1 Introduction

In Asia, rice farming is the single-largest use of land for producing food and rice is the staple food for nearly 4 billion people. The doubling of yields and the tripling of production between 1960 and 2018 was a revolutionary achievement that ensured the supply of rice, meeting the demand of the ever-growing population. The successes of this “green revolution” were widely ascribed to the development of the new semi-dwarf genotypes (Baldwin and Yan 2012). However, it was in fact a combined result of new genotypes with a high yield responsiveness to fertilizer inputs and investments in irrigation infrastructure (Becker and Angulo 2019), that permitted the rice double cropping system to emerge and to dominate irrigated areas.

However, with two (in some cases three) consecutive crops per year, larger amounts of nutrients are being removed from the soil. While fertilizers (organic amendments and/or mineral products) are generally recognized to be a prerequisite to compensate the increased removal of nutrients and to achieve regional food security, the amounts of nutrients applied by farmers in Southeast Asia appear to be unbalanced. Thus, the N:P:K application ratio of about 10:0.3:1 contrasts with a rice demand ratio of 10:1:5 (Syers et al. 2001; Zingore et al. 2022). One consequence of unbalanced fertilizer use is the depletion of P and K reserves in soils (Mutert and Fairhurst 2002). In addition, Zn deficiency is emerging as an increasingly important factor affecting yields in double-cropped systems and limiting rice productivity (Rehman et al. 2012). Most importantly, the imbalanced application of mineral fertilizers is both a fiscal and an environmental problem as high costs combined with low use efficiencies entail large monetary spending as well as pollution problems (Fageria and Baligar 2005). While demographic growth is slowing in most Asian countries, it will continue to impact global rice demand until at least 2035 (Zeigler and Barclay 2008). To meet the projected rise in food demand, an additional production of 76 million tons annually will be needed by 2035 in Asia alone. Particularly in Southeast Asia, the challenge of meeting increased rice demand and protecting environmental quality will be determined in rice-based cropping systems (Cassman et al. 2002), requiring an efficient management and balanced use of both mineral and organic fertilizers (Naher et al. 2019).

The speed and the type of farmers’ adaptive responses to changing on-farm resource equilibria and to policies or environmental pressures are key to sustainable development (Becker and Angulo 2019). Effective technical options can increase input use efficiencies and avoid a further exceeding of the planetary boundaries (O’Neill et al. 2018). Policies addressing these issues require a quantitative understanding of (a) current levels of nutrient uses and their efficiency and (b) benefits accrued from adoption of improved fertilizer and soil fertility management practices (Kishore et al. 2021). These practices and the required application rates and types of fertilizers depend on site attributes, expected yield levels, and farm management strategies (Peng et al. 2010). Thus, organic amendment and mineral fertilizer recommendations should be fine tuned and extrapolated to spatial domains with relatively uniform agro-ecological characteristics, cropping practices, and socioeconomic conditions (Dobermann et al. 2003). Key challenges in rice-based systems of Southeast Asia are thus to achieve regional food security and to increase farm incomes using site-specific integrated crop and nutrient management techniques (Tsujimoto et al. 2019), and to avoid nutrient imbalances in view of achieving a sustainable development (Ullah et al. 2021).

Particularly in the past 20 years, diverse production and land use strategies emerged in Southeast Asia (Becker and Angulo 2019). We assume that the manifestation of soil fertility-improving or crop nutritional practices and their adoption at farm-level depend on rice producers’ adaptive willingness and capacity to respond in a timely manner to changing demands (i.e., soil nutrient depletion, availability of mineral, or organic amendments), expectations (i.e., yield and income targets), and regulatory interventions (i.e., subsidies, technology innovations, extension) (Fig. 1). These will determine the future availability of rice, with implications for well-being at the farm and for food security at the regional scales (Yuan et al. 2021). We surmise that understanding and forecasting trends in fertility management, their determinants, and pathways of transformation can help avoiding undesirable developments and guide policy decisions for a sustained supply of rice. We believe these research areas are still underexploited.

Fig. 1
figure 1

Soil fertility management and nutrient supply in rice-based systems are constantly adapting to site-specific requirements and on-farm resource endowment. Rice straw burning in the Philippines (a) and mineral N application in lowland rice in Cambodia (b). (Photos: M. Becker).

Full size image

With this study, we fill this research gap by i) assessing the rice systems’ evolution, including recent trends and pathways of change during the past 20 years and associated soil fertility management practices; ii) evaluating production trends and processes of adaptation to evolving conditions of resource equilibria (synchrony or mismatch of production factor supply and demand); and iii) quantifying adoption of technological innovations and localizing transformation pathways in fertility management at farm and at regional levels.

To the best of our knowledge, until present only few empirical studies have assessed the dynamics of change in soil fertility and fertilizer management practices in rice over time. The present paper thus reports and discusses for the first time an analysis of change trends in fertility management practices (i.e., input-responsive genotypes, organic amendments, mineral N fertilizers, multiple N split application, other mineral fertilizers, and rice straw return) and their relation to or implications for grain yield over the last two decades (comparison of the years 2000 with 2018). We considered 1024 rice farmers’ households from six regionally representative sites of rice production in Myanmar, Cambodia, and the Philippines. We investigated the applicability of a diachronic approach (comparing recent past and present states) to document agronomic changes at farm level and to localize transformation pathways in the context of regional and global change using the model framework on drivers, pressures, states, impacts, and responses (DPSIR).

2 Material and methods

2.1 Study locations, climate, and edaphic attributes

We studied recent trends in cropping practices in major lowland rice-based systems at six sites in Southeast Asia. National experts in Cambodia, Myanmar, and the Philippines selected the study sites. They represent some of the major rice-growing environments of Southeast Asia, comprising fully irrigated alluvial plains, (partially) irrigated coastal delta areas, and rainfed production sites. The locations of individual households and rice fields at each site are shown in supplementary materials (Figure S1).

These rice-growing environments were differentiated based on the marginality or favorability of climatic and edaphic attributes. They differ in terms of their agro-ecological zones based on the lengths of growing period (LGP) for non-irrigated crops, as defined as the number of days when precipitation exceeds one-half of the potential evaporation plus the number of days needed to lose an assumed amount of 100 mm from the soil (https://gaez-services.fao.org/). The resulting crop-growing periods ranged from a LGP of 210 days in Meiktilia in the Central Dry Zone of Myanmar, over LGPs of 260–280 days at Mawkyun (Ayeyarwady Delta, Myanmar), Pangasinan (Philippines), and Takeo (Cambodia), to LGPs of > 300 days in Nueva Ecija (Philippines) and Svay Rieng (Mekong Delta Region, Cambodia).

In addition, soil physio-chemical properties differed between sites (Table 1). The soil in Nueva Ecija (Philippines) had the highest clay and the soil in Takeo (Cambodia) the highest sand contents. Soil pH (H2O, 1:2.5 ratio of soil to water) was highest in the alluvial plains (pH 6.6) and lowest in the coastal delta areas (pH 5.4–5.7). Soil organic C (C org) and N (Ntot) were highest at the favorable rainfed site in Pangasinan (Philippines) and lowest in Takeo (Cambodia). Related mainly to differences in soil texture, the cation exchange capacity was highest at the two sites in the Philippines (about 30 cmol/kg soil) and lowest at the marginal sites in Myanmar and Cambodia (15–17 cmol/kg). Further production constraints refer to widespread Zn deficiency in Svay Rieng (Cambodia), to moderate soil salinity during the dry season in Mawkyun (Myanmar), and to severe P deficiency in Takeo (Cambodia). Based on combined climatic and edaphic attributes, Nueva Ecija and Pangasinan (both Philippines) and Svay Rieng (Cambodia) were classified as “favorable”, while Takeo (Cambodia), Meiktila, and Mawkyun (both Myanmar) were classified as “marginal.”

Table 1 Selected soil attributes at six lowland rice-growing areas of Southeast Asia. Six composite samples per field plot were taken before soil tillage at the onset of the wet season at 0–15 cm soil depth in 2018. Samples were analyzed for physical-chemical properties at the laboratory of the University of Bonn. STDEV, standard deviation of mean; Ctot, total carbon; Ntot, total nitrogen; CEC, cation exchange capacity, BD, bulk density. Other additional non-listed production constraints based on own field observations and farmers’ reports refer to (a) heat stress during the dry season at Meiktila (Myanmar), (b) wide-spread Zn deficiency at Svay Rieng (Cambodia), (c) soil salinity during the dry season at Mawkyun (Myanmar), and (d) P deficiency at Takeo (Cambodia).
Full size table

2.2 Field survey, data collection, and soil sampling

From available lists of census-based households (sites in the Philippines and Cambodia) or based on recommendations of the national experts (Myanmar), we purposefully selected 150–200 farm households per site (1024 households) from an initial larger random sample. We excluded farmers having abandoned cultivation in 2018 and those not yet cultivating rice in 2000 as well as farmers unwilling to answer the questionnaire or those objecting to research assistants taking soil samples and collecting yield data from harvest areas. A diachronic analysis (baseline year 2000) defined prototypic change patterns in the rice-based production systems with a focus on soil fertility and fertilizer management. At each study site, well-trained field assistants assessed past (year 2000) and current/recent (year 2018/2019) fertility-related production practices, and, where applicable, in both dry and wet seasons, totaling 3264 complete observation datasets. We recognize that a diachronic survey cannot replace several waves of actual farmer surveys over time, relying on personal memories of respondents and their ability to situate past practices in a reliable reference period. However, for data-scarce regions (here all sites in Myanmar and Cambodia as well as the Pangasinan site in the Philippines), a similar approach was used in a study in Bangladesh (Aravindakshan et al. 2020; Emran et al. 2021), and such approach has been recommended for sites where past data are not available.

Direct sampling and measurements in the rice field plots of each household complemented interviews. Soil samples were taken by augers of 5 cm diameter at depths of 0–15 cm, using pooled composites of six samples per rice field plot. Bulk density was determined at the sites or the local institutions in standard 100 cm3 cans (5–10 cm soil depth) after oven drying. Thirty-gram sub-samples of each composite (1024 samples in total) were transferred to the Institute of Crop Science and Resource Conservation of the University of Bonn, Germany, and analyzed for texture, pH (H2O), total C and N contents (C-N Elemental Analyser EuroEA 3000 Series), and the cation exchange capacity using standard methods of analysis (Jones 2000). Deficiencies in soil P and Zn contents were not systematically measured but rather assessed by visual field symptom occurrence and subsequent chemical analysis of few selected soil samples from Svay Rieng (< 2 mg kg−1 HCl-extractable Zn) and Takeo (< 2 mg kg−1 available Olsen-P).

Exact field and plot sizes in each farm were determined by walking around each individual field plot with a GPS device. As nutrient application rates were usually provided in “bags per plot” and as the weight of fertilizer bags differed between countries, their weights were related to the measured sizes of the plots.

Data on past practices, such as fertilizer application rates, were qualitative estimates based on farmers indicating (1) same or similar amounts as in 2018, (2) more, (3) less, or (4) none (no fertilizer applied or no rice grown in 2000). Regarding past grain yields, only 30% (12–42% of the cases, depending on the site) had written yield records, the remaining ones indicated either same or lower yields in 2000 than in 2018. When no perceived or remembered changes in grain yield were indicated, the 2018 yield were applied to the respective fields/households for the year 2000. With farmers indicating “lower yields in the past,” we applied the percentage yield reduction (mean from hard records of each site) to the corresponding 2018 yield records, thus estimating the yield data for the year 2000. Due to the resulting uncertainty in past yield levels, we undertook diverse quality check measures by assessing the validity of farmers’ remembered and our calculated finding against independent yield records from 2000 at the different sites. In the case of the Nueva Ecija and Pangasinan sites in the Philippines, we were able to confirm the accuracy of the 2000 yield data against the Rice-Based Farm Household Survey—RBFHS, conducted by the PhilRice in 2000 (Eligio 2010) and the Central Luzon Loop Survey (Moya et al. 2017). For the other sites, we assessed available records of scientists, extension agents, or ministries in Cambodia and Myanmar. However, these data records were often restricted to the amount of rice sold (tax records) or the amounts of rice milled (miller records), and both disregard the amount of rice retained for home consumption. Accordingly, the reported values for 2000 were on average 5–10% lower than the data obtained from our calculations of rice yields for 2000. We can thus conclude that the diachronic survey permits a reliable and robust change detection for qualitative attributes and for fertilizer application rates. The approach appears less suited for detecting changes in actual yield and data shown here, while validated, may be only estimates with a 10% error. However, as we applied the same approach to all six sites, the systematic error of estimations was similar, allowing a comparison between sites.

We calculated the agronomic use efficiency of applied mineral N (AEn) as the yield increase above a non-amended control divided by the mineral N rate applied. No paired yield data from plots with or without mineral N were available, and the number of farms applying no mineral N varied between sites and periods. Yields from non-amended plots were available from 5 (Nueva Ecija) to 130 (Svay Rieng) rice fields in 2000 but only for 3–40 plots in 2018. Thus, only mean data of AEn (mean yield with mineral N – mean yield without mineral N/mean amount of mineral N applied) were calculated for each period, season, and site, and further statistical differentiations were non-realistic.

The response of grain yield to applied mineral N was differentiated by period (2000 vs. 2018), by season (dry vs. wet seasons), and by the application (or not) of fertilizer nutrients other than N. For calculating yield responses, we applied the “boundary concept” (van Ittersum et al. 2013), using the highest grain yield obtained for each individual N application rate. The response model was a quadratic plateau function (Sripada et al. 2008), using the nls functions in R-packages Ime4 and nlme as suggested by Nyiraneza et al. (2021).

2.3 Data and statistical analysis

For quantitative data on yields and mineral N rates, we removed some extreme values, assuming them non-realistic. We considered such outliers as data points outside the overall distribution pattern, identified by values being the 1.5 times the interquartile range (Vinutha et al. 2018). Thus, out of 3454 data points 28 yield data (< 1%) and 39 N application rate data (about 1%) were removed. We assessed “cleaned” data by central tendency parameters such as means and standard deviations as well as to pairwise comparisons of data vectors by paired t-tests. Subsequent change detection was done by contrast analysis based on the number of cases (or the percentage share of those) showing absolute differences (present-absence data or yes-no questions) or showing significant differences in quantitative attribute means (fertilizer rate, number of N splits, grain yields) when comparing past (year 2000) with present/recent (year 2018). The statistical analysis was performed using SPSS v27 (IBM SPSS Inc., USA).

2.4 DPSIR framework model for qualitative cause-effect relationship

DPSIR (drivers, pressures, state, impacts, and responses) is a causal framework or model of intervention for describing the interactions between society and the environment (Zhou et al. 2013). The European Environment Agency (EEA) and the United Nations Environment Program (UNEP) have advocated it since its first publication (Smeets and Weterings 1999) for assessing the environmental impacts of human activities. The DPSIR model is considered suited to describe qualitatively cause-effect relationships in past, most recently, as well as future developments (Miloslavich et al. 2018). We implemented DPSIR to elucidate causal pressure-state-responses, and to visualize the impacts of system-immanent drivers and external pressures on transitions in fertilizer use and fertility management practices. Thus, the drivers of change and the resulting states and responses concerned mainly farm-internal or system-immanent attributes, while pressures and impacts refer to the regional or national scales, involving policy interventions or other larger-scale pressure attributes (i.e., subsidies and rice and fertilizer pricing). Based on quantitative response attributes (percentage changes in key agronomic practices between 2000 and 2018) and on prevailing on-farm resource endowment, we determined likely changes in states and impacts at the farm and regional scales to derive required policy interventions. These are visualized in five sequential response loops.

3 Results and discussion

3.1 Fertility-related cropping practices

During the past two decades (difference between two discrete time steps, namely, 2000 vs. 2018), the number and the type of cropping practices applied in rice production have changed across sites. Among these, twelve common practices showed significant trends for individual sites (Table 2). Generally, the cropping intensity has increased at all sites, but in Nueva Ecija (Philippines) and in Takeo (Cambodia) mainly by the extension of cultivation into the dry season (double cropping). In rainfed areas and generally at water-strapped sites, farmers tend to opt for intensification of the non-rice component in the system (Table 2). Thus, farm income might be increasingly generated by cultivating, i.e., mungbean in the pre-rice (Cambodia—Ro et al. 2016), tobacco, watermelon, and maize in the post-rice niches (Philippines—Haefele et al. 2013), or oilseeds (sesame and sunflower) in the Central Dry Zone (Myanmar—Herridge et al. 2019). This trend was also observed in other regions of Asia where perishable high-value vegetables are increasingly cultivated in peri-urban zones with access to urban markets (Pokharel 2019; Shrestha et al. 2021). Similarly, at most sites, intensification was associated with a shift toward tractor tillage (in contrast to manual or animal tillage) and the use of modern (input-responsive) rice genotypes (1280 out of 3264 observed cases across sites). The different dimensions concerning the management of soil fertility and nutrient application strategies comprise (a) farmyard manure, (b) rice straw return, (c) mineral N fertilizer rate, (d) number of N split applications, and (e) use of non-N mineral fertilizers (P, K, Zn) (Table 2).

Table 2 Changes in intensification-related production practices during the past two decades (2000 vs. 2018) in six representative systems of lowland rice production in Southeast Asia. a Double rice in irrigated systems and rice-upland crop rotations in rainfed environment or sites with poor irrigation water control; b improved high-yield genotypes or hybrid rice; c dry or wet direct seeding, mainly in the dry season; d application of fungicides and/or insecticides; e basal application of PK fertilizers and/or Zn; f mainly axial flow threshers; g incorporation of rice straw into the soil (as against use for sale, as animal bedding, or straw burning); h average mean of the share of farmers adopting each of the 12 individual intensification practices; i average mean of the share of farmers adopting soil fertility/fertilizer management practices.
Full size table

3.1.1 Organic amendments (farmyard manure and rice straw)

Most farmers at the three marginal sites applied small to moderate amounts of mainly farmyard manure in the past, and except for the site at Mawkyun (Myanmar), still use it today. On the other hand, we observed a significant decline from 2000 in the use of farmyard manure at Svay Rieng (Cambodia) and Meiktila (Myanmar), with declines from about 47% of all farmers to 2 and 18%, respectively, applying manures. According to farmers’ feedback in the open section of the questionnaire, this shift appears to be linked to a declining number of farm animals and to labor limitations for manure application (all sites) and to the sale of manure to intermediaries from neighboring Vietnam (Svay Rieng). In Takeo (Cambodia) farmers own 3–7 livestock units and thus all farmers used and still use farmyard manure, though at highly variable rates of application (data not shown). On the other hand, at the two sites in the Philippines, less than 13% of the farmers apply organic amendments.

The return of rice straw and its incorporation into the soil can return substantial amounts of carbon, potassium, and silica. However, the practice is restricted to Mawkyun (Myanmar) with 56% and to Nueva Ecija (Philippines) with 78% of the farmers returning rice straw. This is largely associated with the use of combine harvesters (straw remaining in situ) and tractor tillage (straw incorporation). Otherwise, straw return is low to non-existent at most other sites. In Pangasinan, most farmers burn rice straw, while in Meiktila (Myanmar) and at the two sites in Cambodia, most farmers remove the rice straw for use as animal bedding or as feed stock (Meiktila) or for sale to mushroom growers and commercial large-scale cattle farms in Svay Rieng (Cambodia). Thus, the application of farmyard manure is restricted to the sites with high cattle numbers and low soil fertility (Meiktila and Takeo), while straw return is largely limited to the highly mechanized production systems (Nueva Ecija, Philippines).

3.1.2 Mineral fertilizers

The vast majority of farmers apply mineral N fertilizers (mainly urea and some di-ammonium-phosphate in the Philippines), with substantial recent increases in the share of urea users from 16 to 59% in Mawkyun and from 59 to 96% in Meiktila (both Myanmar), and from 48 to 85% in Svay Rieng (Cambodia). At the two sites in the Philippines, most farmers applied mineral N already in the past. The rates of mineral N applied differed strongly between sites, but also between farmers at each site, ranging from a mean value of 37 kg N/ha (range: 0–95) in Takeo (Cambodia) to > 100 kg N/ha (range: 35–215) in Nueva Ecija and Pangasinan (Philippines). While most farmers at the three marginal sites did apply only one single dose of N in 2000, the share of farmers topdressing N (2–3 split applications) has recently increased at all sites. Finally, the (mostly basal) application of non-N mineral fertilizers (P, K, and Zn) has also increased between 2000 and 2018 at all sites, with a low of 23% of all farmers in Mawkyun (Myanmar) and a maximum of 100% of all farmers applying PKZn-containing products in Pangasinan (Philippines) in 2018. However, the application rates tended to remain low and are actually below the rates of crop removal, with 7–23 kg P/ha, 12–35 kg K/ha, and 2–4 kg Zn/ha. In summary, N dominates the use of mineral fertilizers, it is widespread, and it has been recently increasing. Supplementary applications of P, K, and Zn are gaining importance at all sites.

3.2 Aggregated changes in fertility management

Considering all fertility management practices together in an aggregated analysis of change, we recognize that fertility management started at very different initial levels in 2000 and has evolved to different extents by 2018 (Fig. 2). While changes in adopting agronomic practices occurred at different speeds and intensities at the six study sites, the general trend appears rather homogenous with more mineral fertilizer use and the adoption of practices that increase the use efficiency, particularly of applied mineral N. While in Nueva Ecija in 2000 a majority of farmers already applied a set of diverse fertility management measures (aggregated share of 60%), this further increased to the level of 75% in 2018. At the other extreme, we see that Mawkyun started in 2000 at an aggregate level of fertility management of only 28% and increased the adoption of improved fertility management measures by only 5 percentage points to 33%. Largest gains in fertility management between 2000 and 2018 were observed at Meiktila (Myanmar) with an aggregate increase from 40 to 58% and in Pangasinan (Philippines) with an increase from 45 to 68%. Privileged determinants for improved fertility management were increased shares of farmers topdressing mineral N (multiple N splits) and an increased use of non-N mineral fertilizers. The two Cambodian sites take intermediate positions with mean increases by 15 percentage points. Changes in fertility management, however, occurred at much higher rates at favorable (Svay Rieng) than at marginal sites (Takeo), but was in both cases associated with higher N application rates and more farmers applying non-N mineral fertilizers.

Fig. 2
figure 2

Aggregate changes in practices of soil fertility and fertilizer management from “traditional” to “intensive” lowland rice at the six study sites in Southeast Asia.

Full size image

3.3 Perceived key production constraints

Farmers’ perceived key constraints within the agronomic portfolio of applied practices (referring here to single-most expensive cropping practices—Table 3), both in the past and at present, differed between sites and years, however, with a rather consistent trend towards mineral N fertilizers being considered the most or at least second-most expensive practice in lowland rice production in 2018. The trends in perceived constraints largely reflect the evolutionary state of the production system or of aggregated fertility management practices (Fig. 2). Thus, mineral N fertilizer use was already in 2000 the single-most expensive rice production practices in Nueva Ecija (Philippines), while at most other sites the labor costs for rice transplanting and harvesting/threshing dominated the list of main constraints. With the adoption of laborsaving production strategies (tractor tillage at all sites, crop establishment by direct seeding instead of transplanting in Svay Rieng, Cambodia, and Pangasinan, Philippines, and the adoption of combine harvesters at all sites), the high price of agrochemicals moved to the forefront of key constraints, irrespective of the site. Mineral N has been replaced only recently by the cost for pest and disease control as most expensive crop management practice in Svay Rieng (Cambodia), a trend that may well be reversed with the expected rising fertilizer prices. We thus analyzed the data set with special emphasis on changes regarding soil N fertility management, particularly applied N rates, rice responsiveness, and use efficiencies of mineral N.

Table 3 Changes in the farmer-perceived single-most and second-most expensive cropping practices between 2000 and 2018 in six representative lowland rice-based production systems of Southeast Asia.
Full size table

3.4 Mineral N management and responses

The shares of farmers applying mineral N, the application rates, yield responses to applied mineral N, and the agronomic N use efficiency (AEn) differed between production environments (sites) and observation period (past vs. present), and are summarized in Table 4. Apart from the two sites in Myanmar, all farmers apply mineral N today, compared to a share of 60% N users in 2000. Mean N application rates were lowest at the marginal sites with on average 58 kg N/ha and highest at the favorable sites with on average 95 kg/ha. The N rates have slightly increased from 59% in 2000 (n = 1042) to 79% in 2018 (n = 1558) and tend to be higher in the dry than in the wet season. The rice grain yields (here only data from field plots having received mineral N) were again lowest at the marginal sites (average yield of 3.3 t/ha) and highest at the favorable sites (average of 5.4 t/ha). They have increased from an overall mean of 3.9 t/ha in 2000 to 4.5 t/ha in 2018 and they were higher in the dry (5.2 t/ha) than in the wet season (3.9 t/ha). Based on the mean yield of plots having received none or only low rates of < 30 kg/ha of mineral N (highly variable in numbers, ranging from 84 cases applying no N in Mawkyun in 2000 to only 3 cases applying < 30 kg N/ha in Nueva Ecija in 2018), we calculated the mean agronomic efficiencies of applied mineral N for each site and period. In general, AEn was slightly higher in 2018 with 17 than in 2000 with 15 kg/kg N and tended to be higher in the dry (18) than in the wet season (17 kg/kg N).

Table 4 Changes in mineral fertilizer N application rates, grain yields, and agronomic nitrogen use efficiency during the past two decade (years 2000 vs. 2018) in six representative systems of rice production of Southeast Asia. Ns, not significant; *, **, *** significantly different at p = 0.05, 0.01, 0.005. NA, not applicable. a agronomic N use efficiency = mean grain yield increase above mean yield of non-amended rice (yield at 0–30 kg ha−1) divided by kg (additional) mineral N applied; WS, wet season; DS, dry season. b N use efficiency data for 2018 have to be taken with care as calculations are in some instances based on a very small sample (n = 3 for the two sites in the Philippines in 2018, < 40 for the other sites).
Full size table

The grain yield responses to applied mineral N were first plotted individually per site and season, and further differentiated into past vs. present and dry vs. wet seasons. In no single case did a linear regression provide significant trends, suggesting a weak or non-existent relationship between N application rate and rice yield (data not shown). However, such analytical responses changed when plotting all data (6 sites, 2 seasons, past and present, n = 3454 data pairs) and applying a boundary (quadratic plateau) function (Fig. 3a).

Fig. 3
figure 3

Rice yield response to applied mineral nitrogen at three marginal and three favorable sites (both past and present as well as dry and wet season data; n = 3206) (a) and rice yield response to applied mineral N fertilizer differentiated by the additional use (or not) of P, K, and/or Zn fertilizers (b) in lowland rice-based systems of Southeast Asia.

Full size image

Data suggest a positive response to applied mineral N up to 150 kg/ha with an N0 intercept at 3.7 and a maximum yield plateau at around 12.0 t/ha for the favorable sites. At the marginal sites, yields responded up to 95 kg N/ha with an N0 intercept at 2.9 and a plateau at around 6 t/ha.

An interesting additional relationship in rice yield response became apparent when differentiating N-induced yield responses by the application of other mineral fertilizers. These non-N fertilizers were not identical across sites and concerned primarily P in Takeo (Cambodia), Zn in Svay Rieng (Cambodia), and PKZn at the sites in the Philippines. The responsiveness of rice to applied mineral N improved when applying other mineral fertilizers in addition to N (Fig. 3b). Thus, mean yields increased by about 0.4 t/ha from sole N to additional NPKZn (data only for 2018). The boundary function (quadratic plateau) reflected this higher base yield an N0 (intercept at 2.4 t/ha without and 2.8 t/ha with PKZn) and with plateau values of 10.4 without and 12.1 t/ha with supplemental application of other mineral fertilizers.

While PKZn application was associated with an improved grain yield response and a higher agronomic N use efficiency – AEn (data not shown), it cannot be excluded that these effects were associated with a generally higher technicity level (high share of farmers applying all six fertility-related management practices) of those farmers willing to apply other mineral nutrients, i.e., soil leveling or improved weed management.

Given that both N application rates and grain yields have increased in the past decades, and that N rates and yields tended to be higher in the dry than in the wet seasons, we further differentiated the yield response to applied mineral N accordingly. A differentiation by decade (Fig. 4a) and by season (Fig. 4b) highlights an upward shift of the centroids (median yield response to median N application rates) in both cases. Thus, both the amount of N applied and the rice grain yield have increased from a centroid of 59 kg N/ha and 4.4 t/ha in 2000 (1695 observations) to 86 kg N/ha and 5.6 t/ha in 2018 (1712 observations). In addition, between seasons, we observe a change of the centroid from 58 kg N/ha at 4.2 t/ha in the wet season (2064 observations) to 87 kg N/ha and 5.6 t/ha in the dry season (1392 observations).

Fig. 4
figure 4

Rice yield response to applied mineral nitrogen in the past (2000) and present (2018) (a) and in the dry and wet season (b).

Full size image

3.5 System transitions in fertility management and their implications

Our study showed an increasing number of fertility management practices being applied, though to variable extents and at different speeds between sites (Fig. 2). In Southeast Asia, such intensification was associated with concomitant increases in regional rice production from 426 million tons of dehulled rice annually in 2000 to nearly 500 million tons in 2018 (FAO-STAT 2020). Despite this encouraging upward trend, the annual rate of yield increase reduced from 2.7% in the 1980s to 1.1% in the1990s and < 1% in the 2000s. With the forecast continued increase in world population, current rates of yield and production increases will not suffice, requiring resumption of the previous rate of > 2% annual increase (Horie et al. 2005). Such productivity gains under a sustainable food production scenario have been projected to be possible by improved soil fertility and nutrient supplying strategies, and by respecting four interlinked relevant planetary boundaries (N and P nutrition, biodiversity, and trace gas emissions/climate change; Gerten et al. 2020). Given the still large differences between experimental and farmers’ actual yields (Mueller et al. 2012), this expected increase could be met by adopting technology innovations, foremost site-specific nutrient management strategies. Achieving synchrony between nutrient supply and crop demand without excess or deficiency is seen to be the key to optimizing trade-offs among yield, profit, and environmental protection (Cassman et al. 2002). Narrowing the large prevailing yield gaps on currently available agricultural land requires a site-specific dissemination of integrated crop and nutrient management practices, giving priority to strategies improving soil properties and organic as well as mineral fertilizer uses (Dossou-Yovo et al. 2020).

3.5.1 Organic amendments

Organic amendments have been the dominant strategy for maintaining or improving soil fertility and for adding nutrients until the advent of the “green revolution” (King 2004). Most farmers at the marginal sites in the present study did apply organic amendments in the past, and many still use them today. However, even at these marginal sites, only few farmers continue to use organic amendments as a stand-alone practice of soil fertility management, most opting for integrating mineral N and some P in combination with farmyard manures. On the other hand, the use of organic amendments in the form of green or farmyard manures has nearly disappeared at the sites with more fertile soils, and particularly, with severe labor limitations for manure application. Thus, among 1024 households across the six study sites, just 203 apply manures as source of nutrients or for soil fertility improvement today. This trend of declining use of green and farmyard manures has been observed already in the 1990s (Becker et al. 1995) and appears to continue with progressing intensification of rice production systems (Selvi and Kalpana 2009) and despite widely recognized benefits on rice soil fertility (Yang et al. 2019) and soil health (Kumar et al. 2020).

Rice straw makes up about half of the total biomass in rice production and contains nearly 70% of the total K and > 60% of the Si accumulated in the biomass of rice, thus largely meeting the K and Si demand of rice upon straw return to the field (Timsina et al. 2013). Despite a low lignin content of only 5–8% (Becker et al. 1994), rice straw is a potentially valuable source of carbon that can possibly maintain or build soil organic matter with positive effects on soil physical attributes (Paul et al. 2021). However, the return of rice straw in submerged soil systems is a double-edged sword, and reports on benefits and trade-offs of straw return are contradictory. Thus, with its low N content and C:N ratios of 40–60, straw incorporation can induce N deficiency by temporary microbial immobilization (Becker et al. 2007), requiring larger amounts of added mineral N. On the other hand, the nitrogen-fixing community in soils, a key functional player to replenish N pools from atmospheric N2, can radically change with the addition of organic substrate, with rice straw amendments reportedly increasing the number of diazotrophic bacteria in rice soils (Yang et al. 2019). In continuously flooded rice fields, emissions of the climate-relevant trace gas methane can increase with straw incorporation by up to 90% (Tan et al. 2018), an effect that has been linked to an increase in the labile (permanganate-oxidizable) C fraction in soils, which reportedly drives methane emissions (Bhattacharyya et al. 2012). Such reports, however, have been contradicted by Zhang et al (2013) who reported a reduced global warming potential with straw incorporation compared to open-field burning, while improving soil fertility.

In the present study, a majority of farmers considered rice straw a waste product, hampering tillage operations and delaying the establishment of the second rice crop. The resulting continuous removal of rice straw and other crop residues has been associated with depleting soil organic matter and soil nutrient reserves, for example, in the Central Dry Zone of Myanmar (Herridge et al. 2019). At all other study sites, rice straw was mainly disposed of by burning, either in situ or next to the threshing floor. Burning of rice straw has been widely judged as an environmental concern, contributing to global CO2 and black carbon emissions, and eventually toward climate change (Singh et al. 2021). Consequently, many countries have recently banned straw burning (i.e., India and China), or at least advised farmers to refrain from straw burning (i.e., the Philippines and Cambodia). The ban is however rarely enforced, and in 2018, > 20% of all farmers in the sample still practiced open-field burning of rice straw.

In response to some of the reported negative effects of direct straw incorporation, some authors suggested the composting of straw (Jusoh et al. 2013), or alternative economic uses for power supply (Logeswaran et al. 2020) or the sale of straw as substrate for various activities (Herridge et al. 2019). A recent global meta-analysis using life cycle assessment studies concluded that soil incorporation and electricity generation are among the most environment-friendly alternatives to rice straw burning (Singh et al. 2021), though rarely applied in the observed sample of farmers. A game changer in recent rice straw use has been the availability and wide use of combine harvesters at most sites, leaving most of the straw evenly scattered in the paddy field. Thus, apart from straw collection by bale pressing and subsequent straw sales in Svay Rieng (Cambodia), rice straw remains in situ and is incorporated during tillage operations in 35% of farms in Mawkyun (Myanmar) and in nearly 80% of farms in Nueva Ecija (Philippines). Hence, the combined effect of labor shortage and technology innovations has increased the application of rice straw at favorable sites with intense double cropping, while the recognition of a need for soil fertility improvement has possibly contributed to a continued application of farmyard manure at the marginal sites.

3.5.2 Mineral fertilizers

While much recent research aimed at increasing the yield potential of rice, such approaches are no stand-alone option as long as current yield potentials are far from being achieved (Neumann et al. 2010). Thus, development research focuses on attempting to close the large yield gap by fostering the site- and system-specific adoption of available soil fertility-improving and nitrogen-supplying technologies in view of achieving productivity gains and sustainable intensification (Yuan et al. 2021). The dynamic nature of N and its propensity for loss from soil‐plant systems creates a unique and challenging environment for its efficient management. Recovery of N in rice is usually less than 50% and is associated with its loss by volatilization, leaching, surface runoff, and denitrification. Low recovery of N is not only responsible for higher costs of production but also for environmental pollution. At present, mineral N fertilizer is not only the single-most expensive practices in most sites (Table 3), mineral N synthesis is also a highly energy-consuming process (Smith et al. 2019). Hence, improving N use efficiency (AEN) is desirable to improve crop yields, reducing cost of production and maintaining environmental quality (Fageria and Baligar 2005).

Instruments to achieve gains in productivity comprised the multiple split application of mineral N fertilizers (Sui et al. 2013), but also sulfur coating (Yasmin et al. 2007) and N placement strategies (Linquist et al. 2013), aiming to achieve synchrony of N supply with crop demand (Becker and Ladha 1997). Thus, the agronomic efficiency of N increased from about 12 kg kg−1 for a single dose of N at transplanting in the 1990s, to 15 (rainfed rice) and 20 (irrigated rice) in the 2000s (Peng et al. 2010), and up to 40 kg kg−1 for banded controlled release fertilizer in modern irrigated systems using hybrid rice varieties (Djaman et al. 2018). This is also reflected in the present study, where more modest increases in AEN during the past two decades were associated with an increase from 43 to 78% of farmers applying multiple N splits. Further gains in increasing AEN and the economic benefits from N application concern (1) the balancing of deficiencies in nutrients other than N, (2) improved crop management, and (3) the development and use of resource efficient rice genotypes (Hirel et al. 2011).

3.6 Drivers and pathways of change

As the intensification of rice-based systems involves changes in several cultivation steps over time (Struik et al. 2014), an assessment not only of such changes and of their drivers but also of the effects at both farm and regional scales is needed to understand and develop sustainable food systems (Smith et al. 2017). The present study showed that during the past two decades, several fertility-related practices have evolved from rather labor-intensive medium-input to highly mechanized (laborsaving) high-input production systems. These transitions occurred and still do occur at different speeds and intensities in different production systems, but appear to follow a rather consistent evolutionary trend across regions. Among agronomic practices that have recently changed, several concern directly or indirectly soil fertility management and nutrient supply. These refer to the use and management of organic amendments, the amount and the types of mineral fertilizers, as well as to their application management. All these strategies, their adoption, and their effects on yield and regional production are closely linked to (1) system-immanent attributes such as the resource endowment of individual farms and to changing expectations and aspirations, to (2) larger-scale external pressures such as the policy environments and global trends, and (3) to the availability and applicability of technical innovations.

We used the DPSIR model to assess and visualize changes in soil fertility- and fertilizer management-related cropping practices regarding effects of system-immanent drivers and external pressure forces on attributes on the farm and on wider implications at the regional/country scales, conceptualizing five consecutive cause-response loops (Fig. 5), starting from farm-scale drivers, regional-scale pressures, over farm-scale states and regional scale impacts to responses.

Fig. 5
figure 5

Applying the DPSIR model frame (drivers, pressures, states, impacts, and responses) to changing fertilizer use and soil fertility practices in rice-based systems of Southeast Asia.

Full size image

Loop 1 refers to an increasing need for rice (food and income demand at farm level), which under the pressure of land scarcity (small/declining farm sizes) increased the land use intensity (double cropping) with implications on soil nutrient depletion and the need for more application of amendments. The model explains the observed shift toward rice double cropping in the fully irrigated environments (31% of all irrigated cases) and the diversification by growing (high-value) upland crops in rotation with rice in rainfed and water-scarce environments (nearly 50% of all rainfed cases).

Loop 2 starts from the required increase in rice yields that, in combination with the availability of N fertilizers, leads to high N input systems with concomitant yield increases (nearly 90% of all irrigated cases in favorable environments).

Loop 3 depicts the response to perceived soil fertility declines in combination with low rice prices at the farm gate, resulting in higher spending for agrochemicals and particularly of increased mineral N application rates. Thus, the number of N users has increased from 60 to nearly 100% of all farmers and the applied rates have increased from a mean of 54 kg N/ha in 2000 to a mean of 86 kg/ha in 2018.

Loop 4 describes farmers’ responses to a rather low use efficiency, particularly of mineral N (AEn of about 12 kg/kg in 2000), to high and raising fertilizer prices on the world market (increase from 612 in 2000 to 716 €/t of urea-N in 2018), and to the removal of input subsidies in some countries. These attribute changes fostered the adoption of input-efficient production strategies such as the multiple split application of mineral N (21% of all observed case), the supplemental application of non-N mineral fertilizers (12% of all observed case), and the use of input-responsive genotypes, mainly hybrid seeds (28% of all observed cases).

Loop 5 depicts the emerging labor scarcity (declining household sizes, increased costs for hiring external labor forces) was associated with the wide adoption of labor-saving production strategies, involving to direct seeding replacing the transplanting of rice seedlings, but mainly referring to mechanization with the adoption of combine harvesters (straw remaining in situ) and tractor tillage (straw incorporation). Thus, substantial amounts not only of organic C but also of K and Si contained in rice straw are restituted to the soil, potentially contributing to improving soil fertility, while likely increasing methane emissions. These latter changes were at least in some instances stimulated by policies imposing minimum wages, thus further increasing the costs of hired labor.

In summary, rice farmers in Southeast Asia have responded to changing system-immanent drivers and to external pressures by gradually adopting land use- and input-intensive production strategies, addressing nutrient removal and declining fertility as well as the need to boost on-farm production and yield. While the pathways of change appear rather homogenous, the extent and intensity of such changes, however, differed between sites. Policy interventions accelerating the transition toward sustainable intensification may involve subsidies for mineral fertilizers, an improved availability of affordable technical innovations (farm machinery, (micro)nutrient fertilizers, and input-efficient genotypes), and research on production strategies contributing to curb trace gas emissions such as temporary soil drying and alternative straw management (i.e., energy, biochar; Singh et al. 2021). A balanced nutrient supply needs to consider spatial variations in soil-related factors such as P, K, Zn, and Si deficiencies (Tsujimoto et al. 2019). Thus, recent increases in the use of non-N mineral fertilizers from 42 to 65% of all farmers across the study sites were associated with higher yields and increased N use efficiencies (Table 4).

In addition to the mineral fertilizer, the use of input-responsive seeds (here mainly hybrid rice) can increase the production potential, maximizing output per unit area or per hour of labor investment (Ju et al. 2015). Thus, compared to “classical” high-yielding genotypes, hybrid seeds reportedly increase the yield potential and the crop’s input response under favorable conditions (Yan et al. 2008), and this improved yield potential is realized particularly during the dry season (Yuan et al. 2021). This observation is in line with the recent increase in the number of farmers using hybrid seeds at those sites where such seeds are available (favorable sites in the Philippines and Cambodia) with 30 and up to 95% of farmers using hybrid seeds in the wet and dry seasons, respectively. Additional strategies to increase nutrient uptake and use efficiencies are under research and concern foremost the development of genotypes possessing a generally improved root architecture for nutrient acquisition (Gonzalez et al. 2021). Such strategies aim primarily at increasing the use efficiency of the single-most expensive production input, mineral N fertilizer. Mineral N fertilizer prices on the world market have recently increased from 716 in 2018 to nearly 1000 €/t of urea N in January 2022 (ChemAnalyst 2022—https://www.chemanalyst.com). Expected further price increases, associated with current and expected future crude oil and natural gas price spikes, will increase the importance of the costs of N applications in the overall cost layout of rice farms and is likely to limit or reduce N application rates in the foreseeable future. Without input subsidies balancing the recent dramatic increases in input prices, current forms of conventional rice farming may no longer be profitable, with national food sovereignty declining, and with rice imports and thus the spending of scarce foreign exchange increasing. However, with subsidies, it is questionable if mineral N alone can achieve the yield increases that they showed in the 1980s. In addition, producer prices must increase to enhance the profitability and future rice production and to support sustainable developments in the rice sector (FAO 2020).

4 Conclusion

We analyzed fertility-related agronomic practices in rice-based systems at six representative sites in Southeast Asia in view of identifying recent trends, the driving forces for change, and their effectiveness to harness them for regional food security and sustainable development. Our findings highlight that rice-based production systems throughout Southeast Asia have followed comparable trends in soil fertility and fertilizer management. The timing and the extent of changes in agronomic practices applied are site specific and are linked to a combination of the favorability or marginality of the site, with on-farm production factor availability (land, labor, capital, knowhow), and farmers’ capability to adopt technological innovations. These lead over time to a convergence of system attributes and a homogenization of practices toward high-input and mechanized production systems, at least in fully irrigated environments. Irrigated rice double cropping is still considered the prime system to meet the growing future demand for rice, and it is the target of recent technological innovations. On the other hand, rainfed production covers an area of 18 million ha in Southeast Asia alone and is the dominant rice production system in Myanmar (78%) and Cambodia (85% of the rice-growing area). In rainfed areas and generally at water-strapped sites, farmers tend to opt for intensification of the non-rice component in the system. Such rotation crops grow under aerobic soil conditions and differ substantially from rice in terms of crop management and the nutrient demand. However, neither crop-specifically adapted fertilizer formulations nor the required recommendations are available to date.

Beyond the conclusions that are supported by our reported findings, additional issues of concern refer to global natural gas price spikes and recent political developments. These issues will affect the availability of inputs and increase the price of most farm inputs, while national pricing policies (minimum wage and import tariffs) will regionally or nationally differentiate the economics of rice production. There is a need to further increase the agronomic efficiency of mineral fertilizer use in rice while making N more affordable. New site- and crop-specifically adapted fertilizer formulations and the breeding of crops adapted to lower mineral N inputs are needed. This will require much greater investments in research and extension over the coming decades. Fertilizer sector reforms need to be fiscally sustainable and politically feasible. Rationalizing subsidies, while necessary, may not suffice to ensure balanced use of fertilizers for the required yield increases. Changing farmers’ practice requires combining the right incentives with the right information to farmers but also to deciders and policy makers. Finally, higher rice prices for producers will increase the profitability of rice production, thus sustaining and increasing future rice supply.