- Email: [email protected]
An analysis of wheat yield and adaptation in India
Field Crops Research 219 (2018) 192–213
Contents lists available at ScienceDirect
Field Crops Research journal homepage: www.elsevier.com/locate/fcr
An analysis of wheat yield and adaptation in India a,⁎
b
b
T b
b,f
Richard Trethowan , Ravish Chatrath , Ratan Tiwari , Satish Kumar , M.S. Saharan , Navtej Bainsc, V.S. Sohuc, Puja Srivastavac, Achla Sharmac, Nitish Ded, Surya Prakashe, G.P. Singhb, Indu Sharmab, Howard Eaglesg, Simon Diffeyh, Urmil Bansala, Harbans Barianaa a
The Plant Breeding Institute Sydney Institute of Agriculture, The University of Sydney, 107 Cobbitty Road, Cobbitty, NSW, 2570, Australia Indian Institute for Wheat and Barley Research, Agarsain Road, Karnal, 132001, India c Punjab Agricultural University, Ludhiana, 141004, India d Bihar Agricultural University, Sabour, Bhagalpur, Bihar, 813210, India e Birsa Agricultural University, Kanke, Ranchii, 834006, Jharkhand, India f Division of Plant Pathology, Indian Agricultural Research Institute, New Delhi, India g CSIRO Agriculture and Food, Black Mountain Science and Innovation Park, GPO Box 1700, Canberra, ACT, 2601, Australia h Centre for Bioinformatics and Biometrics, University of Wollongong, Northfields Avenue, Wollongong, NSW, 2522, Australia b
A R T I C L E I N F O
A B S T R A C T
Keywords: Genotype x environment interaction Wheat Multi-environment trials Factor analytic models
Multi-environment wheat trials provide valuable information on the extent of genotype x environment interaction, the stability of genotypes and define and confirm agro-ecological regions through associations among sites. The All India Coordinated Crop Improvement Project on wheat evaluates candidates for release across the wheat growing regions of India. To facilitate this process the wheat area is divided into six agro-ecological zones; the northwestern plains zone (NWPZ), the northeastern plains zone (NEPZ), the central zone (CZ), the peninsular zone (PZ), the northern hills zone (NHZ) and the southern hills zone (SHZ). Factor analytic (FA) models were used to analyze the genotype x environment interaction for yield of 813 wheat genotypes evaluated at 136 locations across the six agro-ecological zones in 1307 individual advanced variety trials between 2008/09 and 2012/13. Genotype x environment interaction was firstly assessed separately within each of the six established agroecological zones. Key locations with a high genetic correlation with all other locations within each zone were identified. Predicted genetic values of important cultivars that were represented in a wider range of environments within each zone were estimated and highly stable genotypes were found. Genotype x environment interaction was subsequently assessed across agro-ecological zones. Only those environments where the models accounted for > 99% of the genetic variance were retained for further analysis and two smaller zones (NHZ and SHZ) with little or no genotype congruence with other agro-ecological zones were removed. Thus 476 genotypes from 488 environments were included in the analysis. Fifteen clusters of environments with similar patterns of adaptation were found. These clusters were then characterized based on zonal classification, sowing time, irrigation regime, latitude and year and three regions broadly representing the main wheat growing areas of India were identified. These regions represent a combination of the NWPZ and NEPZ defined by latitude, a central region that combines CZ locations with northern PZ locations and a southern region comprised of southern PZ sites. Further stratification of these zones was then possible based on sowing time and irrigation practice. One cluster of 29 environments had a high average genetic correlation (r = 0.75) with most other environments and production zones. These represent key locations where larger numbers of entries might be grown in future seasons as they are the best predictors of yield across cropping zones.
1. Introduction Yield evaluation of new candidate crop genotypes in multi-environment trials is a key component of most varietal release systems ⁎
globally. These multi-environment experiments also provide estimates of genotype x environment interaction, genotype stability and the genetic relationships among varieties and environments (Cooper and DeLacy, 1994). However, large multi-environment data are rarely
Corresponding author. E-mail address: [email protected] (R. Trethowan).
https://doi.org/10.1016/j.fcr.2018.01.021 Received 20 November 2017; Received in revised form 18 January 2018; Accepted 22 January 2018 0378-4290/ © 2018 Elsevier B.V. All rights reserved.
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Fig. 1. Indian cropping zones and the location (blue dots) of AVT trials, 2008–2013. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
balanced. Factor analytic (FA) models have been used to estimate genotype x environment interaction in large unbalanced multi-environment data (Smith et al., 2015). The most general model for
estimating the between environment genetic variance matrix is an unstructured form that estimates a genetic variance for each environment and covariance between each pair of environments. However, as the
193
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
zones in the Advanced Variety Trial (AVT). These production zones include the central zone (CZ), the northeastern plains zone (NEPZ), the northern hills zone (NHZ), the northwestern plains zone (NWPZ), the peninsular zone (PZ) and the southern hills zone (SHZ) (Fig. 1). These data are not only used to make decisions on cultivar release in India but also provide a basis for understanding genotype x environment interaction and further refinement of India’s varietal release strategy. This study estimated genetic correlations between such environments to identify key sites representative of a larger group of environments. This information can be used to reduce the number of experimental sites and increase genotype concurrence among environments both within and between agro-ecological zones.
Table 1 Numbers of unique environments and genotypes evaluated for yield in AVT cropping zone trials, 2008–2013. Cropping zone
Unique Environments
Genotypes
CZ NEPZ NHZ NWPZ PZ SHZ
222 161 138 354 146 18
139 105 226 215 114 37
Note: CZ, NEPZ, NHZ, NWPZ, PZ and SHZ is the central zone, the northeast plains zone, the northern hills zone, the northwest plains zone, the plains zone and the southern hills zone, respectively.
2. Materials and methods Table 2 The range on yield and mean square errors of AVT trials within each cropping zone 2008–2013.
The yield data of AVT trials 2008–2013 were compiled by the Indian Institute for Wheat and Barley Research (IIWBR). The supplied data comprised trial mean yields (quintals/ha), coefficients of variation (CV’s) and predicted variety means and associated standard errors from the analysis of individual trial data. Yield data captured on 813 wheat genotypes evaluated at 136 locations in 1307 individual AVT yield trials between 2008/09 and 2012/13 were analyzed to estimate genotype x environment interaction using FA mixed models. The AVT yield data was captured from trials conducted in the six agro-ecological zones (CZ, NEPZ, NHZ, NWPZ, PZ and SHZ). Trials were managed using standard Indian government management practices recommended for each zone. Each trial/location/year combination is considered an environment and will henceforth be referred to as such. The distribution of environments within each zone classification is outlined in Fig. 1.
Range Cropping zone CZ NEPZ NHZ NWPZ PZ SHZ
Trial mean yield (q/ha) 10–75 14–60 10–63 12–68 10–61 27–63
Mean square error 59.3–0.17 38.8–0.05 59.3–0.37 31.5–0.16 28.9–0.04 85.1–5.87
Note: CZ, NEPZ, NHZ, NWPZ, PZ and SHZ is the central zone, the northeast plains zone, the northern hills zone, the northwest plains zone, the plains zone and the southern hills zone, respectively.
2.1. Analysis of multi-environment trials within specific agro-ecological zones
size of the multi-environment data increases it becomes difficult to fit the model and estimate variance parameters. The FA model provides an alternative to the unstructured form and can efficiently account for the genetic covariance between environments (Smith et al., 2001; Kelly et al., 2007). India is the second biggest producer of wheat globally and annual production regularly exceeds 90 million tonnes (Oladele and Kenamara, 2015). A number of Indian wheat breeding programs produce new improved cultivars for different agro-ecological zones and candidate genotypes with potential for release to farmers are evaluated nationally in coordinated trials. The official testing of candidate wheat cultivars in India is conducted by the All India Coordinated Crop Improvement Project on wheat (Chauhan et al., 2016). This body oversees the conduct of multi-environment wheat yield trials managed by the Indian Council of Agricultural Research (ICAR) Institutes, State Agricultural Universities and institutes governed by state agriculture departments. The top-tier of this yield evaluation is conducted across six production
Separate analyses within each agro-ecological zone were performed as genotypes were specifically deployed in trials in each zone. A summary of the numbers of unique environments and genotypes evaluated in 2008–2013 is given in Table 1. The allocation of genotypes to trials was highly structured and groups of trials within each year and zone had satisfactory genotype concurrence. However, limited concurrence was available between pairs of zones and across years. For this reason FA mixed models were used to estimate the genetic covariance between environments. Weights were applied to the data because of the variability in trial mean square error (Table 2). The weights used were the reciprocal of the estimated variances associated with the predicted variety means from individual trial analyses. The FA mixed models were then fitted to the data from each zone in two stages. In the first stage a simple variance components model was fitted to the data (referred to as the ID model) followed by a model which accommodated variance heterogeneity between
Table 3 Summary of models fitted FA1 through to FA6, the residual log-likelihood (LL) and percentage of genetic variance accounted for (vaf) in each cropping zone. CZ Model FA1 FA2 FA3 FA4 FA5 FA6
LL −5156 −4850 −4619 −4335
NEPZ vaf 35 56 68 80
LL −3036 −2877 −2725 −2575
NHZ vaf 33 56 71 80
LL −3351 −3214 −3093 −2961
NWPZ vaf 41 62 77 88
LL −8907 −8484 −8095 −7701 −7280 −6902
PZ vaf 32 50 61 72 79 86
LL −2589 −2389 −2246 −2109
SHZ vaf 44 66 80 89
LL −268 −253
vaf 52 82
Note: CZ, NEPZ, NHZ, NWPZ, PZ and SHZ is the central zone, the northeast plains zone, the northern hills zone, the northwest plains zone, the plains zone and the southern hills zone, respectively.
194
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
environments, but not the covariance between environments (referred to as the DIAG model). Based on the DIAG model, environments that had no genetic variance were removed from the analyzed data set. In the second stage a series of FA mixed models starting with an FA model of order 1 through to an FA model of order 6, if required, was fitted to the data. A summary of the FA models fitted for each zone is provided in Table 3. The predicted regression part of the FA model was used to estimate genotype performance in each environment using the same principle as Smith et al. (2015). These predictions are referred to as genetic values (GV’s). All models were fitted using the statistical software package asreml (Butler, 2009) within the R (R Core Team, 2015) computing environment. The genetic correlation matrix between environments was then estimated. However, due to the large number of environments within each zone, an agglomerative hierarchical clustering algorithm was used to reorder the rows and columns of the genetic correlation matrix to better observe any structure. The environments within each zone that most closely represented the wider zone were then identified by constructing dendrograms of those environments were the final FA model accounted for at least 99% of the genetic variance.
Table 6 Summary of factor analytic (FA) mixed models fitted to the four AVT cropping zones.
NHZ
NWPZ
PZ
SHZ
CZ NEPZ NHZ NWPZ PZ SHZ
139 5 1 22 29 0
5 105 0 32 7 0
1 0 226 1 0 0
22 32 1 215 13 1
29 7 0 13 113 0
0 0 0 1 0 37
Cropping Zone NWPZ 198
42 65 80 92
Trial mean square error and trial mean yield varied considerably within cropping zones (Table 2). The biggest difference in mean square error was observed in the NEPZ where the largest error (38.8) was approximately 800 times the smallest mean square error (0.05). The smallest range in mean square errors was observed in the SHZ. The genetic variance accounted for by the FA4 model varied from 6% to 100% across all zones. There were 125 environments across all zones where the percentage variance accounted for was < 50% and the genetic correlations and predicted genotype performance associated with these environments was reduced compared to 592 environments where the FA4 model accounted for > 99% of the genetic variance. The FA4 model accounted for between 72 and 89% of the average genetic variance at an environment (Table 3). However, the FA4 model was less efficient in the NWPZ (72%) and the FA6 model (86%) was used instead. The reordered genetic correlation matrix was constructed for each zone (not shown) and some underlying structure in the genetic correlation between environments was observed due to the limited genotype concurrence between many environments. Dendrograms were constructed (not shown) based on the hierarchical clustering of the dissimilarity matrix associated with the genetic correlations among environments to examine site/experiment similarities. Those sites, represented by more than one experiment, among the group least different to all other sites within each zone were identified (Table 8). In the CZ, Indore and Bhopal were the best predictors of zone performance. Kalyani, Kapurthala and Niphad best represented the NEPZ, NWPZ and PZ. Two sites, Wellington and Thalawadi, were the most representative sites in the SHZ. Important widely grown cultivars in each of the different cropping zones were subsequently selected for further analysis (Table 9) and the
Table 5 The number of environments in the Indian AVT cropping regions, 2008–2013 .
NEPZ 88
−9592 −8985 −8330 −7565
3.1. Multi-environment trials within agro-ecological zones
Note: CZ, NEPZ, NHZ, NWPZ, PZ and SHZ is the central zone, the northeast plains zone, the northern hills zone, the northwest plains zone, the plains zone and the southern hills zone, respectively.
CZ 107
FA1 FA2 FA3 FA4
3. Results
Table 4 Genotype concurrence between Indian AVT wheat cropping zones 2008–2013. NEPZ
vaf
weights based on the standard errors associated with the predicted variety means to model genetic covariances between environments. A series of FA mixed models were fitted starting with a FA model of order 1 through to a FA model of order 4. A summary of the FA models fitted is provided in Table 6. The FA4 model accounted for, on average, 92% of the genetic variance at an environment. However, the genetic variance accounted for by the FA4 model varied from 10% to 100%. There were 30 environments where the genetic variance accounted for was less than 50% (Table 7). Twenty-six of these 30 environments were located in the NWPZ and these environments were subsequently removed from the analysis. However, 385 environments where the FA4 model accounted for > 99% of the genetic variance were retained. Of these environments, 135 were late sown and irrigated, 127 timely sown and irrigated and 123 were timely sown in rainfed conditions. All models were fitted using the statistical software package asreml (Butler, 2009) within the R (R Core Team, 2015) computing environment.
Genotype concurrence between agro-ecological zones was relatively low (Table 4). A FA mixed model was fitted to the trial data from each zone (data not shown) and those environments where the models accounted for more than 99% of the genetic variance were retained for analysis. The data comprised 7502 yield observations on 736 genotypes grown in 592 environments. As little or no genotype concurrence between NHZ and SHZ and the other cropping zones was observed (Table 4), these two zones were removed from the analysis. After dropping NHZ and SHZ the data file comprised 476 varieties over 488 environments. The number of genotypes in CZ, NEPZ, NWPZ and PZ are the respective diagonal values in Table 4 and the number of environments in each zone is presented in Table 5. Data were analyzed using a two-stage FA mixed model following the approach of Smith et al. (2015). This approach used the predicted variety means from analyses of individual trial data and calculated
CZ
LL
Note: “LL” refers to the residual (or restricted) log-likelihood and “vaf” refers to the percentage of genetic variance accounted for by each model.
2.2. Analysis of multi-environment trials across agro-ecological zones
Zone
Model
PZ 95
Note: CZ, NEPZ, NWPZ and PZ is the central zone, the northeast plains zone, the northwest plains zone and the plains zone, respectively.
195
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
yielding (Fig. 2B). In the CZ, the genotypes GW322 and GW366 were not as stable as HI1544 (Fig. 2C). Only two genotypes were selected in the small SHZ (HW5216 and CO(W)1). There was significant cross-over interaction observed and while HW5216 was more stable, the genotype CO(W)1 showed specific adaptation to three key SHZ environments (Fig. 2D). In the NHZ, the genotype VL907 was generally below the mean across all sites and HPW349 and HS507 generally superior or equivalent to the mean and were therefore considered more stable (Fig. 2E). The genotypes MACS6222, NIAW917 and RAJ4037 were selected in the PZ (Fig. 2F). However, no genotype was particularly stable, although NIAW917 had generally higher predicted values across environments.
Table 7 Thirty environments from the MET analysis across cropping zones and years that explained < 50% of the genetic variance. Location
Zone
Sowing time
Irrigation
Year
Genetic variance%
Ambala Balachaur Bareilly Bawal Bulandshahar Bulandshahar Bulandshahar Bulandshahar Bulandshahar Delhi Diggi Durgapura Durgapura Gurdaspur Gurdaspur Gwalior Hisar Kalyani Karnal Karnal Kaul Ludhiana Malda Modipuram Modipuram Rauni Tabiji Tabiji Ujhani Varanasi
NWPZ NWPZ NWPZ NWPZ NWPZ NWPZ NWPZ NWPZ NWPZ NWPZ NWPZ NWPZ NWPZ NWPZ NWPZ CZ NWPZ NEPZ NWPZ NWPZ NWPZ NWPZ NEPZ NWPZ NWPZ NWPZ NWPZ NWPZ NWPZ NEPZ
TS TS LS TS LS LS LS TS TS TS TS TS TS LS TS LS TS TS LS TS TS TS TS TS TS TS LS TS TS LS
RF RF IR IR IR IR IR IR IR IR RI IR IR IR IR IR RF IR IR IR IR IR IR RF RF IR IR IR IR IR
2009 2011 2011 2008 2009 2012 2010 2012 2010 2008 2012 2009 2010 2008 2010 2011 2008 2008 2009 2012 2009 2011 2009 2011 2009 2010 2011 2008 2011 2012
13 49 46 37 35 40 45 26 49 32 10 34 47 43 18 30 27 29 20 19 34 26 45 48 49 46 44 46 20 17
3.2. Multi-environment trials across agro-ecological zones A plot of trial mean square error against trial mean yield is presented in Fig. 3. Both the trial mean square error and trial mean yield varied considerably, with the largest mean square error (38.79) more than 800 times larger than the smallest mean square error (0.045). A heat-map of the genetic correlations between the 385 environments where the FA4 model accounted for > 99% of the genetic variance is presented in Fig. 4 and the details of each environment in Appendix A Table A1. The trials specific to individual cropping zones are represented by boxes within the heat-map. There appeared to be little pattern to the genetic correlations across the 385 environments. A dendrogram was constructed (Fig. 5) based on the hierarchical clustering of the dissimilarity matrix associated with the genetic correlations in Fig. 4. Fifteen environment clusters representing similar patterns of adaptation were identified.
Note: CZ, NEPZ, NWPZ and PZ is the central zone, the northeast plains zone, the northwest plains zone, and the plains zone, respectively. TS, LS, IR, RF and RI and timely sown, late sown, irrigated, rainfed and restricted irrigation, respectively.
3.2.1. Classification of all environment clusters The 15 clusters defined in Fig. 5 were then characterized on the basis of dominant zonal representation, sowing time, irrigation, latitude and year (Table 10). A full description of each environment is found in Appendix A Table A1. A single zonal classification dominated in 7 of the 15 clusters (clusters 2, 5, 6, 10, 12, 13 and 14). The NWPZ dominated in cluster 12, the NEPZ in cluster 10, the PZ in clusters 5 and 6 and the CZ in clusters 2, 13 and 14. However, 5 clusters showed no zonal influence (3, 7, 8, 9 and 11) and 3 were dominated by two zones (1, 4 and 15). Sowing time also influenced the clustering of environments. Seven clusters were predominately timely sown (5, 7, 9, 10, 12, 14 and 15) and 3 late sown (2, 6 and 13). Five were classified as both timely and late sown (1, 3, 4, 8 and 11). Nine clusters were dominated by irrigated sites (1, 2, 3, 6, 7, 8, 9, 12 and 13) and only one by drought affected locations (cluster 5). A further five clusters comprised a significant proportion of both irrigated and drought affected environments (4, 10,
predicted genetic value of these genotypes in specific groups of environments was estimated (Fig. 2A–F). These trials/environments were chosen because, apart from the selected genotypes, there were other genotypes in common. Differences in the stability of different genotypes were noted in each zone. Four common and frequently repeated genotypes were selected in the NWPZ (Fig. 2A). Of these, PBW343 was the most unstable genotype and HD2967 the most stable and high-yielding. The adaptation of the three selected genotypes in the NEPZ (HD2967, K0307 and PBW343) was generally similar and the yield relatively stable with the exception of PBW343 in one environment, where this genotype was very high-
Table 8 Key locations least different to all others within each agro-ecological zone; those locations represented by more than one experiment are listed. Agro-ecological zone CZ
NEPZ
NHZ
NWPZ
PZ
SHZ
Indore (3) Bhopal (2)
Kalyani (4) Faizabad (2) Deegh (2) Shillongani (2)
Shimla (7) Hawalbagh (5) Bajuara (4) Kalimpong (2) Majhera (2) Rajouri (2)
Kapurthala (3) New Delhi (2) Gurdaspur (2) Balachaur (2) Karnal (2) Uchani (2)
Niphad (5) Kalloli (4) Karad (3) Prabhani (3) Muhol (2) Pune (2) Arbhavi (2)
Wellington (2) Thalawadi (2)
Note: CZ, NEPZ, NHZ, NWPZ, PZ and SHZ is the central zone, the northeast plains zone, the northern hills zone, the northwest plains zone, the plains zone and the southern hills zone, respectively. Numbers in parentheses indicate the number of occurrences.
196
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
4. Discussion
Table 9 Selected widely grown genotypes in each cropping zone (year of release in brackets).
Very little has been published on the extent of genotype x environment interaction in major crops in India. Those studies that have been reported are generally balanced, small in scale and assess interactions using GGE or AMMI analyses (Bhartiya et al., 2017; Verma et al., 2016; Kota et al., 2013). Trethowan et al. (2001, 2003) used the shifted multiplicative model to examine associations among global wheat yield testing environments including sites in India. While such analyses identified site associations between Indian locations and some global sites and identified Pirsabak in Pakistan as a key predictor of global wheat yield, very little is known about the extent of genotype x environment interaction for wheat on the Indian subcontinent. The distribution of AVT trial sites across India between 2008 and 2013 sampled most of the national wheat growing environments. However, the multi-environment analysis across agro-ecological zones revealed a lack of genotype concurrence across zones and years. This lack of connectivity impeded the identification of key sites and germplasm. However, limited concurrence between trials did not preclude estimating the genetic covariance between trials using FA mixed models, although those sites that accounted for < 99% of the genetic variance were excluded. Twenty-six of the 30 environments that accounted for < 50% of the genetic variance were located in the NWPZ and half were timely sown and irrigated. A lack of range in genotype performance in these high-yielding environments (ie all genotypes were high-yielding) may have accounted for the lower than expected genetic variances in the NWPZ; India’s highest yielding production environment. Nevertheless, 385 environments of 488 (excluding the NHZ and SHZ because of lack of genotype concurrence with other zones) accounted for > 99% of the genetic variance and thus provided sufficient scale for the estimation of genetic correlations. The 15 individual site clusters generated across the entire data set partially confirmed the ICAR classification of agro-ecological zones.
Cropping Zone CZ
NEPZ
GW322 (2002) HD2967 (2011) GW366 (2006) KO307 (2006) HI1544 (2007) PBW343 (1995)
NHZ
NWPZ
PZ
SHZ
HPW349 (2012) HS507 (2010) VL907 (2010)
DBW17 (2006) HD2967 (2011) PBW343 (1995) PBW550 (2006)
MACS6222 (2001) NIAW917 (2005) RAJ4037 (2002)
CO(W)1 (2005) HW5216 (2012)
Note: CZ, NEPZ, NHZ, NWPZ, PZ and SHZ is the central zone, the northeast plains zone, the northern hills zone, the northwest plains zone, the plains zone and the southern hills zone, respectively.
11, 14 and 15). Latitude was also influential on site clustering with four clusters on average < 23° of latitude (5, 6, 8 and 13), eight between 23° and 25.9° (1, 2, 3, 4, 7, 11, 14 and 15) and three > 25.9° (9, 10 and 12). The year of assessment did not generally influence site clustering.
3.2.2. Identification of key environments Twenty-nine environments in cluster 8 located the middle of the dendrogram in Fig. 5 were identified as the group which best correlated with most other environments. Within this group the genetic correlations between environments ranged from 0.22 to 0.99 with an average of 0.75. These 29 environments are located in 26 physical locations based on longitude and latitude (Table 11) and are possible locations where larger numbers of entries might be grown in future seasons as they are the best predictors of yield across cropping zones.
Fig. 2. Predicted genetic values (q/ha), and associated standard errors for key cultivars in each (A) NWPZ, (B) NEPZ, (C) CZ, (D) SHZ, (E) NHZ and (F) PZ in common experiments.
197
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Fig. 3. Trial mean square error (MSE) and trial mean yield (TMY) from individual trial analyses.
Fig. 4. Heatmap of the reordered genetic correlation matrix between the 385 environments where the FA4 model accounted for more than 99% of the genetic variance.
198
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Fig. 5. Dendrogram based on the hierarchical clustering of the dissimilarity matrix associated with the genetic correlations between the 385 environments where the FA4 model accounted for more than 99% of the genetic variance. The key group of 29 locations (cluster 8) indicated with a blue line. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
Table 10 Summary of the dendrogram groupings in Fig. 4 (from left to right) based on zonal representation, sowing time, irrigation, latitude and year. Group number
Number of sites
Zonal representation
Percentage timely sown
Percentage Irrigated
Average Latitude
Year
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
19 8 45 34 19 34 26 29 19 39 9 17 19 29 42
74% (NWPZ/NEPZ) 50% CZ No zonal influence 65% (PZ/CZ) 58% PZ 44% PZ No zonal influence No zonal influence PZ not represented 46% NEPZ; 5% PZ No zonal influence 47% NWPZ 47% CZ 41% CZ 71% (NWPZ and PZ)
63 38 53 59 84 41 69 62 68 84 66 71 36 72 83
89 100 71 65 26 70 77 92 74 56 66 88 79 59 47
25.3 23.8 23.9 23.1 21.9 22.1 24.8 22.6 26.0 26.3 25.8 26.1 22.5 24.4 24.2
No clear influence No clear influence No clear influence 35% of sites 2012 No clear influence 35% of site 2012 54% of sites 2010 No clear influence No 2011 sites No clear influence No clear influence 2008 not represented No clear influence Few 2012 sites No clear influence
Note: CZ, NEPZ, NWPZ and PZ is the central.
differentiation based on sowing time and irrigation. Five clusters did not have a dominant zone including the key cluster 8 that correlated best with all other environments. These sites tended to cluster on the basis of sowing time, irrigation and latitude but could be found spread across the zonal classifications. The multi-environment analysis suggests that three primary regions of adaptation exist across the main wheat growing areas of India. These are a combined NWPZ/NEPZ region defined by latitude, a central region that combines CZ locations with northern locations in the PZ and a southern region comprised of southern PZ sites. Further stratification of these regions is then possible based on sowing time and irrigation practice. The multi-environment analysis of individual zones provided additional and valuable information on key zonal locations and genotype adaptation. Due to the large number of genotypes evaluated in each
Cluster 10, dominated by NEPZ locations, was generally timely sown with a mix of irrigated and drought stressed conditions, all at higher latitude (> 26°). Timely sown sites in this zone also clustered with NWPZ sites (cluster 1) which was predominately timely sown and irrigated, indicating that sowing time and latitude were probable drivers of adaptation across the NWPZ and NEPZ. The three primary clusters of CZ sites (clusters 2, 13 and 14) could be differentiated into those that were late sown and irrigated (2 and 13) and those that were timely sown and a mixture of irrigated and drought stressed conditions (14). Interestingly, the CZ sites in cluster 4 (comprised of PZ and CZ sites) are primarily late sown and irrigated, similar to clusters 2 and 13. Similarly, the two PZ clusters (5 and 6) are differentiated by sowing time and irrigation with cluster 5 predominantly timely sown and rainfed and cluster 6 late sown and irrigated. Low latitude appears to be a primary driver of adaptation in the PZ zone with further
199
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Table 11 Twenty nine environments identified as the cluster (8) which best correlates with most other environments and whose locations are possible candidates for growing larger numbers of entries in future seasons. Superscripts a, b and c identify three pairs of trials in the same location. Zone
Sowing time
Irrigation
Year
Location
Latitude
Longitude
TMY
CZ CZ CZ CZ CZ CZ CZ CZ CZ CZ CZ CZ NEPZ NEPZ NEPZ NEPZ NEPZ NEPZ NWPZ NWPZ NWPZ NWPZ NWPZ NWPZ PZ PZ PZ PZ PZ
TS TS TS TS TS TS TS LS TS LS LS LS TS TS TS TS TS TS TS TS TS LS LS TS LS TS TS TS LS
IR IR IR RI RI RF RF IR RI IR IR IR RF IR IR RF IR RF IR RF RI IR IR IR IR RF RF IR IR
2011 2009 2011 2008 2011 2008 2010 2009 2008 2008 2009 2011 2011 2012 2011 2010 2010 2009 2008 2010 2009 2008 2009 2010 2008 2008 2008 2010 2008
Amreli SK Nagar Junagarh Dhandhuka Dhandhuka Indore Indore Anand Arnej Vijapur Banswara Rewa Varanasi Saini Kanpur Faizabad Araul Barabanki Bulandshahar Hisar Hisar Kashipur Kaul Chatha Karad Parbhani Washim Niphad Amravati
20.86 21.18 21.31 22.37 22.37 22.37 22.37 22.55 22.58 23.35 23.55 24.53 25.20 25.66 26.29 26.47 26.92 26.92 28.41 29.10 29.10 29.22 29.84 32.43 17.28 19.08 20.10 20.60 20.92
70.75 72.86 70.33 71.98 71.98 75.75 75.75 72.95 72.26 72.55 74.45 81.30 83.03 81.32 80.18 82.80 80.01 81.20 77.83 75.46 75.46 78.95 76.66 74.54 74.20 76.50 77.13 74.60 77.76
63.88 47.28 42.78 26.00 24.36 10.70 22.29 37.18 27.60 36.60 51.15 49.72 26.81 52.13 48.08 33.08 40.56 14.58 43.20 31.17 40.17 41.80 43.87 40.73 47.80 13.10 13.20 39.59 28.60
Note: CZ, NEPZ, NHZ, NWPZ, PZ and SHZ is the central zone, the northeast plains zone, the northern hills zone, the northwest plains zone, the plains zone and the southern hills zone, respectively
(Table 7) and was subsequently discarded. For this reason the selection of Bulandshahar as a key location must be treated with caution. Hisar could be considered a suitable location for expanded evaluation of materials for drought adaptation as two water-limited trials in consecutive years explained > 99% of the genetic variance. Five key sites were identified in the PZ and of these only one was timely sown and irrigated (Niphad). Sites in this zone tend to better predict yield under stress across the Indian cropping zones. In summary, 9 of the 29 key locations represented the highest yielding conditions, considered to be timely sown and irrigated, with at least one site located in each of the 4 key agro-ecological zones. Twelve timely sown and water restricted environments (either rainfed or restricted irrigation) were identified in all four primary agro-ecological zones. However, of the 8 late sown/irrigated key environments, none were located in the NEPZ and most were found in the CZ. Nevertheless, the distribution of these 3 environment types (timely sown/irrigated; timely sown/water restricted; late sown/irrigated) across all cropping zones further emphasizes the value of these locations as predictors of yield in the Indian AVT network.
zone, the key genotypes, selected on the basis of their higher concurrence and general importance as major regional cultivars, provided a summary of adaptation in each zone. While little concurrence of major cultivars among cropping zones was evident, there were two exceptions. The cultivars PBW343 and HD2967 occurred in both the NWPZ and NEPZ. In both environments the yield of PBW343 was least stable and HD2967 most stable. PBW343 is an older cultivar (released in 1995) and is now susceptible to stripe rust and this most likely explains the lack of stability in both zones. In general, the most stable cultivars were more recently released and included HI1544, HD2967, K0307, HPW349, HS507, NIAW917 and HW5216. Given the expense of sowing advanced variety trials nationally and the general lack of genotype concurrence in the AVT, the identification of key locations that correlate well with the wider production area could be used to evaluate greater numbers of genotypes and to provide greater connectivity. Of the 29 locations identified in Table 10, twelve were located in the CZ. Of these, 6 environments representing 4 locations (Dhandhuka, Indore, Anand and Arnej) were water stressed and suitable for expanding genotype evaluation for rainfed conditions, whereas the remaining locations (Amreli, Nagar, Junagarh, Vijapur, Banswara and Rewa) best represented irrigated areas. The 6 NEPZ locations were all timely sown and of these 3 represented rainfed and 3 irrigated conditions. Of the 6 NWPZ key locations, only two (Bulandshahar and Chatha) were timely sown and irrigated − representing the most probable environment type in this high-yielding region of India. Interestingly, the key Bulandshahar trial was sown in 2008, however, this location explained < 50% of the genetic variance in all other years
Acknowledgements The authors wish to acknowledge the funding received from the Australian Centre for International Agricultural Research and the Indian Council for Agricultural Research that made this work possible. We also acknowledge the contribution of the many wheat researchers of India who all contributed to the AVT database over many years.
200
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Appendix A
Table A1 All sites/experiments used to define the 15 clusters defined in Fig. 5.
(continued on next page)
201
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Table A1 (continued)
(continued on next page)
202
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Table A1 (continued)
(continued on next page)
203
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Table A1 (continued)
(continued on next page)
204
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Table A1 (continued)
(continued on next page)
205
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Table A1 (continued)
(continued on next page)
206
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Table A1 (continued)
(continued on next page)
207
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Table A1 (continued)
(continued on next page)
208
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Table A1 (continued)
(continued on next page)
209
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Table A1 (continued)
(continued on next page)
210
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Table A1 (continued)
(continued on next page)
211
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Table A1 (continued)
212
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
References
Int. Res. J. Agric. Econ. Stat. 6 (1), 189–192. R Core Team, 2015. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria Retrieved from. https://www. R-project.org/. Smith, A., Cullis, B., Thompson, R., 2001. Analyzing variety by environment data using multiplicative mixed models and adjustments for spatial field trend. Biometrics 57, 1138–1147. Smith, A., Ganesalingam, A., Kuchel, H., Cullis, B., 2015. Factor analytic mixed models for the provision of grower information from national crop variety testing programs. Theor. Appl. Genet. 128 (1), 55–72. Trethowan, R.M., Crossa, J., van Ginkel, M., Rajaram, S., 2001. Relationships among Bread wheat international yield testing locations in dry areas. Crop Sci. 41, 1461–1469. Trethowan, R.M., van Ginkel, M., Ammar, K., Crossa, J., Payne, T.S., Cukadar, B., Rajaram, S., Hernandez, E., 2003. Associations among twenty years of Bread wheat yield evaluation environments. Crop Sci. 43, 1698–1711. Verma, R.P.S., Kharab, A.S., Singh, J., Kumar, V., Sharma, I., Verma, A., 2016. AMMI model to analyse GxE for dual purpose barley in multi-environment trials. Agric. Sci. Digest 36 (1).
Bhartiya, A., Aditya, J.P., Singh, K., Pushpendra, J., Purwar, P., Agarwal, A., 2017. AMMI & GGE biplot analysis of multi environment yield trial of soybean in North Western Himalayan state Uttarakhand of India. Legume Research-An International Journal 40 (2). Butler, D., 2009. ASReml-R Reference Manual. (Retrieved from). www.vsni.co.uk. Chauhan, J.S., Pal, S., Choudhury, P.R., Singh, B.B., 2016. All india coordinated research projects and value for cultivation and use in field crops in India: genesis, outputs and outcomes. Indian J. Agric. Res. 50 (6), 501–510. Cooper, M., DeLacy, I.H., 1994. Relationships among analytical methods used to study genotypic variation and genotype-by-environment interaction in plant breeding multi-environment experiments. Theor. Appl. Genet. 88 (1994), 561. Kelly, A., Smith, A., Eccleston, J., Cullis, B., 2007. The accuracy of varietal selection using factor analytic models for multi-environment plant breeding trials. Crop Sci. 47, 1063–1070. Kota, S., Singh, S., Mohapatra, T., Brajendra, Singh AM., Bhadana, V.P., Ravichandran, S., 2013. Genotype x environment interaxction analysis for grain yield in new plant type (NPT) wheat dwerivatives. SABRAO J. Breed. Genet. 45 (3), 382–390. Oladele, T.A., Kenamara, D.M., 2015. Trends in production and export of wheat in India.
213
Contents lists available at ScienceDirect
Field Crops Research journal homepage: www.elsevier.com/locate/fcr
An analysis of wheat yield and adaptation in India a,⁎
b
b
T b
b,f
Richard Trethowan , Ravish Chatrath , Ratan Tiwari , Satish Kumar , M.S. Saharan , Navtej Bainsc, V.S. Sohuc, Puja Srivastavac, Achla Sharmac, Nitish Ded, Surya Prakashe, G.P. Singhb, Indu Sharmab, Howard Eaglesg, Simon Diffeyh, Urmil Bansala, Harbans Barianaa a
The Plant Breeding Institute Sydney Institute of Agriculture, The University of Sydney, 107 Cobbitty Road, Cobbitty, NSW, 2570, Australia Indian Institute for Wheat and Barley Research, Agarsain Road, Karnal, 132001, India c Punjab Agricultural University, Ludhiana, 141004, India d Bihar Agricultural University, Sabour, Bhagalpur, Bihar, 813210, India e Birsa Agricultural University, Kanke, Ranchii, 834006, Jharkhand, India f Division of Plant Pathology, Indian Agricultural Research Institute, New Delhi, India g CSIRO Agriculture and Food, Black Mountain Science and Innovation Park, GPO Box 1700, Canberra, ACT, 2601, Australia h Centre for Bioinformatics and Biometrics, University of Wollongong, Northfields Avenue, Wollongong, NSW, 2522, Australia b
A R T I C L E I N F O
A B S T R A C T
Keywords: Genotype x environment interaction Wheat Multi-environment trials Factor analytic models
Multi-environment wheat trials provide valuable information on the extent of genotype x environment interaction, the stability of genotypes and define and confirm agro-ecological regions through associations among sites. The All India Coordinated Crop Improvement Project on wheat evaluates candidates for release across the wheat growing regions of India. To facilitate this process the wheat area is divided into six agro-ecological zones; the northwestern plains zone (NWPZ), the northeastern plains zone (NEPZ), the central zone (CZ), the peninsular zone (PZ), the northern hills zone (NHZ) and the southern hills zone (SHZ). Factor analytic (FA) models were used to analyze the genotype x environment interaction for yield of 813 wheat genotypes evaluated at 136 locations across the six agro-ecological zones in 1307 individual advanced variety trials between 2008/09 and 2012/13. Genotype x environment interaction was firstly assessed separately within each of the six established agroecological zones. Key locations with a high genetic correlation with all other locations within each zone were identified. Predicted genetic values of important cultivars that were represented in a wider range of environments within each zone were estimated and highly stable genotypes were found. Genotype x environment interaction was subsequently assessed across agro-ecological zones. Only those environments where the models accounted for > 99% of the genetic variance were retained for further analysis and two smaller zones (NHZ and SHZ) with little or no genotype congruence with other agro-ecological zones were removed. Thus 476 genotypes from 488 environments were included in the analysis. Fifteen clusters of environments with similar patterns of adaptation were found. These clusters were then characterized based on zonal classification, sowing time, irrigation regime, latitude and year and three regions broadly representing the main wheat growing areas of India were identified. These regions represent a combination of the NWPZ and NEPZ defined by latitude, a central region that combines CZ locations with northern PZ locations and a southern region comprised of southern PZ sites. Further stratification of these zones was then possible based on sowing time and irrigation practice. One cluster of 29 environments had a high average genetic correlation (r = 0.75) with most other environments and production zones. These represent key locations where larger numbers of entries might be grown in future seasons as they are the best predictors of yield across cropping zones.
1. Introduction Yield evaluation of new candidate crop genotypes in multi-environment trials is a key component of most varietal release systems ⁎
globally. These multi-environment experiments also provide estimates of genotype x environment interaction, genotype stability and the genetic relationships among varieties and environments (Cooper and DeLacy, 1994). However, large multi-environment data are rarely
Corresponding author. E-mail address: [email protected] (R. Trethowan).
https://doi.org/10.1016/j.fcr.2018.01.021 Received 20 November 2017; Received in revised form 18 January 2018; Accepted 22 January 2018 0378-4290/ © 2018 Elsevier B.V. All rights reserved.
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Fig. 1. Indian cropping zones and the location (blue dots) of AVT trials, 2008–2013. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
balanced. Factor analytic (FA) models have been used to estimate genotype x environment interaction in large unbalanced multi-environment data (Smith et al., 2015). The most general model for
estimating the between environment genetic variance matrix is an unstructured form that estimates a genetic variance for each environment and covariance between each pair of environments. However, as the
193
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
zones in the Advanced Variety Trial (AVT). These production zones include the central zone (CZ), the northeastern plains zone (NEPZ), the northern hills zone (NHZ), the northwestern plains zone (NWPZ), the peninsular zone (PZ) and the southern hills zone (SHZ) (Fig. 1). These data are not only used to make decisions on cultivar release in India but also provide a basis for understanding genotype x environment interaction and further refinement of India’s varietal release strategy. This study estimated genetic correlations between such environments to identify key sites representative of a larger group of environments. This information can be used to reduce the number of experimental sites and increase genotype concurrence among environments both within and between agro-ecological zones.
Table 1 Numbers of unique environments and genotypes evaluated for yield in AVT cropping zone trials, 2008–2013. Cropping zone
Unique Environments
Genotypes
CZ NEPZ NHZ NWPZ PZ SHZ
222 161 138 354 146 18
139 105 226 215 114 37
Note: CZ, NEPZ, NHZ, NWPZ, PZ and SHZ is the central zone, the northeast plains zone, the northern hills zone, the northwest plains zone, the plains zone and the southern hills zone, respectively.
2. Materials and methods Table 2 The range on yield and mean square errors of AVT trials within each cropping zone 2008–2013.
The yield data of AVT trials 2008–2013 were compiled by the Indian Institute for Wheat and Barley Research (IIWBR). The supplied data comprised trial mean yields (quintals/ha), coefficients of variation (CV’s) and predicted variety means and associated standard errors from the analysis of individual trial data. Yield data captured on 813 wheat genotypes evaluated at 136 locations in 1307 individual AVT yield trials between 2008/09 and 2012/13 were analyzed to estimate genotype x environment interaction using FA mixed models. The AVT yield data was captured from trials conducted in the six agro-ecological zones (CZ, NEPZ, NHZ, NWPZ, PZ and SHZ). Trials were managed using standard Indian government management practices recommended for each zone. Each trial/location/year combination is considered an environment and will henceforth be referred to as such. The distribution of environments within each zone classification is outlined in Fig. 1.
Range Cropping zone CZ NEPZ NHZ NWPZ PZ SHZ
Trial mean yield (q/ha) 10–75 14–60 10–63 12–68 10–61 27–63
Mean square error 59.3–0.17 38.8–0.05 59.3–0.37 31.5–0.16 28.9–0.04 85.1–5.87
Note: CZ, NEPZ, NHZ, NWPZ, PZ and SHZ is the central zone, the northeast plains zone, the northern hills zone, the northwest plains zone, the plains zone and the southern hills zone, respectively.
2.1. Analysis of multi-environment trials within specific agro-ecological zones
size of the multi-environment data increases it becomes difficult to fit the model and estimate variance parameters. The FA model provides an alternative to the unstructured form and can efficiently account for the genetic covariance between environments (Smith et al., 2001; Kelly et al., 2007). India is the second biggest producer of wheat globally and annual production regularly exceeds 90 million tonnes (Oladele and Kenamara, 2015). A number of Indian wheat breeding programs produce new improved cultivars for different agro-ecological zones and candidate genotypes with potential for release to farmers are evaluated nationally in coordinated trials. The official testing of candidate wheat cultivars in India is conducted by the All India Coordinated Crop Improvement Project on wheat (Chauhan et al., 2016). This body oversees the conduct of multi-environment wheat yield trials managed by the Indian Council of Agricultural Research (ICAR) Institutes, State Agricultural Universities and institutes governed by state agriculture departments. The top-tier of this yield evaluation is conducted across six production
Separate analyses within each agro-ecological zone were performed as genotypes were specifically deployed in trials in each zone. A summary of the numbers of unique environments and genotypes evaluated in 2008–2013 is given in Table 1. The allocation of genotypes to trials was highly structured and groups of trials within each year and zone had satisfactory genotype concurrence. However, limited concurrence was available between pairs of zones and across years. For this reason FA mixed models were used to estimate the genetic covariance between environments. Weights were applied to the data because of the variability in trial mean square error (Table 2). The weights used were the reciprocal of the estimated variances associated with the predicted variety means from individual trial analyses. The FA mixed models were then fitted to the data from each zone in two stages. In the first stage a simple variance components model was fitted to the data (referred to as the ID model) followed by a model which accommodated variance heterogeneity between
Table 3 Summary of models fitted FA1 through to FA6, the residual log-likelihood (LL) and percentage of genetic variance accounted for (vaf) in each cropping zone. CZ Model FA1 FA2 FA3 FA4 FA5 FA6
LL −5156 −4850 −4619 −4335
NEPZ vaf 35 56 68 80
LL −3036 −2877 −2725 −2575
NHZ vaf 33 56 71 80
LL −3351 −3214 −3093 −2961
NWPZ vaf 41 62 77 88
LL −8907 −8484 −8095 −7701 −7280 −6902
PZ vaf 32 50 61 72 79 86
LL −2589 −2389 −2246 −2109
SHZ vaf 44 66 80 89
LL −268 −253
vaf 52 82
Note: CZ, NEPZ, NHZ, NWPZ, PZ and SHZ is the central zone, the northeast plains zone, the northern hills zone, the northwest plains zone, the plains zone and the southern hills zone, respectively.
194
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
environments, but not the covariance between environments (referred to as the DIAG model). Based on the DIAG model, environments that had no genetic variance were removed from the analyzed data set. In the second stage a series of FA mixed models starting with an FA model of order 1 through to an FA model of order 6, if required, was fitted to the data. A summary of the FA models fitted for each zone is provided in Table 3. The predicted regression part of the FA model was used to estimate genotype performance in each environment using the same principle as Smith et al. (2015). These predictions are referred to as genetic values (GV’s). All models were fitted using the statistical software package asreml (Butler, 2009) within the R (R Core Team, 2015) computing environment. The genetic correlation matrix between environments was then estimated. However, due to the large number of environments within each zone, an agglomerative hierarchical clustering algorithm was used to reorder the rows and columns of the genetic correlation matrix to better observe any structure. The environments within each zone that most closely represented the wider zone were then identified by constructing dendrograms of those environments were the final FA model accounted for at least 99% of the genetic variance.
Table 6 Summary of factor analytic (FA) mixed models fitted to the four AVT cropping zones.
NHZ
NWPZ
PZ
SHZ
CZ NEPZ NHZ NWPZ PZ SHZ
139 5 1 22 29 0
5 105 0 32 7 0
1 0 226 1 0 0
22 32 1 215 13 1
29 7 0 13 113 0
0 0 0 1 0 37
Cropping Zone NWPZ 198
42 65 80 92
Trial mean square error and trial mean yield varied considerably within cropping zones (Table 2). The biggest difference in mean square error was observed in the NEPZ where the largest error (38.8) was approximately 800 times the smallest mean square error (0.05). The smallest range in mean square errors was observed in the SHZ. The genetic variance accounted for by the FA4 model varied from 6% to 100% across all zones. There were 125 environments across all zones where the percentage variance accounted for was < 50% and the genetic correlations and predicted genotype performance associated with these environments was reduced compared to 592 environments where the FA4 model accounted for > 99% of the genetic variance. The FA4 model accounted for between 72 and 89% of the average genetic variance at an environment (Table 3). However, the FA4 model was less efficient in the NWPZ (72%) and the FA6 model (86%) was used instead. The reordered genetic correlation matrix was constructed for each zone (not shown) and some underlying structure in the genetic correlation between environments was observed due to the limited genotype concurrence between many environments. Dendrograms were constructed (not shown) based on the hierarchical clustering of the dissimilarity matrix associated with the genetic correlations among environments to examine site/experiment similarities. Those sites, represented by more than one experiment, among the group least different to all other sites within each zone were identified (Table 8). In the CZ, Indore and Bhopal were the best predictors of zone performance. Kalyani, Kapurthala and Niphad best represented the NEPZ, NWPZ and PZ. Two sites, Wellington and Thalawadi, were the most representative sites in the SHZ. Important widely grown cultivars in each of the different cropping zones were subsequently selected for further analysis (Table 9) and the
Table 5 The number of environments in the Indian AVT cropping regions, 2008–2013 .
NEPZ 88
−9592 −8985 −8330 −7565
3.1. Multi-environment trials within agro-ecological zones
Note: CZ, NEPZ, NHZ, NWPZ, PZ and SHZ is the central zone, the northeast plains zone, the northern hills zone, the northwest plains zone, the plains zone and the southern hills zone, respectively.
CZ 107
FA1 FA2 FA3 FA4
3. Results
Table 4 Genotype concurrence between Indian AVT wheat cropping zones 2008–2013. NEPZ
vaf
weights based on the standard errors associated with the predicted variety means to model genetic covariances between environments. A series of FA mixed models were fitted starting with a FA model of order 1 through to a FA model of order 4. A summary of the FA models fitted is provided in Table 6. The FA4 model accounted for, on average, 92% of the genetic variance at an environment. However, the genetic variance accounted for by the FA4 model varied from 10% to 100%. There were 30 environments where the genetic variance accounted for was less than 50% (Table 7). Twenty-six of these 30 environments were located in the NWPZ and these environments were subsequently removed from the analysis. However, 385 environments where the FA4 model accounted for > 99% of the genetic variance were retained. Of these environments, 135 were late sown and irrigated, 127 timely sown and irrigated and 123 were timely sown in rainfed conditions. All models were fitted using the statistical software package asreml (Butler, 2009) within the R (R Core Team, 2015) computing environment.
Genotype concurrence between agro-ecological zones was relatively low (Table 4). A FA mixed model was fitted to the trial data from each zone (data not shown) and those environments where the models accounted for more than 99% of the genetic variance were retained for analysis. The data comprised 7502 yield observations on 736 genotypes grown in 592 environments. As little or no genotype concurrence between NHZ and SHZ and the other cropping zones was observed (Table 4), these two zones were removed from the analysis. After dropping NHZ and SHZ the data file comprised 476 varieties over 488 environments. The number of genotypes in CZ, NEPZ, NWPZ and PZ are the respective diagonal values in Table 4 and the number of environments in each zone is presented in Table 5. Data were analyzed using a two-stage FA mixed model following the approach of Smith et al. (2015). This approach used the predicted variety means from analyses of individual trial data and calculated
CZ
LL
Note: “LL” refers to the residual (or restricted) log-likelihood and “vaf” refers to the percentage of genetic variance accounted for by each model.
2.2. Analysis of multi-environment trials across agro-ecological zones
Zone
Model
PZ 95
Note: CZ, NEPZ, NWPZ and PZ is the central zone, the northeast plains zone, the northwest plains zone and the plains zone, respectively.
195
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
yielding (Fig. 2B). In the CZ, the genotypes GW322 and GW366 were not as stable as HI1544 (Fig. 2C). Only two genotypes were selected in the small SHZ (HW5216 and CO(W)1). There was significant cross-over interaction observed and while HW5216 was more stable, the genotype CO(W)1 showed specific adaptation to three key SHZ environments (Fig. 2D). In the NHZ, the genotype VL907 was generally below the mean across all sites and HPW349 and HS507 generally superior or equivalent to the mean and were therefore considered more stable (Fig. 2E). The genotypes MACS6222, NIAW917 and RAJ4037 were selected in the PZ (Fig. 2F). However, no genotype was particularly stable, although NIAW917 had generally higher predicted values across environments.
Table 7 Thirty environments from the MET analysis across cropping zones and years that explained < 50% of the genetic variance. Location
Zone
Sowing time
Irrigation
Year
Genetic variance%
Ambala Balachaur Bareilly Bawal Bulandshahar Bulandshahar Bulandshahar Bulandshahar Bulandshahar Delhi Diggi Durgapura Durgapura Gurdaspur Gurdaspur Gwalior Hisar Kalyani Karnal Karnal Kaul Ludhiana Malda Modipuram Modipuram Rauni Tabiji Tabiji Ujhani Varanasi
NWPZ NWPZ NWPZ NWPZ NWPZ NWPZ NWPZ NWPZ NWPZ NWPZ NWPZ NWPZ NWPZ NWPZ NWPZ CZ NWPZ NEPZ NWPZ NWPZ NWPZ NWPZ NEPZ NWPZ NWPZ NWPZ NWPZ NWPZ NWPZ NEPZ
TS TS LS TS LS LS LS TS TS TS TS TS TS LS TS LS TS TS LS TS TS TS TS TS TS TS LS TS TS LS
RF RF IR IR IR IR IR IR IR IR RI IR IR IR IR IR RF IR IR IR IR IR IR RF RF IR IR IR IR IR
2009 2011 2011 2008 2009 2012 2010 2012 2010 2008 2012 2009 2010 2008 2010 2011 2008 2008 2009 2012 2009 2011 2009 2011 2009 2010 2011 2008 2011 2012
13 49 46 37 35 40 45 26 49 32 10 34 47 43 18 30 27 29 20 19 34 26 45 48 49 46 44 46 20 17
3.2. Multi-environment trials across agro-ecological zones A plot of trial mean square error against trial mean yield is presented in Fig. 3. Both the trial mean square error and trial mean yield varied considerably, with the largest mean square error (38.79) more than 800 times larger than the smallest mean square error (0.045). A heat-map of the genetic correlations between the 385 environments where the FA4 model accounted for > 99% of the genetic variance is presented in Fig. 4 and the details of each environment in Appendix A Table A1. The trials specific to individual cropping zones are represented by boxes within the heat-map. There appeared to be little pattern to the genetic correlations across the 385 environments. A dendrogram was constructed (Fig. 5) based on the hierarchical clustering of the dissimilarity matrix associated with the genetic correlations in Fig. 4. Fifteen environment clusters representing similar patterns of adaptation were identified.
Note: CZ, NEPZ, NWPZ and PZ is the central zone, the northeast plains zone, the northwest plains zone, and the plains zone, respectively. TS, LS, IR, RF and RI and timely sown, late sown, irrigated, rainfed and restricted irrigation, respectively.
3.2.1. Classification of all environment clusters The 15 clusters defined in Fig. 5 were then characterized on the basis of dominant zonal representation, sowing time, irrigation, latitude and year (Table 10). A full description of each environment is found in Appendix A Table A1. A single zonal classification dominated in 7 of the 15 clusters (clusters 2, 5, 6, 10, 12, 13 and 14). The NWPZ dominated in cluster 12, the NEPZ in cluster 10, the PZ in clusters 5 and 6 and the CZ in clusters 2, 13 and 14. However, 5 clusters showed no zonal influence (3, 7, 8, 9 and 11) and 3 were dominated by two zones (1, 4 and 15). Sowing time also influenced the clustering of environments. Seven clusters were predominately timely sown (5, 7, 9, 10, 12, 14 and 15) and 3 late sown (2, 6 and 13). Five were classified as both timely and late sown (1, 3, 4, 8 and 11). Nine clusters were dominated by irrigated sites (1, 2, 3, 6, 7, 8, 9, 12 and 13) and only one by drought affected locations (cluster 5). A further five clusters comprised a significant proportion of both irrigated and drought affected environments (4, 10,
predicted genetic value of these genotypes in specific groups of environments was estimated (Fig. 2A–F). These trials/environments were chosen because, apart from the selected genotypes, there were other genotypes in common. Differences in the stability of different genotypes were noted in each zone. Four common and frequently repeated genotypes were selected in the NWPZ (Fig. 2A). Of these, PBW343 was the most unstable genotype and HD2967 the most stable and high-yielding. The adaptation of the three selected genotypes in the NEPZ (HD2967, K0307 and PBW343) was generally similar and the yield relatively stable with the exception of PBW343 in one environment, where this genotype was very high-
Table 8 Key locations least different to all others within each agro-ecological zone; those locations represented by more than one experiment are listed. Agro-ecological zone CZ
NEPZ
NHZ
NWPZ
PZ
SHZ
Indore (3) Bhopal (2)
Kalyani (4) Faizabad (2) Deegh (2) Shillongani (2)
Shimla (7) Hawalbagh (5) Bajuara (4) Kalimpong (2) Majhera (2) Rajouri (2)
Kapurthala (3) New Delhi (2) Gurdaspur (2) Balachaur (2) Karnal (2) Uchani (2)
Niphad (5) Kalloli (4) Karad (3) Prabhani (3) Muhol (2) Pune (2) Arbhavi (2)
Wellington (2) Thalawadi (2)
Note: CZ, NEPZ, NHZ, NWPZ, PZ and SHZ is the central zone, the northeast plains zone, the northern hills zone, the northwest plains zone, the plains zone and the southern hills zone, respectively. Numbers in parentheses indicate the number of occurrences.
196
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
4. Discussion
Table 9 Selected widely grown genotypes in each cropping zone (year of release in brackets).
Very little has been published on the extent of genotype x environment interaction in major crops in India. Those studies that have been reported are generally balanced, small in scale and assess interactions using GGE or AMMI analyses (Bhartiya et al., 2017; Verma et al., 2016; Kota et al., 2013). Trethowan et al. (2001, 2003) used the shifted multiplicative model to examine associations among global wheat yield testing environments including sites in India. While such analyses identified site associations between Indian locations and some global sites and identified Pirsabak in Pakistan as a key predictor of global wheat yield, very little is known about the extent of genotype x environment interaction for wheat on the Indian subcontinent. The distribution of AVT trial sites across India between 2008 and 2013 sampled most of the national wheat growing environments. However, the multi-environment analysis across agro-ecological zones revealed a lack of genotype concurrence across zones and years. This lack of connectivity impeded the identification of key sites and germplasm. However, limited concurrence between trials did not preclude estimating the genetic covariance between trials using FA mixed models, although those sites that accounted for < 99% of the genetic variance were excluded. Twenty-six of the 30 environments that accounted for < 50% of the genetic variance were located in the NWPZ and half were timely sown and irrigated. A lack of range in genotype performance in these high-yielding environments (ie all genotypes were high-yielding) may have accounted for the lower than expected genetic variances in the NWPZ; India’s highest yielding production environment. Nevertheless, 385 environments of 488 (excluding the NHZ and SHZ because of lack of genotype concurrence with other zones) accounted for > 99% of the genetic variance and thus provided sufficient scale for the estimation of genetic correlations. The 15 individual site clusters generated across the entire data set partially confirmed the ICAR classification of agro-ecological zones.
Cropping Zone CZ
NEPZ
GW322 (2002) HD2967 (2011) GW366 (2006) KO307 (2006) HI1544 (2007) PBW343 (1995)
NHZ
NWPZ
PZ
SHZ
HPW349 (2012) HS507 (2010) VL907 (2010)
DBW17 (2006) HD2967 (2011) PBW343 (1995) PBW550 (2006)
MACS6222 (2001) NIAW917 (2005) RAJ4037 (2002)
CO(W)1 (2005) HW5216 (2012)
Note: CZ, NEPZ, NHZ, NWPZ, PZ and SHZ is the central zone, the northeast plains zone, the northern hills zone, the northwest plains zone, the plains zone and the southern hills zone, respectively.
11, 14 and 15). Latitude was also influential on site clustering with four clusters on average < 23° of latitude (5, 6, 8 and 13), eight between 23° and 25.9° (1, 2, 3, 4, 7, 11, 14 and 15) and three > 25.9° (9, 10 and 12). The year of assessment did not generally influence site clustering.
3.2.2. Identification of key environments Twenty-nine environments in cluster 8 located the middle of the dendrogram in Fig. 5 were identified as the group which best correlated with most other environments. Within this group the genetic correlations between environments ranged from 0.22 to 0.99 with an average of 0.75. These 29 environments are located in 26 physical locations based on longitude and latitude (Table 11) and are possible locations where larger numbers of entries might be grown in future seasons as they are the best predictors of yield across cropping zones.
Fig. 2. Predicted genetic values (q/ha), and associated standard errors for key cultivars in each (A) NWPZ, (B) NEPZ, (C) CZ, (D) SHZ, (E) NHZ and (F) PZ in common experiments.
197
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Fig. 3. Trial mean square error (MSE) and trial mean yield (TMY) from individual trial analyses.
Fig. 4. Heatmap of the reordered genetic correlation matrix between the 385 environments where the FA4 model accounted for more than 99% of the genetic variance.
198
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Fig. 5. Dendrogram based on the hierarchical clustering of the dissimilarity matrix associated with the genetic correlations between the 385 environments where the FA4 model accounted for more than 99% of the genetic variance. The key group of 29 locations (cluster 8) indicated with a blue line. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
Table 10 Summary of the dendrogram groupings in Fig. 4 (from left to right) based on zonal representation, sowing time, irrigation, latitude and year. Group number
Number of sites
Zonal representation
Percentage timely sown
Percentage Irrigated
Average Latitude
Year
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
19 8 45 34 19 34 26 29 19 39 9 17 19 29 42
74% (NWPZ/NEPZ) 50% CZ No zonal influence 65% (PZ/CZ) 58% PZ 44% PZ No zonal influence No zonal influence PZ not represented 46% NEPZ; 5% PZ No zonal influence 47% NWPZ 47% CZ 41% CZ 71% (NWPZ and PZ)
63 38 53 59 84 41 69 62 68 84 66 71 36 72 83
89 100 71 65 26 70 77 92 74 56 66 88 79 59 47
25.3 23.8 23.9 23.1 21.9 22.1 24.8 22.6 26.0 26.3 25.8 26.1 22.5 24.4 24.2
No clear influence No clear influence No clear influence 35% of sites 2012 No clear influence 35% of site 2012 54% of sites 2010 No clear influence No 2011 sites No clear influence No clear influence 2008 not represented No clear influence Few 2012 sites No clear influence
Note: CZ, NEPZ, NWPZ and PZ is the central.
differentiation based on sowing time and irrigation. Five clusters did not have a dominant zone including the key cluster 8 that correlated best with all other environments. These sites tended to cluster on the basis of sowing time, irrigation and latitude but could be found spread across the zonal classifications. The multi-environment analysis suggests that three primary regions of adaptation exist across the main wheat growing areas of India. These are a combined NWPZ/NEPZ region defined by latitude, a central region that combines CZ locations with northern locations in the PZ and a southern region comprised of southern PZ sites. Further stratification of these regions is then possible based on sowing time and irrigation practice. The multi-environment analysis of individual zones provided additional and valuable information on key zonal locations and genotype adaptation. Due to the large number of genotypes evaluated in each
Cluster 10, dominated by NEPZ locations, was generally timely sown with a mix of irrigated and drought stressed conditions, all at higher latitude (> 26°). Timely sown sites in this zone also clustered with NWPZ sites (cluster 1) which was predominately timely sown and irrigated, indicating that sowing time and latitude were probable drivers of adaptation across the NWPZ and NEPZ. The three primary clusters of CZ sites (clusters 2, 13 and 14) could be differentiated into those that were late sown and irrigated (2 and 13) and those that were timely sown and a mixture of irrigated and drought stressed conditions (14). Interestingly, the CZ sites in cluster 4 (comprised of PZ and CZ sites) are primarily late sown and irrigated, similar to clusters 2 and 13. Similarly, the two PZ clusters (5 and 6) are differentiated by sowing time and irrigation with cluster 5 predominantly timely sown and rainfed and cluster 6 late sown and irrigated. Low latitude appears to be a primary driver of adaptation in the PZ zone with further
199
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Table 11 Twenty nine environments identified as the cluster (8) which best correlates with most other environments and whose locations are possible candidates for growing larger numbers of entries in future seasons. Superscripts a, b and c identify three pairs of trials in the same location. Zone
Sowing time
Irrigation
Year
Location
Latitude
Longitude
TMY
CZ CZ CZ CZ CZ CZ CZ CZ CZ CZ CZ CZ NEPZ NEPZ NEPZ NEPZ NEPZ NEPZ NWPZ NWPZ NWPZ NWPZ NWPZ NWPZ PZ PZ PZ PZ PZ
TS TS TS TS TS TS TS LS TS LS LS LS TS TS TS TS TS TS TS TS TS LS LS TS LS TS TS TS LS
IR IR IR RI RI RF RF IR RI IR IR IR RF IR IR RF IR RF IR RF RI IR IR IR IR RF RF IR IR
2011 2009 2011 2008 2011 2008 2010 2009 2008 2008 2009 2011 2011 2012 2011 2010 2010 2009 2008 2010 2009 2008 2009 2010 2008 2008 2008 2010 2008
Amreli SK Nagar Junagarh Dhandhuka Dhandhuka Indore Indore Anand Arnej Vijapur Banswara Rewa Varanasi Saini Kanpur Faizabad Araul Barabanki Bulandshahar Hisar Hisar Kashipur Kaul Chatha Karad Parbhani Washim Niphad Amravati
20.86 21.18 21.31 22.37 22.37 22.37 22.37 22.55 22.58 23.35 23.55 24.53 25.20 25.66 26.29 26.47 26.92 26.92 28.41 29.10 29.10 29.22 29.84 32.43 17.28 19.08 20.10 20.60 20.92
70.75 72.86 70.33 71.98 71.98 75.75 75.75 72.95 72.26 72.55 74.45 81.30 83.03 81.32 80.18 82.80 80.01 81.20 77.83 75.46 75.46 78.95 76.66 74.54 74.20 76.50 77.13 74.60 77.76
63.88 47.28 42.78 26.00 24.36 10.70 22.29 37.18 27.60 36.60 51.15 49.72 26.81 52.13 48.08 33.08 40.56 14.58 43.20 31.17 40.17 41.80 43.87 40.73 47.80 13.10 13.20 39.59 28.60
Note: CZ, NEPZ, NHZ, NWPZ, PZ and SHZ is the central zone, the northeast plains zone, the northern hills zone, the northwest plains zone, the plains zone and the southern hills zone, respectively
(Table 7) and was subsequently discarded. For this reason the selection of Bulandshahar as a key location must be treated with caution. Hisar could be considered a suitable location for expanded evaluation of materials for drought adaptation as two water-limited trials in consecutive years explained > 99% of the genetic variance. Five key sites were identified in the PZ and of these only one was timely sown and irrigated (Niphad). Sites in this zone tend to better predict yield under stress across the Indian cropping zones. In summary, 9 of the 29 key locations represented the highest yielding conditions, considered to be timely sown and irrigated, with at least one site located in each of the 4 key agro-ecological zones. Twelve timely sown and water restricted environments (either rainfed or restricted irrigation) were identified in all four primary agro-ecological zones. However, of the 8 late sown/irrigated key environments, none were located in the NEPZ and most were found in the CZ. Nevertheless, the distribution of these 3 environment types (timely sown/irrigated; timely sown/water restricted; late sown/irrigated) across all cropping zones further emphasizes the value of these locations as predictors of yield in the Indian AVT network.
zone, the key genotypes, selected on the basis of their higher concurrence and general importance as major regional cultivars, provided a summary of adaptation in each zone. While little concurrence of major cultivars among cropping zones was evident, there were two exceptions. The cultivars PBW343 and HD2967 occurred in both the NWPZ and NEPZ. In both environments the yield of PBW343 was least stable and HD2967 most stable. PBW343 is an older cultivar (released in 1995) and is now susceptible to stripe rust and this most likely explains the lack of stability in both zones. In general, the most stable cultivars were more recently released and included HI1544, HD2967, K0307, HPW349, HS507, NIAW917 and HW5216. Given the expense of sowing advanced variety trials nationally and the general lack of genotype concurrence in the AVT, the identification of key locations that correlate well with the wider production area could be used to evaluate greater numbers of genotypes and to provide greater connectivity. Of the 29 locations identified in Table 10, twelve were located in the CZ. Of these, 6 environments representing 4 locations (Dhandhuka, Indore, Anand and Arnej) were water stressed and suitable for expanding genotype evaluation for rainfed conditions, whereas the remaining locations (Amreli, Nagar, Junagarh, Vijapur, Banswara and Rewa) best represented irrigated areas. The 6 NEPZ locations were all timely sown and of these 3 represented rainfed and 3 irrigated conditions. Of the 6 NWPZ key locations, only two (Bulandshahar and Chatha) were timely sown and irrigated − representing the most probable environment type in this high-yielding region of India. Interestingly, the key Bulandshahar trial was sown in 2008, however, this location explained < 50% of the genetic variance in all other years
Acknowledgements The authors wish to acknowledge the funding received from the Australian Centre for International Agricultural Research and the Indian Council for Agricultural Research that made this work possible. We also acknowledge the contribution of the many wheat researchers of India who all contributed to the AVT database over many years.
200
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Appendix A
Table A1 All sites/experiments used to define the 15 clusters defined in Fig. 5.
(continued on next page)
201
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Table A1 (continued)
(continued on next page)
202
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Table A1 (continued)
(continued on next page)
203
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Table A1 (continued)
(continued on next page)
204
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Table A1 (continued)
(continued on next page)
205
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Table A1 (continued)
(continued on next page)
206
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Table A1 (continued)
(continued on next page)
207
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Table A1 (continued)
(continued on next page)
208
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Table A1 (continued)
(continued on next page)
209
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Table A1 (continued)
(continued on next page)
210
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Table A1 (continued)
(continued on next page)
211
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
Table A1 (continued)
212
Field Crops Research 219 (2018) 192–213
R. Trethowan et al.
References
Int. Res. J. Agric. Econ. Stat. 6 (1), 189–192. R Core Team, 2015. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria Retrieved from. https://www. R-project.org/. Smith, A., Cullis, B., Thompson, R., 2001. Analyzing variety by environment data using multiplicative mixed models and adjustments for spatial field trend. Biometrics 57, 1138–1147. Smith, A., Ganesalingam, A., Kuchel, H., Cullis, B., 2015. Factor analytic mixed models for the provision of grower information from national crop variety testing programs. Theor. Appl. Genet. 128 (1), 55–72. Trethowan, R.M., Crossa, J., van Ginkel, M., Rajaram, S., 2001. Relationships among Bread wheat international yield testing locations in dry areas. Crop Sci. 41, 1461–1469. Trethowan, R.M., van Ginkel, M., Ammar, K., Crossa, J., Payne, T.S., Cukadar, B., Rajaram, S., Hernandez, E., 2003. Associations among twenty years of Bread wheat yield evaluation environments. Crop Sci. 43, 1698–1711. Verma, R.P.S., Kharab, A.S., Singh, J., Kumar, V., Sharma, I., Verma, A., 2016. AMMI model to analyse GxE for dual purpose barley in multi-environment trials. Agric. Sci. Digest 36 (1).
Bhartiya, A., Aditya, J.P., Singh, K., Pushpendra, J., Purwar, P., Agarwal, A., 2017. AMMI & GGE biplot analysis of multi environment yield trial of soybean in North Western Himalayan state Uttarakhand of India. Legume Research-An International Journal 40 (2). Butler, D., 2009. ASReml-R Reference Manual. (Retrieved from). www.vsni.co.uk. Chauhan, J.S., Pal, S., Choudhury, P.R., Singh, B.B., 2016. All india coordinated research projects and value for cultivation and use in field crops in India: genesis, outputs and outcomes. Indian J. Agric. Res. 50 (6), 501–510. Cooper, M., DeLacy, I.H., 1994. Relationships among analytical methods used to study genotypic variation and genotype-by-environment interaction in plant breeding multi-environment experiments. Theor. Appl. Genet. 88 (1994), 561. Kelly, A., Smith, A., Eccleston, J., Cullis, B., 2007. The accuracy of varietal selection using factor analytic models for multi-environment plant breeding trials. Crop Sci. 47, 1063–1070. Kota, S., Singh, S., Mohapatra, T., Brajendra, Singh AM., Bhadana, V.P., Ravichandran, S., 2013. Genotype x environment interaxction analysis for grain yield in new plant type (NPT) wheat dwerivatives. SABRAO J. Breed. Genet. 45 (3), 382–390. Oladele, T.A., Kenamara, D.M., 2015. Trends in production and export of wheat in India.
213