The Economic Research Services of the United States Department of Agriculture (USDA) listed Lesotho amongst the countries where approximately 100% of the population is projected to remain food insecure until 2022 (United States Department of Agriculture 2012). This food insecurity phenomenon has resulted in 80% of the population of this country being hungry because their livelihood means depend on rain-fed agriculture. Given this kind of livelihood dependence, the key threats to development in this country are chronic poverty and a high unemployment rate, to mention but a few, exacerbated by a decrease in remittances from migrant labourers in South Africa because of the closure of mines and climate-related shocks such as frequent droughts (United Nations World Food Programme (WFP) 2015). However, agriculture, which is the main source of livelihood, is hampered by the following: lack of arable land because of increased urbanisation, soil erosion and environmental degradation, an unfavourable climate that exacerbates and increases drought frequency and over-reliance on rain-fed subsistence agriculture (Assessment Capacities Project (ACAP) 2013).

Lesotho is known for its unique and rich cultural practices, such as the initiation of schooling [Lebollo] for both boys and girls. This practice, which is intended to initiate boys and girls into manhood and womanhood, respectively, lasts approximately for 6 months where the initiates are constantly engaged in collecting large amount of firewood (Matobo, Makatsa & Obioha 2009). In addition to this practice, the bulk of domestic energy consumption in Lesotho is based on using wood as fuel. Although there is very limited tree cover in Lesotho, the majority of people rely on wood as a source of energy (United Nations Environment Programme (UNEP) 2013). Limited tree cover and cutting of trees by households for fuel purpose expose the land to an increasing number of drought events. Land-use management is the direct responsibility of local chiefs at village level, some of whom are illiterate. There are also poor land policies and enforcement of management measures, which lead to land degradation, hence increased vulnerability to drought impacts (Maro 2011).

Given that 80% of population’s livelihood depends on rain-fed agriculture, Lesotho is currently facing the following problems: (1) high unemployment and chronic poverty levels (WFP 2015); (2) lack of arable land because of increased urbanisation, soil erosion and environmental degradation (ACAP 2013); (3) a high deforestation rate as a result of cultural practices [Lebollo] (Matobo et al. 2009); (4) wood as the main source of fuel (UNEP 2013); (5) illiterate local chiefs charged with direct responsibility of land-use management (Maro 2011); (6) poor policies and management measures against droughts (Maro 2011); and (7) no drought risk reduction-specific framework to manage drought disaster risks given that 80% of the population depends on rain-fed agriculture. Drought disaster issues are, therefore, addressed haphazardly, with no guiding principles, leading to poor planning and hence unpreparedness that leads to a reactive disaster management approach. This situation adversely does not only affect peoples’ livelihood but also the agribusinesses.

Drought is the most complex of all natural disasters with damaging and severe impact on agricultural production, ecosystems, water resources and society. This complexity makes the detection of its start and end periods difficult (Tan, Yang & Li 2015). This is directly linked with precipitation, the lack of which affects various economic sectors, leading to substantial costs for communities (Haensel, Matschullat & Schucknecht 2015). However, the term ‘drought’ has a larger number of definitions drawn from different perspectives and perceptions with regard to the purpose that is defined (Gregor 2013). This natural phenomenon has various impacts, ranging from direct to indirect, and from different dimensions, as shown in Table 1. Historically, drought has globally caused both direct and indirect economic, social and environmental problems, some of which are inevitable even with early preparations (UCAR 2015). Similarly, Anon (2015) asserts that this natural hazard produces a complex web of impacts that span over several sectors of the economy. This complexity is because of the lack of water, which is integral to communities in producing their own food and services. Despite the social, economic and environmental adverse impacts, droughts pose difficulty in decision-making with regard to water allocation and result in stringent water-use limitations (National Drought Policy Commission 2015). Hazards related to climate and weather, such as drought, affect more people and have larger economic damage worldwide than any other type of hazards; these hazards have killed or affected 70 times as many people and caused twice as much damage worldwide as did any other hazard types (Arnold & Kreimer 2000).

One of the biggest weather hazards, which many people hardly realise, is heat that kills many people silently (NOAA 2015). Globally, the number of weather-related natural disasters has more than tripled since the 1960s. These disaster events claim over 60 000 lives annually in developing countries (WHO 2015). Extreme high air temperatures directly cause cardiovascular and respiratory diseases, especially amongst the elderly and the young: groups of over 65 and younger than 5 years, respectively. A number of people feel comfortable with temperature in the ranges of 20 °C – 27 °C, and relative humidity of 34% – 60%; however, as these temperatures go higher, the body’s coping mechanism becomes overwhelmed, leading to various and possibly fatal conditions (Canadian Center for Occupational Health and Safety 2015). Some of the disorders caused by heat are sunburn, heat cramps, heat exhaustion and heat stroke (NOAA 2015). Drought ranks highest amongst all natural disasters globally in terms of economic impacts and causes 78% of other natural disasters (DIMTEC 2015; Geerts & Linacre 2015).

Finally, drought and its impacts are really two sides of the same coin. We cannot fully understand drought without understanding its impacts, which can affect all parts of our environment and our communities. Understanding drought conditions, societal vulnerability and their related effects on one another provides us with historical lessons that can aid in dealing with future drought conditions (NOAA 2015).

The following countries shown in Table 2 were identified by the World Bank as the most at risk of drought in 2009. Therefore, this study reviewed the impacts from economic, environmental and social spheres.

Disasters reverse development progress by years, sometimes even by decades, and leave the affected countries with monumental debts for projects that were destroyed. Therefore, development organisations responsible for recovery projects must consider an effort to increase local resilience (Coppola 2011). Table 3 presents the impacts of drought in the greater Horn of African countries that are at risk of drought hazard.

Once the impacts are known, it is the responsibility of every country to implement preventative measures. Lessons learnt from those who have experienced such disasters are needed to guide measures for implementation.

This study followed a quantitative method research design where a researcher relies on numeric data in testing relationships between variables (Creswell et al. 2007). They assert that the researcher also relates the variables to determine the magnitude and frequency of relationships. Moreover, the ultimate goal of quantitative research is to describe the trends, if any exist, or explain relationships between or amongst variables. This study used the positivist research paradigm.

Data for this study were collected from at least one weather station from each of the 10 administrative districts of Lesotho or a nearby station. Data collection was mainly focused on monthly precipitation and temperature data from the past 30 years (1985–2015) over all the identified weather stations, with complete available data in Lesotho. The Lesotho Meteorological Services and Water Affairs are the responsible government departments for keeping the meteorological and stream flow data in Lesotho, respectively. Therefore, the two sets of data (precipitation and temperature) were collected from the Lesotho Meteorological Services and Water Affairs. Figure 1 shows the map of Lesotho exhibiting the 10 districts that were used as selected stations from which data were collected for this study.

All stations from which data were completely missing were exempted from participation in the study. However, in the stations where at least 10% of data were missing, gaps were filled by the methods discussed below. In literature, there exist several techniques for estimating missing data. Two of these techniques are discussed and the discussion ends with the one used in the study.

Expectation Maximum (EM) is defined as a statistical algorithm suitable when there are missing or hidden values in the data sets (Hauskrecht 2017). Borman (2006) adds that EM is a popular tool used in statistical estimation problems that involve incomplete data. Similarly, Chuong and Serafim (2008) refer to EM as an algorithm that enables parameter estimation in probabilistic models with incomplete data. Prior to missing values estimation, data collected from the Lesotho Meteorological Services and Water Affairs were compared for quality, reliability and validity. The EM was, therefore, used in the study. Both monthly values of temperature and precipitation were entered into an IBM Statistical Package for the Social Sciences (SPSS) v. 24 where the EM algorithm was applied to estimate all missing values in the data sets. The three complete data sets were then subjected to outliers detection aided by the IBM SPSS v24 program. Prior to any climatological data analysis, data sets must be tested for homogeneity.

The selected indices, SPI and SPEI, were calculated from DrinC computer software. The SPEI that uses precipitation and potential evapotranspiration, was computed from the drought calculator (DrinC) using Hargreaves methods, which use both minimum and maximum temperature values. DrinC is an open access software developed for calculating drought indices such as SPI. Two indices, SPI and SPEI, were then computed to detect drought conditions from only a 3-month time scale. Standardised precipitation-3 was used because it measures agricultural drought on seasonal basis. The dry spell parameters drawn from duration, intensity and frequency were calculated from a customised computer program. These parameters included the number of dry spell events (N) and average dry spell duration, average dry spell intensity (ADSI) and the ratio of number of dry spell events (%) during the 30-year study period (N/30) for frequency. The four dry spell parameters drawn from duration, intensity, frequency and station elevation were in the k-means clustering of stations on temporal scales. This resulted in regions ranked for easy management, planning and prioritisation. Standardised precipitation intensity maps were then generated and displayed through a StatPlanet computer program.

Prior to the above analysis, homogenisation was employed. Homogenisation has to be undertaken prior to any data analysis as part of data quality control measures in order to eliminate any erroneous and non-climatic biases in the time series (Štěpánek et al. 2013). Most climatological time series suffer from in-homogeneities because of (1) changes in instrument settings, (2) changes in observers, (3) changes in formula calculations, (4) changes in observer practices and (5) station relocations. Homogeneity is an important issue in climate change data analysis to detect variability in the data series. This generally means that when data are homogeneous, they were taken at the same time with the same instruments and in the same environments. In this way, homogenisation of time series data ensures reliability of the results (Franz, Steffan-Dewenter & Menzel 2009). Before any analysis begins, exploratory data analysis must be conducted. Its advantage is to provide a preliminary indication of trends present in the data that enable further analysis (Meals et al. 2011). The authors further emphasise that this technique requires fairly long series of data, collected by consistent methods with few gaps. Most trends and analysis procedures require certain assumptions, such as the data must follow a certain probability distribution (Xie et al. 2016). A non-parametric homogeneity test was, therefore, used in monthly precipitation and minimum and maximum temperature.

Rainfall is the most critical and key variable for both hydrological and atmospheric circles, the lack of which leads to extremes such as droughts. The awareness of the characteristics of dry spells over an area, such as source, intensity, duration, variability, distribution and frequency, is essential for proper and efficient control and management of water resources (Takele & Gebretsidik 2015). On SPI-3, Mohale’s Hoek was the only station that showed a statistically significant decreasing trend. Standardised precipitation detected dry spells much earlier than SPEI over all stations, showing a higher sensitivity than SPEI. This situation leads to SPI being the most suitable index for dry spells analysis in the study area. All analyses that followed used SPI only because of its sensitivity to pick up dry spells earlier than a water balance index (SPEI). All stations showed a high and long frequency and duration, respectively. Moreover, the entire study area was in extreme drought during the study period. All stations were under extreme drought, which indicates that the situation of drought condition is yet to stay constant over years to come. This implies that farmers must be encouraged to grow drought-resistant cultivars to keep agribusiness in the market. Rangeland policies and legislation must be enforced for livestock production, especially in the periods when drought events are expected. It is, therefore, recommended that the agricultural sector should remain vigilant at all times as drought episodes may strike at any given period.