Variability and trends of temperature and rainfall over three agro-ecological zones in North Shewa, Central Ethiopia

This study was conducted to analyze rainfall and temperature variability and trends over three Agro-Ecological Zones (AEZs) in Central Ethiopia. Global weather data for Soil and Water Assessment Tool (SWAT) and global mean monthly sea surface temperature (SST) data series were used for analysis. Mann–Kendall (MK) test was utilized to analyze rainfall and temperature trends. Sen’s slope estimator was employed to find out the rate of change. The study detected an annual increase of 0.07 °C (p < 0.001) in mean maximum temperature at Kolla AEZ. It also showed a 0.06 °C rise per annum (p < 0.001) for both Dega and Woina Dega AEZs. The average annual minimum temperature has increased by 0.03 °C per year at Kolla (p < 0.001), Woina Dega (p < 0.05), and Dega (p < 0.01), which means an increase of 1.05 °C between 1979 and 2013. Results from precipitation concentration index (PCI) revealed the highest percentage (97.1%) of irregular distributions in annual rainfall pattern at Kolla AEZ, followed by Woina Dega (82.9%). Standardized rainfall anomalies (SRAs) calculated in the study also showed higher percentage (28.6%) of drought in Kolla AEZ, which experienced drought once in every 3 or 4 years. The study revealed negative annual rainfall anomalies for 18 years in Kolla and 16 years in both Dega and Woina Dega AEZs. Rising temperatures and irregular rainfall patterns may induce some adverse impacts on the livelihoods of rural populations in general, their subsistence agricultural production processes and food availability in particular. Policymakers and stakeholders should give priority in designing and introducing pro-poor plus geographically differentiated adaptive strategies.


Introduction
Today's global community faces a number of critical challenges. Climate change and/or variability are among the major issues of our time (Limjirakan et al. 2010). The problem of climate change is manifested by the ever-changing trends in the earth's natural climate and weather systems: increasing global average temperatures, melting glaciers, rising sea levels, fluctuations in the amount and duration of rainfall, escalation in the frequency and intensity of extreme meteorological events (Limjirakan et al. 2010;IPCC 2013;Chen et al. 2014).…etc. In future climate scenarios, among others, warmer temperatures and drier rainfall regimes or patterns have been envisaged for various areas in the world (IPCC 2013;Longobardi and Villani 2010;Gebrehiwot and Veen 2013). This will further complicate challenges facing the contemporary global community. Threats related to climate change have been felt and are sensed locally, nationally, regionally, and around the world (Wang et al. 2018;IPCC 2001). However, climate change has a greater impact on developing regions than their developed counter groups (IPCC 2001).
Africa has been identified as one of the regions most sensitive to the adverse impacts of climate change and extreme climatic events (Kinyangi et al. 2009). Rapid warming patterns of annual and seasonal temperature, even higher than the global average, have been observed on the continent in recent centuries and are expected to persist in the future (Hulme et al. 2001). Unpredictable and inconsistent rainfall patterns have also been reported as distinctive features of the African climate across periods (Korecha et al. 2013). The negative effects of such events have already been felt in the continent on the livelihoods of rural populations in general, agricultural production and food availability in particular (Gebrehiwot and Veen 2013). The East African region is not an exception to this scenario. Researchers in the region (Kinyangi et al. 2009;Easterling et al. 2007) have reported very high temperature and rainfall variations, as well as associated adverse consequences, particularly the reduction in agricultural production and agricultural yields (Befikadu et al. 2018).
Ethiopia is one of the East African countries with a variable climate (Kinyangi et al. 2009;Hulme et al. 2001). It has an assortment of AEZs and diverse environment, typified by a striking variety of microclimates and corresponding weather patterns (Astawsegn 2014). Ethiopia is in tropical climatic conditions. However, tropical climatic conditions do not have full spatial coverage in the country. As Ethiopia is a highland country, it experiences vast variations induced by topography in the main climatic parameters particularly temperature and precipitation. Large-scale atmospheric and oceanic factors of climate variability also influence Ethiopia (Korecha and Barnston 2007;Bewket et al., 2017;Meron et al. 2020). Seasonal climate trends over large parts of the country are influenced by the SST of the Atlantic, Indian, and Equatorial Oceans of the Pacific, and associated atmospheric circulations. While the El Nino Southern Oscillation (ENSO) in the Nino-3.4 region of the equatorial Pacific strongly influences the main season's (Kiremt) climate, the Indian Ocean Dipole (IOD) has relatively stronger influence on the climates of the dry season (Bega) and small rain (Belg) season (Takele et al. 2020).
Ethiopia, like many African countries, is subjected to and widely exposed to the adverse impacts of climate change as well as extreme weather events (Conway and Schipper 2011). Sources mentioned its name in the list of most vulnerable countries (Kinyangi et al. 2009;Zerihun and Prowse 2017;Choularton et al. 2014). Significant impacts of current climate variability and extreme events are observed in the country (Dazé 2014). The impacts of climate related shocks are most felt in the country's agricultural sector, on which millions of its citizens depend for survival. Poor performance in the Ethiopian agricultural sector is mainly associated with changes in rainfall patterns which cause both droughts and floods (Conway and Schipper 2011;Deressa 2010). Climate-related shocks are expected to continue in the upcoming years, imposing further threats to the country's agricultural sector in general and food security of its citizens in particular (Gebrehiwot and Veen 2013;Zerihun and Prowse 2017;Dazé 2014;Goel et al. 2013). There is, therefore, a need to pay more attention to adaptation and mitigation mechanisms due to the ongoing and anticipated changes in climate.
Various studies (Gebrehiwot and Veen 2013;Korecha et al. 2013;Zerihun and Prowse 2017;Goel et al. 2013;Asfaw et al. 2018;Conway 2000a;Conway 2000b;Meze-Hausken 2004;Suryabhagavan 2016;Lizcano et al. 2008;Teyso and Anjulo 2016;Bewket and Conway 2007;Rosell and Holmer 2007;Kebede and Bewket 2009) have investigated long-term trends and spatiotemporal variations in key climate change proxy indicators, mainly rainfall and temperature over Ethiopia. Regardless of the differences in space and time dimensions they considered, there is no significant deviation in the results obtained by the previous researchers for trends of temperature in the country. Most of them have reported steadily rising trends in mean maximum, minimum, and average temperature. However, previous studies have reported mixed findings on trends and variability of rainfall in Ethiopia. The existing studies are of three types. The first group comprises Goel et al. (2013), Asfaw et al. (2018), Conway (2000a), and Conway (2000b) and reported downward/negative trends in the amount and duration of annual and seasonal rainfall in Ethiopia. The second grouping constitutes Korecha et al. (2013), Meze-Hausken (2004), Suryabhagavan (2016), and Lizcano et al. (2008). This group of studies could not find in their inquiries a downward or negative trend of rainfall in the country. They documented a non significant trend in annual and seasonal rainfall amount in Ethiopia. The third group includes Teyso and Anjulo (2016), Bewket and Conway (2007), and Rosell and Holmer (2007) and observed that the annual and seasonal rainfall show great spatiotemporal variations in the country. Many studies (Chen et al. 2014;Kinyangi et al. 2009;Hulme et al. 2001) have also been conducted to reveal the extent of climate change at global and regional levels. There is, however, paucity of studies at local places and differentiated by AEZs (Easterling et al. 2007;Bewket and Conway 2007;Kebede and Bewket 2009;Befikadu et al., 2018). The previously conducted aggregate analyses do not reflect local conditions, which are extremely divergent, and where the negative upshots of climate variability are most noticeable.
This makes it difficult to understand the extent to which climate parameters vary in those localities and their diverse AEZs across periods (Gebrehiwot and Veen 2013;Bewket and Conway 2007;Kebede and Bewket 2009). North Shewa Administrative Zone (NSAZ) of the Oromia National Regional State (ONRS) represents one of the places within Ethiopia where trends in climate parameters remain largely unexamined (Asaminew and Diriba 2015) and hence, in bits and pieces. The purpose of this study was to analyze trends and spatio-temporal variations in rainfall and temperature in NSAZ disaggregating by AEZs and to attempt to fill the above deficit. We believe that the results of the study are relevant for predictive adaptation to the impacts of climate variability on agriculture and food security in the face of anticipated changes in climate.

Description of the study area
North Shewa Administrative Zone, with an area of 10,322.48 square kilometers, is located in the Oromia National Regional State of Ethiopia. Its latitudinal and longitudinal locations extend from 9°08′52′′-10°35′17′′N and 37°56′13′′-39°34′47′′E, respectively. The altitude of the area extends from about 1000 m above sea level, located in Abay gorge in Wara Jarso district to over 3500 m above sea level, located in Dagam district (Asaminew and Diriba 2015).
NSAZ is characterized by varied topographic settings: plain features, highland, and pocket lowland areas. It is dissected by high plateaus and mountains associated with hills, valleys, and gorges. The study area's varied topographic features resulted in micro climatic variations in its several localities. The monthly minimum, maximum, and average temperature of NSAZ ranges from about 7.4 to 11.2 °C, 20.8 to 28.5 °C, and 15.6 to 19.8 °C, respectively (Table 2). Annual rainfall received in the area understudy also ranges from about 532.4 to 1287.8 mm (Table 3). There are two distinct rainy seasons in the study area, namely Kirtemt (summer) and Belg (spring). There are also three AEZs in NSAZ: highland (Dega), midland (Woina Dega), and lowland (Kolla).
Dega occupies over 50% of the area. Woina Dega makes the second highest proportion of the area. Kolla occupies small portion of the study area (Fig. 1). Vertisols, Cambisols, and Lithosols are major soils within the study area. The study area's economy mainly depends on rain-fed agricultural activities. It is characterized by mixed farming systems: crop and livestock production.

Data types and sources
All scientific research requires reliable data. Datasets observed at stations are primarily needed to analyze trends and changes in precipitation and temperature. In Ethiopia, however, weather records are rarely complete. Weather stations are rare and unevenly distributed and have poor data quality (Chang et al., 2018). The area understudy also suffers from the same problems: sparse and unevenly distributed weather measurement stations, high missing values, and poor data quality. Problems with conventional weather data have forced researchers to look for other options. Climate Forecast System Reanalysis (CFSR), a freely accessible, high-resolution global dataset, has become one of the accepted alternatives to the problem (Arnold et al. 2007). The application of global CFSR weather data for SWAT to Ethiopian conditions and its validity in relation to stationbased observed series has been evaluated by the previous studies in the country (Chang et al., 2018;Dile and Srinivasan 2014). These studies have supported its applicability in the regions of Ethiopia scaring the data. Thus, the CFSR Fig. 1 Location of the study area data was obtained from the National Centers for Environmental Prediction (NCEP) and used for this study (Table 1).
Data obtained from the 30 stations indicated in Table 1 were averaged in two ways. First, the entire stations' data were all averaged to get an aggregated picture on rainfall and temperature of the study area (North Shewa Administrative Zone). Second, the 30 stations indicated in Table 1 were classified into three elevation zones: highland (Dega), midland (Woina Dega), and lowland (Kolla). The first 10 stations' elevation ranges from 2400 to 2844 masl and categorized under Dega/cool to humid (2300-3200 m) AEZ. The succeeding 16 stations' altitude ranges from 1602 to 2296 masl and classified under Woina Dega/cool sub humid (1500-2300 m) AEZ. The last 4 stations' elevation ranges from 1272 to 1372 masl and placed under Kola/warm semi arid (500-1500 m) AEZ. Data obtained from stations placed under similar elevation zones were averaged so as to represent their respective AEZ and this was used for special analysis (a disaggregated analysis by AEZs).
Global mean monthly SSTs reanalysis time series datasets  in the Nino3.4 and IOD regions were also used to conduct the current study. These datasets were obtained from the National Oceanic and Atmospheric Administration (NOAA). Using the datasets, we computed indices of the important SST modes that are known to affect rainfall variability over Ethiopia: Nino3.4 Index and Dipole Mode Index (DMI) or IOD Index. The Nino3.4 Index was calculated by taking the spatial average SST within the Nino3.4 region, which extends from 5°N to 5°S latitude and from 120 to 170°W longitude (in the Pacific Ocean). DMI or IOD Index, the SSTs indices considered over the Indian Ocean, is the average over (10°S-10°N and 50-70°E) minus the average over (10°S-0 and 90-110°E). Global mean monthly SSTs reanalysis time series datasets from NOAA were used by the previous studies in Ethiopia (Korecha and Barnston 2007;Bewket et al., 2017;Meron et al. 2020).

Research design and method of data analysis
In this study, a longitudinal research design was used to detect changes in rainfall and temperature events and to quantify trends during the study period . The following indices and tests were also conducted using change-point detection software packages in R (Version 1.1.1) and Microsoft Excel spreadsheet to analyze spatiotemporal variations in rainfall and temperature at different agro-ecological levels.

Coefficient of variation (CV)
It is a statistical measure of how the value of individual data varies from the average value. A higher CV value is an indicator of greater spatial variability, and a lower CV value is an indicator of smaller spatial variability (Chakraborty et al. 2013). According to researchers' categorization, CV values < 20% indicate less variability; values between 20 and 30% designate moderate variability and values > 30% stand for high variability of rainfall (Chakraborty et al. 2013). It is computed using the following equation: Where, CV is the coefficient of variation, σ is the standard deviation (SD), and x̂ is the mean.

Precipitation concentration index (PCI)
It measures relative distribution, concentration, and spatial variability of rainfall on both annual and seasonal time scales. It is calculated at the annual and seasonal scale using the following equations, respectively.
Where, pi is the monthly precipitation in month i.

Mann-Kendall (MK) test
It is a non-parametric rank based statistical test for trend detection in time series data records (Mann 1945;Kendall 1975). It is often used to find out whether a sequential (time series) rainfall and temperature dataset has statistically significant or insignificant upward or downward tendency overtime (Abdo et al. 2016;Chen et al. 2008;Poudel and Shaw 2016;Jain and Kumar 2012). The null hypothesis (H o ) for this test is that there is no trend, and the alternative hypothesis (H 1 ) is that there is a trend in the two-sided test or that there is an upward trend (or downward trend) in the onesided test. This test is chosen from other statistical methods available to detect trends within time series data for a number of reasons. Among other things, it does not require that data be normally distributed or linear. This implies that MK is distribution-free test (Chen et al. 2008). It is also strong against the effects of extremes and is not as much sensitive to outliers (Mann 1945;Kendall 1975).
According to Mann (1945) and Kendall (1975), the following equations are used to calculate the MK statistic S, the variance statistic Var (σ), and the standard normal test statistic Z.
Where, N is the number of data points; X i and X j are the time series observations.
Assuming (X j -X i ) = θ, the value of sgn (θ) is computed from: Under the hypothesis of independent and randomly distributed random variables, for large samples, when n ≥ 10, the σ statistic is approximately normally distributed, with zero mean and variance: Similarly, Z-statistic is calculated from: When Z value exceeds either of the confidence limit lines, it shows a significant trend at a given significance level. Hence, H o is rejected and in place H 1 is accepted. It should be noted that while positive (+) Z value signifies increasing (upward) trends over time, negative (−) Z value indicates decreasing (downward) trends (Suryabhagavan 2016).

Sen's slope estimator
It is a non-parametric slope based statistical test for trend detection in time series data records. It is often used to find out the rate of change in amount or intensity of hydro-meteorological parameters measured over time (Sen 1968;Theil 1950;Chattopadhyay and Edwards 2016;Jain and Kumar 2012) .This test is more robust than other slope estimating statistical methods because it is strong against the effects of extremes and is not as much sensitive to outliers (Chattopadhyay and Edwards 2016).
According to Sen (1968) and Theil (1950), to derive an estimate of the Slope b i , the slopes of all data pairs are computed as: Where: xj and xi are data values at times j and i; j > i respectively. The Sen's estimator of the slope is the median of these N values of bi: A positive value of b indicates an "upward trend" (increasing values with time) while a negative value of b indicates a "downward trend" (decreasing values with time) (Chakraborty et al. 2013).

Standardized rainfall anomaly (SRA)
It is used as a descriptor of rainfall variability. It has both negative and positive values. While the former represent periods of below-normal rains (droughts), the latter reflect above normal rains (flood risk). Using SRA's values, severity of drought is categorized as follows: extreme drought for SRA values < −1.65; severe drought for 1.28 > SRA Where, SRA is standardized rainfall anomaly, Ρt is annual rainfall in year t, Ρm is long-term mean annual rainfall, and σ is the SD of annual rainfall for the period stretching from 1979 to 2013.

Pearson's correlation coefficient (r)
It was computed to test the association of Kiremt (June-September), Bega (October-January), and Belg (February-May) seasons' rainfall in NSAZ with corresponding 4-month average anomalous SST in the Nino3.4 and IOD regions. The null hypothesis (H o ) for this test is that there is no correlation between the two variables, and the alternative hypothesis (H 1 ) is that there is correlation between them in the twosided test or that there is a negative or positive co-variability in the one-sided test. It is important to note that the monthly rainfall averaged over Ethiopia (Meron et al. 2020), and specifically also over NSAZ, have no statistically significant increasing or decreasing trend within the period of analysis 1979-2013 ( Fig. 3 and Table 3), therefore the data was used directly without the need of de-trending for the correlation analysis. In this research, Pearson's r was defined as: Where r is correlation coefficient, n is the length of the time series, and i is the number of years during the analyzed periods . Xi and Yi are the rainfall and the SST in the year i, respectively, and x̂ and ȳ are the mean rainfall and the mean of SST, respectively during the studied periods.

Results and discussion
Variability and trends of rainfall and temperature in NSAZ

Variability and trends of temperature
Several parameters are used for analyzing temperature variability and trends. Annual maximum, annual minimum, and annual average are parameters computed on a yearly time scale. Monthly maximum, monthly minimum, and monthly average are parameters computed on a monthly time scale. Temperature variability and trends in NSAZ between 1979 and 2013 based on the sex parameters indicated above are shown in Table 2 and Fig. 2. The minimum temperature of the area was 9.5 °C with a maximum of 25.6 °C and an annual average of 17.5 °C. Long-term mean monthly temperature showed statistically significant increase every month. It is clear that monthly maximum temperatures have increased faster than monthly minimum temperatures, which has led to an increase in average monthly temperatures over the period of study. The coefficient of regression of the annual maximum, minimum, and average temperatures revealed increasing trends at the rate of 0.056 °C, 0.022 °C, and 0.039 °C per year, respectively (Table 2 and Fig. 2). Because the rate of increase in the mean annual maximum temperature was faster than the minimum, the overall rise in annual average temperature was attributed more to the first than to the second. Studies in other parts of Ethiopia have also shown an increase in the mean annual maximum, minimum, and average temperature Table 2 Annual and monthly temperature in NSAZ (1979NSAZ ( -2013 * is significant at p < 0.05, ** is significant at p < 0.01, *** is significant at p < 0.001  (Asfaw et al. 2018;Befikadu et al., 2018;Lizcano et al. 2008;Teyso and Anjulo 2016).

Variability and trends of rainfall
A number of parameters are used for analyzing rainfall variability and trends. Monthly, seasonal, and annual long-term averages of rainfall are among the parameters employed for analysis. Monthly, seasonal, and annual rainfall trends in the study area for the years from 1979 to 2013 are shown in Table 3 and Fig. 3. As shown in Table 3, the area's long-term average annual rainfall was 940.2 mm with an SD of 157.8 and a CV of 16.8%. The highest annual rainfall (1287.8 mm) and the lowest (532.4 mm) were recorded in 1998 and 2002, respectively. Total annual rainfall in the area was highly concentrated in 4 months: June, July, August, and September (Kiremt season). Approximately 81.6% of the total rainfall was  Years obtained during these months and as a result, Kiremt season contributed the highest share of the total annual rainfall in the study area. A considerable amount of rainfall was received during the 3/4 months of Belg season (i.e., February/March, April, and May) representing approximately 14.2% of total annual rainfall. Similar results have been obtained from numerous other studies in Ethiopia (Ayalew et al. 2012;Gebrehiwot and Veen 2013;Suryabhagavan 2016).
This study found that the Belg season's rainfall was more erratic than the Kiremt season's rainfall. The CV of the former season's rainfall was much higher (46%) and more variable than the CV of latter season's rainfall (15%) as shown in Table 3. A study conducted in North Shewa (Central Ethiopia) indicated that the rainfall of the Belg season was more variable than the rainfall of the Kiremt season (Abegaz and Mekoya 2020). Studies carried out in other parts of Ethiopia (Bewket and Conway 2007;Rosell and Holmer 2007;Ayalew et al. 2012;Bewket 2009) have also documented that Kiremt rainfall, which is the largest in terms of its amount and geographical coverage, is less variable in most parts of the country compared to the rainfall of the Belg season and supports the main production of the growing season, locally known as Meher. It is reported in Bekele et al. (2018) and Alemayehu and Bewket (2016) that the rainfall of the Belg season is characterized by great temporal and spatial variability, which has implications on Belg's seasonal crops and household food security.
The results of the MK test revealed a downward, but statistically insignificant, trend in annual and seasonal (Kiremt and Belg) rainfall. A statistically significant downward trend in rainfall was observed for only 2 months: February (p < 0.05) and June (p < 0.05). Regression coefficients for annual, Kiremt and Belg rainfall showed downward trends (Fig. 3) at rates of 3.104 mm, 1.707 mm, and 1.431 mm per year, respectively (Table 3). Rainfall reduction showed variations between the years and among the Kiremt and Belg seasons at a rate of 3.104 mm, 1.707 mm, and 1.431 mm, respectively. This result of the study negates with other inquiries (Goel et al. 2013;Asfaw et al. 2018), which found significantly decreasing trends in annual and Kiremt rainfall. The obtained result is, however, similar to other findings (Korecha et al. 2013;Conway 2000a;Conway 2000b;27, Lizcano et al. 2008;Bewket et al. 2004), which found and reported statistically insignificant decline in annual and Kiremt rainfall.

Associations between seasonal rainfall in NSAZ and SST (1979-2013)
Seasonal climate trends particularly rainfall over large parts of Ethiopia is influenced by the SST of the Atlantic, Indian, and Equatorial Oceans of the Pacific, and associated atmospheric circulations. The link between seasonal rainfall in NSAZ andSST (1979-2013) is presented in Table 4. SST in Nin˜o3.4 region showed positive association with mean rainfall both in Belg (February-May) as well as Bega (October-January) seasons. The correlation (r = 0.326611) was statistically significant for the latter season at α = 0.05 level (1-tailed). The positive correlation between Nin˜o3.4 Index and mean rainfall anomalies in Belg plus Bega seasons implies that increases of SST over the equatorial east Pacific region increased the amount of rainfall in these two seasons within NSAZ during the study period: 1979-2013. The correlation between mean rainfall in Kiremt (June-September) season and SST in NINO 3.4 region was negative (r = −0.25487). Yet, it was statistically insignificant. The negative correlation between Nin˜o3.4 Index and mean rainfall anomalies in Kiremt season also implies that increases of SST over the equatorial east Pacific region decreased the amount of rainfall in this season within NSAZ over the period of study. The analysis revealed negative correlation between IOD Index and mean rainfall anomalies in Kiremt as well as Belg seasons. However, the correlation between Bega season rainfall anomalies and the IOD Index was positive (r = 0.153).
Numerous studies (Korecha and Barnston 2007;Bewket et al. 2017;Takele et al. 2020;Black et al. 2011) assessed the link between SSTs and seasonal rainfall in Ethiopia. A study conducted by Bewket et al. (2017) found an inverse association between Kiremt season rainfall over central and western parts of Ethiopia and SSTs over the equatorial east Pacific and Indian Ocean. Same study reported that the negative correlation of Kiremt season rainfall over central and western parts of Ethiopia with Nin˜o3.4 was stronger (r = −0.59) than with the IOD (r = −0.34). Another study by Takele et al. (2020) documented that while the El Nino Southern Oscillation (ENSO) in the Nino-3.4 region of the equatorial Pacific strongly influences the main (Kiremt) season's climate of Ethiopia, the Indian Ocean Dipole (IOD) has relatively stronger influence on the climates of the dry (Bega) season and small rain (Belg)  Korecha and Barnston (2007) reported negative and positive correlation between Niño 3.4 Index and rainfall in the northwestern and southeast parts of Ethiopia, respectively. Another study by Black et al. (2011) reported negative correlation between the equatorial Pacific SST and rainfall in various parts of Ethiopia during the Kiremt season. A study by Besha et al. (2018) revealed that the correlation between SST anomalies and rainfall was negative and positive in Kiremt and Belg seasons, respectively in the Upper Awash basin.

Variability and trends in mean annual maximum temperature
Mean annual maximum temperature is one among parameters employed to analyze temperature variability and trends. The mean annual maximum temperature for Dega, Woina Dega, and Kolla AEZs within NSAZ varies between 22-25 °C, 25-29 °C, and 28-31 °C, respectively (Table 5). This study found a more or less equal CV for Woina Dega (CV = 3.37) and Kolla (CV = 3.34) AEZs in annual average maximum temperature over the 35 years  considered. For highland area (Dega AEZ), however, a higher CV (4.18) was found. This suggests existence of variability in mean annual maximum temperature over those years between AEZs. The year 2002 was observed as the hottest year in all AEZs during the period of observation . The hottest years identified in the present investigation go in line with the result obtained in a study carried out in other part of Ethiopia, which reported that the 2000s was the warmest decade compared to the 1980 and 1990s in three districts (Basona Werana, Efratana Gidim, and Menz Gera Meder) located in different agro-ecological belts in the central highlands of Ethiopia (Arragaw and Woldeamlak 2016). Results obtained in the present study also go in line with the finding by Befikadu et al. (2018). The temporal and spatial variability and trend of mean annual maximum temperature by AEZs are shown in Fig. 4 and Table 6. The mean annual maximum temperature showed an increase in all AEZs over the 35 years of the study (Fig. 4). It exhibited an increase of 0.07 °C per annum   Table 6). The warming trend observed in the study area is relatively higher than the result obtained in other studies conducted in Ethiopia (Bewket et al. 2004;NMA 2007). Despite the difference in the rates of increase, several other studies at various spatial and temporal scales have reported warming trends in mean annual maximum temperature in other parts of the country (Bewket et al. 2004;NMA 2007;Bewket et al. 2014).

Variability and trends in mean annual minimum temperature
Mean annual minimum temperature is used as one of the measures to examine temperature fluctuation and trends. The mean annual minimum temperature for Dega, Woina Dega, and Kolla AEZs within NSAZ varies between 7-10 °C, 9-11 °C, and 10-12 °C over the 35 years  of observation, respectively (Table 7). The long-term annual average minimum temperature was 8.46 °C (CV = 6.12), 9.81 °C (CV = 5.18), and 10.86 °C (CV = 4.49) for Dega, Woina Dega, and Kolla, respectively. It shows relatively high variability in the highland (Dega) area than the other AEZs. All AEZs of the study experienced relatively cold years in the 1980s. This result goes in line with the finding by Befikadu et al. (2018). The result also goes in line with the cold years reported among AEZs of the Upper Nile basin (Bewket et al. 2014) and in three districts (Basona Werana, Efratana Gidim, and Menz Gera Meder) located in different agro-ecological belts in the central highlands of Ethiopia (Arragaw and Woldeamlak 2016).
This inquiry has demonstrated an increasing trend in mean annual minimum temperature in all AEZs over the years stretching from 1979 to 2013 (Fig. 5). Mean annual minimum temperature exhibited an increase of 0.03 °C per year for Kolla (p < 0.001), Woina Dega (p < 0.05), and Dega (p < 0.01) agro-ecologies, which means an increase of 1.05 °C between the afore-indicated years (Table 8). This finding, therefore, shows that the rates with which mean annual maximum temperature increase was far higher than the rate with which mean annual minimum temperature increase in all AEZs over the period considered in the study. The more rapid rise observed for maximum temperatures than minimum might be attributed to local conditions such as soil moisture and evaporative heat. According to Alfaro et al. (2006) while maximum temperatures are influenced by local circumstances, mainly soil moisture and evaporative heat loss as soil water evaporates, minimum air temperatures are affected by large scale adjustments in atmospheric water vapor content. Increased evaporation from soil and plants due to increased temperature should be balanced by more rainfall in a region Table 6 Monthly and annual maximum temperature by AEZs (1979AEZs ( -2013 * is significant at p < 0.05, ** is significant at p < 0.01, *** is significant at p < 0.001 so as to escape from recurrent dry spell and droughts, which are detrimental to many of the agricultural crops. Otherwise, a rise in temperature, mostly in tropical and sub-tropical areas, will decrease agricultural production and productivity. This is because, in these areas, certain crops are already grown near their limit of heat tolerance, and additional rises in temperature will further escalate temperature stress (Maharjan and Joshi 2012). This study's finding on the rates with which mean annual maximum and minimum temperatures increase over AEZs negates the result obtained by Befikadu et al. (2018), who reported that the rate of change for mean annual minimum temperature is faster than the mean annual maximum temperature both in time and space.

Variability and trends in rainfall
Monthly, seasonal, and annual long-term averages of rainfall are used as major parameters to analyze rainfall variability  Table 8 Monthly and annual minimum temperature by AEZs  * is significant at p < 0.05, ** is significant at p < 0.01, *** is significant at p < 0.001 and trends across AEZs. The annual, seasonal, and monthly average rainfall for 35 years  in the highland, midland, and lowland AEZs of NSAZ are presented in Tables 9, 10, and 11. The annual amount of rain varied widely from year to year in all AEZs. It varies between 186.0-584.1 mm, 375.8-1006.5 mm, and 921.6-2019.2 mm over Kolla, Woina Dega, and Dega AEZs, respectively. The mean annual rainfall was 1520.4 mm, 710.6 mm, and 407.9 mm for highland, midland, and lowland AEZs, respectively. The CV ranges from 15.95% in the Dega AEZ to 20.8% in the Kolla AEZ, signifying higher inner-annual rainfall variability in the latter AEZ. All the three AEZs received their maximum annual rainfall in 1998 and minimum in 2002 (Table 9). These were the wettest and the driest years. All AEZs of the study area, as with most other places in Ethiopia, receive large amount of rainfall during summer (Kiremt) season, which contributes 78.8%, 84.1%, and 89.3% of rainfall for Dega, Woina Dega, and Kolla AEZs, respectively. Belg also contributes 16.6%, 11.9%, and 8.6% of rainfall for each, respectively. While rainfall was more concentrated in Kiremt season at lowland AEZ than others, Belg contributed more amount of rain for highland AEZ than for others. The mean monthly maximum rainfall was 423.6 mm, 214.9 mm, and 145.9 mm at Dega, Woina Dega, and Kolla AEZs, respectively. These variations in rainfall concentration between AEZs during Belg and Kiremt seasons are perhaps caused by topographic differences. Table 10 indicates that while highland and midland AEZs get their maximum amount of rainfall on July, lowland receives its maximum amount of rainfall on August. All AEZs of the study receive limited/minimum amount of rain on December. Table 10 also shows trend test results of rainfall on monthly, seasonal, and annual time scales. The MK test results summarized in the table revealed a declining, but statistically insignificant, pattern in mean annual rainfall at all AEZs. The rate of decline was 4.695 mm/year, 2.55 mm/year, and 1.342 mm/year in Dega, Woina Dega, and Kolla AEZs, respectively (Fig. 6). Studies conducted in other parts of Ethiopia also found and reported statistically insignificant diminishing tendencies of annual as well as seasonal rainfall (Conway 2000a;Conway 2000b;Bewket et al. 2004;;Lizcano et al. 2008;Korecha et al. 2013;Suryabhagavan 2016).
The frequency and percentage of annual and seasonal PCI values by agro-ecology for the years ranging from 1979 to 2013 are presented in Table 12. It shows irregular distributions in the annual rainfall patter at all AEZs. While higher frequency of irregular distributions in the annual rainfall patter was seen at Dega agro-ecology (48.6%), more strong irregularity was observed at Woina Dega (82.9%) and Kolla  Table 11 Monthly, seasonal, and annual rainfall variability by AEZs (1979AEZs ( -2013 Month  (97.1%). Table 12 also illustrates a more or less uniform and moderate pattern in Kiremt rainfall at all agro-ecologies. It further shows an increase in the irregularity of Belg rainfall pattern as on moves from Dega down to Kolla AEZs. The SRA computed for mean annual rainfall in three agro-ecological belts in NSAZ are shown in Fig. 7. The upward bar shows positive anomalies and the downward bar shows negative anomalies during the 35 years  of the study. The analysis revealed negative rainfall anomalies for 16 years in both Dega and Woina Dega AEZs and for 18 years in Kolla AEZ between 1979 and 2013. During those years, annual rainfall was below average. Severe deficits of rainfall were observed in 2002 and 2009 at all AEZs of the study. Our analysis also showed positive rainfall anomaly for 19 years in both Dega and Woina Dega AEZs and for 17 years in Kolla AEZ during the study period and this implies that during those years their respective annual rainfall was above average in each belts. Apparently, both shortage and excess rainfall negatively affect crop production.
The frequency and severity of droughts in the study area for the years between 1979 and 2013 are summarized in Table 13. It shows high frequency of drought in Kolla agroecology (28.6%), which experienced drought once in every 3 or 4 years between 1979 and 2013. Dega and Woina Dega experienced drought less frequently (< 17.1%) than Kolla. It is evident that the occurrence of drought and excessive rainfall over this region is somewhat associated with warm (El Nin˜o) and cold (La Nin˜a) events, respectively (Arragaw and Woldeamlak 2016). A study conducted by Seleshi and Camberlin (2006) revealed that warm ENSO periods (El Niño years) are typically associated with lower rainfall and drought years. In contrast, cold periods (La Niña years) are associated with higher rainfall amounts boosting the risk of flooding.

Potential impacts of rainfall and temperature variability
The study area's economy mainly depends on rain-fed agricultural activities. It is characterized by mixed farming  systems: crop and livestock production. These agricultural production processes are intimately tied to the prevailing weather conditions. The occurrences of temperature and rainfall variability would compromise the productive performances of the agricultural sector and makes rural households at risk. Variations and fluctuations in rainfall and temperature repetitively bring devastating repercussions on livestock and crop productions of the area (Million 2007;Messay 2009;Feleke 2018). The productivity of livestock is being adversely affected, among others, through shortages of rain. Deficiency of rain is usually associated with reduced pastures and limited pasture inturn is associated with death of cattle. Loss of livestock holding consequently exposes rural households, who depend on cattle for survival, to food insecurity. It more specifically makes income earning from livestock raring very low and limits the purchasing capacity of the households to have access to food. Erratic rainfall is also responsible for crop failures and apparently reduces productions. The decline in agricultural yield slowly drains food stocks and increases the risk of hunger.
This study has shown prevalence of erratic climatic conditions within NSAZ. The area is prone to extreme climatic events particularly drought, flooding, water-logging, and frost. Kolla AEZ of the area experienced drought once in every 3 or 4 years between 1979 and 2013. During extreme drought conditions, it is common that many smallholder farmers could not adequately feed their family members from own production. They depend on food aid (could be from the government or their relatives or NGOs) to sustain their lives or die due to hunger (Messay 2009). Million (2007) indicated that water-logging in Woin Dega and frost in Dega AEZs devastate 50 to 75% of standing crops during their worst incidence. Kassahun (2011) showed that changes of rainfall amount and pattern significantly reduce total crop yield through affecting planting time, growing stages, and harvesting periods. Another research further showed that a 10% decrease in the amount of rainfall below the long run average leads to a 4.4% reduction in the production of food and makes rural households vulnerable to food insecurity (Von Braun  1991). Bewket (2009) complemented that extreme variability of rainfall is a major cause of variations in crop production making smallholder farmers susceptible to food shortage. This study found increases in mean maximum, minimum, and average temperature at all AEZs. Rising temperatures, like the fluctuating patterns in the seasonal and annual rainfall, could also have an influence on smallholder farmers and their subsistence agricultural production processes (Canziani et al. 2007). Crop productions are sensitive not only to the changes of precipitation but also to the changes of temperature. Increased temperature adversely affects crop production, among others, thorough affecting plant growth and development. It also increases the influence of droughts by increasing atmospheric water demand, which could lead to additional water stress from increased water pressure deficits, subsequently reducing soil moisture and decreasing yield (Biru et al. 2012). Feleke (2018) researching in Kuyu district of NSAZ indicated extreme temperature as one of the principal factors behind crop production failure across Dega, Woina Dega, and Kolla AEZs. Increasing trends of temperatures could also affect livestock production. Rising temperature could negatively affect, among others, animal fitness and products they offer directly through effects on physiology (Abera et al. 2013). A study conducted in northern part of Ethiopia documented an increase in livestock diseases associated with increasing temperatures and explained how higher temperature can create more favorable conditions for the spread of skin diseases among cattle (Zerihun and Prowse 2017).
In a nut shell, the detected rise in temperatures and changes in precipitation patterns as well as in extreme weather events may all result in reduced agricultural productivity. It is reported that as high as 80% of the variability in agricultural production is caused by the disturbance of weather and related factors (Bäckman et al. 2009). It is also predicted that rainfall and temperature variability will continually compromise the life of rural people particularly the poor if not halted. There is, therefore, the need for an increased attention on adaptation and mitigation mechanisms. The impact of climate variability varies based on rural households' ability to respond for and mitigate its adverse effects. Variations in climate elements disproportionately affect poor rural households. Any effort to be exerted by the government or else any other concerned bodies should have to consider the magnified impact of climate variability on the most vulnerable group of the society. Rainfall and temperature variability and their resultant effect on rural livelihoods imply those poor farmers who are dependent on natural weather condition need protection through for example, irrigation scheme, introduction of drought resistant varieties of crops, livelihood diversifications, and other services. This can help farmers to stabilize income during the period of low rain and poor output.

Conclusion
This study analyzed trends and variability of precipitation and temperature across three AEZs in Central Ethiopia. Global Weather Data for SWAT and global mean monthly SST time series datasets were used for analysis. The study found increases in mean maximum, minimum, and average temperature at all AEZs throughout the period of study. Since the rate of increase in the mean maximum temperature was faster than the minimum, the overall rise in average temperature was more attributed to the former than the latter. The study found that annual rainfall performance was very much variable from year to year in the three AEZs. It showed irregular distributions in the annual rainfall patter at all AEZs. Results from PCI revealed the highest percentage (97.1%) of irregular distributions in annual rainfall pattern at Kolla AEZ, followed by Woina Dega (82.9%). SRA computed in the study also showed higher percentage (28.6%) of drought in Kolla AEZ, which experienced drought once in every 3 or 4 years. The detected rise in temperature and irregularity in precipitation has critical implications for rural livelihoods in general and rain-fed agriculture and food security in particular. Variations and fluctuations in rainfall and temperature compromise, among others, the productive performances of the agricultural sector and make rural households at risk. Rainfall and temperature variability will continually compromise the life of rural people if not brought to an end. There is the need for an increased attention on adaptation and mitigation mechanisms. Policymakers and stakeholders should give priority in designing and introducing pro-poor plus geographically differentiated adaptive strategies.
Abbreviations AEZs: agro-ecological zones; CV: coefficient of variation; IPCC: Intergovernmental Panel on Climate Change; MK: Mann-Kendal; NMA: National Meteorological Service Agency; NSAZ: North Shewa Administrative Zone; ONRS: Oromia National Regional State; PCI: precipitation concentration index; SD: standard deviation; SRA: standardized rainfall anomalies Data availability The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.