Hydro-meteorological drought predictions and trend analysis for ungauged watersheds in the Upper Blue Nile basin, Ethiopia, under future climate change impact scenarios

The Earth is currently under increased pressure from climate change to start experiencing catastrophic disasters. The effects of climate change on water, such as floods and droughts, are expected to become more noticeable in the future years (Weir et al. 2017). Regions are already adapting, employing more natural, environmentally friendly tactics to mitigate the effects of such occurrences and managing water in more resource-smart, environmentally friendly ways to assist in dealing with droughts (Dikici 2020). According to the IPCC (2021) report, the rise of sea levels over the globe significantly causes extreme drought and flooding. Hence, since nature is composed of multiple interconnected systems, the scientific community still needs a comprehensive assessment of those causes.

The degree of dryness (concerning a ‘normal’ or average amount) and the length of the dry period are typically used to describe drought. Given the wide regional variations in the atmospheric conditions leading to precipitation deficits and defined as meteorological drought (Tijdeman et al. 2018), droughts can be divided into four categories. Among them, major hydrological and meteorological drought periods, such as those that occurred in East Africa in 2011, the Central Great Plains in the United States in 2012, and California in 2012–2015, have drawn a lot of attention recently because of their profound effects on agriculture and society (Dutra et al. 2014; Hao et al. 2018; Liu et al. 2023). Hydrological drought is the inheritance, continuation, and development of meteorological drought and is also an intermediate link that leads to agricultural and socioeconomic drought (Liu et al. 2023). Since the majority of African nations rely on rain-fed agriculture, they are particularly vulnerable to the threats of climate change, such as floods and droughts (Kalungu et al. 2013). For instance, it is anticipated that between October and December 2022, regions of Ethiopia, Kenya, and Somalia saw their fifth consecutive failing rainy season, putting those who are already experiencing droughts at urgent risk of hunger (Sandstrom & Juhola 2017).

Due to erratic rainfall and climate change, seasonal droughts have long been a prevalent occurrence in Ethiopia. Ethiopia is a country that is vulnerable to drought, particularly in the Somalia region, the eastern portions of the Oromia region, the Upper Blue Nile (UBN) basin, the Northern Tigray, several Amhara regions (like South Wollo, North Wollo, and South Gondar) and Afar among others (Loakes 2016; Bayissa et al. 2018; Melaku 2020). Taye et al. (2020) show that, over 38 years, the severe or extreme meteorological drought frequency varied between 7% and 11% in the UBN highland, midland, and lowland regions using spatial–temporal analyses of Standardized Precipitation Index (SPI) values. Bayissa et al. (2015) also emphasize that seasonal or yearly meteorological drought spells occurred in the UBN basin in 1978/79, 1984/85, 1994/95, and 2003/04. Additionally, Wubneh et al. (2023) show that a selected gauged watershed above Lake Tana experiences drought with a meteorological frequency of 20.69% and a hydrological frequency of 17.24%. Analyzing the ungauged watershed drought is significantly crucial for minimizing the damage of severe drought. Wu et al. (2022) proposed three approaches for diagnosing and forecasting hydrological drought conditions using ungauged watersheds that take in river water, and their utilization can facilitate effective drought responses, based on drought forecasting. Additionally, Rhee et al. (2020) also detected hydrological droughts in ungauged areas, using percentiles from remotely sensed key hydro-meteorological variables, and the output shows there are hydrological droughts on various time-scales in ungauged watersheds of the study area. Hence, there is a huge gap in identifying the drought of the ungauged watersheds on the sub-basin Lake Tana using regionalization approaches. Over Lake Tana, Kim & Kaluarachchi (2008) estimate the usage of parameter estimation and regionalization methodologies for ungauged watersheds in the UBN river basin, and provide reasonable results. According to Wubneh et al. (2022), the UBN basin, including the Lake Tana sub-basin, comprises 18 watersheds, four of which, Gilgel Abay, Gumara, Megech, and Ribb, are classified as ungauged watersheds. It is imperative to analyze these watersheds to forecast impending hydrological and meteorological droughts in the ungauged sub-watersheds of the Lake Tana sub-basin.

Different scholars have studied drought analysis in Ethiopia over the different watersheds; for instance, Wubneh et al. (2023) over the Lake Tana sub-basin using selected watersheds and using hydrological and meteorological drought. Mekonen et al. (2020) also predict the spatial distribution of drought over Ethiopian highlands, but there is a huge gap in the analysis of the watersheds without gauged measurement, especially in the UBN basin. Nannawo et al. (2022) indicate that, to analyze one of Ethiopia's catastrophic events, the dynamics of hydroclimatic factors from the ungauged watershed greatly contribute to the change in dynamics of the one basin. Therefore, to determine the severity of the drought in the UBN basin, an analysis of a particular ungauged watershed is necessary to understand the dynamics of the drought.

Additionally, the monthly frequency performance of the global climate model (GCM) over the UBN river basin and the baseline hydrological drought monitoring trend over the Abay basin are also estimated by Abera Tareke & Awoke (2022) and Bokke et al. (2017), respectively. Ongoma et al. (2019) also analyzed 22 CMIP5-GCMs for East African historical precipitation simulations. Hence, the main objective of this study is to evaluate the future meteorological and hydrological drought in the Lake Tana sub-basin's ungauged watersheds (Gilgel Abay, Gumara, Megech, and Ribb) using GCM temperature and precipitation data from the Coupled Model Inter-comparison Project 5 (CMIP5). Additionally, the SPI and Reconnaissance Drought Index (RDI), under two slice period biannual (SPI6 and RDI6) and annual time-scales (SPI12 and RDI12) are used for identifying the meteorological drought. Streamflow Drought Index (SDI) under the annual time-scale (SDI12) is also used to identify hydrological droughts. Additionally, it finds Representative Concentration Paths (RCPs) in various scenario data, such as RCP 4.5 and RCP 8.5.

Temperature and precipitation data, which are necessary for bias-correction climate models, are primarily obtained from the National Meteorological Agency's Bahir Dar division for this study. The data was available from 1976 to 2005, but the actual analysis in the study area utilized data between 1988 and 2005 to align with historical GCM data from four meteorological station sites on the sub-basin (Chawit, Hamusit, Zegie, and Addis Zemen) (Figure 1(c)). From the Ministry of Water Energy and Irrigation, Addis Ababa, Ethiopia the observations of hydrological data (flow) were gathered. The data gathered on a daily frequency from 1988 to 2005 was used to validate and calibrate Hydrologiska Byråns Vattenbalansavdelning (HBV) models through observed flow.

Future climate projections are usually created by combining or ensembling a variety of GCM data and scenarios to minimize the uncertainty of the model (Christensen & Christensen 2007). The ensemble GCM provides us with high-quality output findings by reducing the uncertainty of the two models (Wang et al. 2020). Other ensemble methods for GCMs, such as those employed by Chen et al. (2011), Dobler et al. (2012), and Dong et al. (2021), take into account the range of each GCM group's averages over several modeling phases. For analyzing GCM models the daily frequency data using baseline meteorological analysis is necessary, and for conducting hydrological drought, similarly, daily hydrological data is also crucial to comprehend the correlation between daily estimated streamflow and actual regionalized ungauged watershed data.

Wubneh et al. (2022, 2023) examined how Lake Tana's water balance is affected by climate change in the UBN basin using a regional CORDEX climate model powered by GCM–regional climate model (RCM) ensembles under the Eastern Africa domain AFR-44 to analyze the frequency of droughts in the Lake Tana sub-basin. These data are also used in this study because they show reasonable drought output in the basin as well as over the Awash basin watershed (Alemu et al. 2023), and the data are found on the portal (https://aims2.llnl.gov/search). Table 1 displays the different candidate GCM models that could be used to research the consequences of climate change. The GCM models were provided from the portal through a 50 × 50 km grid (Ahmadalipour et al. 2017). The Lake Tana sub-basin's grid cells were used to extract the daily precipitation, maximum temperature, and minimum temperature between 1980 and 2100. The scenario data for this study included a total of two time-slices: the near future (2020–2049) and the mid-future (2050–2079) for 29-year intervals. Moreover, the RCM ‘RCA4’, the RCP of daily time-frequency data, and the project degree location CMIP5 (AFR-44) as historical experimental data were included as criteria for potential GCMs (Table 1) (Wubneh et al. 2022).

According to the IPCC (2014) and IPCC (2021) reports, assessment of the four RCP2.6, RCP4.5, RCP6, and RCP8.5 emission scenarios might be used, since it is expected that their near-range and mid-level climate scenarios will span a reasonable range in climatic analysis by 2079 (Fentaw et al. 2018). Additionally, Alemu et al. (2022) and Wubneh et al. (2023) also used those scenarios (RCP4.5 and RCP8.5) for analysis of water level and hydrological impact of climate change over the Awash and Lake Tana sub-basin, respectively. Hence, only the emission scenarios from RCP4.5 (Thomson et al. 2011) and RCP8.5 (Hausfather & Peters 2020) were used for a better range of output in this study.

Climate models are somehow biased in both the global circulation model (GCM) and the regional circulation model (RCM) (Laux et al. 2021). Hence, before being used as an input for regional climate impact analysis, they need to be corrected. Depending on the performance evaluation criteria, the best correction techniques for biased temperature and rainfall adjustment are categorized; each technique performs differently (Gumindoga et al. 2019). Quantile mapping normal distribution function (Maraun 2013) and power transformation (Ghimire et al. 2019) were used in this investigation to correct rainfall and adjust the temperature, respectively, and were similarly used in the previous study (Wubneh et al. 2022). The performance of these correction methods is evaluated by their capacity to replicate temperature, precipitation, and streamflow simulated by the hydrological model, which is powered by biased-corrected climate GCM model simulations. The minimum temperature, maximum temperature, and biased adjusted precipitation model performance were used to assess how well the methods for simulating streamflow using the selected hydrological model performed.

For simulation and prediction of streamflow and conducting hydrological drought, the Hydrologiska Byråns Vattenbalansavdelning (HBV) semi-distributed model was used. A lumped bucket-type catchment model with little forcing input is needed for simulation of temperature compared with other models (Bergström 1995; Alemu et al. 2022) and it consists of only daily temperature, precipitation data, the elevation of the watershed, and monthly evapotranspiration over the long term determined for estimating (Bergström 1995; Ragab et al. 2010).

The HBV-96 model requires calibration of its parameters, much like other conceptual models of a like kind. HBV models are typically calibrated using both automatic and manual methods (Ouatiki et al. 2020). Hence, conducting manual calibration gives the best out-parameter option for the hydrologic result (Ouatiki et al. 2020; Sahraei et al. 2020). One way to calibrate the model and obtain the optimal set of parameters for this model is through manual calibration.

For the simulation of the Lake Tana sub-basin ungauged catchment (Gilgel Abay, Megech, Gumara, and Ribb) two-thirds of the annual data are used for calibration, while one-third is used for validation. The results concerning the metrics used to assess the model's performance were then displayed by Wubneh et al. (2022), and these were done for 18 ungauged and gauged catchments. Following that, optimizing the baseline and future flow parameters of each model, the most sensitive and moderate models were examined (Wubneh et al. 2022). Finally, the result was examined for the future streamflow of the Lake Tana sub-basin using a semi-distributed hydrologic model (HBV).

To begin, establish a strong association between the optimized model parameters and the PCCs including meteorological data and hydrological data, which is statistically significant and beneficial from a hydrological standpoint for the ungauged catchment's streamflow estimation. The t-test was utilized to ascertain the significance following the optimization of the simple relationship between the model parameters (Allen 1997). The critical value of the t-test must be calculated to test the hypothesis, and since experiments typically use either a 5% (0.05) or 10% (0.1) significance level, the critical value for this study is set at t_cr = 2.132 (critical value from the t-distribution table) (Smucker et al. 2007). Wubneh et al. (2022) found out the correlation coefficient is higher than 0.73, t_cr = t_cor, r = 0.73, among optimized ungauged watersheds. Hence, the value is significant at a 95% significance level.

There are many methods for monitoring droughts that have been described; they can be broadly divided into four categories: meteorological, agricultural, hydrological, and socioeconomic. The first through third strategies concentrate on defining drought as a physical event. The fourth technique examines drought in terms of water supply and demand, and following that, the effects of water shortages as they extend through socioeconomic sectors (Ruwanza et al. 2022). It is difficult to compare the efficacy of one indication to that of others due to the complexity of drought (Kchouk et al. 2021). Tracking meteorological drought can be done using straightforward to complex indicators (Yacoub & Tayfur 2017). Due to their feasibility and ability to capture drought, the SPI and the RDI are both helpful monitoring indicators for the Abay River basin (Abera Tareke & Awoke 2022). The analysis of Wubneh et al. (2023) and Alemu et al. (2023) using those methods on the study area for gauged watersheds gives reasonable output. Hence, the same drought index was also used for this study.

For hydrological drought analysis the Streamflow Drought Index (SDI) (Jahangir & Yarahmadi 2020), the Palmer Hydrology Drought Index (PHDI) (Shin et al. 2018), and the Surface Water Supply Index (SWSI) (Wambua 2019) are a few well-known drought indices. PHDI and SWSI need further details on groundwater levels, reservoir water levels, and soil moisture. So, in this work, the hydrological drought phenomenon was described using the most straightforward and typical SDI.

Consequently, SPI, RDI, and SDI were used in this work to assess the historical hydrological and meteorological droughts in the Lake Tana sub-basin in Ethiopia. Annual and biannual time-slices of drought frequency analysis are mostly better able to capture drought compared with other drought slices (Taye et al. 2020; Maru et al. 2022). Furthermore, both Wubneh et al. (2023) and Alemu et al. (2023) used biannual (SPI6 and RDI6) and annual (SPI12, RDI12, and SDI12) time-frames to analyze the drought and concluded that both frequency periods were mostly able to identify the magnitude of the drought. Hence, in this study, the biannual and annual hydrological droughts were analyzed using Excel functions for the study area, particularly in the Lake Tana sub-basin ungauged watersheds.

The performance of both the Max Planck Institute for Meteorology Earth System Model (MPI-M-MPI-ESM-LR) and Interdisciplinary Research on Climate (MIROC-MIROC5) model was reasonable after bias correction on the study area (Wubneh et al. 2023). The result depicts that RVE becomes 6.44 and NSE shows 0.7, following by Wubneh et al. (2022). The uncertainty of the GCM models was minimized by ensembling or averaging those performed and conducted for further drought analysis.

For the near future (2020–2049) and mid-future (2050–2079) periods of the meteorological drought, four ungauged meteorological stations (Gilgel Abay, Gumara, Megech, and Ribb) were utilized in the Lake Tana sub-basin. We consider the analysis carried out with SPI and RDI for the two time-scales SPI6 and SPI12 and the RCP 4.5 and RCP 8.5 scenarios.

The outcome demonstrates that under both RCPs at a biannual time-scale, 16.38% of the total drought frequency throughout the near timeframe will be experienced on the sub-basin. From those, the maximum drought levels, 8.62% and 7.76%, respectively, occurred at a moderate level under RCP4.5 and RCP8.5 (Table 3A). Furthermore, at this time, the ungauged Megech under RCP8.5 and the ungauged Ribb under RCP4.5 watersheds experienced extraordinary droughts in the years 2046 and 2030, with ranges of 3.2 and 3.23, respectively (Figure 2(a) and 2(b)). Moreover, in RCP4.5 scenarios, the frequency of droughts during the mid-period is 15.52%, while in RCP8.5 scenarios it shows 16.38%. There is 8.62% moderate drought under RCP4.5 for 2020–2049, and 8.62% severe drought under RCP8.5 for 2050–2079 (Table 3A). Moreover, at the same scale, the ungauged Gumara watershed had at 3.5 the greatest range of drought in the year 2060 under RCP4.5, and in 2057 under RCP8.5 (Figure 3(a) and 3(b)). Additionally, the temporal distribution of drought magnitude shows that, under RCP4.5, ungauged Gilgel Abay and Gumara were found to be in a moderate drought level, while ungauged Megech was suspected of being in an extreme drought in the near-period. Figure 4(a) shows that the watershed experienced a normal drought during the mid-period, while the overall drought was slightly reduced during the period.

On the other hand, the near-term RCPs show a 19.83% increase in the frequency of droughts on the annual time-scale (Table 3B). Under the RCP8.5 scenario, 12.07% of the highest drought frequency are at a moderate level. In the year 2048, all ungauged catchments suffered intense drought, although the ungauged Megech watershed had the maximum severity with a range of about 2.57 under RCP4.5. Moreover, in RCP8.5, the maximum drought, which ranges to 2.3, happened in the year 2042 at ungauged Gilgel Abay (Figure 2(c) and 2(d)). In addition, in the mid-period, the proportion of drought decreased somewhat, reaching 14.66% under RCP4.5 and 17.24% under RCP8.5 scenarios (Table 3B). The widest drought at this time occurred in 2070 at the ungauged Ribb watershed, ranging around 3 under RCP4.5, and another occurred at the ungauged Gumara watershed, ranging around 2.6 under RCP8.5 (Figure 3(c) and 3(d)). The spatial distribution SPI12 indicates a similar distribution, but in addition, ungauged Ribb was also included as moderate drought, and ungauged Megech was found at extreme drought in the near-period under RCP4.5 and through time shifted into wet categories (Figure 4(b)).

Under RCP4.5 18.97% of drought frequency occurs during the near-term on a biannual time-scale and 17.24% under RCP8.5. Of those, the largest percentage (11.21%) was found in the RCP4.5 scenarios that occurred during moderate-level drought (Table 4A). All ungauged catchments are predicted to face the greatest drought in 2046 in the near future, with ungauged Ribb expected to be hardest hit, with a range of 3.1 under RCP4.5 and about 3.2 under RCP8.5, respectively. Ungauged Gumara experiences the most extreme drought at a range of 3.5 at the mid-period of a year in 2060 under RCP4.5, whereas under RCP8.5, an intense drought with a 2.9 range magnitude occurs in ungauged Gilgel Abay's watershed in 2056 (Figure 6(a) and 6(b)). In the near-period under RCP4.5, the spatial distribution of ungauged Megech was found at extreme drought and the rest of the watershed was situated under normal drought distribution, and through time at the mid-period, the extreme drought shifted into ungauged Ribb and Gumara under the same scenario (Figure 7(a)).

The distribution of yearly drought frequency in the two near-future eras exhibits a 19.83% rise in trend for both RCPs. Under the RCP8.5 scenario, 15.52% of the greatest frequency of droughts happened at a moderate rate, and in the mid-period, 12.93% and 17.24% of drought occurred under RCP4.5 and RCP8.5, respectively (Table 4B). The severity of the current drought indicates that, according to RCP4.5, ungauged Megech will experience an intense drought that is in the 2.5 range in the year 2048, and ungauged Gilgel Abay will experience a somewhat less severe drought in the year 2042 (around 2.3) (Figure 5(c) and 5(d)). Moreover, ungauged Ribb experiences the most severe drought in the mid-period, with a range of around 2.7 under RCP4.5 in 2070, whereas both ungauged Megech and Gumara catchments experiences a range of drought (around 2.5) by the year 2073 under RCP8.5 (Figure 6(c) and 6(d)). At the annual time-scale in the near-period, all ungauged watersheds are found at extreme drought except the Megech watershed under RCP4.5, and through the time scale the basin drought decreases in both scenarios and periods (Figure 7(b)).

Using the MK test, the drought indices' trend over hydrology (SDI) and meteorology (SPI and RDI) in the Lake Tana sub-basin was determined. The findings demonstrated that the meteorological biannual (SPI6) and (RDI6) time-scale trend in the ungauged catchments of Gumara, Megech, and Ribb indicates a declining tendency toward drought at the close of the RCP4.5 time. Furthermore, all gauged stations on an annual or biannual basis exhibit no discernible meteorological drought pattern at the 5% significance level. Moreover, the sub-basin's near- and mid-periods do not exhibit any significant drought, except for the near-period of RCP4.5 (Table 6). Furthermore, the annual (SDI12) hydrological drought trend research shows no trend at a 5% significance level, as shown in Table 7.

The Lake Tana sub-spatial basin's variability of hydro-meteorological drought variability was examined in this study using data from four meteorological stations and four ungauged stations (Gilgel Abay, Gumara, Megech, and Ribb) under future probability distributions for the near (2020–2049) and mid-future (2050–2079) periods in both scenarios (RCP4.5 and RCP8.5). However, the result hides the drought's regional distribution and gives information on the drought event's size based on the analytical time-scale. Planning successful drought-mitigation strategies and developing drought early warning systems both heavily rely on spatial distribution assessments.

Figures 4, 7, and 10 show the geographic variation of the Lake Tana sub-basin's hydrological and meteorological droughts. It illustrates the incidence of drought in SPI6 and RDI6 under both scenarios in the near-period across all ungauged catchment faces, at some points moderate and extreme drought at ungauged Gumara and Megech, respectively. Continually, under SPI12 and RDI12 extreme and moderate droughts are shown under RCP4.5. By SPI12, the ungauged catchments of Gilgel Abay, Gumara, and Ribb have the moderate drought level but Megech catchment faces extreme drought. Conversely, at RDI12, the Megech catchment's drought distribution changes to a moderate level, while the remaining areas experience extreme drought under RCP4.5 in the near future (Figures 4 and 7). Conversely, when the time scale and scenario increase, the spatial distribution of the drought decreases (Figures 4 and 7).

The occurrence of annual spatial hydrological drought also continues as moderate drought at the four ungauged catchments in the near-period of RCP4.5 (Figure 10). However, the drought decreases as the period increases. At the end of RCP 8.5 in the near-period, Gilgel Abay is located in the wet range, and the distribution of ungauged Gumara, Ribb, and Megech stations reaches a severe drought level (Figure 10).

Drought is caused by a combination of hydrological and meteorological causes of rainfall variation significantly on a daily and regional basis, and SPI and RDI track changes in precipitation over time and space. For the near and mid-future periods, the Lake Tana sub-basin's meteorological drought trend does not exhibit any discernible trend at any of the total ungauged stations for the scenarios. Between the SPI6 and RDI6 time-scales, there is a tendency toward decline at three ungauged sites: Gumara, Megech, and Ribb; however, under the yearly time-scale, there is no variation in the trend. Extreme hydrological drought, on the other hand, is directly correlated with the absence of a meteorological trend. However, when we look at the size and frequency of the relationship at the annual time-scale of the distribution in both indices (meteorological and hydrological), it appears that they are closely related in addition to having a direct relationship with their percentile distribution (Tables 4–6). As a result, depending on the seasonal variation in rainfall throughout the basin of either increasing or decreasing trend, the hydrological dry pattern closely corresponds with the trend of streamflow.

In the RCP4.5 scenario, the RDI meteorological drought index outperforms the SPI index in capturing frequent high drought years when considering the 12-month time-frames and has a specific spatial distribution in the short and medium term. The findings demonstrate that the precipitation–based index is more drought-resistant than the precipitation-temperature-based index and that the precipitation-integrated index, including RDI, is more susceptible to precipitation in the near-period under RCP4.5. Accordingly, the RDI is much more sensitive and better able to capture drought output. Furthermore, SDI12 shows that, in both scenarios, there is a significant magnitude of hydrological drought at the midpoint in time. Overall, the frequency of droughts declines in both space and time.

Ethiopia's climate change varies widely throughout time and space due to increased greenhouse-gas emissions (Zegeye 2018; van Loon et al. 2019; Alemu et al. 2022). One of the causes of such volatility is a negative influence on the country in the form of catastrophic events such as drought (Taye et al. 2020; Wubneh et al. 2023). The UBN and Lake Tana sub-basin are among the basins that are believed to be experiencing drought (Taye et al. 2020; Wubneh et al. 2023). Both hydrological and meteorological droughts occur with varying frequency in some of the ungauged watersheds of the Lake Tana sub-basin, one of the UBN's sub-basins. When watershed parameters are modeled using conceptual HBV and the regionalization approach, gauged catchments are transferred into ungauged catchments to facilitate runoff simulation. This approach demonstrates a significant understanding and performance of the same as the study of Wale et al. (2009). Abera Tareke & Awoke (2022) conducted and evaluated the degree of drought susceptibility over the Abay basin, Ethiopia, using hydro-meteorological modeling methods using a similar approach and gave remarkable output.

Through analysis of future meteorological and hydrological drought, the result shows that the Lake Tana sub-basin data reveals there is no substantial drought trend in the study area. Due to this distribution, this research region has not seen a major increase in drought from the standpoint of future trend analysis, but at the ungauged Gumara, Ribb, and Megech at the biannual time-scale, there has been a modest drop in meteorological drought, which is directly similar to Abera Tareke & Awoke (2022) and Taye et al. (2020) with historical distributions of satellite-based meteorological drought. Also, in the near-period, the occurrence of drought is high in both meteorological and hydrological indices under RCP4.5, which is similar to Wubneh et al. (2023). Findings on the same sub-basin conclude that the basin will experience an extreme drought during this period. Nonetheless, a severe drought is evident in the geographical distribution of meteorological and hydrological drought in the Lake Tana sub-basin ungauged watersheds, which is directly similar to Bayissa et al. (2015) on how the Abay basin is affected by the drought, which is moderate to severe. Yet, in the future, as time lengthens, the distribution of droughts gradually decreases.

After the GCM data bias was corrected the models showed good agreement with the actual data: around 0.45 for NSE and 10.2 for PBIAS (percent bias) for the MIROC model, and 0.57 for NSE and 6.1 for PBIAS for the MPI-M-MPI-ESM-LR model from a previous paper (Wubneh et al. 2022). By ensembling the models, the uncertainty of the model was minimized. This study provides a spatial estimate of the likely hydro-meteorological drought using ensemble future-scenario data from RCP4.5 and RCP8.5 scenarios for the near (2020–2049) and mid (2050–2079) future periods at four ungauged sites (Gilgel Abay, Gumara, Megech, and Ribb). Meteorological drought using SPI and RDI and hydrological drought using SDI were investigated under biannual and annual time-scales. From analysis of the trend variation in the study area using the MK test the result shows that the basin does not experience a 5% significant trend increase in either type of drought at the ungauged stations; however, the spatial distribution of the meteorological drought at Gumara, Ribb, and Megech shows a moderate drought at the biannual time-scale of the near-future period and moderate drought at the biannual time-scale under both SPI and RDI (Figures 4, 7, and 10). For estimation of the SDI for the annual time-slice period, the HBV model was used after calibration of the actual data derived from the gauged watershed, Gumara, Gilgel Abay, Megech, and Ribb, and it shows 0.79, 0.80, 0.68, and 0.80 of NSE, respectively. Such values are used for estimating the hydrological drought of the study area for quantifying the drought.

Overall, 19.83% of the highest drought occurs at SPI12 and RDI12 time-scales under both RCP scenarios. The ungauged Ribb catchment station will experience an exceptional annual hydrological (SDI12) drought shortly under RCP4.5, and as the distribution increases over time, the drought magnitude decreases except during the mid-period of RCP8.5. Overall all RCPs show 18.10% of hydrological drought occurrence, while ungauged Gilgel Abay's distribution falls into the wet category. Both hydrological and meteorological droughts typically occur frequently, and as time scales grow longer, the relationship between the two become stronger. SPI and RDI both have high critical values, which suggests that both indices are appropriate for analyzing meteorological dryness in the Lake Tana sub-basin. The distribution of meteorological drought at ungauged Megech is more vulnerable to extreme drought and the highest magnitude was found under both SPI6 and RDI6 (around 3.5) in the year 2060 under RCP4.5. The hydrological drought of SDI6 also shows ungauged Gilgel Abay experiences extreme drought in the near-period of 2042 (around 2.9). The hydrological drought data's geographical variation indicates that the Lake Tana sub-basin has to prioritize water-resource management. Recently, the Ethiopian government has designed several initiatives to lessen the effects of poverty by utilizing water resources. Thus, climate specialists must create mechanisms for climate-friendly adaptation and forecast future droughts in the research area so that decision-makers, researchers, managers of water resources, and politicians can create drought preparedness and mitigation measures in the Lake Tana sub-basin.

No organization offered the authors support for the material they submitted.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.