Temperature Trends and Accumulation of Chill Hours, Chill Units, and Chill Portions in South Carolina ()
1. Introduction
A recent climate report by the Intergovernmental Panel on Climate Change (IPCC, 2022) [1] expressed with very high confidence that temperature increases reaching 1.5˚C in the near term (2021 to 2040) due to global warming would cause increases in climate-related hazards and risks to ecosystems and humans. The report also expressed, with a high or very high level of confidence, that some of the impacts of the expected warming would include range shifts of terrestrial species and changes in timing (phenology), both globally and in North America. Agricultural production is intimately related to the long-term climate and the short-term weather conditions. Therefore, the expected increases in temperature in the near term could positively and negatively impact local agricultural production systems, depending on specific regional characteristics. For example, a positive impact on local agriculture could be the ability to produce new crops in the warmer climate. On the other hand, dealing with potential new invasive pests prospering in the warmer climate (i.e., weeds, insects, etc.) would be an example of a negative impact.
One of the many potential concerns of a warming climate in agriculture is the potential impact of higher temperatures on the accumulation of chill hours, chill units, or chill portions needed to break the dormancy of many temperate fruit trees. This issue is already creating concerns among many researchers around the world. For example, Asse et al. (2018) [2] observed that spring phenological phases, such as budburst and flowering, have tended to occur earlier in some tree species because of temperature increases in the Alps. Further, they found that although winter warming might be beneficial in reducing the risk of late spring frost, they also warned that this effect was expected to become detrimental if the chilling requirement to break dormancy was not met. Prudencio et al. (2018) [3] observed decreased productivity in extra-late and ultra-late-flowering almond cultivars in a warm season when chilling requirements were not met. Darbyshire et al. (2017) [4] , evaluating different apple phenology models in many locations, found that apple trees flowered later in sites with warmer winters. Delgado et al. (2021) [5] , modeling the potential impact of future climate on apple trees in Spain, found that projected winter chill might decrease by 9 to 12 chill portions under an intermediate global warming scenario and by 9 to 24 chill portions under a pessimistic scenario. This reduction in chill portions is expected to affect the timing of dormancy break in the future significantly. However, Martínez-Lüscher et al. (2017) [6] found that apricot flowering time in the United Kingdom remained relatively unchanged despite significant temperature increases over several decades (1960 to 2014).
The development and selection of appropriate cultivars adaptable to local conditions, among other farming practices, should play a significant role in adapting to future climate conditions. For example, Rouse and Sherman (2003) [7] reported on low-chill peach cultivars adapted to the relatively warm climate prevalent in Florida, USA. Similarly, Delgado et al. (2021) [5] found considerable variability in the chilling requirements of ten apple cultivars (ranging from 59 to 90 chill portions), suggesting that it would be feasible to select appropriate apple cultivars according to current and expected local conditions. However, since planting fruit trees is a long-term investment, local farmers should have a reasonable estimate of the future climate before properly selecting and planting fruit trees.
Peach production, which is affected by chill unit accumulation, is an important economic activity in South Carolina. According to the USDA National Agriculture Statistics Service (NASS, 2022) [8] , South Carolina’s utilized peach production in 2021 was 72,630 tons, valued at US$106.151 million. By comparison, USA’s utilized peach production was 61,890 tons, valued at US$624.366 million. Therefore, South Carolina’s utilized peach production represented around 11% of the USA’s production and 17% of the economic value. Rising temperatures in South Carolina could affect chill accumulation, potentially affecting peach production in the state. It is crucial to investigate the exposure of the peach industry to potential changes in the local climate. Therefore, the objective of this study was to determine if there had been a significant change in air temperatures and chill hours, chill units, and chill portion accumulation in South Carolina over the last two decades.
2. Methodology
2.1. Data Collection and Site Description
Data from a permanent network of electronic weather stations located in each county in South Carolina were used for this study. The data were obtained from the Web Service API made available by the Applied Climate Information System (ACIS) (http://data.rcc-acis.org/). An R script (R Core Team, 2021) [9] was developed to download the required data using HTTP requests following the guidelines described at http://www.rcc-acis.org/docs_webservices.html. The data included the daily maximum and minimum air temperature (Tmax and Tmin) and the weather station’s geographic location (latitude and longitude). The daily Tmax and Tmin data were downloaded in CSV format using a station data request (StnData). The latitude and longitude were downloaded in JSON format using a station metadata request (StnMeta).
The above data were collected from thirty-one of the forty-six counties in South Carolina. These counties were selected because they had at least twenty years of temperature data that were available online. The thirty-one counties included in this study provided a good representation of the state’s different regions, going from the mountainous upper-land areas in the northern part of the state to the lowlands of the coastal areas (Figure 1). The changes in elevations in the selected counties went from around 1000 m above sea level in Pickens County to the lowlands located in the state’s coastal areas, such as Beaufort County, with elevations near sea level. The location of the counties represented a range of latitudes and elevations, which translated into considerable differences in temperatures among counties.
2.2. Estimation of Hourly Temperatures
Hourly temperature data were only available online for a few South Carolina
Figure 1. Locations of temperature measuring sites in South Carolina.
stations. These stations reporting hourly data were typically airport weather stations located away from agricultural settings. Therefore, the hourly temperatures [T(t)] needed to calculate chill hours, chill units, and chill portions were estimated from the daily Tmax and Tmin air temperatures using the procedure proposed by Linvill (1990) [10] and later implemented by Linsley-Noakes et al., (1995) [11] , and Delgado et al., (2021) [5] .
2.3. Calculation of Hourly and Daily Chill Values
The hourly chill hours were calculated from the hourly temperatures using the T(t) < 7.2˚C model (Linvill, 1990 [10] ; Miranda et al., 2013 [12] ; Payero and Sekaran, 2021 [13] ) and the 0˚C < T(t) < 7.2˚C model (Zang and Taylor, 2011 [14] ). The chill units were calculated using the Utah model (Richardson et al., 1974 [15] ; Alburquerque et al., 2008 [16] ). Chill portions were computed using the Dynamic model (Fishman et al., 1987 [17] ; Fishman et al., 1987 [18] ; Zhang and Taylor, 2011 [14] ; Miranda et al., 2013 [12] ; Melke, 2015 [19] ). For the T(t) < 7.2˚C and the 0˚C < T(t) < 7.2˚C model, the hourly chill hours were calculated following the specified temperature ranges. The Utah model’s hourly chill units were calculated according to the temperature ranges in Table 1.
The hourly chill portions using the Dynamic model were calculated using the Dynamic_Model() function from the R package chillR (Luedeling and Fernandez, 2022 [20] ). The daily chill values (hours, units, or portions) for each chill model were calculated by adding the hourly chill values for each 24-hour period (starting from sunrise). The annual chill values were calculated for each year and county by adding the daily chill values between Oct. 1st and Apr. 20th.
2.4. Statistical Analyses
Statistical analyses for this study were conducted using R (R Core Team, 2021 [9] ). Linear regression analyses were used to determine whether the changes in mean temperature (Tmean = [Tmax + Tmin]/2) and annual chill values observed over the previous two decades were statistically significant (α = 0.05). A regression analysis was conducted using year as the independent variable and Tmean as the dependent variable for each county. Regression analyses were performed for each county and chill model using year as the independent variable and annual chill value as the dependent variable. A positive slope of the line indicated that the variable tended to increase over time, and a negative slope showed the opposite trend. The p-value of the slope showed if the increasing or decreasing trend was statistically significant (different from zero at α = 0.05).
Table 1. Definitions of the Utah model.
3. Results and Discussion
3.1. Historical Trends in Temperature
An example of the regression analyses conducted to evaluate the historical trend in temperature (Tmean) for four counties in South Carolina is shown in Figure 2. For Barnwell and Beaufort counties, the slope of the lines was positive, indicating that the temperature tended to increase over the previous two decades at 0.101˚C and 0.117˚C per year, respectively. The slope of the relationship for both counties had a p-value < 0.05, meaning that the observed increase in temperature was statistically significant. Marion and Spartanburg counties, on the other hand, had a p-value > 0.05, which indicates that the observed temperature changes were not statistically significant for these counties.
The weather station coordinates, temperature, and regression results for the temperature trend by county are shown in Table 2. These results indicate that except for Marion County, the regression analyses for all the other counties resulted in a positive slope, signifying that the general tendency had been for the temperature to increase over the previous two decades. However, the increase in
Figure 2. Regression analysis of mean air temperature (Tmean) and year for four counties in South Carolina. The dashed red line is the regression line. (a) Barnwell county; (b) Beaufort county; (c) Marion county; (d) Spartanburg county.
Table 2. Average air temperature data and regression analysis results of mean air temperature (Tmean) and year for each county in South Carolina from 2002 to 2022 (S = significant, NS = not significant at α = 0.05).
temperature was statistically significant for 20 out of the 31 counties (64 %). The overall temperature rise averaged 0.072˚C per year for all 31 counties. The increase was even higher (average = 0.089˚C per year) for the 20 counties with a significant temperature increase over the previous two decades.
3.2. Historical Trends in Chill Values
Examples of the regression analyses conducted to evaluate the historical trend in annual chill hours (using the T(t) < 7.2˚C model) for four counties in South Carolina are shown in Figure 3. Because of the increasing trends in temperature, Barnwell and Beaufort counties had decreasing trends in annual chill hours, resulting in p-values < 0.05. These results indicate that the decreasing trends in chill hours accumulation over the previous two decades were statistically significant for these counties. For Marion and Spartanburg counties, on the other hand, the observed trends in chill hours were not statistically significant (p-values > 0.05).
Figure 3. Regression analysis of annual Chill hours and year for four counties in South Carolina. The dashed red line is the regression line. (a) Barnwell county; (b) Beaufort county; (c) Marion county; (d) Spartanburg county.
The summary statistics for the annual chill values and the results of regression analysis for each county using the four chill models are shown in Table 3 for the T(t) < 7.2˚C model, Table 4 for the 0˚C < T(t) < 7.2˚C model, Table 5 for the Utah model, and Table 6 for the Dynamic model. Because of the different definitions of what a chill value represents for each model, there were considerable variations in the magnitude of the annual chill value results among models. For example, on average, for all counties, the annual chill values were 1536 chill hours, 1178 chill hours, 432 chill units, and 75 chill portions for the T(t) < 7.2˚C, 0˚C < T(t) < 7.2˚C, Utah, and Dynamic model, respectively.
3.2.1. Results of the T(t) < 7.2˚C Model (Chill Hours)
Table 3 shows that the regression analysis for each county always resulted in a negative slope. The negative slope means that the general tendency had been for the annual chill hours in South Carolina to decrease over the previous two decades. Although the slope of the line varied widely by county, the average slope for all counties was −17.7, representing an average decrease of 17.7 chill hours per year. Using this model, the decline in chill hours was statistically significant for 20 of the 31 counties included in this study, which coincided with the results reported for the temperature data.
3.2.2. Results of the 0˚C < T(t) < 7.2˚C Model (Chill Hours)
The linear regressions for the 0˚C < T(t) < 7.2˚C model (Table 4) resulted in a negative slope for all counties except for Marion County. The overall slope for all the counties was −8.6, representing an average decrease in chill hours calculated with this model of 8.6 chill hours per year. However, the p-values indicate that using this model, which only accounts for temperatures above freezing when calculating chill hours, only 11 out of the 31 counties experienced a statistically significant decrease in chill hours over the previous two decades.
3.2.3. Results of the Utah Model (Chill Units)
The results for the Utah model (Table 5) also show a negative slope for all counties except for Marion County. The average slope for all counties was −17.0, similar to the slope obtained with the T(t) < 7.2˚C model. However, since the chill units calculated with the Utah model are not equivalent to chill hours, the two results are not comparable. However, the general tendency for this model had also been for the chill units to decrease over time. Yet, the p-values indicate that only 9 of the 31 counties resulted in a significant decrease in chill units calculated using the Utah model.
3.2.4. Results of the Dynamic Model (Chill Portions)
The regression results of the Dynamic model (Table 6) show that four counties (Darlington, Georgetown, Marion, and Union) resulted in positive slopes. In contrast, the other twenty-seven counties resulted in negative slopes. The overall average slope obtained with the Dynamic model for all counties was −0.395, representing an average decrease of around 0.4 chill portions per year. However,
Table 3. Average chill hours accumulated using the T(t) < 7.2˚C model and regression analysis results between chill hours versus year for each county in South Carolina from 2002 to 2022 (S = significant, NS = not signifi-cant at α = 0.05).
Table 4. Average chill hours accumulated using the 0˚C < T(t) < 7.2˚C model and regression analysis results between chill hours versus year for each county in South Carolina from 2002 to 2022 (S = significant, NS = not significant at α = 0.05).
Table 5. Average chill units accumulated using the Utah model and regression analysis results between chill units versus year for each county in South Carolina from 2002 to 2022 (S = significant, NS = not significant at α = 0.05).
Table 6. Average chill portions accumulated using the Dynamic model and regression analysis results between chill portions versus year for each county in South Carolina from 2002 to 2022 (S = significant, NS = not significant at α = 0.05).
only Beaufort, Berkeley, and Florence counties significantly reduced chill portions calculated with the Dynamic model. These results suggest that the Dynamic model was less sensitive to the observed changes in temperature than the other models.
3.3. Annual Chill Accumulation Maps
Average annual chill accumulation maps of South Carolina using the T(t) < 7.2˚C, 0˚C < T(t) < 7.2˚C, Utah, and Dynamic models, are shown in Figures 4-7, respectively. An equal distance procedure was used to divide the state into five geographical areas (colored zones) according to the average annual chill accumulation. At regular intervals, lines of equal chill accumulation (red lines) were also drawn to create additional zones. The resulting maps show some similarities and some differences among the four models. The zones in each map identify areas with similar chilling conditions, which could be used to locate new areas for growing crops with chilling requirements, as suggested by Linvill (1990) [10] . In general, the chill accumulation in South Carolina tended to increase from south to north and from the coastal areas in the southeast to the mountainous regions in the northwestern part of the state.
Figure 4. Average annual chill hours map of South Carolina using the T(t) < 7.2˚C model. The red lines are the contour lines for annual chill hours.
Figure 5. Average annual chill hours map of South Carolina using the 0˚C < T(t) < 7.2˚C model. The red lines are the contour lines for annual chill hours.
Figure 6. Average annual chill unit map of South Carolina using the Utah model. The red lines are the contour lines for annual chill units.
Figure 7. Average annual chill portions map of South Carolina using the Dynamic model. The red lines are the contour lines for annual chill portions.
4. Conclusion
This study addressed whether air temperatures had increased in different counties in South Carolina in the previous two decades. Another question explored was whether the observed temperature changes have significantly impacted the annual chill hours, chill units, and chill portions calculated using four different chill models. The regression analyses between air temperature and year for the various counties showed that temperatures in South Carolina had tended to increase over time for all counties included in the study. The increase in temperature was significant for most of the counties, representing an average increase of 0.089˚C per year. Consequently, due to the temperature increases, the general tendency was for the annual accumulated chill hours, chill units, and chill portions to decrease over the previous two decades. However, there were variations in the sensitivity of the different chill models to the observed changes in temperature. The T(t) < 7.2˚C model was the most sensitive to the observed temperature changes, while the dynamic model was the least sensitive. The results of this study could be used as a warning for the peach and other fruit tree industries in South Carolina that are sensitive to chilling requirements to evaluate the potential impacts of the observed trends and start planning adaptation strategies. This study also developed annual chill maps for South Carolina using the results of the four chill models. These maps could be used to visualize the different chill regions of the state, which could help farmers determine the most appropriate zones for planting future fruit trees.
Acknowledgements
Technical contribution No. 7280 of the Clemson University Experiment Station. This material is based upon work supported by NIFA/USDA, under project number SC-1700593. Additional funding was provided by USDA-NRCS Project number 69-3A75-17-274.