Climate conditions in the near-term, mid-term and distant future for growing soybeans in Canada

Abstract The soybean industry in Canada aimed to extensively expand soybean production to benefit from new early-maturing varieties and the warming climate. However, setbacks in the soybean industry since 2017 demonstrated the impacts of climate risk and global market uncertainty. Therefore, a better understanding of future climate conditions that will impact soybean growth in Canada is needed for decision-making in the sector, such as prioritizing regions for expansion and developing climate change adaptation strategies through either agronomic management practices or breeding new cultivars. Based on climate projections from a set of global climate models, we analyzed climate conditions for growing soybeans, including growing season start, crop heat units, precipitation, precipitation deficits and climate extremes, in the near-term (2030s), the mid-term (2050s) and the distant future (2070s). We found that a future warmer climate with an increase of 1.6, 2.8 and 4.1 °C in the growing season (May–September) mean temperature averaged over Canada’s land area in the near-term, mid-term and distant future under SSP3-7.0 would favour the expansion of soybean production further north and west. However, an increase of approximately 200 mm in precipitation deficits on the semiarid Canadian Prairies in the mid-term would constrain soybean production unless irrigation could be introduced. Heat- and drought-tolerant cultivars should be developed to adapt soybean production to a changing climate, in addition to the adoption of late-maturing cultivars that would benefit from the lengthened growing season and increased crop heat units.


Introduction
New early-maturing varieties have allowed for an expansion in Canadian soybean [Glycine max (L.) Merr.] production beyond the traditional growing area of southern Ontario. Soybean crop is now a regular part of the rotations on many farms in Manitoba, Quebec and the Maritimes, and more recently, in Saskatchewan and some areas of Alberta (Soy Canada 2022a). Canada's overall soybean production increased steadily from 2.69 million metric tonnes (Mmt) in 2007 to 7.72 Mmt in 2017 due to increased production in Ontario and Manitoba (Soy Canada 2022b). Production in the west decreased in recent years, however, mainly due to poor weather conditions and uncertainty in the market (Hallick and MarketsFarm 2019). While soybean production and the seeded area in eastern Canada were steady, both yield and the seeded area on the Canadian Prairies decreased significantly after 2017 (Statistics Canada 2022). The seeded area decreased from 0.94 million hectare (Mha) in 2017 to 0.53 Mha in 2021 in Manitoba, and dropped from 0.34 Mha in 2017 to 0.03 Mha in 2021 in Saskatchewan. Apparently, in addition to the market uncertainty, climate conditions are still the limiting factor for soybean production in Canada, especially on the Canadian Prairies.
Due to enhanced greenhouse effects, climate change in Canada has lengthened the growing season with a trend of 8.2 days century −1 and increased the heat amount for plant growth with a trend of 228.1 crop heat units (CHU) century −1 (Qian et al. 2010). This trend is projected to continue (Qian et al. 2013). As a warm season crop, soybeans may benefit from warming for the potential expansion from the traditional growing areas in southern Ontario to the Canadian Prairies as this has already happened since the late 2000s. Soy Canada developed an ambitious plan to expand the seeded area of soybeans to 10 million acres (approximately 4 Mha) by 2027, and the expansion was expected mostly on the Canadian Prairies (Soy Canada 2017). Significant decreases in the seeded areas of soybeans on the Canadian Prairies in recent years demonstrated the climate-related uncertainty for growing soybeans and the need for climate projections for the near-term, mid-term and the distant future.
Based on a range of available heat units projected using temperatures from multiple global climate models (GCMs), Bootsma et al. (2005a) estimated an average yield increase for soybeans by 0.6-1.5 t ha −1 (21%-50%) by 2040-2069, compared to 1961-1990, not including the direct effect of increased atmospheric CO 2 concentrations, advances in plant breeding and crop production practices or changes in impacts of weeds, insects and diseases on yield. These estimates might not be applicable to the semiarid prairies under rainfed conditions as they were based on soybean field trials in eastern Canada. Thivierge et al. (2017) projected a yield increase for soybeans in Southwest Quebec using the Integrated Farm System Model (Rotz et al. 2015). Payant et al. (2021) showed the potential of growing soybeans in some regions of Quebec in the future climate where it is currently not suitable. Jing et al. (2017) simulated soybean yields under climate change scenarios using two crop growth models and found that soybean seed yield would not benefit from a prolonged growing season under the projected future climate in eastern Canada unless the harvest index could be maintained. It was estimated that half of the historically increased soybean yield could be attributed to genetic improvement and the other half to enhanced management or other factors including climate (Boote 2011). Therefore, future climate conditions for soybean growth and yield are essential information not only for producers to manage production risks but also for breeders to develop new cultivars that benefit from extended growing seasons and increased heat units while being tolerant to heat stress and drought.
Based on climate scenarios from the up-to-date climate change simulations (phase 6) of the Coupled Model Intercomparison Project (CMIP6; Eyring et al. 2016), we aimed to provide climate projections (i) to address potential changes in growing season (start, end and length), crop heat units (CHU), precipitation, heat stress and related climate extremes across southern Canada and (ii) to discuss the implications for growing soybeans, especially on the Canadian Prairies, in a changing climate.

Climate scenarios
Climate scenarios data used in this study were obtained from phase 3b of the Inter-Sectoral Impact Model Intercomparison Project (ISIMIP3b; https://www.isimip.org/proto col/3/). The ISIMIP3b climate scenarios were derived from CMIP6 output on 0.5 • grids using the bias adjustment and statistical downscaling method ISIMIP3BASD (Lange 2019;, based on the observational data set W5E5 that is version 2.0 of WFDE5 over land merged with ERA5 over the ocean (Cucchi et al. 2020;Lange 2021). Five primary GCMs (GFDL-ESM4, IPSL-CM6A-LR, MPI-ESM1-2-HR, MRI-ESM2-0 and UKESM1-0-LL), which are mandatory to use by the ISIMIP impact modelling teams, were included in ISIMIP3b and used in this study. The selection of primary models was done taking into account process representation, structural independence, climate sensitivity, performance in the historical period as well as the special input data needs of the fisheries and marine ecosystems sector. ISIMIP3b currently includes climate scenarios under three Shared Socioeconomic Pathways (SSPs; Riahi et al. 2017) --SSP1-2.6, SSP3-7.0 and SSP5-8.5 being used in global climate change impact studies (e.g., Jägermeyr et al. 2021). Climate scenarios data, including daily maximum (T max ) and minimum temperature (T min ), and daily precipitation (P), were extracted from ISIMIP3b for grids over land in Canada.

Climate indicators for soybeans
Climate indicators are used in this study to present climate conditions relevant to growing soybeans. Growing season length and its start are essential for growing crops as growing season start (GSS) constrains when to plant and the timing of crop phenology stages. CHU have been widely used to rate the suitability of various regions for the production of corn and soybeans (Bootsma et al. 2005b), especially because CHU is critical for selecting soybean cultivars (Ontario Soybean and Canola Committee 2022) from proper maturity groups (Mourtzinis and Conley 2017). Precipitation and water stress play a critical role for growing soybeans, especially in the semiarid prairies. Climate extremes, such as high temperature and dry spells, are important during the growing season. Climate indicators are defined as follows.
Frost-free period (FFP, days) --number of days between last spring frost (daily minimum temperature ≤0 • C) and first fall frost.
Growing season start (GSS, day of year--DOY) --GSS is also considered as the potential planting date for soybeans and the time to start accumulating CHU. GSS in each year was determined based on the following criteria (Bootsma and Brown 1995): (i) the average daily mean air temperature (based on 30 year average) exceeds 10 • C; and (ii) three consecutive days with mean daily air temperature >12.8 • C occurred after condition (i) was met. The third day after meeting the criteria was counted as GSS.
Growing season end (GSE, DOY) --the date on which the minimum air temperature dropped to ≤0 • C or when the average daily mean air temperature (based on 30 year average) drops to 15 • C, whichever date was earlier (Bootsma and Brown 1995). The 15 • C date corresponds approximately to the time when soybeans are mature (95% of the pods have turned brown). GSE was used to terminate accumulation of CHU.
Crop heat units (CHU) --heat units were calculated from daily maximum and minimum air temperature using the formula in eqs. 1-3 (Brown 1975;Bootsma 1994) on a daily basis and accumulated from GSS to GSE. (1) Growing season precipitation (GSP, mm) --precipitation totals accumulated during the growing season from 1 May to 30 September.
Growing season precipitation deficit (GSPD, mm) -precipitation deficit (PE-P) accumulated during the growing season from 1 May to 30 September. PE is potential evaporation estimated from T max , T min and solar radiation at the top of the atmosphere using the methods of Baier and Robertson (1965) and Baier (1971).
August monthly precipitation (AMP, mm) --monthly precipitation totals in August are critical for pod-filling, especially on the Canadian Prairies, if soybean crops have developed sufficient yield potential. A minimum of one inch (approximately 25 mm) is often required. Precipitation received in August is generally too late to make any differences in spring cereal yields; however, for soybeans and other crops that require a longer growing season, August precipitation may provide a late boost to yield (Ojo 2021). Earlier studies show significant positive correlation between soybean yield in Illinois (USA) and precipitation in August (Goldblum 2009) and from mid-August to mid-September (Runge and Odell 1960). Given the shorter growing season in Canada than Illinois, we adopted AMP in this study.
Maximum dry spell (MDS, days) --maximum number of consecutive dry days with P < 1.0 mm during the growing season.
Number of hot days (NHD, days) --number of days during the growing season with T max ≥ T c . Critical high temperature thresholds (T c ) vary over the vegetative and the reproductive phases (Hatfield et al. 2011). Optimal temperature is considered higher in the vegetative phase than in the reproductive phase (30 vs. 26 • C). Soybeans experience heat stress, and yield reductions can begin to occur regardless of reproductive stage when air temperatures exceed 85 • F (≈30 • C; Farm Progress 2022), especially when soil moisture is limiting and this can be the case on the semiarid Canadian Prairies. Salem et al. (2007) found that the elevated temperature reduced pollen production by 34%, pollen germination by 56% and pollen tube elongation by 33% when they examined the growth of soybeans under two temperature regimes (38/30 vs. 30/22 • C), Baker et al. (1989) and Boote et al. (2005) reported that the harvest index was reduced at temperature above 23 ∼ 27 • C. The timing of soybean growth stages can significantly differ across regions and varieties, as well as year by year. To make it simple, we used T c = 30 • C to calculate NHD for the entire growing season.
Number of days with a warm night (NDWN, days) -number of days with T min between 10 and 25 • C during the growing season. High nighttime temperatures (from 21 to 32 • C) can result in wasteful respiration and a lower net amount of dry matter accumulation in plants (Sharma et al. 2022). Compared to corn, soybeans are less sensitive to high nighttime temperatures. Warm night temperatures do not appear to increase respiration in soybean plants as much as corn. During the day, soybean plants accumulate starch in their leaves. At night, the starch is broken down and exported from their leaves (Onat et al. 2017). When nights are cool, the amount of starch exported is reduced resulting in high leaf starch the following day, which can disrupt photosynthesis. Nighttime temperatures have to exceed 85 • F (approximately 30 • C) before any noticeable reduction in soybean yield is experienced (Farm Progress 2022). Cober and Morrison (2019) found that soybean seed yield was increased significantly by elevated atmospheric CO 2 concentration, higher precipitation and higher minimum temperatures during flowering and podding, although yield decreased with higher minimum temperatures during vegetative growth and seed filling. Given that the respiration rates at 26 and 29 • C were doubled when compared to 20 • C (Djanaguiraman et al. 2013) and the lower cardinal temperature is 10 • C (Souza et al. 2013), we counted warm nights with T min between 10 and 25 • C.

Projecting climate change impacts
Considering the heat units required by soybean crops, our projections cover Canada only south of 60 • N on 0.5 • grids. Some detailed projections are provided for five locations which include locations of potential expansion in western Canada and areas of comparatively high production in eastern Canada (Brian Innes, Executive Director, Soy Canada, personal communication). Projections for future periods in terms of near-term (2030s, 2020-2049), mid-term (2050s, 2040-2069) and distant future (2070s, 2060-2089) are compared with the baseline climate of 1985-2014. Means or other statistics over long term (30 years) are presented for climate indicators related to growing soybeans. Climate conditions may vary year by year and large interannual variability is one of the sources of large uncertainty in climate projections (Qian et al. 2020).
Ensemble means of five ISIMIP3b GCMs are presented but projections based on individual GCMs are available from the authors. We show mainly the projections of SSP3-7.0 representing the medium to the high end of the range of future emissions (i.e., 7.0 Wm −2 at the end of the 21st century) under SSP3, a pathway called a rocky road with high challenges for mitigation and adaptation. SSP3 describes that a resurgent nationalism, concerns about competitiveness and security, and regional conflicts push countries to increasingly focus on domestic or, at most, regional issues. Nevertheless, the global warming levels would be similar under different SSPs for the near-term future and diverge from the mid-term future. SSP1 refers to a world of sustainability-focused growth and equality taking a green road with low consumption and population growth. In contrast, SSP5 depicts a world of fossilfueled development taking the highway (high challenges to mitigation and low challenges to adaptation) with its focus on high economic growth and consumption (Riahi et al. 2017). Projections under SSP1-2.6 and SSP5-8.5 are thus discussed for the mid-term and distant future only.
While projections are often presented by a relative change, such as 5% more precipitation or 10% decrease of crop yield in comparison with the baseline, it is necessary to present the projected values for some climate indicators, such as CHU, because they are critical information for selecting or breeding cultivars from appropriate maturity groups. For additive indicators, such as CHU, projections were derived using eq. 4. On the other hand, projections for multipliable indicators, such as GSP, were obtained using eq. 5.
Where I f is the ensemble mean of the future projection of indicator I and I b is the baseline value derived from the observational data set W5E5; I fmi and I bmi are the future projection and the baseline value of indicator I, respectively, for the i-th GCM; and n is the number of GCMs in the ensemble, that is, 5 in this study.

Longer growing season with an earlier start
The FFP under the baseline climate is mostly between 100 and 130 days on the Prairies but longer in southern Ontario and Quebec, and the coastal regions. It was projected to consistently increase by approximately 10 days in the 2030s, another 10 days in the 2050s and 10 more days in the 2070s (Fig. 1). In the distant future (the 2070s), FFP was projected to reach over 160 days at many locations on the Canadian Prairies and more than 180 days in southern Ontario and Quebec, and the coastal regions. The projected increase would be much smaller under SSP1-2.6 while much larger under SSP5-8.5, especially in the distant future (Fig. S1).
Due to longer FFP resulting from warming, estimated GSS for soybeans was projected to advance from mid-and late-May in the baseline to early and mid-May in the southern Prairies in the 2030s, in most Prairies in the 2050s, and into April in the 2070s in the southern Prairies (Fig. 2). Similar advancement pattern was projected for southern Ontario and Quebec where the GSS in the baseline was in early and mid-May, which is earlier than the southern Prairies. The projected advancement of GSS would be smaller under SSP1-2.6 but much larger under SSP5-8.5 (Fig. S2). Larger areas on the Canadian Prairies and eastern Canada were projected to see GSS for soybeans into April in the 2070s. Projected GSS for selected locations can be seen in Table 1, indicating a similar advancing pace of GSS on the Canadian Prairies and eastern Canada.
Similar to the advancement of the GSS, GSE was projected to be later (not shown).
More crop heat units with lower temperature stress CHU were mostly lower than 2000 except for the current agricultural regions in the baseline climate. CHU accumulation could meet the requirement for growing very short season soybean cultivars, that is, Maturity Group 000 and 00 (MG 000 and 00), in some areas in the southern Prairies and the Atlantic Canada while CHU accumulation was much higher in the traditional soybean growing regions in southwestern Ontario where cultivars up to MG III could grow (Fig. 3). Increased CHU accumulation is the most noticeable character-istic of the growing season climate conditions for soybeans in the projections under a warmer future due to the projected higher temperatures and longer growing seasons. CHU accumulation was projected to increase over time, and more than 3300 CHU, which is required for growing soybean cultivars from MG III, would be available at many locations on the southern Prairies and most of southern Ontario in the 2070s (Fig. 3). Projections for selected locations can be found in Table 1. Similar to the projections for GSS and FFP, the increases in CHU were projected to be much smaller under SSP1-2.6 and larger under SSP5-8.5 in the mid-term and especially in the distant future (Fig. S3).
The NDWN was mostly fewer than 70 days on the Canadian Prairies while this was 20-30 days more in southern Ontario in the baseline (Fig. 4). It was projected to increase significantly and steadily in the 2030s, 2050s and 2070s (Fig. 4). The projection was that it would reach 100 days at most locations in southern Prairies in the 2070s. Therefore, stress due to low temperatures would likely be reduced, which is favorable for the development and growth of soybean crops. The most remarkable difference among the three SSPs would be the largest increase in the 2070s under  showing that the warm nights could reach over 100 days in most locations on the Canadian Prairies (Fig. S4).
Increased potential evaporation with little change in precipitation GSP (1 May-30 September) was mostly over 400 mm in eastern Canada but often below 300 mm on the Canadian Prairies, especially in southern Saskatchewan and Alberta, and between 300 and 400 mm in Manitoba under the baseline climate (Fig. 5). Little change was projected for the future no matter which time horizon and SSPs were considered, although a slight increase/decrease (approximately 5%) might be seen at some locations. Slight decreases, approximately 5-10 mm, were projected for AMP on the Canadian Prairies and noticeable at some locations (not shown). Projections of GSP and AMP for selected locations are listed in Table 2. Noticeable decreases in August precipitation were projected for the Prairie locations where the August precipitation was already low in the baseline climate.
As increasing temperature is more certain and significant than precipitation change under climate change, it would enhance potential evaporation resulting in higher precipitation deficit, especially on the drought-prone Canadian Prairies. GSP deficit was much smaller in eastern Canada than the Canadian Prairies under the baseline climate (Fig. 6). GSPD reached over 400 mm in the southern Prairies. While GSPD would slightly increase in Ontario and Quebec, it was projected to largely increase on the Canadian Prairies and evolve through the near-term, mid-term and distant future. It could reach over 600 mm in the southern Prairies in the 2070s, especially under SSP5-8.5 (Fig. S5).

More stress from climate extremes
In the baseline climate, the NHD (T max ≥ 30 • C) was below 30 in most regions and close to 30 only in the southern Prairies (Fig. 7). Regions with more than 30 hot days would quickly expand over the time horizon and cover all  current agricultural regions and beyond in the 2070s (Fig. 7). It was projected to reach over 40, 50 and 70 days in the nearterm, mid-term and distant future in the southern Prairies. Projected NHD for selected locations was shown in Table 3 and they would all reach 50 days in the 2070s at the three Prairie locations. As expected, the projected numbers would be lower under SSP1-2.6 but higher under SSP5-8.5 (Fig. S6).
The longest mean MDS was observed in the southern Prairies for over 20 days but below 30 days under the baseline climate (Fig. 8). It was projected to be slightly longer by about 5 days in the southern Prairies, reaching close to 30 days in more areas in the 2030s onward (Fig. 8), although the increase in MDS would not be as remarkable and evolve through the time horizon as the case of NHD. Projections for the selected locations (Table 3) are in line with the overall projections. The projected changes under SSP1-2.6 and SSP5-8.5 (not shown) are similar to those under SSP3-7.0 as shown in Fig. 8.

Changing climate variability
Changes in the seasonal pattern of precipitation can alter the timing of precipitation and affect moisture availability for the growth and yield of soybean crops. Monthly  precipitation in the baseline climate and projections for the 2030s, 3050s and 2070s are shown in Fig. 9, for the selected locations. More precipitation would be expected in winter and spring months in both eastern and western Canada with decreased precipitation in summer months especially in July and August at the Prairie locations. Interannual climate variability is critical for crop production, especially in regions prone to droughts in the southern Prairies and areas currently without sufficient CHU. Exceedance probabilities of CHU and GSP in the baseline climate and their projections in the 2030s, 2050s and 2070s under SSP3-7.0 are shown in Fig. 10 for the selected locations on the Canadian Prairies. CHU was projected to steadily increase on the Canadian Prairies. For example, the probability of exceeding 2400 CHU in the baseline was below 10% but would reach over 80% in the 2030s and close to 100% in the 2070s and the probability of exceeding 2800 CHU would be over 90% in the 2070s at Humboldt in Saskatchewan. On the other hand, the probability for GSP exceeding 250 mm was projected to decrease at these Prairie locations, although higher probabilities of exceeding 475 mm were projected for Oakner in Manitoba.

Discussion
An extended FFP with an earlier start in the growing season with increased CHU would potentially expand areas suitable for growing soybeans from the traditional southwestern Ontario to the Canadian Prairies in the west and to further north in Ontario and Quebec. This trend is projected to continue. This trend is also in line with what is happening in Europe (Guilpart et al. 2022). Future climate conditions would also make it possible to grow late-maturing soybean cultivars from MG III or higher in current growing regions and the projected CHU in the 2070s could support soybean cultivars of MG III in the southern Prairies. Bootsma et al. (2005a) estimated an increase of 0.13 t ha −1 of soybean yields for an increase of 100 CHU based on soybean yield data from selected locations obtained for the years between 1996 and 2000 from variety trials conducted in Ontario and the Maritime provinces. Precipitation in these regions was usually sufficient for achieving increased soybean yields. However, crop growth in the Canadian Prairies is constrained by moisture availability thus yields may be less responsive to increased CHU.     Little change or even a slight decrease in the GSP was projected for the Canadian Prairies, even in the current northern agricultural regions. Moreover, critical precipitation in August for realizing soybean yield potential was projected to decrease. Given the potential of earlier planting in a warmer future, critical precipitation could be moved into July. Nevertheless, projections show that July precipitation would also decrease. In addition, higher temperature would largely increase potential evaporation resulting in larger precipita-tion deficit and crop moisture stress. Therefore, expanding soybean production further west on the semiarid Canadian Prairies would be limited by precipitation unless irrigation could be introduced, although moisture deficit might be less detrimental in sub-humid gray and black soil zones since soybean growth may respond favorably to future higher CO 2 concentrations which is known to improve crop water use efficiency and increase photosynthesis (Kothari et al. 2022). A recent study (Ort et al. 2022a) shows that the expansion of Table 3. Number of hot days (T max ≥ 30 • C, NHD, days) and maximum dry spell (MDS, days) in the growing season in the baseline climate  and projections for the 2030s, 2050s and 2070s under SSP3-7.0 at selected locations. Baseline  2030s  2050s  2070s  Baseline  2030s  2050s  2070s   St Hyacinthe (QC)  6  18  28  43  11  12  12  12   Clinton (ON)  9  23  37  50  13  13  14  14 O a k n e r( M B ) 9 2 7 3 9 5 4 1 4 1 4 1 6 1 6

NHD MDS
Melville ( soybean production onto the Canadian Prairies has resulted in new environmental constraints that affect crop phenology, seed yield and quality. Another concern comes from the fact that the soybean crop is a short-day plant. Longer daytime in the north may delay the development and growth thus result in increased risks of crop failure at the higher latitudes with a relatively short growing season before the first fall frost. Current MG III cultivars will likely perform poorly when grown in more northerly latitudes as they are more sensitive to photoperiod than lower MG cultivars (Jiang et al. 2014). Ort et al. (2022b) showed that all cultivars required more time to flowering in the longer photoperiod treatments and the later rated MG had the greatest sensitivity to photoperiod. A delay in time to flowering from a longer photoperiod can delay maturity and expose the crop to fall frost that can reduce seed yield and quality. Heat stress could become a big issue for growing soybeans in the distant future as large increases in the NHD were projected, although warmer climate might be more beneficial to soybeans than spring cereal crops in general and less cold stress would occur for all crops. Therefore, developing heat-and drought-tolerant soybean cultivars could be critical for future soybean production in Canada in the long run.
Plant diseases are a major problem for soybean production in Canada as pathogens, which cause the diseases, are adapted to a wide range of temperature and moisture conditions. For example, phytophthora stem and root rot accounts for approximately $50 M in annual losses in Canadian soybean production (Sepiol et al. 2017). More diseases that affect soybean production in Canada are discussed in Wrather et al. (2001). In addition, climate change coupled with global trade intensification could increase the risk of the introduction and establishment of invasive alien soybean pests in Canada. For instance, the soybean cyst nematode has been discovered in Quebec and was projected to produce up to four or five generations under future climate scenarios, while it can currently complete from one to three generations (St-Marseille et al. 2019). St-Marseille et al. (2019) suspected that the expansion of soybean production to northern areas would be more favorable to the soybean cyst nematode development. Climate conditions projected in this study are in agreement with their projections Fig. 9. Monthly precipitation in the baseline climate  and the near-term (2030s), mid-term (2050s) and distant future (2070s) under Shared Socioeconomic Pathway (SSP) 3-7.0 at selected locations across Canada.
in Quebec but include croplands across Canada. Therefore, preventing and controlling soybean diseases and pests could also be critical to future production in Canada in a changing climate.
This study provides projections of future climate conditions relevant to growing soybeans in Canada. However, the responses of soybean growth and yields are also related to the responses of soybean crops to elevated atmospheric CO 2 concentration (Baker et al. 1989;Boote et al. 2005) accompanying with the warming trend. As a C 3 crop, soybean can benefit from elevated atmospheric CO 2 concentrations with enhanced photosynthesis and increased water use efficiency (Leakey et al. 2009). Improved water use efficiency may be especially important, given the current water stress on the Canadian Prairies and the projected increasing precipitation deficit in a changing climate. In addition, the growing season was projected to extend into April and October on the Canadian Prairies in the warmer future, which could alter the tem-poral distribution of precipitation and precipitation deficits during the growing season. To quantitatively assess climate change impacts on soybean growth and yield in Canada, crop growth models should be used to examine the effects of elevated atmospheric CO 2 concentrations, heat and water stress at critical growth stages on crop growth and development. In agreement with the findings of Baker et al. (1989) and Boote et al. (2005) that the harvest index was reduced at a temperature above 23 ∼ 27 • C, a modelling study by Jing et al. (2017) showed that soybean seed yield in eastern Canada would benefit from an extended growing season under the projected future climate only if the harvest index could be maintained. Jiang et al. (2020), using the Root Zone Water Quality model, predicted that soybean yields may increase by 31% by 2038-2070 in southern Quebec due to increased photosynthesis rates and improved crop water use efficiency under elevated CO 2 . Although soybean growth was negatively affected by increases in temperature (as determined in a Fig. 10. Exceedance probability of crop heat units (CHU) and growing season precipitation (GSP) in the baseline climate  and the near-term (2030s), mid-term (2050s) and distant future (2070s) under Shared Socioeconomic Pathway (SSP) 3-7.0 at selected locations on the Canadian Prairies. sensitivity analysis), this was more than offset by the benefits of higher CO 2 . He et al. (2018), using the DNDC model, also found that soybean yields may benefit from elevated CO 2 in the future in southern Ontario. Relatively cooler regions in present day, such as northern agricultural areas in Ontario and Quebec, might benefit even more under future climate change due to longer growing seasons and higher CHU. It is worth to note that crop models are not perfect as they do not often simulate the impacts of diseases and pests, for example, and require improvements for climate change impact studies (Kothari et al. 2022).
It is worthwhile to mention that future projections of climate conditions are based on the ensemble means of 5 GCMs in this study. The ensemble means provide essential information to policymakers and the general public, as ensemble means usually serve as a best estimate for the impact of climate change (Challinor et al. 2018). Three major sources of uncertainty in climate projections are internal climate variability, climate model response and forcing scenarios (Tebaldi and Knutti 2007;Hawkins and Sutton 2009;Deser et al. 2012). Uncertainty associated with model response can be reduced as climate models improve. However, uncertainty related to internal climate variability is difficult to reduce due to its inherent limits to climate predictability and it can be conveyed into climate projections, especially those for climate extremes. Zhang et al. (2019) claim that the uncertainty in the maize-yield simulations might mostly be from the GCM models. Therefore, multi-GCM ensembles are often recommended for climate change impact studies, for example, 20 GCMs were used by Qian et al. (2019) to account for uncertainty associated with GCMs. The number of GCMs substantially increased from approximately 40 in the CMIP Phase 5 to around 100 in Phase 6, thus often only a subset of GCMs are selected for climate change impact studies. Qian et al. (2021) assessed the effectiveness of using representative subsets of GCMs and found a 5-GCM subset could well capture the full ranges and produce full-ensemble means. In general, temperature-based indicators are more consistent across GCMs than precipitation-based indicators in this study. This is consistent with previous studies such as Li et al. (2018). Climate scenarios under three SSPs are presented in this study to demonstrate uncertainty related to forcing scenarios, that is, greenhouse gases (GHG) emission scenarios.
This study is limited to projected climate conditions for growing soybeans in Canada. Some climate indicators are based on simplified temperature thresholds for the entire growing season while soybean crops respond to air temperature differently at different growing stages. Such information is still useful for planning and developing adaptation strategies to climate change. However, this study does not provide zoning for soybean production across the country as climate is only one of the major factors determining the land suitability for growing soybeans and local soil and landscape can also play an important role (Agronomic Interpretations Working Group 1995). In addition, climate projections cannot be used as climate predictions, especially for the near term, as internal climate variability introduces large uncertainty in climate projections for the near term. As increased CHU and the extended growing season make it more suitable for growing soybeans in the semiarid Canadian Prairies, precipitation deficits may constrain such potential. Therefore, investigation of water availability for irrigated soybean production on the Canadian Prairies is needed, as well as a better understanding of economic performance and environmental impacts. Furthermore, developing cultivars that can benefit from increased CHU but are less sensitive to photoperiod may be critical for growing soybeans at more northerly latitudes in Canada in the distant future. Adopting more heatand drought-tolerant cultivars may be an important adaptation measure but Yu et al. (2021) found that their productivity under normal temperature and precipitation conditions decreased. Nevertheless, N 2 fixing soybeans reduce N 2 O emissions and sometimes reduce CO 2 emissions especially under no-till systems (Almaraz et al. 2009), thus growing soybeans may help the agricultural sector reduce GHG emissions to mitigate climate change. Aforementioned modelling studies may play an important role in quantitatively assessing climate change impacts on soybean production while mitigating climate change.

Conclusions
Climate projections, based on the ensemble of five GCMs used by the ISIMIP, showed a prolonged FFP and a longer growing season with an earlier start and more CHU across southern Canada. The magnitudes of the projected changes depend upon the SSPs used for the projections. These changes would be remarkably larger in the distant future, especially under SSP5-8.5 and SSP3-7.0. Future temperature conditions would make it possible to expand the areas suitable for growing soybeans in Canada further north and west from the current areas limited mostly in southern Ontario and Manitoba where cultivars from late-maturing MGs could grow in the future to attain higher yields. The potential expansion may be constrained in current agricultural regions.
Climate projections also indicated increased precipitation deficit that could cause moisture stress to limit the yield potential of soybeans under rainfed conditions. Combined with increased occurrences of extremely high temperatures, expansion of growing soybeans on the semiarid Canadian Prairies might be constrained by moisture deficit unless irrigation could be applied and irrigation water would be available for soybean production. In contrast, northward expansion in eastern Canada would be more favorable due to moisture availability. However, delayed growth and development may constrain growing soybeans in the high latitudes as soybean is a short-day plant. Developing heat-and droughttolerant soybean cultivars can be critical for soybean production in Canada in a changing climate.
Increasing climate variability, and the possibilities of intensified diseases and invasive pests under future warmer climate conditions, may have large impacts on soybean production in Canada, in either traditional growing areas or new regions through potential expansion. More studies are warranted to improve our understanding of the potential impacts of climate change on soybean growth and yields in Canada for developing adaptation strategies in both the short term and the long run, to reduce climate risks and increase resiliency and sustainability.

Article information
History dates