The Hamassa watershed is located in the Wolaita Zone, Southern Nations and Nationalities and People’s Region, Ethiopia, as indicated in Figure 1. The area extends from Damota Mountain at the North summit to Abaya Lake at the Southeastern tip. The watershed is grouped into three agroecological zones based on the traditional classification of agroecological zones ( Hurni, 1998). The area of the Hamassa watershed is 375.75 km ², of which highland, midland, and lowland have 61.97 km ², 158.60 km ², and 155.17 km ² shares, respectively. The area share of each AEZ is presented in Table 1 below.

The midland and lowland AEZs are dominant in the watershed concerning the area. The elevation in the watershed changes from 2850 mamsl at the Damota mountain area to 1000 mamsl around Lake Abaya. Rainfall varies considerably from one AEZ to another in its intensity, frequency, amount, and period. The mean annual precipitation ranges from 1199 to 956.6 and 723.5 in the highland, midland, and lowland, respectively. There is a bimodal rainfall distribution pattern; the Belgian rainfall is the primary crop production/rainy season, and next to it is the Kiremt rain, in which both rain and crop production are low. The Belg rain commences in February and ends around April or May, and the Kiremt rain, the second in the area, starts in June and ends in August ( Esayas et al., 2018). The mean annual minimum and maximum temperatures range from 14°C to 26°C, respectively. This study has three AEZs by adopting the traditional AEZ clustering technique, where the daily maximum, minimum, and rainfall data are extracted. The soil types in the watershed are Fluvisols, Leptosols, Luvisols, Nitisols, and Verisols. The broader area of the watershed is covered with verisols, which have 325.62 km ².

Cross-sectional data from primary and secondary sources were accessed for this study ( Bergene, 2024a, b, c). FGDs and KIIs were the primary data sources used to collect the information (see dataset ( Bergene, 2024b, c)). Secondary sources (NMA) provided data on daily maximum and minimum temperatures and precipitation. The Institutional Review Board (IRB) of the Addis Ababa University College of Development Studies has authorized all of the methods used for gathering data.

2.2.1 Key informant interviews (KIIs)

Like with the other techniques for gathering data, interviews were undertaken only when all aspects of the research’s purpose and value had been established. Interviews were conducted as part of KIIs to get primary data pertinent to the designated critical explanatory questions. The prominent people who provided the data were development agents, district officers, and well-known village residents ( Bergene 2024c).

2.2.2 Focus group discussions (FGDs)

The primary data from FGDs were collected in each agroecology. Before it commences, as mentioned in the KII, the research objective is communicated to the group to participate as the source of information.

2.2.3 Meteorological Data (Secondary Data)

Ethiopia’s National Meteorological Agency (NMA) collected the daily total rainfall and minimum and maximum temperature data, a gridded dataset (4 km by 4 km spatial resolution) from 1981 to 2018 (see dataset ( Bergene, 2024b)). The gridded climate (rain, maximum, and minimum temperature) dataset is an integration to enhance the quality of data from the national network managed by the Ethiopian NMA and satellite rainfall and temperature estimations from the European Organization for the Exploitation of Meteorological Satellites (EUMETSAT). The data choice of this type is because of the lower quality of station data in Ethiopia, which is substantiated by other scholars ( Mengistu et al., 2014).

The Hamassa watershed, situated in the Wolaita Zone, was chosen as a study area to investigate the impact of climate distress on smallholder farming due to its susceptibility to drought. The study area was divided into three agroecologies to ensure diverse climate perception information was collected from study participants over the analysis period. Respondents were selected from selected sample kebeles (small administrative units) using a simple random sampling procedure. The sample size 328 was determined using Cochran’s sample size determination formula, as described in 1977. Six Key Informant Interviews (KIIs) and three Focus Group Discussions (FGDs), one for each agroecology, were conducted to collect qualitative and quantitative data from household surveys. Participants were surveyed about their perception of climate change and its impact on rainfall and temperature. The survey data were analyzed using STATA version 13, while the qualitative data was transcribed, summarised, and analyzed to triangulate the quantitative data.

Mann-Kendall (MK) statistical test was used for this analysis because it is a widely used statistical test in many research papers and has many advantages. First, it does not require the data to be normally distributed, for it is a non-parametric test, and second, it has low sensitivity to inhomogeneous time series if an abrupt break occurs (outlier). The Mann-Kendall S Statistic is computed as follows: S = ∑ i = 1 n − 1 ∗ ∑ j = i + 1 n sgn ( xj − xi ) (1)

Where xj and xi are sequential data in the series, and sgn ( xj − xi ) = { + 1 , if xj − xi > 0 0 , if xj − xi = 0 − 1 , if xj − xi < 0 (2)

The variance S is computed as: n ( n − 1 ) ( 2 n + 1 ) − ∑ i = 1 m ti ( ti − 1 ) ( 2 ti + 5 ) 18 (3)

Sen’s slop estimator

The magnitude of the precipitation and temperature trend in time of series data is estimated by Sen’s slope estimator ( Sen, 1968). The slope Ti is calculated as the equation given below: Ti = xj − xk j − k for i 1 , 2 , … N (4)

Xj and xk are the data values at times j and k ( j > k), respectively.

The N values of Ti are ranked (smallest to largest), and the median of Sen’s slope estimator is calculated as: Qi = { T ( n + 1 2 ) ifN is odd N 2 + N + 2 2 2 if N is even (5)

The dataset underwent a preliminary evaluation to ensure the temperature and precipitation data quality was satisfactory by using the ClimPACT2 program in R software. Data analysis uses the R software and Microsoft Excel, and the graphs were contemplated using GraphPad Prism 9.5.1 software. The indicators were defined by the Expert Team on Climate Change Detection and Indices (ETCCDI) ( http://cccma.seos.uvic.ca/ETCCDI). For the years 1981 to 2018, the extreme temperature and precipitation indices were calculated. For this study area, 12 extreme temperatures and 10 extreme precipitation indices were chosen based on ETCCDI.

Correlation between major crops produced in each agroecology and Standardized Precipitation–Evapotranspiration Index (SPEI) on an annual basis was made to examine the relationship between climate variables and crop production. This uncovers the climate variability/change condition as a limiting factor for smallholder farming, which in turn causes food insecurity for those whose consumption is solely dependent on farm production.

The research received approval for ethical clearance with No. 035/01/2023 on October, 27, 2023 from the Addis Ababa University College of Development Studies Institutional Review Board (CoDS-IRB) after it was determined that it complied with all requirements. Participants’ informed consent must be obtained by the ethical clearances and criteria of the CoDS-IRB, allowing researchers to use written, verbal, or oral informed consent processes as needed. Oral consent was thus obtained from the participants for this study with approval from the IRB, which was indirectly indicated in the approval letter. Before data collection, the study objectives were made clear to the respondents, guaranteeing that their identities would remain confidential in all documentation. Oral consent was sought due to some respondents’ inability to read and write.

As indicated in Table 2, the SPEI strongly correlates to the primary crop production in the study area. Similar empirical findings uncovered the SPEI and crop production relationship ( Jabbi et al., 2021), other studies focusing on temperature and rainfall, and crops correlation goes the same line that climate affecting crop production in different parts ( Kyei-Mensah et al., 2019; Ndamani & Watanabe, 2015). The crop and SPEI correlation, as indicated in the table hereunder, uncovered that all AEZs have a strong increasing and significant correlation between the production of many crops and drought, which is unreasonably different from highland agroecology. This might be in the midland and lowland, the crop production is more explained by climate elements than other factors, and the highlands few crops positive and robust correlations among 10 major crops might be in the study area context the highland area rugged terrain with steep slopes causing land degradation which might more explain the crop production than climate factors. In connection, the kII and FGD assured the highland’s highly dissected and degraded rugged topography about a decade ago. The carbon project funded by the World Bank and the Australian government is intensively working on the land rehabilitation program. The area is densely forested, and some animals have disappeared, including tigers, which attack our domestic animals. This indicates human-induced land use land cover change alongside climate change, leading highlands to bare (confirmed in the FGD), which is now correctly managed to recover people’s livelihoods. As can be established from the data, though climate affects crop production of all agroecologies in the the study area, its spatial pain to farming is not even.

The interpretation of results below, from 3.2 to 3.3, depends on Table 3 and their subsequent figures.

3.2.1 Warm days (TX90p) and warm nights (TN90p)

All agroecological zones (AEZs), as indicated in Figure 2, have an increasing annual trend in the frequency of warm days (TX90p). However, the magnitude of change is unequal across agroecologies; the Highland and Midland AEZs have positive and significant trends with the same values, 0.033°C/year and 0.03°C/year, respectively. Lowland agroecology also has a positive annual trend, which is not significant. Unlike the case of warm days (TX90p), warm nights (TN90p) have a positive yearly trend in the highland and a negative trend in the midland and lowland AEZs. The decreasing warm nights’ trend is consistent with research output that revealed a significant decreasing trend of minimum temperature over the Northwest Himalayas ( Rahim et al., 2023). The warm days and nights increasing trend overall AEZs and the highlands align with other research results ( Damtew et al., 2022; Esayas et al., 2018). The disparate food insecurity problem in the Sudy area, particularly, and the Wolaita Zone, is probably more explained by climate change than other socioeconomic and physical factors ( Leza & Kuma, 2015).

3.2.2 Cool days (TX10p) and cool nights (TN10p)

Cool days (TX10p) indicate a significant decreasing trend over all three AEZs (see Figure 3). The magnitude of the decreasing trend is about -0.04440c, -0.046990c, -0.03640c in highland, midland, and lowland AEZs, respectively; TN10p is contrasting to Tx10p with its insignificant and varied trend across the three AEZs. Cool night is decreasing in the highlands and increasing in the midlands and lowlands. The studies indicated the same result that the cool days (TX10p) are decreasing ( Damtew et al., 2022; Esayas et al., 2018; Etana et al., 2020; Khan et al., 2022). The decreasing trend of the cool night is observed in the highlands, and it is increasing in the midland and lowland, which is the same with other studies conducted in different parts of the world ( Berhane et al., 2020; Dendir & Birhanu, 2022; Gunawardhana & Al-Rawas, 2014; Wubaye et al., 2023).

3.2.3 Warmest day and coldest day

TXx and TXn (see Figure 4) have an increasing trend in all AEZs. TXx is increasing significantly in all AEZs we have at hand, and its annual magnitude of change is 0.042°C, 0.0396°C, and 0.0385°C in the highland, midland, and lowland, respectively. The change of coldest day (TXN) trend is also significantly increasing in all AEZs except lowland, which is statistically insignificant. The annual amount of trend change is 0.0440c and 0.04340c in the midland and lowland. Other studies showed the same result in different parts of the county ( Gedefaw, 2023; Terefe et al., 2022).

3.2.4 Coldest night and warmest night

Warmest night (Tnx) (as indicated in Figure 5) had a positive and significant change in the midland and lowland, but the positive change in the highland is insignificant. TNn (coldest night) is positive and insignificant in the highlands but negative and insignificant in the midland and lowland agroecologies. The findings of other studies and the result of this study conform ( Terefe et al., 2022; Wubaye et al., 2023). The positive trend of TNn started its change in 1993 in the highland, and it has been the opposite in the midland and lowland since the same year. This year is also the beginning of a trend change in TNx, either negative or positive across AEZs.

3.2.5 Diurnal temperature range and number of summer days

DTR in both highland and lowland indicates a positive and insignificant trend, whereas it is positive and significant in the midland (see Figure 6). An increasing trend of DTR suggests that the daily maximum temperature is greater than the daily minimum temperature; this can imply, along with the increasing DTR of the study area, which is also concurring with the rising mean temperature of the earth’s surface, negatively contributing to agriculture, especially to small scale subsistence farming in the midland area of this study. As earlier studies confirmed, the change in DTR is a sign of climate change, which is devastating the current livelihood of the world, mainly the agriculture sector. Thus, the change in DTR with other climate extreme indices can give a different meaning to Sub-Saharan Africa, including Ethiopia, where the livelihood of the population is solely dependent on agriculture (sensitive to climate change) ( Damtew et al., 2022; Etana et al., 2020). The SU is positive and significant with its magnitude 0.0524°c, 0.058°c, and 0.046°c in the highland midland and lowland AEZS ( Damtew et al., 2022).

3.2.6 Warm spell duration indicator (WSDI) and cold spell duration indicator (CSDI)

The warm spell duration index significantly increases in the highland, midland, and lowland AEZs (see Figure 7); this finding is similar to other studies ( Damtew et al., 2022). However, the cold spell duration index is insignificant and decreasing in the lowland, which agrees with another study ( Esayas et al., 2018), and is insignificant and increasing in the highland and midland AEZs, which has the same meaning as the study in central rift valley area ( Terefe et al., 2022). The WSDI is a count of days at least six consecutive days contributing to the warm condition when the maximum temperature is greater than the 90th percentile, while CSDI is also a count of days at least six consecutive days contributing to the cold condition when the maximum temperature is greater than the 90th percentile.

3.3.1 Trends in precipitation extremes

The results of Consecutive dry days (CDD) and consecutive wet days (CWD) imply no significant trend in all AEZs, as seen in Figure 8. CDD was highest in the highlands in 2008, 2012, and 2007; 1997 and 2012 were the highest CDD recorded in the midlands; and 1997, 2008, and 2015 were also years of higher records in the lowlands. Regarding the CWD, the peak years were 2001, 1997, 2000, and 1997 in the highland, midland, and lowland AEZs. The CDD increasing trend and the CWD decreasing trend of this finding are similar to another research finding ( Teshome & Zhang, 2019; Wubaye et al., 2023).

3.3.2 Number of heavy (R10mm) and very heavy precipitation days (R20mm)

The number of heavy precipitation days decreased in all AEZs of the study area, but the decreasing trend is insignificant (see Figure 9). There are similar studies that found a negative trend of R10mm among AEZs; some were in the line of agreement ( Damtew et al., 2022), and ( Esayas et al., 2018) revealed the same finding; the Wolaita zone is part of the it. Furthermore, the result of very heavy precipitation days (R10mm) increased in both the highland and midland; on the contrary, the lowland had a negative insignificant trend. This finding matches the other study with varied decreasing and increasing trends in different stations ( Kiros et al., 2017). The heavy day precipitation peaked in 1981 in both highland and lowland AEZs, but in the same year, the amount was very low compared to the preceding two. The same year also had very heavy precipitation days (R20mm). In 1996, peak times of R10mm and R20mm in the lowland and midland got higher R20mm (even more than in 1981) in addition to 1981.

3.3.3 Maximum 1-day (RX1day) and 5-day (RX5day) precipitations

As indicated in Figure 10, the positive insignificant trend of RX1day was observed both in the highland and lowland, whereas its trend was negative and insignificant in the lowland. Regarding RX5day, the trend was negative and insignificant in all AEZs. The Rx1day and Rx5day precipitation were highest only in highlands in 1992, 1993(only RX1day), and 1996. 1986, 1996, and 1997 were higher than RX5day in lowland agroecology. The RX1day trend showed quite a spatial variation across AEZ. Similarly, the study found fragmented results across AEZs ( Damtew et al., 2022). The same source found that Rx5day has an increasing trend; of course, the result is more flat than this project did. Another study indicated a decreasing trend of RX5day, which goes together with the findings of this paper ( Khan et al., 2022).

3.3.4 Very wet days (R95p. and extremely wet days (R99p)

Very wet days (R95p) of the study area indicate that all AEZs have an insignificant increasing trend ( Figure 11), the same result from other research outputs ( Worku, 2019). R95p was extremely higher in the highland than any other in the year 1981, next at a lower rate in 1996 and 2005 in the second and third years of the high record. In the same way, the lowland AEZ has the next highest record area compared to the midland, where the years 1996, 1997, 2000, and 2003 are highly considered.

However, R99p had only an insignificant increasing trend in the midland with a similar result of other research ( Worku et al., 2019), but the remaining two agroecologies (highland and lowland) experienced a negative insignificant trend, which shares the same result of other research output ( Kebede & Bewket, 2009; Teshome & Zhang, 2019). Though the higher record of R99p in all AEZs, the years 1992, 1996, 2000, 2001, and 2015 were meaningfully higher in the highland, whereas in the midland and lowland agroecology 1987,1992,1997 and 2003, and 1986,1987,1992, 1996,1997,2002 and 2013 were a year of higher R99p, respectively.

3.3.5 Total Precipitation (PRCPTOT) and Simple Daily Intensity Index (SDDI)

AEZs, there is an insignificant negative trend in extreme rainfall indices ( Figure 12), which has the same analysis as other research results ( Omondi et al., 2014; Santos, 2014). While SDII has an insignificant decreasing trend in the highland and lowland, it has an increasing and insignificant in the midland. The highest SDDI was recorded in 2000 in the lowlands, while the highest records in the midlands and highlands were recorded in 1981 and 1981, respectively. Regarding the PECPTOT, 1996, 1981, 1996, and 1996 had the highest record in the study area’s highland, midland, and lowland agroecology.

3.4.1 Standardized precipitation–evapotranspiration index (SPEI)

The study area has frequent and fast drought occurrences (see Table 4). Drought has a rate of extreme, severe, moderate, and mild, which occurred 4, 9, 8, and 22 times between 1981 and 2018, respectively. This finding is consistent with other research showing that Wolaita is a frequent drought-hit area ( Mohammed & Yimam, 2021). The analysis by Husen in the upper Awash sub-basin found that the area is experiencing crop failure due to drought ( Maru et al., 2021). The same source indicated that the lowland is more sensitive to crop failure and production reduction. In line with this, the recent frequent occurrence of drought in the Hamassa watershed, as seen in Figure 13, might help us conclude the crop and livestock failure similarly. The highland of the Hamassa watershed was attacked by severe and extreme drought in the years 2015, 2009, and 1990. The midland area in the years 1991, 2004, 2009, and 2015 was shocked by severe and extreme droughts, and 1985, 2009, 2012, and 2014 are identified as years of extreme and severe droughts in the lowland area. In all AEZs, the decreasing trend shows the increased frequency of drought in recent years, which might cause a reduction in agricultural production by suppressing soil moisture, concurrent with other findings ( Lu et al., 2020).

Farmers in three agroecologies uncovered the presence of climate change variability and climate extremes, though their tone varied across agroecologies. Of the total surveyed farmers, about 90 percent (296 participants) perceived a change in climate and extreme climate when asked whether they perceived a change in climate and extreme climate. The KII and FGD also revealed the existence of climate change, variability, and extreme events. The discussion yielded broad climate change effects on farming in the area through rain commencement, amount, and length of rainy and dry seasons.

Key informants from the Woreda agriculture office in the midland agroecology stated that.

This shows that the climate is changing in the study area, highly impacting smallholder agricultural production. The converging climate data analysis and climate change perception of farmers in this study align with the findings of another study ( Damtew et al., 2022). Climate change affects all agroecologies but is especially pronounced in midland and lowland regions. The focus group discussion in the lowland agroecology expressed the climate conditions:

Water scarcity and drying conditions have worsened for study participants. The findings align with the droughts noted in Table 4, analyzed through SPEI of meteorological data. The results of the present study align with the existing literature on drought management and call for the development of effective and sustainable strategies to mitigate the adverse impacts of water scarcity and drought on both human welfare and the environment ( Maru et al., 2021). Land productivity decline is linked to precipitation changes, per Tables 5 and 6.

The research received approval for ethical clearance by College of Development Studies Institutional Review Board (CoDS-IRB) at Addis Ababa University with No. 035/01/2023 on October, 27, 2023 after it was determined that it complied with all requirements. Participants’ informed consent must be obtained by the ethical clearances and criteria of the CoDS-IRB, allowing researchers to use written, verbal, or oral informed consent processes as needed. Oral consent was thus obtained from the participants for this study with approval from the IRB, which was indirectly indicated in the approval letter, allows any consent of the three methods mentioned above as the situation allows the researcher. Thus, with the IRB’s approval, oral consent was obtained from the participants for this study. Oral consent was secured from the participants using an audio recording device, specifically the ZTE Blade A71 mobile device (model-ZTE A7030 with Serial number 320225317551). The study objectives were clearly explained to the participants to ensure their understanding, and they were assured that their identities would be kept confidential in all project documentation. Those who volunteered for the study and placed trust in the project’s ethical standards, which were approved by the Institutional Review Board (IRB), provided their oral consent. This oral consent was obtained in accordance with the principles of informed consent, and the use of an audio recording device to document the consent process was explicitly stated in the preceding sentences. Oral consent was sought due to some respondents’ inability to read and write. Thus, the oral consent was chosen to collect the data uniformly for all respondents.