Sustainably feeding the world’s growing population is a challenge1,2,3, and closing yield gaps (that is, differences between farmers’ yields and what are attainable for a given region)4,5,6 is a vital strategy to address this challenge3,4,7. The magnitude of yield gaps is particularly large in developing countries where smallholder farming dominates the agricultural landscape4,7. Many factors and constraints interact to limit yields3,4,5,6,8,9,10, and progress in problem-solving to bring about changes at the ground level is rare. Here we present an innovative approach for enabling smallholders to achieve yield and economic gains sustainably via the Science and Technology Backyard (STB) platform. STB involves agricultural scientists living in villages among farmers, advancing participatory innovation and technology transfer, and garnering public and private support. We identified multifaceted yield-limiting factors involving agronomic, infrastructural, and socioeconomic conditions. When these limitations and farmers’ concerns were addressed, the farmers adopted recommended management practices, thereby improving production outcomes. In one region in China, the five-year average yield increased from 67.9% of the attainable level to 97.0% among 71 leading farmers, and from 62.8% to 79.6% countywide (93,074 households); this was accompanied by resource and economic benefits.
Closing yield gaps could potentially double global food output4,7, substantially enhancing food security. In low-input, low-output areas such as sub-Saharan Africa, considerable yield gains may be achieved by increasing nutrient and/or water inputs11,12. However, there are areas (for example, in India and China) where inputs are already high or excessive, yet yield is substantially below the attainable level, leading to poor resource-use efficiency and environmental consequences12,13,14. For example, the land in Quzhou County (450 km south of Beijing) is irrigated and overuse of nitrogen fertilizer (N) is common, but wheat and maize yields (the main crops) are less than two-thirds of those at the local Experimental Station (the de facto attainable yields; see Methods; Supplementary Table 1). Although knowledge and technologies are available for high-yield, high-efficiency practises8, it is important to establish how, in meaningful scales beyond research-oriented experiments, to equip smallholders with these technologies and enable them to achieve greater performance.
Science and Technology Backyards were established in Quzhou County in 2009 (Extended Data Fig. 1), and staffed with professors and graduate students from China Agricultural University. Living in the villages and working with farmers, we carried out research–education extension activities that centred on technology-transfer and enabled smallholders to achieve higher returns with better environmental stewardship (Fig. 1; Supplementary Table 2). Here we report the processes, outcomes, and lessons from these studies.
We first determined the factors that contributed to the yield gaps by surveying 150 farmers (Extended Data Fig. 2) and conducting 55 single-factor experiments (see Methods). A number of factors were identified, each exhibiting a >5% yield gap (Fig. 2; Extended Data Table 1), as described below.
Crop variety contributed to a yield gap of 9.4% (wheat) and 19.8% (maize; Fig. 2). Many farmers were using seed varieties that were not locally well-suited. Plant density was also shown to be important: wheat over-seeding had a yield gap of 6.3%; maize under-seeding led to a 20.6% yield gap (Fig. 2). The management deficiencies can be attributed to farmers’ lack of knowledge, their risk-aversion tendency, and market confusion (that is, too many products and little help; Methods, Extended Data Fig. 2). Furthermore, deep tillage had a yield advantage of 5.8% for wheat (Fig. 2). The recommended deep ploughing (25 cm) to break the dense soil pan (at 12–15 cm depth, formed as a result of several decades of rotary tillage) for improved productivity and nutrient and water efficiencies15 was adopted by few farmers because of ill-matched plough-cutting width (4 m) against the fields (5–6 m strips).
Time constraints were also found to affect yield. Proper sowing time had a yield advantage of 6.3% (wheat) and 15.0% (maize; Fig. 2). Most of the farmers failed to sow within the recommended time window, mainly owing to labour constraints (many are part-time farmers, and hold additional jobs in nearby cities16).
Irrigation infrastructure is another issue. A single well serves 6–8 ha belonging to dozens of farmers; switching pumps and pipes between farmers (no facility-sharing scheme) delays field operations. In addition, maize-harvest time was associated with a 14.4% yield gap (Fig. 2); 74% of farmers harvested prematurely, not understanding the importance of prolonged grain-filling for achieving yield potential.
Fertilizer management had a marked effect; following the recommended practice exhibited a yield advantage of 11.6% (wheat) with 32% less N applied, and 7.5% (maize) with a similar N rate (Fig. 2, Extended Data Table 1). Farmers’ non-optimal N management is a result of knowledge and market shortcomings. More than two-dozen fertilizer products were marketed (Extended Data Table 2), with neither streamlined nor user-friendly labelling. Only 6.7% farmers could correctly calculate crop nutrient requirement (Extended Data Table 3). The purchasing decisions made by the farmers were influenced by price and marketing schemes, and over-application was considered insurance against potential yield loss. Furthermore, the timing of wheat top-dressing and irrigation had a 6.2% yield gap (Fig. 2); many farmers performed these tasks 7–10 days earlier than recommended, owing to other commitments.
Agronomic, infrastructural, and socioeconomic factors are known to affect smallholder productivity3,4,5,6,8,9,10, but few studies have examined these factors systematically and comprehensively. Furthermore, as specific fundamental barriers (for example, plough width, crop variety, cultivation timing) exist; it was necessary to work directly with the farming community to develop effective solutions.
We developed a set of ten recommended practices that targeted the yield-limiting factors, and then engaged leading farmers for feedback. Leading farmers were defined as those with good farming skills, and were open to new ideas. The initial recommendations were discussed at length by STB staff and leading farmers (45 individuals in 2009, Methods), and revisions were made where required.
We initially recommended deep ploughing plus two-pass rotary tillage before wheat planting, with four in-season irrigations. This was modified, as per lead-farmer input, by adding mechanical compaction after sowing with three instead of four irrigations. The added compaction operation (costing US$36 ha–1) would lower moisture loss and favour uniform germination, thus reducing irrigation frequency (saving US$162 ha−1). Furthermore, wheat-sowing time was postponed for five days to accommodate farmers’ practice of harvesting maize (the preceding crop) by hand. The delayed sowing reduces tiller development, thus necessitating an increase in seeding rate. In addition, maize density was reduced from the recommended 90,000 to 75,000 plants ha–1; farmers were wary of the high cost of seeds and potential lodging. In addition, leading farmers rejected the manuring recommendation (pre-planting wheat), but increased synthetic fertilizer usage (from 200–220 to 250–280 kg N ha–1) to compensate, in order to save money and labour (manure costs 10 times more than fertilizer per unit N; manure-handling requires 60 labour-hours ha–1). Maize fertilization was also modified from three to two applications considering labour shortage.
The collaboration between STB staff and leading farmers and the outcomes achieved demonstrate how participatory innovation, a combination of institutional knowledge with insight from the farming community17,18,19,20, can be put into practice to overcome yield issues. Subsequently, a revised set of ten recommended practices was assembled (right column in Extended Data Table 4). Multi-factor experiments were then carried out by leading farmers (Methods), each employing all ten recommendations. The results are illustrative: leading farmers achieved 97.0% of the attainable yields (average from 2009–2014), compared to 67.9% before STB intervention (Fig. 3; Supplementary Tables 1, 3, 4). Their use of water and N was less efficient, but economic returns were 43.8–195.8% higher than the Experimental Station (Table 1, Extended Data Table 5). Experimental Stations generally operate with all-out efforts regardless of cost, whereas for farmers, economic and labour efficiencies precede yield and environmental goals13,16.
To engage the farming community, we deployed several time-honoured education-extension methods coupled with innovative outreach mechanisms (Methods, Extended Data Table 6). Our approach had four key elements: raising awareness through field demonstrations, farming schools, and yield contests; readily available information with easy-to-understand posters and customized calendars; engaging members of the community via in-person communication and social–cultural bonding; enabling change with on-site advice and reminders dispatched at critical times. Furthermore, to overcome the limitations associated with small fields regarding certain management practices (for example, deep ploughing, irrigation), we pioneered a quasi-farmer co-operative named ‘combining land for uniform practice’ (CLUP). A CLUP consists of 30–40 households with adjoining land (3–4 ha) adopting uniform practices and receiving additional services (Method; Supplementary Discussion). At harvest, the CLUP then ceases existence and each member harvests individually, therefore avoiding complex issues (for example, ownership, rights, and policies) that typically hinder land-use reform21. After the first trial in 2009, farmers enthusiastically embraced the initiative, and 43 new CLUPs were subsequently formed in 2010. CLUPs helped promote adoption of STB recommendations.
The efficacy of STB approach is apparent: compared to those in neighbouring or control villages, farmers from STB villages had (i) better agronomic knowledge (for example, 35.2% can calculate crop nutrient requirements, versus 10.9%; Extended Data Table 3); (ii) higher adoption of STB recommendations (53.5% versus 31.4%; Extended Data Table 7); and (iii) greater yields (5.5–8.5% higher), nutrient- and water-use efficiencies (12.5–47.0%), and economic return (2.9–47.0%; data derived from Table 1). Notably, the effect of STB intervention also permeated to neighbouring villages, which performed better than control villages (Extended Data Tables 3, 5, 7) owing to easier access to STB events/information. Globally, much effort has been made to provide smallholders with advanced management technologies; although some are successful, many fail to scale up or to produce consistent and sustained results1,3,22. The STB model has proved effective, particularly for disseminating comprehensive technology practices.
We also contacted local government and private enterprises, who were encouraged enough by the positive findings of the study to become involved in appropriate niches. Countywide, the Quzhou government allocated US$2,000,000 to subsidize deep-plough acquisition, provided funding to promote STB recommendations in non-STB villages, and took measures to warn farmers against inferior seeds or fertilizer and enable them to identify government-approved products (see Methods). Private enterprises also contributed, for example, by sponsoring farmer yield contests, training field representatives with STBs, and being scrupulous with marketed products (Methods; Extended Data Table 2). Enhanced public and private support contributed to a countywide yield improvement (from 62.8% of the attainable level in 2009 to an average of 79.6% in the subsequent five years), along with substantial increases in resource efficiencies and farmer income (Extended Data Table 8). Cross-institution collaborations are often hindered by issues concerning capital resources, profit-sharing, and intellectual property, for example23,24. As demonstrated in this work, STB provided a platform for the university, industry, and government to work together to improve farming practises, while avoiding such issues.
Following the successful results in Quzhou, 71 STBs now operate in 21 provinces (Extended Data Fig. 1). For future scaling up, approximately US$600,000,000 is required annually to fund STBs for China’s major agricultural regions (Supplementary Discussion), which would be a small amount compared to the US$18-billion-a-year fertilizer subsidies25. Such capacity-building investment could enhance food security and sustainability, while uplifting numerous farmers out of subsistence to a better life.
The STB model combines top-down approaches with bottom-up measures to enable smallholders. Working directly with the farming community enabled us to (i) gain insight into factors that contribute to yield gaps and how to find appropriate solutions; (ii) engage with farmers and the extended support network; and (iii) train students for agricultural services with appropriate skills and attitudes. The STB approach can be adapted in other parts of the world where conditions are permissible to help smallholders achieve greater yield and income with better environmental outcomes (Supplementary Discussion).
Network of Science and Technology Backyards (STB) in China
The conceptualization and deployment of STB was initiated by China Agricultural University (CAU) to test an innovative technology-transfer approach for enabling smallholder farmers. Following the initial success in Quzhou County, more STBs were established by CAU in different farming systems across the country, including large farms in the northeast area (2–20 ha per household, larger than the national average) and the fruit- and vegetable-basket region in southern China. After 2012, other institutions began to adopt the model and establish STBs in their regions. By late 2015, a total of 71 STBs were in place, covering 22 agricultural production systems in 21 provinces (Extended Data Fig. 1). Among these STBs, 25 are overseen by CAU, 26 by other agricultural universities, and the remaining 20 by private enterprises with supervision by academic scientists. The STBs currently cover wheat, maize, rice, soybean, potato, cotton, grape, cherry, apple, strawberry, Chinese date, orange, banana, mango, pineapple, lychee, pitaya (cactus fruit), tomato, garlic, pepper, green onion, and forage production systems. The organizational format and functionalities of all STBs are similar to those in Quzhou, with staff living in the villages year-round and working with farmers to identify yield-limiting factors, revising science-based recommendations for local adoptability, and garnering public and private support.
Agricultural background in Quzhou
Quzhou is a typical agricultural county (114°50′22.3′′E–115°13′27.4′′E, 36°35′43′′N–36°57′N) situated about the centre of the North China Plain (NCP). The latter is a region with intensively managed cereal production systems, producing 38% of agricultural products in China. Rotation of winter wheat (Triticum aestivum L.) and summer maize (Zea mays L.) is the dominant cropping system. Quzhou has a total population of 433,000 and arable land of 66,700 ha, with 93,074 farming families living in 342 villages within 10 townships26. Per capita arable land is about 0.15 ha; per capita, net agricultural income was US$944 in 2008 in rural households (Extended Data Table 8), which is far below the US$2,290 in urban households26.
Quzhou has deep alluvial soils with inherently high salt content, rendering the land barely productive historically. Starting in 1980 s, the saline soil was reclaimed via three measures: (i) deep wells drilled to reach fresh water for irrigation, (ii) deep drainage ditches dug to lower the groundwater table, (iii) crop residue recycled to build up soil organic matter and enhance productivity. Specifically, soil texture comprises light loam, medium loam, sandy loam, clay and salt-affected27. Soil organic matter ranges from 10.3 to 16.0 g kg−1 (mean 13.9 g kg−1), salt content is up to 1.11 g kg−1 at 20–40 cm soil depth28. With steady increases in chemical fertilizer use in the past three decades, soil nutrient content is relatively high: 0.94 g kg−1 for total nitrogen (from 0.70 to 1.06 g kg−1), 23.0 mg kg−1 for available phosphorus (Olsen-P, 1.4 to 50.4 mg kg−1), and 132 mg kg−1 for exchangeable potassium29 (Exc.-K, 67.1 to 190 mg kg−1). The mean annual temperature is 13.1 °C; average annual precipitation is about 500 mm with a variation of 27.5% between years. The area has a distinct wet season from late June to late October with 60% of the annual precipitation falling in those months, and a dry season from November to early June in the semi-arid monsoon climate. The ground water table has been falling by 0.8 m annually in recent decades owing primarily to agricultural irrigation30.
The first STB was established in Beiyou village in 2009 (631 households farming 272 ha). Six additional STBs were established around the end of 2009 and beginning of 2010 in separate villages (Extended Data Fig. 1). Of the seven STBs, four were focused on wheat and maize rotation systems and the remaining three were targeting other cash crops. Staffing included one faculty member and two graduate students per STB. The current paper reports the results based on the four STBs for the wheat and maize rotation systems.
Quzhou Experimental Station, attainable yields
Quzhou Experimental Station was established by CAU in 2007. Located in the northern part of the county and about 20 km from Beiyou village (the first STB village), the Experimental Station occupies 20 ha of cropland. The soil type is Cinnamon (salt-affected, sandy clay loam) with 14.0 g kg−1 organic matter, 0.97 g kg−1 total nitrogen, 16.0 mg kg−1 Olsen-P, and 179 mg kg−1 Exc.-K in the 0–30 cm soil layer31. Soil conditions are similar to those in the STB village area and about average in Quzhou County. The main field study at the Experimental Station relevant to this article was aimed at developing the best management practices applicable in Quzhou on the basis of high-yield, high-efficiency technologies (integrated soil-crop system management, ISSM)32. Plot size was 1800 m2 with four replications. Management practices (treatments) included land preparation, crop variety, sowing date, seeding rate, fertilizer management and harvest date (left column in Extended Data Table 4). During the 5-year span (2009–2014), some practices were modified to incorporate better management options developed in other regions (for example, an improved wheat planter was introduced in 2012), which contributed to the overall trend of yield increase (Fig. 3, Supplementary Table 1). As the field study implemented best management practices according to ISSM guidelines, yields from these studies were used as the de facto attainable yields (which would be lower than potential yields but served our purpose, Supplementary Discussion). Accordingly, yield gaps are operationally defined as the differences between farmers’ yields (through survey and single- and multi-factor experiments, discussed below) and the de facto attainable yields.
Farmer survey, single-factor experiments
A questionnaire survey was conducted in October 2009 to determine the prevailing practices regarding the price of seed, variety selection, fertilizer, manure, irrigation, and pesticide use, machinery services and labour, as well as farmers’ information sources and knowledge base. We selected 10 villages from the high, medium, and low (per capita income) village groups in Quzhou County; 150 farmers randomly selected from the 10 villages were interviewed. The interviews were conducted by STB staff plus additional CAU students, each interview took about 40 min with detailed information on quantitative (for example, fertilizer rate, yields) and qualitative parameters (for example, variety and source of seeds, access to information). Survey results were summarized (Extended Data Fig. 2, Extended Data Tables 2, 3, 7) and examined to identify farmers’ practice and factors constraining their performance.
To quantify yield-gap contribution from each of the main factors, farmers in the four STB villages were solicited to carry out single-factor (paired) experiments (Extended Data Table 1). Selection of participating farmers was based on two considerations. First, his or her field practice was one of the major yield-gap contributing factors. Second, the farmer was willing to try different management per STB recommendations. Each paired experiment was carried out by the farmer, who devoted a parcel of land in which two treatments were laid out side-by-side, one representing the farmers’ practice and the other following STB recommendation. All management remained the same except for the single factor tested. In some cases, participating farmers allocated the least-productive parcel of their fields for the experiment in order to minimize risks. This helps to explain the wide variation shown in Extended Data Table 1. Meanwhile, the substantial yield increase derived from following recommended practices, even on poor land, provided a better service in convincing otherwise-reluctant farmers. Each yield-gap contributing factor was tested in at least three farmers’ fields, which served as the de facto replication. A total of 55 paired experiments were conducted during 2009–2011; plot size ranged from 300 to 600 m2 depending on the individual parcels of land. STB staff provided on-site assistance to make sure that the recommended practice, whether it was sowing date or fertilizer application time or rate, was implemented correctly by the farmers. STB staff measured crop yields and recorded data. Description of treatments and the results are summarized in Extended Data Table 1.
Leading farmers and multi-factor experiments
Leading farmers are those who were, through daily interactions between STB staff and the villagers, recognized leaders in the farming community. Their participation in STB-related work was by volunteering and/or by request. In addition to the critical role in helping STB staff revise the recommended management practices to conform to local farming practises (described in main text), leading farmers also carried out multi-factor experiments in their selected fields, in which all 10 recommended practices were simultaneously implemented. The number of leading farmers participating in multi-factor experiment varied from 45 to 71 per crop-season in 2009–2014 (Supplementary Table 1). Plot size ranged from 600 to 1,200 m2, depending on individual fields. The pool of fields (that is, 45 in 2009, 71 in 2014) served as replications. STB staff provided technical guidance and consultation and recorded yields and relevant data. After harvest, on-farm performance of the improved technologies tried by leading farmers was summarized and the information communicated to the Experimental Station for further innovation. The trend of yield increases during 2009–2014 (Fig. 3) may be attributed to progressive improvement of leading farmers in mastering the recommended technologies.
Knowledge/technology dissemination and outcome evaluation
A variety of methods were used to disseminate the advanced management practices to the farming communities. Mechanisms corresponding to the four key elements summarized are outlined below:
Raising awareness. The single- and multi-factor experiments served as live exhibits. Field days were held each month during the growing season, with leading farmers answering questions and outlining which practices they adopted, why, and the attained or expected benefits. Furthermore, the workshops were offered in winter months to consolidate the outcomes and discuss which practises worked (or did not work). Yield contests were held to engage members of the farming community.
Readily available information. Science-based technologies were presented in eye-catching and comprehensible formats. For example, waterproof posters were erected along the main road, highlighting information on wheat and maize production with graphic illustrations. Customized calendars showing recommended practices were distributed to villagers.
Engaging farming community members. As well as production-oriented events, a variety of social–cultural activities were organized, such as tea gatherings, folk singing and dancing in winter months or around traditional holidays. These events helped build a relationship between STB staff and farmers, while strengthening the community as a whole.
Enabling change. During busy field operations, STB staff were to travel and provide advice on-site when needed. Briefing sessions were offered before each crop-management stage. Reminders were sent via cell phone texting and/or the village’s intra-broadcasting system at the time of important field tasks. See Extended Data Table 6 for photo illustration and Supplementary Table 2 for more details regarding location and service target.
Regarding CLUP (combining land for uniform practice), farmer-members elected a leader who coordinated field tasks and took charge in decision-making with the consultation of STB staff. Uniform cultivation favoured the adoption of deep tillage and eased irrigation schedule (with 20–25% increased efficiency). Furthermore, STB staff designed a fertilizer formula for each CLUP based on soil-test results and crop data, and solicited fertilizer suppliers to blend and deliver the products accordingly; we also helped CLUP determine seed variety and supplies, thus lowering their production cost by about US$100 ha−1 (Extended Data Table 5).
Government officials were interested in strengthening their support and improving services. As the yield was enhanced by deep tillage, the county subsidized the cost of deep-plough purchases and provided US$75 ha−1 to farmers who adopted the tillage method. In addition, two full-time agricultural technicians were added to each township in Quzhou County, starting in 2010, to promote countywide adoption of STB recommendations. They also issued special logos on the packages of trusted seeds or fertilizer products as a measure of protecting farmers from inferior products. In the private sector, seed and fertilizer companies donated money and/or production material (for example, seeds, fertilizers) to farmer yield contests and field demonstrations; they sent their field representatives to STB-organized events. Their service model started to shift from farmers only being able to offer available stock, to being able to supply specific produce on demand. Learning which products are most suitable for a given region, as per STB findings, fertilizer companies began to make crop/region-specific products with improved labelling, for example, specifying target crops and providing instructions by linking application rate with yield.
To assess the effectiveness of education-extension efforts and the outcomes of STB interventions, a follow-up survey was conducted in 2012. Twelve villages were strategically selected, including (i) the four STB villages in which farmers had full access to all STB services, (ii) four neighbouring villages that are adjacent to the STB villages, and (iii) four control villages that are about 10 km away from any of the STB villages. The neighbouring and control villages had similar demographics and cropping systems as STB villages (Extended Data Table 2), but received no direct on-site services by STB staff (no single- or multi-factor experiments), although public events such as training workshops or field demonstrations were open to all. From each of the 12 villages, 30 or more farmers were randomly selected and interviewed. There were a total of 575 households as survey participants (Extended Data Table 2). Survey items were the same as for the 2009 survey. Survey entries of quantitative nature (for example, grain yield, fertilizer rate) were farmers’ self-reporting, that is, not measurements taken by STB staff.
Results of the 2012 survey were summarized and examined to assess the efficacy of knowledge and technology transfer. Clearly, farmers in the STB villages had better understanding of key agronomic parameters and higher adoption of recommended practices than their counterparts (Extended Data Tables 3, 5, 7). The survey results also suggest ‘spill over’ of knowledge and improved management practices from the centres of action (STB villages and STB organized events) to the neighbouring villages as compared to control villages (Extended Data Tables 3, 5, 7).
The outcome of STB intervention can be quantitatively evaluated by comparisons among the three groups of villages in terms of land-use efficiency (crop yield), resource-use efficiency (nutrients and water), and investment and labour productivity (Table 1, Extended Data Table 5, Supplementary Table 5). Partial productivity of nitrogen, calculated as grain yield per kg of chemical nitrogen fertilizer input, was used to provide an estimate of nutrient use efficiency. Water-use efficiency was calculated as grain yield per cubic meter of irrigation water input. Labour productivity was expressed as grain yield per hour labour input (self as well as hired labour). The benefit:cost ratio was the parameter used to evaluate the economic effect of farmers’ practices. Cash expenditure for seeds, fertilizers, machinery, irrigation, and farmers’ labour input were components of the operating costs. Quzhou smallholder farmers rarely rent land, borrow money from the bank, or buy crop insurance (wheat and maize), therefore these were not included in the total costs. Net profit (benefit) was calculated as the product of the grain yield and the market price minus the operational costs.
For all field experiments and farmer surveys, data was analysed by one-way analysis of variance in SAS33. To distinguish STB villages from neighbouring and/or control villages in terms of production and economic performances, two-tailed t-tests were performed. The results were compared using least significant difference at a 0.05 level of significance for grain yield, fertilizer use efficiency, water-use efficiency, labour productivity and the benefit:cost ratio (Table 1, Extended Data Table 5, Supplementary Table 5). In addition, linear regressions of yields over time (2008–2014) were run for the Experimental Station and leading farmers, with statistical significance determined using the F-test (Supplementary Table 4). Also, annual yield comparison between the Experimental Station and leading farmers was made by two-tailed t-test for each year (Supplementary Table 4).
We thank P. M. Vitousek, P. A. Matson, P. Christian, T. H. Misselbrook, G. P. Robertson, D. R. Chadwick, I. Ortiz-Monasterio, X.J. Liu and J. D. Toth for their comments/editing assistance. We thank Q. F. Meng, D. J. Lu, P. Yan, X. Q. Jiao and M. L. Guo for joint field experiment and data analysis. Many individuals were involved in the STB Network in Quzhou and elsewhere in China. This work was sponsored by the China 973 Program (Grant 2015CB150405), the Innovative Group Grant of the Natural Science Foundation of China (Grant 31421092), the Special Fund for Agro Scientific Research in the Public Interest (Grant 201203079), and the Program for New Century Excellent Talents in University (Grant 2016QC125).
Extended data figures
Extended data tables
This file contains Supplementary Tables 1-5, a Supplementary Discussion and additional references.
About this article
Gobi agriculture: an innovative farming system that increases energy and water use efficiencies. A review
Agronomy for Sustainable Development (2018)