Rainfall variability and trends in western Amhara: implication for sustainable water management and agricultural productivity

Climate is a major factor for sustainable development, but the emissions mainly caused by autonomous climate change produced from the production and consumption systems disturb social, economic, and ecological systems (IPCC 2021; Abbas et al. 2022; Bennett et al. 2023; Elahi et al. 2024; Lambert & Deyganto 2024). Humans influence regional, sub-regional, and global climate patterns (Jones et al. 2015; Mahmood & Furqan 2021). Continued high rates of population growth, industry, transportation, growing reliance on fossil fuel-driven growth technologies, and the effects of land use, such as urbanization, agriculture, livestock, and deforestation, are the main causes of global climate change (Chang et al. 2019). Although climate change is global, potential changes and variability are not expected to be uniform worldwide, and there can be dramatic variations from region to region (Abera & Abegaz 2020). However, climate change has historically been perceived as an environmental issue, and its effects can be felt in various areas, including infrastructure, transportation, agriculture, food security, and health (Schellnhuber & Cramer 2006).

Temperature variations and variations in the frequency, intensity, and predictability of precipitation are two of the most severe and potentially catastrophic consequences of climate change and variability in East Africa (Adhikari et al. 2015). According to Michael (2006) and Kisaka et al. (2015), alterations in regional precipitation will eventually influence the spatial and temporal distribution of water resources. This could result in decreased agricultural production and productivity, widespread food shortages, and food insecurity. A good example of a country affected by climate change and variability is Ethiopia, which has gone through several food crises in the past 10 years (Meseret & Belay 2019). This happens due to the majority of Ethiopia's population (around 80%) lives in rural areas (Dube et al. 2019), with agriculture being their primary source of livelihood (Deressa et al. 2008; Dube et al. 2019). This sector employs 85% of the workforce, contributes 50% of the country's gross domestic product (GDP), and generates more than 90% of foreign exchange earnings (Ayele & Tarekegn 2020; Abebe 2024). Additionally, the growth of other sectors is highly dependent on agriculture (Economic Commission for Africa 2016; Welteji 2018). However, agriculture in Ethiopia is highly vulnerable to climate change (Teshome 2017; Abeje et al. 2019). This is because Ethiopia is highly dependent on rain-fed agriculture and is subject to climate change (Geremew et al. 2020; Mesfin et al. 2021) and long-term changes and variability of climatic factors such as temperature and precipitation as frequency and severity (Bahiru & Zewdu 2021; Wossen et al. 2022; Dufera et al. 2023; Tegegn et al. 2024).

The second reason is that Ethiopia has limited resources to adapt to phenomena such as droughts and floods, which lead to reduced agricultural production and productivity and corresponding adverse effects on food security (Masika 2002; Bewket 2009; Alemayehu et al. 2020), irrigation (Döll 2002), and hydroelectric power (Kim & Kaluarachchi 2009). Rural households in the Amhara regional state that rely heavily on weather-sensitive crop and livestock production systems are impacted by climate-related hazards (Bewket 2009; Teshome 2017; Meseret & Belay 2019).

There are a number of studies that assessed the spatial and temporal fluctuations of rainfall in different parts of Ethiopia. However, the results of these studies reveal a range of outcomes, some of which have even been contradictory. For instance, studies such as Ayalew et al. (2012), Jury & Funk (2013), Degefu & Bewket (2014), Mekasha et al. (2014), Mengistu et al. (2014), Kiros et al. (2016), Alemayehu & Bewket (2017), Gedefaw et al. (2018), Getachew (2018), Alemu & Bawoke (2020), and Wubaye et al. (2023) concluded that there are no clear rainfall trends in the northern, central, southwestern, and northwestern parts of the country. However, studies such as Seleshi & Demaree (1995), NMA (2001), Osman & Sauerborn (2002), Seleshi & Zanke (2004), Verdin et al. (2005), Cheung et al. (2008), Wagesho et al. (2013), Addisu et al. (2015), and Mesfin et al. (2021) show that the annual and seasonal rainfall trends in the country's northern, southern, eastern, central, northeastern, northwestern, and southwestern regions are declining. This implies that climatic factors heavily impact the annual and seasonal rainfall trends at a small spatial scale, usually within a small geographic area. Therefore, the detailed spatial and temporal variability of climate parameters across the country, particularly in western Amhara, is very complex and does not show a clear trend. Additionally, studies such as Ayalew et al. (2012), Gedefaw et al. (2018), and Alemu & Bawoke (2020) used limited meteorological station data, which are not representative for generalizing the study area. Furthermore, no previous studies have focused on analyzing the spatial variability and temporal trends of seasonal and annual rainfall and its implication over western Amhara. This study aims to fill those gaps by conducting a more comprehensive assessment and analyzing the spatial variability and temporal trends of seasonal and annual rainfall using dense ENACTS data and a more rigorous methodology.

The study is based on a gridded daily precipitation dataset that was extracted from the Enhancing National Climate Services (ENACTS) for the period 1991–2020. Thirty grid points were purposely selected based on the fair distribution of the grid points in the study area. The Ethiopian Meteorological Institute (EMI) provided the data, which were produced by combining quality-controlled station records with meteorological satellites for rainfall and reanalysis products for temperature. The ENACTS initiative, spearheaded by Columbia University's International Research Institute for Climate and Society (IRI), is a distinctive and comprehensive project that aims to increase the accessibility, availability, and utilization of climate data to integrate climate knowledge into national decision-making. Because ENACTS datasets are geographically and temporally continuous, they provide low-cost, high-impact support for applications and research while enabling the characterization of climate risks at both local and national scales (Dinku et al. 2014). As described by Dinku et al. (2014), this is the most homogeneous dataset for the nation currently available for climate studies. ENACTS's spatial resolution, which is 0.03750 (∼4 km), is the same as that of TAMSAT's satellite rainfall estimates. However, ENACTS datasets typically cover a longer period, often from the 1980s to the present, and this dataset undergoes rigorous quality control and validation procedures to ensure high-quality data. ENACTS datasets are designed to support research on climate trends, variability, and impacts and are used for climate change research, water resource management, and agricultural planning (Dinku et al. 2017).

Numerous evaluations of ENACTS have shown good performance when tested at various station locations around the country, and ENACTS data show superior performance over CHIRPS2.0 and IMERG (Dinku et al. 2014, 2017; Alemayehu et al. 2020). Therefore, ENACTS data was selected for the analysis because (1) stations do not cover all of the study sites and are scattered throughout the study area (Dinku et al. 2014; Asfaw et al. 2018), (2) a lot of missing values are included in station datasets (Alemayehu & Bewket 2017), and since most stations are new, there are not enough data records to support trend analysis (Dinku et al. 2014). After getting ENACTS daily precipitation data from the Ethiopian Meteorology Institute, the next important step was organizing the data in a proper, easy-to-understand format. Next, the Climate Data Tool (CDT) package in the R program was used to extract the data from NetCDF and convert it into a comma-delimited (CSV) format. The CDT is an open-source, R-based software package with an intuitive graphical user interface (GUI) that can run on various operating systems, including Windows and Linux. It was created internally by the IRI and is currently utilized by 24 countries, mostly in Africa (Dinku et al. 2022). Then, data quality control methods were employed to clean and standardize the dataset for analysis. Quartile outlier detection and the Standard Normal Homogeneity Test (SNHT) were utilized to assess data quality. Quartile is one of the best outlier detection methods (Kuppusamy & Kaliyaperumal 2013) and was used to identify outliers in daily precipitation, and the meteorological time-series data were examined for non-homogeneity using the SNHT. The homogeneity test is crucial for climate and trend studies, as non-climatic factors can distort climate data and impact research conclusions. Identifying and removing non-climatic inhomogeneity is another vital task (Costa & Soares 2009). Among statistical homogeneity tests, the SNHT is commonly used to check climate data (temperature and rainfall) because of poor correlations between stations (González-Rouco et al. 2001). The lack of significant change points or inhomogeneity in the test results eliminated the need for mean adjustments; therefore, the original dataset was used for additional analysis. The CDT facilitated the examination of the SNHT and outlier results.

In the ith month, Pi represents the monthly rainfall amount. Oliver (1980) states that PCI values less than 10 demonstrate a uniform monthly rainfall distribution, values between 11 and 16 show a moderate concentration of rainfall, values between 16 and 20 show an irregular monthly rainfall distribution, and values of 20 and above demonstrate a very high concentration of rainfall distribution. Consequently, a rise in the PCI value over time denotes a rise in the monthly rainfall distribution's variability.

To assess statistically the trend's significance, the sample size, n, and the probability associated with S must be calculated.

At a 5% significance level (α = 0.05) for −Z1 − α/2 < Z ≤ Z1 − α/2, in a two-tailed test, the null hypothesis of “no trend” should be accepted. Here, Z1 − α/2 is the standard score (z-score) of the standard normal distribution with the cumulative probability of 1 − α/2. If not, a monotonic trend has been found at α significance level, and the null hypothesis should be rejected. As a result, a positive z_s value denotes an upward trend, a negative z_s value denotes a downward trend, and 0 values denote no trend (Belay et al. 2021). Thus, the R statistical software was used to determine whether or not there would be any significant trends in the rainfall data.

Finally, using a non-parametric model, was computed to determine the trend and slope magnitude at a significance level of 5%. In time series analysis, a positive value of Q_i suggests a rising trend, whereas a negative value indicates a falling or downward trend. Similarly, several zero suggests that the data show no trend.

The inverse distance weighted (IDW) interpolation approach was used in this study with ArcGIS 10.2 to investigate the spatial distribution of annual and seasonal rainfall variability. The weight is inversely proportional to the distance between the interpolated location and the observations or a function of the inverse distance (Birara et al. 2018).

Table 1 displays the results of the computation of rainfall parameters, including mean, CV, and precipitation concentration index (PCI) of the seasonal and annual rainfall over the study area from 1991 to 2020. The grid points at Enjibara (2,250.1 mm) and Mekan Birhan (626.3 mm) record the highest and lowest mean annual rainfall, respectively. The Ethiopian highlands (like Enjibara gird point) receive a high amount of rainfall due to altitude, moisture sources, and the migration of the Inter-Tropical Convergence Zone (ITCZ). On the other hand, like Mekan Birhan gird point, Ethiopian lowland areas receive low rainfall. Hence, the geographic position offers a complex orographic and global air circulation pattern that plays a key role in influencing the amount and distribution of rainfall and the overall climate of the region (Mera 2018). The area mean annual rainfall is 1,178.9 mm, with a corresponding standard deviation and CV of 348.6 mm.

As shown in Table 1, the mean CV values in annual, Kiremt, Bega, and Belg are 17.3, 18.1, 64.9, and 49.5%, respectively. The coefficient variation of most grid points indicated that rainfall in the region has high inter-annual variability (Table 1). As described in Table 1, the CV values in annual rainfall are less than 20%, except for Abrahajira, Debark, Ebinat, Jawi, Mereto Lemariyam, Metema, Nefas Mewucha, and Nefes Gebeya grid points, which had moderate variation (CV between 20 and 30%), whereas Mekan Birhan grid point shows high rainfall variability. All grid points in the Bega and Belg seasons show significant rainfall variability, whereas the majority of grid points (66.6%) in the Kiremt season have low variations in rainfall (CV <20%) (Table 1). The findings align with those of Cheung et al. (2008), who found that Ethiopia experienced rainfall variability exceeding 20%. Furthermore, Ayalew et al. (2012), Alemu & Bawoke (2020), and Mesfin et al. (2021) report that a majority of the rainfall stations in the Amhara Region exhibit moderate variance during annual and Kiremt periods and a very high level of variation during the Bega and Belg seasons. However, in this study, most parts of western Amhara are highly variable (exceeding 70%). Overall, the study's seasonal timeline showed more rainfall variability than its annual scale.

Similarly, during the Belg season, the maximum distributions of rainfall were observed in the highland, southwestern, and southeastern parts (150–400 mm), while the northern, northwestern, and central parts of western Amhara received low amounts of rainfall (0–150 mm) (Figure 2(d)). From February to May, the Arabian High moves toward the northern Arabian Sea and pushes over the water body, causing a moist, southeasterly air current to flow toward Ethiopia (NMSA 1996).

For the Kiremt, Bega, and Belg seasons, the mean rainfall was 937.9, 98.3, and 149.4 mm, respectively. Kiremt, Belg, and Bega contributed the maximum share to the annual rainfall (79.6, 8.0, and 12.4%, respectively). The rainfall during the Kiremt season followed the same pattern as annual rainfall. Similar findings were reported by Seleshi & Zanke (2004) and Ayalew et al. (2012), who demonstrated that in the Amhara Regional State of Ethiopia, the Kiremt and Belg seasons contributed 74.3 and 5–30%, respectively, to the annual rainfall. However, in the northern and northeastern parts of western Amhara, the Kiremt season contributed more than 90%.

In most study areas, the CV value is less than 20% during the annual and Kiremt periods, indicating minimal rainfall variability. The majority of grid points show low variance annually and high to very high variation seasonally, as demonstrated in Figure 3(a)–3(d). In the Bega season, the CV value exceeds 40% across all areas of western Amhara (Figure 3(c)), indicating extremely high rainfall variability. Likewise, during the Belg season, the CV exceeds 40% in nearly all parts of western Amhara.

Generally, in the lowland, northeastern, and northwestern regions of western Amhara that experienced relatively little rainfall distribution, the CV values showed relatively high variability in the annual and seasonal time series. Additionally, Bewket & Conway (2007) and Harka et al. (2021) discovered that regions with elevated precipitation levels demonstrated lower coefficients of variation. Seasons with the highest and lowest CVs were Bega and Kiremt, respectively. As a result, areas with high annual rainfall showed less variation from year to year, whereas areas with low annual rainfall showed significant variation. Compared to places with low CV, the high inter-annual variability in low rainfall locations revealed a bigger disparity in annual rainfall levels from year to year. This suggests that water availability has become more uncertain in these areas. The result implies that the rainfall variability in the Bega season affects the amount of reservoir and groundwater, which is crucial for irrigation and domestic consumption. Such kinds of prolonged dry spells can worsen water shortage in agrarian communities. Likewise, rainfall variability in the Bega season implies that the smallholders should have to plant short-mature crops since its variability affects the growth and productivity of long-mature crops. Furthermore, solving these challenges requires adopting drought-tolerant crops, soil and water conservation practices, and strengthening early warning systems customized to local needs.

Table 2 shows the area (in percentage) of the different classes of PCI in the study area. The table reveals that 3.4% of the study area experienced moderate rainfall distribution, 13.4% had irregular rainfall distribution, and the majority (83.3%) exhibited strong irregular rainfall distribution. This result slightly contradicts the research conducted by Alemu & Bawoke (2020) in the Amhara region. They found that most of the region has an irregular distribution of rainfall.

Table 3 displays the findings on the annual and seasonal precipitation grid points, which show positive and negative trends using the modified Mann-Kendall package in R software. Six of the 30 grid locations had an insignificantly decreasing trend in the mean annual rainfall. At a 5% significance level, Ayehu, Dangila, Enjibara, Mereto Lemariyam, Motta, and Yetenora grid points show insignificantly decreasing trends, while Aberahijira, Adet, Debark, Nefes Gebeya, Quara, and Sanja display significantly increasing trends with rainfall amounts of 10.3, 7.2, 19.0, 8.2, 14.3, and 9.5 mm/year, respectively. In the annual period, 24 out of the 30 rainfall grid points showed statistically non-significant rising or declining trends. Thus, the hypothesis that there is no trend (H₀) is accepted because the computed P-value is greater than the significance level of α = 0.05 (Table 3). The annual rainfall Sen's slope estimators are shown in Table 3. SSE magnitudes that are positive and negative are recorded at 24 and 6 rainfall grid points, respectively. Sen's slope estimators for the annual rainfall were found to have a maximum magnitude of +19.0 mm/year in Debark and a minimum magnitude of −4.3 mm/year in Ayehu grid points. The outcomes align with the studies conducted by Mengistu et al. (2014) in the Ethiopian Upper Blue Nile River Basin. The majority of annual rainfall stations displayed an increasing trend. Mesfin et al. (2021) revealed that most of the Amhara region showed an insignificant trend, whereas the western parts of the study area showed a significantly increasing trend. Furthermore, according to the IPCC (2021), rainfall has increased statistically significant in northern and central Africa but not in southern or eastern Africa. Ogega et al. (2020) note that the eastern Horn of Africa and areas with high/complex topography will experience a longer rainfall season and higher rainfall of over 100 mm on average at groundwater level (GWL) 4.5°C during the short rainy season.

Table 3 presents the statistics for the seasonal rainfall derived from Sen's slope estimators and the MK trend test using the modified MK package of R software. The findings show that during the Kiremt season, almost all of the rainfall grid points (93.3%) are an increasing trend. Among those, Aberahijira, Debark, Metema, Nefes Gebeya, Quara, and Sanja grid points display significantly increasing trends with rainfall amounts of 10.1, 16.2, 7.2, 7.6, 13.5, and 8.6 mm/year, respectively. During the Belg season, nearly all rainfall grid locations (96.6%) displayed a trend that was not statistically significant. A statistically significant increasing trend was observed in only Quara grid points during the Bega season, but other rainfall grid points during that season showed an insignificant increasing trend (Table 3).

In general, MK and SSE findings are consistent with rainfall patterns and trends during the annual and Kiremt periods. This similarity further indicated the high contribution of Kiremt rainfall to annual rainfall in all the grid points in western Amhara (79.6%) during the last 30 years. The outcome supports the findings of Bewket & Conway (2007), which show that Kiremt rainfall accounts for a significant amount of annual rainfall. Furthermore, studies such as Bewket & Conway (2007) and Ayalew et al. (2012) prominently emphasized the significant impact of Kiremt rainfall on annual rainfall. These investigations generally demonstrated that, in other regions of Ethiopia, the Kiremt season is the primary source of annual rainfall.

Generally, climate change in the western Amhara region has critical negative impacts on various sectors such as agriculture, water, health, and forestry. The region is the most vulnerable to climate and ecological changes, given that only a small proportion of its cultivated land is irrigated, and food production is highly dependent on traditional rain-fed agriculture (crop production and livestock keeping) (NMA 2007). Rain-fed agriculture is highly sensitive to climatic fluctuations and exacerbates poverty in the region as well as in the country (Tesfaye et al. 2016; Teshome 2017). Over 85% of the active population is engaged in this sector. However, the north and northeast parts of the region are more exposed to a shortage of rainfall, receiving less than 1,000 mm annually (Figure 2), and a shortage of food throughout the year. Many surplus-producing districts in the administrative zones of North and South Gondar, East Gojjam, and North Shewa that produce excess have already been severely vulnerable to poverty and food insecurity during the past 20 years (Teshome 2017). In western Amhara, temporal rainfall variability during the cropping season induces the main challenges to crop production through flooding, insect outbreaks, spreading of alien weeds, disease, and pests (Mengesha et al. 2023). Hence, the relationship between rainfall variability and crop production in western Amhara is a significant correlation and leads farmers to be vulnerable to food insecurity (Bewket 2009).

On the other hand, Ethiopia, including western Amhara, already experiences a high flood risk. During the 2020 Kiremt season, numerous rivers flooded, including Gumera, Megech, Rib, and surrounding Lake Tana, Akobo, Alwero, Awash, Baro, Bilate, Genale Dawa, Gilo, and Wabe Shebelle rivers, and several dams, including Kesem, Koko, Kuraz, and Tendaho; these factors have affected over a million people and forced 292,863 of them to relocate (Government of Ethiopia and OCHA 2020). These fluctuations in weather and climate may increase climate-related health risks like non-communicable diseases, medical emergencies, the emergence and spread of infectious diseases, morbidity, and mortality.

Based on the analysis, there is a tendency for rainfall to increase during the Kiremt season, which is favorable for irrigation, drinking, and farming (Table 3). Mekoya et al. (2024) similarly show that the Amhara region's annual and Kiremt rainfall projections under representative concentration pathway (RCP) scenarios also indicate an increase in the near term (2021–2040) and mid-term (2041–2060). Furthermore, as stated by Alaminie et al. (2021), the Abay Basin's predicted rainy season (Kiremt) precipitation is expected to increase by 14 point 4 mm, 30 point 5 mm, 31 point 8 mm, and 46 point 4 mm under SSP1-2.6, SSP2-4.5, SSP3-3.7, and SSP4-8.5, respectively. These increases are statistically insignificant. A significant portion of the flow of the Upper Blue Nile River comes from rainfall in the western Amhara region. Rainfall in the western Amhara area contributes significantly to the flow of the Upper Blue Nile River. The observed and predicted rainfall indicates a rise in the occurrence and seriousness of extreme events in the catchment regions, impacting water and sediment flows into the river and agriculture in Sudan and Egypt downstream (Mengistu et al. 2014). Similarly, this result shows that lowland areas have a significant increasing trend during the Kiremt season and have an excellent chance for agricultural production in western Amhara. For many rural smallholder farmers, agriculture is their main source of income; therefore, any variation in the quantity, distribution, or patterns of rainfall will have a direct influence on agricultural output and, as a result, significantly impair their quality of life. This finding indicates that the Belg season has an increasing rainfall trend, and it is a good opportunity for land preparation over western Amhara (Table 3). The Bega season, which runs from October to January, will see a significant decrease in rainfall, which will have a negative effect on legume crops. Farmers in the western Amhara region are unable to cultivate legumes on their small pieces of land in a given year due to this season's results, which show that they receive little rain, which causes food insecurity and poverty.

On the other hand, excessive rainfall combined with insufficient water management during July and August causes more frequent floods and soil erosion, which puts agricultural productivity at risk in western Amhara. Anaerobic stress in the roots brought on by flooding and waterlogging in farmlands, particularly in lower-slope sites, significantly lowers agricultural productivity. Without adaptation measures, temperature and precipitation changes will raise food costs globally by 2050, with estimates ranging from 3 to 84% (IPCC 2021). The availability and accessibility of high-quality water for household use must be improved by sustainable agricultural practices, especially irrigation farming and flood control through the channelization of water passages, in response to increased surface runoff that introduces pollutants into water sources. Furthermore, the results of this study suggest that increasing water availability and flow during the rainy season can help accommodate surplus storage for irrigation during the dry season. This should incentivize the adoption of water-stress crops and the establishment of a watershed-level future development strategy in western Amhara.

In the western Amhara region, there was heterogeneity in the spatial distribution of the seasonal and annual rainfall. During the annual and Kiremt rainfall, the CV showed near-to-moderate variability over the study area. On the other hand, the rainfall during the Bega and Belg seasons showed extremely high and above-limit variability. In most areas of the study area, the PCI data showed extremely uneven distribution with high concentrations and volatility in the annual and seasonal time series. The northern and northeastern parts of the study area had the highest PCI values, while the southern and southwest regions had the lowest PCI values. The negative SAI values were 46.7, 53.3, 60, and 50% for the annual, Kiremt, Bega, and Belg seasons, respectively. The findings demonstrated that the majority of the study area had negative anomalies. During the Kiremt season, six grid points showed a significant increase trend, while one grid point did a significant increase trend during the Bega and Belg seasons.

The study area is susceptible to climatic variation and fluctuation, and these alterations may probably result in a rise in the frequency and severity of natural disasters. The year-to-year variability in the annual and seasonal rainfall suggests that climate-dependent sectors such as agriculture (both crop and livestock) and water resource developments are already highly exposed to current climate-related risks. Therefore, this study recommends that to solve the obstacles caused by rainfall variability in the study region, it is crucial to improve the localized climate advisory service that produces real-time and tangible weather and seasonal rainfall predictions. Enhancing water conservation practices, like rainwater harvesting technology and terracing, with supplemental and deficit irrigation systems can minimize dependencies on erratic rainfall. Furthermore, appreciating the plantations of drought-tolerant and early-maturing crops will improve agricultural resilience, improving soil and water conservation, particularly during the Belg season. Scholars might conduct further investigations to determine the impacts of climate change in western Amhara using official yield statistics and rural households' perceptions of crop yield trends over the past few years. They will also associate other global meteorological forcing factors, such as the ITCZ, El Niño, and La Niña events, with rainfall trends.

T.S.A.: Conceptualization, data collection, statistical analysis, and writing – original draft preparation. Z.Y.A.: Conceptualization, supervision, writing – review & editing. B.B.G.: Analysis support and writing – review and editing. M.G.A.: Writing – review and editing. After reading the final draft of the manuscript, all authors agreed to the published version of the manuscript.

The authors thank the Ethiopian Meteorology Institute (EMI) for allowing us to access its data without any restrictions.

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

The authors declare there is no conflict.