Sahbhagi Dhan has shown relatively high and stable yield across a range of stress-prone and irrigated environments, which has thus far been documented in more detail in managed drought trials in transplanted and rainfed lowland conditions. In advanced yield trials at IRRI and eight institutes in India, Verulkar et al. (2010) reported average yields of Sahbhagi Dhan that were consistently higher than those of standard checks in irrigated, moderate drought, and severe drought conditions. Based on genotype plus genotype × environment plots, Kumar et al. (2012) observed that Sahbhagi Dhan was the most stable yielding out of about 40 entries across 16 rainfed stress and irrigated environments at three locations in eastern India in wet-season trials. In a comparison of multiple drought-tolerance indices, Raman et al. (2012) reported a high ranking for Sahbhagi Dhan and consistently higher yield than IR64 and MTU1010 (popular high-yielding but drought-susceptible varieties) across irrigated and drought-stress environments. According to wet-season data, Sahbhagi Dhan also stood out for stability across environments, resulting in yields of >3 t ha−1 in a simulation of yield in 669 environments conducted using the crop model ORYZA2000, although several other genotypes in the group of 69 genotypes were estimated to be higher ranking in more environments for stability (Li et al., 2013).
Observations in farmers’ fields, results of participatory variety trials for farmer preference, and farmer testimonials have indicated a number of advantages of growing Sahbhagi Dhan. Sahbhagi Dhan has been observed to perform well in both direct-seeded upland and unfavorable transplanted lowland environments. Under moderate to severe drought where IR64 failed to flower and provide any yield, Sahbhagi Dhan has reportedly still yielded 0.8 to 1.0 t ha−1 (Dar et al., 2014). In addition to its higher yield under drought stress, Sahbhagi Dhan can increase profits of farmers because it requires less irrigation, and its short duration allows additional crops to be grown per year (Reyes, 2014). The shorter duration of Sahbhagi Dhan allows farmers to use remaining moisture in the field to plant and grow subsequent summer season legume or pulse crops, thereby not only providing additional returns but also better nutritional security to poor farmers of rainfed ecosystems. Other benefits to farmers by growing Sahbhagi Dhan include earlier access to food, good straw yield for animal feed (Dobermann, 2012), tolerance of drought stress at multiple growth stages, and pest resistance (Reyes, 2009). For these reasons Sahbhagi Dhan seed kits have already been distributed to and grown by hundreds of thousands of farmers in South Asia (Dar et al., 2014). Because of its adaptability, semitall stature, and short duration, crop management recommendations for Sahbhagi Dhan allow for different seed establishment methods (direct-seeded or transplanted) to be used depending on the amount of early-season rainfall (Variar et al., 2010).
Since Sahbhagi Dhan has been observed to provide a yield advantage under drought at multiple growth stages and adaptability to different establishment methods, we hypothesized that this response is conferred by multiple physiological mechanisms. In an effort to explain the adaptability and yield stability of Sahbhagi Dhan across multiple environments and crop establishment methods, this study was conducted to understand the physiological characteristics of Sahbhagi Dhan in comparison with currently cultivated, high-yielding but drought-susceptible varieties as well as more recently developed drought-tolerant breeding lines.
MATERIALS AND METHODS
This study evaluated 12 rice genotypes, including several that have now been released as varieties: Sahbhagi Dhan (IR74371-70-1-1), drought-tolerant breeding lines that were developed subsequent to Sahbhagi Dhan, and checks IR55419-04 and IR64 (Table 1). IR64 is a high-yielding but drought-susceptible variety that is popular in South and Southeast Asia, and IR55419-04 is a drought-tolerant genotype with high potential for release as a variety in drought-prone regions. Another genotype, R1837-RF-40, is a drought-tolerant breeding line derived from IR55419-04 that was included in the Raipur 2011 dry-season vegetative-stage stress experiment only. IR83388-B-B-108-3 was missing from some IRRI trials. All genotypes showed early-to-medium maturity duration and ranged in time to flowering under well-watered conditions from 76 d after sowing (DAS) for IR74371-70-1-1, IR74371-54-1-1, and IR55419-04 to 87 DAS for IR83388-B-B-108-3.
|Genotype||Year of original cross||Male parent||Female parent||Varietal release information|
|IR55419-04||1986||UPLRi 5||IR12979-24-1||Evaluated in the India drought breeding network, showed promising performance in Chhattisgarh (not yet released as variety)|
|IR64||1977||IR2061-465-1-5-5||IR5657-33-2-1||Released as IR64 in 1985 in the Philippines, in 1988 in Bhutan, in 1992 in India. Also released as IR64 in Cambodia, China, Gambia, Indonesia, Mauritania, and Mozambique; FKR42 in Burkina Faso; INIAP11 in Ecuador; and OM89 in Vietnam (Khush and Virk, 2005).|
|IR74371-54-1-1||1996||IR55419-04||Way Rarem||Released in 2010 as Sukha Dhan 2 in Nepal|
|IR74371-70-1-1||1996||IR55419-04||Way Rarem||Released in 2010 as Sahbhaghi Dhan in India, BRRI Dhan 56 in Bangladesh, and Sukha Dhan 3 in Nepal|
|IR83380-B-B-124-3||2005||PSBRC 18||IR72022-46-2-3-3-2||Promising drought-tolerant line, sister line of IR83380-B-B-124-1, which was released as CR Dhan 201 in India|
|IR83388-B-B-108-3||2005||Swarna||IR72022-46-2-3-3-2||Released in 2014 as Sukha Dhan 5 in Nepal|
Seedling-Stage Stress Experiment
A seedling-stage stress experiment was conducted to characterize genetic differences in response to sowing under direct-seeded conditions at IRRI, Los Baños, Laguna, Philippines (14°10′11.81″ N, 121°15′39.22″ E), in the 2012 dry season (DS; January–April). Soil particle size distribution was 36% sand, 40% silt, and 24% clay at the 0- to 15-cm depth. Plots in three rows of 3 m in length (2.25 m2) were sown in four replicates at a depth of about 2.5 cm in a randomized complete block design, with each of the three treatments located in a separate area of the same experimental field. Basal fertilizer was applied at the rate of 40:40:40 kg NPK ha−1. All plots were sprinkler irrigated twice within the first 3 DAS before initiating the three treatments for approximately 1 mo: Rainfed Treatment 1, which did not receive any additional inputs during the treatment period; Rainfed Treatment 2, which received one additional fertilizer (ammonium sulfate at 50 kg N ha−1) and irrigation treatment at 23 DAS; and a well-watered treatment, which was sprinkler irrigated up to three times per week as needed to supplement rainfall. Weeds were controlled manually as needed. A total of 82.5 mm of rainfall occurred during the treatment period, with only 5 mm of rainfall occurring during the first 12 DAS. At 29 DAS, all treatments were sprinkler-irrigated and were maintained as in the well-watered treatment for the remainder of the experiment. An additional topdressing of ammonium sulfate at the rate of 50 kg N ha−1 was applied to all treatments at 33 DAS. At the end of the season, an area of 1.5 m2 per plot was harvested, which was set at a fixed location to account for genotypic differences in germination and emergence that could later be reflected in the total straw mass and grain yield.
Lowland Drought Stress Experiments
Field experiments were conducted across five locations: IRRI, Los Baños, Laguna, Philippines (14°10′11.81″ N, 121°15′39.22″ E) and at four sites in India: Central Rice Research Institute (CRRI; Cuttack, Odisha, 20°27′10.99″ N, 85°56′26.46″ E), Central Rainfed Upland Rice Research Station (CRURRS; Hazaribag, India, 23°57′39.30″ N, 85°22′5.07″ E), Indira Gandhi Agricultural University (IGKV; Raipur, Chhattisgarh, 21°14′2.72″ N, 81°43′1.67″ E), and Narendra Dev University of Agriculture and Technology (NDUAT; Masodha (Faizabad), Uttar Pradesh, 26°43′22.80″ N, 82°8′7.33″ E). The percentage of sand, silt, and clay, respectively, averaged 18, 44, and 38% at IRRI; 52, 22, and 26% at Cuttack; 24, 47, and 29% at Hazaribag; 34, 12, and 54% at Raipur; and 34, 58, and 8% at Faizabad.
A total of 27 lowland trials were conducted across locations in both the DS (December–April) and wet season (WS; July–November) from 2011 to 2014 (Table 2). The experiments were transplanted at 15 to 20 DAS into puddled soil in plots of 4- by 3-m rows (with 0.2 m between hills and 0.20–0.25 m between rows) arranged in four replicates in a randomized complete block design. Fertilizer was applied at a rate of 80:40:40 kg NPK ha−1 for the trials in Cuttack, Faizabad, and Hazaribag; 120:80:40 kg NPK ha−1 under irrigated conditions at Raipur; 80:60:40 kg NPK ha−1 in the drought-stress treatments at Raipur; and 40:40:40 kg NPK ha−1 before transplanting with a topdressing of (NH4)2SO4 at a rate of 40 kg N ha−1 3 to 4 wk after transplanting for the trials at IRRI. Drought-stress and well-watered treatments were planted in separate locations at each site and were considered as separate experiments. The vegetative-stage drought-stress experiments were drained 15 to 20 d after transplanting (35–40 DAS), and irrigation was withheld until the first genotype reached reproductive stage (∼60 DAS). Reproductive-stage stress experiments were drained when the first genotype reached reproductive stage. Well-watered experiments were maintained flooded throughout the season. The Raipur 2011 wet-season experiments included four treatments: rainfed direct seeded, rainfed transplanted, transplanted terminal-stage drought, and irrigated transplanted. Weeds were controlled manually as needed in all experiments. The drought-stress treatments were interrupted by varying amounts of rainfall in each experiment (Table 2).
|Experiment type and no.||Location†||Season||Rainfall during stress treatment|
|1||CRURRS, Hazaribag, India||2011 DS‡||21|
|2||CRURRS, Hazaribag, India||2011 WS||235|
|3||CRURRS, Hazaribag, India||2012 DS‡||27|
|4||IGKV, Raipur, India||2011 DS||60|
|5||IRRI, Philippines||2011 DS‡||4.7|
|6||IRRI, Philippines||2011 WS||185|
|7||IRRI, Philippines||2012 DS‡||81|
|8||CRRI, Cuttack, India||2011 DS||500|
|9||CRURRS, Hazaribag, India||2011 WS||44|
|10||IGKV, Raipur, India||2011 WS‡||514|
|11||IGKV, Raipur, India||2011 WS‡§||410|
|12||IGKV, Raipur, India||2011 WS‡¶||514|
|13||IRRI, Philippines||2011 DS||72|
|14||IRRI, Philippines||2011 WS||525|
|15||IRRI, Philippines||2012 DS||185|
|16||IRRI, Philippines||2014 DS‡||25|
|17||NDUAT, Faizabad, India||2011 DS||152|
|18||NDUAT, Faizabad, India||2012 DS||154|
|19||CRRI, Cuttack, India||2011 DS||na#|
|20||CRURRS, Hazaribag, India||2011 WS||na|
|21||IGKV, Raipur, India||2011 WS||na|
|22||IRRI, Philippines||2011 DS||na|
|23||IRRI, Philippines||2011 WS||na|
|24||IRRI, Philippines||2012 DS||na|
|25||IRRI, Philippines||2014 DS‡||na|
|26||NDUAT, Faizabad, India||2011 DS||na|
|27||NDUAT, Faizabad, India||2012 DS||na|
Physiological Measurements in the Lowland Drought-Stress Experiments
Shoot biomass was sampled at the end of the vegetative-stage treatments at IRRI and Hazaribag (65–91 DAS) by cutting three hills per plot at the base of the plant, drying at ∼65°C, and weighing.
Root growth was evaluated in nine experiments: four at IRRI, Philippines (2011 DS vegetative-stage stress [73 DAS], 2012 DS vegetative-stage stress [70 DAS], and 2014 DS reproductive-stage stress and well-watered treatments [118 DAS]); two at CRURRS Hazaribag (2011 DS [71 DAS] and 2012 DS [113 DAS] vegetative-stage stress trials); and three at IGKV Raipur (54–80 DAS; 2011 WS rainfed direct seeded, rainfed transplanted, and transplanted terminal-stage drought). A core sampler of 4-cm diameter and 60-cm length was used to collect soil samples for root measurements to depths of 45 cm at Hazaribag and 60 cm at Raipur and IRRI. The core sampler was fabricated at IRRI. Root sampling was done from three locations in each plot in between two rows. The soil core was divided into equal parts corresponding to 15-cm soil depths. Roots were washed from the soil immediately after sampling or stored at −4°C until washing (within 3 wk). The roots were washed by repeatedly mixing the soil with water in a container and pouring the root water suspension over a 1-mm plastic sieve. All samples were stored in 25% ethanol until scanning. Root samples were scanned at 400 dpi (Hazaribag 2011 DS) or 600 dpi (Epson V 700). Scanned images were analyzed for total root length and length within diameter classes using WinRhizo v.2007d (Regent Instruments). The proportion of total root length as lateral roots (percentage lateral roots) was evaluated in each sample as lateral root length (<0.2-mm diam./total root length × 100). The proportion of lateral roots was not determined from the images from the Hazaribag 2011 experiment because of the lower resolution of those images. The percentage of deep roots in the soil core was calculated as root length below 30 cm/total root length in the soil core (0–60 cm).
Canopy temperature was measured by infrared thermography (NEC TH7800 infrared camera, NEC Avio Infrared Technologies Co. Ltd.) at midday from a 3-m ladder placed 5 to 10 m in front of the plots in the Hazaribag, Raipur, and IRRI 2011 DS vegetative-stage stress experiments. Images were analyzed using Report Generator software v.1.7 (NEC Avio Infrared Technologies Co. Ltd.). Sequential canopy temperature measurements across the drought-stress period were conducted in all IRRI drought-stress experiments by infrared thermography or with an infrared sensor (three locations per plot; Apogee Instruments). Sequential normalized difference vegetation index (NDVI) measurements across the drought-stress period were conducted in all IRRI drought-stress experiments using a Greenseeker handheld sensor (NTech Industries).
At the end of each season (except in the 2011 DS at Raipur, in which high temperatures necessitated the experiment to be ended after the vegetative stage), a 1.5- to 2.4-m2 area was harvested to determine straw biomass, grain yield, and harvest index (grain yield/[grain yield + straw biomass]).
Mean values for each physiological parameter measured were analyzed per experiment by ANOVA in R v. 2.15.2 (R Development Core Team, 2010) or Statistical Tool for Agricultural Research (STAR ver. 2.0.1; IRRI), using the Shapiro–Wilk test for normality, Bartlett’s test for homogeneity of variances, and Tukey’s honest significant difference pairwise mean comparison as the post hoc test used to identify which genotypes were significantly different. The percentage of deep roots and percentage lateral roots (grouped by seasonal conditions) were analyzed in combined analyses with experiment and genotype as factors by ANOVA in STAR. The combined ANOVA model was as follows:where µ is the grand mean for the trait, si is the environmental effect for the ith experiment, b(k)i is the kth replicate effect within the ith experiment, τj is the ith genotype effect, sτij is the interaction effect between the ith experiment and jth genotype, and eij is random error associated with the trait. A similar model was used for the one-factor randomized complete block design analysis, but with the experiment factors omitted. The one-factor ANOVA model was used for comparing genotypes in the seedling-stage stress experiment (straw mass and grain yield) and for canopy temperature under vegetative-stage stress (2011 DS), final NDVI, and harvest index in the lowland drought-stress experiments. For NDVI and canopy temperature that were measured repeatedly, genotypic differences were estimated using the ASReml (Butler et al., 2009) package with Wald’s test in R, with DAS as a fixed variable and replicate as a random variable.
The additive main effects and multiplicative interaction (AMMI) model in multienvironment trials and the genotype main effects and genotype × environment interaction effects model were used to analyze grain yield data from all sites (drought and well-watered experiments, with wet and dry seasons analyzed separately) for yield stability using Plant Breeding Tools (PBTools ver. 1.3, IRRI). The AMMI model was as follows:where µ is the grand mean for the trait; τi is the genotypic effect; δj is the environmental effect; the constant λk is the singular value of the kth bilinear (multiplicative) component that is ordered λ1 ≥ λ2 ≥ … ≥ λt; αik are elements of the kth left singular vector of the true interaction and represents genotypic sensitivity to hypothetical environmental factors represented by the kth right singular vector with elements γjk. The terms αik and γjk satisfy the following constraints:
Seedling-Stage Stress Experiment
The direct-seeded seedling-stage stress experiment showed different emergence rates among genotypes (Fig. 1A) and treatments that were reflected in the straw biomass and grain yield per plot area at harvest (Fig. 1B, C). Sahbhagi Dhan showed higher rates of emergence based on visual observations and was included in the highest shoot biomass significance groups in the two rainfed treatments (p < 0.001; Fig. 1).
Lowland Drought-Stress Experiments
In the lowland drought experiments, Sahbhagi Dhan showed variable performance for midseason shoot biomass compared with other genotypes under vegetative-stage stress, and this appeared to coincide with variable climate conditions. Lower-than-expected midseason biomass was observed in the 2011 DS in Hazaribag, in which early-season temperatures were low (25.5°C during the first 2 wk after sowing; Fig. 2A, B). Although seasonal differences in temperature at IRRI were small (data not shown), lower-than-expected midseason biomass was observed in the 2012 DS and 2014 DS at IRRI, in which early- and midseason solar radiation values were low (12.4–13.8 MJ m−2 d−1; Fig. 2C, D).
Although genotypes differed significantly for the percentage of total root length in the soil core below 30 cm (percentage deep roots) in a combined analysis across experiments, no genotype stood out as having consistently greater percentage deep roots across experiments (Fig. 3A). A significantly higher proportion of total root length as lateral roots (percentage lateral roots) was observed in Sahbhagi Dhan than in other genotypes in seasons with more typical air temperature and solar radiation in which it did not show a midseason shoot biomass reduction (Hazaribag 2012 DS and IRRI 2011 DS; p < 0.05; Fig. 3B). However, in the vegetative-stage stress treatments at Hazaribag and IRRI, differences in lateral root growth were linked to the climate-related responses in midseason shoot biomass (Fig. 2, 3B); no differences in lateral root growth were observed among genotypes in seasons resulting in relatively low shoot biomass of Sahbhagi Dhan (Hazaribag 2011 DS and IRRI 2012 DS).
Sahbhagi Dhan showed the lowest average canopy temperatures at Raipur and IRRI, although no significant genotypic differences in canopy temperature at vegetative stage were observed in the 2011 DS experiments. Sahbhagi Dhan showed the highest canopy temperature at Hazaribag (at which the shoot biomass was lower than expected) (Fig. 4). Sahbhagi Dhan did not stand out from the other genotypes across repeated measurements of canopy temperature (Table 3) or NDVI (Table 4) and did not show high NDVI at the end of the seasons at IRRI (Fig. 5).
|2011 DS||2011 DS||2011 WS||2012 DS||2014 DS|
|Stress treatment||Vegetative stage||Reproductive stage||Reproductive stage||Reproductive stage||Reproductive stage|
|No. dates measured||2||5||3||5||6|
|Days after sowing p-value||<0.001||<0.001||<0.001||<0.001||<0.001|
|2011 DS||2011 DS||2011 WS||2011 WS||2012 DS||2012 DS||2014 DS|
|Stress treatment||Vegetative stage||Reproductive stage||Vegetative stage||Reproductive stage||Vegetative stage||Reproductive stage||Reproductive stage|
|No. dates measured||4||6||5||5||3||6||5|
|Days after sowing p-value||<0.001||<0.001||<0.001||<0.001||<0.001||<0.001||<0.001|
Genotypic differences in harvest index in the lowland experiments varied by season and location (Fig. 6). Sahbhagi Dhan was included in the highest harvest index significance groups in one out of eight well-watered experiments, three out of five vegetative-stage drought-stress experiments, and four out of eight reproductive-stage drought-stress experiments at IRRI and in India.
The AMMI analysis conducted separately for the wet seasons and dry seasons with the different treatments and locations together revealed Sahbhagi Dhan to be more stable during the wet season than the dry season. In the wet seasons, principal component 1 (PCA1) explained 50.6% of the total variation for grain yield. Sahbhagi Dhan (Fig. 7; genotype G4) was the most stable variety in wet seasons based on its location closest to the intercept for PCA1 in the biplot (Fig. 7A), with an average grain yield of 326 g m−2, nearly equal to the wet-season grand mean of the 12 genotypes across eight locations (331 g m−2). IR83380-B-B-124-3 (Fig. 7, genotype G7) was the next most stable genotype, with a higher average grain yield of 357 g m−2. The next most stable genotypes in the wet seasons were IR83376-B-B-130-3 (G5) and IR55419-04 (G1) with average grain yields of 318 and 283 g m−2, respectively. IR64, the drought-susceptible check (genotype G2), was one of the genotypes with the least stable yield across wet seasons.
In the dry seasons, PC1 explained 40.2% of the total variation for grain yield (Fig. 7B). Among the 12 genotypes, IR64 (G2) was the most stable with an average grain yield of 220 g m−2, followed by IR83383-B-B-141-4 (G9), which showed the highest average grain yield of 254 g m−2. Other stable genotypes in the dry season were IR83387-B-B-27-4 (G11; average yield 230 g m−2), IR83383-B-B-141-2 (G8; average yield 244 g m−2), and IR83388-B-B-108-3 (G12; average yield 224 g m−2).
The high yield of Sahbhagi Dhan in previous research experiments as well as the evidence for good performance of Sahbhagi Dhan and popularity among farmers of rainfed rice in South Asia are likely due to multiple drought response mechanisms of Sahbhagi Dhan that were observed in this study at a range of growth stages: high emergence under direct seeding, adaptable root growth in terms of lateral root production at vegetative stage, and high harvest index.
Emergence ability is a key trait for rice adaptation to direct-seeded conditions that can result in improved crop stand and competition with weeds (Kumar and Ladha, 2011). The emergence ability of Sahbhagi Dhan under direct-seeded conditions (Fig. 1) is a likely reason that this variety is being promoted as suitable for either direct-seeding or transplanting in rainfed farmers’ fields (Variar et al., 2010), both of which may be drought-prone at germination and seedling stage. The high seedling emergence leading to early vegetative growth, together with its semitall height, likely help Sahbhagi Dhan compete with weeds and maintain better plant populations under stress compared with other currently grown rice varieties under dry direct seeding.
Another favorable drought response trait observed for Sahbhagi Dhan was the proportion of total root length as lateral roots, which appeared to increase in seasons that were generally more favorable for growth of Sahbhagi Dhan (Fig. 3B). The detection of this trait during vegetative-stage stress likely explains the relatively low canopy temperatures of Sahbhagi Dhan during the vegetative-stage stress experiments (Fig. 4), except at Hazaribag, where the biomass of Sahbhagi Dhan was low and the canopy temperature measurement was likely also influenced by high interception by the soil in the infrared images. Such plasticity in lateral root growth has been highlighted as an important trait for rice adaptation to the fluctuating soil moisture conditions characteristic of rainfed lowland fields (Bañoc et al., 2000) and has been identified as a mechanism by which several drought-tolerant rice genotypes maintain relatively high yields, including CSSL 47 from a ‘Nipponbare’ × ‘Kasalath’ cross (Suralta et al., 2010) and NPT × IR64 INL YTH 183 (Kano-Nakata et al., 2013). Lateral root growth under drought is also one of the most consistent mechanisms observed to be conferred by rice drought-yield quantitative trait locus qDTY12.1 (Henry et al., 2014), which has also been observed to be present in the genome of Sahbhagi Dhan and its sister line IR74371-46-1-1 (A. Kumar, personal communication, 2015; Mishra et al., 2013). Interestingly, Sahbhagi Dhan did not stand out as having a greater proportion of root growth at depth (Fig. 3A), which may explain its lack of effect on canopy temperature across dates measured throughout the seasons (Table 3).
The relatively high harvest index of Sahbhagi Dhan, particularly under drought stress (Fig. 6), also helps to explain its high and stable yield across environments. The ability to mobilize stored C during grain filling—which a high harvest index indicates—can help offset the reduced assimilation of C caused by closed stomata under water deficit (Yang and Zhang, 2010). High harvest index is a recommended trait, particularly for late-season drought in rainfed lowland rice (Fukai et al., 1999). The lack of a stay-green phenotype in Sahbhagi Dhan, as evidenced by the low end-of-season NDVI (Fig. 5), further points to remobilization of stored carbohydrates as a potentially important mechanism for grain filling in Sahbhagi Dhan.
Although Sahbhagi Dhan showed several favorable traits for drought response, some unfavorable aspects of Sahbhagi Dhan were also observed. In the dry seasons, low temperatures or low solar radiation resulted in relatively poor vegetative growth of Sahbhagi Dhan compared with other genotypes (Fig. 2), which was related to a lack of lateral root response to drought (Fig. 3B). Such a variable response of Sahbhagi Dhan was also evident in its low yield stability in the dry seasons (Fig. 7); this may be expected since Sahbhagi Dhan was bred for growth during the wet season, which is the main growing season for rainfed rice in South Asia. Dry seasons with initially low temperatures during seedling stage, followed by heat stress during grain-filling stage, are characteristic in many rainfed rice-growing regions of South Asia. Since Sahbhagi Dhan was bred for wet-season cultivation, it may not be a suitable variety to be grown in the dry season in areas with low temperatures during germination and seedling stage unless it is improved for tolerance of seedling-stage cold stress.
Fortunately, as drought breeding continues beyond the development of Sahbhagi Dhan, newer lines may prove to show even better performance across rainfed lowland environments. In this study, more recent breeding lines were observed to outperform Sahbhagi Dhan in terms of yield, for example, IR83380-B-B-124-3, which showed similar stability but higher yield across wet seasons, and IR83383-B-B-141-4, which showed higher and more stable yield across dry seasons. Based on the physiological measurements conducted, this yield advantage in IR83380-B-B-124-3 may be due to higher shoot biomass, as evidenced by high midseason biomass (Fig. 2) and high NDVI across the seasons (Table 4) as well as high harvest index (Fig. 6) but probably not increased root growth or water uptake as indicated by canopy temperature (Table 3). For genotype IR83383-B-B-141-4, more stable deep root and lateral root production across seasons (Fig. 3A, B) as well as high harvest index (Fig. 6) may explain the improved yield and stability compared with those of Sahbhagi Dhan.
The approach of direct selection for grain yield under drought and well-watered conditions resulted in the combination of traits identified in this study to be expressed by Sahbhagi Dhan (better ability to germinate in dry soil, plasticity in lateral root growth, and higher harvest index under drought) at a range of growth stages, together with high yield. The increasing intensity and severity of drought predicted under climate change, the delayed onset of monsoons observed to occur more frequently, as well as variable rainfall patterns (Wassmann et al., 2009) will necessitate further development of rice varieties tolerant to multiple-stage drought. Indeed, some of the more recently developed breeding lines in this study were observed to be promising material for use as drought-tolerant varieties beyond Sahbhagi Dhan. Rigorous long-term efforts are needed to combine tolerance of seedling-stage and reproductive-stage drought-stress, tolerance to low temperatures at seedling stage, the adaptability of the crop to grow under dry direct-seeded or transplanted conditions based on rainfall received at the beginning of the season, high yield potential, and tolerance of biotic stress to provide farmers of rainfed rice with more alternatives to achieve better yield. Such progress will undoubtedly require continued incorporation of combinations of physiological traits into single rice varieties.