Screening of drought-tolerant rice landraces using various drought indices in Nepal Authors: Nabin Lamichhane, Urmila Dhami, Sudan Bhandari, Gunanand Pant, Lal Bahadur Thapa, Chandra Prasad Pokhrel, Ram Kailash Prasad Yadav Published: 2025 ABSTRACT AbstractDrought-resilient varieties are of great significance in the context of climate change and the increasing frequency of droughts. This study evaluated the drought tolerance in twelve rice landraces from three distinct agro-climatic zones (Tarai, Inner-Tarai, and Mid-Hill) in Far-west Nepal. The landraces were subjected to control, intermittent drought, and complete drought treatments to assess their yield and drought tolerance indices. The yield reductions were observed across all varieties. Based on the generalized linear model (GLM), varieties like Chamade, Ratomarso, Jhumke from Mid-Hill, Jhini, Shanti from Inner-Tarai, Lalchand and Anjana varieties from Tarai have significantly higher yield loss. Varieties like Temase, Chiudi, Batebudo, Sauthyari and Ghiupuri have less yield reduction than the baseline mean of the studied system i.e., 15.9%. Drought tolerance indices effectively differentiated landrace responses. Ghiupuri, Lalchand, and Sauthyari exhibited superior tolerance based on consistently higher index values. Principal component analysis and cluster analysis revealed distinct patterns of response among landraces, highlighting the importance of considering multiple traits and employing multivariate methods for comprehensive drought tolerance evaluation and selection of resilient varieties for water-scarce environments. By identifying drought-tolerant landraces, this research supports sustainable agriculture and food security, aligning with Sustainable Development Goals 2 (Zero Hunger), 13 (Climate Action), and 15 (Life on Land) by promoting climate-resilient crop production and biodiversity conservation.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-28661-8.Keywords: Agroecological region, Water stress, Rice landraces, Drought index, YieldSubject terms: Abiotic, Drought Introduction Rice (Oryza sativa L.) is a staple food crop for majority of the world’s population. It faces increasing threats from drought stress, particularly in Asia, where 90% of global rice is produced and consumed1. In Nepal, rice plays a crucial role in agriculture, both in terms of cultivated area and overall production, and is vital for ensuring food and nutritional security2. Rice is grown across Nepal’s three agroecological regions, Tarai, Mid-Hill, and High Hills (up to 3050 m asl) under both irrigated and rain-fed conditions3. Nepal has 8,389 rice accessions distributed across diverse agroecological regions, each adapted to varying climatic and topographic conditions4. Nepal’s heavy reliance on rain-fed rice production makes it particularly vulnerable to drought, with a significant portion of rice-growing areas experiencing water deficits5. In Nepal, rain-fed rice accounts for 57% of the total rice cultivation area5. The rain-fed varieties are transplanted with the onset of the monsoon season (June to August) across all the agro-ecological regions, which are harvested during the post-monsoon season. Climate change has altered the rainfall patterns and temperature trends in Nepal, i.e., the monsoon onset is occasionally delayed, and short-term erratic rainfall has become a pronounced feature during the season6,7. This rainfall pattern has created direct negative impacts on the productivity of both irrigated and rain-fed rice accessions8,9. The rain-fed rice varieties have gradually adapted to fluctuating water availability over time, and therefore, they might have become more drought-tolerant compared to the irrigated varieties10. Among these, the rain-fed landraces (locally adapted varieties) are often more resilient than the improved varieties as they have evolved in stress-prone environments in the farmers’ fields11. The landraces, possessing inherent genetic diversity, are valuable resources for developing drought-tolerant cultivars12. In Nepal, over 2500 indigenous landraces are anticipated; however, many of these remain uncharacterized in terms of their agromorphological traits and biochemical properties4,13. Although efforts have been made to introduce drought-tolerant rice genotypes, the adoption of improved varieties has led to the extinction of highly valuable landraces without their characterization. Identifying the drought tolerance in crops like rice is difficult due to its complex heritability14. Nevertheless, some statistical parameters, such as drought indices, are being used to select the genotypes with high yield and drought tolerance14,15. Several drought tolerance indices – including tolerance level (TOL), mean productivity (MP), geometric mean productivity (GMP), stress tolerance index (STI), and stress susceptibility index (SSI) have been proposed to assess and compare the resilience of rice varieties under water-limited conditions5,14. Considering such indices and integrating them with multivariate analysis enables the accurate identification of superior genotypes under drought conditions across diverse lines and varieties16,17. The purpose of this study is to evaluate the drought tolerance of twelve rice landraces from three distinct agro-climatic zones (Tarai, Inner-Tarai, and Mid-Hill) from Far-west Nepal. The study primarily focused on the yield components and the indices derived from them, and therefore, the morphological and physiological responses are not included. By employing a combination of yield measurements, drought tolerance indices, and multivariate analyses, this study aimed to: (1) quantify the impact of drought stress on yield performance (Yield reduction), (2) assess the effectiveness of various drought tolerance indices in differentiating landrace responses, and (3) identify landraces with superior drought tolerance ability. Materials and methods Study sites and climate This study was conducted across three distinct agro-climatic zones in Far-West Nepal: Tarai, Inner-Tarai, and Mid-Hill (Fig. 1, Supplementary Figure S1). Each zone possesses unique climatic and topographical characteristics, influencing rice cultivation practices and drought stress exposure. The research stations in Tarai, Inner-Tarai, and Mid-Hill zones are located at elevations of approximately 180 m, 353 m, and 1550 m asl., respectively. Climatically, both the Tarai and Inner-Tarai represent subtropical zones, whereas the Mid-Hill region is characterized by a lower temperate climate with comparatively cooler conditions. The mean annual temperature recorded in Tarai is 27.1 °C and the annual rainfall is 1593 mm. In Inner-Tarai, the mean annual temperature is 20.8 °C with rainfall of 1523 mm. Mid-Hill experiences a mean temperature of 16.9 °C and rainfall of 1156 mm23 (Supplementary Figure S3). Study area maps and experimental station locations. (A) Nepal showing Far-west region. (B) Farwest Nepal. (C) Dadeldhura district, showing the experimental station in Parshuram Municipality, (D) Baitadi District, showing the experimental station in Patan Municipality. (E) Kanchanpur district, showing the experimental station in Belauri Municipality. Rice landraces Twelve rice landraces, four from each agro-climatic zone, were selected for the experiment based on preliminary surveys and local knowledge regarding drought tolerance characteristics. Landraces included Sauthyari, Ghiupuri, Lalchand, and Anjana from Kanchanpur (Tarai); Chiudi, Jhini, Batebudo, and Shanti from Dadeldhura (Inner-Tarai); and Chamade, Jhumke, Temase, and Rato marso from Baitadi (Mid-hill). Seeds were collected from local farmers; they employed traditional seed storage methods. Experimental design Seedlings were raised in nursery beds prepared using traditional methods and transplanted into experimental plots of size 1.5 m × 2 m within rainout shelter (Supplementary Figure S2). Each of the landraces was subjected to three treatments as (i) control: the plots were continuously irrigated since the seedlings transplanted maintaining full saturation of soil moisture, (ii) intermittent drought: the plots were irrigated intermittently, i.e., irrigation was done when 50% plants wilt and the soil moisture decreases up to 25–35% (iii) complete drought: the plots were irrigated for 15 days after seedling transplantation and then irrigation was stopped, where the moisture remained < 20%. Soil moisture was measured using digital sensors (model: WTPH01803, probe length 20.5 cm, 35*8*6 cm) at randomly selected locations within each plot. The sensor was inserted to a depth of 10–15 cm and readings were recorded twice per week. Each treatment had five replications, resulting in a total of 180 plots. The experiment was conducted in 2022, and the same experiment was repeated in 2023. Yield and tolerance indices calculation Yield data were collected from each plot, and yield reduction percentages were calculated for intermittent and complete drought treatments compared to the control. Various drought tolerance indices, including mean productivity (MP), geometric mean productivity (GMP), yield index (YI), stress tolerance index (STI), tolerance index (TOL), drought resistance index (DRI), and modified stress tolerance index (MSTI). The following equations were used to calculate indices. Mean Productivity (MP) = Hossain et al.24 Geometric mean productivity (GMP) = Fernandez25. Yield index (YI) = Gavuzzi et al.26 Stress tolerance index (STI) = Fernandez25. Drought resistance index (DRI) = Lan27. Tolerance index (TOI) = Rosielle and Hamblin28. Modified stress tolerance index (MSTI) = Farshadfar and Sutka29. Where, Yc = yield of test variety in control treatment. Yd = Yeid of test variety in drought treatment. YDs = Average yield of all the varieties in drought treatment. Ys = Average yield of all the varieties in control treatment. Data collection and analysis To assess the relationship between yield loss and the predictors (treatment and variety), we employed a generalized linear model (GLM) with a Gaussian error distribution and an identity link function, fitted using the gam() function from the mgcv package in R. The model evaluated the yield loss along categorical variables (treatment and variety) and did not incorporate any smoothing terms. Although the gam() function supports generalized additive modeling with automatic selection of smoothing parameters via Generalized Cross Validation (GCV), these features were not utilized in our analysis. Diagnostic checking using the gam.check() function confirmed the absence of smoothing terms. Consequently, the model is equivalent to a standard linear model. Two-way Analysis of Variance (ANOVA) was performed to observe yield loss variation across locations and varieties. Since no significant variation was detected among the locations, but a significant effect was observed among varieties, one-way ANOVA was applied to evaluate the variation in drought indices among the varieties. Principal component analysis (PCA) and hierarchical cluster analysis were performed to explore relationships among landraces and their responses to drought stress. All statistical analyses were done in RStudio (R Core Team, 2024). Results and discussion The average yield reduction in the landraces ranges from 1.92 to 30% (Fig. 2). A larger variation in yield reduction is found within the treatments for the same variety; therefore, no significant difference in reduction of yield is found between the two drought treatments (Table 1; Fig. 2). Two-way analysis of variance shows that yield reduction varied significantly among landraces; however, there is no significant variation across the agroecological regions, treatment types, or their interactions (Table 1). Generalized Linear Model (GLM) shows that landraces such as Chamade, Ratomarso, and Jhumke from Mid-Hill; Jhini and Shanti from Inner-Tarai; and Lalchand and Anjana from Tarai exhibited significantly greater yield loss under drought conditions compared to the baseline mean (Table 2). Landraces like Temase, Chiudi, Batebudo, Sauthyari and Ghiupuri have lower yield reduction than the baseline mean of the studied system i.e., 15.9% (Table 2; Fig. 2). In the intermittent drought, there is no yield reduction in Batebudo and Chiudi landraces (Fig. 2) and the highest reduction was observed for Anjana (24.7%) (Fig. 2). Similarly, for the complete drought, there was no reduction in the mean yield in Ghiupuri landrace whereas, the highest reduction was reported for the landrace Shanti (36.7%). Yield loss (%) across rice varieties and drought treatments. Black points represent predicted yield loss with 95% confidence intervals based on generalized linear models fitted using the gam() function (mgcv package). The dashed red line indicates the overall mean yield loss. Colored points (blue for Intermittent and red for Drought) show treatment-specific observed mean yield loss with standard errors. Table 1Results of Two-way analysis of variance (ANOVA) for yield loss, showing the effects of Location, Variety, Treatment, and their interactions.DfSum SqMean SqF valuePr (> F)Location2813406.40.4470.6403Variety919,0522116.92.327 0.0161* Treatment1930929.71.0220.3132Location × Treatment226231311.71.4420.2388Variety × Treatment9968107.50.1180.9992Residuals216196,519909.8Significant values are in [bold] (* p < 0.05). Results of Two-way analysis of variance (ANOVA) for yield loss, showing the effects of Location, Variety, Treatment, and their interactions. Significant values are in [bold] (* p < 0.05). Table 2Generalized linear model (GLM) results showing the effect of rice variety on yield loss, including estimates, standard errors, t-values, and significance levels (* p < 0.05, ** p < 0.01 and *** p < 0.001).VarietyEstimateStd. Errort valuePr (>|t|)Chamade17.5016.642.6360.0089**Temase8.5496.641.2880.1991Rato marso19.8646.642.9920.0031**Jhumke21.9286.643.3020.0011**Jhini28.7976.644.337< 0.001***Shanti29.9746.644.514< 0.001***Chiudi8.2176.641.2380.2172Batebudo2.7186.640.4090.6827Lalchand16.5696.642.4950.0133*Sauthyari10.7676.641.6220.1063Anjana23.9816.643.612< 0.001***Ghiupuri1.9256.640.290.7722 Generalized linear model (GLM) results showing the effect of rice variety on yield loss, including estimates, standard errors, t-values, and significance levels (* p < 0.05, ** p < 0.01 and *** p < 0.001). The lack of significant differences in yield reduction between intermittent and complete drought treatments suggests that landraces are resilient and employ different strategies to cope with water stresses. However, significant difference observed across drought indices such as MP, GMP, YI, STI, and MSTI indicates that these traits contribute differentially to overall drought tolerance. The selection based on drought tolerance indices can be an efficient tool in the identification of superior drought-tolerant cultivars with higher yield and stability15,18. The landraces Ghiupuri, Lalchand, and Sauthyari consistently exhibited higher values for these indices, suggesting their superior ability to maintain yield under drought stress. Due to the complexity of the dataset, yield related traits were analyzed using drought indices in this study by excluding morpho-physiological traits. This focused analysis limits a more holistic understanding of the drought response mechanism across the studied landraces. There are numerous morpho-physiological traits in rice19 and often they are interrelated, which makes it difficult to determine whether such traits individually or interactively influence rice yield negatively or positively. Therefore, this study focuses on yield-related traits under droughts that is expected to provide a more straightforward way to assess the response of the landraces against drought in terms of actual productivity. All the drought indices show significant differences among the rice landraces both in the intermittent drought and complete drought (Tables 3 and 4) except the TOL. Generally, MP, GMP, YI were higher for the landraces Ghiupuri, Lalchand and Sauthyari in both intermittent and complete droughts. Based on the MSTI, the landraces Sauthyari and Ghiupuri showed higher tolerance as compared to other varieties. Table 3Drought tolerance indices for the intermittent drought.LocationVarietyMPGMPYISTIDRITOLMSTIBaitadi(Mid-Hill)Chamade0.66 ± 0.19 c0.91 ± 0.12 f0.32 ± 0.14 d0.44 ± 0.18 c0.23 ± 0.12 e0.12 ± 0.22 a0.07 ± 0.06 cJhumke0.80 ± 0.10 c1.02 ± 0.08 ef0.42 ± 0.18 d0.55 ± 0.20 c0.33 ± 0.28 de0.09 ± 0.16 a0.14 ± 0.19 cRato marso1.03 ± 0.64 c1.11 ± 0.33 def0.47 ± 0.18 d0.61 ± 0.23 c0.41 ± 0.37 de0.15 ± 0.38 a0.18 ± 0.19 cTemase1.08 ± 0.59 c1.15 ± 0.35 def0.56 ± 0.34 d0.65 ± 0.24 c0.51 ± 0.45 cde0.02 ± 0.28 a0.36 ± 0.44 bcDadeldhura(Inner-Tarai)Batebudo2.49 ± 1.21 ab1.76 ± 0.40 abc1.39 ± 0.71 ab1.86 ± 1.21 ab1.10 ± 0.55 abcd0.19 ± 0.57 a6.69 ± 8.25 abcChiudi1.49 ± 0.54 bc1.38 ± 0.24 cde0.81 ± 0.37 bcd1.09 ± 0.60 bc0.68 ± 0.47 bcde0.13 ± 0.41 a1.18 ± 1.45 bcJhini1.57 ± 0.47 bc1.39 ± 0.18 cde0.68 ± 0.21 cd1.04 ± 0.45 c0.46 ± 0.33 cde0.38 ± 0.53 a0.59 ± 0.50 baShanti1.75 ± 0.42 bc1.48 ± 0.18 bcd0.81 ± 0.14 bcd1.12 ± 0.20 bc0.52 ± 0.19 cde0.31 ± 0.24 a0.77 ± 0.34 bcKanchanpur(Tarai)Anjana2.75 ± 0.32 ab1.87 ± 0.09 ab1.30 ± 0.28 abc1.83 ± 0.54 ab0.82 ± 0.21 abcde0.52 ± 0.27 a3.59 ± 2.35 abcGhiupuri3.27 ± 1.40 a2.04 ± 0.49 a1.75 ± 0.84 a2.04 ± 0.49 a1.58 ± 1.34 a0.00 ± 0.83 a9.05 ± 10.62 aLalchand3.47 ± 1.41 a2.08 ± 0.38 a1.68 ± 0.45 a2.18 ± 0.50 a1.24 ± 0.54 abc0.47 ± 0.77 a7.17 ± 5.40 abSauthyari3.73 ± 1.57 a2.15 ± 0.44 a1.80 ± 0.48 a2.29 ± 0.44 a1.34 ± 0.54 ab0.50 ± 0.81 a8.52 ± 5.83 aOne-way ANOVAF15.0820.9117.2717.646.731.346.22p value<0.001 ***<0.001 ***<0.001 ***<0.001 ***<0.001 ***0.209<0.001 ***MP: mean productivity, GMP: geometric mean productivity, YI: yield index, STI: stress tolerance index, DRI: drought resistance index, TOL: tolerance index, MSTI: modified stress tolerance index.Different letters in each column indicate significant differences in measured parameter between varieties. Results of one-way ANOVA for each index are presented, with F-values and significance levels, *p < 0.05, ** p < 0.01, *** p < 0.001, ns non-significant. Drought tolerance indices for the intermittent drought. Baitadi (Mid-Hill) Dadeldhura (Inner-Tarai) Kanchanpur (Tarai) MP: mean productivity, GMP: geometric mean productivity, YI: yield index, STI: stress tolerance index, DRI: drought resistance index, TOL: tolerance index, MSTI: modified stress tolerance index. Different letters in each column indicate significant differences in measured parameter between varieties. Results of one-way ANOVA for each index are presented, with F-values and significance levels, *p < 0.05, ** p < 0.01, *** p < 0.001, ns non-significant. Table 4Drought tolerance indices for the complete drought.LocationVarietyMPGMPYISTIDRITOLMSTIBaitadi(Mid-Hill)Chamade0.66 ± 0.18 d0.91 ± 0.12 f0.33 ± 0.15 e0.44 ± 0.18 c0.23 ± 0.14 d0.12 ± 0.23 ab0.07 ± 0.07 dJhumke0.78 ± 0.07 d0.99 ± 0.06 ef0.38 ± 0.14 e0.52 ± 0.15 c0.24 ± 0.12 d0.15 ± 0.15 ab0.10 ± 0.08 dRato marso1.01 ± 0.65 d1.09 ± 0.34 def0.45 ± 0.20 e0.59 ± 0.24 c0.38 ± 0.39 cd0.17 ± 0.38 ab0.18 ± 0.20 dTemase1.08 ± 0.60 d1.15 ± 0.35 def0.57 ± 0.32 e0.65 ± 0.24 c0.50 ± 0.45 cd0.02 ± 0.27 ab0.36 ± 0.44 cdDadeldhura(Inner-Tarai)Batebudo2.39 ± 1.05 bc1.72 ± 0.33 abc1.24 ± 0.53 bcd1.73 ± 1.02 ab0.91 ± 0.51 bcd0.39 ± 0.88 ab4.11 ± 4.30 bcdChiudi1.49 ± 0.51 cd1.38 ± 0.22 cde0.83 ± 0.34 cde1.08 ± 0.57 bc0.67 ± 0.43 bcd0.14 ± 0.41 ab1.13 ± 1.11 cdJhini1.54 ± 0.48 cd1.37 ± 0.19 cde0.64 ± 0.18 e1.00 ± 0.43 c0.39 ± 0.33 cd0.45 ± 0.52 ab0.49 ± 0.46 cdShanti1.73 ± 0.44 cd1.47 ± 0.20 bcd0.80 ± 0.19 de1.09 ± 0.22 bc0.49 ± 0.22 bcd0.34 ± 0.24 ab0.78 ± 0.51 cdKanchanpur(Tarai)Anjana2.77 ± 0.32 abc1.88 ± 0.10 ab1.39 ± 0.41 bc1.85 ± 0.56 a0.88 ± 0.38 bcd0.48 ± 0.35 ab4.43 ± 3.60 bcdGhiupuri3.37 ± 1.30 ab2.09 ± 0.43 a2.00 ± 0.70 a2.17 ± 0.41 a1.91 ± 1.21 a-0.19 ± 0.82 b10.68 ± 9.79 aLalchand3.40 ± 1.47 ab2.04 ± 0.41 a1.59 ± 0.45 ab2.09 ± 0.52 a1.02 ± 0.44 bc0.61 ± 0.66 a6.32 ± 5.58 abcSauthyari3.69 ± 1.60 a2.13 ± 0.46 a1.78 ± 0.55 ab2.24 ± 0.47 a1.26 ± 0.67 ab0.57 ± 0.80 ab8.44 ± 6.15 abOne-way ANOVAF-value15.7322.2122.1420.798.952.318.231p value<0.001 ***<0.001 ***<0.001 ***<0.001 ***<0.001 ***0.032*<0.001 ***MP: mean productivity, GMP: geometric mean productivity, YI: yield index, STI: stress tolerance index, DRI: drought resistance index, TOL: tolerance index, MSTI: modified stress tolerance index.Different letters in each column indicate significant differences in measured parameter between varieties. Results of one-way ANOVA for each index are presented, with F-values and significance levels, *p < 0.05, ** p < 0.01, *** p < 0.001, ns non-significant. Drought tolerance indices for the complete drought. Baitadi (Mid-Hill) Dadeldhura (Inner-Tarai) Kanchanpur (Tarai) MP: mean productivity, GMP: geometric mean productivity, YI: yield index, STI: stress tolerance index, DRI: drought resistance index, TOL: tolerance index, MSTI: modified stress tolerance index. Different letters in each column indicate significant differences in measured parameter between varieties. Results of one-way ANOVA for each index are presented, with F-values and significance levels, *p < 0.05, ** p < 0.01, *** p < 0.001, ns non-significant. The yield loss was positively correlated with the TOL (r = 0.7) and it was negatively correlated with the DRI (r = -6) (Fig. 3). The TOL did not significantly differ among the varieties in the intermittent drought (Table 3). In the complete drought, Ghiupuri landrace had the lowest (-0.19) and Lalchand had the highest (0.61) values (Table 4). The DRI was the highest in the Ghiupuri (1.58) and lowest in the Chamade (0.23) in the intermittent drought (Table 3). In the complete drought also, the lowest value was found in the landrace Chamade (0.23) and Jhumke (0.24) and the highest value was recorded for the Ghiupuri (1.91) (Table 4). A positive correlation between yield loss and TOL suggests that the landraces with higher TOL experience higher yield reductions. The lower value of TOL indicates the higher stress-tolerant ability of a given cultivar5. Conversely, the negative correlation between yield loss and DRI implies that a higher DRI indicates ability of varieties to minimize yield losses. Pearson correlation coefficients between drought indices and rice yield. Positive and negative correlations are indicated by the color gradient, with the strength of the correlation reflected by the intensity of the color. (A) Under intermittent drought condition, (B) Under complete drought condition. There was a positive correlation between the DRI with the other parameters like YI, MP, GMP and MSTI, and this DRI is negatively correlated with the yield loss, so in our study, the value of the DRI is likely to explain the drought-tolerant genotypes. Previous studies Adhakari et al.18, Mau et al.20 and Raman et al.21 showed a similar scenario, showing positive correlation between the MP, GMP, YI, MSTI, and DRI. In our study, although there is positive correlation between these parameters, only DRI has shown highly negative correlation with the yield losses. These findings emphasize the importance of considering multiple drought tolerance traits when selecting or developing resilient varieties. The genotypes with high DRI and STI and low TOL can be identified as drought-tolerant genotypes5,22. In this study, Temase from Mid-Hill, Batebudo from Innter-Tarai and Ghiupuri from Tarai were found to have higher DRI (Tables 3 and 4). The principal component analysis (PCA) showed that the other varieties are more or less correlated with each other (Fig. 4). This correlation was also further proved by the positive correlation between most of the indices (Fig. 4). The PCA graph revealed that TOL and Yield loss is negatively correlated with other parameters and showed a distinct pattern of response among the varieties under both intermittent and complete drought conditions. The landrace Anjana exhibited a strong association with TOL under intermittent drought. Conversely, Jhini and Shanti landraces were primarily characterized by higher yield losses. Principal Component Analysis (PCA) biplot depicting the relationship between rice varieties (represented by points) and drought indices (indicated by arrows). The arrows represent the contribution of each drought index to the principal components, with longer arrows indicating stronger influence. The proximity of the variety points illustrates their similarity in yield loss patterns and affinity to drought indices. (A) Under intermittent drought condition, (B) Under complete drought condition. Hierarchical cluster analysis further classified the varieties into groups based on their similarities in response to drought. Six distinct clusters were found under intermittent drought, while four clusters were identified under complete drought (Fig. 5). In the intermittent drought scenario, Temase and Chiudi, Jhini and Shanti, Batebudo and Ghiupuri, and Lalchand and Sauthyari landraces formed respective groups. The landraces Chamade, Jhumke, and Rato marso clustered together; however, Anjana remained distinct from other varieties (Fig. 5). Under complete drought conditions, Ghiupuri landrace formed an independent cluster, while Jhini and Shanti landraces are associated together. Chamade, Jhumke, Rato marso, Temase, and Chiudi clustered together in one group. Similarly, Batebudo, Sauthyari, Anjana, and Lalchand formed a final group (Fig. 5). Hierarchical clustering of rice varieties based on principal component analysis (PCA) of drought indices and yield loss. The dendrogram illustrates the grouping of varieties with similar responses to drought conditions. (A) Under intermittent drought condition, (B) Under complete drought condition. The STI, YI, MP, DRI, GMP have positive correlation with the PCA1 and the yield loss and the TOL has positive correlation with the PCA2. Overall, the PCA 1 and 2 explained 95.9% of the variance among the indices in the intermittent drought and the 95.8% variance in the complete drought (Fig. 4). Similar higher variation was also explained by the PCA in the study carried out by Mau et al.20 in the selected upland rice genotypes. They also identified that the STI, GMP, and YI as useful indices for selecting drought-tolerant rice genotypes20. The clustering patterns observed in the dendrogram were largely consistent with the groupings visualized in the PCA biplots, supporting the validity of the identified clusters. The cophenetic correlation coefficient was 0.82 for intermittent drought and 0.69 for complete drought. Distinct clustering of Anjana landrace under both drought conditions indicates its unique tolerance characteristics and higher association with TOL. From this result, we can infer that Anjana is more drought-susceptible as compared to other varieties. Similarly, the consistent clustering of Jhini and Shanti landraces across treatments underscores their shared susceptibility to drought stress. Identified drought-tolerant landraces, like Ghiupuri and Sauthyari could serve as valuable genetic resources for developing improved varieties with enhanced resilience to water scarcity. There was a lack of significant yield reduction differences among landraces in drought conditions, but the result shows varying drought tolerance indices among the landraces, which reveals limitations in relying solely on these indices for assessing drought resistance. Indices may be biased towards high-yielding varieties, even if they exhibit similar resilience under drought. Therefore, caution is needed when using these indices for screening diverse landraces. Our study suggests combining multivariate methods with drought tolerance indices for a more comprehensive evaluation of drought resistance in genetically diverse germplasm. This study also proves the resilience of selected rice landraces to drought. Thus, identifying, documenting and characterizing such types of landraces is recommended. Additionally, exploring the potential of marker-assisted selection and other breeding techniques can enhance the development of drought-adapted rice varieties, contributing to food security in regions prone to water scarcity. These efforts support the Sustainable Development Goals (SDGs) of Zero Hunger and Climate Action. SDG 2 aims to reduce hunger and improve food security through promoting resilient agricultural practices. SDG 13 focuses on climate adaptation strategies to reduce the effects of drought on crop production. Further research involving a wider range of genotypes and field-based evaluations is needed to validate the findings and assess the stability of drought tolerance across diverse environments. Investigating the genetic basis of drought tolerance in these landraces would provide valuable information for developing molecular markers for breeding purposes. Moreover, exploring the interaction between drought stress and other environmental factors, such as temperature and soil fertility, would provide a more comprehensive understanding of rice responses to complex abiotic stress scenarios. Conclusion Based on our experiment, we concluded that drought reduces yield in all varieties of landraces. The average yield reduction ranges from 1.92% in the landrace Ghiupuri to 30% in Shanti. The variation in yield loss differs according to variety. Using the Generalized Linear Model (GLM), we found that landraces such as Chamade, Ratomarso, and Jhumke from the Mid-Hill; Jhini and Shanti from the Inner-Terai; and Lalchand and Anjana from the Tarai exhibit significantly higher yield loss. In contrast, varieties such as Temase, Chiudi, Batebudo, Sauthyari, and Ghiupuri show less yield reduction than the baseline mean of the studied system (15.9%). Based on drought indices and clustering analysis, we identified Temase from the Mid-Hill, Batebudo from the Inner-Terai, and Ghiupuri from the Tarai as the most drought-tolerant landraces in this study. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements This work was supported by the University Grants Commission, Nepal (Grant No. CRG-77/78- S&T-2). Authors are highly thankful to Mr. Dammar Singh Thwal (Principal, Bhuwaneswari Secondary School, Dadeldhura); Mr. Prem Bahadur Thapa (Principal, Shanti Secondary School, Kanchanpur); Mr. Sagar Bista, Mr. Prasad Singh Bohara, Mr. Keshab Raj Bhatt, Mr. Daljit Bhatt, Mr. Uddhav Raj Bhatt, Mr. Khemraj Bhattrai, and Mr. Dev Raj Bhattrai for their support during the experiment. Author contributions N.L., U.D. and S.B.: Investigation, data curation, data analyses, original manuscript preparation. G.P., L.B.T. , C.P.P. and R.K.P.Y.: Conceptualization, methodology, reviewing and editing, project administration, funding acquisition, supervision. Data availability The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.