could have a very careful look at the manuscript before the revised version is uploaded again.

Reviewer #2:

The revised version of this manuscript is clearly improved and, in my opinion, the authors satisfactorily addressed the previous concerns by all the reviewers. I think this is an interesting study on the benefits of diversification in urban agriculture and shows results of an impressive amount of data. The design of the observational study is not perfect, but the authors managed to alleviate my main concerns with this. I therefore have no further major comments.

Reviewer #3:

General comments:

The topic of the paper is very important. The paper basically states that if plants (in fact, other crops) are planted around rice fields on the dams surrounding the fields, this reduces the need to use insecticides.

I concur with the reviews on the first version of the manuscript – the details of the study, in particular the experimental design and the statistical analysis were unclear. I did not assess the statistical analysis in detail.

Main points:

1) I find the title of the paper misleading. The manuscript does not analyse if increasing plant species richness increases biocontrol services etc., because the authors did not analyse diversity in the surroundings of the rice fields. Basically, they show that going away from the monoculture can be beneficial. Thus, the title should be "diversifying agriculture…" or "planting border crops enhances…"

2) Nowhere in the Results are a test statistic or the number of replicates are given.

3) Materials and methods: it is unclear how information on insecticide use was obtained.

4) Materials and methods: unfortunately, I find that still many details are unclear in the revision, see below.

5) Results: the effect on yield is only marginally significant and should not be overstated. Also, the authors provide no mechanism why yield was higher – do the insecticides not work?

Materials and methods:

Subsection “Monitoring sites”: the entire section should be restructured, the design of the diversified fields is described twice (first and third paragraphs), the same is true for other information. Start with the region, then describe the farms, then how rice is grown and then the different treatments.

Subsection “Monitoring sites”, first paragraph: please provide a drawing of a diversified field, and a photograph, how large is the area where other crops are grown.

Subsection “Monitoring sites”, second paragraph: the pest control guidance by SATESC is not known to the reader – what are the insecticides used, what active ingredients etc.?

Subsection “Monitoring sites”, last paragraph: what is a "Z" subplot? Again, use a drawing.

Subsection “Common location experiments”, first paragraph: design unclear: what is the replicate – a block, i.e. three pairs of?

Subsection “Common location experiments”, fourth paragraph etc.: Please provide more detail of the pest assessments: I assume the light kills insects attracted to the light? Were they collected in alcohol? Who identified the species? Why was the interval for the planthopper different if the same lamp in the same field is used? When you say each rice field, do you mean one in each of 9 plots per farm, or just one?

Subsection “Monitoring and sampling methods”, last paragraph: did you only count predators on plants?

Subsection “Common location experiments”, fourth paragraph: what is the correlation between number of pests and number of predators? I assume it is positive? Did you also count lacewing larvae?

I also do not understand how the farmers decided when to spray – you state that there are economic thresholds, so there must be data on how often the threshold was reached? Were the pest data used for the threshold analysis the same that you used?

Figures:

Figure 1: I thought the 28 farms were mono-rice and only 6 diversified, not the other way round?

Figure 2: were pest insects sampled all year round, or only from June onwards? This is not stated in the manuscript. In the first paragraph of the subsection “Monitoring and sampling methods” it states sampling was done May to September (April to September).

Figure 5: 9-14 kg insecticide per hectare is a lot – what is this number referring to, the liquid that is brought to the field? Details on the insecticides should be given and the amount plotted in active ingredient rather than some other unit.

[Editors’ note: the author responses to the first round of peer review follow.]

We thank the editors for their comments and suggestions, which helped us greatly to improve the quality of the manuscript. We have carefully studied the reviews and revised our manuscript accordingly.

Reviewer #1:

As mentioned at the pre-submission stage, I find this study potentially very relevant and interesting for a broad readership because it shows that farm-level plant diversity may increase crop protection by encouraging predators of pests such that less chemical pest control will be necessary without yield losses.

My problems with the full submission remain, namely that the authors do not clearly enough explain the lack of randomization in the 15-year study (was it their choice not to randomize or was it because they met a "natural experiment" that had been started by others with lack of knowledge about statistical design) and use inappropriate, simple statistics. The first problem may be resolved by explaining in more detail who prepared the study and why it was not randomized.

We thank the editor for these important comments. We now explain fully the origin and background of the study, and the nature of the data. The answers to the question are found in the revised manuscript:

“To test our hypothesis, we refer to monitoring data collected from 34 community farms in 9 districts in suburbs of Shanghai, and we conducted two more controlled experiments for further verification. […] Since the diverse and mono-rice farms were located in different areas of the city, group members from the Shanghai Academy of Agricultural Sciences and Fudan University supplemented the monitoring data with two controlled experiments, in which both treatments were performed at two locations over two years in a complete random design.”

The second problem can be easily resolved by requesting the help of someone who knows how to properly analyze data from such experiments as the authors carried out.

One possible statistical model for the 15-year study is:

y ~ farm type + farm identity within farm type

+ year linear contrast + year as factor

+ farm type x linear year + farm identity x linear year

+ farm type x factorial year

Here "farm identity within farm type" is a random-effects term and must be included in the model to serve as an error term for "farm type" (which is now tested wrongly against the residual). Similarly, "farm identity x linear year" can be used as error term to test "farm type x linear year" if the authors decide to make this probably useful linear contrast.

It seems the authors also used t-tests in which they used the year-averages of y for the two farm types. However, this is not allowed because those are repeated measures of the same set of control and treatment farms. It is like comparing the height of a single tree given nutrients with that of a single control tree, using 15 years of growth data as replicates to test if there are significant effects of nutrients on tree height in more than the two particular tree individuals.

The model suggested by the reviewer has formed the basis of the re-analyses we had carried out. An appropriate statistical model for testing the farm-type effect should account for two design aspects of the 15-year monitoring studies: that farm identity is nested within farm type, and that there are longitudinal repeated measurements within each farm identity across the 15-year study period. To this end, we have re-analyzed the monitoring data based on the following model:

yijk=β0+β1farmTypei+β2yearFactorialj+β12farmTypei×yearFactorialj+μijk

μijk=b0,ik+b1,ikyearj+b2,ikyearj2+b3,ikyearj3+ϵijk,

i nested within farm type i. Finally, ϵijk is the random noise term for each individual measurement. Note that while year is a continuous term in the random-effects component, year is treated as a categorical factor (yearFactorial) in the fixed-effect component of the model.

We have applied a 3rd-degree polynomial transformation for the year variable as we have observed non-linear yearly trends among the various endpoints considered. This leads to a mixed-effect model with varying intercepts and varying coefficients for the polynomial year terms.

In the random-effects component term μijkfrom the model above, the random intercept term and the polynomial year terms are used to model the farm-specific effects and the farm-specific yearly trends. Together with the individual noise term ϵijk, the random component μijk is the error term used for testing the farm type x year interaction effect and the farm type main effect in the mixed-effects model.

To determine the significance of the farm type effect and the farm type x year interaction effect, we have employed a model comparison test, where the full model above was first compared against the model without the interaction terms. If the interaction effect was not significant, the interaction terms were removed and we then tested whether the farm type factor can be further removed from the model. For the test of significance, we have chosen the parametric bootstrap tests for mixed-effect model comparison. This test is preferable for small sample and/or unbalanced data, and was chosen here as the 15-year monitoring data was unbalanced (6 plant-diversified locations and 8 mono-rice locations).

A brief description of the above model has been added to the revised article (subsection “Data analysis”, first paragraph).

The use of the bootstrap test is also described in the revised manuscript (subsection “Data analysis”, third paragraph).

The same problem occurs with the randomized experiment, where the paired-sample t-test is again wrongly applied and the 2-way Anova too simplistic because it ignores "hierarchical structure", i.e. farm identity as random-effects factor. Here a possible statistical model is:

y ~ site + farm type + farm identity within farm type

+ year + farm type x year

Once again we thank the reviewer for pointing out the issues and the suggested model refinement. We have devised another mixed-effects model tailored for assessing the significance of the farm type effect while accounting for the design aspects of the common-location experiments. In detail, the common-location experiments involve two farms, with two consecutive years nested within each farm. For each experimental year, both farms contained two treatments, and three replicated measurements were retrieved from each treatment. This design inspired a model where the random-effects component contains varying intercepts, farm type effects, year effects, and farm type x year effects for the farm identities. Specifically, the model is:

yijkl=β0+β1farmTypei+μijkl

μijkl=b0,k+b1,kfarmTypei+b2,kyerj+ϵijkl,

where yijkl is a response variable, i∈ (mono-rice, plant-diversified) indicates the farm type, j∈ (first year, second year) specifies the year, k∈ (1, 2) indicates the farm identity, and l∈ (1, 2, 3) represents the replicates. μijkl is the random-effects component of the model, where b0,k, b1,ik, and, b2,jk are respectively the intercept, farm type effect, and year effect coefficients for farm identity k. Furthermore, ϵijkl is the noise term for the individual measurements. For the test of significance, we have chosen the parametric bootstrap tests for mixed-effect model comparison. This test is preferable for small sample and/or unbalanced data, and is chosen here due to the small sample size.

The revised manuscript now contains a brief description of the model above.

Also, several speculative statements in the Discussion could be supported with more sophisticated statistical analysis of the data. In particular, the authors should use the fact that they have evaluated different variables at the same farms to calculate multiple linear regressions or include co-variates in the above models and perhaps integrate these analyses all into a structural equation model, that will allow the authors to explore their hypotheses how the different variables affected each other in a path-analytic diagram. At present these are speculations, even though the authors often present them as facts in the Discussion.

Thanks for the pertinent suggestions and useful questions. In the Discussion, we have hypothesized that plant-diversification farming can increase predator abundance. The increased abundance leads to increased predation and thus decreased pest abundance. The decreased pest abundance further leads to reduced insecticide usage and increased grain yield. A possible path-analytic diagram describing the above hypothesis is shown in Figure 11—figure supplement 1 where “+” indicates increase and “- “indicates decrease. Through the use of structural equation modeling techniques, the fitness of the above model can be assessed, and alternative models could also be compared. Currently, however, the predator abundance data were not well matched with the other variables’ measurements. Pest abundance, insecticide usage and yield measurements were collected for 15 years and across all 34 farms, while predator abundance measurements were collected for 9 years at 16 of the 34 farms (4-6 plant-diversified farms and 8-10 mono-rice farms were selected to monitor the predators from 2007 to 2015). This would induce difficulty analyzing the relationship between the predator abundance variable and the other variables inside the structural equation model. The common-location data is another possible data set for path analysis, but potentially have insufficient sample size for fitting structural equation models. While the causal relationships among the variables considered in this study can be further explored through the use of more advanced modeling techniques, but the current data may prohibit us from drawing sound conclusions from such analyses. The path diagram has been added to the revised Supplementary Information (Figure 11—figure supplement 1; Discussion, third paragraph).

Reviewer #2:

[…] My main concerns are related to (1) the selection of the farms, (2) the set-up of the diversified farms, and (3) the insecticide use.

We are grateful for the positive comments, and we have revised our paper in response to the concerns addressed. Two statisticians have been added to our team to perform the statistical model analysis, and we have paid attention to the above concerns proposed by the reviewers in the revised text.

1) It is not clear to me how the farms were selected for this study. Some more information on the selection procedure would be highly appreciated. The sampling design is very unbalanced (28 conventional vs. 6 diversified farms) and not well distributed across Shanghai (the 6 diversified farms were all located on Chongming Island whereas none of the 28 conventional farms was located in this suburb). I guess there must be a good reason for this, which would clarify this slightly awkward sampling design.

As stated above, we have rewritten the text to make all this as clear as possible (Introduction, seventh and last paragraphs).

2) It is unclear to me, how the 6 diversified farms were selected, in particular if these farms were set-up as such through an initiative by those farmers before the observations started or if they were set up specifically for this study through initiative of the researchers. In the latter case, the selection would be highly questionable and could potentially confound most of the results.

Again, we address thus fully in the revised text (Introduction, seventh and last paragraphs).

3) The results of the insecticide use (in particular the number of sprays) seem awkward, given that farmers "applied pesticides according to pest forecast information offered by the Plant Protection Station". I guess this forecast is the same for all farmers and would therefore suggest the same application regime for all farms. I therefore wonder to what extent the differences in pesticide use are simply caused by the contrasting interests of the different farmer communities. A proper pesticide use study would rather apply similar amounts of pesticide and assess their impact on the pest or reduce the pest to similar amounts and quantify pesticide use. However, in this study it seems that both pesticide use and pest abundance varied among farming types, resulting in difficulties to infer sound conclusions.

The use of insecticides in 34 community farms is now clearly described in the text. In this study, pesticide application was not an independent variable, but it is still a very useful variable because pesticides were applied according to the same protocol.

Reviewer #3:

The study analyses a long-term dataset of functional arthropod groups of herbivores and predators (herbivores: three different species of specialist herbivores on rice crop; predators: three generalist predators groups, namely ladybird beetles, lacewings and spiders), replicated over multiple farms and several years. In addition, they have data on insecticide use and yield. The core idea is to test whether a diversified farming system leads to an increased top-down control compared to a monoculture situation, followed by a reduced insecticide use and increased yield. They analysed a total of 6 diversified farms and compared it to 28 control farms. One farm was approximately 2 ha in size. The diversified farm consisted of 9 rice fields surrounded by a stripe of soybeans and an adjacent stripe of another crop. Accordingly, the monoculture farms had 9 rice fields which were surrounded by bare ground. I agree with the authors that we have a lack of knowledge on factors driving ecological intensification in urban systems. However, I have a couple of major concerns with the study.

We appreciate the reviewer’s interest in our paper. We have changed the text in response to the comments.

First, the authors claim that the reduced insecticide use on diversified farms was a consequence of the increased top-down control due to increased predator densities. I wonder how they can rule out that it is not the opposite mechanism, namely that there was a reduced insecticide use on the diversified farms and as a consequence there was an increased predator density which then led to increased top-down control? In other words, how can they rule out that the result they find is not simply a consequence of reduced insecticide use but an effect of having multiple crops on a farm (i.e. diversified farm)?

We appreciate these comments, but there are many reasons to think that our inferences are the most reasonable:

1) When pesticide disturbance was excluded, higher plant species diversity in farming systems has hosted more insect predators, which is supported by a previous study (Wan NF, et al. Chinese Journal of Applied Entomology, 2012, 49: 1604–1609 (in Chinese)). Similar results excluding pesticide disturbance, were reported in our other papers: Wan NF, et al. Ecological Engineering, 2016, 90: 427–430; Wan NF, et al. Ecological Engineering, 2014, 64: 62–65; Wan NF, et al. Ecological Indicators, 2018, in press.

2) In this study, the farmers or workers applied pesticides according to pest abundance that reached the Economic Injury Level (EIL), as we explain in the revised manuscript (subsection “Monitoring sites”, second paragraph). Pesticide application is a response variable, not a true independent variable. Therefore it is extremely unlikely that pesticide application is driving the other results.

A second major concern is the statistical approach used. First, the data are clearly time-series with nicely oscillating densities indicating lag-effects (Figure 2). In addition to lag-effects, the models should include more variables. For example, for explaining predator densities it would be important to incorporate the prey densities in the model.

Thank you for a good question. As stated above, we have redone the analyses with the help of two statisticians. We agree that the farm type effect can be better elucidated by including more predictor variables into the model and by further accounting for possible lag-effects. Unfortunately, the available data may not be sufficient for such comprehensive analyses on the farm-type effect. For instance, the differences in location and time points where the predator abundances and other variables were measured preclude us from obtaining a data set with sufficiently-matched measurements suitable for analyzing the confounding effect of predator abundances on the other variables: pest abundances, insecticide usages, and grain yield were measured across 34 sites over 15 years, yet predator abundance were measured at 16 of the 34 sites across 9 years. Other questions along similar line of thoughts, such as the effect of insecticide usage on the other variables, were ruled out based on arguments mentioned in the response of the reviewer’s previous question. More complex associations among all the variables studied can be analyzed through alternative analytical approaches, and such analyses may be pursued in the future when suitable data can be obtained.

Second, the model (no matter whether it is a time-series analysis or an ANOVA) needs to incorporate the fact that repeated measures were obtained from the same location. Also, to analyse long-term data with multiple t-tests (for each year one test) is not appropriate.

Again, the new statistical analyses respond to these points.

Third, the description of the statistical methods is confusing and incomplete.

See above.

A third major concern is that the main conclusion of higher predator control is not inferred from an experiment (for example, by exposing a standardized number of prey to predators) but based on a correlation.

The concern of the reviewer is correct. This study did not expose a standardized number of prey to predators, but our former study has testified the main conclusion of higher predator in this study (Wan NF, et al. Chinese Journal of Applied Entomology, 2012, 49: 1604–1609). In this literature, when pesticide disturbance was excluded, higher plant species diversity in farming systems has hosted more insect predators.

Finally, the Introduction should be more put into a theoretical context, i.e. why and how do diversified farming systems drive biodiversity and ecosystem functioning. Further, the quality of the literature cited should be improved and all literature cited should be included in the reference list, e.g. Tscharntke et al., 2005.

In response to the helpful comments and those of reviewer 1. The reference (Tscharntke et al., 2005) is very important and we cited it in lots of places. Also, to improve the quality of the literature cited, we have deleted and added some references.

[Editors' note: the author responses to the re-review follow.]

Reviewing Editor (Reviewer #1):

[…] I have no further major comments. It would be good if co-author Jacob Weiner could have a very careful look at the manuscript before the revised version is uploaded again.

Thanks for the encouragements. From the start, Jacob Weiner has been contributing to the manuscript and had a very close look at it before submission.

Reviewer #3:

General comments:

[…] I concur with the reviews on the first version of the manuscript – the details of the study, in particular the experimental design and the statistical analysis were unclear. I did not assess the statistical analysis in detail.

We now provide additional information on the statistical analyses in the revision.

Main points:

1) I find the title of the paper misleading. The manuscript does not analyse if increasing plant species richness increases biocontrol services etc., because the authors did not analyse diversity in the surroundings of the rice fields. Basically, they show that going away from the monoculture can be beneficial. Thus, the title should be "diversifying agriculture…" or "planting border crops enhances…"

Thanks for the useful suggestion. We changed the title to “Increasing plant diversity with border crops reduces insecticide use and increases crop yield in urban agriculture”.

2) Nowhere in the Results are a test statistic or the number of replicates are given.

We now give this information.

3) Materials and methods: it is unclear how information on insecticide use was obtained.

In the 15-year monitoring data, the farmers in the 34 community farms uploaded the pest information through the server into the Shanghai Pest Monitoring System which was accessible to the monitoring locations, 9 districts of Agricultural Technology Extension and Service Centers and SATESC (subsection “Monitoring sites”, fourth paragraph). In the common-location experiments, we kept a written record of the insecticide input for each plot, and this was used to obtain the information on insecticide use (subsection “Common-location experiments”, third paragraph).

4) Materials and methods: unfortunately, I find that still many details are unclear in the revision, see below.

We have made the Materials and methods section much more detailed.

5) Results: the effect on yield is only marginally significant and should not be overstated. Also, the authors provide no mechanism why yield was higher – do the insecticides not work?

In this revision, we have mentioned that “marginally higher grain yield”. Besides, we have provided the mechanism why yield was higher in the Discussion: “In all likelihood, the reduced pest abundance contributed to rice growth, development and reproduction, and thus increased rice yield”. “Thus, the most likely explanation is that the increase in the abundance of invertebrate predators spread from border and neighboring crop areas onto the rice fields, increasing predation on pest populations”.

Materials and methods:

Subsection “Monitoring sites”: the entire section should be restructured, the design of the diversified fields is described twice (first and third paragraphs), the same is true for other information. Start with the region, then describe the farms, then how rice is grown and then the different treatments.

Thanks for the suggestion. We have restructured this section (subsection “Monitoring sites”).

Subsection “Monitoring sites”, first paragraph: please provide a drawing of a diversified field, and a photograph, how large is the area where other crops are grown.

We have added a drawing of a diversified field, and a photograph (see Figure 13—figure supplement 1).

Subsection “Monitoring sites”, second paragraph: the pest control guidance by SATESC is not known to the reader – what are the insecticides used, what active ingredients etc.?

The names and active ingredients of the insecticide used were listed in this revision (see Figure 1—source data 2).

Subsection “Monitoring sites”, last paragraph: what is a "Z" subplot? Again, use a drawing.

The drawing related to "Z" subplot has been presented in this revision (see Figure 12—figure supplement 1).

Subsection “Common location experiments”, first paragraph: design unclear: what is the replicate – a block, i.e. three pairs of?

For each site, the experiment was a randomized block design with three replicate blocks. Each block contained two plots – one for plant-diversified farming and the other for mono-rice farming (50-55 m × 30-35 m plot size in Xinchang, and 55-60 m × 35-40 m in Sanxing) (subsection “Common-location experiments”, first paragraph).

Subsection “Common location experiments”, fourth paragraph etc.: Please provide more detail of the pest assessments: I assume the light kills insects attracted to the light? Were they collected in alcohol? Who identified the species?

Insect trapping lamps were installed in the periphery of 9 paddy plots of each community farm to measure the abundance of the main herbivore pests. The lamps utilized light to attract the pests to fall into the cloth bags or cylindrical iron buckets of the lamp, and the workers in each farm identified the species and counted the attracted pests in the bags or buckets each day (subsection “Monitoring and sampling methods”, first paragraph). The sets of trapping insect herbivores are as shown in Author response image 1.

Author response image 1

Download asset Open asset

Why was the interval for the planthopper different if the same lamp in the same field is used?

The life histories and dates of occurrence on rice fields of pink rice borers and planthoppers are different. Thus, the monitoring intervals are different. From 2001 to 2015, we monitored the number of trapped pink rice borers and rice brown planthoppers from 10 April to 30 September and from 11 May to 30 September, respectively, as this was when they occurred in the rice fields (subsection “Monitoring and sampling methods”, first paragraph).

When you say each rice field, do you mean one in each of 9 plots per farm, or just one?

There were some errors in the previous submission. Insect trapping lamps were installed in the periphery of 9 paddy plots of each community farm to measure the abundance of the main herbivore pests (pink rice borer, rice brown planthopper, etc.) (subsection “Monitoring and sampling methods”, first paragraph).

Subsection “Monitoring and sampling methods”, last paragraph: did you only count predators on plants?

We only counted predators of the pests on the plants.

Subsection “Common location experiments”, fourth paragraph: what is the correlation between number of pests and number of predators? I assume it is positive? Did you also count lacewing larvae?

We analyzed the correlation between number of pests and number of predators, and found that the relationship to be significantly positive (see Author response image 2). However, we think that this figure has no role in this study, so we did not present it in the revision. Additionally, we also counted lacewing larvae (subsection “Common-location experiments”, last paragraph).

Author response image 2

Download asset Open asset

I also do not understand how the farmers decided when to spray – you state that there are economic thresholds, so there must be data on how often the threshold was reached? Were the pest data used for the threshold analysis the same that you used?

Economic Injury Levels of pink rice borers, rice brown planthoppers and rice leaf rollers are listed in Figure 2—source data 4.

Figures:

Figure 1: I thought the 28 farms were mono-rice and only 6 diversified, not the other way round?

We have revised the caption of Figure 1.

Figure 2: were pest insects sampled all year round, or only from June onwards? This is not stated in the manuscript. In the first paragraph of the subsection “Monitoring and sampling methods” it states sampling was done May to September (April to September).

We have added the corresponding information in Figure 2.

Figure 5: 9-14 kg insecticide per hectare is a lot – what is this number referring to, the liquid that is brought to the field? Details on the insecticides should be given and the amount plotted in active ingredient rather than some other unit.

The insecticide per hectare that is brought to the field is the liquid or powder (amount of commercial insecticide). We think that commercial and active ingredient insecticide uses are both important indicators to analyze the ecological intensification. Thus, we provided another Figure 5C.

Author details

1. Nian-Feng Wan

   1. Eco-environmental Protection Research Institute, Shanghai Academy of Agricultural Sciences, Shanghai Key Laboratory of Protected Horticultural Technology, Shanghai Engineering Research Centre of Low-carbon Agriculture, Shanghai, China
   2. Ministry of Education Key Laboratory for Biodiversity Science and Ecological Engineering, Shanghai Chongming Dongtan Wetland Ecosystem Research Station, Institute of Biostatistics, Shanghai Institute of Eco-Chongming, (SIEC), Fudan University, Shanghai, China

   Contribution

   Conceptualization, Resources, Data curation, Software, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing, Performed the statistical model analysis

   Contributed equally with

   You-Ming Cai

   Competing interests

   No competing interests declared

2. You-Ming Cai

   Eco-environmental Protection Research Institute, Shanghai Academy of Agricultural Sciences, Shanghai Key Laboratory of Protected Horticultural Technology, Shanghai Engineering Research Centre of Low-carbon Agriculture, Shanghai, China

   Contribution

   Conceptualization, Resources, Data curation, Software, Formal analysis, Supervision, Funding acquisition, Validation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing

   Contributed equally with

   Nian-Feng Wan

   Competing interests

   No competing interests declared

3. Yan-Jun Shen

   Chongming Agricultural Technology Extension and Service Center, Shanghai, China

   Contribution

   Conceptualization, Resources, Data curation, Software, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing

   Competing interests

   No competing interests declared

4. Xiang-Yun Ji

   Eco-environmental Protection Research Institute, Shanghai Academy of Agricultural Sciences, Shanghai Key Laboratory of Protected Horticultural Technology, Shanghai Engineering Research Centre of Low-carbon Agriculture, Shanghai, China

   Contribution

   Conceptualization, Resources, Data curation, Software, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing

   Competing interests

   No competing interests declared

5. Xiang-Wen Wu

   Shanghai Agricultural Technology Extension and Service Center, Shanghai, China

   Contribution

   Conceptualization, Resources, Data curation, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing

   Competing interests

   No competing interests declared

6. Xiang-Rong Zheng

   Ministry of Education Key Laboratory for Biodiversity Science and Ecological Engineering, Shanghai Chongming Dongtan Wetland Ecosystem Research Station, Institute of Biostatistics, Shanghai Institute of Eco-Chongming, (SIEC), Fudan University, Shanghai, China

   Contribution

   Conceptualization, Data curation, Software, Formal analysis, Supervision, Validation, Methodology, Writing—original draft, Writing—review and editing, Performed the statistical model analysis

   Competing interests

   No competing interests declared

7. Wei Cheng

   Shanghai Agricultural Technology Extension and Service Center, Shanghai, China

   Contribution

   Conceptualization, Resources, Data curation, Software, Formal analysis, Supervision, Validation, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing

   Competing interests

   No competing interests declared

8. Jun Li

   Climate Center of Shanghai, Shanghai, China

   Contribution

   Conceptualization, Resources, Data curation, Software, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing

   Competing interests

   No competing interests declared

9. Yao-Pei Jiang

   Shanghai Agricultural Technology Extension and Service Center, Shanghai, China

   Contribution

   Conceptualization, Resources, Data curation, Formal analysis, Supervision, Validation, Investigation, Visualization, Writing—original draft

   Competing interests

   No competing interests declared

10. Xin Chen

    College of Life Sciences, Zhejiang University, Hangzhou, China

    Contribution

    Conceptualization, Data curation, Software, Formal analysis, Supervision, Methodology, Writing—original draft, Writing—review and editing

    Competing interests

    No competing interests declared

11. Jacob Weiner

    Department of Plant and Environmental Sciences, University of Copenhagen, Frederiksberg, Denmark

    Contribution

    Conceptualization, Data curation, Software, Formal analysis, Supervision, Validation, Methodology, Writing—original draft, Writing—review and editing

    Competing interests

    No competing interests declared

    "This ORCID iD identifies the author of this article:" 0000-0002-0736-7943

12. Jie-Xian Jiang

    Eco-environmental Protection Research Institute, Shanghai Academy of Agricultural Sciences, Shanghai Key Laboratory of Protected Horticultural Technology, Shanghai Engineering Research Centre of Low-carbon Agriculture, Shanghai, China

    Contribution

    Conceptualization, Resources, Data curation, Software, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing

    For correspondence

    jiangjiexian@163.com

    Competing interests

    No competing interests declared

    "This ORCID iD identifies the author of this article:" 0000-0003-3567-8201

13. Ming Nie

    Ministry of Education Key Laboratory for Biodiversity Science and Ecological Engineering, Shanghai Chongming Dongtan Wetland Ecosystem Research Station, Institute of Biostatistics, Shanghai Institute of Eco-Chongming, (SIEC), Fudan University, Shanghai, China

    Contribution

    Software, Methodology, Writing—original draft

    Competing interests

    No competing interests declared

14. Rui-Ting Ju

    Ministry of Education Key Laboratory for Biodiversity Science and Ecological Engineering, Shanghai Chongming Dongtan Wetland Ecosystem Research Station, Institute of Biostatistics, Shanghai Institute of Eco-Chongming, (SIEC), Fudan University, Shanghai, China

    Contribution

    Conceptualization, Data curation, Supervision, Validation, Methodology

    Competing interests

    No competing interests declared

15. Tao Yuan

    Eco-environmental Protection Research Institute, Shanghai Academy of Agricultural Sciences, Shanghai Key Laboratory of Protected Horticultural Technology, Shanghai Engineering Research Centre of Low-carbon Agriculture, Shanghai, China

    Contribution

    Made the survey maps

    Competing interests

    No competing interests declared

16. Jian-Jun Tang

    College of Life Sciences, Zhejiang University, Hangzhou, China

    Contribution

    Conceptualization, Methodology

    Competing interests

    No competing interests declared

17. Wei-Dong Tian

    Ministry of Education Key Laboratory for Biodiversity Science and Ecological Engineering, Shanghai Chongming Dongtan Wetland Ecosystem Research Station, Institute of Biostatistics, Shanghai Institute of Eco-Chongming, (SIEC), Fudan University, Shanghai, China

    Contribution

    Software, Methodology, Performed the statistical model analysis

    Competing interests

    No competing interests declared

18. Hao Zhang

    Eco-environmental Protection Research Institute, Shanghai Academy of Agricultural Sciences, Shanghai Key Laboratory of Protected Horticultural Technology, Shanghai Engineering Research Centre of Low-carbon Agriculture, Shanghai, China

    Contribution

    Made the survey maps

    Competing interests

    No competing interests declared

19. Bo Li

    Ministry of Education Key Laboratory for Biodiversity Science and Ecological Engineering, Shanghai Chongming Dongtan Wetland Ecosystem Research Station, Institute of Biostatistics, Shanghai Institute of Eco-Chongming, (SIEC), Fudan University, Shanghai, China

    Contribution

    Conceptualization, Resources, Data curation, Software, Formal analysis, Supervision, Validation, Visualization, Methodology, Writing—original draft, Writing—review and editing

    For correspondence

    bool@fudan.edu.cn

    Competing interests

    No competing interests declared

    "This ORCID iD identifies the author of this article:" 0000-0002-0439-5666

• 3,710

  Page views

• 457

  Downloads

• 31

  Citations

Article citation count generated by polling the highest count across the following sources: Scopus, Crossref, PubMed Central.