Breeding indicators for high-yielding field pea under normal and heat stress environments

Abstract The warming Canadian summers have become a major abiotic stress to crops, including pea. In the past decade, attempts were made in the understanding of heat stress effect and genomic mapping for heat-responsive traits in field pea. In this study, a new recombinant inbred line population (PR-24) consisting of 39 lines was tested in 6 trials in the summers of 2020 (near normal weather conditions) and 2021 (hot/dry conditions). PR-24 was phenotyped for days to flowering (DTF), days to maturity, plant height, lodging, yield components, plot yield, and seed quality traits. Plant height could be an effective indicator for yield prediction, because its correlation with plot yield was significantly positive in all six trials despite varying degrees of heat and drought stress. Under normal summer weather conditions in 2020, relatively late maturity was correlated with greater seed yield; under heat/drought stress conditions in 2021, successful pod development on the main stem was important for final plot yield. Linkage mapping was used to dissect the genomic regions associated with the measured traits. Four QTLs were identified over multiple trials, one each for DTF (chromosome 7), reproductive node number (chromosome 5), pod number (chromosome 2), and seed protein concentration (chromosome 5). Furthermore, two indices, i.e., stress tolerance index and geometric mean yield, previously used in drought tolerance assessment were validated as useful criteria for heat tolerance assessment in this study.


Introduction
Heat stress (HS) refers to the detrimental effects of temperatures beyond the upper-temperature threshold of a plant's normal growth and development. In western Canada, the threshold maximum temperature for pea yield reduction in the field was approximately 28 • C daily maximum at reproductive stage and above 17.5 • C mean seasonal daily temperature (Bueckert et al. 2015). Huang et al. (2017) also observed that pea yield was negatively correlated with mean daily maximum temperatures at the flowering stage (r = −0.79) ranging from 23 to 26 • C among five field trials in 2013 and 2014. Vegetative traits and reproductive traits of pea, e.g., leaf surface wax, chlorophyll concentration, days to flowering, and pod number, were highly responsive to HS and varied greatly among pea accessions (Huang et al. 2017;Tafesse et al. 2020Tafesse et al. , 2021. Reproductive development is exposed to a higher risk of HS damage than vegetative growth, because reproductive organs are more heat susceptible (Luo 2011), and the time window of flowering and pod development often coincides with heat waves in July and August in pea-growing regions in Canada. High ambient temperature at flowering stage shortened the flowering duration of pea crops, causing young pod abortion, and thereby reduced seed production (Huang et al. 2017). Lodging tolerance contributes to heat tolerance in pea germplasm. Tafesse et al. (2019) concluded that semi-leafless pea accessions had cooler canopy temperatures under HS than normal leaf-type accessions due to better lodging resistances. However, these researchers did not further elucidate the likely relationship of these heat-responsive traits with plot yield, which is generally the core interest in practical pea breeding.
Yield performance of crop varieties can be quite inconsistent among different stress environments. Based on yield, Fernandez (1992) divided genotypes into four groups: (a) genotypes with consistently high yields under both stress and control environments, (b) genotypes with desirable yields only under control condition, (c) genotypes having relatively higher yields under stress conditions, and (d) low yield under both conditions. Several indices for drought stress were suggested by different researchers to distinguish the four groups and were validated for their effectiveness, for instance, stress susceptibility index in pea (Grzesiak et al. 1996). However, an index is still lacking in HS tolerance assessment. Drought often occurs with HS in Canadian peaproduction regions, as well as in most cool-season legume production areas in other countries. Thus, the potential would be high that indices for drought could be applied in HS.
Traits related to flowering, yield, and other agronomic traits were confirmed as quantitative traits that were controlled by multi-gene action (Huang et al. 2017;Jiang et al. 2017;Tafesse et al. 2020). To gain genetic knowledge of quantitative traits, a useful method is to build a genetic linkage map and thereby characterize corresponding QTLs via the integration of phenotypic data. Huang et al. (2017) identified 10 consistent QTLs associated with flowering and yieldcomponent traits, in particular, a QTL for DTF on chromosome 1 consistent across four field trials varying in temperature stresses. The QTLs for flowering duration, thousand seed weight, and reproductive node number (RNN) were different between normal and late seeding, which implies that different mechanisms were involved under contrasting temperature environments. However, different genomic loci were identified for the same traits in different genetic backgrounds (Jiang et al. 2017;Tafesse et al. 2020).
In the current study, a new pea recombinant inbred line (RIL) population, namely PR-24, was used to evaluate associations of various field traits with seed yield in field trials varying in heat and drought stresses. Knowledge of the genetic control of these key traits will inform pea-breeding strategies. Additionally, PR-24 was used for QTL analysis of days to flowering, yield component traits, other agronomic traits, as well as seed composition traits. It was hypothesized that some of the identified QTLs would be at similar genome loci reported in other pea populations, especially in Canadian pea populations. We also expected that novel QTLs would be identified in the current population.

RIL population development
A population of 39 RILs, namely PR-24, was developed from the cross of PR11-2 and CDC Amarillo at the Crop Development Centre, University of Saskatchewan. Each line was derived from a single F2 seed, and the generations were advanced to F7 by single seed descent in the Agriculture Greenhouse, University of Saskatchewan. PR11-2 is a green pea variety developed from the cross of yellow pea cultivar CDC Centennial (Warkentin et al. 2007) and green pea cultivar CDC Sage . It is considered to be a heattolerant variety due to consistently high yield potential under normal and HS field environments . CDC Amarillo is one of the best-yielding yellow pea varieties in western Canada (Warkentin et al. 2014).

Field trials
Evaluation of PR-24 was conducted in field trials starting in 2020 at Saskatoon (52 • 12 N, 106 • 63 W; Dark Brown Chernozem), Rosthern (52 • 66 N, 106 • 33 W; Orthic Black Chernozem), and Lucky Lake (50 • 98 N, 107 • 13 W; Dark Brown Chernozem) in Saskatchewan, Canada. Within each site, a randomized complete block design with three blocks was used. Eighty-four seeds of individual RILs were planted in 1 m × 1 m micro-plots with 3 rows with 30 cm row spacing. The field trial was managed via best management practices for pea in western Canada. Due to unexpected early season sandblasting at Saskatoon, plants were heavily damaged and dis-carded from further analysis. The field trial was repeated in 2021 at Saskatoon and Rosthern. At each location, a normal seeding date and late seeding date trial with three replicates were planted. The late seeding date was expected to shift the flowering period to a higher temperature/more stressful period in late July. Collectively, six sets of field data were produced over two years' field experiments, namely 2020 Rosthern (2020_ROS), 2020 Lucky Lake (2020_LL), 2021 Saskatoon early (2021_SAS_E), 2021 Saskatoon late (2021_SAS_L), 2021 Rosthern early (2021_ROS_E), and 2021 Rosthern late (2021_ROS_L).

Phenotyping
Starting from mid-June, the field trials were visited on a weekly basis. DTF was noted when half of the plants in a microplot started to flower. When plant development was near physiological maturity, three single plants with average height were randomly picked within individual micro-plots, whose RNN and pod number (PN) on the main stem were manually counted. Days to maturity (DTM) was noted for each RIL when 80% of the pods in a microplot had turned brown. In addition, the grain yield of each RIL micro-plot was measured after combine harvest. Then, seed protein and starch concentration were predicted by an FOSS NIR System Model 6500 near infrared spectrophotometer (FOSS, Hilleroed, Denmark) using in-house equations for protein and starch. Moisture was tested on a subset of samples (AACC Method 44.17.01), and all data were reported on a dry matter basis (Arganosa et al. 2006).

Heat-tolerance indices' calculation
Based on their reported effectiveness for predicting crop tolerance under abiotic stress in previous literature, five abiotic stress indices were evaluated in this study, and their calculation formulas are listed below.

Genotyping
Genotyping was conducted at the Pulse Crop Molecular Breeding Lab, University of Saskatchewan. DNA of individual RILs and two parents was extracted using DNeasy Plant Mini Kit (Qiagen, Hilden, Germany), and DNA concentration was quantified at optical density 260 nm using a Nanodrop 8000 UV spectrophotometer. Then, DNA stock was diluted to 10 ng μL −1 standard concentration for KASP array genotyping (LGC, Teddington, UK).
One hundred and seventy-seven genome-wide KASP markers were tested for polymorphism between the two parental lines, and 88 markers were shown as being putatively polymorphic. These 88 markers were subsequently used to genotype the 39 RILs. KASP assays were prepared in 384-well plates, and in a single well, 3 μL DNA of an individual RIL was mixed with 3 μL KASP master mix, as well as 0.084 μL assay mix consisting of a combination of allele-specific primers and a common reverse primer. KASP genotyping was conducted using a BIO-RAD CFX384 instrument, and the PCR conditions were implemented according to the recommended protocol. For each locus, RILs were assigned to either the allele of CDC Amarillo, the allele of PR11-2, or heterozygote based on their grouping into the respective clusters.

Statistical analysis
Analysis of variance (ANOVA) was conducted for individual traits of interest via PROC MIXED program (SAS version 9.4), where genotypes, location, and genotype by location were considered as fixed effects and block as a random effect.
After genotyping on 88 SNP markers, 64 polymorphic markers were selected for the construction of a linkage map, based on the quality of PCR amplification and separation between parental lines and among RILs, and were used to construct the genetic map. For the linkage-map construction, polymorphic SNP markers were clustered into different linkage groups based on a minimum logarithm of odds (LOD) ratio value of five using QTL IciMapping (Meng et al. 2015). Then, the map order of each linkage group was finalized with the use of regression mapping. The recombination frequencies were converted into centimorgan (cM) through the Kosambi mapping function. Additionally, individual SNP sequences were blasted onto the pea reference genome (Kreplak et al. 2019) to retrieve corresponding physical locations on the pea genome.
For individual traits, QTL analysis was conducted on the phenotypic data of separate field trials via inclusive compositive interval mapping program in QTL IciMapping. The QTLs were filtered to select those where the LOD score was above the threshold value 2.0.

Weather summary of field locations in 2020 and 2021 summers
Summer weather data of 2020 and 2021 Saskatoon and Lucky Lake were separately retrieved from the meteorological stations Saskatoon Rcs (Meteorological Station ID: 4057165) and Lucky Lake (Meteorological Station ID: 4024714) from Environment Canada's historical weather database. For Rosthern, its historical weather data were retrieved from Worldweatheronline at the following link: https://www.worldweatheronline.com/rosthern-weathe r-history/saskatchewan/ca.aspx. In 2020, PR-24 population experienced similar HS at the two locations, and the HS level was similar to the previous pea trials in 2013-2017, whose environmental conditions were grouped as "control HS" environment (Huang et al. 2017;Tafesse et al. 2020). Therefore, data arising from 2020 Lucky Lake and 2020 Rosthern were considered as "control conditions" in this study. In 2021, pea plots experienced a greater number of extremely hot days at both vegetative and reproductive stages than pea plots in 2020; thus, all four trials in 2021 were grouped as "HS condition" (Table 1). Along with the hot temperature and drying wind in the summer in 2021, precipitation was also limited, especially rainfall at the reproductive stage. Insufficient topsoil moisture and high temperature resulted in lower pea yield in 2021 than 2020.
ANOVA for agronomic, yield, and seed quality traits To assess the environmental and genotypic effects on phenotypic traits related to flowering, yield components, agronomy, and seed quality, a combined ANOVA with six environments, 39 RILs, and their interaction as fixed effects was conducted on individual traits. All traits varied depending on field environment ( Table 2). All traits except for lodging were also significantly affected by genotypic variability within PR-24 population. In addition, a significant interactive effect between environment and genotype was observed for DTF, DTM, RNN, PN, and seed starch percentage.
Plant height, RNN, PN, plot yield, seed protein, and starch concentration demonstrated a significant difference between control environment in 2020 and HS condition in 2021 (Table 2). PR-24 had a much greater average height, RNN, PN, plot yield, and seed starch concentration in 2020 field trials compared to 2021 trials. Inversely, HS environment in 2021 induced a significantly higher seed protein concentration compared to 2020 trials.

Correlation analysis between plot yield and phenological and yield component traits
With a primary focus on yield performance, it was interesting to find the relationship of plot yield with other measured traits among the 39 RILs in the population. Because most of the phenotypic traits were significantly affected by environment and genotype by environment (Table 2), correlations between yield and other traits were analyzed for individual trials. Plant height had a significantly positive correlation with plot yield in all six trials (Table 3). Lodging had a negative correlation with plot yield in the two control trials of 2020 and the two HS trials of 2021, suggesting that varieties with less lodging would have promising yield potential regardless of climatic summer conditions in Saskatchewan. Positive correlation between DTF and yield was significant under control environment (2020_LL and 2020_ROS) and HS environment with less drought (2021_SAS_E and 2021_SAS_L), but the correlation became significantly negative at 2021_ROS_E and 2021_ROS_L, where heat and drought were both extremely stressful. Under control environments, favorable for pea growth, RILs with relatively late maturity are likely to yield more, and this trend was also reflected by the significantly positive correlation between DTM and yield at 2020_LL and 2020_ROS. Under the four stressful trials, DTM was positively correlated with yield at 2021 SAS_E, and the correlations were not statistically significant at the other trials. PN Table 1. Seeding dates, average PR-24 population flowering dates, number of days with daily maximum temperature above 28 • C, and cumulative precipitation at vegetative and reproductive stages.   was positively correlated with yield in three of the four HS trials, indicating it could be an effective indicator for yield prediction in HS conditions.

Characterization of genomic loci associated with measured traits
To dissect the genomic regions associated with the measured phenotypic traits, QTL analysis was conducted based on the association of phenotypic data with genotypic data. First, each RIL was genotyped by 88 KASP markers that tested poly-morphic between the two parental varieties, PR11-2 and CDC Amarillo, where 64 KASP markers were eventually polymorphic among RILs. Sixty out of these 64 KASP markers were successfully clustered into seven pea linkage groups (LGs) and used to construct the PR-24 genetic map (Table 4). Although marker density is low, 55 common markers to the pea consensus linkage map of Sindhu et al. (2014) were successfully aligned, confirming the quality and accuracy of the PR-24 linkage map built. In the PR-24 linkage map, markers in both LG VI and VII were aligned to the markers of LG VI in the pea Table 4. PR-24 genetic map and its alignment with pea reference genome (Kreplak et al. 2019) and the earlier pea consensus map (Sindhu et al. 2014  consensus map; however, due to the lack of anchor markers, these two LGs were assigned as separate groups. For the same reason, LG VIII and IX in this map both aligned with LG VII of the pea consensus map but were assigned into separate groups. Based on the linkage map and phenotypic data, QTL analysis of individual traits was conducted on each location to dissect the underlying genetic mechanism in trait development. Particular attention was paid to flowering, yield components, and plot yield, as these traits were more heat susceptible and closely linked with heat tolerance. Various QTLs associated with individual traits were found among different field experimental environments, which could be partially reflected by the significant environmental and G × E effects in many traits (Table 2).
For DTF, different QTLs were identified between control and HS environment groups, suggesting that different genetic mechanisms were involved in the control of flowering time under contrasting temperature conditions. No common QTL was found between the two control environments, 2020_LL and 2020_ROS, but a stable QTL for DTF at HS environment was found on pea LG VII, corresponding with pea chromosome 7 (Table 5). This QTL was detected in three out of four HS trials, and it was also found at one control trial, 2020_LL. Four markers were within this QTL in the genetic map, which were between 6744481 and 169190602 base pair (bp) on pea chromosome 7. Among the six QTLs for DTM, two QTLs overlapped at LG III and shared the same right flanking marker "PsC5404p543", which was positioned at 490393084 on pea chromosome 5. One of the two QTLs was detected in the 2021 Saskatoon early seeding trial, and the other QTL was found in the 2021 Saskatoon late seeding trial, with each accounting for ∼20% of the phenotypic variation. For lodging, height, and plot yield, no consistent QTL was detected over multiple trials for these traits.
Likewise, different QTLs for PN were characterized between control and HS environment groups. A QTL on LG I, QTL-PN-1, was considered most potentially linked with the genetic control of pod number under HS, as it was identified in two HS trials (Table 5). The left flanking marker was at nucleotide position 314673840 on chromosome 2, and the other flanking marker was at nucleotide position 386850165 on the same chromosome. Similarly, a common QTL associated with RNN, QTL-RNN-1, was identified on LG III in multiple trials, which corresponded with the physical interval on chromosome 5 from nucleotide position 198235688 to 415679416. Another QTL at LG VI (chromosome 1) was interesting and seemed to link with PN and RNN at 2021 Rosthern early seeding trial. It accounted for 16% of the phenotypic variation of RNN and 22% of the variation of PN in 2021_ROS_E. For seed protein concentration, one QTL at LG III, QTL-protein-1, was identified in two trials, and it explained 27% of the variation of the phenotypic variation among the RILs at 2020_LL and 18% of the variation at 2021_SAS_L.

Selection of stable indices as valid criteria for pea heat tolerance
Ideally, a desirable heat-tolerant variety is expected with promising yield potential in both normal and HS environments. In practical studies of crop abiotic stresses, not limited to HS, it was common to see that varieties could have contrasting yield performance between control and stressful environments, and stress-tolerant varieties often had relatively high yield under a stress environment but relatively low yield under a normal environment. Likewise, the yield data of 2020_ROS (control environment) and 2021_ROS_L (HS environment) among 39 RILs in PR-24 were poorly correlated (R = 0.09; Table 7). As a result, numerous indices were proposed as potential criteria for heat tolerance instead of yield alone. In this study, five indices were evaluated based on their reliability reported in previous literature. According to HS and precipitation conditions (Table 1), yield data at 2020_ROS (Y2020_ROS) were considered as yield potential under control environment, and yield data of 2021_ROS_E (Y2021_ROS_E) and 2021_ROS_L (Y2021_ROS_L) were separately used as yield potential under HS. GMP, SSI, STI, YSI, and ATI were calculated for each RIL between 2020_ROS and 2021_ROS_E and between 2020_ROS and 2021_ROS_L. Among the five indices, only GMP and STI had significant positive correlations with yield under both control and HS environments (Tables 6 and  7). These two indices could be suitable standards for the Table 5. Summary of identified QTLs for the measured phenological, agronomic, and seed quality traits in pea PR-24 population at Saskatoon, Rosthern, and Lucky Lake in 2020 and 2021.   selection of pea varieties with consistently high yield potential under both normal and HS environments.

Trait selection for yield prediction
Field pea is historically known as a cool-season legume crop, but its production has been expanding to drier and hotter prairie regions in North America in the last three decades. Thus, it is important to understand how the pea crop can phenologically adapt to a warming environment. Pea varieties with resistance to HS damage at flowering stage are most desirable, as extremely hot days at anthesis seems to be the major temperature cause for yield reduction (Bueckert et al. 2015;Huang et al. 2017). In our study, 2021 had a much hotter and drier summer than 2020, which caused a sharp yield reduction in 2021 compared with the average yield of RILs in 2020 (Table 1). Taking the Rosthern data for example, the population-yield average of early and late seeding trials at Rosthern was 365 and 233 kg ha -1 in 2021, which was only 26% and 16% of the 2020 Rosthern yield average, 1420 kg ha -1 . The reduced yield was partially due to an impaired pod development (Table 2) and HS-induced pod abortion, as was also seen in western Canadian field trials in other pea populations (Tafesse et al. 2020). Notedly, pea seed protein concentration tended to increase as a response to greater HS (Table 2). This trend was in agreement with Nosworthy et al. (2021), where peas had the highest seed protein in 2018, the hottest and driest year in their study, and the lowest protein percentage was found in 2016, the wettest year. However, HS-induced yield loss outweighed an accumulated protein concentration in this study. The two trials at control environments produced a mean protein yield of 281 kg ha -1 , compared with a mean protein yield of 156 kg ha -1 across the four trials in HS environments.
Selection of agronomic traits that can aid in yield gain is always the core interest in breeding work. In this study, we examined the relationship among six agronomic traits (i.e., DTF, DTM, plant height, lodging, RNN, and PN) and the plot yield at different HS and soil moisture conditions. Plant height seemed to a useful indicator for yield prediction, as it was positively associated with plot yield in all six trials regardless of temperature and soil moisture conditions (Table 3). It is worth noting that all RILs were derived from two semi-leafless pea varieties with good lodging re-sistance. Semi-leafless leaf type is the commercially desirable leaf type, and this trait contributes to reduced lodging and cooler canopy temperature under heat and drought conditions (Tafesse et al. 2019). At controlled heat conditions (2020_LL and 2020_ROS) and heat condition with sufficient soil moisture (2021_SAS_E and 2021_SAS_L), DTF was positively associated with plot yield in PR-24 population, whereas the correlation became negative in the trials where heat and drought were confounded (2021_ROS_E and 2021_ROS_L). A similar trend was previously reported in another RIL population, whose both parents were also semi-leafless pea cultivars (Huang et al. 2017). It seems to be a dilemma for pea breeding; breeders may need to develop relatively late-maturing varieties for environments where growing conditions are typically favorable but develop early-maturing varieties for growing regions where terminal heat and drought are typically severe.
For the two yield component traits, pod number on the main stem was positively correlated with plot yield in three out of the four HS trials (Table 3). A similar correlation was also seen in the other RIL population (Huang et al. 2017) and a set of 24 varieties tested under heat stressful environments (Tafesse et al. 2019). Higher pod number per plant was also a desirable parameter for heat-tolerant pea in subtropical climates (Parihar et al. 2022). Flower abortion is considered as the main reason for yield loss in HS; as a result, peas with better pod set success would potentially yield more under HS. In control environments, the correlation between pod number on the main stem and plot yield was less significant, presumably because under favorable growth conditions, basal branches are also an important source for seed production (Singh et al. 2011;Huang et al. 2017). Field pea basal branching was complexly affected by genotype, plant density, site year, and their interactions (Spies et al. 2010).

QTL comparison
Linkage analysis, also known as QTL mapping, is a common method for genetic dissection of important agronomic and seed quality traits in field pea (for example, Huang et al. 2017;Gali et al. 2018). In this manner, promising QTLs for DTF, DTM, PN, and RNN were characterized over multiple field trials from this work.
The QTL for DTF was located at LG VII, i.e., chromosome 7, flanking from Chr7_67444481 to Chr7_169190602 (Table 5). This QTL explained 20% of the overall DTF variation in PR-24. One QTL associated with flowering time at LG VII was previously reported by Klein et al. (2014); however, the precise positions of these two QTLs were not comparable due to the lack of common markers between the two studies. Consistent QTLs for DTF were also characterized at LG I, III, V, and VI in other pea RIL populations from our group, which were derived from various genetic backgrounds (Huang et al. 2017;Gali et al. 2018). What's more, in a recent review paper, Weller and Ortega (2015) summarized 20-plus genome loci that were putatively associated with pea's flowering time and inflorescence.
In this study, a consistent QTL for PN, QTL-PN-1, was identified at LG I (chromosome 2), which accounted for 28% of the phenotypic variation at 2021_ROS_L and 15% variation at 2021_SAS_E (Table 5). Our finding corroborated previous studies, where stable QTLs for PN were also characterized at this LG (Irzykowska and Wolko 2004;Krajewski et al. 2012;Guindon et al. 2019). In an association mapping study relating to 135 diverse pea accessions (Tafesse et al. 2020), 6 SNP markers on LG I had strong association with PN over multiple site years, and 3 of those 6 markers were within the QTL region described here.
Separate consistent QTLs for DTM and RNN were found at LG III (chromosome 5). QTL-DTM-1 was mapped to a large genomic region (around 80 cM), partly due to the low marker abundance of the current genetic map. The QTL for RNN was co-localized with QTL-PN-2 identified at 2020_LL. A stable QTL for RNN was also characterized at LG III in a previous linkage mapping study (Huang et al. 2017). In another association mapping study, which was made up of a set of 92 diverse pea cultivars, several markers at LG III also seemed strongly associated with the number of reproductive nodes (Jiang et al. 2017). Another consistent QTL for seed protein, namely QTL-protein-1, was also identified at LG III, which was also reported in other pea populations (Tar'an et al. 2004;Klein et al. 2014;Gali et al. 2018). Notedly, eight genes (5g161560, 5g165160, 5g171400, 5g198960, 7g051680, 7g091560, 7g091680, and 7g093240) within QTL-DTF-1 and QTL-RNN-1 were differentially expressed between two parental varieties of PR-24 in the transcriptome study ).

Assessment of heat-tolerance indices
Tolerance to stress is measured by the performance differential of genotypes between stress and non-stress conditions. Based on yield, several indices have been proved to valid criteria for drought stress-tolerance assessment, for example, SSI in spring wheat (Fisher and Maurer 1978) and in pea (Grzesiak et al. 1996), YSI in soybean (Bouslama and Schapaugh 1984), STI in common bean (Schneider et al. 1997), ATI and SSPI in wheat (Mousavi et al. 2008), and STI and GMP in chickpea (Farshadfar and Geravandi 2013). Drought often occurs with HS in Canadian pea-production regions, as well as in most cool-season legume production-areas in other countries. The above indices might be useful for HS in field pea. As a result, they were validated based on two years' yield data at Rosthern in this study, where 2020_ROS was considered as the control environment, and 2021_ROS_E and 2021_ROS_L were regarded as two HS environments. Only GMP and STI were positively correlated with yield at control and HS conditions in both sets of data (Table 6 and 7). GMP and STI were previously reported effective indices to select drought-tolerant chickpea varieties by Farshadfar and Geravandi (2013). Individual top 10 RILs with highest GMP and STI were filtered out from 2020_ROS/2021_ROS_E and 2020_ROS/2021_ROS_L, and four RILs were common, which were PR-24-03, PR-24-08, PR-24-10, and PR-24-12. These four RILs were considered as best heat-tolerant RILs in PR-24 population.

Conclusion
Plant height was positively correlated with plot yield in all six trials, implying an effective yield indicator regardless of environments. Our results also suggest that later maturity and high main-stem pod number can be the selection criteria for yield improvement under normal and HS summer climates. Furthermore, our work assessed five abiotic stress tolerance indices in pea HS for the first time and provided evidence that GMP and STI are suitable criteria for breeding heat-tolerant pea varieties with high yield potential under both normal and HS environments. The four consistent QTLs characterized in PR-24 over multiple trials are worth fine mapping and further validating in other pea genetic populations prior to the application of corresponding markerassisted selection.