Spatiotemporal Variability and Trends in Rainfall and Temperature in South Ethiopia: Implications for Climate Change Adaptations in Rural Communities

Climate change is an environmental challenge for rural communities that rely heavily on rainwater-based agriculture. Te main goal of this study is to investigate spatiotemporal variability and trends in rainfall and temperature in southern Ethiopia. Extreme temperature and rainfall indices were computed using the ClimPACT2 software. Te detection and quantifcation of trends in rainfall and temperature extremes were analyzed using a nonparametric modifed Mann–Kendall (MMK) test and Sen’s slope estimator. Results indicated that the mean annual rainfall has a declining trend at Boditi School and Mayokote stations with a statistically signifcant amount at magnitudes of 0.02mm and 0.04mm, respectively. Te highest average monthly rainfall in the catchment was observed in the months of April, May, June, July, and August up to maximum rainfall of 117.50mm, 177.43mm, and 228.84mm in Bilate Tena, Boditi, and Mayakote stations, respectively. On a seasonal scale, rainfall in Bilate Tena station was highly variable in all months, ranging from 49.54% to 126.92%, and three seasons except spring which showed moderate variation at 40.65%. In addition, the three locations over the catchment exhibited varied drought signs such as severe (1.28 < SRA < 1.65) and extreme drought (SRA > 1.65). Te temperature indices, on the other hand, exhibited a warming trend over the catchment which was observed through an increased annual number of warm days (TX90p) and warm nights (TN90p) ranges from 0.274 to 6.03 and 0.274 to 3.16, respectively. Te annual maximum value of the daily maximum temperature (TXx) ranges from 30.10 to 33.76 ° C in the three agroecological zones and showed low, medium, and high values in Dega, Woyna Dega, and Kola agro-ecologies, while the annual maximum value of the daily minimum temperature (TNx) ranged between 17 and 17.44 ° C at Dega and Kola, respectively. Terefore, based on trends in rainfall variability and persistent temperature rise, appropriate adaptation strategies should be adopted.


Introduction
Climate change is one of the global environmental changes which closely relates to the agricultural sector.Over the past few decades, the global climate has shown unprecedented change and continues to change in the future at an unparalleled rate [1,2].According to the latest report of the Intergovernmental Panel on Climate Change [3], the global average annual surface temperature has increased by 0.3 to 0.6 °C since the late 19th century and is expected to increase by 1.0 to 3.5 °C over the next 100 years.Nowadays, changes in the climate are recognized as one of the greatest environmental challenges of our time, which calls for a concerted efort by the international community to develop diverse adaptation and mitigation plans.
Climate change is a global threat causing severe, crosssectoral, long-lasting, and, in some cases, irreversible impacts on agriculture [2].However, the countries of the world do not withstand equally in front of climate change challenges.Climate change is leading to changes in the frequency, intensity, spatial extent, and time scale of extreme climate conditions, which can lead to unprecedented extreme events [4][5][6][7].Low-income and fragile countries, including Ethiopia, are heavily dependent on rainwater-based agriculture, which is highly sensitive to climate change and has been hardest hit [8].
Among developing countries, Africa has several climate change hotspots, where the physical and environmental impacts of climate change cross with a large number of poor and vulnerable communities [9,10].Te IPCC Assessment Report Six (AR6) states that sub-Saharan Africa is still experiencing a warming trend, with an average rate of change of roughly +0.3 °C/decade from 1991 to 2021, compared to +0.2 °C/decade from 1961 to 1990, 0.04 °C/decade between 1931 and 1960, and +0.08 °C/decade from 1901 to 1930 [11].On the other hand, rainfall in the area varies greatly over both space and time due to a variety of complicated topographical factors and circulation patterns.Climate change adversity coupled with low adaptive capacity poses unanticipated threats to the majority of people in Africa, which afects the most vulnerable people by increasing food insecurity, population displacement, stress on water resources, and disease outbreaks [12][13][14].
As confrmed by previous studies, climate change in Ethiopia is fast-tracking and will lead to wide-ranging shifts in climatic conditions [15,16].Te country's long-term climate data show that Ethiopia's climate, especially the distribution of rainfall and temperature, which has been relatively static for many years, has become very dynamic and unpredictable [17][18][19].According to a report by the World Bank Group, over the past decades, the temperature in Ethiopia has increased by about 0.2 °C per decade, while the average annual temperature is 22.6 °C, with monthly temperatures ranging from 20.9 °C to 23.9 °C [19].Despite the fact that the general trend for rainfall is still quite consistent when compared to the yearly average [20], there is a signifcant amount of variation in rainfall throughout both time and space.Te average annual rainfall in Ethiopia is 815.8 mm, with a range from 0 mm to more than 4,000 mm per year [2].Tis illustrates a high degree of regional variability and fuctuation over time.
Te global climate model (GCM) predicts that the national average annual temperature will increase by 3.1 °C by 2060 and 5.1 °C by 2090, and rainfall will decrease from the annual average of 2.04 mm/day ) to 1.97 mm/day (2070-2099).
Along with predicted rainfall, the future temperature change is one of the most important indicators of ongoing global climate change [21].Changes in temperature, which cause alterations in rainfall patterns, are important for water resource management and water-related natural hazards [22].
In Ethiopia, several studies have assessed the variations in rainfall and temperature over a wide range of geographic areas and time scales [23][24][25][26][27][28].Despite the fact that these studies ofered an important foundation for understanding climate trends and variability, the majority of the previous studies were limited to data from a few selected meteorological stations.Te regional weather stations have encountered a number of difculties when attempting to acquire data, including low data quality (discontinuous data), lack of availability and accessibility, and unevenly distributed stations [29,30].Instead of station-based temperature and rainfall data, which can be obtained from local weather stations along with the satellite-based data, which can be downloaded from the Coordinated Regional Climate Downscaling Experiment (CORDEX) Africa programs.Using station-based data and household surveys at the national and subnational levels, other research studies have looked at the relationship between long-term trends in climate parameters; however, these studies may not fully describe the situation at the local level [23,[26][27][28].
Te local level climate analysis on which this study is centered focused on the seasonal rainfall variability, including onset, cessation, and length of the growing season, and extreme events such as drought statistics generated for both long and short rainy seasons, rather than long-term trends, provides helpful insights into the current trajectory of various climate variables, particularly on shorter-term planning horizons.To characterize and comprehend the present and past climatic conditions, a detailed climate study at a higher temporal and spatial resolution is necessary.Tis will make it possible to ofer data that will correctly guide the creation of programs for adaptation and mitigation.Tis type of research is, therefore, topical and appropriate for identifying regional and local patterns of climate change, developing treatments, and disseminating useful information about climate adversity.
Detecting area-specifc spatiotemporal trends in meteorological time series is important for understanding the evidence and efects of climate change adversity across the country.Tus, our study investigates spatiotemporal variation and trends in daily, monthly, and yearly rainfall and temperature (maximum, minimum, and mean values) in the Wolaita zone, southern Ethiopia, for the year 1990-2021.In order to achieve the goal, the local level rainfall (annual, seasonal, and daily) as well as the daily maximum and minimum temperatures was hypothesized to be neither regularly distributed nor independent of the mean and standard deviation.Tis study is, therefore, very crucial to understand the spatial and temporal variations of climate change within a study area and its efect on farmers life.Te study will also help to monitor and design natural resources management systems, such as environmental planning, land use planning, water resources planning, and irrigation planning while implementing sustainable agricultural development in the area.

Area under Study.
Te study area is situated in Wolaita zone which is one of the administrative areas of the Southern Nations, Nationalities, and Peoples Regional State (SNNPR) of Ethiopia.Tis zone is located 390 km southwest of the country's capital, Addis Ababa, along the main road from Shashamane to Arba Minch [29].Figure 1 depicts the study area, Wolaita zone, which is astronomically positioned between 6.4 °-7.1 °N latitude and 37.4 °-38.2 °E longitude.
Wolaita zone covers a total area of 4,511.7 km 2 and is structured with 16 districts and 6 towns.Among the 16 districts of the Wolaita zone, this study was conducted on 2 Advances in Meteorology three sample areas under the Bilate Wolaita subwatershed, which are the main hotspot areas of the zone in terms of climate change extremes, food security, and land degradation [5].Te selection criteria for the study district were the predominant agroecology of the districts, the presence of meteorological stations, and agroecological location.
According to Ethiopia's classifcation of agroecological zones, the region is mostly characterized by mid-highland agroecology (1500-2300 m.a.s.l.).As per Ethiopia's traditional agroecological classifcations, the study area is divided into three zones, with Woyna Dega making up the majority of the overall area at around 56%; the remaining 35% and 9% are referred to as Kola and Dega, respectively [30] (Table 1).Te study extracted three meteorology stations based on the Ethiopian traditional AEZ grouping approach by employing latitude and longitude, elevation, and patterns of rainfall and temperature to represent the alpine vegetated zone known as "Dega," the temperate zone known as "Woyna Dega," and the hot zone known as "Kola" agroecological zones (AEZs), respectively.
It was quite challenging to forecast the patterns of rainfall and temperature in the study area.However, according to the Wolaita Zone Plan Department's annual report, the study area has two main rainy seasons: the long rainy season (Belg), which lasts from February to May, and the short rainy season (Kiremt), which lasts from June to September.According to the authors in [16], in the Wolaita zone, the average annual maximum and minimum temperatures range from 31.4 °C to 15.2 °C, while the zone's total mean annual rainfall ranges from 1000 mm to 1270 mm, with August often having the highest rainfall records.As a result, the climate change extremes severely afected the study area, particularly for people whose livelihoods heavily depend on subsistence agriculture.

Te Existing Data from Tree Selected Meteorological
Stations.To represent variation over space and time, three meteorological stations' recent records of daily rainfall and temperature data from 1990 to 2021 were used.Te time series data for the chosen stations were supplied by the Ethiopian National Meteorological Agency, which is in task of gathering and studying meteorological data on a countrywide scale.As station-based data alone has low data quality and measurement mistakes in terms of consistency and precision [16,30], this study used integrated quality-controlled station data from the national observatory network with locally calibrated satellite data.Due to the small number of stations, regional variability of the agroecology, and data gaps, better spatiotemporal climatological information is urgently needed to support climate services and related decision-making processes.To fnd mistakes and outliers, manual and automated data checks were also carried out.[32].All RCMs simulate the fundamental climatic variables, including daily precipitation, maximum (T max ), and minimum (T min ) surface air temperatures; however, biases, including data irregularities, outliers, and missing values, exist across the models, making bias correction necessary before using the climate data for any analysis.
Four regional climatic models, namely, the Regional Atmospheric Climate Model Version 22 (RACMO22T), Rossby Center Atmospheric Version 4 (RCA4), Regional Model (REMO2009), and Community for Limited-Area Climate Modeling (CCLM4), were used in this study in order to better describe the spatiotemporal characteristics of climate change extremes.However, there are many criteria by which a subset of models can be selected, for instance, based on the skill in reproducing past climate [33] and the range of projected climate changes [34].Others have implemented automated algorithms based on the clustering of climate extreme indices to identify a representative subset of climate models.Terefore, no single climate model can capture the entire range of possibilities for all factors, regions, or seasons.Working with a restricted selection of models can result in inconsistencies in climate change signals [35].Te simplifcation of extremely complex atmospheric physics in GCMs results in intrinsic errors and uncertainties.Due to the correction of individual errors, they discovered that a multiple model ensemble was a better ft for the situation than individual GCMs [36].
Te climate models chosen for this study were based on earlier research in the catchment that outperformed other climate models [37].Te previous knowledge of selecting climate models from multiple GCM-RCMs is better to limit the number of climate models.Te hydrostatic RCA4 model, which can generate data with a variety of horizontal resolutions but only with 0.44 grid resolution, was employed in this study.Te study also utilized the ffth-order upwind nonhydrostatic regional climate model, Climate Limited-Area Modeling Community Version 4 (CCLM4), which ofers a more accurate representation of the spatiotemporal variability of precipitation and temperature [38].Te Regional Model (REMO2009), a three-dimensional hydrostatic atmospheric regional climate model developed by the Max Planck Institute of Meteorology, was also utilized [38].Moreover, the Regional Atmospheric Climate Model Version 22 (RACMO22T), the hydrostatic KNMI regional climate model and the latest iteration of the RACMO2, was also used.Climate variables are simulated by all regional climate models, but their magnitudes vary.Te climate models used in this study revealed variations in capturing the observed rainfall and temperature in previous studies [37].Te models utilized in this study had a grid resolution of 0.44 by 0.44.

Bias Correction of Climatic
Variables.Bias correction was applied to reduce overestimation or underestimation of the mean of downscaled variables (i.e., temperature and precipitation).Bias correction factors were computed from the statistics of observed and historical variables.Te power transformation/nonlinear method was used to correct both the mean and variance of precipitation [37].
(1) Precipitation.Prior to using the climatic data for analysis of each station, the bias correction for the temperature and precipitation data was performed.Te CV and mean are both corrected using a power transformation approach.Using the formula given in [39], each daily precipitation amount P was converted to a proper P * using the following equation: ( Te parameters a and b were determined for every month of the year, including data from all years available. (2) Temperature.When monthly mean values are included, this method is capable of perfectly adjusting climatic infuence.For each station, the corrected daily temperature T * is given as follows: where T obs is the observed daily temperature from the National Meteorological Agency (NMA) dataset and T rcm is the uncorrected daily temperature from RCP.  Advances in Meteorology (i) Estimating and flling missing data: Inverse distance or weighting method is most commonly used for the estimation of missing precipitation [40][41][42].Rainfall data of interpolation using inverse distance weighting (IDW) can obtain more accurate results [43][44][45][46].Te rainfall at a station was estimated as a weighted average of the observed rainfall at the neighboring stations.Te weights are equal to the reciprocal of the distance or some power of the reciprocal of the distance of the estimator stations from the estimated stations.
where P * is the rainfall of the missing station, n is the number of index stations, and where D i is the distance between the estimator station and the estimated station.Ten, the estimator station's coordinates are x and y, whereas the estimated station's coordinates are x i and y i .In order to test consistency for some stations, the nondimensional sign of rainfall data was computed by dividing monthly time series data by the average rainfall amount of the respective month which is calculated as follows (Figure 2): where p i is the nondimensional value of rainfall for month i, p i is the over monthly rainfall at the station i, and p is the over yearly rainfall of the station.(ii) Checking consistency of the rainfall data: Inconsistency may result from the unreported shifting of the rain gauge in the gauging station.Doublemass curve analysis was used to adjust inconsistent data.Te change in the regime of the curve of the inconsistency was adjusted by using the following equation (Figure 3): where P a is the adjusted precipitation, p o is the observed precipitation, a o is the slope of graph at time p o is observed, and b a is the slope of graph to which records are adjusted.(iii) Accuracy of rainfall simulations from climate models: In this work, the outputs of the model simulation of rainfall data were assessed using statistical techniques such as P Bias , RMSE, Correl, and coefcient of variation (CV).Percent of bias can be estimated by using the following equation: where P Bias is the percent of bias; R obs is the average observed rainfall data; R RCM is the rainfall data over the catchment; and R RCM is the average rainfall data.
Te coefcient of variation (CV) can be estimated by using the following equation: where CV is the coefcient of variation in %; δ is the standard deviation; R is the rainfall over the catchment; and R is the average rainfall over the catchment.Te root mean square error of a model prediction with respect to the estimated variable R RCM is defned as the square root of the mean squared error which is given as follows: where RMSE is the relative mean square error in mm year −1 ; R RCM is the rainfall data over the catchment; R obs is the observed rainfall data over the catchment; and N is the number of years that rainfall observed.Correlation coefcient (Correl) can be estimated as follows: where correl is the correlation coefcient (−); R RCM is the rainfall data over the catchment; R obs is the observed rainfall data over the catchment; n is the number of observations; R obs is the average observed rainfall data; and R RCM is the average rainfall data of the climate.

Data Analysis.
When bias was corrected, the actual analysis of the climatic data was carried out using parametric and nonparametric trend tests, which measure the magnitude of the trends in the extremes of climate change.Te R and R Studio software packages (version 4.2.2) were used for data analysis.Several actual climate data analysis techniques have been developed for the analysis of rainfall and temperature, often falling under the category of variability and trend analysis.Te analysis of variability involves using the coefcient of variation (CV) and the percentage deviation from the mean (anomaly), while nonparametric MMK trend test along with Sen's slop estimator, which is more robust for trend detection in time series, was used [46,47].Te following indices and tests were conducted to analyze the spatiotemporal variations in rainfall and temperature within the study area.

Advances in Meteorology
(1) Analysis of Rainfall Variability (i) Te coefcient of variation in rainfall (CV): Te coefcient of variation (CV) is a widely used technique to analyze interannual variability of rainfall computed as the ratio of the standard deviation to the mean value over the given period.Te advantage of using the coefcient of variation (CV) is that it is a useful statistic for comparing the variability of one data series with another, even when the means are signifcantly diferent.However, CV is sensitive to small mean values and is unable to determine mean intervals, and it is not helpful to analyze the rainfall variability in a specifc season.Despite this limitation, CV can be used to calculate the annual and interseason variability in rainfall.Te greater variability is indicated by a higher coefcient of variation (CV), and vice versa: where CVx is the coefcient of variation for the given month or year, σx is the standard deviation for the given month or year, and μx is the mean value for the given month or year.
Tis study used the classifcation of CV in [48], which classifes CV values as less variable for values less than 0.20, moderately variable for values between 0.20 and 0.30, and highly variable for values greater than 0.30.(ii) Standardized rainfall anomaly (SRA): SRA is commonly used as a simple index to characterize drought at diferent time scales or to identify abnormal wetness or dryness [49].Te standardized rainfall anomaly is calculated as the diference between long-term mean annual rainfall and observed annual rainfall to the ratio of standard deviation which is given as follows: where SRA is the standardized rainfall anomaly, p t is the annual rainfall in the year, p m is the long-term mean annual rainfall for the study period, and σ is the standard deviation of annual RF for the study period.Te classifcation of drought severity given in [50], which is extreme drought (SRA > 1.65), severe drought (1.28 < SRA < 1.65), moderate drought (−0.84 > SRA > −1.28), and no drought (SRA > −0.84), was adopted in this study.Advances in Meteorology (iii) Standardized anomaly index (SAI): A standardized anomaly index (SAI) is a commonly used index for regional climate change analysis [51].For each station, the series of mean annual temperature, mean annual minimum temperature, and average annual maximum temperature were analyzed to identify variations using a standardized anomaly index.Station temperature is expressed as a standardized departure x i from the long-term mean and was calculated as follows: where r is the mean temperature of the year, r i is the long-term mean, and σ is the standard deviation of the annual mean temperature for the long term.(iv For this purpose, a modifed variance of S, designated as Var(S) * , was computed by using the following equation: where n * is the efective sample size and n/n * is the ratio computed directly from the following equation: where "n" is the number of observations, "n * " is the efective number of observation counts for autocorrelation, and "k" is the autocorrelation function for the rank of the observations.(ii) Sen's slope estimator: One of the most common models for identifying linear trends is simple linear regression.But this approach needs to be predicated on residual normality [54,55].Tus, Sen's slope estimator is found to be a powerful tool to develop linear relationships.Sen's slope has an advantage over the regression slope, in that raw data series errors and outliers do not have much efect.In the nonparametric statistical tool, the magnitude of the trend that exists in the time series is estimated by Sen's slope estimator [56].So, this study used Sen's slope to estimate the magnitude of the trends in the time series data in three selected stations.When the trend can be considered to be linear and equal, Sen's method can be utilized by using the following equation: where f (t) is a continuous monotonic increasing or decreasing function of time, Qt is the slope, and β is a constant.
Te slopes of all data value pairs were calculated to get the slope estimate Q:

Advances in Meteorology
where x j and x k are the data values at times j and k (j > k) and N is computed as follows: where n is the number of periods, the N values of Q i were ordered from smallest to largest, and the median slope or estimate of Sen's was calculated by using the following equation: A positive value of Q i indicates an increasing trend and a negative value of Q i gives a decreasing trend in the time series [57].

RCM Model Performance Evaluation Result.
Previously, for any impact assessment in the context of climate change, the accuracy of climate models in terms of statistical measures such as correlation (−) was used to assess the relationship between observed and modeled rainfall with a value of 1 and 0, suggesting a perfect linear relationship; bias indicates a systematic error in rainfall.A value of zero shows no systematic diference between simulated and observed rainfall amounts, whereas a large bias indicates that the RCM rainfall amount largely deviates from the observed rainfall amount.Positive bias indicates underestimation, while negative bias indicates overestimation.A root mean squared error (RMSE) number close to zero denotes the RCM model's optimal performance; the coefcient of variation (CV) and its ability to reproduce annual rainfall cycles must be evaluated.Te observed catchment-averaged annual rainfall amount was 1185.28 mm year −1 .Te accuracy of rainfall for the overlapping period from four GCM-RCMs (1990-2005) is shown in Table 3.
Te rainfall in the catchment may not be accurately represented by all models.As shown in Table 4, some models slightly underestimate the observed rainfall, while others somewhat overestimate.Te most precise estimating model should be used to calculate the observed rainfall in the catchments.Te ensemble means performed the best in terms of bias (PBias � −0.0004%), whereas ICHEC-REMO2009 performs the worst (PBias � −0.003%).Te value in PBias denotes that there was a systematic discrepancy between simulated and observed rainfall levels.Te GCM-RCMs rainfall amount signifcantly difers from the observed rainfall amount, as shown by the high bias (PBias � −0.003%).Te ensemble mean performed best in terms of CV (CV � 0.37%), while MPI-CCLM4 performed worst (CV � 0.75%).Te ensemble mean also performs best (RMSE � 130.03 mm year −1 ), while MPI-CCLM4 performed worst (RMSE � 194.5 mm•year −1 ).However, the MPI-CCLM4 model performs best in terms of correlation coefcient (Correl � 0.43).CNRM-RCA4 has the worst performance (correl � −0.28).To estimate the observed mean annual rainfall quantity, this study used an ensemble mean, which performed 99.98% better than those four models.
Moreover, the monthly rainfall varies by a maximum of up to 173.93 mm from April to August.In the remaining months, the catchment received up to 48.59 mm in January, February, March, November, and December seeing the least amount of rainfall.Tus, RCMs model simulations reasonably reproduced the observed annual rainfall over the catchment after bias correction.Te observed annual cycle of rainfall's amount and pattern were quite well captured.

Rainfall Variability on a Monthly Basis.
Te catchment's mean monthly rainfall variability, standard deviation (SD), and coefcient of variation (CV) are provided in Table 4. Table 4 indicates that rainfall peaked at the Mayokote and Boditi School stations in May with 228.84 mm and 177.43 mm, respectively, while it peaked at the Bilate Tena stations in April with 117.50 mm.With the exception of spring, where there is a moderate variation, rainfall at Bilate Tena station fuctuates greatly throughout the year.Except for July, all other months at the Boditi School station had moderate rainfall variability, accounting for 28.59%, which ranges from 20% to 30%.Every month and season experienced a substantial variation in rainfall at Mayakote station.

Rainfall Variability on a Seasonal Basis.
Te catchment's seasonal variation in rainfall from 1990 to 2021 is presented in Table 4 and Figure 4. Te Mayokote meteorology station recorded the highest summer rainfall variability with a magnitude of 44.21%, while the Bilate Tena and Boditi School meteorology stations varied with magnitudes of 39.17% and 23.77%, respectively.Te Mayakote meteorology station in autumn (Belg) also recorded the highest rainfall variability (63.53%), followed by the station at Bilate Tena (47.79%) and Boditi School (37.59%).In the spring, rainfall varies by the magnitude of 48.17%, 40.65%, and 31.90% at the Mayakote Bilate Tena, and Boditi School meteorological stations, respectively.In the winter, rainfall varied considerably in all three stations with magnitudes ranging from 63.5% to 64.4%.Among the three stations, Mayokote exhibited the highest variability in all seasons and winter was the season with the very high variability recorded in all stations.Summer was found to be a season when high amounts of annual rainfall were received with a somewhat moderate variation.Tis result is consistent with recent fndings by the authors in [23,26,46,58,59], which demonstrated that summer is primarily responsible for the highest annual rainfall received in diferent parts of the country with relatively low variation.
Figure 5 depicts the catchment's spatial distribution of seasonal rainfall, which varied from 23.77 to 64.38%, with 23.77 to 44.21% in the summer, 63.8 to 64.38% in the winter, In contrast to this result, the authors in [16,31,55] reported that the kola agroecology had signifcantly higher seasonal rainfall variability than the rest by taking elevation into account as one of the main determining variables for seasonal rainfall distribution.According to [60], a number of climatological factors, including the southerly/south westerly cross-equatorial moisture fow from the Southern Indian Ocean and the seasonal northward advance of the intertropical convergence zone that persisted over Ethiopia, govern the spatial distribution of seasonal rainfall in Ethiopia.In accordance with the fndings of this study, the authors in [8,24,56,61] showed that Woyna Dega agroecology had extremely high seasonal rainfall variability, whereas Kola agroecology had the highest relative to Dega agroecology.Consequently, the adaptive response to extreme climate event done in the area should be feasible with agroecological heterogeneity.

Standardized Rainfall Anomaly (SRA).
Using the standardized rainfall anomaly (SRA) fndings, a study of the yearly rainfall variability over each station is shown in (Figure 5).As shown in Figure 6, over the study period, 56.25% negative deviations from the norm to 43.75% positive deviations were observed at the Bilate Tena station, 43.75% negative anomalies compared to 56.25% positive anomalies at the Boditi School station, and 53.125% negative anomalies compared to 46.875% positive anomalies at the Mayakote station.Hence, in contrast to Bilate Tena and Mayokote stations, where the negative anomaly was dominant and substantial cooling was evident, Boditi School station exhibited the positive anomaly, demonstrating the persistence of the warming period over the years 1990-2021.

Standardized Anomaly Index (SAI).
A popular index for analyzing local climate change is the standardized anomaly index (SAI) [51].In this study, the standardized anomaly index (SAI) was used to characterize the distribution of temperature in the study area for the years 1990-2021.Te study analyzed the series of mean annual temperature, mean annual minimum temperature, and average annual maximum temperature for each station.Te fnal depiction of the outcome showed a cooling time where the long-term average prevails out and a warming phase when the long-term average dominates.As illustrated in Figure 6, the three stations showed the mixed signal of the standardized anomaly index (SAI) on an annual basis.Te patterns of temperature anomalies showed a time when the below long-term average predominated (cooling) and a time when the above lasting average persisted (warming).Te outcome revealed that the cooling phase started in 1990 and continued uninterruptedly through 2005.On the other hand, from 2006 to 2021, above- Advances in Meteorology 11 average mean annual temperatures were recorded.With little breaks, temperatures rose gradually, and from 2006 to 2021, they were consistently above-average levels.

Rainfall Indices in terms of Intensity.
As can be seen in Table 5, the modifed Mann-Kendall's (MMK) trend test was used to examine the PRCPTOT, RX1day, RX5day, R95P, R99P, and SDII rainfall intensity indices for three stations (Bilate Tena, Boditi School, and Mayakote) throughout the period of 1990-2021.Te trend in the annual total wet-day precipitation (PRCPTOT), annual maximum 1-day precipitation (RX1day), annual maximum 5-day precipitation (RX5day), very wet days (R95P), annual total precipitation on days when daily rainfall is greater than 99th percentile (R99p), and an index of the number of wet days (SDII) was observed for the Bilate Tena station to be increasing.On the days when daily precipitation exceeded the 99th percentile (R99p), the annual total precipitation was negligible.All rainfall intensity indices displayed a growing trend, with the exception of total yearly wet-day precipitation (PRCPTOT), which exhibited a barely dropping trend for the Boditi School station.At the Mayakote station, it was determined that the total yearly wet-day precipitation (PRCPTOT), very wet days (R95P), and number of wet days index (SDII) were all dropping at a 5% level whereas the annual maximum 1day precipitation (RX1day) and the annual maximum 5-day precipitation (RX5day) had increased in Mayakote station.Te details of each rainfall index are presented in Table 5. Figure 7 illustrates the spatial distribution of rainfall intensity indices, which revealed that the total yearly wet-day precipitation (PRCPTOT) ranged from 1013 to 1412 mm and was found to be highest in the southwestern part of the catchment and lowest in the central and extreme northeastern parts.It was also found to be all rainfall indices except SDII, which have high rainfall intensity in the southwestern part of the catchments; others such as RX1day, RX5day, R99P, and SR95P have high rainfall intensity in the northeastern parts of the catchment.Agroecologically, the highest intensity was recorded in the Kola agroecology (Duguna Fango district), while the lowest intensity was recorded in the Dega agroecology (Damote Gale area), and Woyna Dega received the medium rainfall intensity.
In terms of the annual maximum 5-day precipitation (RX5day), the Kola and Woyna Dega agroecologies had the highest intensity (141-165 mm), while the Dega agroecology had the lowest intensity (88.8-140 mm).Te Kola and Woyna Dega agro ecologies received the highest intensity (49 to 57 mm) of the annual maximum 1-day precipitation (RX1day), while the Dega agroecology recorded the lowest intensity (39 to 48 mm).Likewise, the Kola and Woyna Dega agroecologies experienced the maximum intensity (243-252 mm) of extremely wet days (R95P), with the lowest intensity recorded (232-242 mm) in Dega agroecology.In addition, the Dega, Woyna Dega, and Kola agroecologies were shown from lowest to highest in terms of annual total precipitation on days with daily rainfall over 99 percentiles (R99p), which ranged from 39.5 to 146 mm.However, the number of wet days index (SDII) was observed to be high in Dega, medium in Woyna Dega, and low in Kola agroecologies.Te Dega agroecology displayed the lowest intensity, whereas the Kola and Woyna Dega agroecologies received the maximum intensity in all rainfall indices, with the exception of the number of wet days index (SDII).Tis result is consistent with the recent studies [16,55,56,58], which identifed that Kola agroecology has a higher rainfall intensity than Dega agroecology in terms of all rainfall intensity indices.

Rainfall Indices in terms of
Frequency.Table 6 and Figure 8 present the statistical results of the temporal trends and spatial distribution of frequency indices composed of R10 mm, R20 mm, CWD, and CDD.Te modifed Mann-Kendall test-based rainfall frequency analysis displayed a signifcant positive trend in the number of days with substantial precipitation (R10 mm), the number of days with extremely substantial precipitation (R20 mm), and the number of consecutive dry days (CDD), but no trend in the number of days with consecutive wet days (CWD) for the catchment.12 Advances in Meteorology Except for consecutive wet days (CWD), which had no trend and hence have a Sen's slope of 0.00, the number of consecutive dry days (CDDs), heavy precipitation days (R10 mm), and very heavy precipitation days (R20 mm) all exhibited considerably rising trends for the Bilate tena station that was at Duguna Fango agroecological area.Consecutive dry days (CDDs) and consecutive wet days (CWDs) in Boditi School station showed signifcantly increasing trends; however, the increase in the number of days with heavy precipitation (R10 mm) showed a decreasing trend but was statistically insignifcant.Te trends in the number of extremely heavy precipitation days (R20 mm) in Boditi School station showed no trend.At Mayakote station, trends in the frequency of days with heavy precipitation (R10 mm) and very heavy precipitation (R20 mm) showed declining trends but were statistically insignifcant.Both the consecutive dry days (CDDs) and the consecutive wet days (CWDs) at Mayakote station showed an upward trend, while the increase in the consecutive wet days (CWDs) is statistically insignifcant (Table 6).
Figure 9 depicts the spatial distribution of rainfall frequency indices, with the Woyna Dega agroecology having the highest number of days with very heavy precipitation (R20 mm), the Kola agroecology having the fewest days, and the Dega agroecology having a medium number of days within the catchment.Te quantity of days with a lot of rain (R10 mm), in catchments ranged from 33 to 38 days, with the Kola and Woyna Dega agroecologies recording the highest frequency and the Dega agroecology exhibiting the lowest frequency.
Continuous wet days (CWDs) ranged in length from 8 to 24 days in the watershed.Kola agroecology recorded the highest number of consecutive wet days (CWDs), while Dega agroecology recorded the fewest number of CWD and Woyna Dega agroecology had a medium number of CWD.Te catchment's consecutive dry days (CDDs) ranged from 17 to 22 days, with Dega agroecology recording the highest number of days and Woyna Dega and Kola agroecology recording the lowest number of days.Te rainfall frequency indices in Kola agroecology revealed a rising frequencies,   with the exception of consecutive dry days (CDDs), which were high in Dega agroecology and supported by past studies [62].However, conficting results were found by the authors in [60,[62][63][64], and the disparity with these fndings may be attributed to variations in the study period and place.
(1) Temperature Indices Trend Analysis.Te temperature indices of the three stations represented the overall rising trend for the years 1990 to 2021, according to Table 7 analysis of the modifed Mann-Kendall's (MMK) trend test results.Te fndings showed that for three stations across the study period, the annual warm day (TX90p), warm night (TN90p), the annual maximum value of the daily minimum temperature or warmest night (TNx), and the annual maximum value of the daily maximum temperature (TXx) were all observed to be increasing.However, the trends for cool days (TX10P) and cool nights (TN10P) varied across the three stations over the study period.In contrast to the trend for cool nights (TN10p), which exhibited decreasing trends in all three stations with statistically signifcant only at Bilate Tena stations, the trend for cool days (TX10p) increases in Boditi School stations and decreases in both Mayokote and Bilate Tena stations, which was consistent with the recent fndings [8,23,25,30,65] (Table 8).
As illustrated in Figure 8, all the temperature indices in the catchment varied spatially over the study period.Te catchment's annual warm days (TX90p) varied spatially, ranging from 0.27 to 6.027 days, with more heating in the southerly part of the study area.Likewise, warm nights (TN90p), which ranged from 0.274 to 3.163 days, showed long days of heating at night in the southern place of the catchments.Over the study period, the frequency of cool nights (TN10p), which shows 22.148 to 67.853 days, decreased in the southeast of the catchment and increased in the northeast and the opposit is true for cold days (TX10p).Te frequency of the warmest nights (TNx) and warmest days (TXx) showed increasing and decreasing trends in the southern and northern parts of the catchment, respectively.Tree agroecologies recorded the warmest night (TNx), with Kola and Woyna Dega having the highest warm night and Dega agroecology recording the lowest.Regarding the warmest days (TXx), Kola and Woyna Dega agroecologies recorded the maximum warmest days, whereas Dega agroecology reported the minimum.Similarly, the cool days (TX10p) varied agro ecologically, with the maximum at Kola and Woyna Dega agroecologies and the minimum at Dega agroecology.However, the reverse is true for cool nights (TN10p) which showed maximum cool night in Dega and minimum cool night in Kola and Woyna Dega agroecologies.Consistent with annual warm nights (TN90p), warm days (TX90p) varied agro ecologically, with maximum in Kola and Woyna Dega agroecologies and a low or minimum at Dega agroecology.Tis was consistent with past research results that showed a general trend of rising warm and falling cold extremes in the studied area [8,16,62,66].Te observed fuctuations in extreme temperature could be attributed to climate change, which is mostly brought about by human activities like deforestation and greenhouse gas emissions from industry and agriculture [7,8].7.In the frst two decades, the mean annual temperature is likely to increase to 0.86 °C and 0.85 °C under RCP4.5 and RCP8.5 scenarios between the years 2020 and 2039, respectively.Under the RCP4.5 and RCP8.5 scenarios, the annual average temperature will rise to a maximum of 0.91 °C and 0.94 °C, respectively, from 2040 to 2059.Under the RCP4.5 and RCP8.5 scenarios, the temperature is projected to rise by 0.91 °C and 1.04 °C from 2060 to 2079 and by 0.95 °C and 1.17 °C from 2080 to 2099 in comparison to the baseline period .However, the average annual temperature will increase by more than 1 °C between 2060 and 2079 and 2080 and 2099 years under the RCP8.5 high emission scenario.Tis is consistent with the study outputs of [67][68][69][70][71][72][73][74].Moreover, the projections of future climate change indicate that continued GHGs emissions will lead to further warming and changes in climate conditions which share the truth [2,7,11,23].16

Climate Change Projections in the Study
Advances in Meteorology

Conclusions
Tis study investigated space-time trends and variations in duration, intensity, and frequency of climate extremes using diferent meteorological indices at three stations in the Wolaita zone southern Ethiopia, for the period 1990-2021.Te results revealed that over the study period, the main climatic variables, temperature and rainfall, changed both spatially and temporally in the study area.Te fnding further indicated that maximum and minimum temperatures showed high spatiotemporal anomaly with overall signifcant warming, but on an annual basis, the three stations showed a mixed signal of anomaly.Similarly, the results of the modifed Mann-Kendall's trend test also supported the notion that nearly all temperature indices from the three stations during the research period represented the overall upward trend both during the day and at night.Te catchment's spatial distribution of seasonal rainfall varied from 23.77 to 64.38%, with 23.77 to 44.21% in the summer, 63.8 to 64.38% in the winter, 37.59 to 63.53% in the autumn, and 31.9 to 48.17% in the spring over the study period.Agroecologically, the summer rainfall varied within the same range in the Woyna Dega and Kola agroecologies (32.86 to 44.21%).In the same way, the winter rainfall varied between 63.84 and 64.2% in the Woyna Dega and Kola agroecologies.Autumn rainfall, however, varied substantially at diferent agro ecologies, with ranges of 46. 25   Generally, the Woyna Dega agroecology (Damote Woyde district) has demonstrated extremely high seasonal rainfall variability, followed by Kola (Duguna Fango district) and Dega (Damote Gale district), respectively.Te modifed Mann-Kendall test also revealed that while there was no trend for consecutive wet days (CWDs) in the watershed, there were positive trends in the number of heavy precipitation days, the number of very heavy precipitation days, and consecutive dry days across the study period.
Te projections from the selected model outputs under both scenarios revealed a signifcant increase in temperature over the study period, compared to the baseline period .Te average annual temperature will rise by more than 1 °C under a high emission scenario.However, the variation in annual precipitation will increase more in the medium term and decrease in the far future under both emission scenarios.Warming temperatures and unpredictable rainfall timing and distribution make their choice of management practices more difcult and directly afect the productivity of rain-fed agriculture and the livelihoods of rural farmers.In order to address these challenges, farmers, agricultural researchers, and extension experts must work together.Tey also need to localize climate trend analysis to identify the similarities and contrasts in the climatic extremes that farmers experience in various agroecological settings.Te study also recommended the establishment of timely and accurate climatic information, such as seasonal forecasts and early warning systems, which can be used as a reference for decision-making, planning, and policy implications on agriculture and climate change adaptation.Tus, this study urges policy-driven initiatives to convert climate-sensitive livelihood systems into climate-smart alternatives, thereby overcoming the difculties associated with the efects of extreme climate change.Te limited number of rainfall gauging stations was used in this study that conveys less information.Terefore, it is strongly advised that weather station quality and quantity be enhanced in order to increase the model's performance by include a sufcient number of highly efective hydrometeorological stations.

5 Figure 5 :
Figure 5: Anomaly of standardized rainfall over the catchment at the yearly base.

Figure 6 :
Figure 6: Standardized anomaly index (SAI) on an annual basis over the catchment.
Area.Te projections of climate change over two decades under the RCP4.5 and RCP8.5 scenarios compared with the baseline (1976-2005) are presented in Table

Figure 9 :
Figure 9: Spatial distribution of rainfall indices in terms of frequency.

Table 1 :
Te study area's chosen meteorological stations with their annual average rainfall (mm) and temperatures ( °C) from 1990 to 2021 years of observed data points in each gauging station.
37º50′0″E Figure 4: Spatial distribution of rainfall variability on a seasonal basis (1990-2021).10Advances in Meteorology 37.59 to 63.53% in the autumn, and 31.9 to 48.17% in the spring over the study period.Agroecologically, the summer rainfall varied within the same range in the Woyna Dega and Kola agroecologies (32.86 to 44.21%).In the same way, the winter rainfall varied between 63.84 and 64.

Table 5 :
MMK's trend statistics of rainfall intensity indices summary.

Table 6 :
Te modifed Mann-Kendall's trend statistics of rainfall frequency indices.

Table 7 :
Projected changes in the climate under low emission (RCP4.5)andhighemission (RCP8.5)scenarios in comparison with the baseline period.