Spatiotemporal variability and trends in extreme rainfall and temperature indices in Southeastern Oromia, Ethiopia

Abstract

Extreme climatic events have become a significant global concern, exacerbating drought and flood risks in Ethiopia’s agricultural regions. This study presents the first multi-index analysis of 23 ETCCDI indices in southeastern Oromia (1994–2023) to address hydrological challenges. Using data from nine meteorological stations, we analyzed spatiotemporal trends via the Modified Mann-Kendall test with Hamed and Rao’s autocorrelation correction, Theil–Sen slope estimator, descriptive statistics, linear regression, and Standardized Anomaly Index (SAI). Spatial interpolation employed Inverse Distance Weighting (IDW). The results show that the rainfall extremes (CWD, R20mm, R25mm) and temperature indices (TNn, SU25, CSDI, WSDI) exhibited high variability (CV > 30%). Significant declines occurred in RX1day, R95p, and R99p, while CDD increased by + 0.4 to + 1.17 days/year at 77.8% of stations. Warming trends dominated: TX90p and TN90p rose by 0.10 to 0.57 and 0.20 to 0.60 days/year, respectively; TXx and TNx increased by 0.03 to 0.2 °C/year and 0.05 to 0.12 °C/year at all stations. All stations showed rising TXx, TNx, TX90p, and TN90p, with 66.7% recording more SU days. Hot indices (TXx, TNx, TX90p, TN90p, SU) increased as cold indices (TNn, TXn, TX10p, TN10p) declined, highlighting warming asymmetry. Spatial analysis revealed station-specific variability, emphasizing localized climate risks. These trends demand urgent adaptive strategies to mitigate escalating climate threats in vulnerable regions.

Introduction

In recent decades, climate change and its variability have emerged as major scientific concerns^1,2 Climate change is the most critical threat to the global food system³. Extreme climatic and weather events have become a global issue because of their severe effect on human life⁴. Recent advances in climate extremes research highlight the hydrological implications of these events, with studies such as real-time GNSS tropospheric parameter prediction for extreme rainfall events in China⁵, wavelet-based hybrid models for seasonal rainfall forecasting⁶, and machine learning approaches to predict wet-period rainfall using climate indices^7,8 demonstrating the growing sophistication of extreme event analysis.

Temperature and rainfall are two fundamental climatic variables that determine the climate of a region, and their alterations directly or indirectly affect many hydro-climate systems. Changes in the frequency and intensity of extreme climate change event significantly impact the environment and human populations^9,10. Extreme climate events, such as flooding, drought, and heat waves, are currently occurring with greater intensity than previously observed^11,12,13. Variations in extreme weather events exhibit substantial regional differences owing to the effects of climate change on diverse geographical areas¹².

Despite global and regional studies on climate extremes are abundant, southeastern Oromia a key agro-ecological zone and Ethiopia’s first largest wheat producing region has received limited attention, particularly in terms of long-term, multi-index analyses¹⁴. While ETCCDI-based assessments are common in Ethiopia¹⁵ prior work has focused on coarser spatial scales or fewer indices, limiting localized insights for adaptation planning¹⁶.

According to¹⁷, human activity has significantly impacted ecosystems, causing complex environmental hazards that can harm both nature and people. One such impact is the change in the Earth’s atmospheric composition, which has altered the global climate system and caused the planet’s temperature to rise by 1.1 °C in the last century¹¹. Climate scientists estimate that temperatures in the globe will increase by 2.1 °C to 2.6 °C by 2050 and by approximately 3.3 °C to 3.8 °C by 2080¹⁸. This warming has intensified extreme climate events, with severe consequences for society and the environment^10,19. In southeastern Oromia, these changes are exacerbated by local drivers such as orographic effects from the Ethiopian highlands and ENSO-mediated rainfall variability^20,21, yet their role in extreme events remains poorly quantified a gap this study addresses by integrating topographic and climatic drivers into trend analyses. To mitigate these risks, it is imperative to comprehend the spatial and temporal distribution of these events^22,23.

Ethiopia is particularly vulnerable to climate change due to its low adaptive capacity, reliance on rain-fed agriculture, and exposure to recurrent droughts^9,24. Southeastern Oromia a predominantly wheat producing region faces disproportionate risks from temperature extremes and rainfall variability²⁵, Unlike prior studies that treated extreme indices in isolation¹⁶, this work directly links ETCCDI trends (consecutive dry days, warm spells) to wheat yield losses²⁶, providing actionable insights for adaptation. These extremes exacerbate water scarcity, with declining river flows and groundwater recharge rates^27,28, while wheat producing communities face heightened livelihood insecurity²⁹.

Globally, cold days and nights have declined, while warm extremes have increased since 1950^11,30. In Ethiopia, similar trends are documented¹⁴, but critical gaps persist like prior studies focus on national or basin-scale analyses, masking local agro-ecological variations¹⁶; the interplay between extreme indices and crop-specific impacts remains unquantified²⁶; and localized climate drivers (Ethiopian Highlands’ rain shadow effects, ENSO teleconnections) are rarely integrated into extreme event assessments^20,31. Our study advances past trend analyses by combining high-resolution ETCCDI indices with crop-climate modeling, offering a template for microclimatic extreme event assessments in data-scarce regions.

Ethiopia’s complex topography creates microclimates that modulate temperature extremes³². Rapid land-use changes (deforestation) exacerbate local warming³³. Leeward slopes show more hot days due to warming, while windward slopes remain cooler³⁴. Therefore, more accurate datasets are needed to examine changes due to the high spatiotemporal variability of extreme climate events. The present study investigates climate extremes and associated trends to climate change and bridge critical gaps between climate indices and livelihood outcomes by employing a 23-index ETCCDI framework tailored to southeastern Oromia’s agro-climatic zones, quantifying ENSO and topographic influences on extremes, and linking trends to wheat yield vulnerability a novel approach for Ethiopia. The study’s objectives are (i) to quantify changes in rainfall and temperature extremes using 23 ETCCDI indices and (ii) to identify trends and changes associated with 23 extreme climate indices in southeastern Oromia, Ethiopia.

Materials and methods

Description of the study area

The research was carried out in three administrative zones situated in southeastern Ethiopia, namely Arsi, Bale, and Guji Zones of the Oromia Regional State (as shown in Fig. 1). Arsi zone is located between 7^o 32’15“N and 8^o 32’45“N, as well as 38^o 42’30“E and 40^o48’10“E. The area has an altitude range of 600 to 4100 m above sea level and receives an average annual rainfall of 920 mm, with an average minimum and maximum temperature of 12 ℃ and 20 ℃, respectively.

Bale Zone is located between 5^o 36’ 32” to 8^o 12’ 57” N and 39^o 21’ 27” to 42^o 23’ 27” E and has an elevation range of 880 to 3554 m above sea level. The area receives a mean annual precipitation of 1200 mm and exhibits mean minimum and maximum temperatures of 6 °C and 21 °C, respectively. Multiple soil types are present across the region. These soil types exhibit variations in their composition, fertility, and suitability for diverse agricultural activities. The prevalent soil types identified in the Bale Zone include Luvisols, Vertisols, Nitisols, and Cambisols³⁵. Each soil type possesses distinct characteristics and properties that can influence the productivity and potential utilization of land^36,37. Understanding soil types in the Bale Zone is crucial for effective land management and agricultural planning.

Guji Zone is located between 8^o 11’ 5” to 9^o 50’ 42” E and 37^o 4’ 22” to 38^o 49’ 25” N, with an elevation range of 916 to 3602 m above sea level. The region experiences a mean annual precipitation of 1100 mm, with mean minimum and maximum temperatures of 11 and 26 °C, respectively. The Ethiopian climate is characterized by three distinct seasons. The primary rainy season occurs from June to September (JJAS), followed by the dry season from October to January (ONDJ), and the secondary rainy season from February to March (FMAM), locally designated as Kiremt, Bega, and Belg, respectively³⁸. The southeastern highlands of the country receive the highest (> 2000 mm) and lowest (< 300 mm) rainfall. The mean annual temperature ranges from 15 °C in the highlands to 25 °C in the lowlands³⁹.

Fig. 1

Map of the study area.

Data sources

This study utilized high-quality historical daily rainfall and temperature (maximum and minimum) data spanning a 30-year period (1994–2023) obtained from the Ethiopia Meteorological Institute (EMI). The daily data were processed to derive monthly, seasonal, and annual rainfall totals, ensuring temporal consistency across stations. Nine meteorological stations were carefully selected based on strict criteria including spatial representation of the Arsi-Bale-Guji highlands wheat belt, data completeness (> 95%), and agroecological relevance to major wheat-growing districts. Each station met rigorous quality standards, allowing no more than 5 missing days per month and a maximum of 5 incomplete years over the study period. The selected stations provide comprehensive coverage of the southeastern Oromia region, which produces over 40% of Ethiopia’s wheat⁴⁰, with Arsi and Bale zones alone contributing more than 30%⁴¹. This region faces significant climate risks, with projections indicating potential 10–20% yield declines by 2050 without adaptation measures^24,42. The station locations, shown in Fig. 1, capture critical elevation gradients and microclimates essential for wheat production. A comprehensive quality control process involving outlier detection and homogeneity testing was implemented to ensure data reliability. This robust methodological approach provides a solid foundation for assessing climate trends and their implications for Ethiopia’s vital wheat sector.

Data quality control

Various investigative methodologies were implemented to assess data quality following established meteorological data standards (WMO). A multi-stage approach was adopted, aligning with World Meteorological Organization (WMO) standards⁴³ and best practices in climate indices analysis^44[,45.

Outlier handling

Initial physical plausibility checks removed records with impossible values (maximum temperatures < 0 °C, maxima < minima, minima > 20 °C)⁴³. Statistical outliers (> 3σ from station-group means) were cross-verified against maintenance logs and historical weather reports⁴⁶.

Homogeneity testing

Temporal consistency was evaluated using Pettitt’s test⁴⁷ for breakpoint detection (± 5 °C thresholds). Double-mass curve analysis against neighboring stations (within 50 km radius) identified inhomogeneities⁴⁸, corrected via quantile mapping⁴⁹ while preserving natural variability. Furthermore, this study analyzed annual trends of minimum and maximum temperatures and indicators utilizing extreme data calculated via RClimDex 1.1 built-in checks for temporal coherence and threshold adherence⁵⁰. Additionally, rainfall and temperature data from nine stations spanning a period from 1994 to 2023 were employed for the analysis. The geographical coordinates and detailed information pertaining to the selected meteorological stations are presented in Fig. 1.

Methods of data analysis

Descriptive statistics

After selecting the agroclimatic extreme indicators, we calculated descriptive statistics mean, minimum, maximum, and coefficient of variation using Instat + v3.37 software. According to⁵¹ classification, a coefficient of variation below 20% indicates low variability, while 20–30% suggests high variability, and above 30% signifies severe variability. We also examined the spatial distribution patterns of each indicator through spatial trend analysis and the inverse distance weighted (IDW) interpolation technique in ArcGIS. We applied IDW interpolation to spatialize climate extreme indices following standard geostatistical practices⁵². Our IDW implementation (power = 2, 12 neighbors) followed established interpolation standards for Ethiopian highlands⁵³, producing errors < 10% of index means.

Trend analysis of rainfall and temperature extremes

Daily precipitation, maximum temperature, and minimum temperature data from nine climate stations in southeastern Ethiopia (Arsi, Bale, and Guji zones) were analyzed for trends over 30 years (1994–2023). The analysis utilized the RClimDex 1.1 package in R⁵⁴, a widely used tool for climate trend analysis^55,56. Extreme climate indices established by the Expert Team on Climate Change Detection and Indices (ETCCDI) were used to assess various aspects of precipitation and temperature extremes^50,57. Eleven extreme precipitation indices were selected, representing intensity, frequency, and duration. Trend analysis of these indices was conducted using RClimDex 1.1. To detect trends and assess their magnitudes, the modified Mann-Kendall (MMK) test⁵⁸ and Theil–Sen slope estimator⁵⁹ were applied via the Modified Mann-Kendall test with⁶⁰ autocorrelation correction, MK R package as temperature and rainfall extremes in highland regions often exhibit serial dependence⁶¹. Theil-Sen Slope Estimator quantified trend magnitudes. For multi-station consistency, regional trends were evaluated using the Regional Kendall Test⁶². This non-parametric approach was chosen for trend analysis due to its robustness with small samples (n = 30 years) and non-normal data⁶³. Unlike OLS, it resists outliers and heteroscedasticity⁶⁴, making it ideal for skewed extreme rainfall indices. It provides reliable slope estimates⁶⁵ and complements the Mann-Kendall test by quantifying trend magnitude rather than just significance and aligns with WMO⁴³ standards for climate trends. We selected 23 ETCCDI indices (Table 1) to assess climate extremes critical to southeastern Oromia’s semi-arid agriculture. The analysis prioritized indices most relevant to rainfed farming, omitting unsuitable metrics (e.g. snow indices), ensuring region-specific insights. Table 1 presents the details of eleven extreme precipitation and twelve of temperature indices.

Table 1 Indices of air temperature and rainfall developed by ETCCDI, are used for analyze data from daily rainfall and maximum and minimum temperatures, along with their definitions and units.

Anomaly analysis of precipitation and temperature extremes

A Standardized Anomaly Index (SAI) was utilized to analyze the frequency of dry and wet years from 1994 to 2023. The SAI was carefully adapted to each index type: absolute temperature indices were directly standardized, while count indices required Poisson variance stabilization before SAI calculation. For duration indices, we applied log-transformation to normalize distributions prior to standardization, following⁶⁶. The standardized anomaly of the index is calculated as follows:

$$SAIi=\frac{{{X_t} - {X_{mean}}}}{\mu }$$

In this context, SAIi represents the standardized anomaly index for year i. Xt refers to the annual index value in year t, while Xmean is the long-term average of the yearly index over the observation period. Additionally, µ denotes the standard deviation of the annual index during the same timeframe. This study conducted a trend analysis of rainfall and temperature extremes, as well as the magnitudes of those trends, at the station level following the rating outlined in Table 2.

Table 2 Standardized anomaly indices (SAI) values of extreme Climatic variables and Interpretation.

Result and discussion

Descriptive statistics of agroclimatic extremes

Precipitation extreme indicators

Table 3 presents the mean descriptive statistics for extreme rainfall agroclimatic indicators collected from nine meteorological stations in the study area. The results reveal significant differences in the intensity of these indicators among the stations. On average, the stations recorded 52 consecutive dry days (CDD), with a coefficient of variation (CV) of 12.7%. This indicates that, over the past 30 years, the area has experienced an average of 52 days per year with less than 1 mm of rainfall. The annual minimum and maximum CDD values recorded at the Bore and Asela stations were 44 days and 61 days, respectively. The study region has shown significant spatial variability in consecutive dry days (CDD) in addition to statistical variation. Figure 2 illustrates the geographical variations of minimum rainfall-related extremes, with similar color indicating areas of similar values for the variables. In particular, Ginir, Wadera, and Asela districts recorded a higher number of CDD days, ranging from 58 to 61 per year. In contrast, Bore, Dodola, and Bokoji experienced fewer CDD days, a range of 44 to 46 days per year. Agarfa, Adola, and Sinana districts reported medium CDD values, ranging from 50 to 56 days per year. Overall, there is a noticeable pattern of higher CDD values in some of the studied stations of southeastern Oromia. This spatial variability can be attributed to several factors, including local topography, proximity to water bodies, and prevailing wind patterns. The impacts of climate change and land use practices may also influence these CDD patterns (Fig. 2a). Understanding these spatial variabilities is vital for effective agricultural planning and management.

The mean number of consecutive wet days (CWD) recorded at the nine meteorological stations included in this study was 16 days, with a coefficient of variation of 34.2% (Table 3). This observation indicates that, over the past 30 years, the region has experienced an average of 16 days per annum with precipitation exceeding 1 mm. The annual minimum and maximum values of CWD at the Agarfa and Bokoji stations were 10 and 26 days, respectively. This variation showcases significant fluctuations in CWD during the study period in the southeastern Oromia region. Alongside statistical variation, the study region has shown notable spatial variability in CWD. Figure 2b shows the spatial variations of CWD rainfall extremes.

The average value of the Simple Daily Intensity Index (SDII) is 7.1 mm per day per year, reflecting the typical annual rainfall. However, the high coefficient of variation (CV = 20.3%) indicates significant variability in the data, suggesting that the real SDII can differ considerably from year to year. The minimum and maximum values recorded were 5.3 mm at the Sinana station and 9.0 mm at the Adola station, respectively. This means that the lowest SDII measurement of 5.3 mm was observed at the Sinana station, while the highest measurement of 9.0 mm was recorded at the Adola station. Figure 2c clearly displays the spatial variability of minimum rainfall-related variables associated with SDII in the studied region. The geographical analysis of SDII showed that there was highly positive variability in areas around Bore, Agarfa, Ginir, Wadera, and Adola, whereas, SDII was small variability around Sinana, Dodola, Bokoji, and Kulumsa.

Fig. 2

The geographical variability of minimum annual precipitation-related agroclimatic extreme indicators.

Extreme maximum rainfall indicators

Table 3 summarizes the average statistics for maximum rainfall extreme indices from nine weather stations in the studied area. The average 1-day and 5-day rainfall values recorded were 37 mm and 78 mm, with high coefficients of variance of 25.8 and 25.9, respectively. The minimum and maximum R1xdays of rainfall were recorded at Dodola and Adola stations. For R5xdays, the minimum and maximum were observed at Wadera and Adola, measuring 44.6 mm and 115.3 mm, respectively. Figure 3 demonstrates the spatial variability of maximum rainfall-related variables in the study stations. Specifically, Fig. 3a and b depict the geographical distribution of 1-day and 5-day, respectively.

Rest on the recorded data, the average number of heavy precipitation days (R10mm) is 27 days per annum, which indicates the typical frequency of such events annually. However, the coefficient of variation (CV = 24.4%) reveals a significant level of variability, revealing that the actual number of heavy precipitation days can differ highly from year to year. The minimum and maximum number of heavy rainfall days recorded at the Sinana (17) and Bore (40) stations per year. Thus, all the studied districts experienced heavy rainfall for a range of 17 to 40 days per year, and the spatial variability of the R10mm indices is shown in Fig. 3c.

The results of the descriptive statistics analysis revealed that the number of very heavy precipitation days (R20mm) in the study area is highly variable. On average, there were 6.6 days of heavy rainfall (R20mm) each year, with a severe coefficient of variation (CV = 55.56%), implying a significant degree of variability. The lowest and highest recorded results were 2 days at the Sinana station and 12 days at the Adola station, respectively and the spatial variation was observed in Fig. 3d.

The descriptive statistics revealed significant variation in very heavy rainfall days (R25mm) in the study area. On average, there are 4.6 days of heavy rainfall (R25mm) each year, with a severely high coefficient of variation (CV = 81.0%), indicating considerable variability. The minimum recorded value was 1 day at the Sinana, Adola, and Bokoji stations, while the maximum was 11 days recorded at the Ginir station. Similarly, the spatial interpolation on R25mm displayed that there was high spatial variability in the study area from 1994 to 2023 (Fig. 3e).

The mean annual value of very heavy precipitation (R95P) is 175.4 mm, which represents the typical amount of precipitation received in a given year. However, the coefficient of variation (CV = 16.2%) indicates a low level of variability in this data, suggesting that the actual amount of very heavy precipitation can fluctuate minimally from one year to another. The lowest and highest recorded values for this indicator were 132.67 mm at the Sinana station and 223.48 mm at the Adola station, respectively. This shows that Sinana experienced the lowest annual rainfall, while Adola had the highest. The distribution of R95P is uneven across the study region, with higher rainfall volumes observed in districts such as Adola, Bore, and Agarfa, where annual precipitation ranges from 181.3 mm to 223.76 mm. In contrast, districts like Sinana, Dodola, and Bokoji reported lower rainfall volumes, ranging from 132.7 mm to 163.8 mm per year. The spatial analysis of R95p revealed variability across the study areas; however, the variability was comparatively small at the Wadera, Bore, and Adola stations (Fig. 3f).

According to the provided data, the average annual R99P precipitation over the past 30 years was 50.6 mm. The lowest and highest recorded amounts were 38.4 mm at the Sinana station and 65.7 mm at the Adola station, respectively. The coefficient of variation (CV) for these figures was 16.4%, suggesting low variability overall. Higher R99P concentrations, between 53.8 mm and 65.7 mm annually, were noted in the Wadera, Bore, and Adola districts. Conversely, lower concentrations, ranging from 38.4 mm to 41.6 mm per year, were found in the Sinana and Dodola districts. Overall, the data indicate that while the R99P precipitation in the region shows minimal variability, some areas have notably higher levels than others. The study area displayed considerable spatial variability in R99P from 1994 to 2023 (Fig. 3g). The highest variability levels were recorded in Ginir, Wadera, Bore, and Adola, while Sinana, Dodola, Bokoji, and Kulumsa showed relatively lower variability in R99p (Fig. 3g).

The mean annual total wet-day precipitation (PRCPTOT) in the study area was 877.7 mm, with a low coefficient of variation (CV) of 15.4%. This suggests that there was relatively little variability in PRCPTOT from year to year. The lowest and highest recorded values were observed at the Sinana and Bore stations, measuring 687.8 mm and 1151.8 mm, respectively. Conversely, the spatial analysis of the total rainfall extreme indices displays that there was low variability across the stations from to 1994–2023 (Fig. 3h). The spatial analysis of PRCPTOT revealed variability within the study areas.

Fig. 3

The spatial variability of maximum annual precipitation-related agroclimatic extreme indicators.

Table 3 Descriptive statistics results summary of extreme rainfall agroclimatic indicators (1994–2023).

Extreme minimum temperature indicators

The descriptive statistics analysis revealed that the average percentage of cool nights (TN10P) across the study area was likely quite similar. The average TN10P rate was 11.0% (equivalent to 40 days per year), with a low coefficient of variation (CV = 6.8%). This suggests that the number of cool nights (temperatures below 10 °C) per year was fairly uniform throughout the region. Figure 4 shows the spatial variability of minimum temperature-related variables in the area. The spatial variability of cool nights (TN10P) appeared consistent, with minimal variation. However, Adola exhibited a lower percentage of TN10P compared to other districts, at less than 11.0% (30 days) per year (Fig. 4c).

Likewise, the annual average value of TN90P was found to be 10.9% (or 40 days per year), with a low coefficient of variation (CV = 7.2), indicating consistent values across the study area. The lowest and highest TN90P values were observed at the Bore and Adola stations, with annual percentages of 9.7% (35 days) and 12.0% (43 days), respectively. TN90P displayed a nearly uniform pattern throughout the research area. The Ginir, Dodola, and Agarfa districts experienced more than 10.9% (40 days) of TN90P, whereas the Sinana, Asela, and Bokoji districts had less than 10.7% (39 days) annually (Table 4). Similarly, the spatial analysis of extreme TN90P indices revealed minimal variability across stations from 1994 to 2023 (Fig. 4d).

The average TNn value was 3.9 °C per year, with a coefficient of variation of 67.1%. The recorded minimum and maximum TNn values were 0.3 °C at the Dodola station and 8.6 °C at the Wadera station, respectively. The districts, like Adola, Bore, and Asela, exhibited higher TNn values ranging from 5.0 to 5.7 °C/year. Conversely, districts, like Agarfa, Ginir, Sinana, and Bokoji, reported lower TNn values ranging from 1.3 to 3.5 °C per year (Table 4). There was a noticeable spatial variability of TNn across the study area, as shown in Fig. 4b showed high variability. This spatial variability in TNn could have important implications for local climate patterns and ecosystem responses. Grasping these variations is essential for creating targeted adaptation strategies and improving regional climate models. Such insights can inform policymakers and stakeholders in implementing location-specific measures to address climate-related challenges.

Comparable results were observed for TNx, with the mean annual temperature calculated at 13.6 °C, and a lower coefficient of variation (CV) of 14.4%. The recorded lowest and highest TNx values were 10.8 °C at the Agarfa station and 16.6 °C at the Wadera station, respectively. The TNx pattern was inconsistent, as illustrated in Figure. 4a. A notably higher TNx value, ranging from 13.7 °C to 16.6 °C, was predominantly observed in the southeastern Oromia districts, particularly in Adola, Ginir, and Asela. Moderate TNx values were recorded in the Bore, Sinana, and Bokoji districts, while relatively low TNx values (below 12.4 °C per year) were noted in the Dodola stations. This indicates that the study area encountered inconsistent periods of cold minimum temperatures that were below the thresholds needed for most crops.

Fig. 4

Spatial distribution of minimum temperature-related agroclimatic indicators.

Extreme maximum temperature-related indicators

The yearly mean percentage of cool daytime temperatures below the 10th percentile (TX10p) was 11.4%, corresponding to 42 days/year. This figure had a lower coefficient of variation (CV = 8.1%) and narrow yearly ranges, indicating a consistent occurrence of these cool temperatures. The minimum percentage of TX10p recorded was 9.9% (36 days) at Dodola station, while the maximum was 12.8% (47 days) at Ginir station. Figure 5 illustrates the spatial distribution related to indicators of extreme maximum temperature. The distribution of TX10p is relatively uniform across the study area (Fig. 5c). However, the Bokoji and Bore districts exhibited slightly higher percentages of TX10p, ranging from 12.4 to 12.5% (45–46 days/year). Conversely, the Ginir district regularly demonstrated a higher percentage of TX10p (12.8%), whereas the districts of Wadera, Agarfa, Adola, and Sinana showed lower percentages, ranging from 10.9 to 11.3%.

Analysis of hot daytime temperatures above the 90th percentile (TX90p) revealed a lower coefficient of variation (CV = 3.3%) and an annual mean of 10.2% (37 days). The minimum percentage of TX90p was recorded at the Adola and Bore stations, whereas the maximum was observed at the Agarfa station. TX90p is distributed almost evenly throughout the study area. However, the Sinana and Dodola districts had relatively high levels of TX90p, varying from 10.5 to 10.6% (36–37 days per year). Bokoji and Ginir received lower annual percentages of TX90p, ranging from 9.9 to 10.1% (Fig. 5d).

The mean daytime cold temperature (TXn) was recorded at 18.4 °C, with lower and higher readings of 16.9 °C and 21.2 °C, at the Bore and Adola stations respectively. These data showed a minimal coefficient of variation CV (7.2%). The spatial distribution of temperature indicators displayed valuable variations in the study area. Most stations experienced TXn values varying from 16.9 °C to 21.2 °C. However, districts like Kulumsa, Ginir, and Adola reported higher annual temperatures, with values between 18.7 °C and 21.2 °C. Spatial analysis of daytime cold temperature (TXn) showed that lower spatial variability was observed in these areas. This lower spatial variability in daytime cold temperatures may indicate more uniform climatic conditions across these regions. Such uniformity might be linked to factors such as similar topography, plant cover, and land-use patterns. Understanding these patterns of spatial variability in temperature extremes is crucial for climate monitoring and predicting the potential impacts on ecosystems and human activities. Furthermore, the area showed a temperature gradient with a mean annual extreme maximum temperature (TXx) of 32.7 °C and a coefficient of variation (CV) of 10.2%. This indicates that daytime temperature can be characterized as warm. The spatial pattern of the annual TXx and TXn indicators exhibited the same pattern, although with varying magnitudes, as illustrated in Fig. 5a and b. Notably, the geographical variability of TXx varies significantly across districts, with values ranging from 29.5 °C to 32.7 °C annually. Some districts, particularly Bokoji, Bore, Agarfa, and Sinana, recorded lower temperatures, between 25 °C and 29 °C. These spatial variations in extreme temperature indicators suggest potential impacts on agriculture, ecosystems, and human health across different districts. Areas experiencing higher maximum temperatures may face increased risks of heat stress, drought, and crop failures, while regions with lower extremes could be more suitable for certain agricultural practices or vulnerable species.

Fig. 5

Geographic distribution of maximum temperature-related agroclimatic indicators.

The Coefficient of Variation (CV) values for DTR, SU25, CSDI, and WSDI (Table 5; Fig. 6 illustrates the spatial distribution) reveal significant spatial variability in extreme temperature agroclimatic indicators across the study region (1994–2023). The Diurnal Temperature Range (DTR) exhibits moderate variability (CV = 17.2%), indicating relatively consistent differences between daily maximum and minimum temperatures (Fig. 6a), though localized deviations may arise from urbanization and land-use changes⁶⁹. In contrast, the Summer Days Index (SU25) displays exceptionally high variability (CV = 92.4%), reflecting pronounced disparities in the frequency of days exceeding 25 °C, likely driven by elevation gradients and microclimatic influences⁶⁶. For example, Adola (SU25 = 309.1) and Agarfa (SU25 = 6.3) exhibit extreme differences, potentially attributable to altitudinal variations or regional warming trends (Fig. 6b). Similarly, the Cold Spell Duration Index (CSDI) and Warm Spell Duration Index (WSDI) both show high variability (CV = 43.4% and 45.6%, respectively) (Fig. 6c&d). The CSDI’s variability aligns with global declines in cold spell persistence, though local topography such as Bokoji’s high CSDI (23.6) compared to Bore’s (11.7) may alter cooling patterns⁷⁰. Conversely, the WSDI’s variability highlights regional disparities in warming, with Ginir (12.7) and Dodola (4.6) potentially differing due to humidity or land-cover effects, consistent with observed heatwave intensification in warmer zones⁷¹.

Fig. 6

Geographic distribution of additional extreme temperature-related indicators.

Table 4 Summary of descriptive statistics for extreme temperature agroclimatic indicators (1994–2023).

Temporal trends in extreme precipitation indicators

The analysis of extreme precipitation trends from 1994 to 2023 reveals significant spatial variability across the southeastern Ethiopian highlands (Tables 5 and 7). While this study employs statistical trend analysis, recent advancements in machine learning (LSTM and hybrid wavelet-ANN models) demonstrate superior capability in capturing non-linear rainfall patterns and climate index influences 7,6. Such approaches could refine trend classification by accounting for complex interactions between local extremes and large-scale climate drivers (ENSO, IOD).

Annual total wet-day rainfall (PRCPTOT) demonstrates divergent patterns, with 33.3% of stations showing increasing trends - most notably Bokoji station exhibiting a significant increase of 3.045 mm/year/decade (p < 0.01). Conversely, 44.4% of stations displayed decreasing trends, particularly at Agarfa (−2.526 mm/year/decade, p < 0.05) and Wadera (−3.621 mm/year/decade, p < 0.01), while 22.2% of stations (Dodola and Sinana) showed no significant trend (Table 5). Bore and Agarfa districts showed the highest magnitudes of annual PRCPTOT trends, with increases of 9.27 mm and decreases of 13.20 mm, respectively (Supplementary Material Table S1). These findings align with previous studies documenting Ethiopia’s complex precipitation dynamics under climate change^14,72. The significant PRCPTOT reductions observed at Wadera and Agarfa stations (p ≤ 0.01) suggest emerging water security challenges, as decreased annual rainfall may increase irrigation demands for rainfed agriculture while reducing groundwater recharge potential. The spatial heterogeneity of trends, with some stations showing resilience (no significant change) while others exhibit pronounced drying, underscores the need for localized climate adaptation strategies. Hydrological or crop simulation modeling, could contextualize these trends by quantifying impacts on water availability or crop yields. For instance⁷, linked LSTM-predicted rainfall variability to agricultural water demand, a framework applicable to Ethiopia’s rainfed systems. These results contribute to growing evidence of climate-driven precipitation redistribution in East African highlands⁷³ and highlight the importance of station-specific analyses for water resource management and agricultural planning in the region.

Very and extremely wet days (R95P and R99P, respectively) indicate upward trends at 66.6% and 44.4% of the study stations, respectively, while the remaining stations exhibit decreasing trends. These mixed trends align with⁶, where wavelet-ANN hybrids improved seasonal forecast accuracy by disentangling non-stationary signals a method potentially useful for reconciling spatially inconsistent extremes. These findings highlight regional variability in extreme precipitation, with insignificant increases in R95p/R99p overall but significant declines at specific stations (Agarfa and Wadera for R95p). The results align with⁷⁴ for Robe Station but contrast with⁷⁵ in the Jimma Zone, underscoring localized climate drivers. Notably, the highest magnitudes of R99p trends (Dodola: +3.32 mm/year; Agarfa: −6.85 mm/year) suggest divergent hydrological impacts, potentially linked to elevation gradients or land-use changes (Supplementary Material Table S1).

The observed declines in R95p/R99p in southeastern Ethiopia contrast with⁷⁶, who reported increases in southern Ethiopia. This discrepancy may reflect physiographic influences, such as the rain shadow effect in highland areas versus lowland moisture convergence. The reduction in heavy rainfall events could reduce water availability for rainfed agriculture, particularly for staple crops like wheat, with yield losses of ~ 15–20% projected under prolonged dry spells⁷⁷. Conversely, localized increases in extreme rainfall may elevate flood risks, necessitating tailored infrastructure upgrades in flood-prone districts (Dodola).

The analysis of RX1day (maximum 1-day rainfall) and RX5day (maximum 5-day rainfall) trends in the southeastern highlands of Ethiopia reveals significant spatial variability, consistent with recent studies on intensified hydrological extremes under climate change. While 66.67% of stations showed increasing trends in RX5day, only 44.44% exhibited increases in RX1day (Table 6). Notably, Dodola recorded the largest significant rise in RX5day (4.670 mm/5days/decade), suggesting an increase in prolonged heavy rainfall events, possibly linked to enhanced moisture convergence from the Indian Ocean⁷⁸. Conversely, Agarfa and Wadera experienced significant declines in both RX1day (−3.345 and − 3.092 mm/year) and RX5day (−4.798 and − 3.092 mm/year) (Table 5), which may reflect localized drying trends or shifts in atmospheric circulation patterns⁷⁹.

The contrasting trends such as Asela’s significant increase in RX5day (2.855 mm/year) despite a decrease in RX1day (−1.026 mm/year) highlight the complex interplay between short-duration and multi-day extremes. Ensemble regional climate models (RCMs), could augment this analysis by projecting future extremes under varying climate scenarios. For example⁷, used climate indices (MJO) to validate machine learning forecasts, an approach adaptable to RCM-driven trend projections. This aligns with⁷³ projections that climate change may amplify multi-day rainfall events due to slower-moving weather systems, even where daily extremes weaken. The predominance of significant decreasing trends in RX1day (44.44% of stations, mostly at p ≤ 0.01) suggests that while some areas face heightened flood risks from prolonged rainfall (RX5day), others may experience reduced peak daily intensities, potentially affecting soil moisture and groundwater recharge¹⁶. These findings underscore the need for region-specific disaster preparedness, particularly where RX5day increases could exacerbate flood-related hazards.

For SDII, 55.6% of stations showed significant increases (Bokoji (+ 5.855 mm/day), Ginir (+ 2.963 mm/day), Dodola (+ 2.454 mm/day), Sinana (+ 2.23 mm/day), and Asela (+ 2.21 mm/day), while 44.4% declined (Wadera (− 8.03 mm/day), Agarfa (−3.66 mm/day), Adola (−3.36 mm/day) and Bore (3.27 mm/day), revealing a bipolar intensity pattern. Machine learning classification of trends (RF or LSTM) could elucidate drivers of bipolar intensity patterns, as demonstrated by⁷, for Australia’s Northern Territory, where RF robustly handled heterogeneous climate influences. These trends may stem from convective rainfall enhancement in warmer subregions⁷³ versus deforestation driven moisture declines⁸⁰, emphasizing the need for station-specific adaptation. The trends in heavy rainfall indices (R10mm, R20mm, and R25mm) in the southeastern highlands of Ethiopia reveal notable spatial variability, consistent with recent studies on climate change impacts in East Africa. Overall, 66.67% of stations showed increasing trends in R10mm, while 55.56% exhibited declines in R20mm and R25mm (Table 6). This suggests that while moderate heavy rainfall events (≥ 10 mm) may be increasing, more extreme events (≥ 20 mm and ≥ 25 mm) are decreasing in frequency at many locations.

Notably, Ginir station recorded the most significant increases in R10mm (4.568 days/year) and R20mm (4.076 days/year), aligning with findings that localized convection and topography amplify extreme rainfall in highland regions⁷⁸. Conversely, Wadera experienced sharp declines in all three indices (R10mm: −7.577, R20mm: −4.709, R25mm: −3.679 days/year), possibly due to shifting atmospheric circulation patterns linked to Indian Ocean dipole variability⁸¹. The significant decreases in R20mm and R25mm at Agarfa (−3.538 days/decade) and Sinana (−3.080 and − 2.314 days/year) further highlight regional disparities, possibly tied to land-use changes or weakening monsoon dynamics⁸².

These trends have critical implications: increased R10mm may enhance flood risks in some areas, while reduced R20mm/R25mm could signal declining water availability for agriculture, corroborating concerns raised in⁷⁶ for aridifying regions. The mixed trends underscore the need for localized adaptation strategies in this climatically vulnerable region.

Consecutive dry days (CDD) increased significantly at 77.8% of stations consistent with^14,83. Crop simulation modeling, could quantify CDD impacts on staple crops, building on⁶ hybrid models that integrated climate indices for agricultural drought prediction. The spatial heterogeneity Bore (no trend) versus Sinana likely reflects topographic modulation: highland stations (Bore at 2,800 m) retain orographic rainfall, while Sinana (2420 m) face enhanced subsidence during positive Indian Ocean Dipole events⁸⁴. CDD increases correlate with shortened growing periods, potentially reducing maize yields by 10–30% in mid-altitude zones⁸⁵. Concurrently, the pronounced rise in consecutive dry days (CDD) across most stations (+ 5.119 days/decade at Asela) mirrors global observations of lengthening dry spells, which threaten rain-fed agriculture and amplify irrigation demands⁸⁶. This is particularly critical for Ethiopia’s wheat and maize systems, where CDD prolongation has been linked to yield declines⁸⁷. Regional disparities (Bore’s stable CDD versus Adola’s increase) further underscore the need for localized adaptation planning.

Consecutive wet days (CWD) increased at 66.7% of stations (Agarfa, Wadera, Bore, Adola, and Ginir), aligning with¹⁴ but declined at Sinana and Dodola, mirroring their CDD rises. This aridification signal underscores elevation dependent resilience⁸⁸, where highlands (Bore) buffer rainfall declines.

The analysis revealed that the majority of extreme rainfall indices including SDII, R10mm, CDD, CWD, R25mm, R95p, and RX5day exhibited increasing trends across 55.55–77.78% of the studied stations. However, other indices, such as RX1day, R20mm, R25mm, R99p, and PRCPTOT, showed decreasing trends in > 55.55% of stations. While most indices indicated an overall upward trend, interannual variability led to some years with declining values, suggesting mixed precipitation patterns over the study period. These findings align with¹⁴, who reported a predominance of increasing extreme rainfall trends in Ethiopia. Similarly²¹, analyzed long-term climate data (1980–2010) and found that while some Ethiopian stations experienced rising precipitation extremes, others recorded declines. They attributed the increasing trends to global warming-induced climate variability. The observed trends highlight the need for robust climate adaptation strategies and improved water management practices to address regional disparities in rainfall patterns. Long-term climate monitoring is essential to inform agricultural planning, water resource policies, and disaster risk reduction in Ethiopia. The trend analysis for the nine stations (1994–2023) is summarized in Table 5, which categorizes the frequency percentages of eleven rainfall indices by trend type: This study provides the first 30-year ETCCDI analysis of southeastern Oromia’s highlands, filling a critical gap in regional climate literature.

Table 5 The trends of extreme rainfall indices at station level for 1994–2023 at highlands of southeastern ethiopia.

Table 6 A summary of trend results for extreme rainfall indices in southeastern ethiopia.

Figures 7 present the annual trends of recorded daily precipitation extremes over the study area. The trendlines for the precipitation indices reveal distinct patterns of change over time, shows considerable interannual variability. The outcome of the generalized regression analysis indicates significant upward and downward changes in extreme rainfall-related indicators (p < 0.05) across the climate stations, with the exceptions of RX1day, R95p, and R99p only showed a significantly decreasing trend. Agarfa and Wadera, which exhibited a significant downward trend for R95p, and Wadera, Sinana, and Agarfa which showed significant downward trends for R99p. R95p indicated a non-significant increasing trend at all sites except for Adola, Agarfa, and Wadera, where it declined significantly over the study period. Similarly, R99p demonstrated a significant negative trend at Wadera, Sinana, and Agarfa sites, with a significant change rating from − 0.26 to −2.07 mm/year (Table 5). The examination of linear trends for R95p and R99p displayed a pattern akin to that seen in Fig. 7f and g, although minor variations in magnitude were observed among the extreme agroclimatic indices in the research sites. The two indicators reached their maximum peak values in the years 1996, 1998, 2016, and 2020, while lower values were recorded from 1999 to 2012. RX1day also showed a significantly decreasing trend at Sinana, Wadera, Bokoji, and Agarfa stations the rest stations demonstrated an insignificant increasing and declining trend.

According to the regression analysis results, the Sinana and Bokoji stations indicated a significant rising trend in the number of days with precipitation exceeding R10mm, with changes ranging from 0.48 to 0.52 days/year. However, this indicator dropped significantly at the Agarfa and Wadera stations. The number of days exceeding R10mm was high in 1996, 2007, 2010, and 2014, with a significant drop from 2000 to 2005 (Figure. 7c). The R20mm index, which counts heavy precipitation days (≥ 20 mm), shows a slight declining trend indicating a gradual reduction in heavy rainfall frequency (Fig. 7d). Similarly, the R25P index (very heavy precipitation days ≥ 25 mm) exhibits a more pronounced decrease, implies high uncertainty (Fig. 7e). These trends align with regional observations where convective rainfall events may be becoming less frequent, possibly due to shifting atmospheric circulation patterns⁸⁹. The average annual rainfall for the RPCPTOT index from 1994 to 2023 is 850 mm (Fig. 7h). However, most of these 30 years had rainfall levels lower than this average. The data indicate a downward trend in annual precipitation over this time frame. This indicates that while the long-term mean annual precipitation for this index was 850 mm, most individual years from 1994 to 2023 experienced rainfall below the average level. Additionally, there seems to have been a gradual decline in rainfall over the 30 years studied. The decreasing trend over time is particularly noteworthy, as it implies that rainfall patterns in this region may be shifting, potentially owing to factors such as climate change or changes in local/regional weather systems. The annual maximum 1-day (RX1day) and 5-day (RX5day) rainfall indices (Fig. 7(a & b)) do not show a significant deviation over the 30 years.

The recorded CDD overall studied sites demonstrated a significant increase, except for Asela, which observed an insignificant upward of 0.29 days per year. CDD severity was notably high in the years 2000, 2006, and 2012, while it was low in 1996, 1998, and 2007 (Fig. 7k). Despite the considerable variation in this index shown across the study sites, the rising trends could have serious results for agriculture, water resources, and other sectors.

The distribution of extreme climate indicators collected from all sites was examined for linear trends over the studied years, revealing inconsistent changes over time. Figure 7a illustrate the linear trends of precipitation-related climate indicators. All precipitation-related extreme indices displayed significantly mixed patterns, except for RX1day, R95p, and R99p (Fig. 7afg) which had significantly negative trends. Meanwhile, CDD showed significant positive changes, indicating that climatic extremes derived from precipitation data showed the climate behavior in the region. This showed that the southeastern parts of Ethiopia have encountered an increase in CDD and negligible R95p, and R99p events over the past 30 years.

The SDII, measuring daily precipitation intensity, also displays a weak trend (Fig. 7i), suggesting a subtle decline in rainfall intensity per wet day, which could reflect changes in storm characteristics or moisture availability⁶⁶. In contrast, the CWD index, tracking consecutive wet days, demonstrates a positive trend, indicating longer wet spells over time (Fig. 7j). This increase, may signal a shift toward more prolonged rainfall periods, potentially linked to slower-moving weather systems or increased atmospheric moisture⁹⁰. Collectively, these trends highlight nuanced changes in precipitation regimes, with reductions in heavy rainfall intensity and frequency but an increase in wet spell duration. However, the low explanatory power of the trends underscores the influence of natural variability and local factors, necessitating further analysis to discern robust climate signals from noise¹¹.

Fig. 7

The Mann–Kendall test plots of linear trends for extreme minimum and Maximum rainfall-related indicators.

Temporal trends in extreme temperature indices indicators

Table 7 presents the trends of extreme temperature indices at selected stations in southeastern Ethiopia. All study sites indicated an upward trend in the annual maximum amount of daily maximum temperature (TXx) and annual maximum amounts of daily minimum indices (TNx). Out of all the stations, 66.67% and 11.11% demonstrated a significant increasing trend at p ≤ 0.01 and p ≤ 0.05, respectively for both indices. The Sinana and Ginir climate stations recorded the highest positive trend magnitude (0.17 ℃/year and 0.12℃/year) in TXx and TNx respectively (Supplementary Material Table S2). It is essential to note that all studied sites in southeastern Ethiopia indicated an upward trend in both TXx and TNx. However, the Agarfa and Asela stations indicated a non-insignificant upward trend in TXx, while the Dodola and Bokoji sites exhibited an insignificant upward trend in TNx. This warming trend aligns with findings from⁷, who observed significant increases in temperature-related indices in Australia, suggesting a global pattern of rising temperatures. The current findings harmonies with those of⁹¹, who reported a notable increase in the absolute indices of the warmest day temperature (TXx) and the coldest day temperature (TXn) in the Jimma zone from 1991 to 2020.

Likewise, 44.44% of climate stations recorded a positive trend in the annual minimum value of daily maximum temperature (TXn), while another 44.44% displayed a negative trend (Tables 7 and 8). Among all the stations, 33.33% demonstrated a significant trend in TXn. Specifically, the Ginir and Wadera stations exhibited increasing trends, while the Adola station showed a decrease. This variability in trends is consistent with the findings of⁸, who highlighted the influence of localized land cover changes on temperature patterns, as observed in their hydrological simulations. This result is similar to the¹⁴ study in Ethiopia, which found an upward trend in TXx (94.6%) and TXn (81.1%) intensity temperature indices. Similarly⁹², found that their study observed a statistically significant upward trend in both the TXx (88.5% of stations) and TXn (69.2% of stations) indices in the Upper Blue Nile region. The maximum daily minimum temperatures (TXn) exhibit varying trends across the studied stations. Specifically, Ginir, Wadera, Agarfa, and Bokoji show an increasing trend, while Adola, Dodola, Bore, and Sinana demonstrate a decreasing trend. Asela station, conversely, exhibits no discernible trend. Similarly, the trends in the TNn temperatures, which represent the annual lower value of daily minimum temperature, also differ among sites, with some indicating a decrease and others showing an increase. According to the analysis result revealed (Table 8) 66.67% of stations have an increasing trend, while 33.33% have a decreasing trend. It is important to highlight that the Agarfa station exhibits the strongest increasing trend, with a rate of 0.1 ℃ per year. Conversely, the Bore climate site has the highest magnitude of a negative trend, with a rate of −0.07℃ per year (Supplementary Materials Table (S2).

In contrast, 66.67% of the climate sites noticed decreasing trends for cool days (TX10p) and cool nights (TN10p) alike. Among these, 44.44% and 33.33% of the stations exhibited a significant downward in TX10p at a significance level of p ≤ 0.01 (Tables 7 and 8). Conversely, TN10p extreme index showed significant downward in 22.2% and 11.1% of the stations at significance levels of p ≤ 0.01 and p ≤ 0.05, respectively (Table 7). Dodola and Adola were the sole stations that indicated an increasing trend in TX10p (Tables 7 and 8). For warm days (TX90p) and warm nights (TN90p), all climate sites studied in southeastern Ethiopia exhibited upward trends (Tables 7 and 8). Furthermore, TN90p extreme indices showed a significant positive trend in 77.78% of the stations, indicating warming in the area. The highest and lowest magnitudes of TN90p trends were observed at Adola and Sinana stations, with rates of 0.60 and 0.57 days/year, respectively (Supplementary Material Table S2). Additionally, the greatest increase in TX90p was observed at Sinana station, with a rate of 0.57 days/year. These results are consistent with⁶, who demonstrated the effectiveness of hybrid machine learning models in capturing climate trends, further validating the observed warming patterns in this study. The rise in warm day frequency harms crop production⁹³.

Generally, this study’s results show an upward trend in warm days (TX90p) and warm nights (TN90p) and a downward trend in cool days (TX10p) and cool nights (TN10p) in most studied stations indicated. This aligns with the results of¹⁴ who observed that most climate stations in Ethiopia exhibited a rising trend in warm days (TX90p) and warm nights (TN90p). In contrast, 94.6% of the stations recorded a decrease in cool days (TX10p), while 72.9% experienced a decline in cool nights (TN10p). Similar to the findings of research conducted by⁹⁴, there is a growing trend in warm temperature extreme indices (TN90p and TX90p) and a declining trend in cold temperature extreme indices (TN10p and TX10p) in the Jemma sub-basin of the upper Blue Nile basin. In most Horn of Africa regions, a study by⁹⁵ indicated that the frequency of temperature indices for the warmest day, warmest night, coldest day, and coldest night has been increasing. Specifically, the warmest day has been increasing at a rate of 0.11 °C per year, the warmest night at 0.33 °C per year, the coldest day at 0.10 °C per year, and the coldest night at 0.32 °C per year.

Figure 10a-d summarizes the linear trends of warm and cold spell temperature-related indicators over annual timescales, evaluated at the 95% significance level. The cold spell duration indicator (CSDI) showed declining trends (Fig. 6d), while the warm spell duration indicator (WSDI) exhibited upward trends at most climate stations (Fig. 6c). In our case, 55.56% of the climate sites in CSDI showed a decreasing trend, and 44.44% displayed an upward trend. Agarfa, Bokoji, Bore, Dodola, and Sinana climate sites showed a decreasing trend for CSDI (Tables 7 and 8). WSDI exhibited an upward trend at all study stations in southeastern Oromia (Table 7). The findings from this study are similar to⁷⁶ who also found an upward in trend CSDI in the eastern part of Ethiopia. Moreover, 83.8% of the climate sites in WSDI demonstrated an increasing trend, while the remaining 16.2% exhibited a negative trend.

The outcome of the DTR exhibited that there was a negative trend across all stations from 1994 to 2023 except Bore and Ginir stations (Table 7; Fig. 6a). The trend examined showed that there was a significant (p < 0.01) decreasing trend at Bokoji and Adola. At Bokoji, the DTR change rate was − 0.072 °C per year. This decline in DTR is consistent with global observations, as noted by⁸, who linked such trends to land cover changes and urbanization, which alter local microclimates. A similar finding was reported by⁹⁶, who noted a significant decline in daily temperature range (DTR) across various agroecological stations in Ethiopia⁹⁷. also documented a decreasing trend in DTR in arid and semi-arid regions globally. Furthermore, the average annual DTR decreased by 0.5 to 1 °C in Sudan and Ethiopia from the 1950 s to the 2000s. Most stations exhibited an upward trend in warm temperature indices such as TXx, TNx, TX90p, TN90p, WSDI, and SU25 from 1994 to 2023. Conversely, cold temperatures extreme indices like TX10p, TN10p, and CSDI demonstrated a decreasing trend in most stations during the same period (Table 7). This aligns with previous studies that have reported an increase in warm extremes and a decrease in cold extremes in the eco-environment⁹⁸. The study stations’ trends of extreme temperature indices also indicate a warm trend in most warm extreme temperature indices, whereas cold temperature indices like CSDI, TN10p, and TX10p showed a decreasing trend in many of the years from 1994 to 2023. The change was more significant in cold extreme temperature indices (TN10p and TX10p) compared to other cool temperature indices (CSDI). Similarly⁹⁴, observed a greater rate of change in cold extreme temperature indices in the Jemma sub-basin of the upper Blue Nile basin.

Table 7 The trends of extreme temperature indices at different stations of southeastern ethiopia.

Table 8 Summary of trend results for extreme temperature indices in southeastern ethiopia.

Extreme maximum temperature-related indicators

General regression analysis indicates that all maximum temperature-related extremes, except TXn (the extreme minimum temperature) and TX10p (the number of cool nights), exhibited positive trends. The level of TX10p decreased by 0.34% (equivalent to 1.24 days) per year in the study sites, which was statistically significant (p < 0.05). The decrease rates varied, with the lowest being 0.015% (0.011 days) in Asela and the highest at 0.65% (2 days) in Ginir. As a result of global warming, the frequency of cool nights, defined as those with minimum temperatures below the 10th percentile (under 10 °C), has decreased over the last thirty years.

In contrast, TX90p (the number of very hot days) displayed positive trends overall, with significant changes noted during the study years (see Table 5). TXn also exhibited increasing trends, with notable increases rating from 0.01 °C to 0.11 °C/year observed in Wadera and Ginir (p < 0.05). The extreme minimum temperature (TXn) of less than 20 °C showed an insignificant increase in the locations of Bokoji and Agarfa. On the other hand, extreme maximum temperatures (TXx) significantly increased (p < 0.05), with deviations from 0.01 °C to 0.17 °C/year, except in Agarfa and Asela, where trends were not significant.

In conclusion, maximum temperatures in the study region have exceeded 25 °C, creating considerable challenges for agriculture, water resources, and various industries. Figure 7 displays the linear trend lines for the annual mean extreme maximum temperature indicators. The trend analysis revealed a significant decline for TX10p (p < 0.05), while TX90p exhibited a notable upward trend throughout the study period.

The highest value of TX10p was attained in 1994, 1995, 2022, and 2023, whereas its lowest levels were recorded in 1996, 2004, 2005, 2017, 2019, and 2020 (Fig. 8c). For TX90p, peak values occurred in 1996, 2004, 2005, 2020, and 2021, with its lowest levels recorded in 1994, 2006, and 2011 (Fig. 8d). The analysis of TXx revealed significantly positive trends, and patterns for these indicators consistently shifted throughout the study period. TXx reached its highest value of 32 °C in 1997, while TXn also demonstrated closely related increasing trends (Fig. 8a and b).

Fig. 8

The Mann–Kendall test plots of linear trends for extreme maximum temperature-related indicators.

Extreme minimum temperature‑related indicators

The culmination of the linear regression analysis indicates that the number of cool nights with a minimum temperature (TN10p) exhibited a downward trend, with a change ranging from 0.12 to 0.74% per year at the Asela and Agarfa stations. In contrast, TN10p showed an upward trend in the Wadera and Adola districts, with changes ranging from 0.16 to 0.22%. On the other hand, the number of warm nights with a minimum temperature (TN90p) was substantially elevated across all stations (p < 0.05). The variation in TN90p ranged from 0.01 to 0.60 days/year in the Bokoji and Adola districts, respectively, indicating a significant upward in warm nights over the research period. The time of warm nights (TN90p) exhibits the most uniform pattern of trends among all the indices assessed.

The annual mean of TNn experienced an upward change, except at the Adola and Bore stations, which saw a significant decline of 0.06 °C and 0.07 °C, respectively. The general regression analysis also observed a significant increase in TNx at all stations except Bokoji and Adola, with deviations ranging from 0.01 °C to 0.09 °C/year (Table 7; Fig. 9a). Figure 9 summarizes the linear trends of minimum temperature-related indicators over annual timescales, evaluated at the 95% significance level. Overall, the number of nights with cool minimum temperatures (TN10p) has declined, while the number of nights with warm minimum temperatures (TN90p) has upward over time. Conversely, the linear trends of average warm nighttime temperatures (TNx) were found to be significantly increasing, while nighttime cold temperatures (TNn) showed increasing and decreasing trends (Fig. 9a & b). However, the observed changes in extreme indicators were not consistent over time. The peak TN10p values were recorded in 1995 and 2016, while the minimum occurred in 2000, 2001, 2007,2008, 2011, and 2012 (Fig. 9c). The periods with the highest warm night minimum temperatures (TN90p) were 2002, 2003, 2012, and 2023, whereas the lowest TN90p values were seen in 1995, 1998, 1999, 2000, 2004–2005, 2009–2016, and 2019 (Fig. 9d). The linear trend assessment of cold night temperatures (TNn) exhibited a non-significant increasing trend. As a result, the years with the highest result for cold night temperatures (TNn) were 2002, 2006, 2017, 2013, and 2023, while the lowest rates were recorded in 1995, 1997, 2004, 2005, 2011, 2017, and 2021. Additionally, 1997 and 2023 experienced increasing warm night temperatures (TNx) alongside reducing cold night temperatures (TNn). The frequency of warm days and nights has notably been upward when compared to the 1994–2023 period, particularly in the number of nights each year that are above the 90th percentile threshold. Additionally, there has been a significant decline in the frequency of cold nights and cold days over the last 30 years. Among all the indices considered, the frequency of warm nights (TN90p) indicates the most consistent trends.

We have generated a time series for the regions highlighted in Fig. 9 (d), indicating that all areas have experienced an increase, with the most substantial rise seen in TN90p. This data displays considerable spatial coherence. Overall, throughout the entire region, there has been an increase in the frequency of warm days and warm nights, while cold days and cold nights have become less common. These findings align with results from other studies that have examined similar trends in various parts of the Ethiopia region^10,17,69. The TN90p percentile indices are more reliable over large areas because they consider the impact of local climate variations.

Fig. 9

Linear trend plots of extreme maximum temperature-related indicators.

The linear trend analysis of extreme temperature indices reveals significant patterns consistent with global climate change observations. The DTR shows a slight decreasing trend indicating a reduction in the difference between daily maximum and minimum temperatures (Fig. 10a). While the trend is weak, this aligns with global observations where nighttime temperatures have risen faster than daytime temperatures due to increased cloud cover, urbanization, and greenhouse gas emissions⁶⁹. The minimal trend strength suggests substantial interannual variability dominates over any consistent signal. WSDI exhibits a strong increasing trend, reflecting a clear rise in prolonged warm spells (Fig. 10b). This robust trend matches global findings of more frequent and longer-lasting heatwaves, driven by anthropogenic warming and atmospheric circulation changes⁹⁹. The high explanatory power underscores this as a dominant climate change signal. SU25 (days > 25 °C/year) shows a marked increase, signaling more frequent hot days (Fig. 10c). This aligns with observed global warming impacts on temperature extremes, where heat events have become more common and intense¹¹. The moderate suggests regional factors (land use) may contribute to variability. CSDI has a declining trend observed (Fig. 10d), consistent with global reductions in cold extremes due to warmer amplification and atmospheric warming¹⁰⁰.

Fig. 10

Linear trend plots of additional extreme temperature-related indicators.

Figure 10 summarizes the linear trends of warm and cold spell temperature-related indicators over annual timescales, evaluated at the 95% significance level. Figure 10b, c and d indicates the warm (WSDI and SU25) and cold (CSDI) spell extreme temperature in study years.

Spatiotemporal extreme rainfall anomaly indices

Maximum and minimum extreme rainfall anomaly indices

During the research period, all studied stations showed uneven geographical and temporal distribution of precipitation anomaly indices depicted in Fig. 11a–h. The Standardized Anomaly Index (SAI) values were calculated for the areas under investigation for each year from 1994 to 2023, following the methodology outlined in Table 2. These values represent different periods and can be seen in Fig. 11(a-h). The SAI shows the changes in drought intensity and frequency from year to year throughout the study period. Figure 11 presents the findings, which reveal significant differences in the frequency of wet and dry years in the study locations in time series and geographically. This indicates that extreme conditions were present. The annual rainfall in the study areas indicated extreme fluctuations in the frequency and intensity of dry periods. Figure 11a, which displays the RX1day index, suggests a near-complete absence of wet years. Additionally, the analysis of rainfall extremes identified a pattern of consecutive dry years following a single dry year and a similar trend after a wet year. In the areas examined, drought periods accounted for more than 50% of the 30-year study period, in all studied rainfall extreme indexes except R99p and R95p, which exhibited 36.67% and 43.33% respectively, as shown in Fig. 11. During the research period, significant anomalies in heavy (R10mm) and very heavy (R20mm) rainfall were observed in the Bokoji and Ginir districts. The indices for heavy rainfall increased notably, while those for very heavy rainfall rose particularly in Ginir. Additionally, a severe anomaly index related to dry precipitation was identified in the Agarfa and Wadera districts for heavy rainfall, and in Agarfa, Sinana, and Wadera districts for very heavy rainfall. Extremes of very heavy rainfall (R10mm) occurred in the years 2000, 2001, and 2015, whereas significant droughts were recorded in 1996 and 2014. The standardized anomaly index for R20mm displayed an uneven distribution across the study area. The standardized anomaly index for R95P was statistically significant at P = 0.05, while R99P was found to be insignificant. This study indicates an increase in the extremes of R95P and R99P over the research area throughout the study period. Several previous studies conducted in various regions of the country support these findings^101,102. For instance¹⁰², identified significant increases in R95P and R99P in Addis Ababa, Kombolcha, and Jimma from 1990 to 2010. Furthermore¹⁰³, reported noticeable variability and a slight upward trend in annual R99P in Dessie, Kombolcha, and Mekaneselam, located in Ethiopia’s northern highlands, from 1984 to 1992. In southeastern Ethiopia, annual trends showed mixed significant increases and decreases in the number of wet and dry days, maximum consecutive wet and dry spells, and rainfall intensity on wet days¹⁰¹. However, these findings differ from those of¹⁰⁴, who observed a significant increasing trend in extremely heavy precipitation events (R20mm) in Jinka, Key Afer, and Wolaita Sodo within the Omo-Gibe River Basin of southwest Ethiopia. Furthermore, an upward trend was noted in the number of very wet (R95P) and extremely wet days (R99P) across various regions, including Kenya, southern Ethiopia, and the northern and central parts of Tanzania, from 1981 to 2010 period¹⁵.

Fig. 11

Illustrates the geographical and temporal distribution of maximum rainfall extremes. A positive anomaly indicates rainfall that is wetter than their reference values (averages), while a negative anomaly indicates lower rainfall compared to those referred.

Minimum rainfall extreme indicators

Figure 12 illustrates the distribution of CDD (Consecutive Dry Days) anomaly indices over both space and time across the study area. Over the past 30 years, CDD anomalies have become more pronounced in all sections, with notably extreme dry periods occurring in 2005, 2017, and 2020. However, highly positive CDD anomalies were observed throughout the entire study region, except for the Agarfa and Bore districts (Fig. 12a). Standardized anomaly indices were employed to analyze the frequency of precipitation variables, such as CDD and CWD (Consecutive Wet Days), as shown in Fig. 12a & b. The CDD values derived from 30 years of recorded data show the number of drought days with less than 1 mm of rainfall, while the anomaly index highlights deviations from the average duration of dry spells. In contrast, the anomalies for CWD and the Simple Daily Intensity Index (SDII) displayed uneven spatial distributions throughout the study area (as shown in Fig. 12b and c).

Fig. 12

Illustrates the spatial and time series distribution of minimum precipitation extremes. A positive anomaly indicates rainfall that is wetter than their reference values (averages), while a negative anomaly indicates lower rainfall compared to those reference values.

Maximum temperatures extreme anomaly indices

Figure 13 summarizes the spatial and temporal distribution of temperature extreme anomaly indices tested in this study. The spatial distribution of extremely maximum temperature anomaly indices shows a highly uniform pattern in TXx and TX90P across the study area. In contrast, the TXn and TX10P indices demonstrate an uneven distribution. The TXn index primarily indicates moderately cold conditions in the majority of the study area, except small regions in the southwest and central northeastern sections. Meanwhile, the TX10P index has had a significant, extreme negative impact on the northeastern area, while the southeastern area has experienced extremely positive effects. The southeastern region has seen a moderate impact over the past 30 years, particularly during the years 1994, 1995, and 2023, when it was notably affected. The extreme cold conditions represented by the TX10P index have particularly impacted the Sinana, Agarfa, and Ginir districts during this study period, as shown in Fig. 13a and c. This research revealed a significant increase in the coldest daytime temperature anomaly index, with a p-value of 0.01, as illustrated in Fig. 13b and c. Additionally, the Extreme Maximum Temperature Anomaly Index (TXx) and the TX90P were found to be evenly distributed across the study area. From 1994 through the years 2014 to 2021, TXx consistently exhibited warm and moderately warm temperatures in all studied districts. Both extreme and moderate variations of TXx have been prevalent throughout the study period, with this pronounced variable showing a significant increase over time, as shown in Fig. 12a. Supporting these findings²¹, observed a significant increase in the coldest daytime temperatures (TXn) in Addis Ababa; however, the increases in Kombolcha and Jimma were not statistically significant. There has been a significant upward trend in daily maximum temperatures (TXx) from 1981 to 2010. This trend is consistent with the findings of⁷¹, who observed a notable increase of up to 2.4 °C in extreme daily maximum temperatures across Ethiopia, northern Kenya, and the southern regions of Tanzania. After 1999 and 1995, the annual averages of TXn in Ethiopia and Kenya have been consistently higher than the long-term average. Furthermore, similar patterns of rising annual temperatures and increased occurrences of extreme precipitation have been documented in most parts of the Jemma Sub-basin in Ethiopia’s Upper Blue Nile Basin⁹⁴.

Minimum temperature extreme anomaly indices

The geographical distribution of the hottest night temperatures in the study area has varied significantly over time (P = 0.05), as illustrated in Fig. 13. Warm (TNx) and very warm (TN90p) night temperature anomaly indices have shown a strong positive trend over the last 30 years, indicating an increase in the warmest minimum temperatures at night in the region. However, a negative anomaly for cool night temperatures (TN10p) was observed, except in the Adola, Bore, and Wadera districts. In contrast, positive anomaly indices for the coldest night temperatures (TNn) were recorded, except in the Adola and Bore districts, as well as in Wadera. Extremely high warm TNx values were recorded only in the years 1997, and 2023, whereas moderate cold TNx values occurred in 1994 and 1996 (see Fig. 13a).

The uneven distribution of the coldest minimum temperature (TNn) during the study period is depicted in Fig. 13b. A strong positive anomaly in the coldest night temperatures was prevalent across the study locations. The inconsistent occurrence of extreme temperatures, especially the high prevalence of strong negative anomalies in coldest night temperatures, suggests potential impacts on local ecosystems, agriculture, and human activities. In particular, the Bore and Adola areas (see Fig. 13) experienced extremely cool TNn values, while the Wadera station was slightly cooler than average. These findings align with those of⁷⁶, who reported substantial reductions in the coldest minimum temperatures in western Ethiopia, Afar Zone 2, Somalia, and the eastern part of Tanzania. The warmest night temperatures were recorded in 2002, from 2006 to 2008, 2013, and 2023, while the coldest TNn values occurred in the years 1995, 2018, and 2022. In contrast to TNx, the anomaly index showed a significant increase during the study period, indicating that occurrences of the coldest nighttime temperatures above average were more frequent than those below average, as illustrated in Fig. 13a. Additionally, this variable has declined over time. Similar findings were noted in the Jimma zone, where substantial increases in the coldest night temperatures (TNn) occurred, while non-significant increases were reported in Addis Ababa and Combolcha from 1980 to 2010²¹.

Fig. 13

Illustrates the geographical and temporal maximum, minimum, and additional distribution of temperature extremes. A positive anomaly indicates a temperature warmer than their reference mean, while a negative anomaly indicates a cooler temperature than those reference values.

Changes in extreme temperature

The present study demonstrated continuous warming over the study areas from 1994 to 2023. Recent studies indicate that there was extreme warming in different areas around the globe^{76,94,105,106,107}; which is in line with the results obtained in this study. Machine learning approaches, such as LSTM and hybrid wavelet-ANN models, could further refine temperature trend analyses by capturing non-linear interactions between local warming and large-scale climate indices (e.g., ENSO, IOD), as demonstrated in⁷ for rainfall-temperature linkages in Australia. A recent study indicated that the mean, minimum, and maximum temperatures were linearly increasing by 0.046, 0.067, and 0.026 ℃ per decade in southeastern Ethiopia from 1901 to 2014¹⁰⁸. Similarly, there was also a significant increase in the annual maximum temperature in Tana areas in Ethiopia, with maximum and minimum positive slopes of 0.03 and 0.05 0 C per year, respectively¹⁰⁹. Hydrological or crop simulation modeling, could quantify warming impacts on crop yields, building on⁶ framework linking climate indices to agricultural outcomes. It has also been stated that there was a significant increase in the inter-annual and seasonal trends on maximum and minimum temperatures in different parts of Ethiopia⁷⁶.

Similarly, the results also indicate that warming indicator values of extreme temperature indices such as warm days, warm nights, and WSDI revealed significant (p < 0.01) positive trends throughout the study areas. Figure 13a-d also shows the relationship between spatial and temporal warm and cold spells of the study area. In line with the results obtained in this study, a linear increasing trend in WSDI, warm days, and warm nights, coupled with up to 22% increase in warm days; and a decreasing trend on cold days and cold nights were observed in different parts of Ethiopia⁷⁶. Similarly, there was also a negative trend on cold nights, cold days, and CSDI; and a positive linear trend on warm days, warm nights, and WSDI in different parts of Ethiopia from 1967 to 2008⁹⁸. On the other hand, increasing trends in warm days and warm nights varying from 0.38 to 0.71 days per year, and 0.31 to 0.62 days per year, respectively, were observed over the last three decades in the lowland areas of Tigray⁷⁷. On the other hand, the trends on the DTR also revealed that there was a linearly decreasing trend over the study areas with magnitude varying from − 0.001 ℃ at Sinana to 0.072 ℃ at Bokoji in the time span, while there was a high falloff during the studied time (Fig. 10a). The increase in warm days and warm nights might be due to global warming, deforestation, and/or urbanization. Ensemble regional climate models (RCMs), combined with machine learning classification of trends (RF or LSTM), could disentangle anthropogenic vs. natural drivers, as proposed in⁷. As¹¹⁰ showed that hybrid ANN-wavelet models better resolve non-stationary signals; applying similar preprocessing to temperature data could improve DTR trend attribution, particularly in regions with heterogeneous land cover. In the majority of the study areas, deforestation was extensively increasing due to expansions for arable land and resettlements. In line with this, there is strong evidence that anthropogenic factors affect warm days and warm nights⁷³. Increasing warming indicator values might negatively affect the growth and development¹¹¹, and grain yield of different crops^112,113, and increase in distribution, and severity of infectious diseases like Leishmaniasis^114,115 and malaria¹¹⁶ during the main cropping season in the study areas. Subsequently, such circumstances might affect the socio-economy and multifaceted challenges of the livelihood in the study area.

Changes in extreme rainfall

The current study presented that there was high rainfall variability in the highland areas of the southeastern Oromia, with no uniform pattern. The observed variability aligns with⁶, where wavelet-based hybrid models (W-ANN) improved seasonal rainfall forecasts by addressing non-stationarity a method applicable to Ethiopia’s spatially divergent trends. The linear trend analysis on PRCPTOT, Rx1Day, R20mm, R25mm, and R99p showed that there was a negative trend in the majority of the stations from 1994 to 2023 (Table 7). There was a decreasing trend in the annual rainfall, PRCPTOT, Rx1Day, R20mm, R25mm, and R99p across all stations. Trends on rainfall showed that there was a decreasing tendency over the past three decades in the southeastern highland of Oromia¹¹⁷. Machine learning classification of trends (LSTM’s temporal pattern recognition in⁷ could enhance predictability of extreme rainfall shifts, particularly where traditional statistical methods show limited explanatory power. Recent studies confirm that there was seasonal and inter-annual rainfall variability in different parts of the East African region including Ethiopia^{15,76,107,118,119}. Similarly, trend analysis in the annual rainfall showed that there was strong variability with a coefficient of variation from 33.77 to 233% from 1981 to 2010 in different parts of the country^119,120. It has also been stated that there was a significant reduction in the annual rainfall variability in the Oromia national regional state between − 5.92 to −9.74 mm per year¹⁰⁹, which is similar to the results obtained in this study. In line with the results obtained in this study, research studies indicated that there was a decreasing trend in maximum Rx1day rainfall in different agroecologies in Ethiopia^94,98. In this study, we found that there was no uniform rainfall variability in the study areas in the last three decades, which is similar to the changes and variability on the maximum Rx1day and Rx5day rainfall indices; with negative trends on Rx1day at Sinana, Wadera, Bokoji and Agarfa and positive trends at Bore, Adola, Ginir, and Dodola; whereas, the negative trend was observed on the maximum Rx5Day at Adola, Wadera, and Agarfa, the rest stations showed a positive trend. Overall, the results on extreme temperature and rainfall showed that the southeastern Oromia experienced annual variability, coupled with extreme warming, which could substantially affect crop production and food security in the study areas.

As¹¹⁰ demonstrated for rainfall, our temperature extreme predictions (rising TN90p) could inform agricultural adaptation strategies, such as crop scheduling or heat-resistant varietal selection, in vulnerable regions like southeastern Ethiopia. Future research could integrate ensemble regional climate models (RCMs) with wavelet-based artificial neural network (W-ANN) hybrid approaches⁶ to improve projections of localized climate extremes under changing climatic conditions, thereby addressing the need for more advanced modeling methodologies.

Summary and conclusion

The study reveals alarming shifts in climate extremes across southeastern Ethiopia from 1994 to 2023, with critical implications for Oromia’s agricultural systems. As the first multi-index trend analysis focusing on southeastern Oromia’s wheat belt, this research provides novel insights into compound climate risks that were previously unquantified at this scale. Analysis shows 77.8% of stations experienced increased consecutive dry days (CDD), particularly Adola, Asela and Sinana with 4.7–5.1 additional dry days per year (p < 0.01), while 66.7% of stations simultaneously recorded more intense rainfall events (R10mm-R25mm). These contradictory trends combining prolonged droughts with episodic deluges are further exacerbated by universal warming patterns, including 100% of stations showing increased warm nights (TN90p) and extended warm spells (WSDI). The CDD increases are particularly concerning in study zones where parallel warming accelerates evaporation, threatening to shorten growing seasons and deplete water reserves, seen in Wadera’s significant precipitation decline. For agricultural planners in these areas, we recommend immediate implementation of drought early warning systems paired with adjusted planting calendars. Meanwhile, stations like Ginir face opposite challenges from intensified wet spells of heavy rainfall elevating flood risks. Here, adaptation should focus on improved drainage infrastructure and flood-resistant crop varieties, with nitrogen application rates optimized for key cereals to account for increased nutrient leaching. These findings demand immediate, targeted responses including: climate-smart agriculture with drought-resistant crops for CDD-affected areas like Sinana and improved drainage for flood-prone zones like Ginir; precision water management combining micro-catchment systems for drought zones with large-scale storage for variable rainfall areas; and enhanced climate services with station-specific early warnings that integrate traditional forecasting knowledge with modern meteorological data. Policy implementation should prioritize CDD-affected woredas for drought programs while directing flood resources to high R25mm zones, with irrigation plans addressing concurrent CDD increases and rainfall declines stations. For Oromia’s agricultural stakeholders, these results provide a clear roadmap for climate resilience planning. The regional Bureau of Agriculture should establish climate risk zones based on station-specific trends, Prioritize CDD-vulnerable areas for water harvesting investments, and Develop tailored extension packages for flood-prone versus drought-prone districts. The integration of these climate trends into Ethiopia’s National Adaptation Plan is crucial to safeguard agricultural productivity, particularly through climate-smart crop zoning that accounts for the observed increase in warm days and decrease in cold nights. To strengthen policy relevance, we recommend establishing a technical working group comprising climatologists, agronomists, and regional policymakers to translate these findings into actionable adaptation plans. Without such spatially-explicit interventions, the compounding effects of drying trends, erratic rainfall and warming threaten to reverse agricultural progress in this vital breadbasket region, necessitating urgent coupling of these biophysical insights with vulnerability assessments to ensure equitable adaptation.