Theoretical and Applied Genetics (2023) 136:18 https://doi.org/10.1007/s00122-023-04260-x ORIGINAL ARTICLE 50 years of rice breeding in Bangladesh: genetic yield trends Niaz Md. Farhat Rahman1 · Waqas Ahmed Malik2  · Md. Shahjahan Kabir1 · Md. Azizul Baten3 · Md. Ismail Hossain1 · Debi Narayan Rudra Paul1 · Rokib Ahmed1 · Partha Sarathi Biswas1 · Md. Chhiddikur Rahman1 · Md. Sazzadur Rahman1 · Khandakar Md. Iftekharuddaula1 · Steffen Hadasch2 · Paul Schmidt2 · Md. Rafiqul Islam4 · Md. Akhlasur Rahman1 · Gary N. Atlin5 · Hans‑Peter Piepho2 Received: 13 June 2022 / Accepted: 18 November 2022 / Published online: 21 January 2023 © The Author(s) 2023 Abstract To assess the efficiency of genetic improvement programs, it is essential to assess the genetic trend in long-term data. The present study estimates the genetic trends for grain yield of rice varieties released between 1970 and 2020 by the Bangladesh Rice Research Institute. The yield of the varieties was assessed from 2001–2002 to 2020–2021 in multi-locations trials. In such a series of trials, yield may increase over time due to (i) genetic improvement (genetic trend) and (ii) improved man- agement or favorable climate change (agronomic/non-genetic trend). In both the winter and monsoon seasons, we observed positive genetic and non-genetic trends. The annual genetic trend for grain yield in both winter and monsoon rice varie- ties was 0.01 t ha−1, while the non-genetic trend for both seasons was 0.02 t ha−1, corresponding to yearly genetic gains of 0.28% and 0.18% in winter and monsoon seasons, respectively. The overall percentage yield change from 1970 until 2020 for winter rice was 40.96%, of which 13.91% was genetic trend and 27.05% was non-genetic. For the monsoon season, the overall percentage change from 1973 until 2020 was 38.39%, of which genetic and non-genetic increases were 8.36% and 30.03%, respectively. Overall, the contribution of non-genetic trend is larger than genetic trend both for winter and monsoon seasons. These results suggest that limited progress has been made in improving yield in Bangladeshi rice breeding programs over the last 50 years. Breeding programs need to be modernized to deliver sufficient genetic gains in the future to sustain Bangladeshi food security. Introduction ranks fourth among countries globally in rice consump- tion, with an annual per capita availability of rice of about Rice and food security are synonymous in Bangladesh (Brol- 213.5 kg (FPMU Database 2020). ley 2015). For more than 166.5 million people in Bangla- Rice is Bangladesh’s largest crop, occupying about 76% desh, rice is the main staple food, contributing 97% of total of the total cropped area (15.44 million hectares), of which, food grain production (BBS 2019; FPMU 2020). Bangladesh in 2019–20, about 88% was planted to modern varieties, with traditional landrace varieties covering 12% (BBS 2019). The current intake of rice is about 367 g c apita−1  day−1, which Communicated by Daniela Bustos-Korts. provides approximately 60% of total calories and 50% of total protein for adults (HIES 2016). Nearly 48% of rural * Waqas Ahmed Malik w.malik@uni-hohenheim.de workers are directly or indirectly involved in rice production for their livelihood. Rice is grown on more than 13 million 1 Bangladesh Rice Research Institute (BRRI), Gazipur, farms on approximately 11.77 million hectares, summed Bangladesh over the winter (dry) and monsoon (wet) seasons (DAE 2 Institute of Crop Science, Biostatistics Unit, University 2020). The contribution of rice to the value of the crop sub- of Hohenheim, Fruwirthstrasse 23, 70599 Stuttgart, Germany sector is about 70% (Mottaleb and Mishra 2016). 3 Shahjalal University of Science and Technology, Sylhet, Since the Green Revolution of the 1960s, rice output Bangladesh quadrupled in Bangladesh, increasing from 9.67 million 4 International Rice Research Institute (IRRI), Dhaka, tons in 1971 to 38.70 million tons in 2019, with national Bangladesh average yield more than doubling, from 1.50 to 3.29 t h a−1 5 Bill & Melinda Gates Foundation, Seattle, USA Vol.:(012 3456789) 18 Page 2 of 13 Theoretical and Applied Genetics (2023) 136:18 during the same period (BBS 1972; DAE 2020). Key factors use historical yield data from trials of BRRI’s released rice underlying these improvements were the government’s sup- varieties to assess the genetic progress achieved due to rice port for mechanization and irrigation, controlling fuel and breeding. The key objective of this study was to estimate the fertilizer prices, improved credit policies, well-organized long-term rate of genetic improvement delivered by BRRI fertilizer supply, increased quality seed supply by public and breeding efforts. private sectors, expansion of the high-yield irrigated winter rice system, and genetic improvements for both favorable and stress-prone environments. With the steady yield growth Materials and methods of recent years, Bangladesh has been self-sufficient in rice production since 2012 (Bell et al. 2015). Long‑term multi‑environment trials (BRRI rice In 2050, the population of Bangladesh is predicted to stability trials) reach 215.4 million. An estimated 44.6 million tons (MT) of milled rice will be required to feed the increasing popula- Rice is cultivated in Bangladesh in the monsoon (roughly tion of the country (Kabir et al. 2015). At the same time, rice June-October) and winter (November -April) seasons. The production is hampered by decreasing arable land, increas- seasons differ greatly in growing conditions and conse- ing climate vulnerabilities like drought, salinity, flood, heat quently require different varieties, however, both seasons and cold, tidal submergence, water stagnation, and seawater are equally important. Winter rice, grown in the dry season, intrusion, all of which pose long-term threats to the coun- is irrigated. During the winter, evaporative demand and solar try’s agricultural sector and can hinder food security. Cli- radiation are high, and relative humidity and disease pres- mate change is a great challenge for sustaining Bangladeshi sure are low. However, rice grown in the monsoon season rice production and future food security (World Bank 2016). is rainfed (low solar radiation, high relative humidity, high Over the last 50 years, the Bangladesh Rice Research pest, and disease pressure). BRRI operates large, separate Institute (BRRI) has developed and released new varieties breeding pipelines for each season. BRRI has been conduct- and conducted variety trials to assess their performance. ing multi-location rice variety trials since 2001 for both sea- Within this period, yield has increased due to improved sons. One of the major breeding objectives in connection management (agronomic trend) and plant breeding efforts with farmers’ needs is that varieties developed for favorable (genetic trend) (Masuka et  al. 2017; Rife et  al. 2019). environments should also have the capabilities of perform- Genetic gain can be defined as the rate of increase in per- ing well in less favorable environments. Similarly, stress- formance over a period of time that is achieved through tolerant varieties should also perform well under non-stress artificial selection and breeding programs (Xu et al. 2017). conditions. For example, a flash flood tolerant variety should Over the last 10 to 15 years, there have been several stud- exhibit high yield whenever there is no flash flooding in a ies assessing long-term genetic gains in different crop spe- particular area of a farmer’s land. BRRI has 11 regional cies in different countries (Kumar et al. 2021; Laidig et al. stations, all having favorable lands where the trials were 2014, 2017; Rife et al. 2019). To quantify the increase in performed. Conducting trials under on-station conditions has performance with time that is due to breeding, agronomic generated good quality data. All trials were managed for and genetic gain need to be dissected, which can be done optimum yield, without any stresses deliberately imposed, using models that regress varietal performance for traits of and under generally favorable conditions (Supplementary interest on year of release, in trials conducted across loca- Table S1). These trials are called stability trials and com- tions and years. In “ERA” studies, all varieties in the series prise a fairly consistent set of varieties over the years (strong are included in the same trials (Duvick et al. 2004; Duvick carry-over from year to year), with newly released varieties 2005). Other models allow genetic trend to be estimated added each year and hardly any varieties dropping out. The from ongoing varietal performance trials that do not test all varieties included in the trials were developed by BRRI and varieties in all years, but that retain a sufficient number of representatives of all varieties that are grown in Bangladesh common checks from year to year to allow genetic and non- for both seasons. Thus, these trials form an ideal basis for genetic trends to be estimated (Mackay et al. 2011; Piepho assessing long-term genetic trends for yield achieved by et al. 2014). It is to be expected that breeding and varietal BRRI breeding programs. This paper reports genetic trend selection are not the only relevant factors driving crop yields in favorable environments of varieties developed both for gains when comparing cultivar performance and genetic favorable and less favorable environments. advances across several locations and years (Ahrends et al. The monsoon rice dataset contains the results of trials 2018; Muralidharan et al. 2019). through 2020, while the winter rice dataset contains results Modern rice research in Bangladesh started in the 1960s, through the growing season of 2020–2021. For each season, but long-term genetic gain of rice in the country has not yet trials were conducted each year at nine locations for winter been quantified. Therefore, the objective of this study was to rice and eight locations for monsoon rice (Fig. 1). 1 3 Theoretical and Applied Genetics (2023) 136:18 Page 3 of 13 18 Fig. 1 Multi-location rice vari- ety trials from 2001 till 2020 Rice varieties Experiment and trial management From 1970 to 2020, BRRI released 41 varieties for the The experiments were set up as randomized complete block winter season and 45 varieties for the monsoon season. designs (RCBD) with three replicates in each location. For The first variety, BR1, was released in 1970 and the lat- all varieties, breeder’s seed was used. In each location, est variety, BRRI dhan99, was released in 2020. A list 40-day-old seedlings for winter rice and 25-day-old seed- of released varieties for both seasons is given in Table 1 lings for monsoon rice were transplanted. All transplanting (for details see Supplementary Table S1). As new varie- was conducted manually, and any damaged seedlings were ties were added to the trials each year and old ones were retransplanted by the same aged seedlings within 3–7 days retained, the number of tested varieties increased over the of the first transplanting (see Table 2). The harvested area years. The genotype-year combinations for both seasons was 6 m2 out of 10 m 2 per plot. The spacing between rows are shown in Fig. 2. and between plants within row was 20 cm and 20 cm with 1 3 18 Page 4 of 13 Theoretical and Applied Genetics (2023) 136:18 Table 1 List of rice varieties released between 1971 and 2020 in Bangladesh. Forty-one varieties released for winter season and 45 varieties released for monsoon season Year of release Variety Year of release Variety Winter season Monson season 1970 BR1 1973 BR3 1971 BR2 1975 BR4 1973 BR3 1976 BR5 1977 BR6, BR7, BR8 1980 BR10, BR11 1978 BR9 1988 BR22, BR23 1983 BR12, BR14, BR15, BR16 1992 BR25 1985 BR17, BR18, BR19 1994 BRRI dhan30, BRRI dhan31, BRRI dhan32 1994 BRRI dhan28, BRRI dhan29 1997 BRRI dhan33, BRRI dhan34 1998 BRRI dhan35, BRRI dhan36 1998 BRRI dhan37, BRRI dhan38 2005 BRRI dhan45 1999 BRRI dhan39 2007 BRRI dhan47 2003 BRRI dhan40, BRRI dhan41 2008 BRRI dhan50 2005 BRRI dhan44 2011 BRRI dhan55 2007 BRRI dhan46 2012 BRRI dhan58 2008 BRRI dhan49 2013 BRRI dhan59, BRRI dhan60, BRRI dhan61 2010 BRRI dhan51, BRRI dhan52, BRRI dhan53, BRRI dhan54 2014 BRRI dhan63, BRRI dhan64, BRRI dhan67, 2011 BRRI dhan56, BRRI dhan57 BRRI dhan68, BRRI dhan69 2015 BRRI dhan74 2013 BRRI dhan62 2017 BRRI dhan81, BRRI dhan84, BRRI dhan86 2014 BRRI dhan66 2018 BRRI dhan88, BRRI dhan89 2015 BRRI dhan70, BRRI dhan71, BRRI dhan72, BRRI dhan73 2019 BRRI dhan92 2016 BRRI dhan75, BRRI dhan76, BRRI dhan77, BRRI dhan78 2020 BRRI dhan96, BRRI dhan97, BRRI dhan99 2017 BRRI dhan79, BRRI dhan80 2018 BRRI dhan87 2019 BRRI dhan90, BRRI dhan91, BRRI dhan93, BRRI dhan94, BRRI dhan95 2 or 3 plants hill−1. There were 25 hills per m2, hence 6 m2 However, fitting such a model can become computation- contained 150 hills. The fields were irrigated 4 days after ally demanding or even unfeasible due to memory issues. transplanting and a water depth of approximately 5 to 10 cm Therefore, a weighted two-stage analysis was applied here was maintained until 7 days before physiological maturity. to allow for different error variances between year-location Seven days after transplanting, the missing or damaged hills combinations. It is shown in Damesa et al. (2017) that the were retransplanted with the same aged seedlings. Fungi- results of the two-stage analysis are usually very similar to cides and pesticides were also applied to control major dis- those of a single-stage analysis. eases and insects prevailing in the trials. Hand weeding was In the first stage, the linear model (1) was fitted to the data practiced as frequently as needed. The crop was harvested of each year-location (environment) combination to estimate manually and border rows were not harvested. Grain yield the genotype means and their associated standard errors: (t ha−1) data were assessed discarding borders, amounting to a harvested area of about 2.72 m 2 plot−1 and adjusted to a yil =  + bl + gi + eil (1) moisture content of 14%. The trial data are complete for all where yil is the observation of the ith genotype in the lth year-location combinations. block in a given year-location combination,  is a fixed inter- cept, bl is the fixed effect of the lth block, gi is the fixed effect of the ith genotype, and eil is the error associated with y Statistical analysis il which is assumed to be independent and identically normally distributed. Analysis of genetic and non‑genetic trends In the second stage, following Piepho et al. (2014), linear mixed model (2) was fitted to the genotype means computed In multi-environment trials, the analysis should ideally allow in the first stage: the variances to differ between year-location combinations. 1 3 Theoretical and Applied Genetics (2023) 136:18 Page 5 of 13 18 Fig. 2 Classification of genotype-year combinations for winter rice series (left) and monsoon rice series (right). Cell shades of gray indicate number of trials in a year (color figure online) Table 2 Trial information for two rice seasons Season Seeding period Transplanting period Harvest period Fertilizer application (kg/ha) Urea TSP* MoP* Gypsum Zinc Winter rice 15 Nov–30 Nov 25 Dec–10 Jan 25 April–14 May 160, 160, 140 52, 52, 48 88, 88, 80 60, 60, 60 6, 6, 6 Monsoon rice 15 June–30 June 15 Jul–30 Jul 15 Nov–30 Nov 104 32 56 36 *TSP triple superphosphate, MoP muriate of potash y random main effect of the ith genotype, L ijk = + ri + tk + Gi + Lj + Yk + (LY)jk + (GL)ij j is the random (2) main effect of the jth location, Yk is the random main effect + (GY)ik + (GLY)ijk + eijk of the kth year, (LY)jk is the jkth random location-year inter- where y action effect, (GL)ij is the ijth random genotype-location ijk is the mean yield of the ith genotype in the jth location and kth year,  is a fixed intercept,  is the fixed interaction effect, (GY)ik is the ikth random genotype-year slope for genetic trend, ri is the year of release for the ith interaction effect, (GLY)ijk is the ijkth random genotype- variety,  is the fixed slope for non-genetic trend, tk is the location-year interaction effect, and eijk is a random error and trial calendar year corresponding to the kth year, Gi is the assumed independent and identically normally distributed. In model (2), the variance 2 of eijk is assumed to be known eijk 1 3 18 Page 6 of 13 Theoretical and Applied Genetics (2023) 136:18 and equal to the variance (squared standard error) of the season, which are the release year of first variety for that corresponding genotype-location-year mean estimated from season. the analyses of variance of the individual trials computed in the first stage modeling complete blocks as fixed effects Contribution of variance components to the average (Möhring and Piepho 2009, method 2). Variance compo- marginal variance nents of the other random effects are denoted as 2 , 2 , 2 , Y L G 2 , 2 , 2 and 2 . LY GY GL GLY For each variance component of a model, its contribution to the average variance was evaluated by the ratio of a variance Evaluation of trends component to the mean total variance of the data. Due to the two-stage analysis, the second stage is ideally done treating To estimate the genetic trend of yield, we computed the the variances of the genotype means estimated in the first difference between the expected performance of the oldest stage as known in the second stage. With these settings, the variety in the first year of the trial series (2001 for both sea- average marginal variance of the data is sons) and the expected performance of the youngest variety ∑ in the last year of the trial series (2020 in winter season and AMV = 2 + 2 + 2 + 2 + 2 + 2 1 + 2 + 2 , Y L G LY GY GL GLY n eijk 2020–2021 in monsoon season), y1 . Formally, the overall (8) gain is estimated by where n is the number of genotype means estimated in the ∑ ( ) first stage and 1 2 is the mean variance of the genotype y − y  + t n e 1 0 max + rmax −  + tmin + rmin ijk overall gain = = y  + t 0 min + r means. The contribution of a variance component to the min ( ) ( )  tmax − t average marginal variance was computed as min +  rmax − rmin =  + tmin + rmin 2 (3) E contribution2 = 100% (9) E AMV where tmin and tmax are the first and the last calendar year of the trial series. The value of tmin is 2001 and tmax is 2020 for where 2 is one of the terms in AMV . E both seasons. Accordingly, rmin and rmax represent the year of release of the oldest and the youngest varieties, where rmin Heritability is 1970 and 1973 and rmax is 2020 and 2019 for winter and monsoon season, respectively. As the numerator of the total Heritability is often used by plant breeders as a measure of gain is additive, it can be separated into precision of a series of trials (Piepho and Möhring 2007). ( ) Heritability on a variety mean basis can be estimated accord-  rmax − rmin genetic gain = (4) ing to Cullis et al. (2006) by  + tmin + rmin H2 mvd = 1 − 22 (10) ( ) g  tmax − tmin non-genetic gain = (5)  + tmin + rmin where 2 is the genetic variance and mvd is the mean vari- g ance of pairwise differences between estimated genotype The gains per year due to genetic and non-genetic causes effects (BLUPs). Both 2 and mvd were obtained using (2). were estimated as g  Yearly genetic gain = (6)  + t Results min + rmin  The estimated genotype-location-year means from Model Yearly non-genetic gain =  + tmin + r (7) (1) are plotted against the trial year and the release year of min a variety in Figs. 3, 4, 5 and 6. The range of the genotype- where μ is the overall intercept, β and γ are the regression location-year yields of winter rice varieties was between coefficients for genetic and non-genetic trend, respectively, 1.76 t  ha−1 and 9.37 t  ha−1 (Fig. 3). The yield of monsoon and the start year (tmin, rmin) corresponds to the start of the rice varieties varied between 0.78 t  ha−1 and 7.17 t  ha−1 period over which gain was assessed. To assess the gain for (Fig. 5). The newly registered varieties in winter season the last 50 years for both seasons, the value of tmin and rmin are 1970 for the winter season, and 1973 for the monsoon 1 3 Theoretical and Applied Genetics (2023) 136:18 Page 7 of 13 18 Fig. 3 Genotype means per environment of winter rice seasons plotted against the trial year of the respective genotype. Colors indicate the year of release for the respective genotype (color figure online) Fig. 4 Genotype means per environment of winter rice seasons plotted against year of release for the respective genotype. Colors indicate the calendar year of the respective trial (color figure online) have higher yield as compared to older varieties. However, Genetic and non‑genetic trends monsoon rice varieties have consistent yield over time. The estimate of heritability for the winter rice and mon- The genetic trend was positive, with an increase in yield at soon rice is 0.866 and 0.915, respectively. The high herit- the rate of 0.01 t/ha (confidence interval, CI: 0.006, 0.019) ability estimates indicate the good reliability of the data. per year for winter rice (numbers in parenthesis indicate the 1 3 18 Page 8 of 13 Theoretical and Applied Genetics (2023) 136:18 Fig. 5 Genotype means per environment of monsoon rice seasons plotted against the trial year of the respective genotype. Colors indicate the year of release for the respective genotype (color figure online) Fig. 6 Genotype means per environment of monsoon rice seasons plotted against the year of release for the respective genotype. Colors indicate the calendar year of the respective trial (color figure online) lower and upper confidence limits at α = 0.05). The genetic The non-genetic trend was positive with an increase trend for monsoon rice was also 0.01 t/ha (confidence inter- of 0.02 t/ha (confidence interval, CI: −0.002, 0.049) per val, CI: −0.006, 0.017) per year (Table 3). year for winter rice, and 0.02 t/ha (confidence interval, CI: −0.002, 0.042) per year for monsoon rice. 1 3 Theoretical and Applied Genetics (2023) 136:18 Page 9 of 13 18 The overall percentage change from 1970 till 2020 for genotype-location-year variance in both seasons is larger winter rice was 40.96%, of which 13.91% was genetic gain than the other genotype-environment variances, and as and 27.05% non-genetic gain. Overall percentage change for large  as or larger than the genotypic variances in both monsoon rice was 38.39%, slightly less than for winter rice. seasons. The genetic gain was 8.36%, while non-genetic gain was 30.03% for monsoon rice (Table 4). Discussion Variance components In rice, there has been very limited research on genetic gain The variance components and the contribution to the total in Asia in general, and in Bangladesh, no studies have been variance for location-year interaction effects for both sea- done yet, even though rice is the main staple food. The aim sons were relatively large compared to other variances of this study was to estimate the genetic gain for yield from (Table 5). The genotypic variance component observed the varieties released over the last 50 years for monsoon and for yield of both seasons was larger than that of the geno- winter rice ecosystems. The findings showed that over the type-year and the genotype-location variance components fifty years from 1970 to 2020, the overall yield gain for win- suggesting substantial genetic variability and consistency ter rice was 40.96%, with an increase of 0.82% per year. The of ranking across trials in the varieties used, indicative of genetic gain for winter rice was 0.01 t ha−1 (0.28%) per year. limited genotype-environment interaction. However, the The overall genetic improvement in yield over fifty years Table 3 Estimates and 95% Fixed effect Winter rice Monsoon rice confidence limits (CL) of intercepts (  ), genetic trends Estimate (S.E.) Lower CL Upper CL Estimate (S.E.) Lower CL Upper CL (  ) and non-genetic trend () of winter and monsoon rice Intercept () − 66.48 (27.08) − 119.58 − 13.38 − 47.32 (25.01) − 96.35 1.70 varieties evaluated during the Genetic trend () 0.01 (0.003) 0.006 0.019 0.01 (0.006) − 0.006 0.017 trial conducted from 2001 to Non-genetic trend () 0.02 (0.013) − 0.002 0.049 0.02 (0.011) − 0.002 0.042 2020 Table 4 Overall and yearly Winter rice Monsoon rice genetic and non-genetic gains (%) Total gain (%) Yearly gain (%) Total gain (%) Yearly gain (%) (1970 to 2020) (1973 to 2020) Overall gain 40.96 0.82 38.39 0.82 Genetic gain 13.91 0.28 8.36 0.18 Non-genetic gain 27.05 0.54 30.03 0.64 Table 5 Estimates of variance Variance Winter rice Monsoon rice components, standard error component (S.E.) and the contribution of variance components to mean Estimate S.E. Contribution (%) Estimate S.E. Contribution (%) variance (%) Genetic effects Gi 0.115 0.030 7.93 0.262 0.061 20.09 (GY)ik 0.018 0.004 1.26 0.017 0.004 1.28 (GL)ij 0.042 0.006 2.93 0.046 0.007 3.52 (GLY)ijk 0.254 0.008 17.48 0.270 0.008 20.64 Non-genetic effects Yk 0.050 0.039 3.44 0.017 0.028 1.26 Lj 0.327 0.179 22.57 0.134 0.086 10.29 (LY)jk 0.552 0.065 38.08 0.491 0.062 37.58 eijk 0.092 6.32 0.070 5.34 The variance components of eijk represent the mean variance of the genotype means estimated in the first stage 1 3 18 Page 10 of 13 Theoretical and Applied Genetics (2023) 136:18 was 13.91%. The overall non-genetic percentage change in cycle time (Cobb et al. 2019). In general, low rates of genetic 50 years is 27.05%. Non-genetic gain for winter rice was gain in South and Southeast Asian rice breeding are likely 0.02 t h a−1 per year (0.54% per year), which is larger than mainly due to long breeding cycles caused by repeated use of genetic gain (Tables 3 and 4). older, popular varieties as parents, and by limited selection For monsoon rice, the overall yield gain over the 47 years intensity for yield in multi-location trials. An analysis of the from 1973 to 2020 was 38.39%, with an increase of 0.82% BRRI breeding program, conducted as part of a breeding per year. The genetic gain was 0.01 t ha−1 per year, which modernization project initiated in 2015, indicated that both of is 0.18% change per year. The overall genetic percentage these problems were affecting breeding progress. Key areas change in 47 years is 8.36%. Non-genetic gain for winter of weakness detected by BRRI in its review of its program rice was 0.02 t ha−1 per year (0.64% per year). The overall included inadequate multi-location testing, inadequate selec- non-genetic gain in 47 years is 30.03%, which is larger than tion intensity, and very long breeding cycles due to over-use the genetic gain (Tables 3 and 4). of old parents. In the breeding program for the highly pro- Overall, genetic and non-genetic trend were quite similar ductive irrigated winter rice crop, the most popular varieties for monsoon and winter rice, with non-genetic trend substan- BRRI dhan28 and BRRI dhan29 were repeatedly used as tially exceeding genetic trend. Rates of genetic improvement parents, which resulted in limited improvement in additive for grain yield are low compared to some genetic gain esti- breeding value. The key weaknesses of the BRRI breeding mates reported in rice for other programs (Peng et al. 2000; program that limit the rate of genetic gain are not unusual and Tabien et al. 2008; McKenzie et al. 2014; Zhu et al. 2016; affect many public sector breeding programs. Breseghello et al. 2011), but not relative to gains reported The current rate of genetic gain per year in rice yield of by IRRI for favorable rainfed and irrigated environments. the BRRI released varieties of monsoon and winter season For India, Kumar et al. (2021) observed yearly genetic yield is only 0.01 t ha−1, which is not sufficient to meet the food increases of 0.68% under irrigated conditions, 0.87% under requirements for the projected population of 215 million in moderate reproductive stage drought conditions, and 1.9% 2050. The present rice production is about 38.7 million tons, under acute reproductive stage drought conditions. Juma et al. whereas the population in Bangladesh is growing annually (2021) estimated breeding values for rice grain yield using a by 1.22%, arable land is decreasing annually by 0.4%, and mixed model approach considering the pedigree-based rela- climate vulnerability is also increasing. According to Kabir tionship matrix, which varied from 2.12 to 6.27 t ha−1. In the et al. (2015), genetic gain of at least 0.044 t ha−1 per year period 1964–2014, the average genetic gain of grain yield (approximately 1% annually) will be needed to meet Bang- was 0.01 t ha−1  year−1 (0.23%). When only IRRI developed ladesh’s requirements through 2050. Achieving this rate of cultivars were considered in analysis, the rate rose to 0.02 t genetic gain will be a challenging job for breeders, requiring ha−1  year−1 (0.46%). Peng et al. (2000) assessed the trend a new strategy that involves accelerating the breeding cycle in rice yield of IRRI cultivars released during 1966–1996, and shifting away from use of a few popular varieties and observing a gain of 1% per year. Muralidharan et al. (2002) non-elite breeding pools as parents. examined the grain yield trends of rice cultivars developed BRRI’s breeding programs are being reorganized and and tested during 1976–1997 in METs conducted worldwide accelerated based on optimization strategies derived from under various ecosystems in international rice advancement application of the breeder’s equation to breeding pipeline trials, concluding that there was no scientific proof for either design (Cobb et al. 2019; Atlin and Econopouly 2022). In a genetic gain or yield loss of genotypes released for any recent years, BRRI’s project “Transforming Rice Breeding” of the ecosystems. Muralidharan et al. (2019) estimated the reshaped the breeding programs to shorten the breeding genetic gain for yields in genotypes evaluated in 11 rice eco- cycle time, increase selection accuracy and intensity, and systems in India during 1995–2013 and found a consider- improve selection for breeding value. Breeding cycle time able gap between projected growth in human population and has been reduced by implementing single-seed descent and growth in national yield of rice. To ensure rice supply to increasing the number of generations of advance annually meet the demands of a rising population, integrated genetic from one to two or three, and by selecting parents after only technology and policy interventions are required. one or two stages of replicated agronomic testing. Selec- Overall, the rates of genetic gain achieved in monsoon and tion accuracy has been increased substantially by introduc- winter rice by the BRRI program from 1970 through 2020 ing multi-location testing at the first agronomic testing step have been quite low. These low rates of gain are consistent (Stage 1), a practice which is not yet common in South Asian with those achieved by IRRI in favorable environments in rice breeding programs. In most BRRI pipelines, the number the Philippines, but inadequate to keep pace with population of entries included in Stage 1 testing has increased seven growth, climate change, and loss of land due to urbaniza- to tenfold in the last five years. A trait development pipe- tion. The rate of genetic gain can be improved by increasing line has now also been established parallel to the breeding selection differential with sufficient accuracy and decreasing pipeline to develop elite donors with increased frequency 1 3 Theoretical and Applied Genetics (2023) 136:18 Page 11 of 13 18 of alleles with large, well-validated effects on tolerance to analyzed, and a program is designed to transform rice biotic and abiotic stresses. Forward breeding became routine breeding at BRRI by reducing cycle time, increasing selec- since early 2016 with the inception of the BRRI moderniza- tion accuracy, and improving selection for breeding value. tion project. Forward breeding with trait-specific SNP mark- This program, technically supported by the International ers is also in routine use in BRRI breeding programs. Some Rice Research Institute, is expected to increase the annual of the breeding programs have already applied outsourced rate of genetic gain for yield to 1.5–2.0% in the near future. genome-wide marker systems to facilitate parent selection We hope that the findings presented here will assist based on genomic estimated breeding values. When all these governments and policymakers in achieving higher rates improvements are fully implemented, they are expected to of genetic improvement from rice breeding in South Asia. result in a substantial improvement in the rate of genetic gain Efforts are needed both to improve breeding pipelines and delivered by BRRI’s breeding programs. to accelerate dissemination of newly developed improved The long-term trials summarized in this study are an varieties and withdrawal of old varieties. Now that a high important data resource for dissecting genetic and non- level of fertilizer and crop protection input use has been genetic trend in rice yields in South Asia. The similarity of attained in Bangladeshi rice production, higher rates of non-genetic trend estimates for monsoon and winter yields genetic improvement are likely to be the principal pathway may indicate that some common agronomic factors affected to sustaining rice food security. both seasons. Supplementary Information The online version contains supplemen- tary material available at https://d oi.o rg/1 0.1 007/s 00122-0 23-0 4260-x. Conclusion Acknowledgements All the experimental locations for conducting tri- This is the first time that genetic trend has been estimated als and all the scientists involved in this study are thankfully acknowl- for Bangladeshi rice breeding programs, and the estimates edged. We also like to acknowledge the TRB-BRRI project for all sup- port provided. are among the few to have been published for South Asia. The results indicate that rice yield gains due to breeding Author contributions statement NMFR and WAM formatted the data- have been very limited since the end of the Green Revolu- sets for statistical analysis and performed statistical analyses, prepared tion in the favorable environments in which the experiment the figures and wrote the full manuscript. RA assembled all datasets. MSK supervised the trials. MCR, MRI, MIH, KMI, PSB, MAR and was conducted, amounting to 0.28% and 0.18% annually MSR conducted trials and recorded observations at multiple locations. in irrigated winter and rainfed monsoon rice, respectively. PS, SH and H-PP provided suggestions regarding the statistical models These gains are lower than the non-genetic trend detected and analysis, and conducted analyses. H-PP, MAB and GA helped in (0.54% and 0.64% annually for winter and monsoon rice, editing of the manuscript. All authors have read and approved the final manuscript. respectively) and are less than needed to maintain rice food security in the face of population growth, climate change, Funding Open Access funding enabled and organized by Projekt and land loss to urbanization. The low rates of genetic gain DEAL. This experimental study was conducted by Agricultural Sta- observed in this study appear to be broadly representative tistics Division under the research program titled “Stability Analysis of BRRI Varieties” funded by the Bill & Melinda Gates Foundation of those achieved in the favorable rice production environ- for the period of 2016 to 2020 through Transforming Rice Breeding ments in South Asia (Muralidharan et al. 2019) critical to the (TRB)-BRRI project. We would also like to acknowledge Deutsche region’s food security, although there is some evidence that Forschungsgemeinschaft (DFG – German Research Foundation) grant the rate of gain has been higher in drought-prone environ- PI 377/20-2 for supporting Hans-Peter Piepho and Waqas Ahmed Malik. ments (Kumar et al. 2021). Modest continuing gains in non-genetic trend of 0.54% Declarations and 0.64% annually in winter and monsoon, respectively, may be due to long-term improvements in crop management, Conflict of interest The authors declare that they have no conflict of but also may shed light on the impacts to date of climate interest. change on rice productivity in favorable rice production Ethical standards The authors declare that the experiments comply environments in Bangladesh. Increasing temperatures are with the current laws of the countries in which the experiments were expected to increase climate risk to yields in monsoon pro- performed. duction and decrease them in winter rice production (Sarker Data availability The datasets generated during and/or analyzed during et al. 2017), however, we cannot dissect the effect of agro- the current study are available on reasonable request. nomic practices and impact of climate change. The findings of this study confirm the need to increase Open Access This article is licensed under a Creative Commons Attri- the rate of genetic gain delivered by rice breeding pro- bution 4.0 International License, which permits use, sharing, adapta- grams in Bangladesh. Limiting factors have been carefully tion, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, 1 3 18 Page 12 of 13 Theoretical and Applied Genetics (2023) 136:18 provide a link to the Creative Commons licence, and indicate if changes (2015) Rice vision for Bangladesh: 2050 and beyond. Bangladesh were made. The images or other third party material in this article are Rice J 19(2):1–18 included in the article's Creative Commons licence, unless indicated Kumar A, Raman A, Yadav S, Verulkar SB, Mandal NP, Singh ON, otherwise in a credit line to the material. If material is not included in Swain P, Ram T, Badri J, Dwivedi JL, Das SP, Singh SK, Singh SP, the article's Creative Commons licence and your intended use is not Kumar S, Jain A, Chandrababu R, Robin S, Shashidhar HE, Hit- permitted by statutory regulation or exceeds the permitted use, you will talmani S, Piepho HP (2021) Genetic gain for rice yield in rainfed need to obtain permission directly from the copyright holder. To view a environments in India. Field Crops Res 260:107977. https://d oi.o rg/ copy of this licence, visit http://c reati vecom mons.o rg/l icens es/b y/4.0 /. 10. 1016/j. fcr. 2020. 107977 Laidig F, Piepho HP, Drobek T, Meyer U (2014) Genetic and non-genetic References long-term trends of 12 different crops in German official variety performance trials and on-farm yield trends. Theor Appl Genet 127(12):2599–2617 Ahrends HE, Eugster W, Gaiser T, Rueda-Ayala V, Hüging H, Ewert F, Laidig F, Piepho HP, Rentel D, Drobek T, Meyer U, Huesken A (2017) Siebert S (2018) Genetic yield gains of winter wheat in Germany Breeding progress, variation, and correlation of grain and quality over more than 100 years (1895–2007) under contrasting fertilizer traits in winter rye hybrid and population varieties and national applications. Environ Res Lett 13(10):104003 on-farm progress in Germany over 26 years. Theor Appl Genet Atlin GN, Econopouly BF (2022) Simple deterministic modeling can 130(5):981–998. https:// doi. org/ 10. 1007/ s00122- 017- 2865-9 guide the design of breeding pipelines for self-pollinated crops. Crop Mackay I, Horwell A, Garner J, White J, McKee J, Philpott H (2011) Sci 62:661–678. https:// doi. org/ 10. 1002/ csc2. 20684 Reanalyses of the historical series of UK variety trials to quantify BBS (Bangladesh Bureau of Statistics). (1972) Yearbook of Agricultural the contributions of genetic and environmental factors to trends Statistics of Bangladesh 1972. Ministry of Planning, Government of and variability in yield over time. Theor Appl Genet 122:225–238. the People’s Republic of Bangladesh. https:// doi. org/ 10. 1007/ s00122- 010- 1438-y BBS (Bangladesh Bureau of Statistics). (2019) Yearbook of Agricultural Masuka B, Atlin GN, Olsen M, Magorokosho C, Labuschagne M, Statistics of Bangladesh 2019. Ministry of Planning, Government of Crossa J, Bänziger M, Pixley KV, Vivek BS, von Biljon A, Mac- the People’s Republic of Bangladesh robert J (2017) Gains in maize genetic improvement in Eastern and Bell AR, Bryan E, Ringler C, Ahmed A (2015) (2015) Rice productiv- Southern Africa: I CIMMYT hybrid breeding pipeline. Crop Sci ity in Bangladesh: What are the benefits of irrigation? Land Use 57(1):168–179 Policy 48:1–12 McKenzie K, Sha X, Moldenhauer K, Linscombe S, Lyman N, Nalley Breseghello F et al (2011) Results of 25 years of upland rice breeding in L (2014) Rice. In: Smith S, Diers B, Specht J, Carver B (eds) Yield Brazil. Crop Sci 51:914–923 gains in major U.S. field crops. ASA, CSSA and SSSA, Madison, Brolley, M. (2015) Rice security is food security for much of the world. pp 267–292 Rice Today. International Rice Research Institute (IRRI), DAPO Mottaleb KA, Mishra AK (2016) Rice consumption and grain-type Box 7777, Metro Manila, Philippines. pp. 30–32 preference by household: a Bangladesh case. J Agric Appl Econ Cobb JN, Juma RU, Biswas PS et al (2019) Enhancing the rate of genetic 48:298–319 gain in public-sector plant breeding programs: lessons from the Möhring J, Piepho HP (2009) Comparison of weighting in two-stage breeder’s equation. Theor Appl Genet 132:627–645. https:// doi. analysis of plant breeding trials. Crop Sci 49:1977–1988. https:// org/ 10. 1007/ s00122- 019- 03317-0 doi. org/ 10. 2135/ crops ci2009. 02. 0083 Cullis BR, Smith AB, Coombes NE (2006) On the design of early gen- Muralidharan K, Prasad CSV, Rao CS (2002) Yield performance of rice eration variety trials with correlated data. J Agr Biol Environ Stat genotypes in international multi-environment trials during 1976–97. 11(4):381–393 Curr Sci 83:610–619 DAE (2020) Weekly crop production report. Department of agricultural Muralidharan K, Prasad CSV, Rao CS, Siddiq EA (2019) Genetic gain for extension, Ministry of agriculture, Government people's republic of yield in rice breeding and rice production in India to meet with the Bangladesh, Dhaka, Bangladesh demand from increased human population. Curr Sci 116(4):544–560 Damesa MT, Möhring J, Worku M, Piepho HP (2017) One step at a time: Peng S, Laza R, Visperas R, Sanico A, Cassman K, Khush G (2000) Stage wise analysis of a series of experiments. Agrono J 109:845– Grain yield of rice cultivars and lines developed in the philippines 857. https:// doi. org/ 10. 2134/ agron j2016. 07. 0395 since 1966. Crop Sci 40:307–314. https:// doi. org/ 10. 2135/ crops Duvick DN (2005) Genetic progress in yield of United States maize (Zea ci2000. 40230 7x mays L.). Maydica 50:193–200 Piepho HP, Laidig F, Drobek T, Meyer U (2014) Dissecting genetic and Duvick DN, Smith JSC, Cooper M (2004) Long-term selection in a non-genetic sources of long-term yield trend in German official vari- commercial hybrid maize breeding program. Plant Breed Rev ety trials. Theor Appl Genet 127:1009–1018 24(2):109–152 Piepho HP, Möhring J (2007) Computing heritability and selec- FPMU (2020) Bangladesh Food Situation Report, April-June 2020. Food tion response from unbalanced plant breeding trials. Genetics planning and monitoring unit, Ministry of food, Government Peo- 177(3):1881–1888. https:// doi. org/ 10. 1534/ genet ics. 107. 074229 ple’s Republic of Bangladesh, Dhaka, Bangladesh. Vol. 121 Rife TW, Graybosch RA, Poland JA (2019) A field-based analysis of HIES (Household Income and Expenditure Survey), (2016) Bangladesh genetic improvement for grain yield in winter wheat cultivars Bureau of Statistics. Government of Bangladesh, Dhaka developed in the us central plains from 1992 to 2014. Crop Sci Juma RU, Bartholomé J, Thathapalli Prakash P, Hussain W, Platten JD, 59:905–910 Lopena V, Cobb JN (2021) Identification of an elite core panel as a Sarker Md AR, Alam K, Gow J (2017) Performance of rain-fed Aman key breeding resource to accelerate the rate of genetic improvement rice yield in Bangladesh in the presence of climate change. Renew for irrigated rice. Rice 14(1):1–22 Agric Food Syst. https:// doi. org/ 10. 1017/ S1742 17051 70004 73 Kabir MS, Salam MU, Chowdhury A, Rahman NMF, Iftekharuddaula Tabien R, Samonte S, McClung A (2008) Forty-eight years of rice KM, Rahman MS, Rashid MH, Dipti SS, Islam A, Latif MA, Islam improvement in Texas since the release of cultivar Bluebonnet in AKMS, Hossain MM, Nessa B, Ansari TH, Ali MA, Biswas JK 1944. Crop Sci 48:2097–2106 1 3 Theoretical and Applied Genetics (2023) 136:18 Page 13 of 13 18 World Bank (2016) https:// www. world bank. org/ en/ resul ts/ 2016/ 10/ 07/ Publisher's Note Springer Nature remains neutral with regard to bangl adesh- growi ng- econo my- throu gh- advan ces- in- agric ulture jurisdictional claims in published maps and institutional affiliations. Xu Y, Li P, Zou C, Lu Y, Xie C, Zhang X, Prasanna BM, Olsen MS (2017) Enhancing genetic gain in the era of molecular breeding. J Exp Bot 68:2641–2666 Zhu G et al (2016) Genetic improvements in rice yield and concomi- tant increases in radiation- and nitrogen-use efficiency in middle reaches of Yangtze river. Sci Rep 6:21049. https:// doi. org/ 10. 1038/ srep2 1049 1 3