Analysis of Climate Variability and Trends for Climate-Resilient Maize Farming System in Major Agroecology Zones of Ethiopia

,

regions of Ethiopia since it has a high adaptive capacity to climate stress in tropical and subtropical environments [3]. Tough maize is cultivated intensively in the low-to medium-altitude agroecological zone, it is now competing for land in the high-altitude agroecological zone with wheat and tef [Eragrostis tef (Zuccagni) Trotter] in Ethiopia due to what might be climate change [4].
Te average maize productivity is 3 metric tons ha −1 which is low compared to the world average of 5.6 metric tons ha −1 [5]. Tis may be brought on by a number of factors, such as declining soil fertility, limited input use, poor seed quality, insect and disease infestations, inefective agronomic management techniques, and climate variability. Climate variability, particularly substantial geographical, and temporal variations in rainfall and temperature, has been one of the major causes of maize's low productivity in Ethiopia. Te recent studies [6][7][8][9][10] described in diferent areas would be strong evidence that the temperature and rainfall variability impact on maize production and productivity. Terefore, we choose rainfall and temperature as the major climate variables for this study since they have serious implications for maize vegetative growth and yield [11]. Indeed, some studies [12][13][14][15][16][17] and [18] have assessed climate trends and variability at diferent spatial and temporal scales. However, full comprehensive studies of climate variability, trend, onset date, cessation date, length of growing season have been limited at both the spatial and temporal scales levels in our study area. Tese attributes are extremely relevant for designing climate-resilient farming systems since they greatly infuence the agronomic activities to be done in rain-fed agricultural systems [19]. Hence, this paper examined the spatial variability and trends of the temperature and rainfall by using the Enhancing National Climate Services (EN-ACTS) climate dataset between 1987 and 2018.
Resilience can be understood as the ability or capacity of the farming system to adjust, absorb, or adapt to stress and change following a perturbation [20]. Tus, a "resilient" maize farming system would be capable of providing maize production when challenged by severe temperature and variability in a specifc maize production agroecological zone. Te major climate-resilient agronomic practices that are covered in these studies are the selection of a maize variety and setting planting and harvest dates for the current climate condition. Tese practices involve the entire community, whose livelihood is strongly dependent on maize. Terefore, the spatial climate variability and trend analysis on an agroecology basis were an urgent need and highly relevant for restructuring maize farming systems in the maize-growing agroecology zones in our study area.
Agroecological zones indicate major physical conditions that are grouped into relatively homogenous areas having similar altitudes, climates, soil, and agricultural land uses. We choose agroecology as the most appropriate spatial frame for analysis. As such, it is a more meaningful unit than an administrative boundary or river basin when considering climate-resilient agronomic adaptation in rain-fed subsistence agriculture. Terefore, we followed the climate variability and trend analysis based on a specifc cropagroecological zonation approach, which is innovative in sub-Saharan Africa in general and Ethiopia in particular. Terefore, the aim of this study was to analyze rainfall and temperature variability and trends across the maizeproductive agroecology zones in northwest Ethiopia.

Description of the Study Site.
Tis study has been conducted in the maize production belt of Mecha areas in northwestern Ethiopia's agroecology zones (Figure 1). Te geographic location of these areas extends from 11.30°N latitude to 37.20°E longitude. Te altitude varies from 1000 to 3131 meters above sea level. Te average annual rainfall is 1397.37 mm, and the average annual temperature is 24.17°C. Based on the traditional agroecological classifcations of Ethiopia, the study area is categorized into three zones. Lowproductive agroecology zones (highland), high-productive agroecology zones (midland), and medium-productive agroecology zones (lowland) [4] constitute 32%, 51%, and 17% of the total land of Mecha areas. A mixed subsistence farming system that involves crop production and animal husbandry is the main economic sector on which the majority of the population in the study area primarily depends [21].

Data Type and Source.
Te daily precipitation and the minimum and maximum temperature data for the study areas in the period from 1987 to 2018 were obtained from the Enhancing National Climate Services (ENACTS) dataset. ENACTS is the frst high-resolution gridded (4 × 4 km) surface meteorological dataset developed specifcally for studies of surface climate processes in Ethiopia [22]. EN-ACTS data sources for 30 locations in the northwest Ethiopia were obtained from the Ethiopian National Meteorological Agency. In recent years, ENACTS data have been used for climate analysis because of the following basic reasons: 1. conventional measurements of climate parameters did not represent all the study sites due to a limited number of stations [23]. 2. Te distribution of the ground stations is quite irregular, and thus, the distance between stations could be quite big, sometimes more than 50 km. 3. Most stations did not provide long seasonal data records for trend analysis due to recently established [13,22,23] Te coefcient of variation (CV) is a relative measure of variability that indicates the size of a standard deviation in relation to its mean. Accordingly, the variability of annual and seasonal was determined by calculating the coefcient of variation (CV): where CV is the coefcient of variation, σ is the standard deviation, and µ is the mean precipitation of the recording period. According to Hare [24], the degree of rainfall variability is categorized as follows: CV < 20%, indicates less variable, CV � 20% to 30%, indicates moderately variable, and CV � 30%, indicates highly variable. Precipitation concentration index (PCI) is the major tool that is the indicator of uniform or nonuniform rainfall patterns over the region. Accordingly, precipitation concentration index (PCI) was applied to evaluate the heterogeneity of intra-annual rainfall amount (PCI) [25].
Rainfall anomaly index (RAI) is used as a single hydroclimatic index for estimating climatic change wetness and dryness conditions. Terefore, annual variability of rainfall was evaluated using rainfall anomaly index (RAI) used by Tilahun [26] and is calculated as follows for positive anomalies: And for negative anomalies where RAI � rainfall anomaly index, RF � the actual rainfall for a given year, MRF � mean annual rainfall over the full record of analysis, MH10 � the mean of the 10 highest values of annual rainfall on record, and ML10 � the mean of the 10 lowest values of annual rainfall on record. Years with positive and negative anomalies indicate years of high and low rainfall, respectively, compared to the mean climatology. Annual temperature anomaly was computed as the diference between a year's average temperature and the longterm mean.

Trend Analysis.
All rainfall and temperature time series data were tested to examine the autocorrelation problem before applying the Mann-Kendall test by calculating the acf() function in R-software at a 5% signifcance level. Trend analysis was carried out using Mann-Kendall (MK) trend tests (nonparametric trend tests) and Sens's slope estimator in R-package modifed MK. Te MK trend test is the most appropriate and preferred nonparametric test for fnding trends in time series climate data [27]. Tis method is less infuenced by missing values and uneven data distribution, and it is less sensitive to outliers because it considers the ranks of the observations rather than their actual values [28][29][30]. S-Statists: S-statists was applied to check increasing or decreasing, or no trend on the hydrometeorological data series in each of the selected weather station data in the MK test statistic is given as follows:  Figure 1: Location map of the study area.

Advances in Meteorology
where n = number of data points, Xk and Xj = data values in time series k and j (j > k), and sgn(xj − xk) is defned as follows: where n is the number of data, m is the number of tied groups, and tp is the number of data points in the i th group. Z-statistics test: the test statistics Z was used as a measure of the signifcance of the trend. If the value of Z is positive, it indicates increasing trends, while negative values of Z show decreasing trends. Te values of S and VAR(S) are used to compute the test static Zs as follows: Z0 is a null hypothesis that signifes the trend is not signifcant and is recognized if Z-statistics are insignifcant statistically (Zα/2 < Z < Zα/2), where Zα/2 is the standardized normal deviation (Modarres & da Silva, 2007). Hence, for the purpose of this study, 1, 5, and 10% signifcance levels were considered. Sen's slope estimator and percentage change: the magnitude of the trend was estimated through Sen's slope estimator in a nonparametric procedure. Both the slope (i.e., linear rate of change) and intercepts were computed according to Sen's method [31]. Likewise, the linear model can be calculated as follows: where Q is the slope, B is constant, and a set of linear slopes is calculated as follows: where Q is the slope, X denotes the variable, n is the number of data, and j, k are indices where j > k. Te slope is estimated for each observation, and the corresponding intercept is also the median of all intercepts. Median is computed from N observations of the slope to estimate Sen's slope estimator: where n is the number of time periods. Te N values of Qi were ranked from smallest to largest, and the median of slope or Sen's estimator was computed as follows: Positive/negative values of Qi indicate an increasing/ decreasing trend, respectively.

Analysis of Onset, Cessation Date, and Length of Growing
Period. For this study, determining the onset, cessation date, and length of growing period (LGP) was performed by adapting the defnition from [32]. Instat software version 3.36 was used to analyze and estimate the onset, cessation date, and length of the growing period (LGP). Accordingly, the day with accumulated rainfall of 20 mm over three consecutive days that were not followed by greater than 9 days of dry spell length within 30 days from planting day was said to be the onset date. Te condition of having no dry spell lasting more than 9 days after the start of the growing season eliminates the possibility of a false start to the season whereas the stored soil water and its availability to the crop after the rain stops were a very useful criterion for determining the end of the growing season or rainy season [32]. Te cessation date of the rainy season adopted in this case was defned as any day after the frst of September or October, when the soil water balance reaches zero [32]. As a result, the planting and harvest dates for the current climate condition were determined using the onset and session dates of rainfall, respectively. Te length of growing period (LGP) refers to the average amount of time that rainfall from the onset to the cessation date. Te spatial distribution of crops and farming systems in any region is determined by the LGP. Te maturation duration based on the rainfall regime is important to consider when choosing the cultivars to be produced. Hence, the LGP was calculated between the start and end of the growing season using meteorological data  to recommend climate-resilient maize varieties suitable for each of the three main agroecological zones.

Rainfall
3.1.1. Coefcient of the Variability of Rainfall. Te results of the coefcient of variation (CV) of rainfall at the three maize-growing agroecology zones using annual daily time series climatic data from 1987 to 2018 are presented in Table 1. In comparison among agroecological zones, the annual rainfall variability was higher in the LPZ (CV � 16.4%) than in the HPZ (CV � 14.6%) and MPZ (CV � 14.1%). Tese indicated that the LPZ tends to show a slightly higher coefcient of variation than the MPZ and HPZ study area (Table 1). Tis may be due to the fact that the LPZ is found at a higher elevation than the MPZ and HPZ.

Precipitation Concertation Index.
Te proportions average value of PCI in the LPZ, MPZ, and HPZ maizegrowing agroecology zone were 22.2%, 19.06%, and 19.56%, respectively ( Table 2). According to De Luis et al. [25] PCI classifcation of rainfall variability, LPZ is classifed as signifcantly irregular, whilst MPZ and HPZ are part of the strong irregular rainfall distribution category (Table 2). Tese results in particular showed also that the annual rainfall variability was higher in the LPZ (highland agroecology zone) than in the MPZ (lowland) or HPZ (midland agroecology zone). Te fndings are directly comparable to those of Dawit et al. [33] who discovered a signifcant irregularity in precipitation distribution during the kiremt seasons in the Guna Tana Mountain watershed, Upper Blue Nile Basin, Ethiopia. Figure 2 and Table 3 display the results of the RAI distribution pattern for the main maize-growing agroecology zones in northwest Ethiopia. Out of the studied period (1987-2018), there were 5, 3, and 5 years that were extremely wet and 4, 5, and 3 years that were extremely dry under LPZ, MPZ, and HPZ, respectively (  1989, 1990, 1991, 2005, 2009, 2010, 2011, and 2012 was very dry to extremely dry years with average RAI values ranging from −2.00 to −4.47 (Table 3). Most of the years of negative rainfall anomalies in the study region coincided with El Niño events. For instance, the year 2009 experienced the highest negative anomaly due to the occurrence of El Niño which afected the main livelihood of the rural people in diferent parts of Ethiopia [34]. Terefore, this variability can be attributed to the seasonal movement of the intertropical convergent zone (ITCZ) as well as warm and cold ENSO events (El Nio and La Nia) [35].

Rainfalls Anomaly Index.
According to RAI intensity levels between 1987 and 2000, there was evidence of wetness in all maize-growing agroecology zone of northwest parts of the country. However, after 2000, the intensity has shifted to dry compared to 1987-2000 ( Figure 2). Tese indicated that the highest wet and dry years were observed in the 1990s and 2000s, respectively ( Figure 2). Currently, the implications of the results indicated that a negative anomaly (dryness) was now more prominent in the research area. Terefore, these conditions might result in erratic and unexpected rainfall patterns, which would have an impact on the production of maize and encourage the installation of climate-resilient adaptation measures in each zone.

Comparison between Rainfall Anomaly Index and Standardize Precipitation Index.
Te characteristics of the meteorological drought were also derived from SPI in the study area. Te analyzed SPI value in the study area was evaluated based on McKee and others' [36] classifcation system. Accordingly, an SPI value ranging from −1 to −2 categories indicates moderately to extremely dry drought events, whereas +1 to +2 showed moderately to extremely wet.
Te analysis of results revealed that SPI −1 to −2 categories in the LPZ indicate that moderate to extremely dry  Figure 3). Accordingly, our research shows that droughts now happen every one to three years in every agroecological zone in northwest Ethiopia that grows maize; even consecutive droughts were seen in 2009, 2010, and 2011. Because of this, it could be required to develop a system of climate-resilient maize farming to address the issues of both global warming and the sub-Saharan nations that make up our study area.
A comparison of RAI and SPI analysis was done as part of this study to look at precise and realistic data on the occurrences of meteorological drought episodes in the study area. Te procedure of comparison analysis was performed by overlapping the drought index from the RAI with the SPI in the same year. Figure 3 shows a comparison between SPI and RAI using the annual climate data from 1987 to 2018. According to the chart, distributions of SPI showed a similar tendency of increasing drought after 2000 in each of the maize-growing agroecological zones as revealed by RAI. So, the RAI and SPI displayed a pattern that was comparable, suggesting strong agreement between the two drought indexes. Te results from SPI were mostly similar to those from RAI when they were brought into common discourse in the index. LPZ � low productive agroecology zone, HPZ � high productive agroecology zone, MPZ � medium productive agroecology zone, CV � coefcient of variation; source: own study.  LPZ � low productive agroecology zone, HPZ � high productive agroecology zone, MPZ � medium productive agroecology zone, PCI � precipitation concentration index; source: own study. 6 Advances in Meteorology Agro.eo. � Agroecology, LPZ � low productive agroecology zone, HPZ � high productive agroecology zone, MPZ � medium maize productive agroecology zone, RAI � rainfall anomaly index.   (Table 4). Te trends of the change rate vary widely at spatial levels. Te result showed that annual rainfall has decreased by 6.7, 3.8, and 4.1 mm per annum in the LPZ (highland agroecology zone), HPZ (midland agroecology zone), and MPZ (lowland agroecology zone), respectively. Hence, higher decreasing trends of rainfall were observed in the LPZ (67.1 mm) and MPZ (43 mm). Similar to this result, the declining trend was found in the Woleka sub-basin, northcentral Ethiopia [13]. As a result, less water-required maize varieties may be needed for this agroecology zone in the future climate period.

Temperature Variability. Figures 4 and 5 display the
results of an analysis of the mean annual minimum and maximum temperatures in the research area. Te annual anomalies of maximum temperature showed a little bit of variation across maize production agroecology zones. In general, increasing trends for annual maximum temperature anomalies were recorded for all maize-growing agroecology in the study area ( Figure 4). As a result, the 2000s were warmer than the previous decade. In terms of minimum temperature, the 2000s were also the warmest decade compared to the 1990s ( Figure 5). Tis indicated that the general trends of the annual anomalies in the minimum temperature had similar patterns to those of the maximum temperature.

Temperature Trends.
Trends of minimum and maximum temperatures resulting from the Mann-Kendall test for three maize-productive agroecology zones are presented in Tables5 and 6. Based on S or Z values, statistically signifcant increasing trends for annual maximum and minimum temperatures were recorded from all study areas over the last 32 years (1987-2018) ( Tables 5 and 6). All agroecology zones showed a highly signifcant upward trend in the maximum temperature at P < 0.01 and P < 0.001 (Table 5). Relativity the highest warming trend was observed from LPZ while the lowest warming trend was on display from HPZ all over the study area and period. Tis could explain why maize is being grown in a low-maize productive agroecology zone. Similarly, there is a highly signifcant rising trend in the minimum temperature, with values of P < 0.001 and P < 0.01 (Table 6). Te result of this study agrees with previous studies that indicated the annual maximum and minimum temperatures are currently trending upward in diferent stations in Ethiopia [14,37] and [38].  (Table 7). Terefore, planting maize crops earlier than May 9 is possible in a study area once in fve years' time for long-maturing varieties. On the other hand, planting earlier than June 2 (154 DOY) is possible in four out of every 5 years' time. In general, the median onset date of 140 DOY (18 May) could be taken as a dependable planting date at and around in study area. Furthermore, on average, the long rainy season (Kiremt) starts on DOY 144 (May 22) for northwest Ethiopia with CV of 11.9% (Table 7).
Tus, May 22 was picked as a potential planting date for the long-maturing main-season maize crop for the Mecha district. Te result of the analysis showed that in Table 7, the rainy season terminates in the third dekad of October (298 DOY) once in 5 years time and earlier than the second dekad of November (319 DOY) in four out of fve years (Table 7). Accordingly, on average, the rainy season ends in the frst dekad of November (310 DOY) in the study area with a CV of 5.3% (Table 4). Harvesting, transporting, storing, and marketing of maize crop could thus be more easily accomplished in the study area after the frst dekad of November (310 DOY). Te probability that the length of the growing period (LGP) will be shorter than 186 days is 50%, while the probability that it will be longer than 149 days is 25% (Table 7). Te analysis of precipitation data revealed that the LGP in the main rainy season (kiremt) in the Mecha area ranges from 97 to 232 days with a mean of 180 days, a CV and SD of 9.1% and 32.2 days, respectively. Terefore, if northwest Ethiopia has a 97 to 232-day growing period, researchers need to investigate selecting suitable maturing cultivars of maize for smallholder farmers. For this reason, short, medium, and long-maturing maize cultivars could be    an option for demonstrating, and this will help farmers be able to plant early and produce a crop with less exposure to dry spells.

Conclusions
Te fndings of the spatial research of rainfall from 1987 to 2018 indicate that the LPZ (highland) tends to exhibit slightly larger variation and declining trends than the MPZ (lowland) and HPZ (midland). When RAI intensity levels were taken into consideration, there was evidence of wetness in every agroecological zone that produced maize in the northwest of the country between 1987 and 2000. However, after 2000, the intensity has shifted to dry compared to 1987-2000. Tese indicated that the highest wet and dry years were observed in the 1990s and 2000s, respectively. Currently, the implications of the results indicated that dryness was now more prominent in the research area.
Moreover, our research also shows that droughts now happen every one to three years in every agroecological zone in northwest Ethiopia that grows maize; even consecutive droughts were seen in 2009, 2010, and 2011. Terefore, these conditions might result in erratic and unexpected rainfall patterns, which would have an impact on the production of maize and encourage the installation of climate-resilient adaptation measures in each zone.
In the context of temperature, the annual anomalies of maximum and minimum temperature showed a little bit of variation across maize production agroecology zones, but considerable diferences have been observed over the past decade. As a result, the 2000s were warmer than the previous decade. Moreover, statistically signifcant increasing trends for annual maximum and minimum temperatures were recorded from all study areas over the last 32 years (1987-2018). Particularly, the LPZ (Highland) showed the highest warming trend in maximum temperature.
Onset, cessation date, and LGP are determined by adapting the defnition from Stern et al. [32]. Te lower and upper quartiles of the date of onset of rainfall are in a range of 131 (May 9)-154 (June 2) day of the year (DOY). Terefore, May 9, May 18, and June 2 were chosen as potential planting dates for long, mid, and short-maturing maize varieties for LPZ, HPZ, and MPZ, respectively. Te median onset date of 140 DOY (18 May) could be used as a general guide. In this way, farmers prepare their lands for cropping before the second dekdal of May. In the context of the cessation date, the rainy season typically fnishes in the frst dekad of November (310 DOY). So, after the frst dekad of November, it would be simpler to harvest, move, store, and market the maize crop in the research region. In the research area, the LGP during the primary rainy season (kiremt) ranges from 97 to 232 days, with a mean of 180 days. In light of the 97-232-day growing season in the northwest part of Ethiopia, researchers need to look into the best maturing cultivars of maize for smallholder farmers. As a result, farmers will be able to plant earlier and produce a crop that is less susceptible to dry periods by using short, medium, and long-maturing maize cultivars as a demonstration option.
In conclusion, consistent warming trends and increasingly erratic and concentrated rainfall patterns have been shown in all the northwest Ethiopia agroecology zones. Terefore, we have concluded and recommend that the following climate-resilient agronomic activities should be applied in the study area and similar agroecological zones: maize varieties with a low water requirement and hightemperature tolerance may be required for all agroecological zones in the current climate period; In this manner, lands could be prepared for cropping before the second dekdal of May; frst dekdal of May, second dekdal of May and frst dekdal of June could be adopted as potential planting dates for long, mid, and short maturing maize varieties for low productive agroecology (highland), mid-productive agroecology (midland), and high productive agroecology (lowland), respectively; Optimal plant density and nitrogen fertilizer doses would be applied to shorten the maturity time for all agroecology zones and harvesting, threshing, and storing the maize crop in the study area could following the frst a dekad of November.

Data Availability
Te data used to support the fndings of this study are included in the results part of the manuscript.

Conflicts of Interest
Te authors declare that they have no conficts of interest. LGP refers to length growing period: source; own study.
Advances in Meteorology 11