Introduction

As food security is increasingly vulnerable to rising levels of global food protectionism and unexpected war on the Ukraine, improving food self-sufficiency has resurfaced in the agricultural sector (Kim et al. 2022). Potato (Solanum tuberosum L.) is the second most important food crop in South Korea in terms of production. However, during the last two decades, potato production has decreased by 1% per year due to the reduced cultivation area (KOSIS 2021). At the same time, the volume of potato imported has increased to compensate for reduced production. Unfortunately, these trends are likely to persist, contributing to the projected fall in self-sufficiency of potatoes to levels below 70% in the near future (KREI 2021). To prepare a long-term strategy to improve self-sufficiency, it is necessary to accurately project potato yields under anticipated future climate changes and assess feasible adaptation measures for various cropping systems.

In South Korea, potatoes are grown in four cropping seasons: spring, summer, autumn, and winter (RDA 2018). Spring potato is the main cropping pattern that accounts for more than 60% of the national potato production (KOSIS 2021). Spring potato is planted from late February to late March and harvested from late May to late June in the lowlands across the country. Summer potato is cultivated in highlands, and its tuber yield and quality are higher than those of the other potato crops because of the longer growing season (mid-April to late September) and favourable weather conditions (summer temperatures within the optimum temperature range for potato growth). Autumn potato is cultivated only in the southern parts of the country, and tuber yields are the lowest among the potato crops due to the short growing season and unfavourable weather conditions (hot and wet) during early growth. Winter potato is usually grown in greenhouses and accounts for only a small portion of total production.

Climate change is projected to increase mean temperature by 2.6 and 7.0 ℃ across the Korean peninsula by the end of the twenty-first century (2081–2100) under two contrasting shared socioeconomic pathway-representative concentration pathway (SSP-RCP) scenarios: SSP126 and SSP585, respectively (NIMS 2020). The temperature increase is expected to cause supra-optimal temperatures for tuber development and bulking; thus, without adaptation, potato yield is expected to decrease in South Korea (Kim et al. 2017). Particularly, high temperatures at tuber initiation are more detrimental than those at other growth stages (Chen and Setter 2021; Kim and Lee 2019; Rykaczewska 2017). On the other hand, temperature increases in cooler regions such as northern Europe may increase potato yield by lengthening the frost-free season and potato growing season (Haverkort and Verhagen 2008). Elevated CO2 may offset some negative effects of high temperatures by enhancing photosynthesis, dry matter accumulation, and water use efficiency (Chen and Setter 2021; Leisner et al. 2018).

Process-based crop models can be used to integrate the effects of various environmental changes, including climate change, on plant growth and yield, as well as consider possible changes in agricultural management (Jones et al. 2003). A global modeling study suggested a little change or slight decline in potato yield in South Korea by 2055 (Raymundo et al. 2018). In a regional study from South Korea, summer potato yield was projected to increase in the 2050s and decrease to current levels by the 2090s (Bak et al. 2018). However, these two studies did not consider adaptation measures such as a cropping calendar shift. In a recent study, the yield gain by planting date adjustment was expected to over-compensate the adverse warming effects on spring potato yield in South Korea, and tuber yield was projected to increase up to 80% by the 2080s when CO2 fertilization effects were considered (Kim and Lee 2020).

These previous studies have yet to compare climate change impacts on different potato cropping patterns with a consistent protocol, which is essential for informing a comprehensive self-sufficiency plan. Also, projected climates have different seasonal patterns (e.g. some climate models project predominant temperature increases in summer, while others expect the most warming in winter), which can cause considerable variabilities in future crop growth and yield projections. Unfortunately, none of the previous Korean studies compared consequences of different projected climates on potato growth and yield, although five regional climate models (RCMs) were used by Kim and Lee (2020). To address these knowledge gaps, this study aims to assess and compare the effects of climate change and crop calendar adjustment on growth and yield of spring and summer potatoes. The outputs of 24 general circulation models (GCMs) from CMIP6 under four SSP-RCP scenarios (SSP126, SSP245, SSP370, and SSP585) were used to evaluate the consequences of these projected climates. Potato growth and yield were simulated using the SUBSTOR-potato model (Griffin et al. 1993), one of the most widely used potato models.

Materials and Methods

Study Sites

Study sites were selected based on previous calibration and evaluation studies of the SUBSTOR-potato model: Suwon (37.26°N, 126.98°E, 40 m above sea level) for spring potato (Kim and Lee 2020) and Daegwallyeong (37.68°N, 128.72°E, 772 m above sea level) for summer potato (Bak et al. 2018) (Fig. 1). Due to the difference in altitudes, the mean annual temperature of Suwon (12.8 ℃) is much higher than that of Daegwallyeong (7.2 ℃) (Fig. 2), which results in different potato cropping seasons. Both sites fall into the temperate monsoon climate with a typical precipitation pattern of more than 65% of annual precipitation occurring from June to September. Annual precipitation is 1318 mm and 1696 mm for Suwon and Daegwallyeong, respectively. Mean annual solar radiation is 12.5 and 13.2 MJ m−2 day−1 for Suwon and Daegwallyeong, respectively.

Fig. 1
figure 1

Locations of selected study sites in South Korea

Full size image
Fig. 2
figure 2

Monthly mean maximum and minimum temperatures, monthly mean solar radiation, and monthly total precipitation at Suwon and Daegwallyeong for the baseline period 1991–2020

Full size image

Model Description

The SUBSTOR-potato model (Griffin et al. 1993) is a process-based crop model embedded within the Decision Support Systems for Agrotechnology Transfer, DSSAT v4.7.5 (Jones et al. 2003; Hoogenboom et al. 2019). The model simulates crop development and growth in response to weather data on a daily basis. The phenological stages such as emergence and tuber initiation are simulated using temperature and day length (Raymundo et al. 2017). The maturity date is simulated as the date when the maximum tuber dry weight is reached, which is similar to the definition of physiological maturity in most cereal crops. However, potatoes are often harvested before the physiological maturity, and harvest date should be provided as an input to produce realistic outputs. Biomass production is calculated as the product of intercepted photosynthetically active radiation (PAR, MJ m−2) and radiation use efficiency (3.5 and 4.0 g DM MJ−1 PAR before and after tuber initiation). This value is further modified in response to CO2, temperature, water, and nitrogen. The potential growth rates of crop organs (tuber, leaf, stem, and root) are calculated based on the phenological stage and weather data. In a next step, actual growth of each organ is computed by partitioning the produced biomass first to tubers and then to the other organs. Five cultivar parameters are used in the model: potential leaf expansion rate (G2, cm2 m−2 day−1), potential tuber growth rate (G3, g m−2 day−1), determinacy for tuber growth (PD, dimensionless), sensitivity of tuber initiation to photoperiod (P2, dimensionless), and critical temperature for tuber initiation (TC, ℃).

Model Evaluation

The model has been evaluated for the current study sites previously: spring potato at Suwon (Kim and Lee 2020) and summer potato at Daegwallyeong (Bak et al. 2018). Particularly at Suwon, the model has been tested under broad daily mean temperature range conditions (17.2 to 28.1 ℃), providing a good measure of confidence that the model can satisfactorily capture the mean temperature response under future climates. In the present study, the model was re-evaluated with six independent field trials carried out at Suwon during 2015–2017 using the cultivar parameters of “Superior” from Kim and Lee (2020). Model performance was evaluated using the coefficient of determination (R2), root mean square error (RMSE), and index of agreement (d-index) (Willmott 1981).

Climate Data

For the baseline simulations, historical data on daily maximum and minimum temperatures, precipitation, and solar radiation from 1991 to 2020 were collected for the study sites from the Korea Meteorological Administration. Climate data for the period 2061–2090 were generated by the delta change method using the outputs of 24 GCMs (Table 1) from CMIP6 under the four SSP-RCP scenarios: SSP126, SSP245, SSP370, and SSP585 (Eyring et al. 2016). For each projected climate (24 GCMs × 4 SSP-RCP scenarios) and study site, changes in monthly mean maximum and minimum temperatures (℃), solar radiation (%), and precipitation (%) from those during the baseline period of 1991–2020 were obtained using KNMI Climate Explorer (https://climexp.knmi.nl/). Four different atmospheric CO2 concentrations were used according to the SSP-RCP scenarios, which were 460, 550, 670, and 790 ppm for SSP126, SSP245, SSP370, and SSP585, respectively. For the baseline simulations, an atmospheric CO2 concentration of 380 ppm was used.

Table 1 List of the general circulation models (GCMs) used in this study
Full size table

Soil Properties

Soil data were prepared as in Table 2. For Suwon, soil properties were measured from the field trials used for the model evaluation, whereas soil series data from the Rural Development Administration was used for Daegwallyeong. Soil water parameters required for simulations (water contents at each of the lower limit for plant uptake, the drained upper limit, and soil saturation) were generated from soil texture with the pedotransfer functions contained in the SBuild program embedded within DSSAT.

Table 2 Soil characteristics used in this study
Full size table

Crop Calendar

Planting window and the latest harvesting date were estimated using average weather across the 30 years of each projected climate (24 GCMs × 4 SSP-RCP scenarios) and study site. For the spring potato at Suwon, the earliest planting date was set to a day after the last date of the daily minimum temperature below zero (onset of the frost-free season). The latest planting date was set to the last day when the thermal time from planting to the latest harvest date exceeds the thermal time requirement for early potato, 950 ℃ days (Pulatov et al. 2015). Thermal time was computed with the relative temperature factors for vine growth (RTFVINE) used in the SUBSTOR-Potato model (Griffin et al. 1993). Thermal time was set to zero when daily mean temperature was below 2 ℃, linearly increased to 15 ℃ days at the temperature range of 2–17 ℃, maintained to a plateau value of 15 ℃ days at the temperature range of 17–24 ℃, and then linearly decreased to zero at the temperature range of 24–35 ℃. The latest harvesting date was set to a day before the onset of the rainy season since heavy rainfalls cause the losses of tuber yield and quality. The onset of the rainy season was set to the first date of the 5-day moving average precipitation above 8 mm day−1. For the summer potato at Daegwallyeong, the earliest planting date was defined as the spring potato. The latest planting date was set to the day before the onset of the rainy season since seed tubers can decay after exposure to heavy rainfalls, causing crop establishment failure. The latest harvesting date was set to a day before the first date of daily minimum temperature below zero (end of the frost-free season).

Simulation Setting

The cultivar parameters of “Superior” (1000 for G2, 20 for G3, 0.6 for PD, 0.3 for P2, and 20 for TC) were used for simulations since the cultivar occupies most of the spring and summer potato productions in South Korea. Simulations were carried out with multiple planting dates with an interval of 10 days between the earliest planting date and the latest planting date. The simulations were set to terminate if a harvest index reached 0.75 or day-of-year reached the latest harvesting date. The start date of the simulation was set to a year before the planting to estimate the soil water status at planting rather than using fixed initial values. Planting density was set to 5.7 plants m−2. Simulations were carried out for both full irrigation and rainfed potato under non-nutrient limited conditions.

Data Analyses

For each GCM, SSP-RCP, and site, the climate change impact on dry tuber yield was calculated as the 30-year mean dry tuber yield under the future climate minus that under the baseline climate. The effect of planting date adjustment was computed as the 30-year mean dry tuber yield under the future climate with optimized planting date (planting date with the highest yield for each GCM, SSP-RCP, and site) minus that with the baseline planting date (planting date with the highest yield for the baseline climate and site). The climate change impact and the effect of planting date adjustment are presented in absolute changes and relative changes. Yield variability caused by the projected climates was compared between spring and summer potato using the range and standard deviation of 30-year mean dry tuber yield across 24 GCMs × 4 SSP-RCP scenarios.

Results

Model Performance

The model showed a sufficient capability to predict the dry tuber yield from six field trials differing in planting dates (April 10, May 4, and May 14 in 2015, March 15 and March 30 in 2016, and March 17 in 2017): R2 of 0.85, RMSE of 1.36 t ha−1, and d-index of 0.92 (Fig. 3). These values are similar to those from the previous studies: R2 of 0.89, RMSE of 1.07 t ha−1, and d-index of 0.95 for high-temperature experiments (Kim and Lee 2020), and R2 of 0.93 and RMSE of 2.12 t ha−1 for the worldwide experiments (Raymundo et al. 2017).

Fig. 3
figure 3

Model performance for potato tuber yield at Suwon

Full size image

Projected Climate Changes

The changes in climatic variables in the future period of 2061–2090 compared to the baseline period of 1991–2020 are shown in Fig. 4. The annual mean maximum temperatures averaged across the GCMs under SSP126, SSP245, SSP370, and SSP585 were expected to increase by 1.7, 2.4, 3.1, and 4.0 ℃ at Suwon, respectively, and 1.7, 2.4, 3.1, and 3.9 ℃ at Daegwallyeong, respectively. The annual mean minimum temperatures averaged across the GCMs under SSP126, SSP245, SSP370, and SSP585 were projected to increase by 1.7, 2.5, 3.3, and 4.1 ℃ at Suwon, respectively, and 1.7, 2.4, 3.2, and 4.0 ℃ at Daegwallyeong, respectively. The annual mean solar radiations averaged across the GCMs under SSP126, SSP245, SSP370, and SSP585 were predicted to increase by 6.0, 4.5, 0.2, and 3.8% at Suwon, respectively, and 5.6, 4.4, 0.5, and 3.8% at Daegwallyeong, respectively. The annual precipitations averaged across the GCMs under SSP126, SSP245, SSP370, and SSP585 were expected to increase by 8.4, 10.6, 12.1, 4 and 17.5% at Suwon, respectively, and 6.6, 7.3, 9.6, and 13.5% at Daegwallyeong, respectively. The projected changes in monthly mean maximum and minimum temperatures, solar radiation, and precipitation largely varied with the GCMs, resulting in various seasonal patterns of changes in the climatic variables. For example, maximum temperatures were expected to increase prominently during the warm season (May–September) compared to the cool season (November–March) by UKESM1-0-LL f2 under SSP585, whereas IPSL-CM6A-LR under SSP585 exhibited the opposite pattern (Fig. 5).

Fig. 4
figure 4

Changes in monthly mean maximum and minimum temperatures (℃), solar radiation (%), and precipitation (%) at Suwon and Daegwallyeong for the future period of 2061–2090 under SSP126, SSP245, SSP370, and SSP585. Values are the means of 24 GCMs

Full size image
Fig. 5
figure 5

Comparison of two GCMs under SSP585 on the seasonal patterns of changes in maximum temperatures for the future period of 2061–2090 at Suwon and Daegwallyeong

Full size image

Crop Calendar

Potential growing season lengths were expected to extend under the projected climates (2061–2090) for spring and summer potato (Fig. 6). For spring potato, the earliest and latest planting dates were 73 and 109 day-of-year (DOY) under the baseline climate, and the latest harvesting date was 175 DOY. The earliest planting dates averaged across the GCMs under SSP126, SSP245, SSP370, and SSP585 were expected to be hastened by 8.7, 15.5, 19.7, and 24.8 days, respectively. The latest planting dates averaged across the GCMs under SSP126, SSP245, SSP370, and SSP585 were projected to be accelerated by 0.0, 0.8, 2.2, and 4.7 days, respectively. Meanwhile, the latest harvest dates under the future climates were not significantly different from that under the baseline climate.

Fig. 6
figure 6

Potential potato growing seasons at Suwon and Daegwallyeong under the different climates. Red boxes are potential planting windows. Error bars extend to the minimum and maximum values of the earliest and latest planting dates and the latest harvesting date across 24 GCMs

Full size image

For summer potato, the earliest and latest planting dates were 103 and 172 DOY under the baseline climate, and the latest harvesting date was 302 DOY. The earliest planting dates averaged across the GCMs under SSP126, SSP245, SSP370, and SSP585 were expected to be hastened by 8.8, 13.3, 17.0, and 20.5 days, respectively. The latest planting dates averaged across the GCMs under SSP126, SSP245, SSP370, and SSP585 were projected to be accelerated by 0.7, 1.2, 2.0, and 2.8 days, respectively. The latest harvesting dates averaged across the GCMs under SSP126, SSP245, SSP370, and SSP585 were predicted to be delayed by 14.5, 16.4, 17.5, and 21.1 days, respectively.

Climate Change Impacts on Tuber Yield

Dry tuber yields of spring and summer potato were projected to increase with climate change even without planting date adjustment and were expected to be highest under SSP245 (Fig. 7). For spring potato, dry tuber yields averaged across the GCMs under SSP126, SSP245, SSP370, and SSP585 were expected to increase by 1.0 t ha−1 (16%), 1.9 t ha−1 (30%), 1.6 t ha−1 (25%), and 1.2 t ha−1 (18%) for full irrigation potato, respectively, and 0.7 t ha−1 (16%), 1.6 t ha−1 (33%), 1.5 t ha−1 (33%), and 1.3 t ha−1 (26%) for rainfed potato, respectively. For summer potato, dry tuber yield averaged across the GCMs under SSP126, SSP245, SSP370, and SSP585 was projected to increase by 2.3 t ha−1 (17%), 3.1 t ha−1 (23%), 2.3 t ha−1 (18%), and 2.3 t ha−1 (18%) for full irrigation potato, respectively, and 2.0 t ha−1 (15%), 2.8 t ha−1 (22%), 2.1 t ha−1 (16%), and 2.1 t ha−1 (16%) for rainfed potato, respectively.

Fig. 7
figure 7

Climate change impacts on dry tuber yield of full irrigation and rainfed potato at Suwon and Daegwallyeong for the future period of 2061–2090 under SSP126, SSP245, SSP370, and SSP585 without consideration of planting date adjustment. The boxes indicate the lower and upper quartiles of 24 GCMs, and the lines inside the boxes denote the medians. Whiskers extend to the minimum and maximum values. The numbers below the box and whisker plots are the means

Full size image

The effects of planting date adjustment were significant for spring potato but not for summer potato (Fig. 8). For spring potato, the adjustment effects were largest under SSP 585. The adjustment effects averaged across the GCMs under SSP126, SSP245, SSP370, and SSP585 were expected to be 1.1, 2.3, 2.7, and 3.7 t ha−1 for full irrigation potato, respectively; 0.9, 1.9, 2.3, and 2.8 t ha−1 for rainfed potato, respectively. Meanwhile, for summer potato, the adjustment effects averaged across the GCMs under SSP126, SSP245, SSP370, and SSP585 were projected to be 0.2, 0.2, 0.3, and 0.3 t ha−1 for full irrigation potato, respectively, and 0.1, 0.1, 0.2, and 0.2 t ha−1 for rainfed potato, respectively.

Fig. 8
figure 8

Effects of planting date adjustment on dry tuber yield of full irrigation and rainfed potato at Suwon and Daegwallyeong for the future period of 2061–2090 under SSP126, SSP245, SSP370, and SSP585. The boxes indicate the lower and upper quartiles of 24 GCMs, and the lines inside the boxes denote the medians. Whiskers extend to the minimum and maximum values. The numbers below the box and whisker plots are the means

Full size image

With consideration of planting date adjustment, dry tuber yield of spring potato was projected to be highest under SSP585 (Fig. 9). Dry tuber yield averaged across the GCMs under SSP126, SSP245, SSP370, and SSP585 were expected to increase by 2.2 t ha−1 (34%), 4.2 t ha−1 (66%), 4.3 t ha−1 (67%), and 4.5 t ha−1 t ha−1 (71%) for full irrigation potato, respectively, and 1.7 t ha−1 (35%), 3.5 t ha−1 (73%), 3.9 t ha−1 (82%), and 4.1 t ha−1 (86%) for rainfed potato, respectively.

Fig. 9
figure 9

Climate change impacts on dry tuber yield of full irrigation and rainfed potato at Suwon and Daegwallyeong for the future period of 2061–2090 under SSP126, SSP245, SSP370, and SSP585 with consideration of planting date adjustment. The boxes indicate the lower and upper quartiles of 24 GCMs, and the lines inside the boxes denote the medians. Whiskers extend to the minimum and maximum values. The numbers below the box and whisker plots are the means

Full size image

Yield variability across the projected climates (24 GCMs × 4 SSP-RCP scenarios) did not differ between spring and summer potato when the effects of planting date adjustment were not considered (Fig. 7). However, when the adjustment effects were taken into account, the yield variability of spring potato was twice that of summer potato (Fig. 9). Standard deviations of dry tuber yield among the projected climates were 0.63, 0.50, 0.60, and 0.57 t ha−1 for full irrigation spring potato, rainfed spring potato, full irrigation summer potato, and rainfed summer potato, respectively, when the effects of planting date adjustment were not considered. When the adjustment effects were taken into consideration, standard deviations of dry tuber yield among the projected climates were 1.08, 1.04, 0.55, and 0.53 t ha−1 for full irrigation spring potato, rainfed spring potato, full irrigation summer potato, and rainfed summer potato, respectively.

Effects of Projected Climates on Plant Growth

A comparison of time-series plant growth between the projected climates with the highest and lowest tuber yield is shown in Fig. 10, and the changes in monthly mean maximum and minimum temperatures of the projected climates are presented in Fig. 11. For spring potato, the highest yield was associated with the extended growing season by earlier planting, and this earlier planting was possible because of the prominent increase in minimum temperatures from January to March. However, the projected climates with the highest yields (MIROC6 and INM-CM4-8 under SSP585 for full irrigation and rainfed potato, respectively) were not the projected climate with the highest increase in the minimum temperatures from January to March and the earliest planting date (CNRM-CM6-1-HR under SSP585). Under several climates, spring potato matured (harvest index > 0.75) before the latest harvesting date, indicating that the potential growing season length constrained by climatic conditions would not be a primary limiting factor of yield. Instead, excessive increases in temperatures during the late growth hastened plant maturity, reduced tuber bulking rate, and lowered tuber yield.

Fig. 10
figure 10

Comparisons of time-series tuber dry weight (solid line) and leaf area index (dashed line) between the projected climates (24 GCMs × 4 SSP-RCP scenarios) with the highest (red) and lowest (blue) tuber yield for full irrigation and rainfed potato at Suwon and Daegwallyeong. Lines and shadow areas are the means and standard deviations of the 30 years, respectively

Full size image
Fig. 11
figure 11

Changes in monthly mean maximum (Tmax) and minimum temperatures (Tmin) under the projected climates (24 GCMs × 4 SSP-RCP scenarios) with the highest (red) and lowest (blue) tuber yield for full irrigation and rainfed potato at Suwon and Daegwallyeong

Full size image

For summer potato, the lowest yield was associated with the hastened leaf senescence and reduced tuber bulking rate during the late growth caused by the large increase in temperatures from July to August (Figs. 10 and 11). Full irrigation potato yield was lowest in UKESM1-0-LL f2 under SSP585, of which temperatures from July to August were highest among all projected climates. Meanwhile, rainfed potato yield was lowest in UKESM1-0-LL f2 under SSP370, of which temperatures from July to August were highest among the GCMs under SSP370. Although the GCMs under SSP370 had favourable (lower) temperature conditions for potato growth compared to the GCMs under SSP585, in some cases, the yield gain by cooler temperatures under SSP370 could be over-compensated by the yield loss due to lower solar radiation and atmospheric CO2 concentrations under SSP370 (Figs. 4 and 9). Even so, the difference in rainfed potato yield between UKESM1-0-LL f2 under SSP370 and SSP585 was not significant, 0.2 t ha−1.

Discussion

Multiple CMIP6 model outputs (24 GCMs × 4 SSP-RCP scenarios) were used for the first time in South Korea to compare the effects of climate change and planting date adjustment on spring and summer potato and to explore growth and yield variabilities for a range of projected climates. The current simulations indicated that spring and summer potato yields are expected to increase (Fig. 7). However, the effects of planting date adjustment would be significant only for spring potato, and variabilities in growth and yield across the projected climates for spring potato are larger than those of summer potato (Figs. 8 and 9).

Spring Potato

Previously, Kim and Lee (2020) reported that spring potato yield in South Korea could be increased up to 80% in the 2080s compared to the 2000s using the five RCMs under RCP85, which is similar to the current results of 71% for full irrigation spring potato under SSP585 (Fig. 9). Slightly different values between the studies might be due to the different harvesting dates: fixed harvesting date (173 DOY or 120 days after planting) in the previous study versus harvesting date depending on the precipitation pattern and harvest index in the current study. In addition, a larger number of new CMIP6 GCMs were used in the current study.

For spring potato, yield variability among the projected climates was primarily associated with the growing season length, which is determined by the last frost date and the onset of the rainy season. The last frost date was expected to be accelerated by 8.7, 15.5, 19.7, and 24.8 days at Suwon under SSP126, SSP245, SSP370, and SSP585, respectively (Fig. 6). These values are similar to those from the previous study in South Korea (Bae et al. 2017): the last frost date in the late twenty-first century (2071–2100) is projected to be hastened by 15.0 and 23.5 days compared to the baseline (1981–2010) under RCP45 and RCP85, respectively. The onset of the rainy season was projected to be accelerated by 0.0, 0.8, 2.2, and 4.7 days at Suwon under SSP126, SSP245, SSP370, and SSP585, respectively (Fig. 6). Kwon et al. (2017) reported a similar result for East Asia (South and North Korea, China, and Japan) using 19 GCMs of CMIP5 under RCP85: The onset of the rainy season in the late twenty-first century (2080–2099) was projected to be hastened by 5 days compared to the baseline (1986–2005). Consequently, spring potato’s potential growing season was likely to be extended by 8.7, 14.7, 17.5, and 20.1 days under SSP126, SSP245, SSP370, and SSP585, respectively (Fig. 6).

Under the hottest climate, mean June temperature exceeded the upper limit of optimum temperature for photosynthesis used in the SUBSTOR-Potato model, 25 ℃ (Raymundo et al. 2018), a value similar to the optimum temperature for potato canopy photosynthesis, 24 ℃ (Timlin et al. 2006). Also, June temperature exceeded the upper limit of optimum temperature for tuber bulking rate, which is determined by the cultivar parameters, TC of 20 ℃ (0.25 × maximum temperature + 0.75 × minimum temperature). Therefore, under the hottest climate, spring potato matured before the last harvesting date (onset of the rainy season) and tuber bulking rates were slightly reduced during the late growth, which resulted in lower tuber yield than that under the mild climates.

Based on these results, we suggest that different adaptation strategies should be established for mild and severe climate change conditions. Under mild climate change conditions where potential growing season length is still limiting, genetic improvement should focus on creating frost-tolerant cultivar to extend the planting date windows to the early in the year. Meanwhile, under severe climate change conditions where high temperatures during the late growth reduce potato yield, genetic improvement should concentrate on breeding mid-late cultivars with high-temperature tolerance to delay senescence and enhance late growth.

Summer Potato

According to Bak et al. (2018), summer potato yield at Daegwallyeong is expected to increase in the 2050s while it will decrease to the baseline (2010s) level in the 2090s. This is similar to the current results in that the highest yield will be obtained under mild climate scenarios (SSP245) and slightly decrease under severe climate scenarios (SSP585) (Fig. 9). However, in the current study, summer potato yields were higher under every climate scenario than that under the baseline climate. This discrepancy may have been caused by different simulation settings and climate projections. In the previous study, fixed crop calendars were used for the baseline and future simulations with simple climate projections: + 2 ℃ and 517 ppm CO2 for the 2050s and + 4 ℃ and 895 ppm CO2 for the 2090s. However, the use of multiple GCMs under different climate scenarios, as in the current study, is a more recommended practice in climate change impact studies to illustrate the range of potential impacts and adaptations depending on future trends in global socio-economic development, which determine GHG emission levels, and to reduce uncertainty within each climate scenario (Semenov and Stratonovitch 2010).

For summer potato, the length of the frost-free season in the baseline period was 200 days, a sufficient length for early potatoes to mature. Therefore, yield variability among the projected climates was associated with different CO2 fertilization effect among the SSPs and the accelerated leaf senescence and reduced tuber bulking rate by excessive increases in summer temperatures (Figs. 10 and 11). Under severe climate change conditions, temperatures of July and August exceeded the upper limits of optimum temperature for photosynthesis (25 ℃) and tuber bulking rate (TC of 20 ℃). Based on these results, unlike spring potato, the adaptation strategy for summer potato would not significantly vary among projected climates. Genetic improvement should focus on breeding mid-late cultivars with high-temperature tolerance to extend growing season length and enhance late growth.

Temperature and CO2 Responses

Potato is vulnerable to high-temperature episodes (Obiero et al. 2021), and the magnitude of high-temperature stress varied with growth stages (Chen and Setter 2021; Kim and Lee 2019; Rykaczewska 2017; Struik 2007). Kim and Lee (2019) reported that early potato yield is reduced by high temperatures at tuber initiation but not at tuber bulking. More specifically, high night temperatures at tuber initiation reduced tuber yield by delaying tuber development, reducing the yield proportion of large tubers, and lowering the harvest index. Meanwhile, high day temperatures at tuber initiation reduced tuber yield by decreasing the photosynthetic rate. Similarly, Chen and Setter (2021) reported that the dry matter accumulation rate of potatoes was inhibited by high temperature twice as much at tuber initiation than at tuber bulking. To our best knowledge, potato models, including the SUBSTOR-potato model, have never been tested for simulating stresses from high-temperature episodes, which would be more detrimental than the stresses from increased mean growing season temperatures.

Potato plants exhibit photosynthetic acclimation under long-term exposure to elevated CO2 concentration. Lawson et al. (2001) reported that elevated CO2 initially increases the assimilation rate of potatoes; however, the beneficial effects could not be sustained to maturity due to the photosynthetic acclimation and accelerated senescence. Furthermore, elevated CO2 concentration reduces stomatal conductance and transpiration cooling effect, thereby increasing canopy temperatures and possibly high-temperature stresses (Webber et al. 2018 and references therein). Since the SUBSTOR-Potato model does not yet capture the abovementioned phenomena, the beneficial effects of climate changes on spring and summer potato in the real world might be lower than those in the current study.

Limitations

There are several limitations to the current study. A single early potato cultivar, “Superior,” was used for the simulations, and genetic variability among cultivars was not considered due to the limited information on other cultivars to evaluate the model. Also, the current study did not account for breeding activities that may lead to more resilient potato genotypes, such as frost- and heat-tolerant cultivars. For example, a new cultivar, “Dami,” has been released, which has a higher resistance to high-temperature induced physiological tuber disorder than “Superior” (Park et al. 2019). Further simulation studies should consider a wide range of maturity groups, as in Rana et al. (2020) and climate-resilient cultivars, and field experiments should be conducted with various cultivars to calibrate and evaluate the model before the simulations.

Plastic mulch is a common agricultural management practice in South Korean potato production. Plastic mulch improves early shoot growth, such as emergence and groundcover, by increasing soil temperature, maintaining soil moisture, and controlling weeds (Chang et al. 2016). However, the effects of plastic mulch on soil temperatures and soil water dynamics are not considered by most potato models, including the SUBSTOR-potato model. Therefore, coupling crop models to hydrological models that can simulate plastic mulch effects, such as HYDRUS-2D (Filipović et al. 2016; Zhang et al. 2018), may provide better estimates. Fortunately, there was an attempt to couple the DDSAT crop model with the HYDRUS-1D hydrological model to improve soil water dynamic simulation (Shelia et al. 2018). However, this coupled model has never been tested for simulating soil temperature and water dynamics under the fields covered with plastic mulch.

This study did not use an ensemble of crop models, although this has been reported to provide better estimates than any individual model (Martre et al. 2015). For potato, Fleisher et al. (2017) conducted a multi-model assessment using nine potato models to quantify variations among the models and evaluate the responses to climate change. Their results indicated that temperature increases caused the largest inter-model uncertainty compared to CO2 concentration and precipitation changes. Therefore, the current results should be checked with other potato models to make more realistic predictions since the yield uncertainty caused by projected climates in the current study was primarily related to the temperature increases, which caused the extension of the growing season and high-temperature stresses on potato growth.

Conclusions

The current results demonstrated that spring and summer potato yields in South Korea are expected to increase under future climate change; however, the effects of planting date adjustment would be significant only for spring potato. Furthermore, different breeding goals (frost tolerance or high-temperature tolerance) could be established under different climate change conditions for spring potato, but only high-temperature tolerant cultivars may increase the tuber yield of summer potato. The current optimistic results should be carefully interpreted because the SUBSTOR-Potato model does not fully reflect the effect of high-temperature episodes and the interactive effect between CO2 and temperature on potato growth and yield, which may reduce the beneficial effects of climate change.