Meteorological Drought Variability and Its Impact on Wheat Yields across South Africa

Drought is one of the natural hazards that have negatively affected the agricultural sector worldwide. The aims of this study were to track drought characteristics (duration (DD), severity (DS), and frequency (DF)) in South Africa between 2002 and 2021 and to evaluate its impact on wheat production. Climate data were collected from the South African Weather Service (SAWS) along with wheat yield data from the Department of Agriculture, Forestry and Fisheries (2002–2021). The standard precipitation index (SPI) was calculated on 3-, 6-, 9-, and 12-month time scales, and the trend was then tracked using the Mann–Kendall (MK) test. To signify the climatic effects on crop yield, the standardized yield residual series (SYRS) was computed along with the crop-drought resilience factor (CR) on a provincial scale (2002–2021). The output of the SPI analysis for 32 stations covering all of South Africa indicates a drought tendency across the country. On a regional scale, western coastal provinces (WES-C and NR-C) have been more vulnerable to meteorological droughts over the past 20 years. Positive correlation results between SYRS and wheat yield indicate that the WES-C province was highly influenced by drought during all stages of wheat growth (Apr–Nov). Historical drought spells in 2003, 2009, and 2010 with low CR = 0.64 caused the province to be highly impacted by the negative impacts of droughts on yield loss. Overall, drought events have historically impacted the western part of the country and dominated in the coastal area. Thus, mitigation plans should be commenced, and priority should be given to this region. These findings can assist policymakers in budgeting for irrigation demand in rainfed agricultural regions.


Introduction
Climate change has been observed globally [1]. According to the Intergovernmental Panel on Climate Change (IPCC), the universal mean surface temperature is predicted to rise by 1.5-2 • C by 2100 [2,3]. The atmospheric carbon dioxide (CO 2 ) concentration is expected to rise to the range of 540 to 970 parts per million around the 2100 period [4]. When the concentration levels of greenhouse gases such as carbon dioxide, methane, water vapor, and nitrogen oxide rise in the atmosphere, the Earth's surface temperature increases [5].
Abebe et al. [38] reported that more than 50% variability in crop yield was explained by the SPI on Ethiopian agricultural land. Similar studies assessing drought affecting multiple crop yields have been conducted in Central Malawi and Ethiopia using meteorological indices such as the SPI, SPEI, and PDSI [39,40]. Other than meteorological drought indices, composite drought indices such as the crop drought vulnerability index (CDVI) derived from the SPI are also being utilized to measure the vulnerability of crop yield towards droughts [41].
South Africa is one of the drought-prone African countries and has experienced several interconnected climatic extremes, including drought [42], floods [43], and heat stress [44]. This has resulted in infrastructure destruction, agricultural damage, and the loss of human lives [45,46]. It is also the second-largest wheat-producing region in Sub-Saharan Africa, with a mix of irrigated and dryland production. However, the low relative production of wheat in this region is attributed to droughts and heat stress, along with biotic diseases [47]. The recent drought in 2015-2017 in the Western Cape (i.e., the largest wheat-producing region of South Africa) strongly affected wheat production and reduced its exports [48,49]. Thus, drought is one of the most catastrophic weather events in South Africa, with a drastic impact on the agricultural and environmental sectors [50]. Furthermore, Mpandeli et al. [51] reported a rise in the drought intensity and frequency in South Africa due to increasing temperatures and a rainfall deficit . In addition to this, another study conducted by Lottering et al. [52] also indicated that local areas such as "uMsinga" in the "KwaZulu-Natal" province of South Africa experienced droughtthreatened agricultural productivity on small-scale farms. Thus, climatic pressure induced by droughts is largely affecting the agricultural production system in this region due to inadequate irrigation water management [53]. Future climatic models of the region also reveal an approximate 20% decline in rainfall in the next 3 decades, which will ultimately be responsible for below-average crop yields [54].
Hence, keeping in view the background of the study and region, the rainfall-based meteorological assessment of drought and its interrelationship with wheat yield is muchneeded research of this time. Although some studies have been conducted in this region on a smaller scale using multiple drought indices to monitor agricultural droughts in the region [55,56], to the best of our knowledge, no specific studies have been conducted to address long-term meteorological droughts and their impact on wheat yields on a broader scale in this region. Thus, the aim of this study is two-fold, with a focus on the regional scale covering 31 meteorological stations: 1.
To examine and analyze the variability in short-term drought (SPI-3) occurrence and trends at all meteorological stations in the region over a time period of the past 20 years (2002-2021); 2.
To explore the impacts of short-term meteorological droughts (SPI-3) on wheat yield loss and its resistance using a standardized yield residual series (SYRS) and the crop drought resistance factor (CR) in all provinces in the region.
This study provides fruitful findings on a historical baseline to address this issue for better drought hazard management in the future at the regional level.

Study Area and Data Collection
South Africa is situated between latitudes 22 and 35 • S and longitudes 17 and 33 • E and is neighbored by two oceans, the Indian Ocean in the east and the Atlantic Ocean in the west. The country spans a land area of 1,220,813 km 2 , partitions political boundaries with Mozambique, Zimbabwe, Botswana, Namibia, and Eswatini (Swaziland), and totally landlocks Lesotho [57] (Figure 1). It is described as a country with a semiarid climate that lies in the subtropics and the mid-latitudes [58].
Wheat is the main staple crop of the region after maize, with dryland and irrigated production. Climatic variations in the region make it vulnerable to extreme weather events such as drought and heat stress [47]. Rainfall in the region is quite variable from the eastern to western coastline based on the movement of oceanic currents [60]. The unpredicted rainfall variability in different seasons exposes wheat production to more climatic extremes [56].

Data Collection
For the drought analysis, the available historical climatic data were commissioned from the South African Weather Service (SAWS). Rainfall and temperature data spanning the 2000-2021 period was collected from 31 meteorological stations covering the whole region of South Africa (Table 1). Data quality and homogeneity were ensured by the South African Weather Service. The average temperature was 30.22 °C in January and 23.29 °C in June, while the rainfall ranged between 108.44 mm in January and 2.5 mm in August ( Figure 2). South Africa is populated by 59.62 million people, and the major industrial activities are manufacturing, financial services, mining, tourism and trade, agriculture, and telecommunications [59]. On a regional scale, the country is divided into nine large provinces. Wheat is the main staple crop of the region after maize, with dryland and irrigated production. Climatic variations in the region make it vulnerable to extreme weather events such as drought and heat stress [47]. Rainfall in the region is quite variable from the eastern to western coastline based on the movement of oceanic currents [60]. The unpredicted rainfall variability in different seasons exposes wheat production to more climatic extremes [56].

Data Collection
For the drought analysis, the available historical climatic data were commissioned from the South African Weather Service (SAWS). Rainfall and temperature data spanning the 2000-2021 period was collected from 31 meteorological stations covering the whole region of South Africa (Table 1). Data quality and homogeneity were ensured by the South African Weather Service. The average temperature was 30.22 • C in January and 23.29 • C in June, while the rainfall ranged between 108.44 mm in January and 2.5 mm in August ( Figure 2).
Wheat yield data at the provincial level (Table 1) were collated from the Crop Estimates Committee of the Department of Agriculture, Forestry and Fisheries for the period from 2000 to 2021. Wheat plays a crucial role in the agricultural economy of the country [47]. The wheat growth cycle in the region prevails from April to November, with slight variations in its summer and winter characteristics [61]. Irrigated wheat is planted in summer rainfall regions in eastern states from mid-May to the end of July [62], while in winter rain areas such as WES-C, it is planted from mid-April to mid-June [63]. The harvesting period of wheat in the region runs from October to November [61]. Between 500,000 and 900,000 hectares of cereal crop is cultivated per annum, with a mean yearly output of 1.3 to 2.4 million tons between 2000 and 2021. The area under irrigation produces approximately 5 tons per hectare on an annual basis, whereas the dryland produces 2-2.5 tons per hectare [64].

Standard Precipitation Index (SPI)
McKee et al. [11] prescribe the standard precipitation index (SPI) to capture the spatiotemporal variation in drought properties [65]. The SPI uses long-term monthly rainfall as input data and computes the divergence in rainfall from the average numerical parameter in a certain region during a specific time span [66]. The probability density function, such as the gamma statistical distribution function, is fitted to the rainfall data, which, according to Lloyd−Hughes and Saunders in [67], fits very well. The normalization of the gamma cumulative distribution function then follows [68]. The index can assimilate the drought duration, amount, and intensity on different time scales (3, 6, 9, 12, and 24 months). More details about the SPI calculation and classification can be found in McKee et al. [11]. Table 2 shows the categories of SPI index values [11]. According to Table 2, drought can be categorized according to specific SPI values, with extreme drought corresponding to less than −2 [69,70].  Wheat yield data at the provincial level (Table 1) were collated from the Crop Estimates Committee of the Department of Agriculture, Forestry and Fisheries for the period from 2000 to 2021. Wheat plays a crucial role in the agricultural economy of the country [47]. The wheat growth cycle in the region prevails from April to November, with slight variations in its summer and winter characteristics [61]. Irrigated wheat is planted in summer rainfall regions in eastern states from mid-May to the end of July [62], while in winter rain areas such as WES−C, it is planted from mid-April to mid-June [63]. The harvesting period of wheat in the region runs from October to November [61].
Between 500,000 and 900,000 hectares of cereal crop is cultivated per annum, with a mean yearly output of 1.3 to 2.4 million tons between 2000 and 2021. The area under irrigation produces approximately 5 tons per hectare on an annual basis, whereas the dryland produces 2-2.5 tons per hectare [64].

Standard Precipitation Index (SPI)
McKee et al. [11] prescribe the standard precipitation index (SPI) to capture the spatiotemporal variation in drought properties [65]. The SPI uses long-term monthly rainfall as input data and computes the divergence in rainfall from the average numerical parameter in a certain region during a specific time span [66]. The probability density function, such as the gamma statistical distribution function, is fitted to the rainfall data, which, according to Lloyd−Hughes and Saunders in [67], fits very well. The normalization of the gamma cumulative distribution function then follows [68]. The index can assimilate the drought duration, amount, and intensity on different time scales (3, 6, 9, 12, and 24 Wheat yield data at the provincial level (Table 1) were collated from the Crop Estimates Committee of the Department of Agriculture, Forestry and Fisheries for the period from 2000 to 2021. Wheat plays a crucial role in the agricultural economy of the country [47]. The wheat growth cycle in the region prevails from April to November, with slight variations in its summer and winter characteristics [61]. Irrigated wheat is planted in summer rainfall regions in eastern states from mid-May to the end of July [62], while in winter rain areas such as WES−C, it is planted from mid-April to mid-June [63]. The harvesting period of wheat in the region runs from October to November [61].
Between 500,000 and 900,000 hectares of cereal crop is cultivated per annum, with a mean yearly output of 1.3 to 2.4 million tons between 2000 and 2021. The area under irrigation produces approximately 5 tons per hectare on an annual basis, whereas the dryland produces 2-2.5 tons per hectare [64].

Standard Precipitation Index (SPI)
McKee et al. [11] prescribe the standard precipitation index (SPI) to capture the spatiotemporal variation in drought properties [65]. The SPI uses long-term monthly rainfall as input data and computes the divergence in rainfall from the average numerical parameter in a certain region during a specific time span [66]. The probability density function, such as the gamma statistical distribution function, is fitted to the rainfall data, which, according to Lloyd−Hughes and Saunders in [67], fits very well. The normalization of the gamma cumulative distribution function then follows [68]. The index can assimilate the drought duration, amount, and intensity on different time scales (3, 6, 9, 12, and 24 ) mean; ( ____ ) median. Table 2. Classification of standard precipitation index (SPI) for drought studies.

. Drought Trend and Characteristics
The Mann-Kendall (MK) test [71] is applied to assess the monotonic tendencies of studied variables. This statistical test is nonparametric and assumes no normality but independent data [72]. In this case, the null (H 0 ) assumes that there is no tendency in the studied variable, whereas the alternative hypothesis (H a ) presumes there is a tendency [73]. In addition, the Sen slope estimator was implemented to capture the values of changes during the study period [74].
Moreover, the drought characteristics computed in this study include the drought duration (DD) in number of months and the drought sum (DS), which defines the sum of all SPI values during a particular drought spell in months or years. The frequency of drought events is calculated by employing the equation [75,76] where n s represents the number of drought events during the selected time, and N s represents the total number of months, i.e., 240 in the currently selected period of 20 years (2002-2021).

Drought Impact on the Agricultural Sector
South African cereal production, particularly wheat ( Figure 3), has been steadily increasing, mainly due to the expansion of cultivated land, an increase in irrigated areas, and the adoption of advanced agricultural inputs, such as improved seed cultivars [77]. To investigate the intercorrelation between agricultural drought (SPI−3) and wheat yield, the standardized yield residual series (SYRS) was calculated on a regional scale. A polynomial regression model was deployed to offset climatic, economic, and technological factors [78,79]. Yield variations due to nonclimatic factors (e.g., fertilizers, improved seed breeds, and irrigation operations) were separated using the detrending technique, and the remaining detrended yield was employed [80]. To signify the climatic effects on crop yield, the standardized yield residual series (SYRS) was computed employing the following equation: where represents the residuals of the detrended yield; μ is the mean of the residuals of the detrended yield; and σ denotes the standard deviation (i: year). Table 3 shows the categories of the SYRS. Evolution of wheat yield on provincial scale over the selected time of 20 years is shown in Figure 3 where red line presents the yield trend and black dotted line presents the significance level.
To investigate the intercorrelation between agricultural drought (SPI-3) and wheat yield, the standardized yield residual series (SYRS) was calculated on a regional scale. A polynomial regression model was deployed to offset climatic, economic, and technological factors [78,79]. Yield variations due to nonclimatic factors (e.g., fertilizers, improved seed breeds, and irrigation operations) were separated using the detrending technique, and the remaining detrended yield was employed [80]. To signify the climatic effects on crop yield, the standardized yield residual series (SYRS) was computed employing the following equation: where X i represents the residuals of the detrended yield; µ is the mean of the residuals of the detrended yield; and σ denotes the standard deviation (i: year). Table 3 shows the categories of the SYRS. Table 3. Classification of SYRS and CR.

SYRS Values SYRS Classes CR Values CR Classes
To showcase the impact of seasonal agrarian drought (SPI-3) on wheat production on a regional scale, the crop drought resilience factor (CR) was computed per province across South Africa. Mohammed et al. [81] define the CR as the ability of the cultivated crop to resist environmental pressures (such as drought) while maintaining its physiological, biochemical, and morphological duties. The CR [82] was calculated by the following equation:

SYRS Values SYRS Classes CR Values CR Classes
Acceptable losses due to drought 0.9 < CR < 1 Slightly nonresilient To showcase the impact of seasonal agrarian drought (SPI−3) on wheat production on a regional scale, the crop drought resilience factor (CR) was computed per province across South Africa. Mohammed et al. [81] define the CR as the ability of the cultivated crop to resist environmental pressures (such as drought) while maintaining its physiological, biochemical, and morphological duties. The CR [82] was calculated by the following equation: where denotes values of yield in drought production seasons at the provincial level, and represents values of detrended yield in a similar production season. Table 3 [82] depicts the CR categorization.

Correlation Analysis between Crop Yields and Agricultural Drought Indices
A nonlinear regression model was employed to explore the temporal relationship between drought indices and the SYRS. This relationship was studied to determine the impact of seasonal agrarian drought (SPI−3) on wheat crop yields on a regional scale. The Pearson correlation coefficient values between the SYRS and SPI−3 were calculated to determine the provinces that were severely affected by drought in South Africa (2002-2021).

SPI Trend and Frequency on Regional and Provincial Scales
The output of the SPI analysis reveals that the northern and eastern parts of the country were less prone to drought compared with the southern and western parts ( Figure 4 and Table 4). Both SPI−3 and SPI−6 showed that the majority of the stations experienced a negative MK trend (an increase in drought events) ( Table 4) and shown by inverted red triangles in Figure 4. For SPI−3, only five stations (15.6%) experienced a positive MK trend (decrease in drought events), while eight stations (28.12%) witnessed a significant (p < 0.05) increase in drought events (Table 4) and shown in circular inverted red triangles in Figure 4. For SPI−6, only three stations (9.3%) were less vulnerable to drought (p < 0.05), while the rest (90%) witnessed an increase in the drought trend. In terms of SPI−9, fourteen stations (43%) were more susceptible to drought (p < 0.05), while four stations depicted a significant positive (p < 0.05) MK trend. Like SPI−9, the MK test for SPI−12 revealed that thirteen (40.6%) stations witnessed a significant (p < 0.05) increase in drought events, while only two stations were significantly (p < 0.05) less vulnerable to drought (Table 4).
where D dx denotes values of yield in drought production seasons at the provincial level, and D dt represents values of detrended yield in a similar production season. Table 3 [82] depicts the CR categorization.

Correlation Analysis between Crop Yields and Agricultural Drought Indices
A nonlinear regression model was employed to explore the temporal relationship between drought indices and the SYRS. This relationship was studied to determine the impact of seasonal agrarian drought (SPI-3) on wheat crop yields on a regional scale. The Pearson correlation coefficient values between the SYRS and SPI-3 were calculated to determine the provinces that were severely affected by drought in South Africa (2002-2021).

SPI Trend and Frequency on Regional and Provincial Scales
The output of the SPI analysis reveals that the northern and eastern parts of the country were less prone to drought compared with the southern and western parts ( Figure 4 and Table 4). Both SPI-3 and SPI-6 showed that the majority of the stations experienced a negative MK trend (an increase in drought events) ( Table 4) and shown by inverted red triangles in Figure 4. For SPI-3, only five stations (15.6%) experienced a positive MK trend (decrease in drought events), while eight stations (28.12%) witnessed a significant (p < 0.05) increase in drought events (Table 4) and shown in circular inverted red triangles in Figure 4. For SPI-6, only three stations (9.3%) were less vulnerable to drought (p < 0.05), while the rest (90%) witnessed an increase in the drought trend. In terms of SPI-9, fourteen stations (43%) were more susceptible to drought (p < 0.05), while four stations depicted a significant positive (p < 0.05) MK trend. Like SPI-9, the MK test for SPI-12 revealed that thirteen (40.6%) stations witnessed a significant (p < 0.05) increase in drought events, while only two stations were significantly (p < 0.05) less vulnerable to drought (Table 4).     Three stations, POR, IRE, and JHB BT, exhibited a significant positive (p < 0.05) trend (i.e., less affected by drought) (2000-2020) on the four SPI time scales (SPI-3, 6, 9, and 12) (Figure 4 and Table 4). Interestingly, these four stations were in the northern mountains and received an average monthly rainfall of 43.94-55 mm (Table 1 and Figure 1). In contrast, nine stations, namely, RSB (northern part), SKZ (eastern part), SPB (western part), PTA (southern part), PE (southern part), PTMBG (eastern part), CAL (southern part), CCE (southern part), and CPT (southern part), showed a fixed significant negative (p < 0.05) trend (Figures 1 and 4; Table 4). Overall, drought events historically impacted the western part of the country and dominated in the coastal area. On a regional scale, six provinces exhibited a negative SPI-3 trend (increase in drought) between 2002 and 2021. The highest decrease was recorded in WES-C (western part of the country, p < 0.05), followed by NR-C and NW ( Figure 5). Two provinces experienced a positive SPI-3 value, namely, LPP and GG; however, this trend was not significant ( Figure 5). Looking in depth at the SPI-3 values across all the provinces, most of the provinces experienced at least one event with less than −1.5, categorized as a severe drought event. However, the lowest value (−2.4) was recorded in WES-C ( Figure 5). Figure 6 presents the percentages of drought frequencies categorized from "no drought" to "extreme drought" based on SPI-3 ranges (Table 2) in all provinces of South Africa over a period of 20 years. A total of 1.3% of months had extreme drought events in the FS province, followed by 0.8% of months in the GG province and 0.4 % in NW. Similarly, the highest percentage of moderate to severe drought months was also experienced in the FS province, i.e., 11.3 and 3.4%, followed by 8.8% of months with moderate droughts in the LPP province of the region. The WES-C province of the region experienced 41.6% of months with mild droughts, 5.9% of months with moderate droughts, and only 0.8% of months with extreme drought events over a period of 20 years. Overall, the highest percentage of drought months was experienced by the FS province in the region, which makes it vulnerable to negative impacts. Overall, the highest percentage, i.e., 53.7%, of all droughts (mild to extreme) was experienced by ES-C, followed by 51.2% in NR-C, 50.8% in MP, and 50.4% in LPP. The lowest of all drought percentages, i.e., 46.2%, was experienced in the GG province of the region.

Impact of Drought on Wheat Production (SYRS)
The main idea of implementing the SYRS is to isolate the impact of agricultural development on crop yield due to climate conditions. The output of the SYRS could provide an overview of the direct impact of drought on wheat production in South Africa. In this

Impact of Drought on Wheat Production (SYRS)
The main idea of implementing the SYRS is to isolate the impact of agricultural development on crop yield due to climate conditions. The output of the SYRS could provide an overview of the direct impact of drought on wheat production in South Africa. In this research, SPI-3 was chosen as a representative of agricultural drought, and then the SYRS was calculated on a regional scale (i.e., the nine provinces).
In the NW province, the lowest SYRS value was recorded in 2006, where SYRS SPI-3 = −1.95, indicating the high impact of drought on wheat yield. In 2015, 2016, and 2019, the SYRS SPI-3 values were −1.21 (moderate losses), −0.89 (acceptable losses), and −1.00 (acceptable losses), respectively. However, the impact of drought (SPI-3) in the other years could be neglected (Table 5)  By tracking the impact of drought (SPI-3) on wheat yield across the country, three years can be distinguished based on the values of SYRS SPI-3 (≤−1.5). In this sense, 2003 had a negative impact on wheat production in MP, LPP, and GG provinces. Three provinces, ES-C, KZN, and NW, were affected in 2005, while a recent drought in 2019 had a direct impact on FS, WES-C, and NR-C (Table 5).

Correlation between SYRS and SPI-3 on a Monthly Time Scale
The impact of drought on crop yield varied between the provinces. Table 6 depicts the correlation between SYRS and SPI-3 on a monthly scale in each province across South Africa. For WES-C, the SYRS had a positive correlation for all months. However, the highest correlations were in March (r SPI-3 vs. SYRS = 0.53), July (r SPI-3 vs. SYRS = 0.55), and August (r SPI-3 vs. SYRS = 0.55) ( Table 6). In NR-C, the highest correlation was recorded in the growing cycle (Table 6). Like NR-C, the highest correlation in the FS province was recorded between October (r SPI-3 vs. SYRS = 0.54) and December (r SPI-3 vs. SYRS = 0.54) ( Table 6). For both NS-C and NKZ, there was a low correlation between SPI-3 and SYRS (Table 6). For MP, the highest correlation was recorded in May (r SPI-3 vs. SYRS = 0.41) ( Table 6). The correlation was weak in LPP from January to May and had the lowest value (r SPI-3 vs. SYRS = 0.2, April-May) ( Table 6). In the GG province, the growing cycle reflects a good correlation with SPI-3 in May (r SPI-3 vs. SYRS = 0.35), June (r SPI-3 vs. SYRS = 0.51), and July (r SPI-3 vs. SYRS = 0.53) ( Table 6). The growing cycle in the NW province did not reveal any notable correlation between SYRS values and drought (SPI-3) ( Table 6).

Drought Resilience (CR) of Wheat on a Regional Scale
The CR analysis provides an overview of crop resistance to drought. In this research, the CR was calculated for the nine provinces across South Africa. All results were compared with the threshold CR = 0.8 (Table 3) to indicate whether wheat yield was moderately resilient to drought events or not. The most affected province by drought was WES-C (western part of the country), where the CR value was 0.65, followed by FS with CR = 0.65 (severely nonresilient) (Table 7). Interestingly, the SYRS analysis also revealed that WES-C had a positive correlation with SPI-3 in all months, while FS also had a positive correlation with SPI-3 in the harvesting months of Oct-Nov (Table 6), thus designating these provinces as the most affected by agricultural drought. Regarding the rest of the provinces, the CR values were above the threshold, i.e., ranging from 0.85 to 0.97, revealing a good deal of drought resilience. Table 7 presents an extensive analysis of the yield loss percentage (YL %) in different stages of wheat growth and during different drought events. Consistent with the correlation evaluation, WES-C revealed the highest YL of 35% during the longest SPI-3 drought duration (DD) of 20 months, i.e., from August 2016 to March 2018, in the whole growing cycle (GC) of wheat, with a drought sum (DS) of 13.3. This was followed by a yield loss of 29.5% during another longer DD of 18 months from August 2018 to January 2020 in the GC of wheat with a DS of 14. WES-C also experienced a YL of 29.3% in 3 months of the sowing period (SP) in 2004 and 18.6% in 6 months from the SP to the growing period (GP) and 4 months from the GP to the harvesting period (HP) in 2003 and 2010, respectively.
Like WES-C, another province in the region i.e., FS, also experienced the highest SPI-3 drought-associated YL of 34.5% in 7 months from the growing period to the harvesting period (GP-HP) from July 2019 to Jan 2020. Other significant YLs of 20.4% and 17% in the same (GP-HP) stages of wheat growth were observed in 2010 and 2004, with short DDs of 4 and 5 months. The results also revealed GP to be most closely linked to YL, followed by HP, GC, and SP collectively in both provinces in the region (Table 7).
Similar results are found for other provinces in the region during different growth stages but with a good drought resilience of above 0.8. Another interesting finding is that a minimum of 7% to a maximum of 17.8% YL was found in NW, KZN, NR-C, and ES-C provinces of the region during different wet years when there was no drought identified (Table 7). * Winter wheat growing stages (GSs) include the sowing period (SP), which runs from May to June; the growing period (GP), which runs from June to September; the harvesting period (HP), which runs from September to November; and the entire growing cycle (GC), which runs from May to November. DD is drought duration, DS is drought severity, and gray shading is used to represent the YL (%) during wet events (w) or no drought event occurrences (n).

Current and Future Drought across South Africa
In South Africa, a few studies have examined the drought trend and its intercorrelation with crop production. In this sense, this research was designed to bridge the gap regarding agricultural drought and its impact in South Africa and to highlight the region's most vulnerable provinces to drought. Thus, local governments and policymakers can take action to minimize drought impacts and ensure food security through adaptation and mitigation plans. The output of the SPI analysis for 32 stations covering the whole of South Africa indicates a drought tendency across the country at different drought levels, e.g., SPI-3, SPI-6, SPI-9, and SPI-12 ( Figure 4 and Table 4). Notably, most of the provinces showed a negative trend based on the SPI-3 analysis (Figures 4 and 5). This drought trend could be explained by the El Niño-Southern Oscillation. In this sense, the recent 2016 El Niño indicated that the country is prone to drought trends [83,84]. The Pacific El Niño has a history of causing meteorological variability [85,86]. The El Niño-Southern Oscillation is reportedly causing drought periods in the northern region of the country [87]. In this context, the arid and semiarid climates of the study area, along with fluctuating rainfall, has accelerated the evolution of drought in the country [88]. However, a drought trend has been reported previously across Africa and in South Africa (Table 8).  Future climate projection by GCM models (from the 1960-2000 baseline timeframe) highlighted in the IPCC 4th Assessment Report [100] shows that South African rainfall is predicted to decrease by 2030-2060. The report mentions that the temperature is expected to increase by 1 and 3 degrees Celsius in most inland areas by 2060, while coastal areas will experience lower increases compared with the interior. This indicates that the surface wind direction and speed are predicted to change with high pressures from anticyclones in both the Pacific and Indian Oceans.

Drought Impacts on Wheat and Its Resilience
Since most droughts in Africa occur in the temporal and permanent domains, crop prediction models that are not based on these variable rates may give false positives. Drought has severe agricultural impacts, from the loss of income for farmers due to crop yield and livestock losses [101] to regional food security shortages [102]. Based on the SPI-3 versus SYRS analysis (Tables 5 and 6), drought has had an impact on wheat production, especially in the western parts of the country. Pearson correlations between SYRS and SPI-3 on the monthly scale clearly show the strength of positive and negative relationships between yield loss and drought during stages of wheat growth. A positive correlation between them is found in WES-C during all stages of wheat growth, i.e., from Apr to Nov, over a period of 20 years. Severe droughts in 2003, also studied by Rouault and Richard [103], followed by low to moderate droughts in 2009 and 2010, became a cause of drought-associated yield loss here. Furthermore, after 2015, WES-C faced a continuous long-term drought condition that significantly impacted the whole wheat growing cycle and caused the highest YL of 29-35% in the whole of South Africa (Table 7) [56]. Similarly, LPP, GG, and NW provinces are also impacted by drought during different stages of wheat growth, with positive correlations between SYRS and SPI-3 from June to November, April to July, and January to April, respectively. In contrast, three provinces in the region, i.e., FS, ES-C, and KZN, revealed negative correlations between the SYRS and SPI-3 from Mar to Aug and April to December, respectively. A study by Shew et al. [47] revealed wheat yield loss in the dryland cropping system of FS due to heat stress, which is consistent with our results of significant SPI-3-associated YLs of 10.8 to 34.5% from the sowing to harvesting period (SP-HP) of wheat growth (Table 7).
Drought impacts crop risk profiling and seasonal crop-water requirements, which are fundamental to crop life. [104]. Unfortunately, only 25% of the total cropped area in the region is under irrigation [105]; therefore, the SPI-associated yield loss is quite evident in the results of our study.
A few studies have been carried out in South Africa to highlight the impact of drought on crop production. For instance, Unganai et al. [106] focused on the interaction between drought monitoring and corn yield predictions using remote sensing, with the findings highlighting the need to further investigate the severity of vegetation stress. Since most African countries do not possess long-term instrumental climate records, drought prediction using statistical methods is not prevalent. Senay and Verdin [107] allude to using modern GIS water balance algorithms to forecast seasonal drought to offset the negative impact of drought in agropastoral systems and ensure water allocation.

Strategies for Drought Mitigation in South Africa and Future Steps
The way to combat agricultural drought is to integrate robust sectoral strategies [108]. Adaptation strategies surely depend on the agropastoral practices that farmers are using. For example, cereal farmers may need to grow fall-sown crops and use better cultivars to offset the impact of drought. Other strategies for cereals sown in the winter include cultivars with reduced vernalization periods. Land management practices such as shifting from rainfed agriculture to irrigated agriculture can alleviate water stress in crops during drought seasons [109]. Reducing greenhouse gases by adopting precision agriculture practices will help alleviate carbon escape from most croplands [110]. Crop diversity, stubble residue management, and nutrient recycling procedures may help reduce the impact of drought [111].
Given that agricultural drought results in a persistent deficit in soil moisture content, which is associated with wilting crops [112], it is urgent for farmers and policymakers to adopt measures such as irrigation supply, water demand, and aftermath mitigation measures. The first two measures deal with water shortages, while the last addresses the socio-environmental impact of drought. Earthwork can improve water availability through the abstraction of groundwater and the construction of artificial dams, reservoirs, canals, or rivers. Non-earthwork systems can sound early warning alarms for drought detection and thus enable emergency responses, such as insurance aid, rehabilitation, and recovery measures. These types of measures strengthen institutional structures and their capacity to better prepare for drought [113].
Even though this research was based on input from 32 stations, the distribution of the climate stations covers the whole region and, thus, could be used to reach a robust result. Forthcoming studies will include engaging other climate indices, such as the SPEI and the Palmer drought severity index (PDSI), to identify the drought frequency and intensity based on different inputs. Furthermore, the ecosystem response to drought will be analyzed using satellite images.

Limitations of Study
Our study has some limitations that open new research questions for the future as well. Firstly, the study was limited to analyzing the meteorological drought impact on wheat yield using a single drought index, i.e., the SPI, which is comparable to other indices, such as the SPEI, PDSI, and scPDSI, for incorporating the impacts of temperature and evapotranspiration. A similar type of study for drought assessment in east Africa utilized the SPI and SPEI and presented quite similar results, with some uncertainties in the overand underestimation of drought events [98]. However, the results of multiple indices may vary from region to region and can be utilized in the future to differentiate drought impacts in rainfed and irrigated wheat regions. In the next step, multiple indices, such as the SPEI, PDSI, and composite drought index (CDI) [114], will be used to capture drought events in South Africa and to assess their impacts on the agricultural sector.

Conclusions
Our study examined the temporal interaction of meteorological droughts and their impacts on wheat yield across the nine provinces of South Africa. A widely utilized meteorological drought index, i.e., the SPI, was used to examine the duration, frequency, and trend of meteorological droughts in dryland and irrigated provinces in the region. The SYRS and CR computed from trended and detrended wheat yields across all provinces were revealed to be significantly impacted by meteorological drought over a period of 20 years. The following major conclusions can be drawn from the results of this study:

•
The frequency of drought events revealed that ES-C experienced the highest percentage of drought, i.e., 53.7%, followed by NR-C, MP, and LPP provinces of the region. • SPI-3 trend analysis reveals a significant negative trend across many provinces in the region. Specifically, the western coastal provinces WES-C and NR-C have been more vulnerable to meteorological droughts over the past 20 years. Positive correlation results between the SYRS and wheat yield indicate that the WES-C province was highly influenced by drought during all stages of wheat growth, i.e., Apr-Nov. Historical drought spells in 2003, 2009, and 2010 and a low CR = 0.64 caused the province to be highly impacted by the negative impacts of droughts on yield loss. • Some provinces in the region, including FS, ES-C, and KZN, were not found to be highly impacted by droughts, with negative correlations between the SYRS and SPI-3 during the wheat growth cycle from Apr to Nov.

•
The WES-C and FS provinces of the region experienced the highest yield loss % during the SP-GP-HP of wheat growth stages with a CR of 0.65, indicating extremely low resilience. Overall, the growing period of wheat was found to be the most associated with yield loss, followed by the harvesting and sowing periods. Yield loss in the WES-C province is linked to the whole growing cycle in all months of wheat growth (Apr-Nov).
However, the current study efficiently examined the meteorological drought variations and related wheat yield losses in both rainfed and irrigated dryland provinces of South Africa, but still, the results of the study can be further enhanced by incorporating other indices, such as the SPEI and PDSI. Other than these, the incorporation of remote-sensingbased indices such as NDVI can also be utilized to examine drought-associated stress and yield loss. This research can be helpful in robust yield prediction modeling associated with climatic changes in the region.
Other than this, the major findings of the study also suggest adapting and focusing on drought-resistant agricultural practices in the western coastal parts of South Africa to prevent future yield loss risk. Our study recommends an immediate climate adaptation and mitigation plan to support farmers and stakeholders in combating climate change. A regional plan for climate awareness should also be formulated to protect the country's food security.