Assessment of Two Drought Indices to Quantify and Characterize Drought Incidents: A Case Study of the Northern Part of Burundi

Abstract

Droughts are natural catastrophes that cost the health and wealth of humans due to their harmful effects on the natural environment, ecology, hydrology, and agriculture in particular. Droughts are recurring incidents that last for prolonged periods of time in the northern part of Burundi. Despite the region being prone to drought and often suffering from dry conditions, drought has not been widely investigated. For the quantification and characterization of dryness conditions, this research utilized two drought indices, the Standardized Precipitation Evapotranspiration Index (SPEI) and the Standardized Precipitation Index (SPI), at 2-, 6-, 24-, and 48-month timescales, where 2-, 6-, 24-, and 48-months correspond to agricultural and hydrological droughts, respectively. The two drought indices were compared, and the difference between SPEI and SPI was illustrated by quantifying and characterizing drought incidents. The findings revealed that different types of droughts threatened the northern part of Burundi during the periods of 1993–2000 and 2002–2009. Both indices illustrated that 2005, 2006, and 2007 were extremely dry years. The drought incidents detected by the SPEI index were classified into moderate and severe categories, characterized by long duration and greater magnitude. In contrast, the drought incidents detected by SPI were classified into the “extremely dry” category, characterized by limited duration and lower magnitude but with higher intensities. This research highlighted that SPEI differs from SPI in quantifying and characterizing droughts and highly suggests the use of both SPEI and SPI when assessing droughts. The outcome of this study will be useful in drought prevention and mitigation strategies across Burundi, specifically for agricultural purposes.

1. Introduction

Drought is one of the greatest natural catastrophes, and its effects on both human societies and the environment are greater than any other natural disaster [1]. Drought harms humanity directly or unintentionally through food insecurity, conflicts among individuals or communities, economic losses, and diseases [2,3]. Since drought can develop gradually and softly [4,5], it can be difficult to identify its exact onset and end, unlike many natural disasters that hit the region and have immediate and noticeable consequences, such as floods, landslides, earthquakes, and windstorms. It is indispensable to capture the onset of drought within a week for better extension forecasts that describe changes in drought situations over a sub-seasonal timescale [6] because duration and severity are the characteristics of drought that significantly influence adaptability and preparation for drought [7]. Various drought indexes have been introduced in the academic literature [8,9], however, the Standardized Precipitation Index (SPI) and Palmer Drought Severity Index (PDSI) have been the most widely used worldwide [10,11]. SPI is a simplified index that requires one single parameter, precipitation, while the PDSI integrates different variables such as moisture supply, evaporation, precipitation, and runoff. In recent decades, the Standardized Precipitation Evapotranspiration Index (SPEI), pioneered by Vicente-Serrano et al. [12], has needed two input variables: precipitation, and temperature, which have received more attention in monitoring and the assessment of drought [13,14,15,16]. Droughts are a frequent and recurring event in Eastern Africa, particularly in Burundi, with harmful implications for agricultural production, socio-economic activities, and the natural environment [17]. Even though the northern and northeastern regions of Burundi are susceptible to droughts and continue to suffer from their consequences, the country lacks essential management tools for monitoring and predicting drought phenomena. A significant number of the region’s residents are entirely reliant on rain-fed agriculture for sustenance. In addition to agricultural irrigation water demand, the region’s water resources are under increased pressure because of rapid population growth, development projects, and the expansion of cities [18]. Therefore, it is crucial to have sufficient drought monitoring and analysis tools available that can accurately generate reliable information for the assistance of decision-makers at all levels and in all sectors of the country for better management of water resources. The northern part of Burundi has a considerable impact on Burundi’s economy due to its agricultural production of coffee and tea, which are the main income generators for many farmers, and the hydropower station at Rwegura, which provides electricity for many cities across Burundi. Many Burundians who live in the northern part, such as those who live in other parts of Burundi, rely on natural resources [19], especially water resources, for different purposes, for instance industrialization, agriculture, mining, and daily life. Even though the region is strongly dependent on water resources, the region is threatened by natural climatological variabilities, which are exacerbated by a changing climate, therefore intensifying the occurrence of extremely dry incidents and leading to the occurrence of different types of droughts. Many researchers have employed SPI across East Africa and Burundi in particular [20,21,22,23], and Uwimbabazi et al. [24] used both SPEI and SPI to examine changes in meteorological drought in Rwanda, and the overall analysis of the study highlighted the characteristics of SPEI and illustrated that Rwanda is less disposed to severe and extreme drought incidents than moderate drought events. In the meantime, other studies illustrated the influence of evapotranspiration on drought [25] and the robustness of the SPEI index [14]. Due to global warming and the coincidence of an increase in temperature [26], evapotranspiration significantly influences the conditions of drought. For quantification and characterization of drought, the incorporation of evapotranspiration might consider the impact of other parameters such as temperature, relative humidity, wind speed, and solar radiation. Then, using both evapotranspiration and precipitation parameters to quantify and characterize drought incidents may generate more reliable characteristics of drought compared to precipitation alone [27], since the formulations of the SPEI and SPI indexes differ [28,29]. This research aimed at comparing SPEI and SPI illustrates their performance in quantifying and characterizing droughts. This is the first study that evaluates the performance of drought indices in the assessment, quantification, and characterization of drought incidents over the northern part of Burundi. It will provide information that aids decision-makers in better managing water resources across the region and executing policies and strategies for drought risk management and drought mitigation.

2. Study Area and Methods

2.1. The Study Area’s Description and Datasets

Burundi is one of the landlocked countries covering 27,834 km², situated in the central-east of Africa that lies between 2.3° and 4.45° south latitude and 28.8° and 30.9° east longitude, sharing borders with Tanzania from the east to the south, Rwanda to the north, and the Democratic Republic of the Congo (DRC) to the west.

The northern part of Burundi plays a significant role in the economy of Burundi due to its agricultural production, especially the production of tea, coffee, and beans.

The northern part of Burundi is composed of many provinces (Figure 1). It is defined by latitude 2°30′–3°30′ S and longitude 29°30′–30°30′ E. The climate differs depending on the topography. As a result, the northern part, a mountainous and plateau region, benefits from the Congo–Nile Ridge zone rainfall, where precipitation ranges from 1400 to 1600 mm and above. The study domain benefits from a tropical climate because of Burundi’s special geographic location in the equatorial band. The Inter-Tropical Convergence Zone (ITCZ) crosses Burundi twice a year and is a major factor in determining precipitation seasonality [30]. Similar to the entire country of Burundi, the northern part of Burundi is normally characterized by four distinct seasons classified into A, B, C, and D seasons [31]. Season A runs for four months: September, October, November, and December (SOND), widely known as the short rain season. Season B runs for three months: March, April, and May (MAM), which is typically regarded as the longest rainy season. Both rain seasons are separated by season C, which runs from June to August (JJA), and season D, which runs in mid-January and February (JF), known as the shortest dry season. The region is extremely densely populated, and approximately 90% of the population settles in countryside areas involved in agropastoral activities [32]. The study used meteorological data (monthly average precipitation and maximum and minimum temperatures) ranging from January 1981 to December 2020 (

Table 1. Geographical location of meteorological stations and length of climatic records taken into consideration in this study.

Station	Latitude	Longitude	Elevation (m)	Type of Data	Length
Muyinga	−2.85	30.35	1756	Rainfall &Temperature	1981–2020
Kirundo	−2.58	30.11	1449	Rainfall & Temperature	1981–2020
Cankuzo	−3.31	30.53	1652	Rainfall & Temperature	1981–2020
Rwegura	−2.91	29.51	2302	Rainfall & Temperature	1981–2020

). The data was obtained from the Burundi Geographical Institute (IGEBU). Four meteorological stations (Muyinga, Kirundo, Cankuzo, and Rwegura) were purposely considered based on the availability of both precipitation and temperature data that are required for the computation of the SPEI and SPI indices to analyze and assess the characteristics of drought. Within these selected meteorological stations, no missing data could significantly disturb the analysis, and the missing data were filled under the linear mean. The IGEBU processed the initially observed data through standard quality control.

2.2. Climate and Topography

The research domain is located in the mountainous and plateau regions, which are among the highest altitudes in the country. It comprises the central plateau and the northern part of the Congo–Nile Ridge. The altitudes range from 2000 to 2300 m. The northern and eastern parts of the country are prone to drought in contrast to the western parts, where floods frequently occur [33]. In the area of focus, the average temperature varies between 11 and 21 °C. The region’s precipitation varies between 1300 and 1600 mm, and the dry season runs between three and four months. Figure 2 and Figure 3 present the annual mean precipitation and the maximum and minimum average temperatures, respectively.

3. SPEI and SPI Indices of Drought

The SPEI and SPI were compared and used to identify, characterize, and quantify drought incidents. The analyses were carried out at various timescales (2, 6, 24, and 48 months) from 1981 to 2020. The 2-, 6-, 24-, and 48-month timescales correspond to agriculture based on the fact that most cultivated plants need 2 to 6 months to fully develop and hydrological droughts, respectively [34]. Monitoring drought at different timescales enables the identification of short-term wet periods within long-term droughts and dry spells within long-term wet periods [35,36]. The SPEI and SPI were calculated utilizing the SPEI package found in the R software version 4.1.0 (Auckland, New Zealand, Robert Gentleman and Ross Ihaka) [12,37]. The SPEI package offers various options for generating SPEI and SPI values.

3.1. SPI

The SPI is a major meteorological drought calculation tool. Due to its simplicity and effectiveness in quantifying, identifying, and assessing meteorological drought, the SPI is a well-known and widely used indicator. To begin, SPI, as introduced by Mckee et al. [10], enables the detection and monitoring of various drought variables such as frequency, intensity, and duration over time in a specific region. The use of the SPI to measure drought changes linked to climate variations has gotten a lot of attention in many regions of the world [38,39,40,41]. It has been recommended by the World Meteorological Organization (WMO) as one of the best tools for monitoring drought incidents. It shall be recalled that a drought event begins when the SPI value reaches –1.0 and ends when the SPI value returns to positive again; therefore, monthly rainfall data is collected, and time series are created in order to calculate this index. In this study, SPI values are generated utilizing the following equation:

$$SPI\mathrm{value}=\frac{X_i-\overline{X}}{\sigma }$$

(1)

where $X_i$ represents precipitation of the particular month, $\overline{X}$ represents the longest mean precipitation, and σ represents the standard deviation for the determined period.

3.2. SPEI

The SPEI, an evolution of the well-established SPI recognized by the World Meteorological Organization (WMO), was first developed by Vicente et al. [12]. It is based on precipitation and potential evapotranspiration (PET) input data to calculate climatic water balance, which makes it different from SPI. The SPEI index has been widely used to monitor drought/flood conditions [14,42,43,44,45,46,47], and for instance, ref. [48,49] conducted a study to examine dry and wet conditions in Kenya from 1981 to 2016. Their study highlighted the good performance and effectiveness of SPEI in detecting dry and wet conditions. It is calculated using monthly precipitation and air temperature. The climatic water balance equation is as follows:

$$D_i=P_i-PE{T}_i$$

(2)

where $P$ and $PET$ denote rainfall and potential evapotranspiration, respectively, $D$ denotes the difference between $P$ and $PET$, and $i$ denotes the number of months counted. Several methods have been introduced to calculate PET and some researchers have compared the effectiveness of those methods [49,50]. The Hargreaves method, which does not have the fundamental limits of the Thornthwaite method and whose performance is nearly comparable to the criteria of the FAO Penman-Monteith (PM) methods, was utilized for the calculation of PET [37,42,51].

The determination of the cumulative difference between precipitation and PET at various timescales (2-, 6-, 24-, 48-month) is as follows:

$$\left\{\begin{array}{l}{{\displaystyle X}}_{s,j}^n={\displaystyle \sum _{t=13-n+j}^{12}{{\displaystyle D}}_{s-1},t+{\displaystyle \sum _{t=1}^j{{\displaystyle D}}_{s,j}\mathrm{if}j\le n}}\\ {{\displaystyle X}}_{s,j}^n={\displaystyle \sum _{s=j-n+1}^j{{\displaystyle D}}_{s,j}\mathrm{if}j\ge n}\end{array}\right.$$

(3)

where $X_{s,j}^n$ represents the accumulations of the difference between $P$ and $PET$ at the n-month timescale during the j-month of the s-th year and $D_{s, t}$ is the difference between monthly precipitation and PET in the t-month of the s-th year.

Due to the fact that the original may contain negative values, the SPEI utilizes the three parameters of log-logistic probability distribution proposed by [12]. For the data sequences of all timescales, the cumulative function of the log-logistic probability distribution $F\left(G\right)$ is presented as:

$$F\left(G\right)={\left[1+{\left(\frac{\alpha }{g-\gamma }\right)}^{\beta }\right]}^{-1}$$

(4)

where $\alpha , \beta , \mathrm{and} \gamma$ represent the scale, shape, and locality parameters, respectively. The calculation can be done utilizing the equations introduced by [12].

The probability of a definite $X_{s,j}^n$ value is $p$.

$$p=1-F\left(G\right)$$

(5)

If p ≤ 0.5,

$$w=\sqrt{-2lnp}$$

(6)

$$SPEI=W-\frac{C_0+C_{1}W+C_2{W}^2}{1+d_{1}W+d_2{W}^2+d_3{W}^3}$$

(7)

If p > 0.5

$$w=\sqrt{-2ln\left(1-p\right)}$$

(8)

$$SPEI=W-\frac{C_0+C_{1}W+C_2{W}^2}{1+d_{1}W+d_2{W}^2+d_3{W}^3}$$

(9)

where $C_0$ = 2.515517, $C_1$ = 0.802853, $C_2$ = 0.010328, $d_1$ = 1.432788, $d_2$ = 0.189269, and $d_3$ = 0.001308 [12,52,53].

The generated values of SPEI and SPI are categorized in

Table 2. Classification of dry and wet states based on drought indices values [10].

Categories	SPI, SPEI
Extremely wet	$\ge$2
Severely wet	1.5–1.99
Moderately wet	1–1.49
Near normal	−0.99 to −0.99
Moderately dry	−1 to −1.19
Severely dry	−1.5 to −1.99
Extremely dry	$\le -$2

and are utilized to evaluate the condition of dryness across the study area in terms of drought characteristics such as frequency, durability, severity, and intensity. The number of consecutive months during which SPEI and SPI values are equal to or less than a thresholding level. In this research, the criterion for the indication of dry incidents is when both SPEI and SPI values ≤ −1. See Table 2.

The severity (S) is the accumulative sum of the index values across the time of the event. The intensity (I) of an incident is obtained from the division of its severity by the time. Incidents with short periods and significantly larger severities will always have higher intensities. The calculation is as follows:

$$S={\displaystyle \sum _{i=1}^{Time}Index}$$

(10)

$$I=\frac{Severity}{Time}$$

(11)

4. Results and Discussion

4.1. Presentation of Temporal Variability of Drought Incidence at Different Timescales

In order to illustrate historical variabilities of drought incidents at various timescales (2-, 6-, 24-, and 48- months) in the northern part of Burundi, the two drought indices, SPI and SPEI values, were calculated based on the meteorological station’s records for Muyinga, Kirundo, Cankuzo, and Rwegura, as indicated in Figure 4A,B. According to the findings, the two indices demonstrate similarities in patterns of change over timescales, although they are at variance in terms of persistence and magnitude of drought. Considering that most cultivated crops need 2 to 6 months to develop completely, an accumulation of at least 2 months of water deficit during the crop growth stage will negatively affect crop production, resulting in an agricultural drought. Given that drought expands over time, it is reasonable to assume that the prolonged periods of drought at longer timescales result from the cumulative effects of predecessors’ water shortages over time, which may be exacerbated by the persistent and continuous absence of precipitation. Furthermore, drought occurs when seasonal precipitation patterns shift. For example, the rains are hampered, resulting in threatening activities that are heavily reliant on the start of precipitation incidents. When comparing SPEI and SPI, distinctions were noticed in some years and across timescales. The longest drought incidents that occurred were detected under SPEI at all time scales. The drought lasted longer as the number of drought months increased annually over the time series. For several years, the region had already experienced drought conditions according to SPI, and the drought magnitude was intensified when identified by SPEI. The increased magnitude of droughts generated by the SPEI can sometimes be explained by the fact that the SPEI incorporates evapotranspiration, and evapotranspiration places a demand on water availability. Its effects are experienced mostly in water-scarce circumstances [54,55]. In Burundi, the mountainous regions experience natural climatic changes, which can be the provenance of drought variability. Drought occurs more frequently on a shorter timescale, typically one month of dominating water deficit for a meteorological drought. In contrast to a meteorological drought, a hydrological drought necessitates a prolonged period of water deficit or a considerable decline in water storage to arise. Thus, meteorological droughts on a 2-month timescale are most frequent, followed by agricultural droughts on 6-month timescales and hydrological droughts on 24- and 48-months. Still, at the higher timescale of 24-months, the drought persists longer, and the magnitude expands. On the other hand, on a 48-month timescale, the drought months are more severe and extremely dry, but with a lower magnitude.

4.2. Characteristics of Drought and Date of Occurrence Patterns

The SPEI and SPI indices were utilized to detect drought duration, severity, intensity, and date of occurrence over the past 40 years in the northern part of Burundi. Those characterizations of drought are presented in

Table 3. The characteristics and categories of some main dry incidents (SPI ≤ −1) in the study domain.

Stations	SPI Index	Date of Occurrence	Duration (Month)	Severity	Intensity	Category
Muyinga	SPI 2	October 1998–January 1999	4	−8.00	−2.00	Extreme
		June 2002–Sepember 2002	4	−4.85	−1.21	Moderate
		July 2010–October 2010	4	−5.06	−1.26	Moderate
	SPI 6	December 1993–July 1994	8	−13.53	−1.67	Severe
		August 1998–October 1999	15	−25.72	−1.72	Severe
		April 2000–October 2000	7	−10.83	−1.55	Severe
		August 2010–December 2010	5	−7.55	−1.51	Severe
		April 2016–March 2017	12	−17.50	−1.46	Moderate
	SPI 24	November 1999–August 2001	22	−52.80	−2.4	Severe
		February 2010–October 2010	9	−12.30	−1.37	Moderate
		November 2016–January 2018	15	−19.73	−1.31	Moderate
	SPI 48	November 2001–August 2003	22	−45.43	−2.06	Severe
		November 2016–February 2018	16	− 24.15	−1.51	Severe
Kirundo	SPI 2	August 1992–May 1993	10	−14.22	−1.42	Moderate
	SPI 6	October 1993–January 1998	52	−73.70	−1.42	Moderate
		November 1992–January 2000	87	−132.50	−1.53	Severe
		January 2001–June 2001	6	−7.07	−1.20	Moderate
	SPI 24	October 1993–January 2001	88	−136.13	−1.55	Severe
	SPI 48	November 1994–April 2002	90	−134.65	−1.50	Severe
Cankuzo	SPI 2	September 1993–November 1993	3	−5.05	−1.68	Severe
		August 2000–October 2000	3	−5.00	−1.66	Severe
		September 2005–December 2005	4	−6.82	−1.70	Severe
		January 2012–March 2012	3	−4.47	−1.50	Severe
	SPI 6	October 1993–December 1993	3	−4.90	−1.63	Severe
		November 1998–July 1999	9	−18.00	−2.00	Extreme
		April 2000–November 2000	8	−15.24	−1.90	Severe
		June 2005–February 2006	9	−15.44	−1.72	Severe
		January 2011–May 2011	5	−8.40	−1.68	Severe
		April 2014–September 2014	6	−8.50	−1.42	Moderate
	SPI 24	December 1999–March 2002	28	−52.76	−1.88	Severe
		January 2005–October 2006	22	−36.55	−1.66	Severe
		January 2015–December 2015	12	−13.82	−1.15	Moderate
	SPI 48	November 2001–April 2003	18	−32.50	−1.81	Severe
		August 2005–December 2007	29	−40.00	−1.40	Moderate
		October 2016–February 2018	17	−22.81	−1.34	Moderate
Rwegura	SPI 2	July 2002–November 2002	5	−6.54	−1.30	Moderate
		September 2005–February 2006	6	−11.00	−1.83	Severe
	SPI 6	May 2000–October 2000	6	−11.08	−1.85	Severe
		July 2002–March 2003	9	−12.00	−1.33	Moderate
		September 2005–June 2006	10	−24.70	−2.47	Severe
	SPI 24	May 1994–November 1994	7	−7.72	−1.10	Moderate
		January 2004–November 2004	11	−17.35	−1.58	Severe
		September 2005–January 2008	29	−64.00	−2.20	Extreme
	SPI 48	November 2005–December 2009	50	−108.00	−2.10	Extreme

and

Table 4. The characteristics and categories of some main dry incidents (SPEI ≤ −1) in the study domain.

Stations	SPEI Index	Date of Occurrence	Duration (Month)	Severity	Intensity	Category
Muyinga	SPEI 2	August 1982–October 1982	3	−4.66	−1.55	Severe
		January 1984–March 1984	3	−4.67	−1.55	Severe
		October 1998–January 1999	4	−8.11	−2.03	Extreme
		April 2000–August 2000	5	−7.30	−1.46	Moderate
	SPEI 6	November 1983–July 1984	9	−14.32	−1.60	Severe
		August 1998–October 1999	11	−19.36	−1.76	Severe
		April 2000–October 2000	7	−11.70	−1.67	Severe
		August 2010–December 2010	5	−6.71	−1.34	Moderate
		May 2016–April 2017	12	−16.06	−1.33	Moderate
	SPEI 24	November 1999–August 2001	23	−48.21	−2.10	Extreme
		December 2005–May 2007	18	−27.32	−1.52	Severe
		June 2017–January 2018	8	−10.86	−1.35	Moderate
	SPEI 48	March 2001–May 2003	23	−49.40	−2.15	Severe
		January 2006–December 2008	36	−43.00	−1.20	Moderate
Kirundo	SPEI 2	July 2010–December 2010	6	−8.13	−1.35	Moderate
		June 2014–August 2014	4	−5.48	−1.37	Moderate
		July 2015–October 2015	4	−5.43	−1.35	Moderate
		June 2016–October 2016	5	−7.67	−1.53	Moderate
		April 2020–July 2020	4	−6.00	−1.50	Moderate
	SPEI 6	August 2010–November 2011	16	−22.09	−1.40	Moderate
		June 2016–December 2017	19	−25.70	−1.35	Moderate
	SPEI 24	May 2009–October 2014	61	−62.72	−1.03	Moderate
		September 2016–August 2018	24	−32.26	−1.34	Moderate
	SPEI 48	November 1994–April 2002	105	−120.9	−1.15	Moderate
Cankuzo	SPEI 2	November 1998–October 1999	4	−6.72	−1.68	Severe
		April 2010–October 2010	7	−10.87	−1.55	Moderate
		July 2005–December 2005	6	−10.36	−1.73	Severe
	SPEI 6	January 1999–July 1999	7	−13.20	−1.88	Severe
		February 2000–December 2000	11	−18.04	−1.64	Severe
		February 2005–April 2006	15	−24.38	−1.63	Severe
		August 2008–December 2008	5	−7.72	−1.54	Severe
		January 2011–May 2011	5	−7.10	−1.42	Moderate
	SPEI 24	December 1999–March 2002	28	−44.16	−1.60	Severe
		May 2004–December 2006	32	−50.60	−1.60	Severe
		November 2008–March 2010	17	−21.30	−1.25	Moderate
	SPEI 48	January 2002–March 2004	27	−35.00	−1.30	Moderate
		January 2005–November 2009	61	−83.00	−1.40	Moderate
		January 2011–June 2011	6	−7.02	−1.20	Moderate
Rwegura	SPEI 2	September 1993–November 1993	3	−5.20	−1.73	Severe
		September 2005–April 2006	8	−14.00	−1.75	Severe
	SPEI 6	September 1993–March 1994	7	−11.02	−1.57	Severe
		May 2000–October 2000	6	−11.00	−1.83	Severe
		July 2002–March 2003	9	−12.00	−1.33	Moderate
		June 2005–December 2006	19	−38.00	−2.00	Extreme
		March 2011–July 2011	5	−6.00	−1.20	Moderate
	SPEI 24	April 1994–August 1995	17	−20.30	−1.20	Moderate
		May 2000–October 2000	6	−7.32	−1.22	Moderate
		December 2003–December 2007	49	−81.30	−1.66	Severe
	SPEI 48	October 1996–August 1997	11	−12.43	−1.13	Moderate
		November 2005–December 2009	51	−89.10	−1.75	Severe

based on 2-, 6-, 24-, and 48-month timescales. SPEI- and SPI-series permitted to identify in which months or years the study domain experienced moderately, severely, or extremely dry incidents. The overall analysis demonstrated that both categories of drought had struck the region. Based on the SPEI and SPI on a two-month timescale, however, the SPEI generated many drought months. Within the study period, in Kirundo station, SPEI generates 102 dry months, all classified as moderately dry. The years 1993, 1997, 2004, 2005, 2006, and 2007 are recorded as dry periods. SPEI recorded two dry periods of long duration but low intensity, whereas SPI recorded one dry period of limited duration but high intensity. SPI generates 57 dry months: 31 moderately dry, 16 severely dry, and 10 extremely dry; where the years 2005 and 2006 were identified as dry years. According to the drought events generated by the two drought indices, the SPEI index can more accurately assess meteorological drought compared to the SPI index. According to the results generated by SPEI-2, the years 1993, 2000, 2003, 2005, 2006, 2007, and 2008 were dry. On the other side, SPI-2 illustrated that 1993, 2000, 2005, and 2006 were extremely dry. SPI-2 presented two extremely dry incidents: one of three consecutive months occurred from September 1993 to November 1993 with an intensity of 2.19, and the second appeared from October 2005 to April 2006, seven consecutive dry months with an intensity of 2.19. These two incidents highly affected the agriculture sector. Since September to December and March to May are the rainy seasons in northern Burundi, the second drought incident affects two successive crop-growing seasons. The revelations of both SPEI and SPI on 2- and 6-month timescales have some correspondence with the previous highlights of the study conducted by Nkunzimana et al. [27]. This assistance identified agricultural droughts that resulted in crop failure and livestock losses, particularly in northern Burundi, resulting in considerable family hunger. Thousands of people got sick, and some died because of thirst and hunger-related diseases and there was the immigration of people moving from the provinces that are located in the northern part of Bujumbura. It also strengthens the fact that drought indices on a 2-month timescale characterize a perfectly agricultural drought. With SPEI-6 detection of moderately and severely dry years, the years 1993, 1994, 2000, 2002, 2003, 2005, and 2006 were recorded as extremely dry years, and the dryness incident from June 2005 to December 2006 lasted for 19 months with an intensity of 2.00. The analysis of SPI-6 at Rwegura station showed moderately dry periods during 1999 and 2002; the year 2000 was identified as severely dry, while the years 2005 and 2006 were extremely dry, with a drought incident occurring from September 2005 to June 2006, ten consecutive months with a drought intensity of 2.47 for the station of Rwegura. The analysis of the SPEI and SPI at higher timescales is crucial to identifying the occurrence of hydrological drought. On a 24-month timescale, they generated almost the same number of drought months, for moderately dry and extremely dry conditions, respectively. On 24- and 48-month timescales, both SPEI and SPI indicated that the years 1994, 2003, 2004, 2005, 2006, 2007, 2008, and 2009 were extremely dry. The SPEI index generated more moderately and severely dry months, while SPI detected more extremely dry months with a higher intensity than SPEI. Table 3 and Table 4 show how the two indices differ in characterizing drought incidents.

4.3. Evaluation of Drought on Yearly Scales with SPEI and SPI Indices

The number of drought months annually, together with the drought classifications for the meteorological stations of Rwegura, Muyinga, Kirundo, and Cankuzo, are illustrated for the period under investigation (1981–2020) for both the SPEI index and the SPI index. Significant meteorological, agricultural, and hydrological droughts were noticed in 1993–1994, 1998, 2000, and 2002–2009 years, and Figure 5 shows the quantified drought months during the study period. Dryness incidents that struck the northern part of Burundi during the 2000–2010 decades were also identified in bordering counties, specifically Tanzania [56] and Rwanda [57]. The intensification and prolonged drought incidents across the study domain highly affected Burundi’s economy. A severe hydrological drought caused a severe water depletion of the Rwegura hydroelectric reservoir. The Rwegura hydropower station generates over 30% of Burundi’s electricity [20]. With water shortages in the reservoir leading to an insufficiency of electricity in many cities, most of the societal and economic activities that require electricity were paralyzed.

4.4. The Effectiveness of SPEI and SPI in Identifying Drought Months

The SPEI index identified more drought months over the study domain than the SPI index under both moderately and severely dry classes for all the various timescales used for this research, except for the station of Kirundo, where there was no drought incident that was classified as extremely dry for all timescales, as shown in Figure 6. The findings demonstrate that, even though lack of rainfall is the principal cause of droughts, the consequence of temperature throughout the climatic water balance plays a significant role in drought determination. When only precipitation was considered, more drought months were categorized as “extremely dry” than when potential evapotranspiration (PET) was incorporated, especially for the station of Rwegura under a 48-month timescale. Generally, Figure 6 indicates the number of drought months detected at all timescales in the northern part of Burundi during the study period of 1981–2020.

5. Conclusions

This paper compared the SPEI and SPI indices in quantifying droughts characterizing the extremely dry incidents in the northern part of Burundi based on 2-, 6-, 24-, and 48-month timescales for the period 1981–2020 for both SPI and SPEI. As pointed out by the various timescales, the two drought indices captured historical variations of droughts, and they could detect various types of droughts. Evidently, SPEI identified the occurrence of droughts characterized by a long duration and increased magnitude but with lower intensity. This might be due to the fact that global warming increases the level of potential evapotranspiration and changes precipitation patterns, thereby raising the evaporative water demand. All of these leads to a shortage of water and increases the frequency of extreme climate change-related incidents such as drought. The northern part of Burundi is vulnerable to climate variations, soil degradation, extensive deforestation, and the hilly landscape with its dominant steep slopes. The SPEI was thought to be the best index due to its capability to consider the potential effects of climate change [58]. On the other hand, taking into consideration only precipitation, SPI detected more extreme drought incidents than SPEI. In addition, the considerable difference is evident basically in the higher timescales. It is undeniable that rainfall plays a significant role in determining droughts. A temperature rise has essential effects on the severity of droughts and the amount of soil moisture available, which in turn affect agricultural droughts. The SPEI index is a powerful tool for determining the historical pattern of dry conditions in the northern part of Burundi.

Both the SPEI and SPI indices can quantify droughts and identify the onset and end of dryness incidents. Notwithstanding, this study’s aim was to determine how the SPEI differs from the SPI in the quantification and characterization of drought incidents. All of them can detect different types of droughts, but with different characteristics. SPEI detected more drought incidents characterized by long duration and greater magnitude but with lower intensity, while SPI detected drought incidents characterized by limited duration and lesser magnitude but with higher intensity. Therefore, SPEI is more accurate for quantifying droughts than SPI, while SPI is more accurate for indicating drought severity. Thus, this study suggests that: more attention should be paid when applying one index like SPI; with caution, SPEI can be effective in climatological conditions where the influence of evapotranspiration is noticeable, moreover, when assessing drought, the use of both SPEI and SPI indices is highly suggested. Furthermore, this study will serve as a reference for the studies ahead for as long as this study recommends using multiple indices in characterizing dryness and wetness conditions.