Introduction
Traditional upland rice crops are grown in unflooded fields which are not bunded, where soil conditions remain aerobic through most of the growing season (Atlin et al., Reference Atlin, Lafitte, Tao, Laza, Amante and Courtois2006). Soils are generally acidic and infertile in southeast Asia, and farmers treat upland rice as a subsistence crop, investing little in inputs beyond family labour (Saito et al., Reference Saito, Linquist, Atlin, Phanthaboon, Shiraiwa and Horie2006a). Traditional upland rice is tall, with limited tillering and prone to lodging when grown under favourable conditions of soil water and soil fertility (Asai et al., Reference Asai, Saito, Samson, Songyikhangsuthor, Homma, Shiraiwa, Kiyono, Inoue and Horie2009). Yields of upland rice are thus commonly only 1-2 t ha−1 in most regions (Atlin et al., Reference Atlin, Lafitte, Tao, Laza, Amante and Courtois2006).
Upland rice is typically grown by subsistence farmers under shifting cultivation systems in Laos, the most mountainous country in southeast Asia (Linquist et al., Reference Linquist, Trosh, Pandey, Phouynyavong and Guenat2007). Such systems are considered sustainable when low population pressure makes it possible to have long fallows between crops (Nye and Greenland, Reference Nye and Greenland1960). Under these conditions, upland rice has traditionally been grown without fertiliser under rainfed conditions (Roder, Reference Roder2001), with traditional large-grained glutinous tropical japonica genotypes being preferred by local farmers, which yield well under these initially fertile soil conditions. These genotypes are generally tall, with few tillers and low harvest index, but are suited to long fallow slash-and-burn systems (Saito et al., Reference Saito, Linquist, Atlin, Phanthaboon, Shiraiwa and Horie2006a). When grown in fields after long fallow, their yields reached 3-4 t ha−1 (Asai and Soisouvanh, Reference Asai and Soisouvanh2017), but they performed poorly when grown in fields following continuous upland rice cropping, due to declining soil fertility and increasing weed pressure (Sengxua et al., Reference Sengxua, Toomsan, Linquist, Lefroy and Limpinuntana2007).
The Lao-IRRI (International Rice Research Institute) project has screened over 3000 traditional accessions and identified several traditional indica genotypes that yielded 0.5–1.4 t ha−1 more, on average, than traditional tropical japonica checks under intensive continuous rice cropping systems (Sengxua et al., Reference Sengxua, Toomsan, Linquist, Lefroy and Limpinuntana2007). In addition, two improved indica ‘aerobic rice’ genotypes were identified that were high yielding under both low- and high-fertility conditions (Saito et al., Reference Saito, Linquist, Atlin, Phanthaboon, Shiraiwa and Horie2006a; Asai et al., Reference Asai, Saito, Samson, Songyikhangsuthor, Homma, Shiraiwa, Kiyono, Inoue and Horie2009). Since genotype-by-environment (GxE) interaction for grain yield is often large relative to the effect of genotype in low-yielding rainfed environments (Cooper, Reference Cooper1999; Samson et al., Reference Samson, Voradeth, Zhang, Tao, Xayavong, Khammone, Douangboupha, Sihathep, Sengxua, Phimphachanhvongsod, Bouahom, Jackson, Harnpichitvitaya, Hu and Wade2018), quantitative data are needed to rigorously compare traditional and improved indica and traditional tropical japonica genotypes for identifying optimal genotypes for high yield. Previous studies, however, were conducted in northern Laos, at higher altitude, with lower temperature and lower evaporation than in southern Laos.
The objectives were to (1) examine the nature of GxE interaction for grain yield of rainfed upland rice in southern Laos, (2) identify genotypes with high yield under these conditions, and (3) consider the basis of the genotype responses, and what cultural history and soil management may assist in stabilising and improving upland rice yield in southern Laos.
Materials and methods
The experiments were conducted in Savannakhet Province in southern Lao PDR in the 2011 and 2012 wet seasons. Monthly average long-term (1999–2023) temperatures ranged from a low of 15o C in December-January to a high of 35o C in March-May (Supplementary Table 1a). Temperatures followed similar patterns in the two years, but 2012 was slightly warmer. Of the long-term mean annual rainfall (1452 mm), ninety percent was received from May to October, during the wet season, with 2011 somewhat wetter (+266 mm) and 2012 considerably drier (-418 mm). Although rainfall commenced earlier in 2012, monthly rainfall in June to September was only 654 mm (390 mm below the long-term mean), with this deficiency followed by a rapid decline in September and October (195 mm below average). In contrast, from June to September, 2011, rainfall was 415 mm higher than the long-term mean. When long-term means for northern and southern Laos were compared, however, rainfall in the north (Luang Prabang) was similar to the south (Savannakhet), but solar radiation, and especially evaporation, were greater in the south, especially late in the wet season and into the dry season (October to March; Supplementary Table 1b); so, plant water relations late in the growing season were likely to be less favourable in the south.
In each of the two years (2011 and 2012), experiments were conducted in 2 districts, 2 villages per district, and with 5 farms per village. The 2 districts were Xepon (16o 41’ N, 106o 15’ E, 186 m) and Nong (16o 22’ N, 106o 29’ E, 285m). Villages close to Xepon and Nong were chosen, from opposite sides on the main highway, with 5 adjacent farms per village selected as replicates, from within 3 km radius. In May-June 2011, the emerging seedlings in one village in Xepon were lost to heavy rain (694 mm), as the smallholdings were close together, and all plots were submerged there. Consequently, seven sites remained to explore adaptation of upland rice genotypes in southern Laos. Sites were referred to by their site codes (Table 1).
The seven sites used to discriminate upland rice genotypes. l.s.d. are shown below the column for each variable (P < 0.05)
In 2011, the sites at Xepon had been under continuous cultivation for 5 or more years, while those in Nong were new cultivation after 5 or more years of long fallow. Three soil samples (0-0.15 m) from each of 5 adjacent farms within each village and district were pooled for analysis, with the following soil tests conducted at NAFRI (National Agriculture and Forestry Research Institute) Vientiane in 2011: Soil pH, 1:1.25; Soil organic carbon, Walkley and Black; Total N, Kjeldahl; Available P, Bray II; Exchangeable K, 1 N ammonium acetate. The soil at Xepon had a pH of 4.5, organic C 1.34 g kg−1, total N 0.08%, available P 3.57 mg kg−1, and exchangeable K 17.1 cmol kg−1. At Nong, pH was 5.9, with organic C 13.0 g kg−1, total N 1.90%, available P 30.1 mg kg−1, and exchangeable K 11.0 cmol kg−1. The soils were Acrisols, reddish-brown in colour, and moderately acidic, with the new cultivation in Nong more fertile than the old cultivation in Xepon, especially for Total N.
At each farm, 6 genotypes of upland rice were evaluated, comprising 2 indica cultivars (Laboun and Non), 2 tropical japonica cultivars (Nok and Mak Hin Sung), an improved indica cultivar (B6144F-MR-6-0-0), and a tropical japonica local check, which varied from farm to farm. Genotypes were referred to by their genotype code (Table 2).
Genotypes evaluated in upland rice experiments in southern Laos in 2011 and 2012. l.s.d. are shown below the column for each variable (P < 0.05)
Farmers established 5 m × 4 m plots (20 m−2) of each genotype by dibbling by hand between 11 May and 20 May, 2011, and between 27 April and 7 May, 2012. Holes were made with a dibble stick at approximately 0.25 × 0.25 m, with 2-3 seeds per hole. Consistent with traditional upland rice culture, farmers did not apply fertiliser, but plots were hand-weeded as needed. Neither pests nor diseases were observed. At harvest, grain yield was measured from 1.0 m−2 samples from the middle of each plot.
Farmer preference among the genotypes was also evaluated, using cross-site visits prior to harvest in Xeponkao and Meuangsaen in Xepon, and Thang Alai Neu in Nong, in September 2012. Participating farmers were from neighbouring villages, as well as farmers experienced in upland rice production from Phin district near Xepon. The 51 participants were asked to vote for each genotype, giving each a positive, neutral, or negative score. Mean ratings were calculated and recorded, together with the number of participants for each. Farmer interviews were then conducted to examine the reasons for their preferences, such as tillering and panicle characteristics, grain type, plant size and vigour, and local suitability.
Yield data were extracted from appropriate single-site analyses. GxE interactions were analysed using the pattern analysis tool in CropStat (IRRI, 2007). This method involved the joint application of cluster analysis and ordination to a transformed GxE matrix. Since the objective was to understand genotype adaptation for breeding, the GxE matrix was transformed by environment standardisation (Cooper, Reference Cooper1999). The transformed data were clustered using Ward’s agglomerative clustering algorithm. Scores for both genotypes and environments from the two-component interaction principal components analysis (PCA) were computed for PCA1, PCA2, and PCA3, and plotted as biplots, with environment points at the end of spokes with labels as in Table 1, and genotype points as symbols with labels as in Table 2. Means for main effects and interactions were compared using least significant difference with the appropriate degrees of freedom (l.s.d.; P < 0.05).
Results
For the three sites in 2011, mean time to flowering and mean plant height were consistent at 100 days and 99 cm, respectively (Table 1). Over the four sites in 2012, established plant density was also similar, averaging 7.31 plants m−2, but tiller number and especially panicle number varied widely over environments (42.0 to 89.1 panicles m−2). Over all seven sites, mean grain yield ranged from 1.04 to 3.71 t ha−1, with yields in 2011 being 2.5-fold those in 2012 (Table 1).
Among genotypes, for the three sites in 2011, mean time to flowering ranged from 88 days in Laboun (Early maturity) to 98–103 days in B6144, Nok, Non, and Mak Hin Sung (Medium maturity) and 109 days in the Local Check (Late maturity) (Table 2). Plant height was shorter in the improved B6144 (82 cm), intermediate in Nok, Laboun, and Mak Hin Sung (98 cm), and taller in Non and the Local Check (109 cm). Over four sites in 2012, plant density was similar among genotypes (7.31 m−2), but tiller number and especially panicle number were lower in Mak Hin Sung (44.3 panicles m−2) than in Nok, Non, and the Local Check (60.0 panicles m−2) and Laboun and B6144 (90.9 panicles m−2). Over all seven sites, mean grain yield ranged from 1.37 t ha−1 in Local Check to 2.43 t ha−1 in Laboun (Table 2).
In the combined analysis of variance for grain yield, environment main effects accounted for 74.3% of total sum of squares (TSS), with genotype and GxE accounting for 8.0 and 4.5% (Supplementary Table 2). Stability regression accounted for only 7.1% of GxE-SS. Cluster analysis on environment-standardised residuals was used to identify three site groups and three genotype groups, which preserved 60.3, 85.3, and 61.7% of E-SS, G-SS, and GxE-SS, respectively. Ordination analysis indicated 3 interaction principal component axes (PCA), accounting for 50.0, 37.9, and 10.1% of the GxE-SS, respectively, or 98.0% in total.
The cluster dendrogram for environments (Fig. 1a) initially separated 3 sites in Nong (E1) from the remainder, which then split into two groups: Two sites in Xepon 2012 (E2), and Xepon Kengluang 2011 (XK1) and Nong Thang Alai Kao 2012 (NA2) (E3). In the cluster dendrogram for genotypes (Fig. 1b), B6144 and Local Check (G1) separated from the remainder, which then split to the singleton Mak Hin Sung (G2) and Non, Nok, and Laboun (G3). For the biplots, all sites were positive for PCA1 (Fig. 2a). PCA2 separated the sites, with Xepon 2012 positive (E2), three sites in Nong negative (E1), and XK1 and NA2 intermediate (E3). PCA3 then separated XK1 and NA2 from each other, and from the rest (Fig. 2b). This separation for PCA3 was consistent with the dendrogram for environments, in which the next two sites to separate were XK1 and NA2 (Fig. 1a). For genotypes, PCA1 separated the negative Local Check and B6144 (G1) from the remainder, while PCA2 separated Mak Hin Sung from the rest (Fig. 2a). PCA3 then separated positive B6144 from negative Local Check (G1; Fig. 2b).
Environment (a) and genotype (b) groupings applied to standardised yield data for 6 upland rice (Oryza sativa L.) genotypes over 7 environments. The dendrograms show fusion levels at which the groups join. The fusion level is proportional to the increase in within-group SS at each fusion. The vertical lines represent the truncation (a) of 7 environments into 3 groups, and (b) of 6 genotypes into 3 groups, using Ward’s agglomerative clustering algorithm. Refer to Tables 1 and 2 for environment and genotype abbreviations, respectively. Grain yield (t ha–1) is also shown for each group.
Principal component analysis (location standardised) for the 3 environment group x 3 genotype group interaction for grain yield, for (a) PCA1 and PCA2 and (b) PCA1 and PCA3, from 7 environments and 6 upland rice (Oryza sativa L.) genotypes. Refer to Tables 1 and 2 for environment and genotype abbreviations, respectively. The GxE interactions for PCA1 and PCA2, and for PCA1 and PCA3, accounted for 87.9 and 60.1% of the sum of squares, respectively.
For grain yield, genotypes in G3 (Nok, Non, and Laboun) were highest yielding in E1 (three sites in Nong) (3.09 t ha−1; Table 3). Genotype G2 (Mak Hin Sung) was highest yielding in E2 (two sites in Xepon 2012) (1.62 t ha−1). Genotypes in G1 (B6144 and the Local Check) were the lowest yielding in all environment groups (1.69 t ha−1). Although the three wetter sites in 2011 were higher yielding (3.19 t ha−1) than the four drier sites in 2012 (1.23 t ha−1), environments did not group by year or rainfall alone. Panicle number m−2 was high in Xepon 2012, especially in B6144 and Laboun (Supplementary Table 3). Panicle number m−2 was lower in Nong 2012, especially in Nok and Mak Hin Sung.
Grain yield (t ha-1) of 6 genotypes in each of 7 environments in southern Laos. L.s.d. for E, G, and GxE were 0.25, 0.23, and 0.62, respectively (P < 0.05)
Farmers in Xepon preferred Mak Hin Sung and Laboun, and in Nong preferred Laboun and Nok (Table 4). Conversely, farmers in Nong gave Mak Hin Sung the most negative ranking, followed by B6144 and Nok, while farmers in Xepon were strongly negative for Nok, Non, and B6144. Overall, rankings from Nong district were positive, while rankings from Xepon district were negative on average. At Xepon, farmer preference was related to strong tillering, large grain, and suitability to local conditions. At Nong, the preference was for long panicles, large tillers, and suitability to local conditions. In a joint meeting afterwards, the preference was for long panicles, long grains, and strong culms. In contrast, farmer non-preference was for few tillers, short stature, and uneven panicles.
Farmer preference indices (and number of farmers) for upland rice genotypes evaluated in Xepon and Nong districts in southern Laos in 2012, and the reasons stated by farmers for their preference or non-preference for each genotype
Discussion
Variation among the seven sites was large, contributing 74.3% of the TSS (Supplementary Table 2), which is common in rainfed rice experiments (Cooper, Reference Cooper1999; Samson et al., Reference Samson, Voradeth, Zhang, Tao, Xayavong, Khammone, Douangboupha, Sihathep, Sengxua, Phimphachanhvongsod, Bouahom, Jackson, Harnpichitvitaya, Hu and Wade2018). Nevertheless, genotype effects and the patterns of genotype response over environments assisted data interpretation, with three vectors accounting for almost all GxE, suggesting a high repeatable component, as in other studies of upland rice (Botwright Acuna et al., Reference Botwright Acuna, Lafitte and Wade2008).
Two Nong sites in 2011, NL1 and NS1, were the highest yielding (3.55 t ha−1; Table 3). These sites were new cultivation, so their soil fertility was high, and they also encountered above-average seasonal rainfall in 2011 (Supplementary Table 1). Although the mean yield of NN2, the third site in E1, was lower (1.23 t ha−1), its genotype yield rankings were unchanged, suggesting that yield may have been reduced in response to a more moderate level of soil fertility in the second year, rather than any interactive change such as in seasonal rainfall pattern, as its maturity group rankings were also unchanged. Genotypes Non, Nok, and Laboun (G3) were highest yielding in E1 (3.09 t ha−1), and overall.
The two Xepon sites in 2012, XM2 and XK2 (E2) encountered limited rainfall for much of the growing season (Supplementary Table 1) and were old cultivation, so had limited soil fertility. Consequently, they were lowest yielding on average (1.32 t ha−1; Table 3). Mak Hin Sung was highest yielding there (1.62 t ha−1), and genotype rankings were identical within E2. While early maturity (Laboun, 1.34 t ha−1) was generally advantageous over late maturity (Local Check, 1.08 t ha−1) at these drier sites, this did not seem to fully explain the change in genotype rankings from E1. It is notable that Mak Hin Sung was higher yielding in all Xepon sites, especially in 2012, suggesting that some soil factors may have had a role. Root aphids were reportedly more severe under dry rainfed upland conditions (Asai et al., Reference Asai, Saito, Samson, Songyikhangsuthor, Homma, Shiraiwa, Kiyono, Inoue and Horie2009), which may partly explain this response, but would require verification. None of these genotypes are noted for drought tolerance (Saito et al., Reference Saito, Linquist, Atlin, Phanthaboon, Shiraiwa and Horie2006a), so access to genotypes with some drought tolerance would also have been helpful, especially in the drier year, 2012, in these southern Lao sites.
At XK1 and NA2 (E3), the earliest genotype was highest yielding (Laboun, 2.45 t ha−1; Table 3). The lower panicle number m−2 observed at NA2 (Supplementary Table 3) suggests some intermittent stress was encountered in E3, perhaps due to lower rainfall in 2012. Since the plots in the neighbouring village to XK1 in Xepon 2011 were lost to heavy rainfall (Supplementary Table 1), perhaps XK1 was also affected somewhat by heavy rainfall in the wetter year 2011, thereby restricting its tillering and panicle number m−2 as well.
These adaptive relationships are illustrated in Figs. 1 and 2. PCA1, the X-axes in Figs. 2a and 2b, is interpreted as yield potential, with all environments being positive, and with the high-yielding genotypes Nok, Non, and Laboun (G3) to the right, the intermediate-yielding Mak Hin Sung (G2) neutral, and low-yielding B6144 and Local Check (G1) to the left.
The Y-axis in Fig. 2a (PCA2) then separates Mak Hin Sung (G2) from the other genotypes, perhaps due to root aphids, but also separates the three environment groups. It is notable, however, that all three Xepon sites are positive, while all four Nong sites are negative (Fig. 2a). Perhaps, PCA2 relates to cultivation history, since the new cultivation sites of high fertility in Nong are together, and the older cultivation sites of lower fertility in Xepon are together. PCA2 could also include some contribution from water deficit, with drier sites in Xepon 2012 being together (Fig. 2a). Severity of root aphids was reportedly greater under stress, but was reduced by higher soil fertility (Asai et al., Reference Asai, Saito, Samson, Songyikhangsuthor, Homma, Shiraiwa, Kiyono, Inoue and Horie2009).
The Y-axis in Fig. 2b (PCA3) separates the two E3 sites from each other (positive) and from the rest. PCA3 also separates the positive B6144 and Laboun from neutral Nok and Mak Sin Hung, which in turn separate from Non and Local Check. It is notable that PCA3 separates the genotypes by maturity, height, and panicle number, with B6144 and Laboun (88–98 days, 82–98 cm, and 93–120 panicles m−2) being early and short with more panicles, while Non and Local Check (103–109 days, 108–109 cm, and 69–55 panicles m−2) are late and tall with fewer panicles (Supplementary Table 3). This is consistent with the sites XK1 and NA2 (E3) encountering some intermittent growth restrictions, which favour early maturity, short stature, and higher panicle number m−2 to facilitate yield expression, as was also observed for rainfed lowland rice in southern Laos (Sengxua et al., Reference Sengxua, Samson, Bounphanousay, Xayavong, Douangboupha, Harnpichitvitaya, Jackson and Wade2017). Recent research on the regulation of tillering in upland rice may allow breeders to select for panicle number (Lyu et al., Reference Lyu, Huang, Zhang, Zhang, He, Zeng, Zeng, Huang, Zhang, Ning, Bao, Zhao, Fu, Wade, Chen, Wang and Hu2020).
Overall, upland rice response was related to four cultural history by year groupings: Nong 2011 (E1: new cultivation, wet year, high yield potential), Xepon 2012 (E2: old cultivation, dry year, low yield potential), and intermittent stress (E3), associated with either old cultivation in a wet year (Xepon 2011) or new cultivation in a dry year (Nong 2012).
Farmers preferred Mak Hin Sung for its strong tillering, large grain size, and ability to grow on different soils in Xepon only, but said it was not suited to local conditions in Nong (Table 4). In contrast, Laboun was preferred in both Xepon and Nong for its high panicle number, long panicles, large grain size, and ability to grow in local conditions. Farmers disliked B6144F-MR-6-0-0, as it was the only non-glutinous genotype, and it had smaller panicles and smaller grain size. Following the experiments and the farmer assessment, some upland farmers in Xepon adopted Mak Hin Sung, and some farmers in both Nong and Xepon adopted Laboun.
Farmer interviews conducted seven years later, when the outcomes of this project were reviewed (Wade and Sengxua, Reference Wade and Sengxua2019), revealed that some upland farmers were still growing Mak Hin Sung in Xepon, and some upland farmers were still growing Laboun in both Nong and Xepon. This subsequent adoption was consistent with the original project data on farmer preference (Table 4). Farmers also indicated they had passed seed to relatives and neighbours, and even to friends from a neighbouring village, which was consistent with patterns of farmer behaviour identified in other studies of technology adoption by smallholder farmers in southern Laos (Sengxua et al., Reference Sengxua, Jackson, Simali, Vial, Douangboupha, Clarke, Harnpichitvitaya and Wade2019; Clarke et al., Reference Clarke, Jackson, Koeka, Phimphachanhvongsod, Sengxua, Simali and Wade2018). Continuing farmer adoption long after the completion of a project is a strong affirmation of the results.
It is notable that B6144 did poorly at all sites in southern Laos (Table 3), and with no simple grouping of indica and tropical japonica genotypes (Table 3), in contrast with previous reports from northern Laos (Saito et al., Reference Saito, Linquist, Atlin, Phanthaboon, Shiraiwa and Horie2006a; Asai et al., Reference Asai, Saito, Samson, Songyikhangsuthor, Homma, Shiraiwa, Kiyono, Inoue and Horie2009). The results here demonstrate that rainfall was similar, but evaporation was higher late in the wet season and into the dry season in southern Laos (Supplementary Table 1b), implying conditions during grain filling were less favourable in southern Laos. Consequently, grain yields were lower, especially in 2012 (Table 3), when rainfall deficit was also more severe (Supplementary Table 1a). In contrast, rainfed conditions are generally considered favourable for upland rice in northern Laos, with water deficit being considered mild or intermittent only (Schiller et al., Reference Schiller, Linquist, Douangsila, Inthapanya, Douangboupha, Inthavong and Sengxua2001). Recently, reports from Brazil have also implicated more severe water deficit limiting yields of upland rice there (Heinemann et al., Reference Heinemann, Ramirez-Villegas, Rebolledo, Costa Neto and Castro2019), requiring testing of potential genotypes for their ability to maintain grain yield under water deficit. We conclude that upland rice for southern Laos requires greater tolerance to water deficit, especially during grain filling.
Performance of upland rice was strongly influenced by whether the cropping cycle was new cultivation or following multiple previous rice crops (Asai et al., Reference Asai, Saito, Samson, Songyikhangsuthor, Homma, Shiraiwa, Kiyono, Inoue and Horie2009; Asai and Soisouvanh, Reference Asai and Soisouvanh2017; Sengxua et al., Reference Sengxua, Toomsan, Linquist, Lefroy and Limpinuntana2007). With restrictions on land available for upland cropping and increasing population pressure, fallow periods have reduced from 40 years to only 2 or 3 years since the 1960s (Nye and Greenland, Reference Nye and Greenland1960; Linquist et al., Reference Linquist, Trosh, Pandey, Phouynyavong and Guenat2007). This is putting huge pressures on system sustainability, with soil erosion and nutrient leaching depleting the resources available for subsequent crops (Asai et al., Reference Asai, Saito, Samson, Vongmixay, Kiyono, Inoue, Shiraiwa, Homma and Horie2007; Saito et al., Reference Saito, Linquist, Keobualapha, Phanthaboon, Shiraiwa and Horie2006b). Indeed, grain yields have been observed to have declined linearly with the loss of soil organic carbon (Asai et al., Reference Asai, Saito, Samson, Songyikhangsuthor, Homma, Shiraiwa, Kiyono, Inoue and Horie2009; Saito et al., Reference Saito, Linquist, Keobualapha, Phanthaboon, Shiraiwa and Horie2006b), which is important for soil structure, soil water-holding capacity, and nutrient buffering. Efforts to correct this via application of inorganic or organic fertiliser, legume rotations and fallows, alley farming, and rows of contour forage have been tried (Asai and Soisouvanh, Reference Asai and Soisouvanh2017; Roder, Reference Roder2001; Saito et al., Reference Saito, Linquist, Johnston, Phengchanh, Shiraiwa and Horie2008), but have often been ineffective in improving upland rice yield on these acidic, coarse-textured soils. Perhaps, all these factors need to be combined, as argued recently for the similarly acidic and coarse-textured soils supporting rainfed lowland rice grown in bunded fields lower in the topography (Sengxua et al., Reference Sengxua, Inthavong, Sihatep, Samson, Newby, Jackson, Harnpichitvitaya and Wade2022). A longer-term and moderate input strategy was proposed to gradually improve fine-fraction soil organic carbon, soil nutrient buffering, and soil microbial biomass, in order to improve soil nutrient and water retention and nutrient release characteristics. This strategy should be evaluated in further research, along with selection for yield stability in upland rice under water deficit. Such systems research may also reduce the yield gap, which is regularly identified as a major concern in upland rice (Saito et al., Reference Saito, Senthikumar, Dossou-Yoyo, Ali, Johnson, Mujawamariya and Rodenburg2023; Santosa et al., Reference Santosa, Kurniasih, Alam, Handayani, Supriyanta, Ansari and Taryono2024).
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0014479725100148.
Acknowledgements
We thank the farmers for use of their land and site management, and the Provincial Government of Savannakhet for use of their facilities, labour, and extension support.
Funding statement
This work was supported by the Australian Centre for International Agricultural Research (ACIAR): Grant number CSE/2009/004.
Competing interests
The authors declare that they have no known competing financial or personal relationships that could have influenced this work. ACIAR had no role in the design, analysis, or writing.
Authorship
KS: Formulation, Design, and Interpretation; SV: Implementation; PS: Formulation, Design, and Implementation; BKS: Formulation, Design, and Implementation; TJ: Formulation, Design, and Implementation; DH: Analysis; LJW: Analysis, Interpretation, and Writing.
Abbreviations: B6144, B6144F-MR-6-0-0; GxE, genotype by environment interaction; PCA, principal component axis; TSS, total sum of squares.
Open access
