Assessment of Trends and Spatial Patterns of Agricultural Drought Incidence in Mpumalanga Province, South Africa

1. Introduction

Drought, a pressing global issue, significantly impacts food production, water availability, and economic stability. Despite the lack of a universally accepted definition, it is most clearly understood as a period of arid weather caused by insufficient precipitation, leading to a significant imbalance in water supply and causing moisture deficiency that affects human activity and environmental processes [1]. Changes in precipitation, river flows, groundwater levels, and soil moisture often lead to the interpretation of this natural hazard as a shift in weather and climate [2]. It can be classified as meteorological, hydrological, agricultural, and socioeconomic, each type with distinct impacts and durations [3,4,5,6].

Apurv et al. [7] highlighted that multiple droughts are driven by a combination of reduced precipitation and higher evaporation rates, affecting the entire water cycle. Meteorological drought happens when there is an extended period of minimal to no rainfall in a region [8,9]. Agricultural drought occurs when soil moisture decreases due to increased atmospheric dryness, which results from extended precipitation deficits and rising temperatures [10]. According to Bhardwaj et al. [11], hydrological drought happens when underground water levels decline due to high demand for water that is not adequately supplied. Additionally, socio-economic drought occurs when there is a water shortage for people [12]. A recent paper by

Crausbay et al. [13] argues that ecological drought should be classified as the fifth type of drought. It happens when there is a shortage of water for ecosystems, and this severe lack of water affects the ecosystems’ services and causes harm to both nature and humans. Additionally, Otkin et al. [14] highlight another type of drought called flash drought, characterized by a short-term, severe drought with a rapid onset and involving specific processes.

Droughts are natural phenomena that occur slowly and in complex ways. Droughts have become more prevalent and severe in recent years, primarily due to ongoing climate change caused by humans and natural phenomena like El Niño [15,16,17,18]. The impacts of drought differ significantly across regions, with distinct spatial and temporal patterns. For example, in Pakistan from 1999 to 2002, drought affected 3.3 million people, primarily in Pakistan’s Sindh and Baluchistan provinces, and caused many deaths from starvation [19]. In Namibia in 2019, drought affected more than 500,000 people, and more than 60,000 animals died from drought [20]. The situation is similar in Zimbabwe; the 2018-2019 drought resulted in low crop harvests and livestock losses [21].

In South Africa, a recent study conducted by Diko [22] found that in 2016/17, maize production in Free State Province declined from 7.3 million tons to 5.2 million tons in 2017/18 and further to 4 million tons in 2018/19. Similar declines were observed in the Mpumalanga Province, from 3.4 million tons to 2.8 million tons to 2.6 million tons, and in Northwest Province, from 3.1 million tons to 2 million tons and then to 1.6 million tons due to drought. In Mpakeni, Mpumalanga Province, farmers mention that they had to sell their livestock at low prices because they could not afford to feed them due to the drought that hit them in 2015 and 2017 [23]. Furthermore, drought does not affect food security and the agricultural sector only; it also hinders us from achieving the 2030 Sustainable Development Goals (SDGs), such as SDG 1 (No poverty), SDG 2 (Zero hunger), and SDG 3 (Good health and well-being) [24]. Therefore, these highlight the crucial need to monitor agricultural drought for water management and drought mitigation strategies.

Preventative measures and monitoring can minimize agricultural drought impacts if detected early. Several indices are employed to monitor agricultural drought, including the Standardized Soil Moisture Index (SSMI) [25], Crop Moisture Index (CMI) [26], and Vegetation Condition Index (VCI) [27]. However, due to data limitations in the region, this study used a meteorological drought index, which is the Standardized Precipitation Evapotranspiration Index (SPEI), to assess agricultural drought. The SPEI was proposed to improve the Standardized Precipitation index (SPI) by Vicente-Serrano et al. [28]. It’s a multi-scalar index that offers a flexible way to characterise the combined effects of precipitation and evaporative demand on droughts [29,30].

Numerous studies have used the SPEI to monitor agricultural drought because it incorporates evapotranspiration (ET), an essential component of the agricultural water budget, making it a suitable index for assessing agricultural drought [31]. The effectiveness of SPEI in identifying agricultural drought was confirmed by studies conducted in Pakistan [32], Argentina [33], the South-Central United States [34], and Southwest China [35]. Since the SPEI can be calculated over time scales ranging from 1 to 48 months, it is used to assess different types of droughts, especially in regions with limited data, such as Mpumalanga Province. The SPEI follows the SPI guidelines, indicating that agricultural drought can be detected at 6-month timescales (SPEI6) [36]. Additionally, several scholars, Parsons et al. [37], and Tian et al. [38], demonstrated that the 6-month timescale (SPEI6) is particularly effective for evaluating agricultural drought.

Due to the reasons that are provided above, this study computed the SPEI at a 6-month timescale (SPEI6) to evaluate agricultural drought from 1981 to 2025. This study is of paramount importance as it provides the occurrence, frequency, intensity, trends, and spatial patterns of agricultural drought in Mpumalanga Province. It offers crucial insights into how agricultural droughts develop and evolve over time, providing valuable information for more effective water management and drought prevention.

2. Materials and Methods

2.1. Study Area

This study was conducted in Mpumalanga Province, which is in northeastern South Africa (latitude −29.8167; longitude 30.6167). It covers an area of 76,495 square kilometres and has an average elevation of 494 metres above sea level [39]. It is divided into three districts: Nkangala, Ehlanzeni, and Gert Sibande. These 3 districts are subdivided into 17 local municipalities. The Mpumalanga Province receives high rainfall between October and March, making it a summer rainfall region. The rainfall pattern is influenced by topography, where the highveld receives the most annual rainfall of up to 1000 mm, while the lowveld receives the lowest annual rainfall of 600 mm [40]. According to recent scientific literature, there is a lack of studies that provide a comprehensive understanding of agricultural drought in this region, even though the region is prone to drought, which is why it was chosen.

Figure 1. Geographical location of the selected study area and ERA5 Grids Points, (Source: Author, 2026).

Figure 1. Geographical location of the selected study area and ERA5 Grids Points, (Source: Author, 2026).

2.2. Methodology

This study used the ERA5-Land dataset from the European Centre for Medium-Range Weather Forecasts (ECMWF), which provides hourly climate data from 1950 to the present. The data was collected and processed using Google Earth Engine (Accessed on March 02, 2026). 14 grid points were selected to cover the entire province of Mpumalanga, and we named them G1 to G14. This study focused on the period 1981–2025 (45 years), using monthly precipitation (mm) and maximum and minimum temperatures (°C) derived from hourly data.

Standardized Precipitation Evapotranspiration Index (SPEI) which is widely recognized for assessing agricultural drought by calculating the difference between precipitation (P) and potential evapotranspiration (PET) was computed at a 6-month timescale which is referred to as SPEI6 to identify the agricultural drought occurrence, frequency, and trends over the past 45 years. The SPEI, proposed by Vicente-Serrano et al. [28], is a significant advancement over the Standardized Precipitation Index (SPI). Unlike the SPI, the SPEI does not consider precipitation only as input data; it also incorporates temperature data, making it a more comprehensive and advanced tool for drought analysis. This choice of index ensures that our study is informed by the latest developments in the field.

The SPEI was computed using “SPEI” package in RStudio. The first step is to calculate the Potential Evapotranspiration (PET), which is calculated using different equations such as Penman-Monteith (PM), Thornthwaite (Th), and Hargreaves (HG), and each of these equations requires different data [41]. However, in this study, the Hargreaves equation was chosen due to its simplicity and accuracy in estimating PET, especially in regions with limited meteorological data. This was followed by calculating the Climate Water Balance (CWB), which is the difference between the Precipitation (P) and Potential Evapotranspiration (PET). It was calculated as follows:

CWB = P − PET

CWB represents the climate water balance, P represents the monthly precipitation, and PET represents the monthly potential evapotranspiration. A log-logistic distribution is then applied to fit the CWB values and transform them into a standard normal distribution, which are the SPEI values. The SPEI values were classified according to [42] into seven classes: extremely wet (>2), severely wet (1.5 to 1.99), moderately wet (1.0 to 1.49), near normal (-0.99 to 0.99), moderate drought (-1.0 to -1.49), severe drought (-1.5 to -1.99), and extreme drought (< -2), indicating that wet conditions start when SPEI values reach positive 1.00 and above and drought conditions start when SPEI values reach -1.00 and below.

The Inverse Distance Weighted (IDW) interpolation method in QGIS (Version: 3.32.2) was utilized to map the spatial extent of agricultural drought based on point data. IDW is a deterministic interpolation method used to satisfy specified statistical assumptions. It assumes that an unmeasured point’s value is equal to the weighted average of the nearby known values [43,44]. To identify the trend of the agricultural drought incidence, we used the Mann-Kendall test, which is a non-parametric test. It is used to determine if a time series dataset has a monotonic trend, either rising or dropping. According to Muhammad et al. [45], the Mann–KendallMann-Kendall test is a statistical technique used to examine hydroclimatic series’ temporal trends and spatial variation. It was performed using the “Kendall” package in RStudio.

3. Results

3.1. Historical Agricultural Drought Events

Historical agricultural drought and wet conditions observed by the SPEI at a 6-month timescale within 14 grid points in Mpumalanga Province, covering the period from 1981 to 2025, are shown in Figure 2. The graph uses colour to differentiate between wet and drought events; the blue bars represent wet events, while the red bars indicate drought events. A more significant downward movement of the red bar indicates more severe drought conditions, whereas an upward movement of the blue bar shows a more intense wet period.

The duration of the agricultural drought was determined by identifying when the negative SPEI values began, marking this point as the onset date. The end date was determined by identifying the month when the SPEI values returned to positive values, marking the end of the duration. The frequency of agricultural drought was calculated by counting the total number of drought events over 45 years. To classify the severity of each drought event, we analyzed SPEI values during each event. An event with SPEI values reaching -1.00 to -1.49 was classified as a moderate drought, while values between -1.5 and -1.99 indicated a severe drought, and less than < -2 indicate an extreme drought event, and so on.

The events observed differ significantly in intensity, ranging from extremely wet conditions to extreme drought. The G1 grid point recorded 18 drought events, including 8 moderate, 6 severe, and 4 extreme events. However, areas around this grid point experienced their worst drought, classified as extreme events, from September 1991 to November 1992 and from June 2015 to January 2017. In contrast, the G2 experienced 14 drought events (4 moderate, 6 severe, and 4 extreme). Its worst drought occurred from September 2014 to January 2017 and again from August 2018 to January 2020. On the other hand, the G3 observed 18 drought events, including 7 moderate, 9 severe, and 2 extreme events, with the worst occurring from July 1990 to November 1992 and from October 2015 to December 2016.

In the G4 grid point, a total of 21 drought events were recorded (12 moderate, 5 severe, and 4 extreme). The worst drought event emerged from December 1991 to February 1993 and from January 2015 to January 2017. Similarly, G5 grid point recorded 18 drought events from 1981 to 2025, including 11 moderate, 4 severe, and 3 extreme events. Its worst drought occurred from December 1991 to April 1993 and from June 2014 to January 2017. However, the G6 grid point recorded 15 events, including 3 moderate, 8 severe, and 4 extreme events, indicating that the area surrounding this point is more prone to severe drought events, and its worst drought events occurred from September 2014 to April 2017. Additionally, the G7 noted a total of 20 drought events, including 7 moderate, 10 severe, and 3 extreme drought events, with the worst drought events occurring from May 2015 to June 2016 and from October 2023 to January 2025.

A total of 20 drought events were recorded from the G8, including 8 moderate, 9 severe, and 3 extreme events, with the worst drought occurring from October 2015 to July 2018. In contrast, the G9 observed 16 events (6 moderate, 4 severe, and 6 extreme); its worst drought emerged from May 2015 to June 2018. In contrast, the G10 grid point noted 15 drought events (4 moderate, 7 severe, and 4 extreme), with the worst occurring from September 2018 to June 2020. Overall, the G11 grid point recorded 17 drought events during the study period, consisting of 8 moderate, 4 severe, and 5 extreme events. Its worst drought emerged from September 2018 to May 2021.

The G12 noted a total of 16 events, consisting of 10 moderate, 2 severe, and 4 extreme drought events. Its worst drought events occurred from September 2014 to June 2016 and from September 2018 to August 2020. However, the G13 observed 18 drought events, including 9 moderate, 6 severe, and 3 extreme events. It recorded two of the worst drought events occurring from September 2014 to May 2026 and from September 2018 to March 2020. Furthermore, the G14 grid point recorded 18 drought events, including 8 moderate, 8 severe, and 2 extreme events, with the worst events occurring from September 2018 to December 2019 and from November 2023 to January 2025.

3.2. Spatial Distribution of Agricultural Drought

To show which areas were affected by different drought severities (moderate, severe, and extreme) throughout the study period, we counted the months with moderate, severe, or extreme drought, rather than using drought events, because a drought event can span multiple severity levels. The obtained results were pre-processed using the Inverse Distance Weighted (IDW) interpolation method in QGIS to create a distribution map, and the results are shown in Figure 3.

Figure 3 shows the spatial distribution of monthly agricultural drought occurrence classified into (A) moderate drought, (B) severe drought, (C) extreme drought, and (D) combined drought severities. The results show the spatial variability of agricultural drought conditions across Mpumalanga Province. However, a high frequency of moderate intensity occurred in central and western areas and in a few areas of the southern part of the province, ranging from 59 to 66 months, with a decline in the frequency of moderate droughts towards the northeast (Figure 3A). In contrast, Figure 3B shows that the western side and a few areas in the central and southern parts were most affected by severe droughts, with frequencies ranging from 32 to 38 months, while a few frequent severe droughts were noted in the southern and northeastern parts of the region. On the other hand, a high frequency of extreme drought events was observed in the northeast, ranging from 14 to 17 months, with a decline in their frequency towards the west (Figure 3C).

Furthermore, Figure 3D, which highlights combined drought severities, indicates that most droughts with different intensities occurred in the central area and a few areas on the western side, while a few frequencies were observed on the northern side of the region (Figure 3D). Additionally, the results show that each part of Mpumalanga Province experienced different types of drought severity, but their frequency varies significantly across the region. These spatial patterns show which areas were prone to which levels of drought over the past 45 years.

3.3. Trends of Agricultural Drought Incidence

The Mann-Kendall trend test was used to identify significant increasing or decreasing monotonic trends in the SPEI values from 1981 to 2025 at 14 grid points across Mpumalanga Province. The results of this trend analysis are shown in Table 1. Two statistics from the Mann-Kendall test (Kendall’s tau and the p-value) were used to detect a trend in the data. The Kendall tau statistic was used to identify whether the trend is increasing or decreasing. In contrast, the p-value was used to show whether the trend is statistically significant (p < 0.05) or insignificant (p > 0.05). The results were performed at a 95% confidence interval.

The results reveal that out of the 14 grid points, only one grid point, which is the G14, showed a statistically insignificant trend (p > 0.05), whereas 93% of the grid points exhibit a statistically significant decreasing trend of the SPEI values, indicating that more agricultural drought events are expected in this region soon. This decrease highlighted the need for agricultural planning and water resource management as more drought events are expected in these areas. The potential impact of these findings on agricultural planning cannot be overstated, as they provide crucial insights into future resource management strategies.

4. Discussion and Conclusions

Agricultural drought is a significant threat to food security in Mpumalanga Province, as noted by recent papers on the vulnerability of subsistence farmers and drought risk in South African agriculture [23,24], yet studies providing a comprehensive understanding of this natural phenomenon are lacking. This study aims to fill this gap by assessing the occurrence, frequency, trends, and spatial patterns of agricultural droughts from 1981 to 2025.

The results reveal that this region is prone to various agricultural drought intensity events, providing a significant contribution to understanding this crucial issue. Chivangulula et al. [46] and Chikoore and Jury [47] noted that droughts occur more frequently in southern Africa; this aligns with the results observed in this study, as most grid points recorded high drought events. For example, the highest total number of drought events was recorded in G4 (21 events), followed by G7 (20 events), G8 (20 events), and G1, G3, G5, G13, and G14 (18 events each). In contrast, the fewest drought events were recorded at G11 (17 events), G12 (16 events), G9 (16 events), G6 (15 events), G10 (15 events), and G2 (14 events). These results show that areas in Mpumalanga Province are prone to varying levels of drought intensity. These findings are supported by previous research demonstrating that drought risk varies significantly across municipalities and provinces [47].

In support of the study conducted by Moeletsi and Walker [48], which highlights that droughts are associated with significant variations in both intra-seasonal and interannual rainfall throughout the majority of southern Africa, it was noted in the results that agricultural drought started to be more frequent after the year 2000, when the wet event conditions started to decrease. This indicate that changing in rainfall patterns and increase in temperature variability is making agriculture drought to be worse in the region. Furthermore, recent studies have indicated that South Africa experienced its worst drought since the 1980s and 1990s during the 2015/2016 period [49,50,51]. However, this research reveals that not all areas of Mpumalanga Province experienced their worst drought during the 2015/2016 period, but most did.

For example, the G1 experienced its worst drought from September 1991 to November 1992 and from June 2015 to January 2017, while G3 highlighted that its worst drought occurred from July 1990 to November 1992 and from October 2015 to December 2016. In contrast, G4 noted its worst drought event from December 1991 to February 1993 and from January 2015 to January 2017. However, the G5 experienced its worst drought from December 1991 to April 1993 and from June 2014 to January 2017. The G6 experienced its worst drought events from September 2014 to April 2017. On the other hand, the G7 reported the worst drought events from May 2015 to June 2016 and from October 2023 to January 2025. Overall, the G8 faced its worst from October 2015 to July 2018.

The G9 experienced its worst drought from May 2015 to June 2018. On the other hand, the G10 grid point experienced its worst drought from September 2018 to June 2020. In the G11, the worst drought was faced from September 2018 to May 2021. Similarly, the G12 point noted their worst drought events from September 2014 to June 2016 and from September 2018 to August 2020. However, the G13 recorded the worst drought events from September 2014 to May 2026 and from September 2018 to March 2020. In addition, the G14 grid point observed its worst drought events occurring from September 2018 to December 2019 and from November 2023 to January 2025.

Additionally, the G2 grid point experienced its worst drought events from September 2014 to January 2017 and again from August 2018 to January 2020. The 2014-2017 drought event was supported by a study by Ebhuoma et al. [23] in Mpakeni, an area near the G2 grid point, where farmers reported selling their livestock at lower prices due to the drought that affected them during this period. These findings highlight the necessity for monitoring and management strategies by showing the variation in drought intensity across Mpumalanga Province.

Spatial distribution maps of agricultural drought show that a high frequency of moderate intensity occurred in central and western areas (Figure 3A). In contrast, the western side and a few areas in the central and southern parts were most affected by severe droughts (Figure 3B). A high frequency of extreme drought events was observed in the northeast (Figure 3C). The combined drought severity map highlighted in Figure 3D indicates that most droughts, with varying intensities, occurred in the central area and a few areas on the western side. These spatial patterns show that drought risk varies across the region, influenced by topography and rainfall distribution. Furthermore, Nembilwi et al. [52] reported that wetter areas are frequently more affected by drought than drier areas, which aligns with the results of this study, which reveal that the highveld, which receives high rainfall, experienced more drought frequency than the lowveld, which receives less rainfall. Additionally, spatial distribution maps of agricultural drought can help farmers and policymakers better plan and allocate resources more effectively by identifying the areas most vulnerable to varying drought intensities.

The Mann-Kendall test results reveal that only one grid point shows a statistically insignificant decreasing trend, while 93% of the grid points exhibit a statistically significant decreasing trend in SPEI values, indicating an increase in agricultural drought soon. These results correspond with those of a study by Letswamotse et al. [53], which indicates that agricultural drought is expected to increase in Mpumalanga Province. Furthermore, these results confirmed that agricultural drought is indeed prevalent in this area and underscore the urgent need for proactive measures to minimize its impact, as more drought events are expected at 93% of grid points. These can help farmers and policymakers to make informed decisions.

This research also faced a few limitations. Due to data limitations, the SPEI model, which considers temperature and precipitation data, was used in this study. However, although the SPEI model does provide accurate results, but it overlooks important factors such as vegetation health, soil moisture, and land surface characteristics, which are necessary for a comprehensive understanding of the impacts of agricultural drought. To address these limitations, future studies must integrate the SPEI model with satellite data, such as land surface temperature, vegetation indices, and soil moisture data. This integrated approach would improve the accuracy of agricultural drought assessment and give a more comprehensive picture of agricultural drought conditions. Therefore, it would help close data gaps and improve the accuracy of agricultural drought monitoring and forecasts.