Meteorological drought assessment in north east highlands of Ethiopia Yimer

Purpose – The purpose of this paper is to investigate the patterns and trends of drought incidence in north east highlands of Ethiopia usingmonthly rainfall record for the period 1984-2014. Design/methodology/approach – Standard precipitation index and Mann – Kendal test were used to analyze drought incident and trends of drought occurrences, respectively. The spatial extent of droughts in the study area has been interpolated by inverse distance weighted method using the spatial analyst tool of ArcGIS. Findings – Most of the studied stations experienced drought episodes in 1984, 1987/1988, 1992/1993, 1999, 2003/2004 and 2007/2008 which were among the worst drought years in the history of Ethiopia. The year 1984 was the most drastic and distinct-wide extreme drought episode in all studied stations. TheMann–Kendal test shows an increasing tendencies of drought at three-month (spring) timescale at all stations though significant (p< 0.05) only at Mekaneselam and decreasing tendencies at three-month (summer) and 12-month timescales at all stations. The frequency of total drought was the highest in central and north parts of the region in all study seasons. Originality/value – This detail drought characterization can be used as bench mark to take comprehensive drought management measures such as early warning system, preparation and contingency planning, climate change adaptation programs.


Introduction
Drought is a recurrent climate phenomenon which occurs in most parts of the world, with varying frequency, severity and duration (Wilhite, 1993;Shatanawi et al., 2013).It is difficult to determine the onset and ending time of a drought.It develops slowly, and its impact may The analysis and forecasting of extreme climatic events has become increasingly relevant to make planning effective.Although drought is becoming the most common and damaging natural hazard in Ethiopia, there were no as such detail studies conducted about its magnitude, frequency and spatial extent in different regions of Ethiopia (Degefu and Bewket, 2013).The same is true in this study area.Hence, to reduce the damages from drought, it is crucial to characterize drought.Drought characterization at regional and local scales has significant implications for drought management such as early warning system (farmers could be warned before the advent of drought as to when, what and where to cultivate and when to sell their animals as well as how to conserve water resources and food), preparation and contingency planning (the government is working toward reorganizing its resources before the impact of drought is felt, climate change adaptation programs (introduction of offfarm activities and drought resistant crop and animal varieties).
Different indices have been developed through time to quantify the magnitude of meteorological drought.Most of these indices are based on direct observed measurement of climatic variables such as rainfall, evapotranspiration and temperature (Steinemann et al., 2005).Some of the widely used indices include the Palmer Drought Severity Index (PDSI) (Palmer, 1965), the Percent of Normal, and the Deciles approach (Gibbs and Maher, 1967), the Standard Precipitation Index (SPI) (McKee et al., 1993) and Standard Precipitation Evapotranspiration Index (SPEI) (Vicente-Serrano et al., 2010).
Each index has its own strength and weakness.For example, PDSI is considered as useful in monitoring drought at a regional scale and allows comparisons over relatively large zones (Steinemann et al., 2005).It also uses water balance models to consider the effects of temperature and rainfall.However, it does not work well in mountainous areas where there are extremes in rainfall or runoff variables (Mpelasoka et al., 2008;Ellis et al., 2010).(Degefu and Bewket, 2013;Morid et al., 2006).A brief description of these and other indices was given in Morid et al. (2006).The choice of indices for drought monitoring should be based on the quantity and quality of available climate data, purpose of the study, computational simplicity and its ability to consistently detect spatial and temporal variations of a drought event (Morid et al., 2006).Keyantash and Dracup (2002) also indicated that drought indices must be statistically robust and easily calculated, and have a clear and comprehensible calculation procedure.
In this study, SPI, which is currently used widely for its multiple advantages, has been used.The SPI was proposed by McKee et al. (1993) and has been used frequently during the past two decades (Hirschi et al., 2011;Vicente-Serrano et al., 2015).The robustness of SPI over the other drought indices has been reported in many studies (Vicente-Serrano et al., 2015;Viste et al., 2012;WMO, 2012;Hirschi et al., 2011;Hayes et al., 1999;Guttman, 1998;Mpelasoka et al., 2008).In addition to its simplicity and ease of calculation, in regions like Ethiopia, where the access to data is limited, there are good reasons for choosing a rainfallbased drought measure (SPI) (El Kenawy et al., 2016;Degefu and Bewket, 2013;Viste et al., 2012).Details of SPI computation can be accessed in the user guides of SPI (WMO 2012).
The objective of this paper is therefore to monitor the magnitude, duration, frequency, spatial variability and trends of drought incidence in South Wollo, north east highlands of Ethiopia.

Description of the study area
South Wollo zone is located in the north east part of Ethiopia lying between 10°12 0 N and 11°4 0 0 N and 38°30 0 E and 40°05 0 E. Its zonal capital, Dessie, is found 400 kms North of Addis Ababa.It is one of the drought-prone and aid-dependent areas in Amhara Regional State (Bewket and Conway, 2007).The study area covers a total area of 17053.45km 2 with 18 rural and 2 urban districts.
South Wollo is characterized by diverse topographic features in which high mountainous and deeply incised canyons and gorges, valleys and plateaus with steep slopes dominate its most parts (Coltorti et al., 2007).The elevation ranged from the dry plains at 1,000 m altitude in the east to the high peaks above 3,500 m altitude in the west.However high-land areas ranging between 1,500 and 3,500 m altitude are the dominating feature of South Wollo (Rosell and Holmer, 2007).Because of the diverse topography, the study area experiences different climatic conditions that range from hot and arid lowlands in the eastern part to cold and humid highlands in the western part.
The mean annual temperature varies from less than 5°C in the western highlands to 22°C in the eastern lowlands.The annual rainfall varies between less than 1,000 mm in the western part to more than 1,200 mm in the eastern part.Bimodal rainfall pattern with shorter spring (March-May) and longer summer (June-August) characterizes the study area which leads to two harvest periods.The spring season is very influential for mid to high latitude areas, whereas the mid lands to lowland areas mainly depend on the summer rain.Except a few pocket areas where small-scale irrigation is practiced, crop production is rainfed.
As the topography is very rugged, soil erosion was a critical problem in the study area.It is highly degraded and deforested in terms of indigenous trees but does have considerable eucalyptus plantations.Alpine species unique to extreme highland areas are found in western highland parts of the study area.Local people in the area are engaged in subsistence agriculture for their livelihood.Unlike other areas of East Africa where highlands generally Meteorological drought assessment are the most food secure parts of a country, the opposite often is the case in Ethiopia.In fact, the crowded, steep-sloped highlands above 2,000 metres including large parts of South Wollo, are among the country's most famine-prone areas (Little et al., 2006) (Figure 1).

Data source
Monthly rainfall data from six stations for the period  were obtained from Ethiopian National Meteorological Agency (NMA) and used as an input variable to calculate meteorological drought.Initially, we collected rainfall data from 10 stations found in the study area.However, due to their missing data and shorter length of records, we only selected the six stations.

Standard Precipitation Index computation
SPI was designed to quantify the rainfall deficit for multiple timescales in the studied stations.The SPI is a z-score and represents the drought event departure from the mean, expressed in standard deviation units.SPI is a normalized index in time and space.This feature allows comparisons of SPI values among different locations.
Although SPI can be calculated from 1 month up to 72 months, 1-24 months is the best practical range of application (Guttman, 1999;WMO, 2012).We, therefore, computed the SPI values at two time-scales, i.e. 3 months (SPI-3) and 12 months or annual (SPI-12).The SPI-3 was used to assess droughts during spring (belg) and summer (kiremt) seasons which represent the shorter and longer rain seasons, respectively, and SPI-12 was used to assess the annual drought.Positive SPI values indicated greater than mean rainfall and negative values indicated less than mean rainfall.
For each month of the calendar year, new data series were created, with the elements equal to corresponding rainfall moving sums (Degefu and Bewket, 2013).Then, the SPI value provides a comparison of the rainfall over a specific period with the rainfall totals from the same period for all the years included in the historical record (Shahid, 2008).It is essentially a seasonally normalized, backwards-looking moving average of precipitation.For example, the three-month SPI calculated for August 2005 used the total rainfall of June-July-August 2005 and compared with the June-August rainfall totals of all the years considered in the study.Similarly, the 12-month SPI for December 2005 used the rainfall total for January 2005 to December 2005.
Conceptually, SPI is equivalent to the Z-score used in statistics, and it is calculated as: where SPI ij is the SPI of the ith month at the jth timescale, X ij is rainfall total for the ith month at the jth time scale, l ij and a ij are the long-term mean and standard deviation associated with the ith month at the jth time scale, respectively.In this study, SPI values were produced for each month of the year using SPI_SL.6.exe program developed by Colorado Climate Center (available at http://ulysses.atmos.colostate.edu/SPI.html).
This program classified drought events in four magnitude classes, i.e. if SPI ranges between 0 and À0.99, it is mild drought; À1.00 and À1.49, it is moderate drought; À1.50 and À1.99, it is severe drought; and extreme drought if SPI is À2.00 or less (McKee et al., 1993;WMO, 2012).Each drought event has a duration defined by its beginning and end.

Methods for smoothing time series data and trend detection
The rank-based nonparametric Mann-Kendall (MK) test (Mann, 1945;Kendall, 1975) has been commonly used to assess the significance of monotonic trends in hydrometeorological time series.Another rank-based nonparametric test, the Spearman's rho (SR) test (Lehmann, 1975;Sneyers, 1990), has sometimes been applied to detect trends in hydrological data.The study of Yue et al. (2002) and Viste et al. (2012) noted that these two tests have almost the same power to identify trends in time series.In comparison to the parametric t test, the common use of the nonparametric tests is mainly due to the consideration that they are more suitable for the situations of non-normal data, incomplete data, and missing data problems, which frequently occur in hydrometeorological studies.MK test simply calculates whether the variable is increasing or decreasing with time (Collins, 2009).
MK test, used by many researchers for trend detection due to its robustness for nonnormally distributed data, was applied in this study to assess trends in the time series data (Kendall, 1975;Mann, 1945).However, the MK test requires the time series data to be serially independent (Petrow and Merz, 2009).The standard p-values obtained from it are based on an assumption of independence between observations.The presence of a significant positive serial correlation tends to overestimate the significance of the trend (Petrow and Merz. 2009;Yue et al., 2002).Thus, the time series data should be tested for serial correlation and should be corrected if there is any significant autocorrelation (Petrow and Merz, 2009;Burn et al., 2010) before subjected the data to the MK trend test.To improve the performance of the test (MK), the most widely used procedure of the trend-free pre-whitening (TFPW) method can be applied to remove serial correlation from the time series, and hence to eliminate the effect of serial correlation on the MK test (Yue et al., 2002(Yue et al., , 2003;;Petrow and Merz, 2009).In this study, we found serial correlation of six out of eighteen time scales.However, only Ambamariam and Wereilu at 12-month timescale showed significant autocorrelation (p = 0.59 and 0.55), respectively.TFPW method was applied to remove the autocorrelation of Meteorological drought assessment these two stations following the works of Petrow and Merz (2009), Yue et al. (2002).Finally, we run the MK test.For detailed MK test statistics, see Kendall (1975) and Mann (1945) and the MK test manual prepared by Finnish Meteorological Institute (2002).

Mapping spatial distribution of drought incidences
The output from the SPI program is used as an input to ArcGIS to generate drought severity maps for the study area at 3 and 12-month time scales.To assess the spatial extent of droughts in the study area, SPI time series values of each meteorological station have been interpolated by Inverse Distance Weighted (IDW) method using the Spatial Analyst tool of ArcGIS.The IDW method gives better representation for interpolation of rainfall distribution over heterogeneous topographic terrain (Tagel et al., 2011).

Results
The results of the study are discussed under three sub-topics: (1) frequency and magnitude of drought events; (2) trends of occurrences of drought events; (3) spatial patterns of drought events.

Magnitude and frequency of drought events
Rainfall is bimodal in the study area (Figure 2).The contribution of summer (long rain period) rainfall to the average annual rainfall across stations was very high ranging from 65.7 per cent at Ambamariam to 53.56 per cent at Dessie.The number of drought months with different magnitude classes at short and long timescales (3 and 12-months) is shown on Figure 3. Drought months for 3 months (belg and kiremt rainfall) and 12 months (annual rainfall) were calculated using SPI.Drought frequency, in this study, was measured by the number of years which experienced negative SPI values in the total time series of 30 years.
3.1.1Spring season.The total number of drought events with mild, moderate, severe and extreme intensities computed at three-month timescale (March-May) was accounted for 46 per cent in spring season in all stations except Combolcha and Ambamariam (not shown here).However, they had varied magnitude classes.Extreme magnitude droughts occurred for 2 months at Ambamariam, Combolcha and Dessie.As the analysis of three-month timescale (March-May) showed, the year 1988, 1999, 2004 and 2007-2009 were drought years across all studied stations.Except Haik and Wereilu which had severe drought magnitude, Meteorological drought assessment extreme drought was recorded in 1999 in all stations ranging from À2.09 to À2.61.In the remaining drought years, the drought magnitude varied from mild to extreme in which Ambamariam and Dessie had extreme drought in 2008 and Combolcha and Wereilu experienced extreme drought in 2009.
3.1.2Summer season.The total number of drought events at the three-month timescale (June-August) in the entire period of analysis was found between 12 months at Dessie and 18 months at Mekaneselam, respectively.In addition to 1984 drought year, Ambamariam had extreme (SPI value À2) drought at this timescale (June-August) in 1987 and Wereilu with SPI value of À2.39 in 1996.The worst drought was recorded in 1984 which had substantial impact on the region.Even though its magnitude varied, drought occurred across the study area at this timescale in 1984, 1987, 1992-1993, 2004 and 2008.The frequency of occurrences of drought at this season was the highest at Mekaneselam (60 per cent) and the lowest at Dessie (53 per cent) implying that Mekaneselam experienced one drought episode in almost every two to three years.
3.1.3Annual drought.The total number of drought months at 12 month timescale (January-December) was found between 12 at Wereilu and 16 at Ambamariam which constitutes 40 and 53 per cent of the total number of drought incidences in the study period, respectively.At this time scale, the highest frequency of extreme drought was recorded at Wereilu in 1984 and 1996 with SPI value of À2.24 and À2.33, respectively.The year 1984 was the worst drought across the region with SPI value ranging from À1.9 to À3.17.Although there was varying degree of severity, drought was recorded across the study region in 1984, 1987, 2004 and 2008-2009.The total number of moderate droughts at 12 month timescale was the highest at Ambamariam and Mekaneselam with 4 and 6 months, respectively.This study confirms that the frequency of occurrences of droughts at summer was highest at Ambamariam.There is a remarkable correspondence between rainfall in Ethiopia and ENSO events (Haile, 1988).When past ENSO events are compared with drought and famine periods in Ethiopia, they show a remarkable association.Recent documented droughts of Ethiopia in 1987Ethiopia in -1988Ethiopia in , 1991Ethiopia in -1992Ethiopia in , 1993Ethiopia in -1994Ethiopia in , 2002Ethiopia in -2003Ethiopia in , 2008Ethiopia in -2009Ethiopia in , 2011Ethiopia in , 2015Ethiopia in -2016 were all strong El Nino years (UN OCHA, 2015;FAO, 2014).This suggests that Ethiopian intense and extended drought periods either coincide or follow El Nino events.In recent years, the meteorological droughts are increasing especially in spring (belg) season (Figure 3) in the study region, and they have strong correlation with the El Nino events.Forecasting the occurrences of El Nino events will provide insights to estimate drought occurrences and plan adaptation strategies to minimize the associated risks.

Trends of drought occurrences
Trends of drought occurrences for 3 months (spring and summer) and 12 months (annual) timescales were shown in Figure 4. SPI values of May, August and December were considered to represent the drought conditions from March-May (spring season), June-August (summer season) and January-December (annual rainfall), respectively.
The computed SPI values for 3 months timescale during spring revealed that the occurrence of negative rainfall anomalies or frequent droughts were observed in the 1980s and 2000s and positive rainfall anomalies in the 1990s at most of the meteorological stations.The SPI value for 3 months timescale during summer showed increasing trends, which indicates declining of occurrences of droughts from time to time across the study stations.The SPI value for 12 months (annual) timescale showed negative rainfall anomalies at 1980s, the beginning of 1990s and after the mid of 2000s, while positive rainfall anomalies were observed at the end of 1990s and beginning of 2000s in the studied period.

Meteorological drought assessment
The MK trend test showed decreasing changes in SPI values in all stations suggesting increasing tendency of drought incidence at the three-month timescales during spring (Figure 4) However, significant trend (p < 0.05) was observed only at Mekaneselam station.On the other hand, the SPI values at the three-month timescale during summer showed increasing changes across stations, however, significant at Dessie, Haik and Mekaneselam (p < 0.05).Similar increasing trend of drought was also observed at 12-month timescale in all stations except Combolcha and Mekaneselam.However, the changes were not significant.
The trend analysis shows that there was no statistical evidence of any positive or negative trend of meteorological drought severity and frequency for the study area except Mekaneselam at the three-month timescale during spring and Dessie, Haik and Mekaneselam at the three-month timescale during summer.Although trends at all timescales were not statistically significant, increasing tendencies of drought were observed during spring season and decreasing tendencies of drought during summer and annual scale in the study region.

Spatial patterns of drought occurrences
This study examined patterns of drought across the study region using time series  standardized precipitation index (SPI).The spatial patterns of drought at different time steps (three-and 12-months) were depicted in Figures 5(a  Except small pocket areas with severe droughts in eastern stations, the other stations were prone to moderate drought magnitude during spring.Extreme drought magnitude was also occurred at Ambamariam, Combolcha and Dessie in the same season.During summer, the western stations (Ambamariam, Mekaneselam and Wereilu) were highly exposed to high frequency of mild and moderate droughts, whereas eastern stations were less exposed to droughts.At annual timescale, extensive areas in the north east part of the region (Ambamariam, Combolcha and Dessie) also experienced high frequency of drought.Generally, the total frequency of drought was the highest at Mekaneselam at spring and summer and at Ambamariam at annual time scale.

Discussions
In this study, we presented a brief drought analysis at annual and seasonal time steps using SPI.We assessed the magnitude, frequency, trend, spatial pattern and probability of drought events over South Wollo from 1984 to 2014 at 3-and 12-month time scales.As the study area experiences bimodal rainfall (spring and summer) occurring between March to May and June to August respectively, we computed the SPI at three-month timescale for both seasons.
Results indicate that occurrences of drought in South Wollo showed temporal variation.The total number of drought events was higher in summer season than spring at Ambamariam, Combolcha and Mekaneselam.Summer is the main rainy season in the study  (Viste et al., 2012).Failing to rain at this season means failure of 85-95 per cent of the Ethiopian food crop which is produced in this season (Degefu, 1987).
Most of the studied stations experienced extreme magnitude droughts.The year 1984 was the most drastic and distinct-wide drought episode.All stations experienced extreme magnitude drought ranging from À2.18 to À3.44 at 3-month (summer) and from À1.9 to À3.03 at 12-month (annual) timescales.Except Wereilu and Haik (which had severe magnitude), extreme magnitude droughts were also observed at all stations in 1999 during spring season.This finding agrees with the findings of Viste et al. (2012) who found driest conditions all over Ethiopia during spring season in 1999.Additional extreme magnitude drought also occurred during spring at Ambamariam and Dessie in 2008 with SPI values of À2.28 and À2.22, respectively.There were also extreme magnitude droughts at Combolcha and Wereilu during spring season in 2009 with SPI value of À2.5 and À2.14, respectively.At the three-month timescale during summer, Ambamariam and Wereilu experienced extreme magnitude droughts in 1987 and 1996 with SPI values of À2. and À2.39, respectively.
The frequency of occurrences of drought was the highest (60 and 53 per cent) at Mekaneselam and Ambamariam at 3 month during summer and 12-month annual timescale, respectively.This shows that Mekaneselam and Ambamariam were frequently stricken by drought compared to other studied stations in South Wollo.
The occurrences of extreme magnitude droughts have important implications on rainfed agriculture (Ellis et al., 2010;Morid et al., 2006).The impacts of 1984 drought greatly expanded food insecurity, malnutrition, devastated livelihoods and caused for the loss of millions of life in the region.This was also true all over Ethiopia.This finding agrees with previous research findings of (El Kenawy et al., 2016;Mosaad, 2015;Bayissa et al., 2015;Degefu and Bewket, 2013;Viste et al., 2012;Tagel et al., 2011;Edossa et al., 2010;Bewket and Conway, 2007) who reported extreme drought in different parts of the country in the same period.Segele and Lamb (2005) also extensively demonstrated the severity of the 1984 drought over Ethiopia, particularly during the summer season.
The analysis revealed that in 1996 the SPI values at 3-month (June-August) and 12month timescales at Wereilu and in 1987 at 3-month (June-August) at Ambamariam were unique to other stations which showed extreme droughts with SPI value of À2.39 and À2.33 at Wereilu, respectively, and À2.0 at Ambamariam.The analysis implies that drought magnitude can vary largely over space and this exceptional extreme drought shows a local character, where small area was affected, while others had less severe drought magnitude or even no drought.This finding is supported by the study of El Kenawy et al. (2016) who found similar drought occurrences in studying the changes in the frequency and severity of hydrological droughts over Ethiopia from 1960 to 2013.
The most striking characteristic of a drought is the change in its frequency as the time scale changes (Umran, 1999).On shorter time scales, droughts became more frequent but their duration was short.The opposite is true for longer time scales.Their frequency becomes less but remains for longer time.At the three-month scale, drought frequency increased but its duration decreased.This means that drought became more frequent at shorter time scales but stayed for shorter periods (Degefu and Bewket, 2013;Edossa et al., 2010;Umran, 1999).

Meteorological drought assessment
Generally, the temporal SPI analysis shows that most of the stations measured extreme, severe, moderate and mild drought episodes in the years 1984, 1987/1988, 1992/1993, 1999, 2003/2004 and 2007/08.Indeed, these years were among the worst drought years in the history of Ethiopia which agrees with previous findings of ( Viste et al., 2012, Tagel et al., 2011and Edossa et al., 2010).The drought years identified by this SPI analysis for South Wollo are known for their substantial damage in terms of life and economic losses like other parts of Ethiopia which agrees with the findings of (Bayissa et al., 2015;Tagel et al., 2011).It is very important to remind that SPI used the monthly rainfall to analyze the drought episode in this study.However, the monthly rainfall data did not reflect the daily rainfall characteristics, such as the beginning and end time of dry spells which have a great implication on the effect of droughts especially for agricultural activities.
The MK test results indicate that SPI values were ranging from À0.14 to À0.51 Z-unit per decade at the three-month timescales during spring season suggesting increasing drought episodes.However, it was significant only at Mekaneselam (p < 0.05).On the other hand, the SPI values were increasing at the three-month (June-August) at all stations albeit the trends were statistically significant at Dessie, Haik and Mekaneselam.Generally, significant increasing tendencies of drought were observed during spring season and decreasing tendencies of drought during summer and annual scale in the study region.

Conclusions
In this study, a brief drought analysis was presented using SPI.It is a very important tool for quantifying drought and comparing its characteristics over time and space.We used SPI, in this study, to examine the magnitude and frequency, trend, patterns and probability of drought occurrences.Here, droughts occurrences were analyzed both at 3-and 12-month time scales, and the number of droughts at six stations are presented in Figure 3. Though almost all stations in the study region suffer from drought, it is important to consider that all of the stations did not experience well-defined drought episode during the same periods.In other words, temporal distribution and frequency of droughts varied markedly among each station.
Extreme droughts were more pronounced in stations where their altitude is above 3,000 masl during spring and in stations having altitude of less than 3,000 masl during summer.Similarly, stations found in western part of the study region were exposed to high frequency of severe and extreme droughts at annual timescale.The drought years in the study region identified by this SPI analysis were among the worst drought years in the history of Ethiopia.The year 1984, for example, was the most drastic and distinct-wide drought episode.Almost all stations experienced extreme magnitude drought at 3-month (summer) and 12-month timescales in the specified year.
Generally, the entire study area can be considered as drought prone area.Increasing tendencies of drought were observed during spring and decreasing tendencies at summer and annual timescales.The patterns of drought events in the study area are highly localized.Special attention (local-scale planning) should be given, while decision makers plan to effectively manage drought.The findings of this study have implications for drought management, early warning system, preparedness and contingency planning and climate change adaptation.In real sense, drought is a climatic event that cannot be prevented very easily, but interventions and preparedness to drought can help to cope with drought by developing more resilient ecosystems, improving resilience to recover from drought and taking various adaptation strategies like water harvesting, making irrigation system more efficient and a geographical shift of agricultural system.IJCCSM 10,1 Figure 1.Location map of South Wollo and meteorological stations used in this study Figure 2. Bimodal rainfall distribution of the study area Figure 3.
Figure 4.The Mann-Kendall's trend test for spring, summer and annual time scales )-(d) to 8(e1)-(e3).The Figure 5. Spatial distribution of drought events at the three-month time scale in spring (March-May) season in South Wollo, north east highlands of Ethiopia: (A) mild drought, (B) moderate drought, (C) severe drought, (D) extreme drought Figure 6.Spatial distribution of drought events at the three-month time scale in summer (June-August) season in South Wollo, north east highlands of Ethiopia: (A) mild drought, (B) moderate drought, (C) severe drought, (D) extreme drought Figure 7. Spatial distribution of drought events at 12month time scale in South Wollo, north east highlands of Ethiopia: (A) mild drought, (B) moderate drought, (C) severe drought, (D) extreme drought Figure 8. Spatial distribution of total drought events in, seasons and at annual timescales in South Wollo, north east highlands of Ethiopia: (E1) spring (March-May); (E2) summer (June-August); (E3) annual drought