Climate change and its implications for rainfed agriculture in Ethiopia

This study aims to investigate the spatio-temporal variability and trends in climate and its implications for rainfed agriculture in the Rib watershed, north-western highland Ethiopia from 1986 to 2050. The daily rainfall and temperature records for the period 1986–2017 were used to detect the variability and trends of the current climate using the coefficient of variation, precipitation concentration index, Mann–Kendall test, and Sen’s slope estimator. On the other hand, future climate changes (2018–2050) were analyzed based on the Coupled Model Intercomparison Project version 5 (CMIP5) model outputs under under two representative concentration pathway (RCP) scenarios, RCP 4.5 and 8.5. The results showed high inter-seasonal and inter-annual variability of rainfall and temperature in the studied watershed over the last four decades. The annual and Kiremt (June–September) rainfall showed a generally increasing trend, while the Belg (March–May) rainfall exhibited a decreasing trend between 1986 and 2017. Conversely, the minimum, maximum and mean temperature demonstrated increasing trends over the study period although most of the detected trends were statistically insignificant at 5 and 10% level of significance. Future climate analysis results showed an increase in future temperature and annual and Kiremt rainfall while Belg rainfall declined.


GRAPHICAL ABSTRACT INTRODUCTION
The world's climate has been changing for several thousands of years with a widespread impact on human and natural systems (Kotir ; Marohasy et al. ; Birara et al. ; Yadav ). However, its change has become more rapid and unusual in recent decades relative to that of the past, which can be shown through observations of increasing temperature, sea-level rise, increase in the emission of greenhouse gases (GHGs), frequent floods and droughts, and changes in the amount, distribution and patterns of rainfall (Asare-Nuamah & Botchway ).
Although climate change is global in its extent and impacts, Africa has been identified as the most vulnerable continent to climate change due to low adaptive capacity and high reliance on climate-sensitive sectors such as rain-fed agriculture (Conway &  Oceans significantly affect the rainfall distribution in Ethiopia by displacing and weakening the rain-producing air mass. This evidence provides a picture of how Ethiopia is highly exposed to changes and variations in rainfall and temperature at varying space and time. It has also been predicted that increasing rainfall and surface temperature in most of the East African countries will highly affect the availability of water and agriculture in Ethiopia (Fentaw et al. ; Muhati et al. ). To this end, a clear understanding of the temporal trends and spatial distribution of past and projected rainfall and temperature is crucial for proper planning and decision making. As the government of Ethiopa strives to expand agricultural production, reliable and timely climate change information is essential for planning and formulation of appropriate mitigation mechanisms.
While several studies have been conducted on observed climatic varability in many parts of Ethiopia ( Therefore, the main aim of this study is to understand and analyse the variability and changes of past and future temperature and rainfall conditions and their implications for rainfed agriculture in the Rib watershed, northwestern highland Ethiopia. The specific objectives include: (1) to assess the variability and trends of rainfall and temperature over the last few decades; (2) to project the rainfall and temperature changes in the next few decades; and (3) to analyze the possible impact of climate variability and changes on the rain-fed agriculture in the Rib watershed.

STUDY AREA
The Rib watershed is located in north-western highland Ethiopia between 11 40 0 and 12 20 0 north latitude and 37 30 0 and 38 20 0 east longitude with a drainage area of 1,975 km 2 . The area is characterized by irregular topography, valleys, and gorges with elevation ranging from 1,758 m in the western part of the watershed to over 4,100 m in the south-eastern part (Mount Guna Massif) (Figure 1). This diversity in topography enables the watershed to receive adequate rainfall to satisfy crop water demand and produce different crops and livestock. The mean annual rainfall in the watershed is about 1,503 mm, with a mean annual temperature of 15.6 C. The climate of the watershed can be divided into three seasons: summer or main rainy season, locally known as Kiremt season (June-September), dry or Bega season (October-February), and short rainy or Belg season (March-May). The rainfall in the study area is generally erratic and unevenly distributed with more than 80% of rainfall occurring during the main rainy season. The agriculture of the watershed is characterized by mixed croplivestock systems. The major crops grown include maize, barley, wheat, beans, and rice.

DATASETS AND METHODOLOGY
The long-term monthly rainfall and temperature data for both current and future timescales were considered in this study. The daily precipitation and temperature records for twelve weather stations located within and around the Rib watershed were obtained from the National Meteorological Agency (NMA) of Ethiopia for the period 1986-2017. After testing the consistency, spatial representativeness, and continuity of data for each station, only six stations (Deberetabor, Kimirdingay, Yifag, Alember, Gassay, and Addiszemen), which have a relatively long period of data (at least 30 years) and have no more than 10% missing values, were selected and used for further analysis. Table 1

Root mean square error
The RMSE was used to evaluate the difference between observed and simulated rainfall data where the values where R sim and R obs denote the simulated and observed rainfall data, respectively, and N represents the analysis period (32 years).

Coefficient of variation
The temporal variability in the annual and seasonal rainfall was assessed using the CV. The CV was used to measure the dispersion of seasonal and annual rainfall from the mean values. It represents the ratio of the standard deviation to the mean and can be computed using the following formula: where σ is the standard deviation, and μ is the mean rainfall where R sim and R obs denote the simulated and observed rainfall data, respectively, and the bar over the variables denotes the average over the analysis period .

Correlation coefficient
The correlation coefficient ( where R sim and R obs denote the simulated and observed rainfall data, respectively, N represents the analysis period (32 years), and the over-bar symbol denotes the mean statistical operation over the analysis period (1986-2017).

Precipitation concentration index
Oliver's () Precipitation Concentration Index (PCI) was used to assess the seasonal and annual variability of rainfall in the study area. The PCI measures the temporal distribution of monthly rainfall and helps to assess the seasonal where pi denotes the monthly precipitation in the month i.
The constant multiplier of 33.3 indicates the share of rainfall that occurred in four months (June-September) out of the 12 months' total rainfall (i.e. 33.3% kiremt rain out of the 100% total annual rainfall) and, similarly, 25 infers the rainfall that occurred in three months (March-May), i.e. 25% of the total annual rainfall. Since the rainfall in the study area commonly occurs during the kiremt and belg seasons, the bega season (dry period) was not considered in the analysis.

Mann-Kendall (MK) test
The long-term annual temperature trends across the water- where x j and x k are the sequential data values of the time series in the years j and k ( j > k), and n is the length of the time series. In cases where the sample size is greater than 10 (n > 10), the statistic 'S' is approximately normally distributed with mean, and 'S' becomes zero (Kendall ). In this case, the variance statistic is given by: where n is the number of observations, q is the number of tied groups, and t p is the number of data values in the pth group. Then, the values of 'S' and 'VAR(S)' are used to compute the test statistic 'Z' as follows: where the test statistic Z follows a normal distribution, the positive and negative values of Z indicate the increasing and decreasing trends, respectively. The presence of a statistically significant trend was evaluated using Z values at 5 and 10% significance levels.

Sen's slope estimator
A non-parametric method known as Sen's slope estimator where xj and xk are data values at times j and k ( j > k), respectively. The median of these 'N' values of Q i is Sen's estimator of slope and is calculated using the following formula:

Variability and trends of observed rainfall
The seasonal and annual variability of rainfall in the watershed was analyzed using the CV and PCI. Table 3 summarizes the results of seasonal and annual rainfall variability across the Rib watershed over the past few decades.
The mean rainfall for Kiremt season varies from 881 to 1,214 mm while the mean Belg rainfall varies from 84 to 184.6 mm. This shows that much of the rainfall in the watershed comes during the Kiremt season (June-September), which contributes more than 80% of the rainfall. Most of the analyzed stations show moderate to high seasonal rainfall variations. As shown in Table 3, the Kiremt season shows relatively lower variability and low concentration of rainfall compared to that of Belg and annual periods.
The trends in annual and seasonal rainfall were assessed by applying Sen's slope and MK tests at 5 and 10% levels of  scales, we noticed that the highest values of PCI belong to less rainfall receiving areas, implying more rainfall irregularity in small rainfall receiving areas compared to that of high rainfall receiving areas.

Variability and trends of observed temperature
The changes in long-term minimum, maximum and mean monthly and annual temperatures were analyzed to detect the variabilities and trends across the watershed for the period 1986-2017. The results of descriptive statistics of monthly minimum, maximum, and mean temperature calculated for the average of six stations are summarized in Table 5. The results indicate that the monthly temperature in the watershed has exhibited significant variation during the last few decades. The temperature in the watershed was relatively low during the monsoon season (June-September) and high during pre-and post-monsoon months.

Change in projected rainfall
Projected changes in rainfall and temperature until the middle (2050) of the 21st century under RCP4.5 and 8.5 were analyzed based on the outputs of the five most commonly used GCMs. The projected mean annual and seasonal rainfall for the year 2050 is presented in Figure 3.

Change in projected temperature
The projection results from the selected model outputs under both scenarios reveal a significant increase in temperature over the Rib watershed in the middle of the 21st century, compared to the baseline period (1986-2017) T minimum , T maximum , and T mean indicate minimum, maximum and mean temperature, respectively.    The higher increase in growing season temperatures may adversely affect crop production, farm income, and food security in many ways, basically when combined with high inter-annual and intra-seasonal variability of rainfall.
The projected warming would reduce the grain yield of cereal crops of the watershed which is already experiencing significant reduction due to human-induced soil erosion (Moges & Bhat ). A study by Maharjan & Joshi () indicates that an increase in temperature significantly affects mean yield responses as well as yield variability of maize, millet, and sorghum in many African countries.
Heat stress reduces grain yield by raising evaporation and reducing water availability (

CONCLUSIONS
This study has presented a detailed analysis of the current and future rainfall and temperature variability and trends in the Rib watershed, north-western highland Ethiopia. A focus has been also given to describe the possible implications of observed and projected climate variability and changes in the farming system of the watershed. The results show that the watershed is highly exposed to spatiotemporal variations in the magnitude and direction of rainfall and temperature. Similar trends of change were found for the current and future rainfallan increase in annual and Kiremt rainfall and a decrease in Belg rainfall. Similarly, the analyses of recorded and projected temperature data indicate increasing trends in mean minimum and mean maximum temperature in the study area. The variation in the amount, distribution, and trends of rainfall and warming temperature could have a direct implication for the productivity of rainfed agriculture and the livelihoods of rural farmers who depend largely on agriculture. The observed rainfall and temperature variability and changes would cause excessive runoff and soil loss, exaggerate crop and livestock diseases, and increase evaporation and reduce water availability, thereby significantly influencing agricultural productivity, food security, and rural livelihoods. The findings of the present study would offer useful information to better understand the spatial variability and temporal trends of rainfall and temperature which are essential for water resource management and farming practices.

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