Two Australian locations with varying climates were chosen to collect physiological data. The first was in Kununurra, Western Australia (15.7783°S, 128.7439°E), the second in Narrabri, New South Wales (30.2737°S, 149.7350°E). The recorded minima and maxima temperature for these locations are shown in Appendix 1, and these data were collected from stations managed by ozforcast.com.au and weather.agric.wa.gov.au, respectively. In summary, for the purpose of this trial, Kununurra is considered an environment in which the trials experienced only high temperature stress, as the trials were flood irrigated and regularly treated with nitrogen. Due to this, the Kununurra site could be considered an “ideal” environment when studying heat stress, as other confounding stressors are absent. Narrabri can be considered a more typical environment for chickpea production, experiencing the milder mid-season temperatures, with the crop receiving rain and supplementary irrigation as required.

Field experiments were conducted at these locations in 2019, at two sowing times. Experiment refers to the year and environment in which the trial was executed. The first time of sowing (TOS) was based on typical commercial growing times within each region, and the second was delayed so that flowering would occur when average temperatures were higher, thus generating heat stress from reproductive development onwards. In Kununurra, this delay was 1 month, as the higher temperatures caused a shorter season length. In Narrabri it was 2 months, due to the milder conditions. This resulted in a total of 4 experiments across 2 environments and 2 TOS.No soil analyses were conducted as the focus was on the temperature stress (manipulated through location and time of sowing). Experimental plots were located in the same field with the same cropping history/agronomic practices minimizing soil variability.

The 148 elite chickpea genotypes evaluated included 10 Desis and 138 Kabulis, originating from Syria (129), India (10), and Australia (9) (Appendix 2). Those genotypes not from Australia were provided by ICARDA and ICRISAT and were considered high yielding in hot environments. Australian genotypes were chosen as comparison, being already used in Australian environments. Genotypes were sown according to commercial standards (~169 seeds and ~381 seeds per plot in Narrabri and Kununurra respectively) in randomized complete block designs of 2 replicates for each TOS in plots of 8 m². Randomized complete block designs were used to control for variability across the field. Differences between replicates were considered through randomized complete block design and treating replicates as random effects. Main effects of genotype and environment (location and time of sowing) were significant, while differences between replicates were inconsequential ( Tables 1 – 3 ). Each trial was prepared and treated using standard agronomic practice including fertilizer, herbicide and pesticide recommendations at each location (Appendix 3). Plots were inspected daily to identify date of flowering and maturity. This was recorded as the number of days from sowing to each stage of development. According to standard commercial practices, the growth season for Chickpea in Kununurra is typically shorter than in Narrabri (GRDC, 2016; Regan, 2015). All plots were harvested when the entire experiment attained harvest ripeness. Gross yield and 100 seed weight (HSW) was assessed in grams post-harvest. Weather data was collected using an on-site weather station which sampled every 15 minutes. This was collated into daily averages across the season. Across all trials, data can be categorized into plant physiological and climate data. Plant physiological data comprised number of days to flowering, and maturity, the weight of 100 seeds in grams, and gross plot yield in grams. Climate data refers to the temperature regimes at each location over the duration of the trial.

There was greater range and variability in the SDD and GDD data when compared to Grow Days ( Table 2 ). This was particularly notable when considering the difference in grow days between locations for maturity, compared to their respective GDD and SDD values. The GDDs for Kununurra at maturity were triple those of Narrabri, and the SDDs tenfold, whereas the grow days (# Days) were 30% higher at most. Additionally, there was a greater proportional difference between flowering GDD at Narrabri (TOS2 was 35.3% larger than TOS1), than between TOS in Kununurra (TOS2 was 18.1% larger than TOS1). There was greater variation for age at flowering between TOS in Narrabri (23.5%) compared to Kununurra (3.5%), and both TOS in Kununurra reached flowering in a much shorter time compared to Narrabri (32.9% and 15.4% faster, respectively). However, the time between flowering and maturity was shorter for both TOS in Narrabri (49 and 37 days, respectively) compared to Kununurra (71 and 76 days, respectively).

A large difference in yield was observed between locations, with mean yield at Kununurra (x = 3.4 T/ha) 50% higher and HSW (x = 38.0 g) 12.1% higher than Narrabri. However, there was less variation between TOS in Kununurra for both yield (6%) and HSW (5.4%), than Narrabri (48.5% and 10.2% respectively). There was also less variation for HSW than Yield, (19.2% and 67.3% respectively) across all experiments.

GDD and SDD are predominately significantly correlated with HSW throughout the lifespan of the crop ( Table 3 ). This correlation was mostly negative, particularly at Narrabri (R² = -0.333, -0.29, -0.182, P < 0.001, P < 0.001, P = 0.002), though the SDD value in Narrabri was consistently much lower than Kununurra ( Table 2 ). However, there was a significant positive correlation between HSW and SDD observed during the reproductive stage in Kununurra (R² = 0.164, P = 0.005). There was no significant correlation between GDD and SDD for HSW in the vegetative period at Kununurra (P = 0.479, P = 0.158), nor was there for GDD during the reproductive period (P = 0.802).

From Table 3 , it can be noted that both GDD and SDD are highly significantly correlated with Yield (P < 0.05): particularly in Narrabri (P < 0.001). This correlation is exclusively negative in the vegetative growth period (R² from -0.807 to -0.180) but positive in Kununurra during the reproductive period, with a stronger correlation between yield and SDD (R² = 0.187, P < 0.001) than GDD (R² = 0.183, P = 0.002).

Figure 1 continues to show the negative correlation between GDD, SDD and yield between sowing and flowering. This trend is strongest in Narrabri for both SDD and GDD compared to Kununurra, with negative exponential trends and higher R² values. For Narrabri, the trend between GDD and yield is stronger than SDD and yield (R² = 0.66 and 0.58 respectively), whereas for Kununurra, the reverse is true, with SDD having a stronger negative correlation compared to GDD (R² = 0.04 and 0.03 respectively) ( Figure 1 ).

The GGE effects of each environment on yield and HSW are given in Figures 2A–F . HSW was more influenced by genotype than environment, evidenced by high PC1 scores (89.63 – 94.29%) and acute angles between each environmental vector. Genotypes that clustered towards the origin demonstrated a similar performance to the virtual genotype.

Yield was impacted more significantly by the environment in comparison to HSW. There are wider angles between environmental vectors, and lower PC1 scores (61.87 – 83.29%). Genotypes are spread further from each other, with some genotypes associated with specific environmental vectors.

Variance in yield at Narrabri was more strongly related to genotype (83.29%) than Kununurra (73.25%). In Kununurra, a larger number of genotypes aligned more closely with specific environmental vectors than Narrabri, demonstrating superior performance in this environment.

For HSW, 60% of top performers were common to both environments, and 10% were top performers or highly stable across traits in each environment ( Table 1 ). For yield, 30% of top performers were common to both environments and 10% were highly stable ( Table 1 ). In terms of consistent stability across traits, genotypes 39 and 68 had high stability for both HSW and yield in Narrabri and Kununurra ( Table 1 ).

No genotypes were both high performers and highly stable in each environment.

Desi genotypes represent 6.76% of the genotypes assessed in these experiments but represented 30% of the genotypes ranked in the top 10 for yield ( Table 1 ). Desi genotypes represented 20% of the genotypes ranked in the top 10 for HSW stability, but these genotypes were only considered highly ranked based on performance in Kununurra ( Table 1 ).

HSW had a higher heritability than yield overall (H² = 0.916 – 0.970 and 0.378 – 0.956, respectively). HSW had a more consistent heritability estimated across environments when compared to yield (3.6% and 60.4% variation, respectively). HSW had a more consistent heritability estimate in Narrabri compared to Kununurra (0.2% and 4.6% variation respectively), whereas yield showed consistently lower heritability in Narrabri compared to Kununurra (17.9% and 4.3% variation respectively). The experiment with the highest heritability across both traits was Narrabri TOS 2, and the lowest was Kununurra TOS 2.

This study showed that GDD and SDD provide greater variation in field data compared to a simple count of the number of days that were at suitable temperature for growth. Data for both geographical locations were vastly different for each calculation, however this was to be expected, considering the vast differences overall climate (Appendix 2), and the differing agronomy practices at each location, such as more extensive watering at Kununurra. However, this variation allowed detailed investigation of genotypic and environmental influences on key traits ( Figure 2 , Table 2 ), due to the bespoke environment of each trial. This diversity of conditions highlights the need to develop location-specific cropping strategies and genotypes for farmers. This study helps provide the ability to develop these strategies with the capacity to predict the impact of climate on yield potential, given a single year of data ( Figure 1 ; Table 3 ).

Both TOS in Kununurra reached flowering sooner than Narrabri, likely due to a larger accumulation of GDD in Kununurra over both TOS ( Table 2 ). Additionally, yield in Kununurra was 50% higher on average than Narrabri ( Table 2 ). This high GDD indicates a consistently warm environment, with reduced vegetative development and earlier flowering. Furthermore, several studies have shown a correlation between early flowering and high yield (Kumar et al., 2013; Rubio et al., 2004; Siddique et al., 2003). Apart from higher GDD, Kununurra also had higher SDD across the season compared to Narrabri. As the SDD calculation refers to the stress experienced at this location, one may expect a decrease in yield in Kununurra compared to Narrabri. However, commercial irrigation standards in Kununurra are considerably more generous, thus plants had greater capacity for stress-management via transpiration cooling, compared to Narrabri, where trials were irrigated according to typical rainfed conditions for the region. Previous research on the impacts of temperature and water stress on chickpea, found that sites with high rainfall typically demonstrated high yield performance (Dreccer et al., 2018), and this was likely due to the plant’s ability to reduce canopy temperatures by 6-10°C (Mathur et al., 2014). Interestingly, HSW in TOS2 was also 5.4% larger than TOS1 in Kununurra ( Table 2 ), likely due to a longer period between flowering and maturity. This longer grain filling period results in larger seeds (Davies et al., 1999; Zaiter and Barakat, 1995), as well as the consistent development of new seeds/pods due to the indeterminate nature of the species (Singh, 1997). There is a delicate balance here, as it has been shown that in Australian conditions, early maturing chickpeas with limited water have higher yields only when temperatures are optimal (Sabaghpour et al., 2003). In this study, SDD and GDD were both negatively correlated with yield throughout the season in Narrabri, particularly in TOS2 ( Figure 1 ; Tables 2 , 3 ). Thus, it can be seen that even though SDD was lower than Kununurra, stress management is often reliant on resource availability (Mathur et al., 2014).

Australian genotypes represented 5 of the 10 highest yielding genotypes with highest HSW across environments HSW ( Table 1 ). However, in terms of static stability, only one genotype was represented, ranking 10^th in yield, and this was predominately observed only at Narrabri ( Table 1 ). This suggests that the Australian genotypes are well adapted to this environment, although the average yield loss between Narrabri times of sowing was 32.34% ( Table 2 ). It can therefore be concluded that the heat tolerance of Australian cultivars could be improved by utilizing some of the international genotypes that ranked highly in static stability as parents ( Table 1 ; Appendix 1 ).

There is clearly a trade-off between seed size and seed number with consequences for yield. However, this allows for tailoring of the product to several markets. Farmers will be driven by price. If gross margins are greater for high yield, then smaller seeded types will be produced. However, a premium price for large seed may change this equation. Unfortunately, no genotypes that ranked highly for both stability and yield. However, some genotypes had both HSW and yield in specific environments, suggesting these may be good choices as breeding stock. Due to the vast differences in environments both between and within locations, it would be unlikely that high performers for either trait would also be stable. However, stable genotypes ( Table 1 ) may be a good choice for farmers in highly variable environments with unpredictable seasons.