Abstract
Drought assessment is necessary for creating adaptation and resilience measures for the livelihoods of the affected communities. This study assessed drought trends in Kenya's Upper Ewaso Ng'iro River Basin (UENB) from 1981 to 2020. A Standardized Precipitation Index (SPI), a precipitation-based index, and a Standardized Precipitation Evapotranspiration Index (SPEI), a multivariate index that considers the difference between precipitation and potential evapotranspiration (PET), were used to evaluate drought severity and frequency over varying timescales. Monthly rainfall and temperature data for 10 stations within the basin were analyzed to calculate the SPI and SPEI time series values for 3, 6, and 12 months. The results demonstrate an alarming increase in the severity and frequency of drought events in the UENB since 1999. Additionally, the study reveals that the SPI and SPEI indices differ in identifying temporal and spatial drought characteristics, with longer timescales showing improved accuracy. Notably, the SPEI identifies more extensive and severe drought periods in the region compared to the SPI. The research findings are crucial in guiding policy decisions related to SDGs as they provide valuable information on drought trends necessary for implementing effective drought adaptation and resilience measures and promoting sustainable development in the UENB.
HIGHLIGHTS
The study describes the temporal and spatial evolution characteristics of meteorological drought in the region.
There has been an increase in drought frequency in the basin over the 40 years, and it has significantly increased since 2000.
The study shows the relationship between the SPI and SPEI in identifying drought events and that the SPEI performed better than the SPI in identifying drought.
INTRODUCTION
Drought is a natural climatic hazard considered a relatively long-term average condition of a region's balance between precipitation and evapotranspiration (Livneh & Hoerling 2016). Other climatic factors such as temperature, wind, and below-average relative humidity are also used to identify drought periods (Van Loon 2015). Drought events are hard to detect since they develop slowly and can affect a large area (Khan et al. 2020). They are among the most devastating natural disasters affecting water resources, ecosystems, and people and are considered crises worldwide (Hoegh-Guldberg et al. 2018). According to the Intergovernmental Panel on Climate Change report (IPCC 2021), global changes have become more frequent in recent years, especially in vulnerable areas, due to increased global warming, drought, and other climate systems. Therefore, the global impact of droughts will keep growing and affecting more regions worldwide.
There are increased drought events in Africa, for example, in the Horn of Africa and Kenya, causing famine, diseases, and leading to floods after the drought. Around 84% of Kenya's occupied land is Arid and Semi-Arid Land (ASAL) (Government of Kenya, Gok, 2021) and has increasingly been experiencing drought events, the frequency and severity of which are increasing. As a result, the ASALs of Kenya, which hosts 38% of the people, 70% of national livestock herds, and 90% of the wildlife, are experiencing severe effects (GoK 2021). Additionally, agricultural production, the country's largest foreign exchange earner, with tea and horticultural production being the most exported products, will be highly impacted (Lanari et al. 2018). Therefore, drought will affect the economy, retarding development in the country.
The Upper Ewaso Ng'iro River Basin (UENB), one of the regions practising large-scale commercial horticultural production, has been experiencing increased demand for scarce natural resources, i.e., fresh water and land (Lesrima et al. 2021). In addition, according to Famine Early Warning Systems Network (FEWSNET) (2020), the basin is also experiencing more extreme droughts. The significant impacts of drought in the catchment are increased water stress, reduced agricultural production, increased food insecurity, malnutrition, and conflicts among the communities (Lanari et al. 2018; Lesrima et al. 2021).
According to the World Meteorological Organization (WMO) & the Global Water Partnership (GWP) (2016), the main drought characteristics are duration, intensity, and severity. Drought duration is the period when drought occurs continuously; drought intensity is the drought magnitude, and severity is the cumulative occurrence of drought over a period expressed as the product of intensity and duration. Droughts are classified into four types: meteorological, hydrological, agricultural, and socioeconomic.
To effectively manage droughts, early warning systems that rely on accurate and timely drought development information are necessary (Moghimi et al. 2020; Yue et al. 2022). Drought assessment and forecasting are crucial in monitoring drought conditions as they provide insights into droughts' historical patterns and impacts (Mishra & Singh 2010). However, due to drought complexity, drought monitoring and forecasting have challenged climatologists and decision-makers (Hao et al. 2018). Consequently, drought assessment has been achieved using indices describing and quantifying drought conditions (WMO & GWP 2016). These indices encompass climate variables that capture droughts' key aspects, including duration, severity, intensity, and spatial extent (Mishra & Singh 2010; Dalezios 2014). They can also be used, depending on the index, to provide historical drought information (WMO & GWP 2016).
The Standard Precipitation Index (SPI) and Standard Precipitation and Evapotranspiration Index (SPEI) are considered universal meteorological drought indices that allow comparisons of drought conditions across different climate regions (Ojha et al. 2021). The SPI is a precipitation-based index developed to measure the deficiency of precipitation at several time scales (Mishra & Singh 2010). It is the most extensively applied drought index worldwide (Mutiga et al. 2011). The WMO identified it as the most suitable drought monitoring and forecasting index (WMO & GWP 2016). On the other hand, the SPEI, a multiscalar drought index, incorporated temperature and precipitation. It was developed by Vicente-Serrano et al. (2010) to identify drought periods, utilizing the same concepts as the SPI but including temperature (WMO & GWP 2016). The calculation of the SPEI is similar to that of the SPI but uses the monthly (or weekly) difference between precipitation and Potential Evapotranspiration (PET). By leveraging the SPI and SPEI, decision-makers gain valuable tools for comparing and assessing drought conditions across diverse climate regions, allowing for better planning and resource allocation (Ojha et al. 2021). These indices provide a standardized framework for evaluating and understanding drought severity, aiding in effective drought management strategies, and mitigating the potential socioeconomic and environmental impacts caused by drought events.
In the UENB, resources, particularly water, are unevenly distributed due to the large amounts required to sustain agriculture, increased climatic changes, and drought occurrence (Kimwatu et al. 2021; Lesrima et al. 2021). This will leave the marginalized communities suffering more significant impacts for a prolonged period. Therefore, there is a need to develop drought resilience and transformative adaptation strategies for the communities affected by drought. The formulation of these strategies will depend on a timely, accurate, and reliable understanding of the characteristics and features of lack (Pei et al. 2020).
Huho et al. (2010) and Karanja (2018) used the SPI to assess the drought and its impacts in Laikipia County. Their findings showed that frequent droughts in the county hindered sustainable rain-fed agriculture, the primary livelihood source. Odhiambo et al. (2018) also conducted a study assessing the basin's drought and flood hazards between 2004 and 2014. Using the SPI to evaluate the meteorological droughts and the SPEI for agricultural droughts, their results indicated the usefulness of combined drought and flood hazards assessment for planning for flood and drought risk reduction activities. In their study, Kimwatu et al. (2021) developed a Socioeconomic Drought Index (SeDI) to monitor the evolution of drought in the basin. The domestic water deficit index, bareness index, normalized difference vegetation index, and water accessibility index were the input variables. The study revealed that the basin manifested a moderate severity level between 85 and 96%, a severe level between 2.2 and 13.3%, and an extreme level between 0.73 and 1.17%.
Additionally, Kigumi (2014) examined the use of Tropical Rainfall Measuring Mission (TRMM) precipitation estimates in meteorological drought monitoring alongside stream flow modelling for drought identification in the UENB's Narumoro sub-catchment. The analysis found TRMM data suitable for streamflow modelling since the identified hydrological drought pattern was similar to precipitation. However, despite all these studies done in the region, no study has been done to analyze the evolution of drought in the basin considering the different climatic zones. Also, there is no study done to examine the applicability of the SPI and SPEI and identify a suitable drought index for meteorological drought monitoring in the basin.
Drought monitoring is vital in informing policies by developing a generalized understanding of droughts' severity, frequency, and duration. This study will contribute towards meeting the SDG targets set by the United Nations. In particular, it informs the SDGs 13 and 11.5 and 11.B (The United Nations Department of Economic & Social Affairs (UN DESA) 2022), by providing information essential for developing drought risk resilience strategies and promoting mechanisms for raising capacity for effective climate change-related planning and management.
This study calculated and analyzed drought years in the basin between 1980 and 2020. The main objective is to provide information about drought occurrences from 1981 to 2020 in the UENB, to support the early warning system and drought management. The specific objectives include (i) assessing the temporal and spatial drought characteristics at 3-, 6-, and 12-month time scales as shown by the SPI and SPEI and (ii) comparing the consistency and applicability of the SPI and SPEI in drought monitoring in the UENB.
MATERIALS AND METHODS
Study area
The annual temperatures in the upper parts of the catchment range between 9 and 22 °C, while at lower parts range between 15 and 29 °C (Mutiga et al. 2011). The basin's annual PET ranges between 1,200 and 1,800 mm/year. The basin has a yearly rainfall spatial and temporal variation of 300 mm in the Northeast areas to 1,500 mm at the source in the Nyandarua Ranges (Ericksen et al. 2012). The rainfall pattern is tri-modal, experiencing long rains from April to June, short rains in October and December, and a third rainfall season in July and August.
The basin's primary land uses are livestock rearing, agriculture, and wildlife and forest conservation. Pastoralists and commercial enterprises manage the cattle ranches, and agriculture is operated and owned by smallholder farmers and commercial agriculture enterprises (Omwoyo et al. 2017). Wildlife and forest conservation are managed by both government and private wildlife conservancies.
Data collection
Rainfall and temperature data from 1981 to 2020 for 10 stations were obtained from the Kenya Meteorological Department (KMD) and the Centre for Training and Integrated Research in ASAL Development (CETRAD). The stations named according to the CETRAD are given in Table 1. The study uses the average monthly rainfall data to calculate the SPI. In contrast, in calculating the SPEI, the same stations' maximum and minimum temperature and rainfall data from 1981 to 2020 were used.
Station ID . | Station name . | Longitude . | Latitude . |
---|---|---|---|
1 | Archer's Post | 37.6681 | 0.6335 |
9 | Colcheccio | 36.80318 | 0.61932 |
10 | Dol Dol Dao | 37.15697 | 0.38849 |
22 | Isiolo Dao | 37.58509 | 0.35384 |
51 | Mukenya Farm | 36.82054 | 0.24204 |
69 | Mukongondo Farm | 37.29269 | 0.09152 |
80 | Rumuruti Mow | 36.54844 | 0.26748 |
83 | Segera Plantations | 36.88782 | 0.1689 |
89 | Suguroi Estate | 36.64205 | −0.02835 |
94 | Wamba Do | 37.33198 | 0.98218 |
Station ID . | Station name . | Longitude . | Latitude . |
---|---|---|---|
1 | Archer's Post | 37.6681 | 0.6335 |
9 | Colcheccio | 36.80318 | 0.61932 |
10 | Dol Dol Dao | 37.15697 | 0.38849 |
22 | Isiolo Dao | 37.58509 | 0.35384 |
51 | Mukenya Farm | 36.82054 | 0.24204 |
69 | Mukongondo Farm | 37.29269 | 0.09152 |
80 | Rumuruti Mow | 36.54844 | 0.26748 |
83 | Segera Plantations | 36.88782 | 0.1689 |
89 | Suguroi Estate | 36.64205 | −0.02835 |
94 | Wamba Do | 37.33198 | 0.98218 |
Standard Precipitation Index Analysis
Based on Equation (1), the 3-, 6-, and 12-month SPI values in this study were calculated using the SPI packages available at Comprehensive R Archive Network.
An event is regarded as drought if the SPI value is constantly negative. Conversely, when the value becomes positive, it is considered the end of the event (Mishra & Singh 2010; Khan et al. 2020). Table 2 shows the classification of drought based on the SPI.
SPI/SPEI values . | Drought category . |
---|---|
Index ≥ 2 | Extremely wet |
1.5 ≤ Index < 1 | Severely wet |
1 ≤ Index < 1.5 | Moderately wet |
−1 ≤ Index < 1 | Near normal/mild drought |
−1.5 ≤ Index < −1 | Moderately dry |
−2 ≤ Index < −1.5 | Severely dry |
Index < −2 | Extremely dry |
SPI/SPEI values . | Drought category . |
---|---|
Index ≥ 2 | Extremely wet |
1.5 ≤ Index < 1 | Severely wet |
1 ≤ Index < 1.5 | Moderately wet |
−1 ≤ Index < 1 | Near normal/mild drought |
−1.5 ≤ Index < −1 | Moderately dry |
−2 ≤ Index < −1.5 | Severely dry |
Index < −2 | Extremely dry |
Standard Precipitation Evapotranspiration Index Analysis
The negative SPEI values indicated drought conditions; positive values corresponded to wetter or above-normal conditions (Table 2). The study utilized the SPEI packages available at Comprehensive R Archive Network to compute the SPEI values for all the meteorological stations at 3-, 6- and 12-month time scales.
Spatial analysis of the SPI and SPEI
Geographic Information Systems (GIS) was used for isopleth mapping. All the SPEI and SPI data from the 10 stations were interpolated using the kriging interpolation in the Surfer Mapping Software to estimate their respective amounts in the UENB. In kriging, the weights of the 10 known station points were fitted into a function within a specified radius to determine the output value of each station.
The Mann–Kendall trend test method
To analyze the trend of the drought characteristic based on the SPI and SPEI, the Mann–Kendall (M–K) trend test and Sen's slope (SS) estimator were used. The M–K trend test is a non-parametric statistical testing method that the WMO recommends to study hydrological and meteorological variable trends (Pei et al. 2020). Based on the formula described by Khan et al. (2020), Pei et al. (2020), and Mehta & Yadav (2022), the Modified M–K (MMK) trend test package in R-programming language was used over the SPI and SPEI indices to assess the drought trend in the study area for the 10 stations. With a 95% significance level and a p value of ≤0.05, in the test, positive values of SS indicate an increasing trend, meaning increases in wet conditions. In contrast, the negative values show a decreasing trend meaning increases in dry conditions. The null hypothesis H0 of no trend is rejected when p < 0.05, meaning there is a significant trend.
Assessment of the linear relationship between the SPI and SPEI
The Pearson correlation coefficient (r) tests the linear correlation between variables with continuous data from the same experiment – SPI and SPEI, in this case – and variables that follow a bivariate normal distribution (Mehr et al. 2020). It also shows how pairs of variables are related to each other. For example, values of +1 show a perfect positive correlation, −1 shows a perfect negative correlation, and 0 shows no linear relationship. Statistical Package for the Social Sciences (SPSS) was used in this study to calculate the coefficient and test the SPI/SPEI relationship.
RESULTS
Temporal drought variations in the UENB
From the results, the SPEI identified more drought years than the SPI in the Northern (stations 1, 22, and 94) and Central (10, 9, 80) parts of the basin. SPEI 3 indicated more droughts years not shown by the SPI in 2002, 2007, and 2012 and SPEI 6 and 12 in 2017 and 2018 (Figures 2 and 3). In the Southern (stations 51, 69, 83, and 89) parts of the basin, closer to the river's source, both indices indicated fewer drought years than the North and Central parts. In the South, the SPI and SPEI revealed similar drought years (Figure 4). The results also show that between 2013 and 2019, the basin experienced a long drought with extreme periods. The SPEI identified extreme droughts in 1987, 2000, 2004, 2006, 2009, 2014, 2018, and 2019, while the SPI identified extreme droughts in 2000, 2004, 2009, 2014, 2018, and 2019 (Figures 2–4).
The summary statistics results in Table 3 show the standard deviation (SD) and the standard error (SE) values indicating that the variation of the SPI and SPEI values from their mean values was almost similar. For example, for 6-month SPI and SPEI, the SD was 0.971115 and 0.972326, and the SE was 0.014017 and 0.014034. The values for other timescales are also close (Table 3). Additionally, the SPI and SPEI mean, minimum and maximum values are relatively relative at all time scales. The table indicates that the SPI and SPEI differed in identifying dry years for the different timescales. For example, the SPI 3 identifies 2000 as the driest year, while SPEI 3, 2011. However, it is worth noting for both SPI and SPEI, the 6 and 12 months showed similar dry years. For example, SPEI 6 and 12 identified 2006 as the driest year, while the SPI 6 and 12 in 2000.
Time scale . | Index . | Mean . | Median . | SD . | SE . | MIN . | Year . | MAX . | Year . |
---|---|---|---|---|---|---|---|---|---|
3-month | SPI | 0.006059 | −0.02545 | 0.974392 | 0.014064 | −3.61091 | 2009 | 3.956279 | 1998 |
SPEI | −0.00672 | −0.02694 | 0.981245 | 0.014163 | −3.85381 | 2011 | 2.744269 | 1998 | |
6-month | SPI | 0.003492 | −0.00894 | 0.971115 | 0.014017 | −2.79673 | 2000 | 3.842416 | 1998 |
SPEI | −0.00589 | −0.01859 | 0.972326 | 0.014034 | −3.51613 | 2006 | 2.756177 | 1998 | |
12-month | SPI | −0.00503 | −0.02425 | 0.960006 | 0.013856 | −2.77908 | 2000 | 3.542871 | 1998 |
SPEI | −0.0106 | −0.00407 | 0.962691 | 0.013895 | −2.82916 | 2006 | 2.594733 | 1998 |
Time scale . | Index . | Mean . | Median . | SD . | SE . | MIN . | Year . | MAX . | Year . |
---|---|---|---|---|---|---|---|---|---|
3-month | SPI | 0.006059 | −0.02545 | 0.974392 | 0.014064 | −3.61091 | 2009 | 3.956279 | 1998 |
SPEI | −0.00672 | −0.02694 | 0.981245 | 0.014163 | −3.85381 | 2011 | 2.744269 | 1998 | |
6-month | SPI | 0.003492 | −0.00894 | 0.971115 | 0.014017 | −2.79673 | 2000 | 3.842416 | 1998 |
SPEI | −0.00589 | −0.01859 | 0.972326 | 0.014034 | −3.51613 | 2006 | 2.756177 | 1998 | |
12-month | SPI | −0.00503 | −0.02425 | 0.960006 | 0.013856 | −2.77908 | 2000 | 3.542871 | 1998 |
SPEI | −0.0106 | −0.00407 | 0.962691 | 0.013895 | −2.82916 | 2006 | 2.594733 | 1998 |
SD, standard deviation; SE, standard error; Min, minimum; Max, maximum
Spatial Drought Evolution in UENB
For the SPI, the North-eastern UENB tended towards extreme drought at all time scales, while the Central and Southeast tended towards mild drought. For the SPEI, extreme to severe droughts were consistently seen in the North, Northwest, Northeast, and Central part of the basin from 1981 to 2020, while the East and West showed moderate drought. For the 6- and 12-month scales, the Northeast shows extreme drought consistently from 1981 to 2020 and mild drought to near normal from the Central part to the Southeast from 1981 to 2000. From 2001 to 2020 at all timescales, the Central part gradually shows an increase in severe to extreme drought. Overall, the two indices indicated the North part of the basin, to the river's outlet, to be severe to highly drought-prone.
As for the drought duration (Figure 7), the SPEI showed the region with the most prolonged period being the North and Central regions, with occurrence in more than 410 months. The South parts had the shortest drought periods, with 380 months of drought, while the East, West, and parts of the lower Central region indicated drought periods between 385 and 400 months. The SPI also showed a similar trend but with a smaller number of drought months compared to the SPEI. The North and West had 400 months, the Central, and East, between 395 and 360 months, and the South, less than 360 months.
The intensity of drought events, i.e., the number of events less than −1.5, is shown in Figure 8. The drought intensity in the region has been increasing from South to North and with increasing timescales. The number of months with high drought intensity shown by the SPI is less than those by the SPEI. The SPI indicated 16–31 months, 18–38 months, and 22–38 months, while the SPEI showed 26–73 months, 26–40 months, and 26–42 months of high intensity for the 3-, 6,- and 12-month timescales. For all the indices, it was consistent that the number of drought events with high intensity increased from South to North of the basin, with the areas around the mountains showing fewer months of high intensity (22–26 months) while the lowlands showed higher number (up to 42 months). Drought severity, duration, and intensity in the UENB increase from the South and the river's source to the North in the lowlands and the outlet.
Between the two indices, the drought intensity indicated by the SPI is milder than that shown by the SPEI. The SPEI performs better than the SPI in detecting the spatial evolution of drought because it considers both the temperature and rainfall changes. In summary, due to the differences in the value of the SPI and SPEI in time series, their drought characteristics differed in space at different timescales.
Consistency of the SPI and SPEI
Further investigation of the drought features through correlation analysis between the SPI and SPEI showed temporal variations in each station's correlation coefficient (r) (Table 2). Generally, the SPI and SPEI showed a significant (p < 0.01) positive correlation in all stations (Table 4). The strongest relationship (r = 0.876) was observed between the SPI 12 and SPEI 12 in stations 51, 83, and 89, and the weakest (r = 0.476) between the SPI 6 and SPEI 6 in station 10. It was relatively close at longer timescales and decreased at shorter timescales. Similar to this study, Mehr et al. (2020) and Ojha et al. (2021) found that the longer the timescale, the higher the correlation, and the shorter the timescale (in this case, 3- and 6- months), the lower the correlation. The analysis shows that the indices identify drought better at longer than shorter timescales. The results also reveal that in an arid area such as the Northern side, i.e., stations 1, 9, 10, and 94, where there is a higher average temperature and lower average precipitation, the correlation between the SPI and SPEI is lower than in other climatic areas. This result is similar to those shown by Homdee et al. (2016) and Lotfirad et al. (2022), indicating poor SPI performance in ASAL regions. This could be explained by the fact that the SPEI considers evapotranspiration, which is higher in the Northern (annual average 1,800 mm) than Southern parts (1,200 mm) (Mehr et al. 2020).
Station ID . | SPI3/SPEI 3 . | SPI6/SPEI6 . | SPEI12/SPEI12 . |
---|---|---|---|
1 | 0.737** | 0.613** | 0.813** |
9 | 0.729** | 0.775** | 0.793** |
10 | 0.754** | 0.476** | 0.771** |
22 | 0.677** | 0.761** | 0.807** |
51 | 0.722** | 0.786** | 0.876** |
69 | 0.771** | 0.797** | 0.823** |
80 | 0.755** | 0.794** | 0.792** |
83 | 0.722** | 0.786** | 0.876** |
89 | 0.722** | 0.786** | 0.876** |
94 | 0.581** | 0.664** | 0.703** |
Station ID . | SPI3/SPEI 3 . | SPI6/SPEI6 . | SPEI12/SPEI12 . |
---|---|---|---|
1 | 0.737** | 0.613** | 0.813** |
9 | 0.729** | 0.775** | 0.793** |
10 | 0.754** | 0.476** | 0.771** |
22 | 0.677** | 0.761** | 0.807** |
51 | 0.722** | 0.786** | 0.876** |
69 | 0.771** | 0.797** | 0.823** |
80 | 0.755** | 0.794** | 0.792** |
83 | 0.722** | 0.786** | 0.876** |
89 | 0.722** | 0.786** | 0.876** |
94 | 0.581** | 0.664** | 0.703** |
**Correlation is significant at the 0.01 level (two-tailed).
SPI and SPEI trends
Table 5 presents the drought trends and magnitude based on the SPI and SPEI in the UENB basin. The M–K trend test and Sen's slope estimator showed a significant downward drought trend at stations 1, 9, 10, 22, 69, and 94 for all the time series for the SPEI but no significant trend at the same stations for the SPI. At station 80, there was an increasingly significant trend for the SPI at all time scales but no significant trend for the SPEI. There was no trend for any SPI or SPEI timescale at stations 51, 83, or 89, but SS shows negative values meaning that drought events were increasing there. Based on SS, most stations showed a decreasing trend for both SPI and SPEI for all the time series, i.e., stations 1, 22, 51, 69, 83, and 89, while other stations showed both increasing and decreasing trends for the indices at different time scales. For example, stations 9 and 10 indicated a decreasing trend for both the SPEI time series and SPI at three months while indicating an increasing trend for the SPI at the 6- and 12-month scales. Station 94 told an increasing wetness trend for the SPI at all time scales. Generally, the trend in the basin shows an increase in dry periods, which could also mean these periods could keep increasing over the coming years.
Station . | Test . | SPI3 . | SPI6 . | SPI12 . | SPEI3 . | SPEI6 . | SPEI12 . |
---|---|---|---|---|---|---|---|
1 | P value | 0.54811 | 0.35387 | 0.29584 | 0.02910 | 0.02574 | 0.00916 |
Sen's value | −0.00035 | −0.00069 | −0.00085 | −0.00199 | −0.00239 | −0.00291 | |
9 | P value | 0.80148 | 0.91203 | 0.61914 | 0.01940 | 0.02033 | 0.01931 |
Sen's value | −0.00015 | 0.00007 | 0.02667 | −0.00120 | −0.00144 | −0.00152 | |
10 | P value | 0.80148 | 0.91203 | 0.61914 | 0.01940 | 0.02033 | 0.01931 |
Sen's value | −0.00015 | 0.00007 | 0.00032 | −0.00120 | −0.00144 | −0.00152 | |
22 | P value | 0.49384 | 0.31089 | 0.24660 | 0.02523 | 0.02227 | 0.00772 |
Sen's value | −0.00039 | −0.00074 | −0.00089 | −0.00196 | −0.00237 | −0.00289 | |
51 | P value | 0.14418 | 0.14863 | 0.16067 | 0.13768 | 0.13413 | 0.11378 |
Sen's value | −0.00107 | −0.00150 | −0.00173 | −0.00169 | −0.00193 | −0.00227 | |
69 | P value | 0.10999 | 0.09884 | 0.10818 | 0.03956 | 0.07002 | 0.08340 |
Sen's value | −0.00118 | −0.00174 | −0.00209 | −0.00220 | −0.00244 | −0.00251 | |
80 | P value | 0.00640 | 0.00478 | 0.00055 | 0.09978 | 0.06137 | 0.00932 |
Sen's value | 0.00156 | 0.00222 | 0.00286 | 0.00122 | 0.00154 | 0.00201 | |
83 | P value | 0.14418 | 0.14863 | 0.16067 | 0.13768 | 0.13413 | 0.11378 |
Sen's value | −0.00107 | −0.00150 | −0.00173 | −0.00169 | −0.00193 | −0.00227 | |
89 | P value | 0.14418 | 0.14863 | 0.16067 | 0.13768 | 0.13413 | 0.11378 |
Sen's value | −0.00107 | −0.00150 | −0.00173 | −0.00169 | −0.00193 | −0.00227 | |
94 | P value | 0.54648 | 0.82724 | 0.79423 | 0.00014 | 0.00000 | 0.00000 |
Sen's value | 0.00034 | 0.00013 | 0.00015 | −0.00232 | −0.00284 | −0.00343 |
Station . | Test . | SPI3 . | SPI6 . | SPI12 . | SPEI3 . | SPEI6 . | SPEI12 . |
---|---|---|---|---|---|---|---|
1 | P value | 0.54811 | 0.35387 | 0.29584 | 0.02910 | 0.02574 | 0.00916 |
Sen's value | −0.00035 | −0.00069 | −0.00085 | −0.00199 | −0.00239 | −0.00291 | |
9 | P value | 0.80148 | 0.91203 | 0.61914 | 0.01940 | 0.02033 | 0.01931 |
Sen's value | −0.00015 | 0.00007 | 0.02667 | −0.00120 | −0.00144 | −0.00152 | |
10 | P value | 0.80148 | 0.91203 | 0.61914 | 0.01940 | 0.02033 | 0.01931 |
Sen's value | −0.00015 | 0.00007 | 0.00032 | −0.00120 | −0.00144 | −0.00152 | |
22 | P value | 0.49384 | 0.31089 | 0.24660 | 0.02523 | 0.02227 | 0.00772 |
Sen's value | −0.00039 | −0.00074 | −0.00089 | −0.00196 | −0.00237 | −0.00289 | |
51 | P value | 0.14418 | 0.14863 | 0.16067 | 0.13768 | 0.13413 | 0.11378 |
Sen's value | −0.00107 | −0.00150 | −0.00173 | −0.00169 | −0.00193 | −0.00227 | |
69 | P value | 0.10999 | 0.09884 | 0.10818 | 0.03956 | 0.07002 | 0.08340 |
Sen's value | −0.00118 | −0.00174 | −0.00209 | −0.00220 | −0.00244 | −0.00251 | |
80 | P value | 0.00640 | 0.00478 | 0.00055 | 0.09978 | 0.06137 | 0.00932 |
Sen's value | 0.00156 | 0.00222 | 0.00286 | 0.00122 | 0.00154 | 0.00201 | |
83 | P value | 0.14418 | 0.14863 | 0.16067 | 0.13768 | 0.13413 | 0.11378 |
Sen's value | −0.00107 | −0.00150 | −0.00173 | −0.00169 | −0.00193 | −0.00227 | |
89 | P value | 0.14418 | 0.14863 | 0.16067 | 0.13768 | 0.13413 | 0.11378 |
Sen's value | −0.00107 | −0.00150 | −0.00173 | −0.00169 | −0.00193 | −0.00227 | |
94 | P value | 0.54648 | 0.82724 | 0.79423 | 0.00014 | 0.00000 | 0.00000 |
Sen's value | 0.00034 | 0.00013 | 0.00015 | −0.00232 | −0.00284 | −0.00343 |
DISCUSSION
Since the beginning of the 21st century, drought events in the UENB have increased. The variation of the SPI and SPEI were noticeable at various time scales. They reflected a change in the degree of dryness and wetness each month, especially after 1999, when the degree of dryness in some months increased. The droughts reflected by the SPI and SPEI were slightly different each month at various time scales (Liang et al. 2021). However, after 1999, the two indices showed almost similar degrees of dryness and wetness each month, showing that the SPI and SPEI consistently reflected drought periods.
The study witnessed moderate to extreme droughts in 1983/84, 1987, 1991/92, 1995/96, 1999, 2001, 2004/2005, 2006, 2007, 2008, 2009, 2011, 2014, 2016, 2017, 2019. Among these, the droughts in 2007, 2011, and 2016/2017 were identified as highly severe, which is consistent with studies by Karanja (2018) and Mwangi et al. (2014); National Drought Management Authority (NDMA) (2017) and FEWSNET (2020). The 2011 and 2016/2017 droughts were mainly considered national disasters by the Kenyan government, with the 2016/2017 drought causing a loss of up to US$ 12.1 billion in the country (NDMA 2017). Mbogo et al. (2014) also mentioned that the 2011 drought affected more than 13.3 million in Kenya, Ethiopia, and Somalia. As shown by both the SPI and SPEI, 2020 was a moderately wet year with an SPI/SPEA magnitude of approximately 1, which is consistent with the reports by FEWSNET (2020) that it was a relatively wet year.
Neither the SPI nor SPEI recognized 2010 as a drought year. Karanja (2018) and Odhiambo et al. (2018) also found that, in station 80, 2010 was a very wet year which caused floods in the region. In the study, the SPEI identified moderate droughts in 2008/2009, similar to the findings by Odhiambo et al. (2018), Mwangi et al. (2014), and NDMA (2017). While the SPEI indicated a moderate drought in this study in 2007, Karanja (2018) found the areas around stations 80 and 10 experienced a wet season in the same year. Moreover, in this study, the SPI also identified a moderately wet season in 2007. Considering that Karanja (2018) also used the SPI to identify drought, the study concludes that the index showed wetness because it did not account for PET. FEWSNET (2020) and Mbogo et al. (2014) also identified a severe drought in 2007, which led to the government of Kenya spending approximately 7 billion Kenya Shillings on relief food for affected people. The other drought years identified in the study that were identified by earlier studies are 1983/84, 1987/1988, 1991/92, 1995/96, 1999/2000, 2004/2005 (Huho et al. 2010; Mbogo et al. 2014; Karanja 2018; Odhiambo et al. 2018; FEWSNET 2020). The present study shows drought events have become extremely severe over the past 40 years. FEWSNET (2022) and Mwangi et al. (2014) also revealed an increase in drought severity and predicted more future drought events.
In this study, the meteorological drought events in the UENB were increasing. Like Pei et al. (2020) findings, the increase was much more evident after 1999. After 2004 the drought frequency increased further, with all stations indicating SPEI and SPI values between −1 and −3. The results showed that the intensity and severity of the droughts in the basin varied with stations 51, 69, 83, and 89. These stations are a tropical forest region, experiencing lower intensity and less severe droughts than stations 1, 9, 10, 22, 80, and 94, which are in arid and semi-arid regions.
Due to the highland-lowland topographical nature of the basin, PET increases with decreasing altitude while precipitation increases (Mutiga et al. 2010). This means that the Northern downstream part of the basin has a higher water deficit than the upstream, so it experiences extremely high drought events and is more vulnerable to drought. Consistent with this study, Omwoyo et al. (2017) also found that the amount of precipitation received upstream in the basin is higher than that received. However, both parts of the basin are experiencing increasing drought events and their agricultural and socioeconomic activities are affected mainly by droughts (Huho et al. 2010; Karanja 2018).
When analyzing the applicability of the SPI and SPEI, the two indices had their advantages and can generally monitor regional drought. However, they will always have differences due to climate change and the different climatic conditions in the various regions (Pei et al. 2020). In the study, the SPEI reflected more drought events and an increasing drought trend in all stations in all time scales. The SPEI detects more droughts in arid areas because of the PET parameter, which helps detect high evaporation caused by higher temperatures and low rainfall (Vicente-Serrano et al. 2010; Homdee et al. 2016). Because the SPEI accounts for both precipitation and evapotranspiration and as some studies suggest (Mehr et al. 2020; Pei et al. 2020), it is more suitable for drought monitoring in arid and semi-arid areas, especially in the context of global warming. For example, in station 10 on the Central side, 94 in the North, and 80 in the basin's East, the drought monitoring based on the SPI did not reflect severe and extreme drought, unlike the SPEI, which showed increasing trends at all time scales. This is because the stations are in a warmer and drier climate, and the high evapotranspiration with the rising temperatures in the areas (Omwoyo et al. 2017; Kimwatu et al. 2021) accounted by the SPEI. Even though the SPI described the drought variations, it did not consider the effect of evaporation on drought and could not be sufficient to detect the drought evolution.
When comparing the drought indices used in different studies, Kigumi (2014) investigated the performance of downscaled and original TRMM in identifying droughts. Both datasets showed similar characteristics to those identified by rain gauge data. However, the downscaled TRMM computed SPIs exhibited a stronger correlation with rain gauge data SPIs than the original TRMM. The study concluded that both downscaled and original TRMM could be utilized for meteorological drought monitoring. However, using them together would be more advantageous as they complement each other in detecting droughts.
In a different study conducted by Khan et al. (2020), agricultural and meteorological drought variations were examined using the SPI and Soil Saturation Index (SSI), while composite drought anomalies were analyzed using the Moisture–Precipitation Anomaly Index (MSDI). Additionally, the study investigated the relationship between MSDI and the Rainfall Anomaly Index (RAI), SPI, and SSI within the study area. The study observed that MSDI outperformed SSI and SPI in detecting droughts due to its consideration of precipitation and soil moisture, while SSI only considers moisture and SPI only precipitation. Thus, MSDI behaves like a multiscalar index, enabling earlier detection of drought onset compared to the other indices. However, in arid areas, all three indices (RAI, SSI, and SPI) showed greater consistency, and MSDI did not provide significant additional information about drought. This study explains why the SPEI performed better than the SPI.
Both this study, and those by Kigumi (2014) and Khan et al. (2020), demonstrate the importance of employing multiple drought indices instead of relying on one. This recommendation is because the various indices possess distinct characteristics, making them sensitive to different aspects of drought monitoring. As a result, combining multiple indices allows for a more comprehensive and nuanced understanding of drought conditions, leading to a more effective and accurate drought monitoring approach.
CONCLUSIONS
The study assessed the evolution of drought events from 1981 to 2020 in UENB using the SPI and SPEI at different time scales. In addition, the performance of the SPI and SPEI was also analyzed in assessing the temporal and spatial variation characteristics of drought and its consistency.
The temporal and spatial evolution of drought based on the SPI and SPEI had some differences at various time scales and some continuity similarities. For example, the SPEI identified extreme droughts in 1987, 2000, 2004, 2006, 2009, 2014, 2018, and 2019, while the SPI identified them in 2000, 2004, 2009, 2014, 2018, and 2019. The SPEI indicated drought occurrence in more months than the SPI in all the time scales. In the Northern part, the SPEI showed 10 more drought months than the SPI. The magnitude of drought shown by the SPEI was higher than that by the SPI, especially in the lowland areas in the Northern part of the basin. On average, the SPEI showed 7–15 more months with a drought magnitude of less than −1.5 compared to the SPI. However, both drought indices indicated increased drought occurrence from 1999 and showed a consistent drought from 2012 to 2019. The spatial variations maps indicate that in the UENB, the North, Northeast, Northwest, and Upper Central part of the basin is experiencing extreme to severe droughts. Although the indices showed a significantly high correlation, the SPEI performed better in reflecting the drought events than the SPI. The highest correlation was in stations closer to the river's source (stations 51, 83, and 89) between SPI 12 and SPEI 12 (876). The lowest was in station 10 between SPI 6 and SPEI 6 (476). Generally, there was a more negligible correlation between the stations on the lowlands (stations 10, 94, and 80) experiencing severe droughts. It was concluded that the results showed the dependency of the SPEI on evapotranspiration.
When analyzing the applicability of the indices in UENB, each had its advantage and both can generally assess droughts in the UENB region. The study shows that using multivariate index like the SPEI is more effective in detecting special and temporal drought conditions than using the stand-alone indicator of the SPI. However, the study recommends that even though the SPEI in the long term was selected as the suitable index in arid areas, using multiple indices to assess drought is necessary.
In alignment with the water-related disaster management goal (SDG-11.5), addressing the challenges posed by climate change and meteorological droughts is imperative. The findings indicate that climate change and meteorological droughts will result in the general reduction of the water in the Ewaso Ngiro River. The paper suggests implementing watershed management operations that prioritize soil and water conservation to counteract the negative impacts of climate change and human activities in the basin. Additionally, it advocates for establishing early warning systems to detect potential threats in advance. By doing so, the possible damages of environmental changes to the socioeconomic livelihoods of the communities can be minimized. Furthermore, future studies are recommended to examine the specific impacts of the meteorological droughts on the communities in the different regions of the basin, considering the varying drought characteristics in the basin regions. This holistic approach will contribute to a more comprehensive understanding of the communities' challenges and aid in developing targeted strategies to enhance resilience and adaptability.
ACKNOWLEDGEMENTS
This work was supported by the UKRI GCRF Equitable Resilience grant ES/T003006. The authors wish to thank the Jomo Kenyatta University of Agriculture and Technology (JKUAT) in Kenya and Cranfield University UK for their support during the study.
DATA AVAILABILITY STATEMENT
Data cannot be made publicly available; readers should contact the corresponding author for details.
CONFLICT OF INTEREST
The authors declare there is no conflict.