Partially mechanized non-puddled rice establishment: on-farm performance and farmers’ perceptions

ABSTRACT Transplanting rice (Oryza sativa L.) seedlings into non-puddled soils with minimum soil disturbance (by mechanized strip tillage) is an opportunity to expand Conservation Agriculture (CA) adoption in rice-based cropping systems. However, the farmer acceptance and on-farm profitability of this technology has not been assessed. Here, we analysed farmer-managed non-puddled transplanting (NT) of rice seedlings in paired comparisons with rice seedlings transplanted in puddled (PT) soil. Sixty-six rainfed monsoon (aman) and 84 dry-season irrigated (boro) rice crops were established by NT and compared with paired PT crops during 2013–2015 in north and north-west Bangladesh. Non-puddled fields were prepared by firstly making 40–60-mm-wide tilled strips with a Versatile Multi-crop Planter, then by 18–24-h inundation with water to soften soils in the strip, followed by transplanting seedlings into strips. The NT crops produced similar or significantly greater grain (boro season of 2015) and straw yields than paired PT crops. The minimum soil disturbance NT did not hinder seedling transplanting of rice or increase the total labour costs for transplanting and weeding compared to PT, rather it reduced the cost of production, and increased net benefit for rice in both seasons relative to PT. Under farmer management on a range of soils and in the two main rice-growing seasons, NT performed as well or better than PT suggesting that this is a feasible option for farmers planning to adopt CA in rice-based cropping systems. Over 3 years, farmers’ perception of NT performance shifted from scepticism to mostly favourable.


Introduction
Rice is the most important staple food for about 3 billion people (Maclean, Dawe, Hardy & Hettel, 2002) and contributes significantly to food security among millions of farms households. Rice area, production and yield increased rapidly from the 1960s to the late 1980s (Khush, 1995) due to adoption of high-yielding varieties, together with increased use of irrigation water, chemical fertilizers, herbicides, pesticides, etc. However, sustainability threats have to be addressed for the future of wetland rice cultivation. Rice seedling establishment in fully tilled and puddled saturated soils is a major source of greenhouse gas (GHG) emission, particularly methane (Pathak et al., 2011). Avoidance of puddling of soils for establishment of rice can decrease water requirements for growing rice (Mahajan, Timsina & Kuldeep-Singh, 2011;Mahmud, Bell, Vance & Haque, 2017) and decrease GHG emissions (Alam, Biswas & Bell, 2016).
Soil tillage followed by ponding of water and puddling is practised to control weeds, facilitate transplanting and reduce soil permeability for wetland rice establishment (Sharma, Ingram & Harnpichitvitaya, 1995). However, puddling of soil consumes about 200-250 mm of water, equivalent to 17% of the total water use by rice (Mahajan et al., 2011). Puddling also destroys soil aggregates, breaks capillary pores, disperses clay particles (that can form an impermeable clayey layer on the surface of coarse-textured soil) and creates a plough pan that impedes root penetration for following crops (Haque, Bell, Islam & Rahman, 2016). Soil puddling for rice cultivation results in low wheat (Triticum aestivum L.) grain yields in the rice-wheat system, mainly due to weakening of soil structure and the development of sub-surface hardpans and increased risk of water-logging (Fujisaka, Harrington & Hobbs, 1994). Continuation of soil puddling for rice transplanting will negate the benefits of zero tillage practised in other crops in the rotation as is reported for the rice-wheat system .
Conservation Agriculture (CA) helps farmers to reduce production costs while improving soil health, crop diversity and timeliness of cultivation (Johansen, Haque, Bell, Thierfelder & Esdaile, 2012). However, the adoption of CA is still low by rice-based smallholders (Johansen et al., 2012), even though two-wheel tractor (2WT)-based implements have created new avenues for the pursuit of CA in rice-based smallholder farming systems (Bell et al., 2019;Haque, Bell, Islam, Sayre & Hossain, 2011). In rice-based cropping systems, implementation of CA is hampered due to the practice of soil puddling for rice seedling transplanting. Initial development of non-puddled transplanting (NT) of rice seedlings systems in Bangladesh was described in Haque (2009). Various forms of NT of rice seedling, as well as benefits and limitations of NT, were further elaborated on by Ladha et al. (2009), Saharawat et al. (2009), Johansen et al. (2012, and Haque et al. (2016). Generally, the methodology of NT of rice seedling is the same as for rice seedlings transplanting in puddled (PT) soil except for the different land preparation. Haque et al. (2016) evaluated four options of land preparation for NT of rice seedling in 2 long-term experiments and 27 on-farm sites in Rajshahi district of Bangladesh. In this study, NT establishment had no negative effect on rice yields across seasons and years, but it reduced cost of production and increased the gross margin of NT rice relative to PT, as well as decreasing the time taken for land preparation and crop establishment and decreasing the number of irrigation events required to saturate the soil prior to transplanting. However, despite these obvious benefits of NT from research on farms, the acceptance and profitability of the NT technology when managed by farmers is not well understood.
Several hundreds of farmers have adopted NT rice establishment methods where the technologies were demonstrated earlier in Bangladesh. The aim of the present study was to determine the crop growth, yield and profitability of NT rice crops managed by farmers. A total of 150 on-farm studies on NT aman (hot, monsoon season growing period July to October-November) and boro (cool, dry season growing period February to April-May) season rice were conducted in eight locations (Table 1) and compared with paired crops of conventional PT rice establishment during 2013, 2014 and 2015 in varied Agro-ecological Zones (AEZ) of Bangladesh ( Figure 1). Focus group discussions (FGD) were also held over three successive years to elicit farmers' opinions about the suitability, particularly, of the mechanized strip tillage NT and how their perceptions changed over time.

Experiment description
To account for the range of soils used for rice cultivation in north and northwest Bangladesh, 150 on-farm paired comparisons in the aman ( Figure 2) and boro ( Figure 2) seasons between NT and conventionally puddled PT rice establishment were conducted in eight locations (Table 1)

Farmer selection
Officers of the Department of Agricultural Extension (DAE), local leaders and farmers of the study locations (Table 1) were invited to field days, FGD and introduced to the past years' NT rice experiments (Haque et al., 2016). Each year, motivated farmers were provided with 1-day orientation training on NT rice production methods. From each location, six farmers were pre-selected to conduct on-farm experiments of NT of rice seedling in boro and aman seasons ( Table 1). Performance of each NT rice field (Table 1) was closely supervised to compare with a paired field established by conventional PT of rice seedlings.

Experiment location characteristics
During 2013-2015, a total of 66 and 84 paired trials for aman and boro rice, respectively, were conducted in Durgapur and Godagari upazila (sub-district) under Rajshahi district; Sadar and Gouripur upazila of Mymensingh district; Sadar upazila of Thakurgaon district; and Baliakandi upazila of Rajbari district (Table 1). These locations are known as High Ganges River Floodplain, High Barind Tract, Old Himalayan Piedmont Plain and Old Brahmaputra Floodplain (FAO-UNDP, 1988) and classified as AEZ 11, 26, 1 and 9, respectively. The soil types were varied in the experimental plots from alluvial in Durgapur upazila, silty-clay loam in Godagari upazila, to light loamy in Gouripur upazila and Sadar upazilas of Thakurgaon and Mymensingh (Huq & Shoaib, 2013). Among these locations, the highest average annual rainfall (2522 mm) was recorded in Gouripur and Sadar upazilas of Mymensingh followed by Sadar upazila of Thakurgaon (2218 mm) and the lowest rainfall (1581 mm) was recorded in Godagari and Durgapur upazilas in Rajshahi district ( Figure 2) but in all cases about 80% of the precipitation occurs during June-September. The minimum (in December) and maximum (in April) average temperature in Durgapur and Godagari upazilas were 11-36°C, 10-33°C in Sadar upazila of Thakurgaon and 12-32°C in Gouripur and Sadar upazilas of Mymensingh ( Figure 2).

Treatments and design of on-farm trials
The study was conducted in farmers' fields with NT and PT treatments each applied to half of a field.  (2) MoP (3) Gypsum (4) Zinc (5) Boron (  Each farmer's field (plots size was ranging from 660 to 1330 m 2 ) was treated as a replicate for both methods of rice crops establishment types. Treatments consisted of two rice establishment methods: (i) puddled transplanted (PT) and (ii) non-puddled transplanted (NT) of rice seedlings. All farmers used exactly the same management for the paired fields apart from crop establishment (PT vs. NT).

Land preparation
In case of land preparation for PT, three to four rotary tillage passes and one to two land levelling passes were done by a locally hired 2WT. Initially, farmers used two rotary tillage passes by 2WT to break up the soil which was left for 7-15 days, and then inundated by irrigation or rain water for 2-3 days. Initial tillage and water inundation for PT helps to kill germinated weeds and soften the soils. Finally, lands were puddled for PT with additional one to two rotary tillage operations and similar numbers of land levelling passes by 2WT in the inundated field. The lands were left for 1-3 days with standing water to allow suspended sediment to settle and to firm up the surface before manual rice seedling transplantation.
In case of NT, the land preparation was done by locally hired Versatile Mutli-crop Planter (VMP) (Haque et al., 2011). Generally, in dry land (having gravimetric soil water content from 15% to 40%), 50-70-mm-deep and 40-60-mm-wide strips (that preserved about 70-80% undisturbed soil) were made by VMP in a single pass operation. In boro season, the irrigation water was applied to the fields after strip tillage to inundate the field for 18-24 h before two to three rice seedlings hill −1 were transplanted manually in rows with hill spacing of 200 × 200 mm (Hossen, Hossain, Bell & Haque, 2017). In aman season, the strips were made prior to the onset of monsoon rain; however, in two cases in aman season the farmers used the VMP as single-pass shallow tillage where it was not possible to make the mechanized strip tillage due to standing water in the fields and because it was not possible to spray knockdown herbicide to control pre-germinated weeds due to standing water in the field.

Rice variety
In these trials, we did not influence the farmers' rice variety selection. Farmers used their preferred rice varieties in both PT and NT establishments.  (Table 1). The lifespan of hybrid-Tej, local-Jirashail, local-Swarna and HYV-BRRIdhan-48 in aman season is 130-140, 125-130, 135-145 and 105-115 days, respectively, and the potential grain yield is 6.5-7.0, 6.0-6.5, 5.5-6.0, 4.5-5.0 t ha −1 , respectively; the lifespan of HYV-BRRIdhan-28 in boro season is 140 days, and yield potentiality is up to 6.0 t ha −1 (Bangladesh Rice Research Institute [BRRI], 2018).   Table 1). The seedling numbers hill −1 were one, two to three, and three to five for hybrid, HYV and local varieties, respectively; and seedlings were transplanted manually at the row and hill spacing of 200 mm. Both PT and NT on each farm took place on the same day.

Nutrient management
The farmers were instructed to apply locally practised doses of fertilizers and apply the same rates on the paired PT and NT plots. The sources of added nutrients were urea for N, diammonium phosphate (DAP) for N and P, muriate of potash (MoP) for K, gypsum for S, zinc oxide (ZnO) for Zn and boric acid (H 3 BO 3 ) for B. In all aman seasons, the use of urea fertilizer was varied 150 kg ha −1 (in Choighati and Alipur during 2015) and more than 200 kg ha −1 in Jogonnathpur for 2014 and 2015 and Durbachor in 2015 (Table 1). The use of DAP and MoP fertilizers were higher in case of Jogonnathpur during 2014 and 2015 aman seasons (Table 1). Urea fertilizer rate varied from 150 to 240 kg ha −1 , DAP from 113 to 210 kg ha −1 , MoP from 75 to 113 kg ha −1 in boro seasons (Table 1). The dose of Zn and B fertilizers was 7.5 kg ha −1 but was not commonly used in all locations and years for aman season rice; by contrast, farmers commonly used the Zn and B fertilizers for boro season rice (Table 1). Except urea, all fertilizers were applied basally. The total urea was applied as three splits -50% at 3-5 days after transplanting (DAT), 25% at 30-40 DAT and 25% at 50-60 DAT (panicle initiation stage) in both seasons of all three years for on-farm experiments. The full amount of basal fertilizers, viz., DAP, MoP, gypsum, Zn and B, were hand broadcast for PT during final land preparation. For the NT plots, only DAP fertilizer was banded in strips during preparation of strip tillage by VMP; and other fertilizers, viz., MoP, gypsum, Zn and B, were hand broadcast for 4-24 h prior to rice seedling transplanting, when there was 2-3 cm of standing water in the fields.

Crop protection
In each season, two insecticides, viz., Virtako 40 WG (Thiamethoxam + Chlorantraniliprole), were applied at the rate of 75 g ha −1 to control brown plant hopper; and Imidacloprid to control aphid at the rate of 450 ml ha −1 .

Weed control
Non-selective herbicide glyphosate (3.5 l ha −1 ) was sprayed at 1.85 kg a.i. ha −1 for NT treatment in most of the on-farm experiments 1-7 days prior to preparing fields for strip tillage. However, as primary tillage and standing water in the field during puddling operation in PT helps to control exposed weeds, non-selective herbicide was used in PT plots. Most of the farmers used herbicide for both PT and NT to control posttransplanting weeds followed by a hand weeding at 25-30 and 35-40 days, in the case of aman and boro season rice, respectively. The pre-emergence herbicides, Pretilachlor, at the rate of 450 g ha −1 , and Pyrazosulfuron ethyl 20 WP, at the rate of 200-375 g ha −1 , were applied at 4-7 DAT and 15-20 DAT, respectively, for most of the plots.

Irrigation water management
The aman season rice was mostly dependent on rain water; however, supplementary irrigation water was applied if necessary to keep at least 2-3 cm ponding water during the rice-growing season. The lands were irrigated 1 day prior to final land preparation if there was not sufficient rain water for land preparation and transplanting of rice seedling. The boro season rice was grown in fully irrigated condition. In boro season, the irrigations were applied from land preparation and transplanting to grain-filling stage. The plots were initially supplied with 30-50 mm standing water at each of the irrigation events, and at the later stage 50-70 mm cm of standing water was supplied up to grain-filling stage. The lands were reirrigated after disappearance of standing water.

Crop harvest
In the boro and aman seasons, rice was harvested during the second week to end of May, and during mid-October to mid-November, respectively.

Data collection
At the beginning of each season, project field staff (PFS) were trained on the data collection methodology and provided an individual structured proforma for each onfarm experimental plots to collect the following details: general information of the location, crop establishment and management, various crop economic data, etc. The inputs and economic data were recorded based on the whole experimental plots, whereas the grain and straw yields were recorded from randomly pre-marked quadrats; and data on yield contributing characters were collected from randomly pre-selected hills. Within 3-5 DAT, the PFS randomly selected three quadrats of 1 m −2 from both NT and PT treatments of each plot and premarked them with pegs and strings. At the same time, three hills were tagged from each quadrat to collect yield contribution characters, e.g. effective and noneffective tiller number per hill, average plant height (cm) before harvest, panicle length (cm) and thousand grain mass (g). Grain yield at 14% moisture and sundried straw yields were collected from three premarked quadrats and converted to ha −1 basis. At the end of each season, the farmers were invited in FGD to collect the filled-up proforma, and to validate the recorded data; and compile farmers' feedback, experiences and perceptions about the NT rice crop performance. The FGD collected overall responses from each group and hence the information could not be dis-aggregated into individual responses.

Crop economic data
Total production cost was calculated considering all input costs, e.g. tillage and land levelling; seedling; fertilizers; irrigation; herbicide (knockdown, pre-and postemergence) and pesticides; labour (hired and family) for transplanting, weeding, irrigation, harvesting; and rent of land (Table 10). The gross production cost and gross income were calculated considering the estimated sale price of grain and straw (Table 10). Net return or net loss was calculated as gross incomegross production cost.

Data analysis
All statistical analysis on grain, straw and yield components, and input use were carried out by MSTAT-C statistical software (Michigan State University, U.S.A.) and Statistics 10. For each year and season, a one-way analysis of variance was carried out, treating each field (with paired NT and PT crops) as a replicate. Duncan's multiple range test (DMRT) was applied using the same programs and was used at the p < 0.01, p < 0.05 and p < 0.10 level to test the differences among the treatment means.

Crop production inputs
3.1.1. Number of tillage passes and cost for land preparation Across all years, seasons and locations, a single-pass operation was used for NT, but 3-5 tillage (average 3.94) passes and 1-2 levelling operations by 2WT were reported ( Figure 3) for land preparation in the case of PT. In aman seasons of 2013, 2014 and 2015, across all locations on average US$ 84, US$ 91, and US$ 93 ha −1 , respectively, were reported for land preparation cost for PT which were 65%, 63%, 61%, respectively, lower than for NT. Across all the locations, in aman season of 2014 and 2015, the highest land preparation cost was US$ 72 ha −1 for PT at Digram in 2014.

Labour use
The average total labour use in aman seasons of 2013 and 2015 for transplanting and weeding was higher (p < 0.01) for PT than NT (Table 2). Similarly, total labour use for transplanting and weeding in boro season of 2013, 2014 and 2015 was also higher (p < 0.01) for PT than NT (Table 3) Between NT and PT, no differences were observed for the total labour cost of aman season rice cultivation during 2013 and 2014; but higher (p < 0.01) cost for total labour use was recorded in 2015 in PT than NT (Table 2). Total labour cost for boro season rice cultivation was reported higher for PT than NT during 2013, 2014 and 2015 (Table 3).  (Table 5).          recorded for NT in aman season of 2015 over PT (Table 4). Thousand grains mass was similar for NT and PT during aman season of 2013. In the aman season of 2014, the thousand-grain mass was higher with NT in some locations but lower in others (Table 6). During aman season on 2015, heavier thousand grain mass was recorded after NT (Table 6 and Figure 5).

Grain and straw yield
In aman seasons, there was no effect of NT on grain yield of rice (Table 6)  had no effect on boro grain yield. The interaction effect between location and treatments (NT and PT) was observed in the boro season of 2015 where highest (p < 0.05) grain yield (6.10 t ha −1 ) was recorded in NT, but among the locations it was only statistically higher at Alipur (Table 7).
During the aman and boro seasons of 2015, higher straw yield was recorded for NT treatment than PT (Tables 6 and 7).

Gross production cost
The average gross production cost of aman season rice in 2014 for NT treatment was US$ 998 ha −1 which was US$ 15 ha −1 higher (p < 0.05) than PT (Table 8). In the aman season of 2015, gross production cost was US$ 59 ha −1 lower (p < 0.01) for NT than PT (Table 8). Lower (p < 0.01) gross production cost for NT over PT was consistently found in boro seasons of 2013, 2014 and 2015 (Table 9).

Gross income
Higher (p < 0.01) gross income (earning from grain and straw sales) was reported for NT over PT during aman season of 2014 (Table 8). No difference on gross income was reported between two treatments (NT and PT) in the aman season of 2015 (Table 8). Higher (p < 0.01) gross income was attained from NT than PT during the boro season of 2015 (Table 9).

Net return or loss
The net returns of 2013 aman season rice for NT and PT were US$ 109 and US$108 ha −1 , respectively, which were statistically similar. The net loss for cultivation of aman season rice in PT treatment was US$ 139 ha −1 in    2015, which was 59% greater (p < 0.01) than NT (Table  8) in the same year. The average net return for boro season rice cultivation with NT was US$ 411, US$ 82 and US$ 408 ha −1 during 2013, 2014 and 2015, respectively, which was higher (p < 0.01) than PT in all three years (Table 9). In aman season, 53 out of 66 farmers who practised NT reported higher net returns than in PT, while 49 out of 66 farmers reported higher yield with NT ( Figure 6). In boro season of 2013, 2014 and 2015, the net return was higher in 90-92% of cases in NT than PT, while 75% had the same or higher grain yield ( Figure 6).

Farmers' experiences and perception on NT
At the end of each aman and boro season, farmers' experiences and feedback on NT vs. PT were captured through farmers' meetings and FGD events. A total of 21 farmers' meetings and FGDs were conducted in different locations of the on-farm experiments where a total of 703 farmers (282 aman and 421 of boro seasons) attended and shared their experiences and perceptions as a group on the benefit and bottlenecks of NT. For consistency among the years, collected information from Alipur, Choighati and Digram locations were reported in Figure 7. Several farmers of Alipur locations were aware about the NT since 2008 (Haque et al., 2016), though the information on their perception and experience of NT was not previously collected. In this study, data were collected from Alipur location from the boro season of 2013 to aman season of 2015. After the first boro season, a total of 20 farmers attended the FGD at Alipur and reported various benefits of the NT over PT ( Figure 7); however, the percentage of farmers reporting the benefits increased over time (Figure 7). During the boro season of 2013, about 55% of farmers reported that the adoption of NT could reduce land preparation cost, which was increased (R 2 = 0.85) up to 92% after the sixth season (aman season of 2015) (Figure 7). Similarly, 50% farmers in the boro season of 2013 reported higher grain yield which increased up to 70% of farmers at the end of aman season of 2015. Farmers of Alipur also similarly reported on the use of reduced labour numbers ha −1 for weeding and less weed infestation in the NT plots ( Figure 7). Farmers' perception and experience on the negative aspects of NT declined over time at Alipur location (Figure 7). At the beginning (from boro season of 2013), 50-60% of farmers of Alipur reported that: it was difficult to transplant rice seeding in NT; labourers were less willing (and sometimes charged more) to transplant rice seedlings in NT than PT; it was difficult to make strips in hard setting dry soil for NT; initial water requirement (number of irrigation events prior to transplanting of rice seeding in NT, but not total irrigation amount) was higher; farmers were discouraged by the neighbouring farmers to try NT for rice seedling transplanting; and land appeared untidy due to transplanting of rice seedling in retained residue under NT systems (Figure 7). However, after six cropping seasons, the negative perception had declined (Figure 7). Similar trends on the farmer's experience and perception for adoption of NT  were reported from Choighati and Digram locations (Figure 7).

Discussions
From the 150 farmer-managed experiments during aman and boro seasons of three consecutive years, it was confirmed that rice cultivation commencing with NT of rice seedlings provided similar or greater grain yield of rice to that under full tillage and puddling of soil followed by transplanting (PT). Indeed 83%, 61% and 75% farmers obtained higher or similar boro rice grain yield in NT over PT during 2013, 2014 and 2015, respectively ( Figure 6). A similar trend was also found in the case of aman season where 67%, 75% and 75% farmers received higher or similar rice grain yield in 2013, 2014 and 2015, respectively ( Figure 6). This result indicated that although there were no substantial rice grain yield increases in the case of NT, the majority farmers obtained higher rice yield in NT (Figure 6).       whose potential benefit may take time to realize. There was also a substantial increase in positive perceptions and decline in negative perceptions about NT among farmers over the three years of gaining experience with the technology and gathering evidence of its benefits. This adds to previous results on NT (e.g. Ladha et al., 2009;Saharawat et al., 2009) which found that NT into zero tilled soil increased average rice yields on farmer's fields by 0.3 t ha −1 , whereas Ladha et al. (2009) reported a range of rice yield responses to zero tillage NT from −0.8 to +1.5 t ha −1 . Many previous studies on minimum soil disturbance crop establishment have also reported equivalent crop yields to traditional full tillage systems. However, most of the previous researchers had examined minimum soil disturbance vs. full tillage with rainfed crops (e.g. Sharma, Abrol & Sharma, 2011) rather than transplanted rice. For rainfed crops, Baker and Saxton (2007) report that it is common to experience some yield reduction in the first few no-tillage years, largely because it takes time for the soil to re-establish favourable soil structure after minimum tillage is implemented. If a transition period of yield reduction occurs, it can often be overcome or even averted with increased fertility, fertilizer banding with drill openers and careful crop selection (Baker & Saxton, 2007). Sharma et al. (1995) examined transplanted rice after one full tillage pass in wet soil to puddle the soil and reported similar rice yield to transplanting after full soil puddling involving several tillage operations; however, single-pass full tillage in wet land while reducing the number of tillage passes failed to take advantage of minimum soil disturbance and the benefits it affords. The present study now shows that transplanting rice seedlings into the narrow strips of disturbed soils (that covered about 20-25% of the soil) without puddling (i.e. NT) produces greater or at least similar yield to the conventional crop establishment by transplanting on fully puddled (i.e. PT) soils. The increases in grain yield of boro rice in 2015 with NT were associated with higher number of effective tillers hill −1 and higher grain −1000 weight ( Figure 5). This suggests that less post anthesis crop stress, along with farmers' improved skills of NT management, enhanced effective tiller numbers and grain mass which might explain the increased rice grain yield in the case of NT. In two experiments where soils have been continuously managed by CA practices (SP + NT plus increased crop residue retention) for 5 years, rice crop N uptake was increased in the post-tillering phase of growth (Md. K. Alam, personal communication). If the same improvement in post-tillering N uptake occurred in the NT crops in 2015, that might explain the increased grain weight, effective tiller number and grain yield. The beneficial effects of strip placement on rice crop N fertilizer availability are discussed below. However, further research is still needed to pinpoint the main physiological or nutritional factors that increase grain yield of NT rice crops.
Several studies, including Haque et al. (2016), reported that poor weed control was a threat to rice grain yield under NT. However, in the present study,  the application of the non-selective herbicide, glyphosate, prior to land preparation for NT was practised sufficiently well by farmers to control pre-germinated weeds in NT establishment of rice. Indeed, the labour requirement for weeding was lower in PT fields than NT rice despite the fact that the full tillage help to kill weeds before transplanting. In NT plots, on average US$ 4.5 ha −1 was required for killing pre-emerged weed (Table 10), whereas substantial amount of money (US$ 83 to US$ 103 ha −1 ) was required for land preparation (Table 10) including killing preemerged weeds of PT plots. Application of knockdown herbicide and minimum soil disturbance for NT might help to reduced post-transplanting weeding labours for aman and boro rice (Tables 2 and 3). Weed infestation is also a persistent constraint in direct seeding of rice in both wet and dry seeding that results in lower grain yield  and hinders widespread adoption by farmers (Farooq et al., 2011). Hand weeding in these farming systems is hampered by the increasing scarcity of labour and its high cost. Other than hand weeding, there are many pre-and post-transplanting herbicides available Zahan et al., 2014). Selection of effective herbicides will play a key role in maintaining weed control for PT and NT rice (Zahan, Hashem, Rahman, Bell & Begum, 2018). It remains unclear whether residue retention and the lower suspended sediment load in standing water of non-puddled soils alters the efficacy of the pre-emergence and early post-emergence herbicides commonly used in wetland rice. The use of herbicide in conventionally puddled rice fields (e.g. PT) to control weed is rapidly increasing in Bangladesh particularly for rice crops due to limited availability of labour . The farmers in the study areas used knockdown herbicide for NT, and selective herbicides for both NT and PT to control weeds, which ensured satisfactory weed control. Hossain et al. (2017) also confirmed that the higher level of straw mulching or appropriate selection of herbicide could successfully suppress weeds in NT rice fields. Specific crops and varieties might be better adapted to particular levels of soil disturbance and agronomic practices. Sayre (1998) reported that the crucial step in promoting bed-planting of wheat was to test a wide spectrum of varieties with differing heights, tillering abilities, phenology and canopy architectures in order to identify the plant phenotype for optimum performance on beds. The same is probably true for NT rice. The present rice cultivars have been selected under PT, but it remains unclear whether these cultivars are most suitable for NT. Hossain et al. (2017) reported that the rice cv. BR11 and BRRIdhan-46 produced higher grain yield and benefit cost ratio (BCR) in the NT than other cultivars but more rigorous assessment of cultivar suitability for NT is still needed. For minimum tillage methods, early vigour of transplanted seedlings would be beneficial for suppressing weeds (Dingkuhn, Johnson, Sow & Audebert, 1999). Hence, there is potential to achieve greater productivity of the NT rice through research to identify varieties better adapted to this planting approach.
Changes in nutrient availability and soil strength under NT may also favour cultivars with greater early rooting vigour or nutrient uptake efficiency (Borrell, Garside, Fukaj, & Reid, 1998). Initial shorter term studies on the N fertilizer requirements for rice under NT had shown no change in fertilizer N requirement vs. PT (Jahiruddin, Islam, Haque, Haque & Bell, 2014). These findings need to be verified in longer term studies in farmer-managed NT. However, the addition of DAP in the strip means that NT seedlings have greater access to N and P fertilizer after transplanting which may increase the early vigour of NT seedlings. Observations of NT seedlings in farmers' field suggest that the seedlings recover faster after transplanting and have darker green leaves and more early vigour (M.E. Haque, personal communication). This may be related to the close proximity of NT seedling roots to strip-placed DAP, as compared to the broadcast and incorporated DAP for PT seedlings. On the other hand, K, S, Zn and B fertilizers are incorporated into soil during the final tillage operation, while in the NT crop those fertilizers are broadcast on the soil surface but not incorporated. Hence, plant availability of K, S, B and Zn fertilizers applied may be greater under PT than NT.
In the study areas, cultivation of boro rice fully depends on irrigation water. The charge for irrigation water in the study areas was based on lump-sum fixed cost per ha. Hence, there is presently no decrease in cost of irrigation water when farmers use reduced volume of irrigation water as was the case for NT in the present study. Cultivation of aman rice in the study areas mostly depended on rain water. In several cases, supplementary irrigation water was required to cultivate aman rice in the study areas, so farmers hired irrigation pump on an hourly basis and recorded irrigation costs for NT and PT separately (Table 10). This record confirmed that the aman rice cultivation with supplementary irrigation water using NT saved about 33% irrigation water and cost over PT (Table 10).
Greater net returns (or minimized losses) were calculated for NT in aman season by 50%, 63% and 92% farmers in successive years. Due to low price of aman rice grain, a declining price trend (256, 231 and 205 US$ t −1 ), and increasing trend of gross production cost during 2013, 2014 and 2015, respectively, net losses occurred for rice cultivation in several cases in 2014, and most cases in 2015 ( Figure 6). However, overall farmers increased profit or minimized the net losses with NT relative to PT ( Figure 6). During harvesting period, the boro rice grain price was 244, 231 and 256 US$ t −1 in 2013, 2014 and 2015, respectively. Overall 92%, 94% and 92% farmers increased their net return (or minimized net losses) where NT were practised in boro season of 2013, 2014 and 2015, respectively ( Figure 6). Above results confirmed that while there were sometimes economic losses in both PT and NT in the farmers' fields due to low grain prices, the majority of farmers increased economic returns from NT, especially in boro season.
The transplanting of rice seedlings in puddled fields by hand is the common practice (about 100% of rice fields) in Bangladesh. However, labourers were initially reluctant to transplant rice seedling in NT fields perceiving that the soils in NT will be harder to transplant. However, the reluctance diminished steadily over time (Figure 7). On a sandy loam soil in south-west Bangladesh, Rashid et al. (2018) found a positive effect on rice yield of zero tillage NT with a mechanized transplanter, but not with manual transplanting. Wider application of mechanical rice transplanters may enhance the adoption of NT rice by removing concerns of labour about the slower process of transplanting in non-puddled soils which took about 2 person-days ha −1 longer and negated some of the decrease in labour required for weeding. Hossen, Hossain, Haque and Bell (2018) successfully developed a conventional rice transplanter of PT which could be used for both NT and PT; and on-station trial results confirmed that there were no significant yield differences using the transplanter for both NT and PT; however, greater cost and time saving, and higher BCR were reported in NT than PT.
Other than PT of rice seedlings, rice seed can be broadcasted directly in moist soils or sown by zero tillage. Due to labour shortages for rice seedling transplanting, wet direct seeding has become popular in Vietnam, Thailand, Lao PDR, Haryana and Punjab of India. However, wet direct seeding still needs puddled soil. Risks of severe weed infestation , bird damage, low productivity, uncertain and erratic rainfall , insecure irrigation water supply and high irrigation water requirement during early seedling stage are the major constraints to adoption of zero tillage direct seeded rice (Haque et al., 2016) hindering the replacement of puddled rice transplanting by direct seeding rice in most rice growing areas (Farooq et al., 2011).

Conclusions
Due to farmers' preference for puddled transplanting to establish wetland rice seedlings, adoption of CA in rice-based systems has been slow. From 150 farmermanaged evaluations in both aman and boro seasons during 2013, 2014 and 2015, we conclude that transplanting of rice seedlings in non-puddled soils following strip tillage is feasible as an option. Moreover, the strip tillage non-puddled rice transplanting reduced the cost of rice cultivation and increased gross margin for the rice farmers. Over the 3 years, 69% of boro season crops and 67% of aman season crops had higher gross margin with NT. Within three consecutive years comprising six rice seasons, there was generally no significant yield difference between NT and PT; however, in the boro season of 2015, NT produced significantly higher grain and straw yield of rice than the conventional puddling and transplanting. This suggests that the benefits from nonpuddled transplanting in terms of cost and greater grain and straw yields can be sustained over time. In farmers' fields, strip tillage, flooding soils for 18-24 h and then transplanting rice into non-puddled soil could be an option for rice establishment in ricebased systems which could enhance the adoption of CA.