Precipitation Variability and its Teleconnection with the Global SST and ENSO Indices in the Food Insecure Rural Areas of Tigray


 The impact of precipitation variability on food production is very significant. For food insecure rural areas, understanding the nature of precipitation variability and its teleconnection has paramount importance in guiding regional and local level decisions. In this study, we analyzed the monthly, seasonal and annual precipitation variability and the strength of its teleconnection with the global sea-surface temperature (SST) and El Niño Southern Oscillation (ENSO) indices in the food insecure rural areas of Tigray region, Ethiopia. The precipitation, SST, and ENSO indices data for the study were used from 1979 to 2019. A Summary of descriptive statistics and Mann Kendall test methods were applied to detect existence of trends; and Sen’s Slope and coefficient of variation are used to analyze the magnitude of the trend, and degree of variation in the trend of precipitation. Further, Pearson’s correlation is used to determine the effect of ENSO, and SST variations on the precipitation using the Canonical Correlation Analysis (CCA). The results revealed that the precipitation over the study areas is characterized by a distinctive bi-modal pattern with limited rains in March – May preceding the main rainy season June – September. The limited amount of precipitation, exacerbated by higher degree of variability, makes the food production in the study areas more uncertain. Besides, there was a very significant decline in the trend of March – May average precipitation and a significant decline in the trend of the annual average precipitation of Hintalo area. The SSTs of the central and eastern equatorial Pacific Ocean, and northeast and northwest equatorial Atlantic Ocean was strongly correlated with April’s average precipitation of the study areas. Further, the SST of south, west and southwest of equatorial Indian Ocean, and west equatorial Pacific Ocean were associated with July – September average precipitation with greater variation in strength among of the study areas. Moreover, July’s average precipitation of all the study areas, April’s average precipitation of Atsbi and Eirop, and May’s precipitation of Hintalo are found significantly associated with the ENSO indices of JFM, FMA, MJJ and MAM. Therefore, the task of achieving food security in the study areas should incorporate the design of informed food production strategies that can adapt the limited and variable precipitation based on these SST and ENSO indices.


Introduction
Climate change, the catchy phrase, is the major threat to food security in rural areas. Because of the limited capacities to cope up with the varying climate, food insecurity is higher in rural areas of developing countries, where much of their population depends on rain to produce food. World Bank (2016) reported that rural farmers are more than four times as likely to be food insecure as compared to urban dwellers engaged in non-agricultural sectors. According to the findings of a research conducted across 105 countries by Alkire et al. (2014), 86 percent of food insecure people of Sub-Saharan Africa and South Asia live in rural areas.
Although many factors are associated with food insecurity in rural areas, food shortage, the main feature of food insecurity, is often associated with precipitation shortage and variability. A number of scholars like Darwin (2001), Schmidhuber and Tubiello (2007), Wheeler and Braun (2013), agree on the significant impact of precipitation variability on food production. Kinda & Badolo (2019) showed that precipitation variability has reduced food availability per capita and increased fluctuation in food production for 71 developing countries from 1960 to 2016.
According to Von Braun (1991), a 10 percent decline in the average amount of precipitation leads to a 4.4 percent reduction in the food production.
Ethiopia is predominantly an agrarian nation, in which more than 80 percent of the population relies on agriculture. More importantly, nearly 90 percent of the smallholder farmers mainly depend on rain-fed agriculture (Alhamshry et al., 2020). For this reason, Ethiopia listed as the most vulnerable to adverse impacts of climate variability (World Bank, 2010). Yet, not all parts of the country are equally vulnerable to the impacts of climate variability. Subsistence farmers were relatively the most susceptible to climate variability in Ethiopia (Asfaw et al., 2018).
Subsequently, the problem of food insecurity in Ethiopia is more pronounced in rural areas (World Finance, 2017). Further, an overdependence on the rain-fed agriculture was one of the reasons for the pervasiveness of food insecurity in rural Ethiopia (Mekonnen & Gerber, 2017).
In Ethiopia, the intensity and variability of precipitation have been important determinants of food security in rural areas (Demeke, 2011;Alemayehu & Bewket, 2016;and Agidew & Singh, 2018). During the period of 1983-1985, Ethiopia's Tigray region experienced the severest food insecurity induced by drought which caused an estimated one million people deaths (Reid, 2018). In the period of 2015-2016, an El Niño induced drought took place mainly in the lowlands of the country (Singh et al., 2016). During that time, a quarter of Ethiopian population was food insecure and more than 18 million people were requesting for urgent food aid (Mohamed, 2017).
Tigray is one of the regions in Ethiopia, which, over the past many decades have been affected by recurrent droughts (Endalew et al., 2015). In the region, the average precipitation is very short and variable as compared to the southern and western parts of Ethiopia (Seleshi & Demaree, 1995;Woldehanna, 2000;and Weldearegay & Tedla, 2018). The average annual precipitation of Tigray for the last 20 years was 725mm (Weldearegay & Tedla, 2018). According to Demeke et al. (2011) and Weldearegay & Tedla (2018), precipitation variability was the major cause of food shortage in Tigray region. Having this uncertain and lesser precipitation, the rural farmers in Tigray still depend on the unreliable precipitation to produce food.
Although precipitation variability has many forms, the intensity and timing in the precipitation pattern is the main form of variability. According to Torres et al. (2019), variations in timing and intensity of precipitation are higher in an intra-year than inter-yearly. Intra-year variability matters most as many rural farmers keep months and seasons to do the farming activities.
Precipitation variability is highly determined by global sea-surface temperature (SST) and El Niño Southern Oscillation (ENSO). The SST is among the major drivers of precipitation variability (Dittus et al., 2018), particularly for the Ethiopian precipitation (Alhamshry et al., 2020). Besides, the ENSO is the other most important determinant of precipitation variability, particularly in the precipitation pattern of Ethiopia (Kasie et al., 2019& Tefera et al., 2020.
These two factors highly determine the intra-year and inter-year precipitation variability.
However, different areas have variant sensitivities to the SST and ENSO indices; and the impacts of those indices greatly vary from place to place. Hence, studies of precipitation variability and its associated climate factors have to be conducted at the lowest possible geographical spaces.
Furthermore, how significant is the influence of SST and ENSO in the precipitation variability of the food insecure rural areas of Tigray region has not been well researched before. Therefore, this study will investigate the precipitation time-series trend on monthly, seasonal and annual scales, its degree of variability, and detect the significance of the global SST and ENSO indices impact on the precipitation pattern of the food insecure rural areas of Tigray region, Ethiopia. This would be crucial for guiding local level food production related decisions and indicate feasible adaptation strategies to reduce the risk in food production.

Methodology
The study has constituted top three food insecure rural districts found in Tigray region. All these three districts were purposefully selected based on the recent data from the regional office of food security. Accordingly, rural areas of Atsbi Wenberta, Eirop, and Hintalo Wajerat districts were the three most food insecure rural districts of Tigray region.
Table1: Geographical location of study areas

Name of study areas Elevation (m)
Latitude ( Where is the Mann-Kendal's test statistics, is the number of data points, and are the data values in the time series and ( > ) respectively. The sgn X − X is the sign function as indicated in equation (2): The variance is calculated as equation (3): Where is the number of data points, m is the number of tied groups (a set of sample data that have the same value), the summation sign (Σ) indicates the summation over all tied groups, and t is the number of data point for the th tie. If there are no tied groups, this summation process can be ignored. In the case where the sample size > 10, Z approximates the standard normal distribution with the mean ( ) = 0 and computed using equation (4): The presence of a statistically significant trend is evaluated using the Z value. Positive values of Z indicate increasing trends, while negative Z values show decreasing trends. Testing trends are performed at the specific α (0.05) significance level. When |Z |> Z1-α/2, the null hypothesis is rejected, indicating that a significant trend exists in the time series. Z1-α/2 is the critical value of Z is obtained from the standard normal distribution table which is 1.96.
The Mann-Kendall test only indicates the direction; hence, the magnitude of the trend is usually determined by Sen's test which is defined by calculating the slope. The slope (change per unit time) was estimated based on the procedure in equation (5) and (6).

Q = (5)
Where X and are considered as data values at time i and j (i >j) correspondingly and is the Slope. The Sen's estimator is computed as Q = Q if N appears odd, and it is considered as Q = (Q + Q ) if N appears even which is given as: Q is computed by a two-sided test at 100 (1-α) % confidence interval and then a true slope can be obtained by the non-parametric test. Positive value of Q indicates an upward or increasing trend and a negative value of Q gives a downward or decreasing trend in the time series.
Coefficient of variation of the annual precipitation variation is calculated as equation 7: Where Sd is the standard deviation which is computed by square root of the variance, and X is the sample mean of the annual average precipitation.
In this study, we use Pearson's correlation coefficient with the significance assessed using the precipitation pattern of the study areas and the likelihood global SST and ENSO variations. The Pearson correlation coefficient in this study is defined as follows: Suppose that there are two variables X and Y, each having n values X1, X2, … Xn, and Y1, Y2, … Yn, respectively. Let the mean of X be x̄ and the mean of Y be y ̅ . Then, Pearson's r is given by (8) where the summation proceeds across all possible values of and in this sample.

Results and Discussions
In Ethiopia, particularly in Tigray region, smallholder farming is the common agricultural practice where farmers usually depend on precipitation. For the reason that precipitation is a natural phenomenon that varies timely and spatially, understanding the nature of precipitation variability can minimize the possible risks on rural households. It can also significantly improve the food security status of the vulnerable rural community.
Food insecurity, in many parts of the world, is associated with shortage of rainfall (Afifi et al., 2014); and in Ethiopia, rainfall variability is among the primary drivers of food security (Lewis, 2017). In areas where rainfall is relatively low, efficient utilization of the precipitations of all the time matters most. Atsbi wenberta, Eirop, and Hintalo wajerat are the most food insecure rural areas of Tigray region. Table 1 shows that these areas can get precipitations in any month of the year. This implies for an enhanced utilization of the water precipitated in any time and place of these areas.
Regardless of the variation in the amount of precipitation, figure 2 shows that all the study areas have a bimodal nature of rainfall pattern with higher and lower amount of precipitation during the JJAS and MAM seasons, respectively. This corresponds with findings of Kahsay et al., (2019) & Gebru (2020) who found a bimodal nature of rainfall pattern in Eastern Tigray and southern Tigray. In the study areas, August was found to be comparatively the rainiest month with maximum average amount of precipitations in the time span; and the precipitation during July has never been zero for the last 41 years. The monthly average amount of precipitation was relatively lower from October to February; and it was higher during June to September. Yet, the range between the maximum and minimum amount of monthly precipitations was extremely higher in the rainy months than the drier months. The average annual precipitation of the study areas, i.e. Atsbi, Eirop, and Hintalo was found to be 542.5, 318, and 520.7mm, respectively. This is much lower than the national average and even from the regional average. According to Weldearegay & Tedla (2018), the average annual precipitation of Tigary region for the last 20 years was 725mm. Although crop production with these amounts of rainfall may be possible, it is not sufficient and reliable. In line with this, FAO (1986) suggests irrigation based crop production in areas where average annual precipitation is less than 1200mm. FAO (1986) also suggests that irrigation is a must in areas with less than 400mm of average annual precipitation. In the study areas, let alone the average annual precipitation, the maximum annual precipitation recorded in the last 41 years was less than 1200mm. More extremely, the maximum annual precipitation for Eirop district was only 643.9mm. In this condition, achieving food security would be difficult using the conventional farming system. When we see the variability in the annual precipitation depicted in Figure       In line with this, the variability of the monthly average precipitation depicted in table 3 shows that there was extremely very high variability during all months except for July and August which was relatively much lower than the other months. Seasonally, the coefficient of variation was nearly similar for both MAM and JJAS seasons of Atsbi and Eirop districts. However, the variation during MAM was much higher than JJAS for Hintalo. The coefficient of variation for the annual precipitation was almost similar and high for all the districts, yet it was higher than the regional average, which was 16 percent during 1997 -2017 (Weldearegay & Tedla, 2018).
This implies for a shift from the rainfed to irrigation based agricultural systems. In southern Tigray, the coefficient of variation for the annual precipitation during 1981-2010 was ranging from 33.77 -233 percent (Hayelom et al., 2017). On the other hand, annual precipitation data of 40 years from 109 meteorological stations in Ethiopia showed a coefficient of variation ranging from 20 to 89 percent (Addisu et al., 2015).
In order to monitor this precipitation variability, it is crucial to specify the causal factors and study their correlations. The global SST is among the key factor that plays a significant role in determining the variability of the monthly, annual and decadal precipitation patterns. Alhamshry Previous studies confirm that the spatial and temporal variability in the precipitation of Ethiopia is attributed to the variations in SSTs over the Atlantic, Indian, and Pacific Oceanic indices (Degefu et al., 2017;Zeleke et al., 2017;Dubache et al., 2019;Alhamshry et al., 2020;Molla, 2020;Tefera et al., 2020;& Bayable et al., 2021).
The strength in the statistical association between Ethiopian precipitation and global SST greatly varies with time and space. In the central and western Ethiopia, the equatorial east Pacific and Indian Ocean SSTs were found to be correlated with JJAS precipitation (Degefu et al., 2017).
Besides, the drying trend in the southern and northern of Ethiopia is associated with Atlantic There was a strong and positive correlation between the precipitation in northeastern Ethiopia and southern Indian, the Atlantic, and most of the western Pacific Ocean (Gobie & Miheretu, 2021). In Tigray region, tropical Indian Ocean was identified as statistically significant drought influencing factor (Tefera et al., 2020). The tropical Indian Ocean, tropical Atlantic Ocean, tropical Pacific Ocean, the Red Sea and Nino 3.4 regions were the other drought influencing factors on an annual scale (Tefera et al., 2020); and other events like Pacific Decadal Oscillation, Southern Oscillation Index and Indian Ocean Dipole were the important factors for causing meteorological and agricultural droughts in Tigray region (Molla, 2020).
For the reason that many of the correlation analysis made so far were larger in spatial and time scope, a monthly average precipitation of the study areas was used to test for its correlation with the global SST. This is vital in understanding the correlations more in depth.
As shown in figure 6, the canonical correlation analysis shows that the MAM precipitation of the study areas was correlated to the global SST with different correlational values. But, most of these correlations were not strong enough to determine the average monthly precipitation of the study areas. The central equatorial Pacific Ocean was strongly and positively correlated with April's average precipitation of Atsbi and Eirop. Besides, April's average precipitation of Atsbi and Hintalo was strongly and negatively correlated with the SST of eastern equatorial Pacific Ocean. Further, the April's average precipitation of Hintalo has shown a strong negative correlation with the SST of northeast and northwest equatorial Atlantic Ocean.
This implies that the declined April's average precipitation over Atsbi and Hintalo districts is associated with the warming in the central and eastern equatorial Pacific Ocean. In contrast, the eastern equatorial Pacific Ocean was source of the limited amount of April's precipitation for Atsbi and Hintalo districts. Thus, the projected SST in these regions can be used to predict the precipitation so as to guide food production strategies during the month.  interaction that occurs mainly in the tropical and sub-tropical Pacific Oceans. This naturally occurring phenomenon is the most predictable climate system at the time scales from months to seasons and years (Tang et al., 2018), providing the basis for regional and local level precipitation predictions.
ENSO has a significant climate influence on climate patterns of various parts of the world including Ethiopia. Although the different parts of Ethiopia have different climate sensitivities, the 2015 drought that occurred in most parts of the country and East Africa was associated with ENSO induced rain shortages (Philip et al., 2018;& Bayable et al., 2021). In northern Ethiopia, periodicity in dryness and wetness were largely determined by ENSO variability in both the spring and summer rainy seasons (Zeleke et al., 2017).
In northeast part of Ethiopia, La Nina was associated with increased rainfall in most parts of the region and that El Nino's was associated with decreased rainfall in limited parts of the region (Gobie & Miheretu, 2021). In Tigray region, in addition to the global SST indices, ENSO was identified as drought influencing factor (Molla, 2020& Tefera et al., 2020. Similarly, Looby   -0.190 -0.197 -0.196 -0.256 Hintalo -0.296 -0.270

Conclusion
Food security depends on the four pillars: food availability, access, utilization and stability. And lack of adequate precipitation and its variability is among the main causes of food insecurity.
Nevertheless, food availability is mostly dependent on precipitation compared to the other pillars, particularly in areas where the farming system is dominantly rainfed.
The study areas have showed a similarity in the magnitude and seasonality of precipitation.
March -May and June -September were the wettest seasons of the study areas. While the March -May average precipitation of the study areas have showed a decreasing trend, an insignificant increment was observed in the precipitation pattern of June -September. This implies for careful decisions to be made for any agricultural practices during these seasons.
On the other hand, a marked variation was observed in the trend of the precipitation pattern among the study areas. Although no significant trend was detected for Atsbi and Eirop precipitation pattern, a very significant decline in the trend of March -May average precipitation and a significant decline trend in the annual average precipitation of Hintalo area is observed.
This calls for an immediate action to get alternative sources of water for the rural areas where precipitation is the only source for food productions.
More importantly, not all of the rainy months of the study areas were significantly susceptible to the global SST and ENSO variations. The April's average precipitation of the study areas is found being under the influence of central and eastern equatorial Pacific Ocean and northeast and northwest equatorial Atlantic Ocean SSTs. Further, the SST of south, west, and southwest of equatorial Indian Ocean, and west equatorial Pacific Ocean were associated with July -September average precipitation with greater variation in strength among of the study areas.
The ENSO indices, on the other hand, are found to be significantly unrelated with many of the rainy seasons of the study areas. Nevertheless, July's average precipitation of all the study areas, April's average precipitation of Atsbi and Eirop, and May's precipitation of Hintalo are found significantly associated with the ENSO indices. Therefore, SSTs of central and eastern equatorial Pacific Ocean, northeast and northwest equatorial Atlantic Ocean, southwest of equatorial Indian Ocean, and west equatorial Pacific Ocean, and the ENSO indices of JFM, FMA, MJJ and MAM can be used to develop a skillful precipitation forecast for the areas under the study.
Generally, the limited amount of precipitation, given its higher degree of variability, will be the major challenge in the task of achieving food security in the food insecure rural areas. Therefore, unless coping strategies are arranged that meet the varying monthly, seasonal and annual precipitation, or alternative water sources are used and the rain water is harvested prudently, these areas will continue to face the severe consequences of food insecurity.
Furthermore, the precipitation pattern and its statistical association with the SST and ENSO indices vary greatly in strength among the study areas. Thus, future researchers have to analyze precipitation patterns and any associations with their trend in clusters rather than as a whole. In this way, food security programs and actions can be effectively downscaled to the local level based on their respective precipitation pattern information.