1 Introduction

Rice (Oryza sativa L.) is an important crop that is consumed by over half of the global population as their staple food. The projected growth of the world’s population is expected to exceed nine billion by 2050, adding two billion to current levels (Max Roser et al., 2019; Stuart et al., 2016). This suggests that rice yields must increase globally at an annual rate of 1.5% (or 48.5% by 2050) in order to keep pace with growing demand and to ensure food security (Alam et al., 2013; Stuart et al., 2016). However, crop production faces enormous challenges, with limited availability of land, water, and labour (Cai & Rosegrant, 2002; Carrijo et al., 2017; Lampayan et al., 2015; Stuart et al., 2016; Zhou et al., 2017). Additionally, depletion of natural resources (e.g., soil, groundcover, and biodiversity) together with rising climatic stressors (e.g., droughts, floods, heat, and strong winds) increase pressure on farming systems (Montgomery et al., 2016; Stuart et al., 2016; Touch et al., 2016a, b). In this context, raising agricultural production in a sustainable manner is a critically important issue.

In a capitalist global economy, improving sustainable agricultural production will require that changes include the profitability of smallholders (Goodman & Redclift, 1986), who produce 70–80% of global food (Ricciardi et al., 2018). When aiming to increase crop yields, there is a need to understand and measure ‘yield gaps’, which are defined as yield differences between average farm level production relative to an adapted crop cultivar that is grown under the most favourable conditions without limitations from abiotic and biotic stressors (e.g., soil, nutrients, water, pests, diseases, temperatures, and light) (Fermont et al., 2009). Explanations for the resulting gap between ideal and ‘real-world’ are diverse and multi-faceted; for example, it may not be feasible for farmers to increase yields due to (1) input costs, (2) inefficient use of farm inputs, (3) risks due to exposure to natural hazards, (4) smallholders are often beset by exploitative relations, and (5) environmental degradation are just a few examples of limitations on ‘ideal’ productivity (Cunningham et al., 2013; Mueller et al., 2012). Given such constraints, it is valuable to understand farming systems in the context of farmer characteristics and perceived constraints, which can help to explain why yield gaps persist (Fermont et al., 2009), what shapes farmers’ understandings, decisions, and actions, and, potentially, to suggest feasible and sustainable options to reduce yield gaps. While the connection between closing yield gaps and smallholder profitability is contested (Tittonell & Giller, 2013; Zhang et al., 2016), and the production–distribution-consumption of food will be central to any sustainable system of food security (Taylor, 2018), for the purposes of this analysis, we assume that increased yields will be beneficial for farmers.

Numerous studies have indicated primary input and management interventions and strategies for achieving potential crop yields, including: (1) integrated pest, disease, and weed management; (2) advanced soil improvement and management; (3) market accessibility; (4) mitigation of weather impacts; (5) nutrient management; (6) the use of improved cultivars; and increasing supplementary irrigation (Davis et al., 2017; Mueller et al., 2012; Pradhan et al., 2015; Rosa et al., 2018). While these suggestions are useful conceptualizations, they are too abstract to put into practice for many smallholders, overlooking unique constraints as experienced by farmers.

For smallholders in Cambodia, many farm operations (i.e., land preparation, planting, agrochemical spraying, and harvesting) are performed by service contractors, with farm inputs (i.e., fertilisers, herbicides, insecticides, and so on) purchased on credit. This means that for most smallholder lowland rice farmers, there are extensive service costs coupled with a requirement that they repay borrowing costs as soon after harvesting as possible to avoid further interest charges. With more than 15 years of experience working and living in the region, the authors have noted that many service contractors, in general, rush through contracted jobs to finish them quickly rather than focus on the service quality. In addition, farmers frequently compromise the quality of their farm inputs and practices based on their household financial situation, the terms of input credits, the availability of farm machinery services, and the risks involved. These agricultural systems and arrangements situate farmers’ decision-making in the context of adopting new practices and/or technologies (Bryan et al., 2009; Cook et al., 2021; Esham & Garforth, 2013; Falconer, 2000; Guan et al., 2021; Khan et al., 2021; Stuart et al., 2014).

Many studies provided comprehensive analyses of different factors (e.g., production costs, production risks, complexity and practicability of implementations, product market and prices, etc.) that influence farmers’ uptake of specific interventions (Bartkowski & Bartke, 2018; Brown et al., 2021; Pirmoradi & Rostami, 2022; Van Herzele et al., 2013). Other studies have pointed out the need of recognising the constraints and concerns of farmers (such as financial resources, perceived risks, high transaction costs, limited action space for adopting diverse crop rotation and unrealistic regulations) when intending to sustainably improve agricultural production (Gütschow et al., 2021; Sanou et al., 2022; Tiet et al., 2022). As stated by ESCAP-UN (2021), for example:

“Smallholder farmers are playing an increasingly important role in transforming the food system, but they are often overlooked or under-recognized by institutions and other actors. Instead of being viewed as actors and partners, they are instead viewed as passive recipients” (p.24).

This research examines smallholder farming households’ characteristics and perceptions regarding the challenges that prevent them from adopting improved practices and/or technologies. It draws on these findings to inform the research and development priorities that will enable smallholders to adopt more sustainable and, ideally, more profitable practices. The resulting sustainable increases in productivity and profitability are assumed to contribute to both Cambodian and global food security, while also alleviating poverty amongst smallholders. The original contribution of this article is a real-world account of the constraints and limitations as experienced by smallholder paddy farmers in Northwest Cambodia, as well as their responses in the context of profit- and risk-oriented farming. The research fills these gaps, by highlighting how household demographics and socio-economics, labour availability, farming machinery availability, climate variability, irrigation availability, market volatility, soil degradation, profit- and service contractor-oriented farming, and input credit currently constrain and influence the practices of smallholding farmers. If we are to help farmers close yield gaps, we must address these issues directly.

2 Materials and methods

2.1 Descriptions of study areas

This study was carried out in nine communes in Banteay Meanchey and Battambang provinces in Northwest Cambodia (Fig. 1). In this area, the climate is affected by the Southeast Asian Monsoon (Chhinh & Millington, 2015), with a six-month dry season (November–April) and a six-month wet season (May–October). Based on the analysis of the long-term climate for the region between 1984 and 2019, annual mean rainfall for both provinces is approximately 1300 mm, with September and October generally being the wettest months with the average rainfall of the two months 445.9 mm (or accounted 34.1% of the total annual rainfall). The mean minimum and maximum temperatures for the region are 23.0 °C and 32.1 °C, respectively. March and April are the warmest months of the year with the mean maximum temperature of 36.2 °C, whereas December and January are the coolest months with the mean minimum temperature of 19.7 °C (Fig. 2). The following four seasons can be classified based on the characteristics of temperature and rainfall distributions:

  • November–January: Early dry season (cool and dry)

  • February–April: Late dry season (warm and dry)

  • May–July: Early wet season (warm and wet)

  • August–October: Late wet season (cool and wet)

Fig. 1
figure 1

Heatmap of long-term historical climate for regional northwest Cambodia: rainfall (a), monthly maximum temperature (b) and monthly minimum temperature (c)

Full size image
Fig. 2
figure 2

Study areas in Northwest Cambodia

Full size image

2.2 Data collection and analysis

This study adopted multi-stage and convenience sampling procedures. In our study, site selection processes began with consultations with agricultural research and development professionals with experience working in Cambodia. The northwest provinces of Battambang and Banteay Meanchey were proposed as the research focus areas because they are the country’s most important production areas for lowland rice production, allowing findings to be generalised across the country and serving as a strong case study for other regions with comparable cropping systems.

Due to limited resources and time, we were restricted to nine communes in the provinces of Battambang and Banteay Meanchey and purposely determined a sample size of 524 households to take part in the study. We then applied a random sampling method to select 524 household representatives from the sampling frame based on the records kept by local government authorities (Table 1).

Table 1 Locations and the number of respondents participated in the study
Full size table

We employed a survey methodology designed to gather farmers’ perceptions, attitudes, experiences, and practices (Bryman, 2016; Groves et al., 2009). Semi-structured interviews with a designated list of questions were collected using Android tablets running CommCare software. These surveys were followed by in-depth interviews and field observations (Fig. 3). Ethics approval was obtained from the Human Research Ethics Committee of The University of Sydney (No. 2016/882) with every farmer respondent providing written consent before the interviews began. Field data collection was implemented by 30 enumerators who were enrolled as undergraduate students from two local universities (namely National Banteay Meanchey University and the National University of Battambang). Prior to the interview implementation, a two-day training and a test run were organised for the enumerators. The interviews were conducted (using Android devices) on a two-on-one basis–where two enumerators (one questioner and one recorder) formed a small group to survey one household respondent. On average, it took 45 minutes to complete one interview. At the end of each interview, each enumerator group went through all the questions and checked if the answers given were correctly recorded before submitting the results to the central server.

Fig. 3
figure 3

Schematic diagram of the convergent mixed methods research framework used in this study

Full size image

In addition to the two-on-one interviews, the survey team leaders randomly observed different interviews, and chose to sit in on some interviews to capture key points for later interviews with the respondents if they were willing to spend an additional 10 to 15 minutes with the research team. Additionally, the research team leaders also conducted field observations to gain a better understanding of farming activities and other local contexts–considered to be useful for translating and explaining the survey results.

Survey questionnaires covered demographic and socio-economic characteristics of households, cropping production systems and operations, and constraints and difficulties faced in cropping production. These engagements were followed by semi-structured interviews that explored actions and proposed solutions to production constraints. The follow-up questions primarily revolved around the following themes: tell us more about this or that; why this or that; and what you think about this or that.

Our procedures for qualitative data analysis centered on data gathered in the form of text. We used an inductive approach (bottom-up) to develop themes (nodes) by reading the entire interview transcripts to identify general or specific patterns in the interview data. Then, using a deductive approach (top-down), we extracted textual data from interview transcripts and placed it in relevant themes (notes). Lastly, we developed categories by connecting relevant themes (notes) to construct narratives or answers linked to the specific research questions.

The R statistical program was used for quantitative data analysis and visualisation.

3 Results

3.1 Household demographic and socio-economic profiles

3.1.1 Household characteristic profiles

Collected household characteristics include the number of members, as well as their age, gender, and education levels (Fig. 4). These data were collected to explore the factors that shape farmers’ perceptions and decision-making in relation to farming practices. Distribution of household characteristics showed household size was mostly in the range of 4–6 persons, with the average and median being approximately five people (Fig. 4a). In most cases, one or two people in a family were reported to work full-time on the farms. While the other adult family members assisted occasionally, providing labour during busy periods. For those providing periodic labour, their primary source of income was derived from wage-labour with the aim of supplementing household income. Interviewed respondents explained that household size affected farm operations and represented a source of additional income generation. While it was essential to have some family members working off-farm, in order to diversify household income, farm operations tended to require more labour than was possible for two individuals. This meant that farming families also relied on hired labour and machinery service contractors to accomplish certain operations, namely: land preparation, seed sowing, applications of fertiliser and pesticide, and harvesting.

Fig. 4
figure 4

Household demographic profiles: (a) family members; and (b) age, (c) gender and (d) education levels of household head

Full size image

The data demonstrate that the gender of a household head is an important factor because of its influence on farm management and the income-generating capacity of the household. A large proportion of households (83%) identified males as their head-of-household compared with 17% who identified as female-headed households (Fig. 4c). The female-headed households (mainly widows) highlighted constraints and difficulties with farming labour. Female household heads explained that most farming operations, including land preparation, seed sowing, pesticide and fertiliser applications, and other machinery operations were largely outsourced or performed by male household members. Female-headed households’ higher dependence on hired labour and contractors meant that labour costs/ha/crop were higher than in male-headed households.

“As a female headed household, I find it is so difficult and inconvenient to do farming. Most farming activities (i.e., land preparation, seed broadcasting, spraying, harvesting, loading, transporting grain, etc.) are not suitable for women. It is too heavy for a woman like me to carry a full bag and basket of rice seed and broadcast it on the field. Similar to spraying, it is too hard for me to walk in the muddy and/or flooded field with a tank loaded with pesticide and water. I can do it, but it is not effective and is going to be very slow. For these reasons, I have to depend on service contractors and hired labour to do those heavy duties,” said a 53-year-old female farmer on 11 May 2017, Prey Tortueng Village of Battambang Province.

The data demonstrated that the age of a household head is another important factor, which influences decision-making and the overall management of the household and farm production. Our study found the mean age of the household head was approximately 50 years old (Fig. 4b); with an average level of formal education between 5 and 6 years. Anecdotally, we found that the age of farmers impacted their ability to conduct manual labour, which meant that some farmers reported not intensifying or expanding farm operations due to their physical limitations. Another field observation indicated that younger farmers were more likely to have tried new farming practices and technologies when compared with relatively elderly farmers. However, we also observed some elderly farmers who reported having attempted to improve farm practices and technologies. The findings, therefore, demonstrate variable associations between age and the adoption of improved farming practices and technologies.

In terms of literacy and formal education, the data demonstrated that nearly every household head (> 82%) had completed the first level of formal education (primary level) (Fig. 4d), suggesting that most are able to read and write at a basic level. Between 16 and 18% of the household heads had never attended school, implying that they were functionally illiterate. Very few household heads had been educated at a university level, including 1% from Banteay Meanchey and 2% from Battambang province (Fig. 4d). It was mentioned during the interviews that farmers who were more educated tended to implement new farming practices and technologies as well as communicate updated information associated with farming (e.g., using new crop varieties, adopting drones for fertiliser and pesticide applications, adopting laser land levellers, etc.). Some elderly farmers told us during interviews that farmers with a relatively higher level of education were more confident, creative, and successful in terms of achieving high yield and returns on investment because they understood farming inputs, technologies, and practices, as well as the broader agricultural system.

3.1.2 Farm household income diversification

Distribution and the breakdown of household composition are shown in Fig. 5. The data demonstrated that the average total annual incomes of households in Battambang and Banteay Meanchey (Fig. 5c) were US$3,300 (Std Dev = $2,692) and US$2,953 (Std Dev = $3,690), respectively, and were not significantly different (F = 1.464, p = 0.227). These data indicate that farmers not only depended on farming as their primary source of income but also engaged in different economic activities, sustaining diversified livelihood strategies. Broadly speaking, farmers across both provinces had a similar diversity of household incomes. However, the proportion of incomes from off-farm labour (including remittance) and non-farm income (i.e., small business, rents, services, etc.) was higher among farming households in Banteay Meanchey (39%) than those living in Battambang (23%) (Fig. 5a–b). In villages where on-farm income was low, off-farm income was much higher–with four of five villages in Banteay Meanchey more reliant on off-farm income attributed to household member’s temporary migration for work. Furthermore, in three of these villages, remittance from temporary migration constituted approximately a quarter of total income. Income diversification was explained by respondents as a favourable risk management mechanism in dealing with high variability in farm and non-farm incomes.

Fig. 5
figure 5

Distribution and diversity of annual incomes for smallholder farms in Banteay Meanchey (BMC) and Battambang (BTB) province. Proportion of farm household composition for farmers in Battambang (BTB) (a) and for farmers in Banteay Meanchey (BMC) (b), and total annual income distribution presented in Boxplot with mean value of each site (red diamond) and mean value of all sites (red dashed line)

Full size image

Farmers described their incomes as highly variable, particularly in relation to crop production, which was highly susceptible to stressors such as weather (drought and flood), pests, and grain price, with the average rice price ranging from US$0.12 to US$0.19 per kg in 2016. In a bad year, either from poor crop yield or low grain price, there was an immediate adverse effect on the overall welfare and performance of the household, including their willingness to invest in crop productivity. Most households managed income deficiencies by going into debt, with 37% of all households surveyed relying on microcredit loans. The average loan was approximately US$2,000 with an average interest rate of 2% per month or 24% per annum. The majority of these loans were used for crop inputs and household expenditure. Purchasing inputs on credit reduced opportunities for capturing post-harvest value as debts had to be paid at harvest time. This means that a poor harvest, or successive poor harvests, had the ability to trigger a negative debt cycle (Green, 2020). While savings groups via agricultural cooperatives are an option in some villages, only 9% of respondents participated in such ventures. Anecdotally, to cope with financial stress, some farmers explained that they tended to use less and lower quality farm inputs when they had less money available to spend or could only take on a certain amount of debt. This constrained decision making suggests serious negative implications for farm efficiency and profitability.

With respect to off-farm labour income, it involved temporary or permanent migration to other provinces, major cities, and/or other countries. In Banteay Meanchey, it was common to have some family members, especially the younger generation, working in construction in Thailand (which shares a border with Banteay Meanchey). This is represented in the data with approximately 28% of surveyed households in Banteay Meanchey explaining that they had at least one household member migrating for off-farm work compared with Battambang where approximately 11% (of surveyed households) had at least one member migrating for off-farm work. Non-farm income diversification was explained as a vital measure in maintaining a broad portfolio of household income sources–and that non-farm income was seen as more reliable and regular, especially during bad agricultural years. It was also explained that off-farm labour is available for young and healthy individuals, meaning that they are more likely to return home when they get older and feel unfit for the jobs available to foreign labourers. Remittances sent by household members were often used to pay off debt, to purchase farm inputs (i.e., fertilisers, pesticides), and to cover daily expenses on household needs.

3.1.3 Farmland and irrigation

The data demonstrated that farm size is an important factor in terms of household income. While smaller farms can be more productive than larger farms, they frequently do not generate adequate income to sustain the livelihood and welfare of the household. Alternately, the data demonstrated that large farms provide farmers with greater opportunities to produce more crops and therefore income–although this also means potentially higher debt levels and more risk. This research sought information about ownership and lease of farmland among cropping farmers in Northwest Cambodia (Fig. 6). Of the surveyed households, almost every household (96%) owned farmland distributed in small fields with the average land size per household being 3.57 ha (Std Dev = 4.2 ha) in Banteay Meanchey and 2.8 ha (Std Dev = 2.5 ha) in Battambang province (Fig. 6b). A small proportion of farm households (23%) leased an additional paddy field of less than a hectare (0.74 ha on average Std Dev = 2.8 ha) from other farmers (Fig. 6e). This brought the total land use by each household in Banteay Meanchey and Battambang to 4.25 ha (Std Dev = 5.98 ha) and 3.56 ha (Std Dev = 2.97 ha), respectively (Fig. 6a).

Fig. 6
figure 6

Ownership and lease of farmland amongst the smallholder farms in Banteay Meanchey (BMC) and Battambang (BTB) provinces of the north-western Cambodia. Red dashed lines denote the average in each province: (a) Total farmland (included rented-in land) being cultivated, (b) total farmland ownership, (c) total farmland rent-in, (d) percentage of farmland under rainfed and irrigation condition, and (e) percentage of farmland ownership, rent-in and rent-out. Red dashed lines denote the mean

Full size image

As reflected in the literature (Davis et al., 2017; Mueller et al., 2012; Rosa et al., 2018), irrigation is of paramount importance to enhancing cropping productivity and minimising risk from weather impacts (i.e., drought, flood, etc.) amongst participating farmers. The data demonstrated that irrigated cropland accounted for 26% and 55% of farmland in Banteay Meanchey and Battambang, respectively. Farmers with irrigated cropland explained that they grew two rice crops per year and, anecdotally, felt relatively secure in the context of climate, allowing them to invest greater amounts into farming inputs. Nonetheless, farmers who were able to irrigate retained significant uncertainty concerning the timing and availability of water. For the non-irrigated fields, farmers grew one crop each year and they were averse to investing in inputs. This difference between Banteay Meanchey and Battambang was distinct–in summary, farmers in Banteay Meanchey had larger rainfed farms but, on average, less income from cropping; farmers in Battambang had smaller irrigated farms but, on average, higher incomes from cropping. These data suggest that access to irrigation limits farmers from adopting and implementing cropping options that could potentially reduce environmental burdens, enhance productivity, and generate economic benefits.

In summary, with respect to farm household incomes, the results demonstrate that farm size and ownership, farming type and intensification, farmland being irrigated or not, climate, market price of agricultural produce (viewed as being a good or bad year), household members engaging in on-farm and non-farm activities, and availability of job opportunities were the primary determinants of farm household incomes.

3.2 Lowland crop production

In the lowland, rice was grown commercially by almost every surveyed household (97.5%), while other crops such as vegetables, legumes (e.g., mungbean, peanut, etc.), and sweet corn were reported to be planted by a very small proportion of farming households (Table 2). This practice means that rice was apparently not rotated and/or grown in combination with other crops on paddy fields. These interview findings are supported by data obtained from the field observations, in which non-rice crops were planted by a very small number of households on small fields located next to water sources (mostly ponds and rivers) and farmers’ houses and/or villages. The research team noticed that the farmer surveys did not capture the few farmers who grew watermelon and muskmelon (Cucumis melo L.; locally known as Rice Cucumber), sometimes planted a few months prior to the start of the wet season and/or just after harvesting the main wet reason rice.

Table 2 Number of surveyed households reported to have grown crops commercially
Full size table

Maturities of rice grown in the study areas were of short duration (< 120 days from sowing to maturity), medium duration (120–150 days), and late duration rice (> 150 days) (Fig. 7). A trend towards adoption of a shorter duration rice variety is crucial for implementation of an adaptive strategy in response to climate change and to make more efficient use of water availability. As shown in Fig. 6, there were differences between provinces, with Banteay Meanchey more reliant on late duration crops and Battambang more reliant on early and medium duration crops–which gave households in Battambang the ability to grow 2 rice crops per year. Avoiding flood damage during the peak rainfall months in the wet season and having access to water for irrigation were explained as the main factors for switching to short and medium duration rice varieties.

Fig. 7
figure 7

Proportion of crops grown commercially in the lowland study areas (a) and different rice maturities grown in the study areas (b)

Full size image

Each interviewed household produced an average total paddy production of just below 9 tonnes annually (Fig. 8a). Of the total production, approximately 82% was for sale, and the remaining was kept for household consumption (12%) and seed (6%) (Fig. 8). While Battambang households yielded slightly more rice per ha, the average yield of the two provinces was not significantly different, with the average of 2.818 t/ha (Std Dev = 1. 28 t/ha), with the median of 3 t/ha and the interquartile range (IQR) of 2.0–3.6 t/ha (Fig. 8e). The top 10% of the farmer respondents from both provinces achieved a paddy yield of 4.41 t/ha. This suggests a local attainable minimum yield gap of 1.59 t/ha exists, which translates into US$299.87 of potential additional income per hectare.

Fig. 8
figure 8

Paddy production from the past year harvest 2016 and distributed proportions of paddy used by smallholder farms in Banteay Meanchey (BMC) and Battambang (BTB) of the north-western Cambodia. Each red dash-line denotes the average across both provinces. The red dot-line (e) denotes yield achieved in t/ha by the top 10% of farmers in both provinces. The red dashed line in the boxplot between the 1st quantile and 3rd quantile presents the average value across both provinces

Full size image

3.3 Understanding farm-related problems and farmer adaptation

Following the farm household survey interviews, we conducted in-depth interviews aiming to gain a better understanding of the problems, difficulties, and challenges that farmers confront, as well as their responses to those constraints. The following primary questions were used to frame our in-depth interviews:

  • What have been the significant problems and challenges you have confronted?

  • Where and how have these problems and challenges affected your family and farming?

  • What measures and actions have you taken to prevent, prepare for, and mitigate these problems?

  • Why have you chosen these measures or actions over others?

We found that farmers were confronted with many profound problems, stresses, and pressures while attempting to sustain their livelihoods. We classified their problems and challenges into three main categories: biophysical, socioeconomic, and environmental problems. We acknowledge that these categories may overlap in terms of drivers and effects. Insect pests, diseases, and weeds–for example–may be caused by a combination of biophysical, socioeconomic, and environmental factors. Additionally, their impacts and consequences may overlap and extend beyond the three primary categories. Table 3 summarises the most important farm-related problems and stresses, as reported by farmers in the study areas.

Table 3 Primary farm-related problems and their effects on small-scale farmers in Northwest Cambodia
Full size table

3.3.1 Biophysical constraints and adaptation

Within the biophysical constraint category, important problems and constraints were found to be compacted and infertile soil, weeds, and inefficient crop establishment and harvesting. Each problem and its associated stressors are discussed below.

Soil compaction: During the interviews, farmers claimed that their soil was becoming harder and more infertile over time. They observed changes in soil properties, including substantially lower fertility, lower organic matter, rising compaction, and lower pH. Together, these impacts are likely to limit rice yield (Edralin et al., 2017; Singh et al., 2016; Sumner et al., 2016).

“Like many other farmers, soil fertility of my farm continues to decline; and it becomes harder. In the past, my farm soil was so fertile and not that hard. We did not use any fertiliser. But these days, we need to apply additional fertiliser to obtain good yield,” said a 67-year-old male farmer on 10 May 2017, Svay Cheat village of Battambang Province.

Most farmers perceived that having a deep tillage was important because it enabled the roots of the rice to penetrate and distribute into the soil, which is why they chose to have their paddy fields ploughed by a four-wheel drive tractor for the first ploughing. They emphasised that the main purposes of ploughing and/or rotavating were not only to break down soil mass but also to kill weeds, incorporate crop residues to speed decomposition, and level the soil surface for seedbed preparation. Farmers were charged by a service contractor at an average rate of US$30/ha for each ploughing and US$15/ha for each harrowing. For the second cycle of rice, instead of ploughing, fields were typically rotavated twice–the first and second rotavations cost on average US$46 and US$38 per hectare, respectively.

Participants were asked if they had used fertilisers in their paddy production, with a large proportion of farmers surveyed–90% in Banteay Meanchey and 76% in Battambang–having used inorganic fertilisers in their crop production. They explained that composting fertiliser and cattle manure were not commonly used in paddy production due to the wide availability of inorganic fertilisers and that cattle production is no longer a common practice–only 30% of farmers surveyed kept cattle. These factors have led to dependence on chemical fertilisers to supplement soil nutrient supply and to enhance crop yields.

Farmers from both provinces used a mixture of fertilisers with an average application of 93 kg/ha (Std Dev = 63.82 kg/ha) (Fig. 9). When discussing splitting and timing fertiliser applications, farmers advised that they often had two splits of fertiliser application: first at the tillering phase and second at flowering. Farmers were unable to describe what fertilisers they were using and distinguished the differences between fertiliser types through fertiliser company logos printed on the bags and fertiliser colours (e.g., white, brown, or mixed colours), rather than by chemical compounds. Based on what we saw on the fertiliser bags shown by some respondents during the interviews, as well as information given by input suppliers, the fertilisers used were mostly Urea, DAP, N-P-K, N-P-K-TE and N-P-K-S-TE, where TE refers to trace elements.

Fig. 9
figure 9

Use of inorganic fertiliser in paddy production by farmers in Banteay Meanchey (MBC) and Battambang (BTB). The boxplot provides a compact view of a distribution of fertiliser use rates. The diamond in the boxplot denotes the mean value, and black dots are the outlier

Full size image

Crop establishment: The surveyed data showed that every household (100%) used hand-broadcasting as their rice planting method. The planting method was described as being implemented in two distinct ways: dry seeding and wet seeding, with the method chosen based on field conditions. In the case of dry fields, farmers sowed dry seed by hand uniformly onto the soil surface of the recently ploughed field, then incorporated the seed into the soil by harrowing. The farmers claimed that they adopted hand-broadcasting to reduce production costs (between $5 and $8 per ha), increase speed, and reduce labour requirements. Historically, manual transplantation was prevalent, but this practice has transitioned to hand broadcasting due to high costs, the need for irrigation, and the work involved in nursery preparation for seedlings, pulling, bundling, carrying, and transplanting seedlings. However, some respondents highlighted disadvantages (Fig. 10c) such as high seeding rates (often between 160 and 220 kg/ha), uneven plant distribution, more weeds, increased costs for weed control, and unhealthy rice. This inefficient planting practice has resulted in low yields.

Fig. 10
figure 10

Combine harvesters damage to the soil when harvesting on the wet field (a); combine harvesters spill rice grain back in the field (b); pre-germinated rice seed stays on the soil surface from the use of hand-broadcast sowing method (c)

Full size image

Weeds: The surveyed respondents reported that weeds were among the main production challenges in direct-seeded rice. This study found that herbicides were used primarily to control weeds in the target locations, whereas in the past, when herbicides were unavailable, the fields were weeded exclusively by hand. Almost every interviewed household (92%) reported that they used herbicides as their primary weed control method, with only a small proportion of farmers (6%) removing weeds manually. The central reasons for adopting herbicides were agricultural labour shortages and high labour costs. In addition, hand-weeding in broadcasted paddy was reportedly arduous and time-consuming; however, it was commonly observed that farmers (field owners) would do some hand-weeding while inspecting their fields.

Since many of the interviewed farmers were unable to tell us which herbicides, insecticides, and fungicides they used, we performed impromptu interviews with five agrochemical input suppliers. Data from these engagements conveyed that the most common herbicides used by farmers in northwest Cambodia were either pretilachlor or butachlor as the pre-emergence herbicide (with our findings showing that this was used by few farmers), and broadleaf and grass weed herbicides such as 2,4-D, cyhalofop, metsulfuron, fenoxaprop, and bispyribac-sodium as the post-emergence application (used by most farmers in this study). Pre-mixed products of broadleaf and grass weed herbicides were commonly used. However, depending on the types of weeds that emerged in the field, farmers might add an additional herbicide–either a broadleaf or grass weed herbicide–together with the pre-mixed product in one application. We observed that pre-mixed products were also commonly used. Farmers had generally spent between US$10–15/ha for pre-emergence herbicide application and US$30–50/ha for post-emergence herbicide application. This information provided by input suppliers correlated strongly with the survey results. It appeared that farmers focused on weed control options (i.e., land cultivation and post-emergence herbicide application) rather than a weed management strategy, which is defined as an integrated and long-term approach to reduce, prevent, and manage weeds (Swanton et al., 2008).These findings indicate that the combined constraints of biophysical, socioeconomic, and environmental factors have driven farmers to rely heavily on agrochemicals for weed, pest, and disease control. This reliance on agrochemicals in farming has serious implications for both the agroecosystem and human health.

Harvesting: Traditional hand-harvesting of rice consists of cutting matured rice, bundling, field-drying, staking/pilling, hauling, threshing, and cleaning grain through winnowing. These practices require extensive labour, have high costs, and involve much time. The average breakdown of expenses per hectare using manual harvesting were: hand-reaping US$144 ($6.25 * 23 person-day), collecting and piling the cut rice US$38 ($12.50 * 3 person-day), manual threshing US$75 ($12.50 * 6 person-day), and grain cleaning US$25 ($12.50 * 2 person-day). Relative to these costs, a combine harvester was a one-time cost of approximately US$70 per ha. This, compared to the traditional hand-harvest, indicated that farmers could save approximately US$200/ha by employing the combine. Besides a large cost reduction, farmers described paddy harvesting by the combine harvester as very convenient with no waiting-time required. It combined all necessary harvesting activities into a single operation, and the machine produced clean grain and discharged it into bags.

Of the surveyed farmers, more than 94% employed combine harvesters to harvest their paddy. Our sample of farmers indicated that mechanised harvesters did not cause any problems or threaten job opportunities amongst local people. From their perspectives, these technologies were deemed essential and beneficial in the context of labour shortages and rising labour costs. However, farmers pointed out some drawbacks resulting from the combine harvester (Fig. 10a–b). One was the loss of much grain blown away with rice straw back to the field. The grains blown back into the field cause issues with volunteer rice infestation, which had adverse effects on the yield, quality, and price of the harvested paddy. Another issue arising from the use of combine harvesters was soil compaction when harvesting on wet and soft fields. These findings provide a compelling explanation for why farmers adopt this combine harvester operation, and the trade-offs that they undertake as part of that adoption process. It remains unknown how much rice was lost during harvesting nor the impacts of soil compaction on future productivity, presenting opportunities for future research.

Household debt: Most households managed income deficiencies by going into debt, with 37% of all households surveyed holding loans. These loans were used primarily for crop inputs and household expenditures. Purchasing inputs on credit reduced opportunities for capturing post-harvest value because, per the terms of the agreement, debts must be paid back at harvest. This means that successive poor harvests have the ability to trigger a negative debt cycle.

3.3.2 Socioeconomic constraints and adaptation

Poverty, household debt, high agricultural input costs, and low grain prices were reported as the most significant socioeconomic constraints. Farmers described their incomes as highly variable, particularly in relation to crop production, which was highly susceptible to those stressors. In a bad year, either from poor crop yield or low grain price, there was an immediate adverse effect on the overall welfare and performance of the household, including investment in crop production. Most households managed income deficiencies by going into debt, with 37% of all households surveyed relying on microcredit loans. The average loan was approximately US$2,000 with an average interest rate of 2% per month or 24% per annum. The majority of these loans were used for crop inputs and household expenditure. While savings groups via agricultural cooperatives are an option in some villages, only 9% of respondents participated in such ventures. Anecdotally, to cope with financial stress, some farmers explained that they tended to use less and lower quality farm inputs when they had less money available to spend or could only take on a certain amount of debt. This constrained decision making suggests serious negative implications for farm efficiency and profitability.

“Our family experienced many difficulties and challenges. Our paddy production had been severely affected by prolonged droughts. Agricultural inputs (i.e., fertiliser, pesticide, etc.) are getting incredibly expensive. We have no solution to these critical problems. We need farming methods with low production costs and high yields,” said a 67-year-old male farmer on 10 May 2017, Svay Cheat village of Battambang Province.

3.3.3 Environmental constraints and adaptation

Changing rainfall patterns, unreliable rainfall distribution, prolonged drought periods, and excessive and extended periods of rainfall were identified as the most important environmental factors affecting farming households in the study areas.

“Last year, our paddy production was greatly affected by flood at the mature stage. The grain was soaked with water and had a bad smell. It was so hard on us,” said a 35-year-old male farmer on 5 May 2017, Trapeang Thma village of Banteay Meanchey Province.

The data indicates a trend toward the adoption of a shorter duration rice variety, which was an adaptive strategy to these climatic constraints and an attempt to make more efficient use of available water. Farmers claimed that avoiding flood damage during the peak rainfall months of the wet season and having access to water for irrigation were the primary reasons for adopting short and medium duration rice varieties. They acknowledged that many environmental problems were difficult and, in most circumstances, uncontrollable. They reported experiencing both drought and flooding in the same year. They were uncertain about when to begin planting and carrying out other farm operations because rainfall was perceived to have become variable and unpredictable.

3.4 Farmers’ adaptation constraints and future research and development prospects

The attempted to investigate key drivers and constraints that influenced farmer decisions and adaptation to specific problems, seeking an empirical basis for discussion of gaps for future research and development initiatives.

Farmers often devised and employed various strategies and interventions to avoid and/or lessen the effect of multiple and complex stressors. Their choices of adaptation mechanisms can be driven by multiple factors (e.g., risks, resources, information, markets, etc.), and their interventions may lead to other problems. This study found that, in response to the compaction and soil nutrient depletion problems, the majority of farmers (88%) chose to plough their fields more to break up the soil mass. They perceived that having a deep tillage was important because it enabled the roots of the rice to penetrate and distribute into the soil, which is why they chose to have their paddy fields ploughed by a four-wheel drive tractor for the first ploughing. However, their excessive land cultivation practices may importantly contribute to and accelerate land degradation (Abdi et al., 2013), causing a decline in soil fertility, an increase in soil erosion, and reduced soil water retention (Abdi et al., 2013; Kaiser, 2020). Collectively, these factors have long-term negative implications for farm productivity, profitability, and the environment (Kaiser, 2020).

Chemical fertilizers currently play a vital role in agricultural food production and soil fertility. While chemical fertilizers can increase crop yields, they also impact soil physicochemical and biological characteristics. The continued and excessive use of chemical fertilizers contributes to the decline of soil organic matter content, soil hardness, decreased soil fertility, soil pollution, and depletion of important nutrients and minerals (Pahalvi et al., 2021). Farmers were asked if they had used chemical fertilizers in their paddy production, and most of them (83%) said that they had in order to maintain and/or increase crop yields. The data indicates that farmers from both provinces used a mixture of fertilisers with an average application of 93 kg/ha (Std Dev = 63.82 kg/ha) (Fig. 8).

The findings, as reported on p. 18, suggest that soil compaction and fertility depletion have driven farmers to engage in excessive land cultivation and to apply more fertilisers in order to maintain and/or increase yields. This upward trend of land cultivation and reliance on chemical fertilisers is unsustainable, necessitating research and development to help farmers restore and maintain their soil health. This indicates that these research and development prospects for restoring and preserving soil health are vital for promoting sustainable agricultural and ecological systems.

Regarding crop establishment practices, farmers have adopted manual broadcasting due to labour shortages, labour wage increases, reduced production costs, and climate-related risks. While hand broadcasting has demonstrated to drastically lower planting costs to less than 10 US$/ha, it has also led to increased seeding rates of more than 200 kg/ha, as well as increased incidences of weeds, crop pests, and diseases. These increases in crop pests have driven farmers to rely on agrochemicals, which have degraded agroecological systems and rendered agricultural production unsustainable in the long run. This suggests that future research, refinement, and development initiatives on mechanised direct seeders that are effective and suitable for both wet and dry paddy could have profound impacts on the lives of millions of smallholder farmers in the study areas. Similarly, this recommendation is given for combine harvesters.

Another important problem shared by farmers was weeds. This study discovered that herbicides have been the primary method by which farmers control weeds in their paddy fields. Their experience with weed control has proven that just employing herbicides to control weeds is ineffective due to the wide variety of weed species with distinct life cycles. In addition, controlling weeds with the same herbicides repeatedly can allow weeds to develop resistance to those herbicides (Jabran & Chauhan, 2018; Korres, 2018). In order to effectively manage weeds and other agricultural pests, as well as to minimise herbicide-resistant weeds and reduce the likelihood of human exposure to herbicide residues, this study recommends creating/raising farmer awareness about integrated weed management practices and developing systems that enable them to incorporate multiple methods for weed management (e.g., prevention methods, suppression methods, chemical weed control methods, biological weed control methods, mechanical weed control methods, and so on).

4 Discussion

Farmers were found to be confronted with a range of challenges and concerns, including soil degradation, weeds, diseases, insect pests, high production costs, flooding, droughts, changing rainfall patterns, and unreliable rainfall distribution. To avoid and lessen the effect of those stressors, they have devised and employed various strategies and interventions, particularly a strategy to diversify income from several sources. These findings are consistent with those found in Bojnec and Fertő (2019); Bojnec and Knific (2021); Giaccio et al. (2018); Wu et al. (2018). Their strategy has led to an upward trend of temporary and permanent migration to cities and other countries, which results in farm labour shortages and labour price increases. These findings are reinforced in part by Pender and Gebremedhin (2008)‘s study that managing household income with off-farm activities has reduced agricultural labour intensity. These labour issues, together with high dependencies on hired labour and machinery service contractors, cause a substantial increase in production costs, which may explain the rational adoption of inefficient, low-cost farming practices and low-quality farm inputs (Balwinder et al., 2020; Chauhan et al., 2012).

As shown in the Results, the top 10% of the surveyed households achieved a rice yield of 4.41 t/ha with a few farmers reporting yields greater than 7 t/ha, while the majority received the median yield of 3.0 t/ha, which is 32% less than the yield obtained by the top 10% under general farmer field conditions (Fig. 8). This is very similar to the findings of Chhun et al. (2019) in four communes of Battambang province where the top 10% and average yield of farmers are 4.6 t/ha and 3.3 t/ha, respectively. Surprisingly, a field experiment conducted in 2017 at the Don Bosco farm in Battambang (the same location as the study areas) revealed that the top yield was 7.0 t/ha, giving an exploitable yield gap of 4.0 t/ha (Tan et al., 2018). This indicates the current average farm yield can potentially be doubled.

Farmer households’ income is shown to be highly variable due to reliance on crop production which is highly susceptible to stressors. An immediate adverse effect on the overall welfare and performance of the household (Ali et al., 2019; Wossen et al., 2018), including investment in crop production, is variable income from crop production, which affects future behaviours. This finding underscores a negative feedback relationship in which variable or low yields reduce the capacity and willingness to invest in inputs, which entrenches existing practices and declining soil fertility.

The current hand-broadcasting method is found to be one of the main yield limiting factors. This manual broadcasting method has two limitations: (1) non-uniform plant density distribution; and (2) seeds that stay on the soil surface. Tan et al. (2018) suggests that improving efficiencies in crop establishment (e.g., by using a mechanised seed drill to plant seeds into the soil surface and in rows) and seeding rate (less than 100 kg per ha) could contribute to increased crop productivity and profitability. On-farm trials, in two of the target locations included in this study over three seasons (i.e., Dry Season, Early Wet Season and Main Wet Season), have demonstrated that manual broadcasting can be productively and economically replaced by 4-wheel tractor drill seeders for dry paddy, and 2-wheel tractor mechanised drum seeder and Eli Air seeder for muddy fields (Tan et al., 2018). The research has also shown that the seeding rate could be reduced to well below 100 kg/ha, while maintaining and sometimes increasing productivity. On-farm field research demonstrated that rice seeds planted into the soil surface using the mechanised seed drill retained a high germination rate even after approximately one month without rain. Under the same field and rainfall conditions, the experimental plots and farmer fields seeded by hand-broadcasting where the seeds remained on the soil surface had a very low germination rate, necessitating reploughing and replanting (Tan et al., 2018). This means that by implementing mechanised methods of crop establishment, farmers can potentially reduce the impacts of drought during the early stages of crop growth while also minimizing land preparation operations and input use (seeds, fertilisers, pesticides, and herbicides).

Another major constraint in direct seeded rice production was yield loss from weed competition (Abdullah Al Mamun, 2014). Weed issues in direct seeded rice are greater than in transplanted rice, due to the lack of water to suppress weeds during plant emergence (Chauhan and Johnson, 2012). Application of post-emergence herbicides is the primary weed control practice found in this study. Farmers are not always aware of pesticide active ingredients and safe modes of application. This results in farmers not always following the instructions, raising concerns for safety and correct use of chemicals. Farmers from our sample rarely wore safety equipment while mixing and spraying the pesticides. This suggests serious adverse risks and impacts on human and environmental health. These findings are consistent with the work of Damalas and Abdollahzadeh (2016); Sharifzadeh et al. (2019); particularly in Chhun et al. (2019) where they demonstrated that although 94% of farmers responded that they were aware of pesticide exposure risk, use of protective equipment was low (10% and 54%, respectively). Farmers rely on prior experience, neighbours, and input sellers for information on chemical and fertiliser purchases and applications, which is also consistent with the findings of Chhun et al. (2019); Damalas and Abdollahzadeh (2016); and Sharifzadeh et al. (2019). These researchers demonstrated that the majority of farmers relied on the advice of input dealers for the selection and use of herbicides.

Knowledge gaps were found among farmers concerning herbicide application techniques, including selection of appropriate sprayers, nozzle tips, and sprayer calibration. In response to nearly every question posed, farmers indicated that they “don’t know” much of the technical information concerning chemical use and application. They could only provide the bottles and/or packages of the products that they used. It is important to build farmer knowledge on chemical and fertilizer use for economic, social, and environmental reasons, though it is equally important that companies use consistent terminology and clarity in their instructions (Hameed & Sawicka, 2017; Krasilnikov et al., 2022). There is an opportunity to raise awareness of integrated weed management (Krupnik et al., 2016; Kumar et al., 2017) that combines non-chemical and chemical options to prevent weeds from growing, reduce weed populations, and kill weeds effectively. This, as well as more strategic applications of fertilizer and other integrated approaches such as integrated pest management, have the potential to reduce chemical and fertilizer use in paddy production.

Lack of knowledge and understanding about sustainable strategies and practices for maintaining and improving soil health (i.e., Conservation Agriculture and Climate Smart Agriculture practices) is an additional major production constraint or factor preventing farmers from achieving high yield and profitability (Cárceles Rodríguez et al., 2022). Addressing this knowledge gap, Fisher et al. (2018) has pointed out the important roles of leading farmers and farmer networking as effective extension mechanisms for the adoption of new practices. For example, to enhance the adoption of these sustainable practices (i.e., reduced tillage, crop residue retention, growing cover crops, increasing vegetal biodiversity, etc.), farmers may require a series of training and promotional activities, including on-farm demonstrations, farmer training and consultations, and community networking to raise awareness and build capacity (Mgendi et al., 2021). We found that some farmers attended training sessions, but these did not necessarily lead to the adoption of new practices, indicating that extension efforts may be lacking or may need to be tailored to farmers’ needs while highlighting production benefits. Ashoori et al. (2017), Ghimire et al. (2015), and Touch and Yorn (2019) each suggest that introducing and developing appropriate farming machinery (e.g., direct seeders and crop residue management machines) may effectively facilitate adoption of sustainable agriculture practices and accelerate technology adoption.

This study found farmers relied heavily on agrochemical inputs for crop protection and yield increase. Ineffective and, sometimes, overuse of agrochemicals may lead to depletions of agrobiodiversity and natural resources, which can be harmful to the environment. This implies that their agricultural systems will become increasingly vulnerable to multiple factors (e.g., climate change, pest outbreaks, soil depletion). Because of the high risks involved in farming and low adaptive capacity, the existing production system tends to lock farmers into the predominant industrial model of agriculture and poverty, which may contribute to global food insecurity and lack of food safety.

5 Limitations of the study

While this study offers a comprehensive analysis of the constraints and limitations influencing farmers’ decisions and future priorities for increased sustainable agriculture production, we acknowledge two important limitations presented in the study.

The first important limitation is that this study adopted snowball sampling to identify study sites and purposive sampling to determine sample size. This suggests that not every farm household in the study areas had an equal chance of being selected. Consequently, there is a possibility of bias in the presentation of the research findings.

The second important limitation is that this study is based on the results of a cross-sectional sample survey, which means that data was collected from research participants at a single point in time. We believe a variety of factors can influence the responses of the research participants (e.g., who conducts the interview, who responds to the interview, the time of the interview, etc.). For example, a study carried out during a bad year (e.g., a long period of drought) may be subject to bias in reporting compared with a study undertaken during a good year. We, therefore, believe that gaining a comprehensive analysis of context-, constraint-, and measure-specific aspects remains to be addressed by future research.

6 Conclusions

This study demonstrates that improved crop production practices and better resource use efficiency can be a starting point for sustainable intensification of agriculture production in this region. This transition might become even more effective when further attention is devoted to the holistic understanding of local specific contexts, farm household characteristics, available resources, and farm management practices and constraints faced by farmers. We observed that farmers and their agricultural associations have yet to be fully recognised as partners and actors. They are generally viewed as beneficiaries and recipients of practices and technologies. We, therefore, recommend that farmers and their agricultural associations be acknowledged and included in a process of co-creation of knowledge-practices. These are the critical aspects that influence farmers’ adoption of innovative farming practices and technologies.

This study also indicates that farmers adopted a range of strategies and measures, especially diversifying income from multiple sources, in order to minimise and avoid perceived challenges and threats such as soil degradation, weeds, diseases, insect pests, high production costs, flooding, droughts, changing rainfall patterns, and unreliable rainfall distribution. Their responses have inadvertently led to an upward trend of temporary and permanent migration to cities and other countries, which results in farm labour shortages, soaring labour price, and a heavy reliance on agrochemicals. To address these challenges and threats, there is an urgent need to place research and development priorities on agricultural practices and technologies aimed at reducing the use of farm inputs and costs while increasing productivity. Future research and development leading to alternatives to agrochemical dependence in farming is highly needed.

While socioeconomic and environmental challenges and threats remain to have substantial impacts on smallholder farmers, increasing their capacity to adapt is of utmost importance. This allows farmers to avoid being (re)trapped into poverty, which enhancing the adaptive capacity of farmers contributes substantially to enhancing global food security and food safety.